update HISTORY.md and version.h for 8.3.3

Summary:
This should only affect iterator when
- user uses DeleteRange(),
- An iterator from level L has a non-ok status (such non-ok status may not be caught before the bug fix in https://github.com/facebook/rocksdb/pull/11783), and
- A range tombstone covers a key from level > L and triggers a reseek sets the status_ to OK in SeekImpl()/SeekPrevImpl() e.g. https://github.com/facebook/rocksdb/blob/bd6a8340c3a2db764620e90b3ac5be173fc68a0c/table/merging_iterator.cc#L801

Pull Request resolved: https://github.com/facebook/rocksdb/pull/11786

Differential Revision: D48908830

Pulled By: cbi42

fbshipit-source-id: eb564be375af4e33dc27542eff753260186e6d5d

.circleci/config.yml

@@ -152,15 +152,15 @@ commands:
     steps:
       - run:
           name: "Test low-variance benchmarks"
-          command: ./tools/benchmark_ci.py --db_dir /tmp/rocksdb-benchmark-datadir --output_dir /tmp/benchmark-results --num_keys 10000000
+          command: ./tools/benchmark_ci.py --db_dir /tmp/rocksdb-benchmark-datadir --output_dir /tmp/benchmark-results --num_keys 20000000
           environment:
             LD_LIBRARY_PATH: /usr/local/lib
             # How long to run parts of the test(s)
-            DURATION_RO: 400
-            DURATION_RW: 700
+            DURATION_RO: 300
+            DURATION_RW: 500
             # Keep threads within physical capacity of server (much lower than default)
             NUM_THREADS: 1
-            MAX_BACKGROUND_JOBS: 3
+            MAX_BACKGROUND_JOBS: 4
             # Don't run a couple of "optional" initial tests
             CI_TESTS_ONLY: "true"
             # Reduce configured size of levels to ensure more levels in the leveled compaction LSM tree
@@ -170,7 +170,11 @@ commands:
             # The benchmark host has 32GB memory
             # The following values are tailored to work with that
             # Note, tests may not exercise the targeted issues if the memory is increased on new test hosts.
-
+            COMPRESSION_TYPE: "none"
+            CACHE_INDEX_AND_FILTER_BLOCKS: 1
+            MIN_LEVEL_TO_COMPRESS: 3
+            CACHE_SIZE_MB: 10240
+            MB_WRITE_PER_SEC: 2
 
   post-benchmarks:
     steps:
@@ -269,12 +273,12 @@ jobs:
           ./sst_dump --help | grep -E -q 'Supported compression types: kNoCompression$' # Verify no compiled in compression
       - post-steps
 
-  build-linux-shared_lib-alt_namespace-status_checked:
+  build-linux-static_lib-alt_namespace-status_checked:
     executor: linux-docker
     resource_class: 2xlarge
     steps:
       - pre-steps
-      - run: ASSERT_STATUS_CHECKED=1 TEST_UINT128_COMPAT=1 ROCKSDB_MODIFY_NPHASH=1 LIB_MODE=shared OPT="-DROCKSDB_NAMESPACE=alternative_rocksdb_ns" make V=1 -j32 check
+      - run: ASSERT_STATUS_CHECKED=1 TEST_UINT128_COMPAT=1 ROCKSDB_MODIFY_NPHASH=1 LIB_MODE=static OPT="-DROCKSDB_NAMESPACE=alternative_rocksdb_ns" make V=1 -j24 check
       - post-steps
 
   build-linux-release:
@@ -282,11 +286,21 @@ jobs:
     resource_class: 2xlarge
     steps:
       - checkout # check out the code in the project directory
+      - run: make V=1 -j32 LIB_MODE=shared release
+      - run: ls librocksdb.so # ensure shared lib built
+      - run: ./db_stress --version # ensure with gflags
+      - run: make clean
       - run: make V=1 -j32 release
+      - run: ls librocksdb.a # ensure static lib built
       - run: ./db_stress --version # ensure with gflags
       - run: make clean
       - run: apt-get remove -y libgflags-dev
+      - run: make V=1 -j32 LIB_MODE=shared release
+      - run: ls librocksdb.so # ensure shared lib built
+      - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
+      - run: make clean
       - run: make V=1 -j32 release
+      - run: ls librocksdb.a # ensure static lib built
       - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
       - post-steps
 
@@ -302,27 +316,6 @@ jobs:
       - run: USE_RTTI=1 DEBUG_LEVEL=0 make V=1 -j16 static_lib tools db_bench
       - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
 
-  build-linux-lite:
-    executor: linux-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - run: LITE=1 make V=1 J=8 -j8 check
-      - post-steps
-
-  build-linux-lite-release:
-    executor: linux-docker
-    resource_class: large
-    steps:
-      - checkout # check out the code in the project directory
-      - run: LITE=1 make V=1 -j8 release
-      - run: ./db_stress --version # ensure with gflags
-      - run: make clean
-      - run: apt-get remove -y libgflags-dev
-      - run: LITE=1 make V=1 -j8 release
-      - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
-      - post-steps
-
   build-linux-clang-no_test_run:
     executor: linux-docker
     resource_class: xlarge
@@ -427,7 +420,10 @@ jobs:
     steps:
       - checkout # check out the code in the project directory
       - run: apt-get update -y && apt-get install -y libgflags-dev
-      - run: make V=1 -j8 unity_test
+      - run:
+          name: "Unity build"
+          command: make V=1 -j8 unity_test
+          no_output_timeout: 20m
       - run: make V=1 -j8 -k check-headers # could be moved to a different build
       - post-steps
 
@@ -438,7 +434,7 @@ jobs:
       - pre-steps
       - setup-folly
       - build-folly
-      - run: USE_FOLLY=1 CC=gcc-7 CXX=g++-7 V=1 make -j32 check
+      - run: USE_FOLLY=1 LIB_MODE=static CC=gcc-7 CXX=g++-7 V=1 make -j32 check # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   build-linux-gcc-7-with-folly-lite-no-test:
@@ -484,7 +480,7 @@ jobs:
     resource_class: 2xlarge
     steps:
       - pre-steps
-      - run: CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench
+      - run: LIB_MODE=static CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   build-linux-clang-13-no_test_run:
@@ -503,7 +499,7 @@ jobs:
       - pre-steps
       - setup-folly
       - build-folly
-      - run: CC=clang-13 CXX=clang++-13 USE_CLANG=1 USE_FOLLY=1 COMPILE_WITH_UBSAN=1 COMPILE_WITH_ASAN=1 make -j32 check
+      - run: CC=clang-13 CXX=clang++-13 LIB_MODE=static USE_CLANG=1 USE_FOLLY=1 COMPILE_WITH_UBSAN=1 COMPILE_WITH_ASAN=1 make -j32 check # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   # This job is only to make sure the microbench tests are able to run, the benchmark result is not meaningful as the CI host is changing.
@@ -520,7 +516,7 @@ jobs:
     resource_class: large
     steps:
       - pre-steps
-      - run: ulimit -S -n `ulimit -H -n` && make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=960 --max_key=2500000' blackbox_crash_test_with_atomic_flush
+      - run: ulimit -S -n `ulimit -H -n` && make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=960 --max_key=2500000 --use_io_uring=0' blackbox_crash_test_with_atomic_flush
       - post-steps
 
   build-linux-crashtest-tiered-storage-bb:
@@ -530,7 +526,7 @@ jobs:
       - pre-steps
       - run:
           name: "run crashtest"
-          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS=--duration=10800 blackbox_crash_test_with_tiered_storage
+          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS='--duration=10800 --use_io_uring=0' blackbox_crash_test_with_tiered_storage
           no_output_timeout: 100m
       - post-steps
 
@@ -541,7 +537,7 @@ jobs:
       - pre-steps
       - run:
           name: "run crashtest"
-          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS=--duration=10800 whitebox_crash_test_with_tiered_storage
+          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS='--duration=10800 --use_io_uring=0' whitebox_crash_test_with_tiered_storage
           no_output_timeout: 100m
       - post-steps
 
@@ -821,18 +817,16 @@ workflows:
       - build-linux-cmake-with-folly-coroutines
       - build-linux-cmake-with-benchmark
       - build-linux-encrypted_env-no_compression
-      - build-linux-lite
   jobs-linux-run-tests-san:
     jobs:
       - build-linux-clang10-asan
       - build-linux-clang10-ubsan
       - build-linux-clang10-mini-tsan
-      - build-linux-shared_lib-alt_namespace-status_checked
+      - build-linux-static_lib-alt_namespace-status_checked
   jobs-linux-no-test-run:
     jobs:
       - build-linux-release
       - build-linux-release-rtti
-      - build-linux-lite-release
       - build-examples
       - build-fuzzers
       - build-linux-clang-no_test_run

.gitignore

@@ -95,3 +95,4 @@ fuzz/crash-*
 
 cmake-build-*
 third-party/folly/
+.cache

CMakeLists.txt

@@ -245,41 +245,40 @@ if(CMAKE_SYSTEM_PROCESSOR MATCHES "s390x")
   endif(HAS_S390X_MARCH_NATIVE)
 endif(CMAKE_SYSTEM_PROCESSOR MATCHES "s390x")
 
-option(PORTABLE "build a portable binary" OFF)
-option(FORCE_SSE42 "force building with SSE4.2, even when PORTABLE=ON" OFF)
-option(FORCE_AVX "force building with AVX, even when PORTABLE=ON" OFF)
-option(FORCE_AVX2 "force building with AVX2, even when PORTABLE=ON" OFF)
-if(PORTABLE)
-  add_definitions(-DROCKSDB_PORTABLE)
+if(CMAKE_SYSTEM_PROCESSOR MATCHES "loongarch64")
+  CHECK_C_COMPILER_FLAG("-march=loongarch64" HAS_LOONGARCH64)
+  if(HAS_LOONGARCH64)
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mcpu=loongarch64 -mtune=loongarch64")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mcpu=loongarch64 -mtune=loongarch64")
+  endif(HAS_LOONGARCH64)
+endif(CMAKE_SYSTEM_PROCESSOR MATCHES "loongarch64")
 
-  # MSVC does not need a separate compiler flag to enable SSE4.2; if nmmintrin.h
-  # is available, it is available by default.
-  if(FORCE_SSE42 AND NOT MSVC)
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -msse4.2 -mpclmul")
-  endif()
-  if(MSVC)
-    if(FORCE_AVX)
-      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /arch:AVX")
-    endif()
-    # MSVC automatically enables BMI / lzcnt with AVX2.
-    if(FORCE_AVX2)
-      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /arch:AVX2")
-    endif()
-  else()
-    if(FORCE_AVX)
-      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mavx")
-    endif()
-    if(FORCE_AVX2)
-      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mavx2 -mbmi -mlzcnt")
-    endif()
+set(PORTABLE 0 CACHE STRING "Minimum CPU arch to support, or 0 = current CPU, 1 = baseline CPU")
+if(PORTABLE STREQUAL 1)
+  # Usually nothing to do; compiler default is typically the most general
+  if(NOT MSVC)
     if(CMAKE_SYSTEM_PROCESSOR MATCHES "^s390x")
       set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=z196")
     endif()
+    if(CMAKE_SYSTEM_PROCESSOR MATCHES "^loongarch64")
+      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=loongarch64")
+    endif()
+  endif()
+elseif(PORTABLE MATCHES [^0]+)
+  # Name of a CPU arch spec or feature set to require
+  if(MSVC)
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /arch:${PORTABLE}")
+  else()
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=${PORTABLE}")
   endif()
 else()
   if(MSVC)
+    # NOTE: No auto-detection of current CPU, but instead assume some useful
+    # level of optimization is supported
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /arch:AVX2")
   else()
+    # Require instruction set from current CPU (with some legacy or opt-out
+    # exceptions)
     if(CMAKE_SYSTEM_PROCESSOR MATCHES "^s390x" AND NOT HAS_S390X_MARCH_NATIVE)
       set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=z196")
     elseif(NOT CMAKE_SYSTEM_PROCESSOR MATCHES "^(powerpc|ppc)64" AND NOT HAS_ARMV8_CRC)
@@ -294,25 +293,6 @@ if(NOT MSVC)
   set(CMAKE_REQUIRED_FLAGS "-msse4.2 -mpclmul")
 endif()
 
-CHECK_CXX_SOURCE_COMPILES("
-#include <cstdint>
-#include <nmmintrin.h>
-#include <wmmintrin.h>
-int main() {
-  volatile uint32_t x = _mm_crc32_u32(0, 0);
-  const auto a = _mm_set_epi64x(0, 0);
-  const auto b = _mm_set_epi64x(0, 0);
-  const auto c = _mm_clmulepi64_si128(a, b, 0x00);
-  auto d = _mm_cvtsi128_si64(c);
-}
-" HAVE_SSE42)
-if(HAVE_SSE42)
-  add_definitions(-DHAVE_SSE42)
-  add_definitions(-DHAVE_PCLMUL)
-elseif(FORCE_SSE42)
-  message(FATAL_ERROR "FORCE_SSE42=ON but unable to compile with SSE4.2 enabled")
-endif()
-
 # Check if -latomic is required or not
 if (NOT MSVC)
   set(CMAKE_REQUIRED_FLAGS "--std=c++17")
@@ -485,12 +465,6 @@ if(CMAKE_COMPILER_IS_GNUCXX)
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fno-builtin-memcmp")
 endif()
 
-option(ROCKSDB_LITE "Build RocksDBLite version" OFF)
-if(ROCKSDB_LITE)
-  add_definitions(-DROCKSDB_LITE)
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fno-exceptions -Os")
-endif()
-
 if(CMAKE_SYSTEM_NAME MATCHES "Cygwin")
   add_definitions(-fno-builtin-memcmp -DCYGWIN)
 elseif(CMAKE_SYSTEM_NAME MATCHES "Darwin")
@@ -649,12 +623,14 @@ set(SOURCES
         cache/cache.cc
         cache/cache_entry_roles.cc
         cache/cache_key.cc
+        cache/cache_helpers.cc
         cache/cache_reservation_manager.cc
         cache/charged_cache.cc
         cache/clock_cache.cc
         cache/compressed_secondary_cache.cc
         cache/lru_cache.cc
         cache/secondary_cache.cc
+        cache/secondary_cache_adapter.cc
         cache/sharded_cache.cc
         db/arena_wrapped_db_iter.cc
         db/blob/blob_contents.cc
@@ -741,6 +717,7 @@ set(SOURCES
         db/write_batch.cc
         db/write_batch_base.cc
         db/write_controller.cc
+        db/write_stall_stats.cc
         db/write_thread.cc
         env/composite_env.cc
         env/env.cc
@@ -806,6 +783,7 @@ set(SOURCES
         table/block_based/block_based_table_iterator.cc
         table/block_based/block_based_table_reader.cc
         table/block_based/block_builder.cc
+        table/block_based/block_cache.cc
         table/block_based/block_prefetcher.cc
         table/block_based/block_prefix_index.cc
         table/block_based/data_block_hash_index.cc
@@ -831,6 +809,7 @@ set(SOURCES
         table/get_context.cc
         table/iterator.cc
         table/merging_iterator.cc
+        table/compaction_merging_iterator.cc
         table/meta_blocks.cc
         table/persistent_cache_helper.cc
         table/plain/plain_table_bloom.cc
@@ -872,6 +851,7 @@ set(SOURCES
         util/compression_context_cache.cc
         util/concurrent_task_limiter_impl.cc
         util/crc32c.cc
+        util/data_structure.cc
         util/dynamic_bloom.cc
         util/hash.cc
         util/murmurhash.cc
@@ -979,6 +959,12 @@ if ( ROCKSDB_PLUGINS )
         plugin/${plugin}/${src}
         PROPERTIES COMPILE_FLAGS "${${plugin}_COMPILE_FLAGS}")
     endforeach()
+    foreach (test ${${plugin}_TESTS})
+      list(APPEND PLUGIN_TESTS plugin/${plugin}/${test})
+      set_source_files_properties(
+        plugin/${plugin}/${test}
+        PROPERTIES COMPILE_FLAGS "${${plugin}_COMPILE_FLAGS}")
+    endforeach()
     foreach (path ${${plugin}_INCLUDE_PATHS})
       include_directories(${path})
     endforeach()
@@ -993,12 +979,6 @@ if ( ROCKSDB_PLUGINS )
   endforeach()
 endif()
 
-if(HAVE_SSE42 AND NOT MSVC)
-  set_source_files_properties(
-    util/crc32c.cc
-    PROPERTIES COMPILE_FLAGS "-msse4.2 -mpclmul")
-endif()
-
 if(CMAKE_SYSTEM_PROCESSOR MATCHES "^(powerpc|ppc)64")
   list(APPEND SOURCES
     util/crc32c_ppc.c
@@ -1264,6 +1244,7 @@ if(WITH_TESTS OR WITH_BENCHMARK_TOOLS)
   add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
   add_library(testharness STATIC
   test_util/mock_time_env.cc
+  test_util/secondary_cache_test_util.cc
   test_util/testharness.cc)
   target_link_libraries(testharness gtest)
 endif()
@@ -1309,6 +1290,7 @@ if(WITH_TESTS)
         db/db_bloom_filter_test.cc
         db/db_compaction_filter_test.cc
         db/db_compaction_test.cc
+        db/db_clip_test.cc
         db/db_dynamic_level_test.cc
         db/db_encryption_test.cc
         db/db_flush_test.cc
@@ -1470,6 +1452,7 @@ if(WITH_TESTS)
         utilities/ttl/ttl_test.cc
         utilities/util_merge_operators_test.cc
         utilities/write_batch_with_index/write_batch_with_index_test.cc
+	${PLUGIN_TESTS}
     )
   endif()
 

HISTORY.md

@@ -1,10 +1,160 @@
 # Rocksdb Change Log
+## 8.3.3 (09/01/2023)
+### Bug Fixes
+* Fix a bug where if there is an error reading from offset 0 of a file from L1+ and that the file is not the first file in the sorted run, data can be lost in compaction and read/scan can return incorrect results.
+* Fix a bug where iterator may return incorrect result for DeleteRange() users if there was an error reading from a file.
+* Fixed a race condition in `GenericRateLimiter` that could cause it to stop granting requests
+
+## 8.3.2 (06/14/2023)
+### Bug Fixes
+* Reduced cases of illegally using Env::Default() during static destruction by never destroying the internal PosixEnv itself (except for builds checking for memory leaks). (#11538)
+
+## 8.3.1 (06/07/2023)
+### Performance Improvements
+* Fixed higher read QPS during DB::Open() reading files created prior to #11406, especially when reading many small file (size < 52 MB) during DB::Open() and partitioned filter or index is used.
+
+## 8.3.0 (05/19/2023)
+### New Features
+* Introduced a new option `block_protection_bytes_per_key`, which can be used to enable per key-value integrity protection for in-memory blocks in block cache (#11287).
+* Added `JemallocAllocatorOptions::num_arenas`. Setting `num_arenas > 1` may mitigate mutex contention in the allocator, particularly in scenarios where block allocations commonly bypass jemalloc tcache.
+* Improve the operational safety of publishing a DB or SST files to many hosts by using different block cache hash seeds on different hosts. The exact behavior is controlled by new option `ShardedCacheOptions::hash_seed`, which also documents the solved problem in more detail.
+* Introduced a new option `CompactionOptionsFIFO::file_temperature_age_thresholds` that allows FIFO compaction to compact files to different temperatures based on key age (#11428).
+* Added a new ticker stat to count how many times RocksDB detected a corruption while verifying a block checksum: `BLOCK_CHECKSUM_MISMATCH_COUNT`.
+* New statistics `rocksdb.file.read.db.open.micros` that measures read time of block-based SST tables or blob files during db open.
+* New statistics tickers for various iterator seek behaviors and relevant filtering, as \*`_LEVEL_SEEK_`\*. (#11460)
+
+### Public API Changes
+* EXPERIMENTAL: Add new API `DB::ClipColumnFamily` to clip the key in CF to a certain range. It will physically deletes all keys outside the range including tombstones.
+* Add `MakeSharedCache()` construction functions to various cache Options objects, and deprecated the `NewWhateverCache()` functions with long parameter lists.
+* Changed the meaning of various Bloom filter stats (prefix vs. whole key), with iterator-related filtering only being tracked in the new \*`_LEVEL_SEEK_`\*. stats. (#11460)
+
+### Behavior changes
+* For x86, CPU features are no longer detected at runtime nor in build scripts, but in source code using common preprocessor defines. This will likely unlock some small performance improvements on some newer hardware, but could hurt performance of the kCRC32c checksum, which is no longer the default, on some "portable" builds. See PR #11419 for details.
+
+### Bug Fixes
+* Delete an empty WAL file on DB open if the log number is less than the min log number to keep
+* Delete temp OPTIONS file on DB open if there is a failure to write it out or rename it
+
+### Performance Improvements
+* Improved the I/O efficiency of prefetching SST metadata by recording more information in the DB manifest. Opening files written with previous versions will still rely on heuristics for how much to prefetch (#11406).
+
+## 8.2.0 (04/24/2023)
+### Public API Changes
+* `SstFileWriter::DeleteRange()` now returns `Status::InvalidArgument` if the range's end key comes before its start key according to the user comparator. Previously the behavior was undefined.
+* Add `multi_get_for_update` to C API.
+* Remove unnecessary constructor for CompressionOptions.
+
+### Behavior changes
+* Changed default block cache size from an 8MB to 32MB LRUCache, which increases the default number of cache shards from 16 to 64. This change is intended to minimize cache mutex contention under stress conditions. See https://github.com/facebook/rocksdb/wiki/Block-Cache for more information.
+* For level compaction with `level_compaction_dynamic_level_bytes=true`, RocksDB now trivially moves levels down to fill LSM starting from bottommost level during DB open. See more in comments for option `level_compaction_dynamic_level_bytes` (#11321).
+* User-provided `ReadOptions` take effect for more reads of non-`CacheEntryRole::kDataBlock` blocks.
+* For level compaction with `level_compaction_dynamic_level_bytes=true`, RocksDB now drains unnecessary levels through background compaction automatically (#11340). This together with #11321 makes it automatic to migrate other compaction settings to level compaction with `level_compaction_dynamic_level_bytes=true`. In addition, a live DB that becomes smaller will now have unnecessary levels drained which can help to reduce read and space amp.
+* If `CompactRange()` is called with `CompactRangeOptions::bottommost_level_compaction=kForce*` to compact from L0 to L1, RocksDB now will try to do trivial move from L0 to L1 and then do an intra L1 compaction, instead of a L0 to L1 compaction with trivial move disabled (#11375)).
+
+### Bug Fixes
+* In the DB::VerifyFileChecksums API, ensure that file system reads of SST files are equal to the readahead_size in ReadOptions, if specified. Previously, each read was 2x the readahead_size.
+* In block cache tracing, fixed some cases of bad hit/miss information (and more) with MultiGet.
+
+### New Features
+* Add experimental `PerfContext` counters `iter_{next|prev|seek}_count` for db iterator, each counting the times of corresponding API being called.
+* Allow runtime changes to whether `WriteBufferManager` allows stall or not by calling `SetAllowStall()`
+* Added statistics tickers BYTES_COMPRESSED_FROM, BYTES_COMPRESSED_TO, BYTES_COMPRESSION_BYPASSED, BYTES_COMPRESSION_REJECTED, NUMBER_BLOCK_COMPRESSION_BYPASSED, and NUMBER_BLOCK_COMPRESSION_REJECTED. Disabled/deprecated histograms BYTES_COMPRESSED and BYTES_DECOMPRESSED, and ticker NUMBER_BLOCK_NOT_COMPRESSED. The new tickers offer more inight into compression ratios, rejected vs. disabled compression, etc. (#11388)
+* New statistics `rocksdb.file.read.{flush|compaction}.micros` that measure read time of block-based SST tables or blob files during flush or compaction.
+
+## 8.1.0 (03/18/2023)
+### Behavior changes
+* Compaction output file cutting logic now considers range tombstone start keys. For example, SST partitioner now may receive ParitionRequest for range tombstone start keys.
+* If the async_io ReadOption is specified for MultiGet or NewIterator on a platform that doesn't support IO uring, the option is ignored and synchronous IO is used.
+
+### Bug Fixes
+* Fixed an issue for backward iteration when user defined timestamp is enabled in combination with BlobDB.
+* Fixed a couple of cases where a Merge operand encountered during iteration wasn't reflected in the `internal_merge_count` PerfContext counter.
+* Fixed a bug in CreateColumnFamilyWithImport()/ExportColumnFamily() which did not support range tombstones (#11252).
+* Fixed a bug where an excluded column family from an atomic flush contains unflushed data that should've been included in this atomic flush (i.e, data of seqno less than the max seqno of this atomic flush), leading to potential data loss in this excluded column family when `WriteOptions::disableWAL == true` (#11148).
+
+### New Features
+* Add statistics rocksdb.secondary.cache.filter.hits, rocksdb.secondary.cache.index.hits, and rocksdb.secondary.cache.filter.hits
+* Added a new PerfContext counter `internal_merge_point_lookup_count` which tracks the number of Merge operands applied while serving point lookup queries.
+* Add new statistics rocksdb.table.open.prefetch.tail.read.bytes, rocksdb.table.open.prefetch.tail.{miss|hit}
+* Add support for SecondaryCache with HyperClockCache (`HyperClockCacheOptions` inherits `secondary_cache` option from `ShardedCacheOptions`)
+* Add new db properties `rocksdb.cf-write-stall-stats`, `rocksdb.db-write-stall-stats`and APIs to examine them in a structured way. In particular, users of `GetMapProperty()` with property `kCFWriteStallStats`/`kDBWriteStallStats` can now use the functions in `WriteStallStatsMapKeys` to find stats in the map.
+
+### Public API Changes
+* Changed various functions and features in `Cache` that are mostly relevant to custom implementations or wrappers. Especially, asychronous lookup functionality is moved from `Lookup()` to a new `StartAsyncLookup()` function.
+
+## 8.0.0 (02/19/2023)
+### Behavior changes
+* `ReadOptions::verify_checksums=false` disables checksum verification for more reads of non-`CacheEntryRole::kDataBlock` blocks.
+* In case of scan with async_io enabled, if posix doesn't support IOUring, Status::NotSupported error will be returned to the users. Initially that error was swallowed and reads were switched to synchronous reads.
+
+### Bug Fixes
+* Fixed a data race on `ColumnFamilyData::flush_reason` caused by concurrent flushes.
+* Fixed an issue in `Get` and `MultiGet` when user-defined timestamps is enabled in combination with BlobDB.
+* Fixed some atypical behaviors for `LockWAL()` such as allowing concurrent/recursive use and not expecting `UnlockWAL()` after non-OK result. See API comments.
+* Fixed a feature interaction bug where for blobs `GetEntity` would expose the blob reference instead of the blob value.
+* Fixed `DisableManualCompaction()` and `CompactRangeOptions::canceled` to cancel compactions even when they are waiting on conflicting compactions to finish
+* Fixed a bug in which a successful `GetMergeOperands()` could transiently return `Status::MergeInProgress()`
+* Return the correct error (Status::NotSupported()) to MultiGet caller when ReadOptions::async_io flag is true and IO uring is not enabled. Previously, Status::Corruption() was being returned when the actual failure was lack of async IO support.
+* Fixed a bug in DB open/recovery from a compressed WAL that was caused due to incorrect handling of certain record fragments with the same offset within a WAL block.
+
+### Feature Removal
+* Remove RocksDB Lite.
+* The feature block_cache_compressed is removed. Statistics related to it are removed too.
+* Remove deprecated Env::LoadEnv(). Use Env::CreateFromString() instead.
+* Remove deprecated FileSystem::Load(). Use FileSystem::CreateFromString() instead.
+* Removed the deprecated version of these utility functions and the corresponding Java bindings: `LoadOptionsFromFile`, `LoadLatestOptions`, `CheckOptionsCompatibility`.
+* Remove the FactoryFunc from the LoadObject method from the Customizable helper methods.
+
+### Public API Changes
+* Moved rarely-needed Cache class definition to new advanced_cache.h, and added a CacheWrapper class to advanced_cache.h. Minor changes to SimCache API definitions.
+* Completely removed the following deprecated/obsolete statistics: the tickers `BLOCK_CACHE_INDEX_BYTES_EVICT`, `BLOCK_CACHE_FILTER_BYTES_EVICT`, `BLOOM_FILTER_MICROS`, `NO_FILE_CLOSES`, `STALL_L0_SLOWDOWN_MICROS`, `STALL_MEMTABLE_COMPACTION_MICROS`, `STALL_L0_NUM_FILES_MICROS`, `RATE_LIMIT_DELAY_MILLIS`, `NO_ITERATORS`, `NUMBER_FILTERED_DELETES`, `WRITE_TIMEDOUT`, `BLOB_DB_GC_NUM_KEYS_OVERWRITTEN`, `BLOB_DB_GC_NUM_KEYS_EXPIRED`, `BLOB_DB_GC_BYTES_OVERWRITTEN`, `BLOB_DB_GC_BYTES_EXPIRED`, `BLOCK_CACHE_COMPRESSION_DICT_BYTES_EVICT` as well as the histograms `STALL_L0_SLOWDOWN_COUNT`, `STALL_MEMTABLE_COMPACTION_COUNT`, `STALL_L0_NUM_FILES_COUNT`, `HARD_RATE_LIMIT_DELAY_COUNT`, `SOFT_RATE_LIMIT_DELAY_COUNT`, `BLOB_DB_GC_MICROS`, and `NUM_DATA_BLOCKS_READ_PER_LEVEL`. Note that as a result, the C++ enum values of the still supported statistics have changed. Developers are advised to not rely on the actual numeric values.
+* Deprecated IngestExternalFileOptions::write_global_seqno and change default to false. This option only needs to be set to true to generate a DB compatible with RocksDB versions before 5.16.0.
+* Remove deprecated APIs `GetColumnFamilyOptionsFrom{Map|String}(const ColumnFamilyOptions&, ..)`, `GetDBOptionsFrom{Map|String}(const DBOptions&, ..)`, `GetBlockBasedTableOptionsFrom{Map|String}(const BlockBasedTableOptions& table_options, ..)` and ` GetPlainTableOptionsFrom{Map|String}(const PlainTableOptions& table_options,..)`.
+* Added a subcode of `Status::Corruption`, `Status::SubCode::kMergeOperatorFailed`, for users to identify corruption failures originating in the merge operator, as opposed to RocksDB's internally identified data corruptions
+
+### Build Changes
+* The `make` build now builds a shared library by default instead of a static library. Use `LIB_MODE=static` to override.
+
+### New Features
+* Compaction filters are now supported for wide-column entities by means of the `FilterV3` API. See the comment of the API for more details.
+* Added `do_not_compress_roles` to `CompressedSecondaryCacheOptions` to disable compression on certain kinds of block. Filter blocks are now not compressed by CompressedSecondaryCache by default.
+* Added a new `MultiGetEntity` API that enables batched wide-column point lookups. See the API comments for more details.
+
+## 7.10.0 (01/23/2023)
+### Behavior changes
+* Make best-efforts recovery verify SST unique ID before Version construction (#10962)
+* Introduce `epoch_number` and sort L0 files by `epoch_number` instead of `largest_seqno`. `epoch_number` represents the order of a file being flushed or ingested/imported. Compaction output file will be assigned with the minimum `epoch_number` among input files'. For L0, larger `epoch_number` indicates newer L0 file.
+
+### Bug Fixes
+* Fixed a regression in iterator where range tombstones after `iterate_upper_bound` is processed.
+* Fixed a memory leak in MultiGet with async_io read option, caused by IO errors during table file open
+* Fixed a bug that multi-level FIFO compaction deletes one file in non-L0 even when `CompactionOptionsFIFO::max_table_files_size` is no exceeded since #10348 or 7.8.0.
+* Fixed a bug caused by `DB::SyncWAL()` affecting `track_and_verify_wals_in_manifest`. Without the fix, application may see "open error: Corruption: Missing WAL with log number" while trying to open the db. The corruption is a false alarm but prevents DB open (#10892).
+* Fixed a BackupEngine bug in which RestoreDBFromLatestBackup would fail if the latest backup was deleted and there is another valid backup available.
+* Fix L0 file misorder corruption caused by ingesting files of overlapping seqnos with memtable entries' through introducing `epoch_number`. Before the fix, `force_consistency_checks=true` may catch the corruption before it's exposed to readers, in which case writes returning `Status::Corruption` would be expected. Also replace the previous incomplete fix (#5958) to the same corruption with this new and more complete fix.
+* Fixed a bug in LockWAL() leading to re-locking mutex (#11020).
+* Fixed a heap use after free bug in async scan prefetching when the scan thread and another thread try to read and load the same seek block into cache.
+* Fixed a heap use after free in async scan prefetching if dictionary compression is enabled, in which case sync read of the compression dictionary gets mixed with async prefetching
+* Fixed a data race bug of `CompactRange()` under `change_level=true` acts on overlapping range with an ongoing file ingestion for level compaction. This will either result in overlapping file ranges corruption at a certain level caught by `force_consistency_checks=true` or protentially two same keys both with seqno 0 in two different levels (i.e, new data ends up in lower/older level). The latter will be caught by assertion in debug build but go silently and result in read returning wrong result in release build. This fix is general so it also replaced previous fixes to a similar problem for `CompactFiles()` (#4665), general `CompactRange()` and auto compaction (commit 5c64fb6 and 87dfc1d).
+* Fixed a bug in compaction output cutting where small output files were produced due to TTL file cutting states were not being updated (#11075).
+
+### New Features
+* When an SstPartitionerFactory is configured, CompactRange() now automatically selects for compaction any files overlapping a partition boundary that is in the compaction range, even if no actual entries are in the requested compaction range. With this feature, manual compaction can be used to (re-)establish SST partition points when SstPartitioner changes, without a full compaction.
+* Add BackupEngine feature to exclude files from backup that are known to be backed up elsewhere, using `CreateBackupOptions::exclude_files_callback`. To restore the DB, the excluded files must be provided in alternative backup directories using `RestoreOptions::alternate_dirs`.
+
+### Public API Changes
+* Substantial changes have been made to the Cache class to support internal development goals. Direct use of Cache class members is discouraged and further breaking modifications are expected in the future. SecondaryCache has some related changes and implementations will need to be updated. (Unlike Cache, SecondaryCache is still intended to support user implementations, and disruptive changes will be avoided.) (#10975)
+* Add `MergeOperationOutput::op_failure_scope` for merge operator users to control the blast radius of merge operator failures. Existing merge operator users do not need to make any change to preserve the old behavior
+
+### Performance Improvements
+* Updated xxHash source code, which should improve kXXH3 checksum speed, at least on ARM (#11098).
+* Improved CPU efficiency of DB reads, from block cache access improvements (#10975).
+
 ## 7.9.0 (11/21/2022)
 ### Performance Improvements
 * Fixed an iterator performance regression for delete range users when scanning through a consecutive sequence of range tombstones (#10877).
 
 ### Bug Fixes
-* Fix FIFO compaction causing corruption of overlapping seqnos in L0 files due to ingesting files of overlapping seqnos with memtable's under `CompactionOptionsFIFO::allow_compaction=true` or `CompactionOptionsFIFO::age_for_warm>0` or `CompactRange()/CompactFiles()` is used. Before the fix, `force_consistency_checks=true` may catch the corruption before it's exposed to readers, in which case writes returning `Status::Corruption` would be expected.
 * Fix memory corruption error in scans if async_io is enabled. Memory corruption happened if there is IOError while reading the data leading to empty buffer and other buffer already in progress of async read goes again for reading.
 * Fix failed memtable flush retry bug that could cause wrongly ordered updates, which would surface to writers as `Status::Corruption` in case of `force_consistency_checks=true` (default). It affects use cases that enable both parallel flush (`max_background_flushes > 1` or `max_background_jobs >= 8`) and non-default memtable count (`max_write_buffer_number > 2`).
 * Fixed an issue where the `READ_NUM_MERGE_OPERANDS` ticker was not updated when the base key-value or tombstone was read from an SST file.

INSTALL.md

@@ -20,12 +20,15 @@ There are few options when compiling RocksDB:
 depend on gflags. You will need to have gflags installed to run `make all`. This will compile RocksDB in debug mode. Don't
 use binaries compiled by `make all` in production.
 
-* By default the binary we produce is optimized for the platform you're compiling on
-(`-march=native` or the equivalent). SSE4.2 will thus be enabled automatically if your
-CPU supports it. To print a warning if your CPU does not support SSE4.2, build with
-`USE_SSE=1 make static_lib` or, if using CMake, `cmake -DFORCE_SSE42=ON`. If you want
-to build a portable binary, add `PORTABLE=1` before your make commands, like this:
-`PORTABLE=1 make static_lib`.
+* By default the binary we produce is optimized for the CPU you're compiling on
+(`-march=native` or the equivalent). To build a binary compatible with the most
+general architecture supported by your CPU and compiler, set `PORTABLE=1` for
+the build, but performance will suffer as many operations benefit from newer
+and wider instructions. In addition to `PORTABLE=0` (default) and `PORTABLE=1`,
+it can be set to an architecture name recognized by your compiler. For example,
+on 64-bit x86, a reasonable compromise is `PORTABLE=haswell` which supports
+many or most of the available optimizations while still being compatible with
+most processors made since roughly 2013.
 
 ## Dependencies
 
@@ -48,6 +51,11 @@ to build a portable binary, add `PORTABLE=1` before your make commands, like thi
 * If you wish to build the RocksJava static target, then cmake is required for building Snappy.
 
 * If you wish to run microbench (e.g, `make microbench`, `make ribbon_bench` or `cmake -DWITH_BENCHMARK=1`), Google benchmark >= 1.6.0 is needed.
+* You can do the following to install Google benchmark. These commands are copied from `./build_tools/ubuntu20_image/Dockerfile`:
+
+`$ git clone --depth 1 --branch v1.7.0 https://github.com/google/benchmark.git ~/benchmark`
+
+`$ cd ~/benchmark && mkdir build && cd build && cmake .. -GNinja -DCMAKE_BUILD_TYPE=Release -DBENCHMARK_ENABLE_GTEST_TESTS=0 && ninja && ninja install`
 
 ## Supported platforms
 
@@ -178,7 +186,7 @@ to build a portable binary, add `PORTABLE=1` before your make commands, like thi
         gmake rocksdbjava
 
 * **iOS**:
-  * Run: `TARGET_OS=IOS make static_lib`. When building the project which uses rocksdb iOS library, make sure to define two important pre-processing macros: `ROCKSDB_LITE` and `IOS_CROSS_COMPILE`.
+  * Run: `TARGET_OS=IOS make static_lib`. When building the project which uses rocksdb iOS library, make sure to define an important pre-processing macros: `IOS_CROSS_COMPILE`.
 
 * **Windows** (Visual Studio 2017 to up):
   * Read and follow the instructions at CMakeLists.txt

Makefile

@@ -44,13 +44,6 @@ quoted_perl_command = $(subst ','\'',$(perl_command))
 # Set the default DEBUG_LEVEL to 1
 DEBUG_LEVEL?=1
 
-# LIB_MODE says whether or not to use/build "shared" or "static" libraries.
-# Mode "static" means to link against static libraries (.a)
-# Mode "shared" means to link against shared libraries (.so, .sl, .dylib, etc)
-#
-# Set the default LIB_MODE to static
-LIB_MODE?=static
-
 # OBJ_DIR is where the object files reside.  Default to the current directory
 OBJ_DIR?=.
 
@@ -81,29 +74,23 @@ else ifneq ($(filter jtest rocksdbjava%, $(MAKECMDGOALS)),)
 	endif
 endif
 
-$(info $$DEBUG_LEVEL is ${DEBUG_LEVEL})
+# LIB_MODE says whether or not to use/build "shared" or "static" libraries.
+# Mode "static" means to link against static libraries (.a)
+# Mode "shared" means to link against shared libraries (.so, .sl, .dylib, etc)
+#
+ifeq ($(DEBUG_LEVEL), 0)
+# For optimized, set the default LIB_MODE to static for code size/efficiency
+	LIB_MODE?=static
+else
+# For debug, set the default LIB_MODE to shared for efficient `make check` etc.
+	LIB_MODE?=shared
+endif
 
-# Lite build flag.
-LITE ?= 0
-ifeq ($(LITE), 0)
-ifneq ($(filter -DROCKSDB_LITE,$(OPT)),)
-  # Be backward compatible and support older format where OPT=-DROCKSDB_LITE is
-  # specified instead of LITE=1 on the command line.
-  LITE=1
-endif
-else ifeq ($(LITE), 1)
-ifeq ($(filter -DROCKSDB_LITE,$(OPT)),)
-	OPT += -DROCKSDB_LITE
-endif
-endif
+$(info $$DEBUG_LEVEL is $(DEBUG_LEVEL), $$LIB_MODE is $(LIB_MODE))
 
 # Figure out optimize level.
 ifneq ($(DEBUG_LEVEL), 2)
-ifeq ($(LITE), 0)
 	OPTIMIZE_LEVEL ?= -O2
-else
-	OPTIMIZE_LEVEL ?= -Os
-endif
 endif
 # `OPTIMIZE_LEVEL` is empty when the user does not set it and `DEBUG_LEVEL=2`.
 # In that case, the compiler default (`-O0` for gcc and clang) will be used.
@@ -266,6 +253,7 @@ ROCKSDB_PLUGIN_EXTERNS = $(foreach p, $(ROCKSDB_PLUGIN_W_FUNCS), int $($(p)_FUNC
 ROCKSDB_PLUGIN_BUILTINS = $(foreach p, $(ROCKSDB_PLUGIN_W_FUNCS), {\"$(p)\"\, $($(p)_FUNC)}\,)
 ROCKSDB_PLUGIN_LDFLAGS = $(foreach plugin, $(ROCKSDB_PLUGINS), $($(plugin)_LDFLAGS))
 ROCKSDB_PLUGIN_PKGCONFIG_REQUIRES = $(foreach plugin, $(ROCKSDB_PLUGINS), $($(plugin)_PKGCONFIG_REQUIRES))
+ROCKSDB_PLUGIN_TESTS = $(foreach p, $(ROCKSDB_PLUGINS), $(foreach test, $($(p)_TESTS), plugin/$(p)/$(test)))
 
 CXXFLAGS += $(foreach plugin, $(ROCKSDB_PLUGINS), $($(plugin)_CXXFLAGS))
 PLATFORM_LDFLAGS += $(ROCKSDB_PLUGIN_LDFLAGS)
@@ -336,13 +324,6 @@ endif
 ifeq ($(PLATFORM), OS_SOLARIS)
 	PLATFORM_CXXFLAGS += -D _GLIBCXX_USE_C99
 endif
-ifneq ($(filter -DROCKSDB_LITE,$(OPT)),)
-	# found
-	CFLAGS += -fno-exceptions
-	CXXFLAGS += -fno-exceptions
-	# LUA is not supported under ROCKSDB_LITE
-	LUA_PATH =
-endif
 
 ifeq ($(LIB_MODE),shared)
 # So that binaries are executable from build location, in addition to install location
@@ -356,8 +337,8 @@ ifneq ($(MACHINE), arm64)
 # linking with jemalloc (as it won't be arm64-compatible) and remove some other options
 # set during platform detection
 DISABLE_JEMALLOC=1
-PLATFORM_CFLAGS := $(filter-out -march=native -DHAVE_SSE42 -DHAVE_AVX2, $(PLATFORM_CFLAGS))
-PLATFORM_CXXFLAGS := $(filter-out -march=native -DHAVE_SSE42 -DHAVE_AVX2, $(PLATFORM_CXXFLAGS))
+PLATFORM_CCFLAGS := $(filter-out -march=native, $(PLATFORM_CCFLAGS))
+PLATFORM_CXXFLAGS := $(filter-out -march=native, $(PLATFORM_CXXFLAGS))
 endif
 endif
 endif
@@ -558,7 +539,7 @@ endif
 
 ifdef USE_CLANG
 	# Used by some teams in Facebook
-	WARNING_FLAGS += -Wshift-sign-overflow
+	WARNING_FLAGS += -Wshift-sign-overflow -Wambiguous-reversed-operator
 endif
 
 ifeq ($(PLATFORM), OS_OPENBSD)
@@ -647,10 +628,12 @@ STRESS_OBJECTS =  $(patsubst %.cc, $(OBJ_DIR)/%.o, $(STRESS_LIB_SOURCES))
 ALL_SOURCES  = $(filter-out util/build_version.cc, $(LIB_SOURCES)) $(TEST_LIB_SOURCES) $(MOCK_LIB_SOURCES) $(GTEST_DIR)/gtest/gtest-all.cc
 ALL_SOURCES += $(TOOL_LIB_SOURCES) $(BENCH_LIB_SOURCES) $(CACHE_BENCH_LIB_SOURCES) $(ANALYZER_LIB_SOURCES) $(STRESS_LIB_SOURCES)
 ALL_SOURCES += $(TEST_MAIN_SOURCES) $(TOOL_MAIN_SOURCES) $(BENCH_MAIN_SOURCES)
-ALL_SOURCES += $(ROCKSDB_PLUGIN_SOURCES)
+ALL_SOURCES += $(ROCKSDB_PLUGIN_SOURCES) $(ROCKSDB_PLUGIN_TESTS)
 
+PLUGIN_TESTS = $(patsubst %.cc, %, $(notdir $(ROCKSDB_PLUGIN_TESTS)))
 TESTS = $(patsubst %.cc, %, $(notdir $(TEST_MAIN_SOURCES)))
 TESTS += $(patsubst %.c, %, $(notdir $(TEST_MAIN_SOURCES_C)))
+TESTS += $(PLUGIN_TESTS)
 
 # `make check-headers` to very that each header file includes its own
 # dependencies
@@ -702,6 +685,7 @@ NON_PARALLEL_TEST = \
 	env_test \
 	deletefile_test \
 	db_bloom_filter_test \
+	$(PLUGIN_TESTS) \
 
 PARALLEL_TEST = $(filter-out $(NON_PARALLEL_TEST), $(TESTS))
 
@@ -1078,13 +1062,11 @@ check: all
 	rm -rf $(TEST_TMPDIR)
 ifneq ($(PLATFORM), OS_AIX)
 	$(PYTHON) tools/check_all_python.py
-ifeq ($(filter -DROCKSDB_LITE,$(OPT)),)
 ifndef ASSERT_STATUS_CHECKED # not yet working with these tests
 	$(PYTHON) tools/ldb_test.py
 	sh tools/rocksdb_dump_test.sh
 endif
 endif
-endif
 ifndef SKIP_FORMAT_BUCK_CHECKS
 	$(MAKE) check-format
 	$(MAKE) check-buck-targets
@@ -1240,9 +1222,9 @@ clean: clean-ext-libraries-all clean-rocks clean-rocksjava
 clean-not-downloaded: clean-ext-libraries-bin clean-rocks clean-not-downloaded-rocksjava
 
 clean-rocks:
-	echo shared=$(ALL_SHARED_LIBS)
-	echo static=$(ALL_STATIC_LIBS)
-	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(ALL_STATIC_LIBS) $(ALL_SHARED_LIBS) $(MICROBENCHS)
+# Not practical to exactly match all versions/variants in naming (e.g. debug or not)
+	rm -f ${LIBNAME}*.so* ${LIBNAME}*.a
+	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(MICROBENCHS)
 	rm -rf $(CLEAN_FILES) ios-x86 ios-arm scan_build_report
 	$(FIND) . -name "*.[oda]" -exec rm -f {} \;
 	$(FIND) . -type f \( -name "*.gcda" -o -name "*.gcno" \) -exec rm -f {} \;
@@ -1355,6 +1337,14 @@ db_sanity_test: $(OBJ_DIR)/tools/db_sanity_test.o $(LIBRARY)
 db_repl_stress: $(OBJ_DIR)/tools/db_repl_stress.o $(LIBRARY)
 	$(AM_LINK)
 
+define MakeTestRule
+$(notdir $(1:%.cc=%)): $(1:%.cc=$$(OBJ_DIR)/%.o) $$(TEST_LIBRARY) $$(LIBRARY)
+	$$(AM_LINK)
+endef
+
+# For each PLUGIN test, create a rule to generate the test executable
+$(foreach test, $(ROCKSDB_PLUGIN_TESTS), $(eval $(call MakeTestRule, $(test))))
+
 arena_test: $(OBJ_DIR)/memory/arena_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
@@ -1490,6 +1480,9 @@ db_compaction_filter_test: $(OBJ_DIR)/db/db_compaction_filter_test.o $(TEST_LIBR
 db_compaction_test: $(OBJ_DIR)/db/db_compaction_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
+db_clip_test: $(OBJ_DIR)/db/db_clip_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 db_dynamic_level_test: $(OBJ_DIR)/db/db_dynamic_level_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
@@ -2058,7 +2051,7 @@ JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/linux
 ifeq ($(PLATFORM), OS_SOLARIS)
 	ARCH := $(shell isainfo -b)
 else ifeq ($(PLATFORM), OS_OPENBSD)
-	ifneq (,$(filter amd64 ppc64 ppc64le s390x arm64 aarch64 sparc64, $(MACHINE)))
+	ifneq (,$(filter amd64 ppc64 ppc64le s390x arm64 aarch64 sparc64 loongarch64, $(MACHINE)))
 		ARCH := 64
 	else
 		ARCH := 32
@@ -2079,7 +2072,7 @@ ifneq ($(origin JNI_LIBC), undefined)
 endif
 
 ifeq (,$(ROCKSDBJNILIB))
-ifneq (,$(filter ppc% s390x arm64 aarch64 sparc64, $(MACHINE)))
+ifneq (,$(filter ppc% s390x arm64 aarch64 sparc64 loongarch64, $(MACHINE)))
 	ROCKSDBJNILIB = librocksdbjni-linux-$(MACHINE)$(JNI_LIBC_POSTFIX).so
 else
 	ROCKSDBJNILIB = librocksdbjni-linux$(ARCH)$(JNI_LIBC_POSTFIX).so
@@ -2445,6 +2438,8 @@ checkout_folly:
 	@# NOTE: this hack is required for gcc in some cases
 	perl -pi -e 's/(__has_include.<experimental.memory_resource>.)/__cpp_rtti && $$1/' third-party/folly/folly/memory/MemoryResource.h
 
+CXX_M_FLAGS = $(filter -m%, $(CXXFLAGS))
+
 build_folly:
 	FOLLY_INST_PATH=`cd third-party/folly; $(PYTHON) build/fbcode_builder/getdeps.py show-inst-dir`; \
 	if [ "$$FOLLY_INST_PATH" ]; then \
@@ -2455,8 +2450,8 @@ build_folly:
 	fi
 	# Restore the original version of Invoke.h with boost dependency
 	cd third-party/folly && ${GIT_COMMAND} checkout folly/functional/Invoke.h
-	cd third-party/folly && MAYBE_AVX2=`echo $(CXXFLAGS) | grep -o -- -DHAVE_AVX2 | sed 's/-DHAVE_AVX2/-mavx2/g' || true` && \
-		CXXFLAGS=" $$MAYBE_AVX2 -DHAVE_CXX11_ATOMIC " $(PYTHON) build/fbcode_builder/getdeps.py build --no-tests
+	cd third-party/folly && \
+		CXXFLAGS=" $(CXX_M_FLAGS) -DHAVE_CXX11_ATOMIC " $(PYTHON) build/fbcode_builder/getdeps.py build --no-tests
 
 # ---------------------------------------------------------------------------
 #   Build size testing
@@ -2477,18 +2472,6 @@ build_size:
 	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.shared_lib $$(stat --printf="%s" `readlink -f librocksdb.so`)
 	strip `readlink -f librocksdb.so`
 	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.shared_lib_stripped $$(stat --printf="%s" `readlink -f librocksdb.so`)
-	# === lite build, static ===
-	$(MAKE) clean
-	$(MAKE) LITE=1 static_lib
-	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.static_lib_lite $$(stat --printf="%s" librocksdb.a)
-	strip librocksdb.a
-	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.static_lib_lite_stripped $$(stat --printf="%s" librocksdb.a)
-	# === lite build, shared ===
-	$(MAKE) clean
-	$(MAKE) LITE=1 shared_lib
-	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.shared_lib_lite $$(stat --printf="%s" `readlink -f librocksdb.so`)
-	strip `readlink -f librocksdb.so`
-	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.shared_lib_lite_stripped $$(stat --printf="%s" `readlink -f librocksdb.so`)
 
 # ---------------------------------------------------------------------------
 #  	Platform-specific compilation

PLUGINS.md

@@ -5,3 +5,4 @@ This is the list of all known third-party plugins for RocksDB. If something is m
 * [ZenFS](https://github.com/westerndigitalcorporation/zenfs): a file system for zoned block devices
 * [RADOS](https://github.com/riversand963/rocksdb-rados-env): an Env used for interacting with RADOS. Migrated from RocksDB main repo.
 * [PMEM](https://github.com/pmem/pmem-rocksdb-plugin): a collection of plugins to enable Persistent Memory on RocksDB.
+* [IPPCP](https://github.com/intel/ippcp-plugin-rocksdb): a plugin to enable encryption on RocksDB based on Intel optimized open source IPP-Crypto library.

README.md

@@ -1,8 +1,6 @@
 ## RocksDB: A Persistent Key-Value Store for Flash and RAM Storage
 
 [![CircleCI Status](https://circleci.com/gh/facebook/rocksdb.svg?style=svg)](https://circleci.com/gh/facebook/rocksdb)
-[![Appveyor Build status](https://ci.appveyor.com/api/projects/status/fbgfu0so3afcno78/branch/main?svg=true)](https://ci.appveyor.com/project/Facebook/rocksdb/branch/main)
-[![PPC64le Build Status](http://140-211-168-68-openstack.osuosl.org:8080/buildStatus/icon?job=rocksdb&style=plastic)](http://140-211-168-68-openstack.osuosl.org:8080/job/rocksdb)
 
 RocksDB is developed and maintained by Facebook Database Engineering Team.
 It is built on earlier work on [LevelDB](https://github.com/google/leveldb) by Sanjay Ghemawat (sanjay@google.com)

ROCKSDB_LITE.md

@@ -1,21 +0,0 @@
-# RocksDBLite
-
-RocksDBLite is a project focused on mobile use cases, which don't need a lot of fancy things we've built for server workloads and they are very sensitive to binary size. For that reason, we added a compile flag ROCKSDB_LITE that comments out a lot of the nonessential code and keeps the binary lean.
-
-Some examples of the features disabled by ROCKSDB_LITE:
-* compiled-in support for LDB tool
-* No backup engine
-* No support for replication (which we provide in form of TransactionalIterator)
-* No advanced monitoring tools
-* No special-purpose memtables that are highly optimized for specific use cases
-* No Transactions
-
-When adding a new big feature to RocksDB, please add ROCKSDB_LITE compile guard if:
-* Nobody from mobile really needs your feature,
-* Your feature is adding a lot of weight to the binary.
-
-Don't add ROCKSDB_LITE compile guard if:
-* It would introduce a lot of code complexity. Compile guards make code harder to read. It's a trade-off.
-* Your feature is not adding a lot of weight.
-
-If unsure, ask. :)

TARGETS

@@ -11,6 +11,7 @@ load("//rocks/buckifier:defs.bzl", "cpp_library_wrapper","rocks_cpp_library_wrap
 cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "cache/cache.cc",
         "cache/cache_entry_roles.cc",
+        "cache/cache_helpers.cc",
         "cache/cache_key.cc",
         "cache/cache_reservation_manager.cc",
         "cache/charged_cache.cc",
@@ -18,6 +19,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "cache/compressed_secondary_cache.cc",
         "cache/lru_cache.cc",
         "cache/secondary_cache.cc",
+        "cache/secondary_cache_adapter.cc",
         "cache/sharded_cache.cc",
         "db/arena_wrapped_db_iter.cc",
         "db/blob/blob_contents.cc",
@@ -104,6 +106,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "db/write_batch.cc",
         "db/write_batch_base.cc",
         "db/write_controller.cc",
+        "db/write_stall_stats.cc",
         "db/write_thread.cc",
         "env/composite_env.cc",
         "env/env.cc",
@@ -180,6 +183,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "table/block_based/block_based_table_iterator.cc",
         "table/block_based/block_based_table_reader.cc",
         "table/block_based/block_builder.cc",
+        "table/block_based/block_cache.cc",
         "table/block_based/block_prefetcher.cc",
         "table/block_based/block_prefix_index.cc",
         "table/block_based/data_block_footer.cc",
@@ -198,6 +202,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "table/block_based/reader_common.cc",
         "table/block_based/uncompression_dict_reader.cc",
         "table/block_fetcher.cc",
+        "table/compaction_merging_iterator.cc",
         "table/cuckoo/cuckoo_table_builder.cc",
         "table/cuckoo/cuckoo_table_factory.cc",
         "table/cuckoo/cuckoo_table_reader.cc",
@@ -245,6 +250,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "util/concurrent_task_limiter_impl.cc",
         "util/crc32c.cc",
         "util/crc32c_arm64.cc",
+        "util/data_structure.cc",
         "util/dynamic_bloom.cc",
         "util/file_checksum_helper.cc",
         "util/hash.cc",
@@ -348,351 +354,14 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "//folly/synchronization:distributed_mutex",
     ], headers=None, link_whole=False, extra_test_libs=False)
 
-cpp_library_wrapper(name="rocksdb_whole_archive_lib", srcs=[
-        "cache/cache.cc",
-        "cache/cache_entry_roles.cc",
-        "cache/cache_key.cc",
-        "cache/cache_reservation_manager.cc",
-        "cache/charged_cache.cc",
-        "cache/clock_cache.cc",
-        "cache/compressed_secondary_cache.cc",
-        "cache/lru_cache.cc",
-        "cache/secondary_cache.cc",
-        "cache/sharded_cache.cc",
-        "db/arena_wrapped_db_iter.cc",
-        "db/blob/blob_contents.cc",
-        "db/blob/blob_fetcher.cc",
-        "db/blob/blob_file_addition.cc",
-        "db/blob/blob_file_builder.cc",
-        "db/blob/blob_file_cache.cc",
-        "db/blob/blob_file_garbage.cc",
-        "db/blob/blob_file_meta.cc",
-        "db/blob/blob_file_reader.cc",
-        "db/blob/blob_garbage_meter.cc",
-        "db/blob/blob_log_format.cc",
-        "db/blob/blob_log_sequential_reader.cc",
-        "db/blob/blob_log_writer.cc",
-        "db/blob/blob_source.cc",
-        "db/blob/prefetch_buffer_collection.cc",
-        "db/builder.cc",
-        "db/c.cc",
-        "db/column_family.cc",
-        "db/compaction/compaction.cc",
-        "db/compaction/compaction_iterator.cc",
-        "db/compaction/compaction_job.cc",
-        "db/compaction/compaction_outputs.cc",
-        "db/compaction/compaction_picker.cc",
-        "db/compaction/compaction_picker_fifo.cc",
-        "db/compaction/compaction_picker_level.cc",
-        "db/compaction/compaction_picker_universal.cc",
-        "db/compaction/compaction_service_job.cc",
-        "db/compaction/compaction_state.cc",
-        "db/compaction/sst_partitioner.cc",
-        "db/compaction/subcompaction_state.cc",
-        "db/convenience.cc",
-        "db/db_filesnapshot.cc",
-        "db/db_impl/compacted_db_impl.cc",
-        "db/db_impl/db_impl.cc",
-        "db/db_impl/db_impl_compaction_flush.cc",
-        "db/db_impl/db_impl_debug.cc",
-        "db/db_impl/db_impl_experimental.cc",
-        "db/db_impl/db_impl_files.cc",
-        "db/db_impl/db_impl_open.cc",
-        "db/db_impl/db_impl_readonly.cc",
-        "db/db_impl/db_impl_secondary.cc",
-        "db/db_impl/db_impl_write.cc",
-        "db/db_info_dumper.cc",
-        "db/db_iter.cc",
-        "db/dbformat.cc",
-        "db/error_handler.cc",
-        "db/event_helpers.cc",
-        "db/experimental.cc",
-        "db/external_sst_file_ingestion_job.cc",
-        "db/file_indexer.cc",
-        "db/flush_job.cc",
-        "db/flush_scheduler.cc",
-        "db/forward_iterator.cc",
-        "db/import_column_family_job.cc",
-        "db/internal_stats.cc",
-        "db/log_reader.cc",
-        "db/log_writer.cc",
-        "db/logs_with_prep_tracker.cc",
-        "db/malloc_stats.cc",
-        "db/memtable.cc",
-        "db/memtable_list.cc",
-        "db/merge_helper.cc",
-        "db/merge_operator.cc",
-        "db/output_validator.cc",
-        "db/periodic_task_scheduler.cc",
-        "db/range_del_aggregator.cc",
-        "db/range_tombstone_fragmenter.cc",
-        "db/repair.cc",
-        "db/seqno_to_time_mapping.cc",
-        "db/snapshot_impl.cc",
-        "db/table_cache.cc",
-        "db/table_properties_collector.cc",
-        "db/transaction_log_impl.cc",
-        "db/trim_history_scheduler.cc",
-        "db/version_builder.cc",
-        "db/version_edit.cc",
-        "db/version_edit_handler.cc",
-        "db/version_set.cc",
-        "db/wal_edit.cc",
-        "db/wal_manager.cc",
-        "db/wide/wide_column_serialization.cc",
-        "db/wide/wide_columns.cc",
-        "db/write_batch.cc",
-        "db/write_batch_base.cc",
-        "db/write_controller.cc",
-        "db/write_thread.cc",
-        "env/composite_env.cc",
-        "env/env.cc",
-        "env/env_chroot.cc",
-        "env/env_encryption.cc",
-        "env/env_posix.cc",
-        "env/file_system.cc",
-        "env/file_system_tracer.cc",
-        "env/fs_posix.cc",
-        "env/fs_remap.cc",
-        "env/io_posix.cc",
-        "env/mock_env.cc",
-        "env/unique_id_gen.cc",
-        "file/delete_scheduler.cc",
-        "file/file_prefetch_buffer.cc",
-        "file/file_util.cc",
-        "file/filename.cc",
-        "file/line_file_reader.cc",
-        "file/random_access_file_reader.cc",
-        "file/read_write_util.cc",
-        "file/readahead_raf.cc",
-        "file/sequence_file_reader.cc",
-        "file/sst_file_manager_impl.cc",
-        "file/writable_file_writer.cc",
-        "logging/auto_roll_logger.cc",
-        "logging/event_logger.cc",
-        "logging/log_buffer.cc",
-        "memory/arena.cc",
-        "memory/concurrent_arena.cc",
-        "memory/jemalloc_nodump_allocator.cc",
-        "memory/memkind_kmem_allocator.cc",
-        "memory/memory_allocator.cc",
-        "memtable/alloc_tracker.cc",
-        "memtable/hash_linklist_rep.cc",
-        "memtable/hash_skiplist_rep.cc",
-        "memtable/skiplistrep.cc",
-        "memtable/vectorrep.cc",
-        "memtable/write_buffer_manager.cc",
-        "monitoring/histogram.cc",
-        "monitoring/histogram_windowing.cc",
-        "monitoring/in_memory_stats_history.cc",
-        "monitoring/instrumented_mutex.cc",
-        "monitoring/iostats_context.cc",
-        "monitoring/perf_context.cc",
-        "monitoring/perf_level.cc",
-        "monitoring/persistent_stats_history.cc",
-        "monitoring/statistics.cc",
-        "monitoring/thread_status_impl.cc",
-        "monitoring/thread_status_updater.cc",
-        "monitoring/thread_status_updater_debug.cc",
-        "monitoring/thread_status_util.cc",
-        "monitoring/thread_status_util_debug.cc",
-        "options/cf_options.cc",
-        "options/configurable.cc",
-        "options/customizable.cc",
-        "options/db_options.cc",
-        "options/options.cc",
-        "options/options_helper.cc",
-        "options/options_parser.cc",
-        "port/mmap.cc",
-        "port/port_posix.cc",
-        "port/stack_trace.cc",
-        "port/win/env_default.cc",
-        "port/win/env_win.cc",
-        "port/win/io_win.cc",
-        "port/win/port_win.cc",
-        "port/win/win_logger.cc",
-        "port/win/win_thread.cc",
-        "table/adaptive/adaptive_table_factory.cc",
-        "table/block_based/binary_search_index_reader.cc",
-        "table/block_based/block.cc",
-        "table/block_based/block_based_table_builder.cc",
-        "table/block_based/block_based_table_factory.cc",
-        "table/block_based/block_based_table_iterator.cc",
-        "table/block_based/block_based_table_reader.cc",
-        "table/block_based/block_builder.cc",
-        "table/block_based/block_prefetcher.cc",
-        "table/block_based/block_prefix_index.cc",
-        "table/block_based/data_block_footer.cc",
-        "table/block_based/data_block_hash_index.cc",
-        "table/block_based/filter_block_reader_common.cc",
-        "table/block_based/filter_policy.cc",
-        "table/block_based/flush_block_policy.cc",
-        "table/block_based/full_filter_block.cc",
-        "table/block_based/hash_index_reader.cc",
-        "table/block_based/index_builder.cc",
-        "table/block_based/index_reader_common.cc",
-        "table/block_based/parsed_full_filter_block.cc",
-        "table/block_based/partitioned_filter_block.cc",
-        "table/block_based/partitioned_index_iterator.cc",
-        "table/block_based/partitioned_index_reader.cc",
-        "table/block_based/reader_common.cc",
-        "table/block_based/uncompression_dict_reader.cc",
-        "table/block_fetcher.cc",
-        "table/cuckoo/cuckoo_table_builder.cc",
-        "table/cuckoo/cuckoo_table_factory.cc",
-        "table/cuckoo/cuckoo_table_reader.cc",
-        "table/format.cc",
-        "table/get_context.cc",
-        "table/iterator.cc",
-        "table/merging_iterator.cc",
-        "table/meta_blocks.cc",
-        "table/persistent_cache_helper.cc",
-        "table/plain/plain_table_bloom.cc",
-        "table/plain/plain_table_builder.cc",
-        "table/plain/plain_table_factory.cc",
-        "table/plain/plain_table_index.cc",
-        "table/plain/plain_table_key_coding.cc",
-        "table/plain/plain_table_reader.cc",
-        "table/sst_file_dumper.cc",
-        "table/sst_file_reader.cc",
-        "table/sst_file_writer.cc",
-        "table/table_factory.cc",
-        "table/table_properties.cc",
-        "table/two_level_iterator.cc",
-        "table/unique_id.cc",
-        "test_util/sync_point.cc",
-        "test_util/sync_point_impl.cc",
-        "test_util/transaction_test_util.cc",
-        "tools/dump/db_dump_tool.cc",
-        "tools/io_tracer_parser_tool.cc",
-        "tools/ldb_cmd.cc",
-        "tools/ldb_tool.cc",
-        "tools/sst_dump_tool.cc",
-        "trace_replay/block_cache_tracer.cc",
-        "trace_replay/io_tracer.cc",
-        "trace_replay/trace_record.cc",
-        "trace_replay/trace_record_handler.cc",
-        "trace_replay/trace_record_result.cc",
-        "trace_replay/trace_replay.cc",
-        "util/async_file_reader.cc",
-        "util/build_version.cc",
-        "util/cleanable.cc",
-        "util/coding.cc",
-        "util/compaction_job_stats_impl.cc",
-        "util/comparator.cc",
-        "util/compression.cc",
-        "util/compression_context_cache.cc",
-        "util/concurrent_task_limiter_impl.cc",
-        "util/crc32c.cc",
-        "util/crc32c_arm64.cc",
-        "util/dynamic_bloom.cc",
-        "util/file_checksum_helper.cc",
-        "util/hash.cc",
-        "util/murmurhash.cc",
-        "util/random.cc",
-        "util/rate_limiter.cc",
-        "util/ribbon_config.cc",
-        "util/slice.cc",
-        "util/status.cc",
-        "util/stderr_logger.cc",
-        "util/string_util.cc",
-        "util/thread_local.cc",
-        "util/threadpool_imp.cc",
-        "util/xxhash.cc",
-        "utilities/agg_merge/agg_merge.cc",
-        "utilities/backup/backup_engine.cc",
-        "utilities/blob_db/blob_compaction_filter.cc",
-        "utilities/blob_db/blob_db.cc",
-        "utilities/blob_db/blob_db_impl.cc",
-        "utilities/blob_db/blob_db_impl_filesnapshot.cc",
-        "utilities/blob_db/blob_dump_tool.cc",
-        "utilities/blob_db/blob_file.cc",
-        "utilities/cache_dump_load.cc",
-        "utilities/cache_dump_load_impl.cc",
-        "utilities/cassandra/cassandra_compaction_filter.cc",
-        "utilities/cassandra/format.cc",
-        "utilities/cassandra/merge_operator.cc",
-        "utilities/checkpoint/checkpoint_impl.cc",
-        "utilities/compaction_filters.cc",
-        "utilities/compaction_filters/remove_emptyvalue_compactionfilter.cc",
-        "utilities/convenience/info_log_finder.cc",
-        "utilities/counted_fs.cc",
-        "utilities/debug.cc",
-        "utilities/env_mirror.cc",
-        "utilities/env_timed.cc",
-        "utilities/fault_injection_env.cc",
-        "utilities/fault_injection_fs.cc",
-        "utilities/fault_injection_secondary_cache.cc",
-        "utilities/leveldb_options/leveldb_options.cc",
-        "utilities/memory/memory_util.cc",
-        "utilities/merge_operators.cc",
-        "utilities/merge_operators/bytesxor.cc",
-        "utilities/merge_operators/max.cc",
-        "utilities/merge_operators/put.cc",
-        "utilities/merge_operators/sortlist.cc",
-        "utilities/merge_operators/string_append/stringappend.cc",
-        "utilities/merge_operators/string_append/stringappend2.cc",
-        "utilities/merge_operators/uint64add.cc",
-        "utilities/object_registry.cc",
-        "utilities/option_change_migration/option_change_migration.cc",
-        "utilities/options/options_util.cc",
-        "utilities/persistent_cache/block_cache_tier.cc",
-        "utilities/persistent_cache/block_cache_tier_file.cc",
-        "utilities/persistent_cache/block_cache_tier_metadata.cc",
-        "utilities/persistent_cache/persistent_cache_tier.cc",
-        "utilities/persistent_cache/volatile_tier_impl.cc",
-        "utilities/simulator_cache/cache_simulator.cc",
-        "utilities/simulator_cache/sim_cache.cc",
-        "utilities/table_properties_collectors/compact_on_deletion_collector.cc",
-        "utilities/trace/file_trace_reader_writer.cc",
-        "utilities/trace/replayer_impl.cc",
-        "utilities/transactions/lock/lock_manager.cc",
-        "utilities/transactions/lock/point/point_lock_manager.cc",
-        "utilities/transactions/lock/point/point_lock_tracker.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/concurrent_tree.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/keyrange.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/lock_request.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/locktree.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/manager.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/range_buffer.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/treenode.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/txnid_set.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/wfg.cc",
-        "utilities/transactions/lock/range/range_tree/lib/standalone_port.cc",
-        "utilities/transactions/lock/range/range_tree/lib/util/dbt.cc",
-        "utilities/transactions/lock/range/range_tree/lib/util/memarena.cc",
-        "utilities/transactions/lock/range/range_tree/range_tree_lock_manager.cc",
-        "utilities/transactions/lock/range/range_tree/range_tree_lock_tracker.cc",
-        "utilities/transactions/optimistic_transaction.cc",
-        "utilities/transactions/optimistic_transaction_db_impl.cc",
-        "utilities/transactions/pessimistic_transaction.cc",
-        "utilities/transactions/pessimistic_transaction_db.cc",
-        "utilities/transactions/snapshot_checker.cc",
-        "utilities/transactions/transaction_base.cc",
-        "utilities/transactions/transaction_db_mutex_impl.cc",
-        "utilities/transactions/transaction_util.cc",
-        "utilities/transactions/write_prepared_txn.cc",
-        "utilities/transactions/write_prepared_txn_db.cc",
-        "utilities/transactions/write_unprepared_txn.cc",
-        "utilities/transactions/write_unprepared_txn_db.cc",
-        "utilities/ttl/db_ttl_impl.cc",
-        "utilities/wal_filter.cc",
-        "utilities/write_batch_with_index/write_batch_with_index.cc",
-        "utilities/write_batch_with_index/write_batch_with_index_internal.cc",
-    ], deps=[
-        "//folly/container:f14_hash",
-        "//folly/experimental/coro:blocking_wait",
-        "//folly/experimental/coro:collect",
-        "//folly/experimental/coro:coroutine",
-        "//folly/experimental/coro:task",
-        "//folly/synchronization:distributed_mutex",
-    ], headers=None, link_whole=True, extra_test_libs=False)
+cpp_library_wrapper(name="rocksdb_whole_archive_lib", srcs=[], deps=[":rocksdb_lib"], headers=None, link_whole=True, extra_test_libs=False)
 
 cpp_library_wrapper(name="rocksdb_test_lib", srcs=[
         "db/db_test_util.cc",
         "db/db_with_timestamp_test_util.cc",
         "table/mock_table.cc",
         "test_util/mock_time_env.cc",
+        "test_util/secondary_cache_test_util.cc",
         "test_util/testharness.cc",
         "test_util/testutil.cc",
         "tools/block_cache_analyzer/block_cache_trace_analyzer.cc",
@@ -723,6 +392,7 @@ rocks_cpp_library_wrapper(name="rocksdb_stress_lib", srcs=[
         "db_stress_tool/db_stress_test_base.cc",
         "db_stress_tool/db_stress_tool.cc",
         "db_stress_tool/expected_state.cc",
+        "db_stress_tool/expected_value.cc",
         "db_stress_tool/multi_ops_txns_stress.cc",
         "db_stress_tool/no_batched_ops_stress.cc",
         "test_util/testutil.cc",
@@ -5080,6 +4750,12 @@ cpp_unittest_wrapper(name="db_bloom_filter_test",
             extra_compiler_flags=[])
 
 
+cpp_unittest_wrapper(name="db_clip_test",
+            srcs=["db/db_clip_test.cc"],
+            deps=[":rocksdb_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="db_compaction_filter_test",
             srcs=["db/db_compaction_filter_test.cc"],
             deps=[":rocksdb_test_lib"],

USERS.md

@@ -15,6 +15,28 @@ At Facebook, we use RocksDB as storage engines in multiple data management servi
 
 [2] https://code.facebook.com/posts/357056558062811/logdevice-a-distributed-data-store-for-logs/
 
+## Bilibili
+[Bilibili](bilibili.com) [uses](https://www.alluxio.io/blog/when-ai-meets-alluxio-at-bilibili-building-an-efficient-ai-platform-for-data-preprocessing-and-model-training/) Alluxio to speed up its ML training workloads, and Alluxio uses RocksDB to store its filesystem metadata, so Bilibili uses RocksDB.
+
+Bilibili's [real-time platform](https://www.alibabacloud.com/blog/architecture-and-practices-of-bilibilis-real-time-platform_596676) uses Flink, and uses RocksDB as Flink's state store.
+
+## TikTok
+TikTok, or its parent company ByteDance, uses RocksDB as the storage engine for some storage systems, such as its distributed graph database [ByteGraph](https://vldb.org/pvldb/vol15/p3306-li.pdf). 
+
+Also, TikTok uses [Alluxio](alluxio.io) to [speed up Presto queries](https://www.alluxio.io/resources/videos/improving-presto-performance-with-alluxio-at-tiktok/), and Alluxio stores the files' metadata in RocksDB.
+
+## FoundationDB
+[FoundationDB](https://www.foundationdb.org/) [uses](https://github.com/apple/foundationdb/blob/377f1f692da6ab2fe5bdac57035651db3e5fb66d/fdbserver/KeyValueStoreRocksDB.actor.cpp) RocksDB to implement a [key-value store interface](https://github.com/apple/foundationdb/blob/377f1f692da6ab2fe5bdac57035651db3e5fb66d/fdbserver/KeyValueStoreRocksDB.actor.cpp#L1127) in its server backend.
+
+## Apple
+Apple [uses](https://opensource.apple.com/projects/foundationdb/) FoundationDB, so it also uses RocksDB.
+
+## Snowflake
+Snowflake [uses](https://www.snowflake.com/blog/how-foundationdb-powers-snowflake-metadata-forward/) FoundationDB, so it also uses RocksDB.
+
+## Microsoft
+The Bing search engine from Microsoft uses RocksDB as the storage engine for its web data platform: https://blogs.bing.com/Engineering-Blog/october-2021/RocksDB-in-Microsoft-Bing
+
 ## LinkedIn
 Two different use cases at Linkedin are using RocksDB as a storage engine:
 
@@ -26,6 +48,9 @@ Learn more about those use cases in a Tech Talk by Ankit Gupta and Naveen Somasu
 ## Yahoo
 Yahoo is using RocksDB as a storage engine for their biggest distributed data store Sherpa. Learn more about it here: http://yahooeng.tumblr.com/post/120730204806/sherpa-scales-new-heights
 
+## Tencent
+[PaxosStore](https://github.com/Tencent/paxosstore) is a distributed database supporting WeChat. It uses RocksDB as its storage engine.
+
 ## Baidu
 [Apache Doris](http://doris.apache.org/master/en/) is a MPP analytical database engine released by Baidu. It [uses RocksDB](http://doris.apache.org/master/en/administrator-guide/operation/tablet-meta-tool.html) to manage its tablet's metadata.
 
@@ -79,6 +104,18 @@ quasardb uses a heavily tuned RocksDB as its persistence layer.
 ## TiKV
 [TiKV](https://github.com/pingcap/tikv) is a GEO-replicated, high-performance, distributed, transactional key-value database. TiKV is powered by Rust and Raft. TiKV uses RocksDB as its persistence layer.
 
+## TiDB
+[TiDB](https://github.com/pingcap/tidb) uses the TiKV distributed key-value database, so it uses RocksDB.
+
+## PingCAP
+[PingCAP](https://www.pingcap.com/) is the company behind TiDB, its cloud database service uses RocksDB.
+
+## Apache Spark
+[Spark Structured Streaming](https://docs.databricks.com/structured-streaming/rocksdb-state-store.html) uses RocksDB as the local state store.
+
+## Databricks
+[Databricks](https://www.databricks.com/) [replaces AWS RDS with TiDB](https://www.pingcap.com/case-study/how-databricks-tackles-the-scalability-limit-with-a-mysql-alternative/) for scalability, so it uses RocksDB.
+
 ## Apache Flink
 [Apache Flink](https://flink.apache.org/news/2016/03/08/release-1.0.0.html) uses RocksDB to store state locally on a machine.
 

buckifier/buckify_rocksdb.py

@@ -26,7 +26,7 @@ from util import ColorString
 # $python3 buckifier/buckify_rocksdb.py \
 #        '{"fake": {
 #                      "extra_deps": [":test_dep", "//fakes/module:mock1"],
-#                      "extra_compiler_flags": ["-DROCKSDB_LITE", "-Os"]
+#                      "extra_compiler_flags": ["-DFOO_BAR", "-Os"]
 #                  }
 #         }'
 # (Generated TARGETS file has test_dep and mock1 as dependencies for RocksDB
@@ -154,16 +154,9 @@ def generate_targets(repo_path, deps_map):
     # rocksdb_whole_archive_lib
     TARGETS.add_library(
         "rocksdb_whole_archive_lib",
-        src_mk["LIB_SOURCES"] +
-        # always add range_tree, it's only excluded on ppc64, which we don't use internally
-        src_mk["RANGE_TREE_SOURCES"] + src_mk["TOOL_LIB_SOURCES"],
+        [],
         deps=[
-            "//folly/container:f14_hash",
-            "//folly/experimental/coro:blocking_wait",
-            "//folly/experimental/coro:collect",
-            "//folly/experimental/coro:coroutine",
-            "//folly/experimental/coro:task",
-            "//folly/synchronization:distributed_mutex",
+            ":rocksdb_lib",
         ],
         headers=None,
         extra_external_deps="",

build_tools/build_detect_platform

@@ -63,13 +63,7 @@ if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
     if [ "$LIB_MODE" == "shared" ]; then
       PIC_BUILD=1
     fi
-    if [ -n "$ROCKSDB_FBCODE_BUILD_WITH_PLATFORM010" ]; then
-      source "$PWD/build_tools/fbcode_config_platform010.sh"
-    elif [ -n "$ROCKSDB_FBCODE_BUILD_WITH_PLATFORM009" ]; then
-      source "$PWD/build_tools/fbcode_config_platform009.sh"
-    else
-      source "$PWD/build_tools/fbcode_config_platform009.sh"
-    fi
+    source "$PWD/build_tools/fbcode_config_platform010.sh"
 fi
 
 # Delete existing output, if it exists
@@ -154,7 +148,7 @@ case "$TARGET_OS" in
         ;;
     IOS)
         PLATFORM=IOS
-        COMMON_FLAGS="$COMMON_FLAGS -DOS_MACOSX -DIOS_CROSS_COMPILE -DROCKSDB_LITE"
+        COMMON_FLAGS="$COMMON_FLAGS -DOS_MACOSX -DIOS_CROSS_COMPILE "
         PLATFORM_SHARED_EXT=dylib
         PLATFORM_SHARED_LDFLAGS="-dynamiclib -install_name "
         CROSS_COMPILE=true
@@ -627,7 +621,7 @@ EOF
   fi
 fi
 
-if test "0$PORTABLE" -eq 0; then
+if [ "$PORTABLE" == "" ] || [ "$PORTABLE" == 0 ]; then
   if test -n "`echo $TARGET_ARCHITECTURE | grep ^ppc64`"; then
     # Tune for this POWER processor, treating '+' models as base models
     POWER=`LD_SHOW_AUXV=1 /bin/true | grep AT_PLATFORM | grep -E -o power[0-9]+`
@@ -650,27 +644,26 @@ if test "0$PORTABLE" -eq 0; then
     COMMON_FLAGS="$COMMON_FLAGS -march=${RISC_ISA}"
   elif [ "$TARGET_OS" == "IOS" ]; then
     COMMON_FLAGS="$COMMON_FLAGS"
-  elif [ "$TARGET_OS" == "AIX" ] || [ "$TARGET_OS" == "SunOS" ]; then
-    # TODO: Not sure why we don't use -march=native on these OSes
-    if test "$USE_SSE"; then
-      TRY_SSE_ETC="1"
-    fi
   else
     COMMON_FLAGS="$COMMON_FLAGS -march=native "
   fi
 else
-  # PORTABLE=1
-  if test "$USE_SSE"; then
-    TRY_SSE_ETC="1"
-  fi
-
-  if test -n "`echo $TARGET_ARCHITECTURE | grep ^s390x`"; then
-    COMMON_FLAGS="$COMMON_FLAGS -march=z196 "
-  fi
-
-  if test -n "`echo $TARGET_ARCHITECTURE | grep ^riscv64`"; then
-    RISC_ISA=$(cat /proc/cpuinfo | grep isa | head -1 | cut --delimiter=: -f 2 | cut -b 2-)
-    COMMON_FLAGS="$COMMON_FLAGS -march=${RISC_ISA}"
+  # PORTABLE specified
+  if [ "$PORTABLE" == 1 ]; then
+    if test -n "`echo $TARGET_ARCHITECTURE | grep ^s390x`"; then
+      COMMON_FLAGS="$COMMON_FLAGS -march=z196 "
+    elif test -n "`echo $TARGET_ARCHITECTURE | grep ^riscv64`"; then
+      RISC_ISA=$(cat /proc/cpuinfo | grep isa | head -1 | cut --delimiter=: -f 2 | cut -b 2-)
+      COMMON_FLAGS="$COMMON_FLAGS -march=${RISC_ISA}"
+    elif test "$USE_SSE"; then
+      # USE_SSE is DEPRECATED
+      # This is a rough approximation of the old USE_SSE behavior
+      COMMON_FLAGS="$COMMON_FLAGS -march=haswell"
+    fi
+    # Other than those cases, not setting -march= here.
+  else
+    # Assume PORTABLE is a minimum assumed cpu type, e.g. PORTABLE=haswell
+    COMMON_FLAGS="$COMMON_FLAGS -march=${PORTABLE}"
   fi
 
   if [[ "${PLATFORM}" == "OS_MACOSX" ]]; then
@@ -704,101 +697,6 @@ EOF
   fi
 fi
 
-if test "$TRY_SSE_ETC"; then
-  # The USE_SSE flag now means "attempt to compile with widely-available
-  # Intel architecture extensions utilized by specific optimizations in the
-  # source code." It's a qualifier on PORTABLE=1 that means "mostly portable."
-  # It doesn't even really check that your current CPU is compatible.
-  #
-  # SSE4.2 available since nehalem, ca. 2008-2010
-  # Includes POPCNT for BitsSetToOne, BitParity
-  TRY_SSE42="-msse4.2"
-  # PCLMUL available since westmere, ca. 2010-2011
-  TRY_PCLMUL="-mpclmul"
-  # AVX2 available since haswell, ca. 2013-2015
-  TRY_AVX2="-mavx2"
-  # BMI available since haswell, ca. 2013-2015
-  # Primarily for TZCNT for CountTrailingZeroBits
-  TRY_BMI="-mbmi"
-  # LZCNT available since haswell, ca. 2013-2015
-  # For FloorLog2
-  TRY_LZCNT="-mlzcnt"
-fi
-
-$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_SSE42 -x c++ - -o test.o 2>/dev/null <<EOF
-  #include <cstdint>
-  #include <nmmintrin.h>
-  int main() {
-    volatile uint32_t x = _mm_crc32_u32(0, 0);
-    (void)x;
-  }
-EOF
-if [ "$?" = 0 ]; then
-  COMMON_FLAGS="$COMMON_FLAGS $TRY_SSE42 -DHAVE_SSE42"
-elif test "$USE_SSE"; then
-  echo "warning: USE_SSE specified but compiler could not use SSE intrinsics, disabling" >&2
-fi
-
-$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_PCLMUL -x c++ - -o test.o 2>/dev/null <<EOF
-  #include <cstdint>
-  #include <wmmintrin.h>
-  int main() {
-    const auto a = _mm_set_epi64x(0, 0);
-    const auto b = _mm_set_epi64x(0, 0);
-    const auto c = _mm_clmulepi64_si128(a, b, 0x00);
-    auto d = _mm_cvtsi128_si64(c);
-    (void)d;
-  }
-EOF
-if [ "$?" = 0 ]; then
-  COMMON_FLAGS="$COMMON_FLAGS $TRY_PCLMUL -DHAVE_PCLMUL"
-elif test "$USE_SSE"; then
-  echo "warning: USE_SSE specified but compiler could not use PCLMUL intrinsics, disabling" >&2
-fi
-
-$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_AVX2 -x c++ - -o test.o 2>/dev/null <<EOF
-  #include <cstdint>
-  #include <immintrin.h>
-  int main() {
-    const auto a = _mm256_setr_epi32(0, 1, 2, 3, 4, 7, 6, 5);
-    const auto b = _mm256_permutevar8x32_epi32(a, a);
-    (void)b;
-  }
-EOF
-if [ "$?" = 0 ]; then
-  COMMON_FLAGS="$COMMON_FLAGS $TRY_AVX2 -DHAVE_AVX2"
-elif test "$USE_SSE"; then
-  echo "warning: USE_SSE specified but compiler could not use AVX2 intrinsics, disabling" >&2
-fi
-
-$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_BMI -x c++ - -o test.o 2>/dev/null <<EOF
-  #include <cstdint>
-  #include <immintrin.h>
-  int main(int argc, char *argv[]) {
-    (void)argv;
-    return (int)_tzcnt_u64((uint64_t)argc);
-  }
-EOF
-if [ "$?" = 0 ]; then
-  COMMON_FLAGS="$COMMON_FLAGS $TRY_BMI -DHAVE_BMI"
-elif test "$USE_SSE"; then
-  echo "warning: USE_SSE specified but compiler could not use BMI intrinsics, disabling" >&2
-fi
-
-$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_LZCNT -x c++ - -o test.o 2>/dev/null <<EOF
-  #include <cstdint>
-  #include <immintrin.h>
-  int main(int argc, char *argv[]) {
-    (void)argv;
-    return (int)_lzcnt_u64((uint64_t)argc);
-  }
-EOF
-if [ "$?" = 0 ]; then
-  COMMON_FLAGS="$COMMON_FLAGS $TRY_LZCNT -DHAVE_LZCNT"
-elif test "$USE_SSE"; then
-  echo "warning: USE_SSE specified but compiler could not use LZCNT intrinsics, disabling" >&2
-fi
-
 $CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS -x c++ - -o test.o 2>/dev/null <<EOF
   #include <cstdint>
   int main() {

build_tools/dependencies_platform009.sh

@@ -1,22 +0,0 @@
-# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
-GCC_BASE=/mnt/gvfs/third-party2/gcc/1795efe5f06778c15a92c8f9a2aba5dc496d9d4d/9.x/centos7-native/3bed279
-CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/7318eaac22659b6ff2fe43918e4b69fd0772a8a7/9.0.0/platform009/651ee30
-LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/4959b39cfbe5965a37c861c4c327fa7c5c759b87/9.x/platform009/9202ce7
-GLIBC_BASE=/mnt/gvfs/third-party2/glibc/45ce3375cdc77ecb2520bbf8f0ecddd3f98efd7a/2.30/platform009/f259413
-SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/be4de3205e029101b18aa8103daa696c2bef3b19/1.1.3/platform009/7f3b187
-ZLIB_BASE=/mnt/gvfs/third-party2/zlib/3c160ac5c67e257501e24c6c1d00ad5e01d73db6/1.2.8/platform009/7f3b187
-BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/73a237ac5bc0a5f5d67b39b8d253cfebaab88684/1.0.6/platform009/7f3b187
-LZ4_BASE=/mnt/gvfs/third-party2/lz4/6ca38d3c390be2774d61a300f151464bbd632d62/1.9.1/platform009/7f3b187
-ZSTD_BASE=/mnt/gvfs/third-party2/zstd/64c58a207d2495e83abc57a500a956df09b79a7c/1.4.x/platform009/ba86d1f
-GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/824d0a8a5abb5b121afd1b35fc3896407ea50092/2.2.0/platform009/7f3b187
-JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/b62912d333ef33f9760efa6219dbe3fe6abb3b0e/master/platform009/c305944
-NUMA_BASE=/mnt/gvfs/third-party2/numa/0af65f71e23a67bf65dc91b11f95caa39325c432/2.0.11/platform009/7f3b187
-LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/02486dac347645d31dce116f44e1de3177315be2/1.4/platform009/5191652
-TBB_BASE=/mnt/gvfs/third-party2/tbb/2e0ec671e550bfca347300bf3f789d9c0fff24ad/2018_U5/platform009/7f3b187
-LIBURING_BASE=/mnt/gvfs/third-party2/liburing/70dbd9cfee63a25611417d09433a86d7711b3990/20200729/platform009/7f3b187
-KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/32b8a2407b634df3f8f948ba373fc4acc6a18296/fb/platform009/da39a3e
-BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/08634589372fa5f237bfd374e8c644a8364e78c1/2.32/platform009/ba86d1f/
-VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/6ae525939ad02e5e676855082fbbc7828dbafeac/3.15.0/platform009/7f3b187
-LUA_BASE=/mnt/gvfs/third-party2/lua/162efd9561a3d21f6869f4814011e9cf1b3ff4dc/5.3.4/platform009/a6271c4
-BENCHMARK_BASE=/mnt/gvfs/third-party2/benchmark/30bf49ad6414325e17f3425b0edcb64239427ae3/1.6.1/platform009/7f3b187
-GLOG_BASE=/mnt/gvfs/third-party2/glog/32d751bd5673375b438158717ab6a57c1cc57e3d/0.3.2_fb/platform009/10a364d

build_tools/fbcode_config.sh

@@ -147,7 +147,7 @@ else
 fi
 
 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DHAVE_SSE42"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT"
 CXXFLAGS+=" $CFLAGS"
 
 EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"

build_tools/fbcode_config_platform009.sh

@@ -1,170 +0,0 @@
-#!/bin/sh
-# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
-#
-# Set environment variables so that we can compile rocksdb using
-# fbcode settings.  It uses the latest g++ and clang compilers and also
-# uses jemalloc
-# Environment variables that change the behavior of this script:
-# PIC_BUILD -- if true, it will only take pic versions of libraries from fbcode. libraries that don't have pic variant will not be included
-
-
-BASEDIR=`dirname $BASH_SOURCE`
-source "$BASEDIR/dependencies_platform009.sh"
-
-CFLAGS=""
-
-# libgcc
-LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/9.3.0 -I $LIBGCC_BASE/include/c++/9.3.0/backward"
-LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"
-
-# glibc
-GLIBC_INCLUDE="$GLIBC_BASE/include"
-GLIBC_LIBS=" -L $GLIBC_BASE/lib"
-
-if test -z $PIC_BUILD; then
-  MAYBE_PIC=
-else
-  MAYBE_PIC=_pic
-fi
-
-if ! test $ROCKSDB_DISABLE_SNAPPY; then
-  # snappy
-  SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
-  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy${MAYBE_PIC}.a"
-  CFLAGS+=" -DSNAPPY"
-fi
-
-if ! test $ROCKSDB_DISABLE_ZLIB; then
-  # location of zlib headers and libraries
-  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
-  ZLIB_LIBS=" $ZLIB_BASE/lib/libz${MAYBE_PIC}.a"
-  CFLAGS+=" -DZLIB"
-fi
-
-if ! test $ROCKSDB_DISABLE_BZIP; then
-  # location of bzip headers and libraries
-  BZIP_INCLUDE=" -I $BZIP2_BASE/include/"
-  BZIP_LIBS=" $BZIP2_BASE/lib/libbz2${MAYBE_PIC}.a"
-  CFLAGS+=" -DBZIP2"
-fi
-
-if ! test $ROCKSDB_DISABLE_LZ4; then
-  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
-  LZ4_LIBS=" $LZ4_BASE/lib/liblz4${MAYBE_PIC}.a"
-  CFLAGS+=" -DLZ4"
-fi
-
-if ! test $ROCKSDB_DISABLE_ZSTD; then
-  ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
-  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd${MAYBE_PIC}.a"
-  CFLAGS+=" -DZSTD"
-fi
-
-# location of gflags headers and libraries
-GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
-GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags${MAYBE_PIC}.a"
-CFLAGS+=" -DGFLAGS=gflags"
-
-BENCHMARK_INCLUDE=" -I $BENCHMARK_BASE/include/"
-BENCHMARK_LIBS=" $BENCHMARK_BASE/lib/libbenchmark${MAYBE_PIC}.a"
-
-GLOG_INCLUDE=" -I $GLOG_BASE/include/"
-GLOG_LIBS=" $GLOG_BASE/lib/libglog${MAYBE_PIC}.a"
-
-# location of jemalloc
-JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
-JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc${MAYBE_PIC}.a"
-
-# location of numa
-NUMA_INCLUDE=" -I $NUMA_BASE/include/"
-NUMA_LIB=" $NUMA_BASE/lib/libnuma${MAYBE_PIC}.a"
-CFLAGS+=" -DNUMA"
-
-# location of libunwind
-LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind${MAYBE_PIC}.a"
-
-# location of TBB
-TBB_INCLUDE=" -isystem $TBB_BASE/include/"
-TBB_LIBS="$TBB_BASE/lib/libtbb${MAYBE_PIC}.a"
-CFLAGS+=" -DTBB"
-
-# location of LIBURING
-LIBURING_INCLUDE=" -isystem $LIBURING_BASE/include/"
-LIBURING_LIBS="$LIBURING_BASE/lib/liburing${MAYBE_PIC}.a"
-CFLAGS+=" -DLIBURING"
-
-test "$USE_SSE" || USE_SSE=1
-export USE_SSE
-test "$PORTABLE" || PORTABLE=1
-export PORTABLE
-
-BINUTILS="$BINUTILS_BASE/bin"
-AR="$BINUTILS/ar"
-AS="$BINUTILS/as"
-
-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE $LIBURING_INCLUDE $BENCHMARK_INCLUDE $GLOG_INCLUDE"
-
-STDLIBS="-L $GCC_BASE/lib64"
-
-CLANG_BIN="$CLANG_BASE/bin"
-CLANG_LIB="$CLANG_BASE/lib"
-CLANG_SRC="$CLANG_BASE/../../src"
-
-CLANG_ANALYZER="$CLANG_BIN/clang++"
-CLANG_SCAN_BUILD="$CLANG_SRC/llvm/clang/tools/scan-build/bin/scan-build"
-
-if [ -z "$USE_CLANG" ]; then
-  # gcc
-  CC="$GCC_BASE/bin/gcc"
-  CXX="$GCC_BASE/bin/g++"
-  AR="$GCC_BASE/bin/gcc-ar"
-
-  CFLAGS+=" -B$BINUTILS"
-  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
-  JEMALLOC=1
-else
-  # clang
-  CLANG_INCLUDE="$CLANG_LIB/clang/stable/include"
-  CC="$CLANG_BIN/clang"
-  CXX="$CLANG_BIN/clang++"
-  AR="$CLANG_BIN/llvm-ar"
-
-  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"
-
-  CFLAGS+=" -B$BINUTILS -nostdinc -nostdlib"
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/9.x "
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/9.x/x86_64-facebook-linux "
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
-  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
-  CFLAGS+=" -isystem $CLANG_INCLUDE"
-  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE/linux "
-  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE "
-  CFLAGS+=" -Wno-expansion-to-defined "
-  CXXFLAGS="-nostdinc++"
-fi
-
-CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DHAVE_SSE42 -DROCKSDB_IOURING_PRESENT"
-CXXFLAGS+=" $CFLAGS"
-
-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS $LIBURING_LIBS $BENCHMARK_LIBS"
-EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform009/lib/ld.so"
-EXEC_LDFLAGS+=" $LIBUNWIND"
-EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform009/lib"
-EXEC_LDFLAGS+=" -Wl,-rpath=$GCC_BASE/lib64"
-# required by libtbb
-EXEC_LDFLAGS+=" -ldl"
-
-PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
-PLATFORM_LDFLAGS+=" -B$BINUTILS"
-
-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS $LIBURING_LIBS $BENCHMARK_LIBS"
-
-VALGRIND_VER="$VALGRIND_BASE/bin/"
-
-# lua not supported because it's on track for deprecation, I think
-LUA_PATH=
-LUA_LIB=
-
-export CC CXX AR AS CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD LUA_PATH LUA_LIB

build_tools/fbcode_config_platform010.sh

@@ -154,7 +154,7 @@ CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE/linux "
 CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE "
 
 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DHAVE_SSE42 -DROCKSDB_IOURING_PRESENT"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_IOURING_PRESENT"
 CXXFLAGS+=" $CFLAGS"
 
 EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS $LIBURING_LIBS $BENCHMARK_LIBS"

build_tools/regression_build_test.sh

@@ -360,7 +360,7 @@ function send_to_ods {
     echo >&2 "ERROR: Key $key doesn't have a value."
     return
   fi
-  curl --silent "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=$key&value=$value" \
+  curl --silent "https://www.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=$key&value=$value" \
     --connect-timeout 60
 }
 

build_tools/update_dependencies.sh

@@ -104,46 +104,3 @@ get_lib_base valgrind   LATEST platform010
 get_lib_base lua        5.3.4  platform010
 
 git diff $OUTPUT
-
-
-###########################################################
-#                platform009 dependencies                 #
-###########################################################
-
-OUTPUT="$BASEDIR/dependencies_platform009.sh"
-
-rm -f "$OUTPUT"
-touch "$OUTPUT"
-
-echo "Writing dependencies to $OUTPUT"
-
-# Compilers locations
-GCC_BASE=`readlink -f $TP2_LATEST/gcc/9.x/centos7-native/*/`
-CLANG_BASE=`readlink -f $TP2_LATEST/llvm-fb/9.0.0/platform009/*/`
-
-log_header
-log_variable GCC_BASE
-log_variable CLANG_BASE
-
-# Libraries locations
-get_lib_base libgcc     9.x     platform009
-get_lib_base glibc      2.30    platform009
-get_lib_base snappy     LATEST  platform009
-get_lib_base zlib       LATEST  platform009
-get_lib_base bzip2      LATEST  platform009
-get_lib_base lz4        LATEST  platform009
-get_lib_base zstd       LATEST  platform009
-get_lib_base gflags     LATEST  platform009
-get_lib_base jemalloc   LATEST  platform009
-get_lib_base numa       LATEST  platform009
-get_lib_base libunwind  LATEST  platform009
-get_lib_base tbb        2018_U5 platform009
-get_lib_base liburing   LATEST  platform009
-get_lib_base benchmark  LATEST  platform009
-
-get_lib_base kernel-headers fb platform009
-get_lib_base binutils   LATEST centos7-native
-get_lib_base valgrind   LATEST platform009
-get_lib_base lua        5.3.4  platform009
-
-git diff $OUTPUT

cache/cache.cc

@@ -16,7 +16,8 @@
 #include "util/string_util.h"
 
 namespace ROCKSDB_NAMESPACE {
-#ifndef ROCKSDB_LITE
+const Cache::CacheItemHelper kNoopCacheItemHelper{};
+
 static std::unordered_map<std::string, OptionTypeInfo>
     lru_cache_options_type_info = {
         {"capacity",
@@ -64,7 +65,6 @@ static std::unordered_map<std::string, OptionTypeInfo>
           OptionType::kBoolean, OptionVerificationType::kNormal,
           OptionTypeFlags::kMutable}},
 };
-#endif  // ROCKSDB_LITE
 
 Status SecondaryCache::CreateFromString(
     const ConfigOptions& config_options, const std::string& value,
@@ -75,7 +75,6 @@ Status SecondaryCache::CreateFromString(
     Status status;
     std::shared_ptr<SecondaryCache> sec_cache;
 
-#ifndef ROCKSDB_LITE
     CompressedSecondaryCacheOptions sec_cache_opts;
     status = OptionTypeInfo::ParseStruct(config_options, "",
                                          &comp_sec_cache_options_type_info, "",
@@ -84,19 +83,13 @@ Status SecondaryCache::CreateFromString(
       sec_cache = NewCompressedSecondaryCache(sec_cache_opts);
     }
 
-#else
-    (void)config_options;
-    status = Status::NotSupported(
-        "Cannot load compressed secondary cache in LITE mode ", args);
-#endif  //! ROCKSDB_LITE
 
     if (status.ok()) {
       result->swap(sec_cache);
     }
     return status;
   } else {
-    return LoadSharedObject<SecondaryCache>(config_options, value, nullptr,
-                                            result);
+    return LoadSharedObject<SecondaryCache>(config_options, value, result);
   }
 }
 
@@ -108,7 +101,6 @@ Status Cache::CreateFromString(const ConfigOptions& config_options,
   if (value.find('=') == std::string::npos) {
     cache = NewLRUCache(ParseSizeT(value));
   } else {
-#ifndef ROCKSDB_LITE
     LRUCacheOptions cache_opts;
     status = OptionTypeInfo::ParseStruct(config_options, "",
                                          &lru_cache_options_type_info, "",
@@ -116,14 +108,51 @@ Status Cache::CreateFromString(const ConfigOptions& config_options,
     if (status.ok()) {
       cache = NewLRUCache(cache_opts);
     }
-#else
-    (void)config_options;
-    status = Status::NotSupported("Cannot load cache in LITE mode ", value);
-#endif  //! ROCKSDB_LITE
   }
   if (status.ok()) {
     result->swap(cache);
   }
   return status;
 }
+
+bool Cache::AsyncLookupHandle::IsReady() {
+  return pending_handle == nullptr || pending_handle->IsReady();
+}
+
+bool Cache::AsyncLookupHandle::IsPending() { return pending_handle != nullptr; }
+
+Cache::Handle* Cache::AsyncLookupHandle::Result() {
+  assert(!IsPending());
+  return result_handle;
+}
+
+void Cache::StartAsyncLookup(AsyncLookupHandle& async_handle) {
+  async_handle.found_dummy_entry = false;  // in case re-used
+  assert(!async_handle.IsPending());
+  async_handle.result_handle =
+      Lookup(async_handle.key, async_handle.helper, async_handle.create_context,
+             async_handle.priority, async_handle.stats);
+}
+
+Cache::Handle* Cache::Wait(AsyncLookupHandle& async_handle) {
+  WaitAll(&async_handle, 1);
+  return async_handle.Result();
+}
+
+void Cache::WaitAll(AsyncLookupHandle* async_handles, size_t count) {
+  for (size_t i = 0; i < count; ++i) {
+    if (async_handles[i].IsPending()) {
+      // If a pending handle gets here, it should be marked at "to be handled
+      // by a caller" by that caller erasing the pending_cache on it.
+      assert(async_handles[i].pending_cache == nullptr);
+    }
+  }
+}
+
+void Cache::SetEvictionCallback(EvictionCallback&& fn) {
+  // Overwriting non-empty with non-empty could indicate a bug
+  assert(!eviction_callback_ || !fn);
+  eviction_callback_ = std::move(fn);
+}
+
 }  // namespace ROCKSDB_NAMESPACE

cache/cache_bench_tool.cc

@@ -16,7 +16,7 @@
 #include "db/db_impl/db_impl.h"
 #include "monitoring/histogram.h"
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/convenience.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
@@ -77,11 +77,9 @@ DEFINE_bool(lean, false,
             "If true, no additional computation is performed besides cache "
             "operations.");
 
-#ifndef ROCKSDB_LITE
 DEFINE_string(secondary_cache_uri, "",
               "Full URI for creating a custom secondary cache object");
 static class std::shared_ptr<ROCKSDB_NAMESPACE::SecondaryCache> secondary_cache;
-#endif  // ROCKSDB_LITE
 
 DEFINE_string(cache_type, "lru_cache", "Type of block cache.");
 
@@ -226,7 +224,7 @@ struct KeyGen {
   }
 };
 
-char* createValue(Random64& rnd) {
+Cache::ObjectPtr createValue(Random64& rnd) {
   char* rv = new char[FLAGS_value_bytes];
   // Fill with some filler data, and take some CPU time
   for (uint32_t i = 0; i < FLAGS_value_bytes; i += 8) {
@@ -236,28 +234,36 @@ char* createValue(Random64& rnd) {
 }
 
 // Callbacks for secondary cache
-size_t SizeFn(void* /*obj*/) { return FLAGS_value_bytes; }
+size_t SizeFn(Cache::ObjectPtr /*obj*/) { return FLAGS_value_bytes; }
 
-Status SaveToFn(void* obj, size_t /*offset*/, size_t size, void* out) {
-  memcpy(out, obj, size);
+Status SaveToFn(Cache::ObjectPtr from_obj, size_t /*from_offset*/,
+                size_t length, char* out) {
+  memcpy(out, from_obj, length);
   return Status::OK();
 }
 
-// Different deleters to simulate using deleter to gather
-// stats on the code origin and kind of cache entries.
-void deleter1(const Slice& /*key*/, void* value) {
-  delete[] static_cast<char*>(value);
-}
-void deleter2(const Slice& /*key*/, void* value) {
-  delete[] static_cast<char*>(value);
-}
-void deleter3(const Slice& /*key*/, void* value) {
+Status CreateFn(const Slice& data, Cache::CreateContext* /*context*/,
+                MemoryAllocator* /*allocator*/, Cache::ObjectPtr* out_obj,
+                size_t* out_charge) {
+  *out_obj = new char[data.size()];
+  memcpy(*out_obj, data.data(), data.size());
+  *out_charge = data.size();
+  return Status::OK();
+};
+
+void DeleteFn(Cache::ObjectPtr value, MemoryAllocator* /*alloc*/) {
   delete[] static_cast<char*>(value);
 }
 
-Cache::CacheItemHelper helper1(SizeFn, SaveToFn, deleter1);
-Cache::CacheItemHelper helper2(SizeFn, SaveToFn, deleter2);
-Cache::CacheItemHelper helper3(SizeFn, SaveToFn, deleter3);
+Cache::CacheItemHelper helper1_wos(CacheEntryRole::kDataBlock, DeleteFn);
+Cache::CacheItemHelper helper1(CacheEntryRole::kDataBlock, DeleteFn, SizeFn,
+                               SaveToFn, CreateFn, &helper1_wos);
+Cache::CacheItemHelper helper2_wos(CacheEntryRole::kIndexBlock, DeleteFn);
+Cache::CacheItemHelper helper2(CacheEntryRole::kIndexBlock, DeleteFn, SizeFn,
+                               SaveToFn, CreateFn, &helper2_wos);
+Cache::CacheItemHelper helper3_wos(CacheEntryRole::kFilterBlock, DeleteFn);
+Cache::CacheItemHelper helper3(CacheEntryRole::kFilterBlock, DeleteFn, SizeFn,
+                               SaveToFn, CreateFn, &helper3_wos);
 }  // namespace
 
 class CacheBench {
@@ -300,7 +306,6 @@ class CacheBench {
       LRUCacheOptions opts(FLAGS_cache_size, FLAGS_num_shard_bits,
                            false /* strict_capacity_limit */,
                            0.5 /* high_pri_pool_ratio */);
-#ifndef ROCKSDB_LITE
       if (!FLAGS_secondary_cache_uri.empty()) {
         Status s = SecondaryCache::CreateFromString(
             ConfigOptions(), FLAGS_secondary_cache_uri, &secondary_cache);
@@ -313,7 +318,6 @@ class CacheBench {
         }
         opts.secondary_cache = secondary_cache;
       }
-#endif  // ROCKSDB_LITE
 
       cache_ = NewLRUCache(opts);
     } else {
@@ -436,7 +440,7 @@ class CacheBench {
     uint64_t total_entry_count = 0;
     uint64_t table_occupancy = 0;
     uint64_t table_size = 0;
-    std::set<Cache::DeleterFn> deleters;
+    std::set<const Cache::CacheItemHelper*> helpers;
     StopWatchNano timer(clock);
 
     for (;;) {
@@ -461,7 +465,7 @@ class CacheBench {
                  << BytesToHumanString(static_cast<uint64_t>(
                         1.0 * total_charge / total_entry_count))
                  << "\n"
-                 << "Unique deleters: " << deleters.size() << "\n";
+                 << "Unique helpers: " << helpers.size() << "\n";
             *stats_report = ostr.str();
             return;
           }
@@ -477,14 +481,14 @@ class CacheBench {
       total_key_size = 0;
       total_charge = 0;
       total_entry_count = 0;
-      deleters.clear();
-      auto fn = [&](const Slice& key, void* /*value*/, size_t charge,
-                    Cache::DeleterFn deleter) {
+      helpers.clear();
+      auto fn = [&](const Slice& key, Cache::ObjectPtr /*value*/, size_t charge,
+                    const Cache::CacheItemHelper* helper) {
         total_key_size += key.size();
         total_charge += charge;
         ++total_entry_count;
-        // Something slightly more expensive as in (future) stats by category
-        deleters.insert(deleter);
+        // Something slightly more expensive as in stats by category
+        helpers.insert(helper);
       };
       timer.Start();
       Cache::ApplyToAllEntriesOptions opts;
@@ -533,14 +537,6 @@ class CacheBench {
     for (uint64_t i = 0; i < FLAGS_ops_per_thread; i++) {
       Slice key = gen.GetRand(thread->rnd, max_key_, max_log_);
       uint64_t random_op = thread->rnd.Next();
-      Cache::CreateCallback create_cb = [](const void* buf, size_t size,
-                                           void** out_obj,
-                                           size_t* charge) -> Status {
-        *out_obj = reinterpret_cast<void*>(new char[size]);
-        memcpy(*out_obj, buf, size);
-        *charge = size;
-        return Status::OK();
-      };
 
       timer.Start();
 
@@ -550,8 +546,8 @@ class CacheBench {
           handle = nullptr;
         }
         // do lookup
-        handle = cache_->Lookup(key, &helper2, create_cb, Cache::Priority::LOW,
-                                true);
+        handle = cache_->Lookup(key, &helper2, /*context*/ nullptr,
+                                Cache::Priority::LOW);
         if (handle) {
           if (!FLAGS_lean) {
             // do something with the data
@@ -579,8 +575,8 @@ class CacheBench {
           handle = nullptr;
         }
         // do lookup
-        handle = cache_->Lookup(key, &helper2, create_cb, Cache::Priority::LOW,
-                                true);
+        handle = cache_->Lookup(key, &helper2, /*context*/ nullptr,
+                                Cache::Priority::LOW);
         if (handle) {
           if (!FLAGS_lean) {
             // do something with the data

cache/cache_entry_roles.cc

@@ -101,34 +101,4 @@ std::string BlockCacheEntryStatsMapKeys::UsedPercent(CacheEntryRole role) {
   return GetPrefixedCacheEntryRoleName(kPrefix, role);
 }
 
-namespace {
-
-struct Registry {
-  std::mutex mutex;
-  UnorderedMap<Cache::DeleterFn, CacheEntryRole> role_map;
-  void Register(Cache::DeleterFn fn, CacheEntryRole role) {
-    std::lock_guard<std::mutex> lock(mutex);
-    role_map[fn] = role;
-  }
-  UnorderedMap<Cache::DeleterFn, CacheEntryRole> Copy() {
-    std::lock_guard<std::mutex> lock(mutex);
-    return role_map;
-  }
-};
-
-Registry& GetRegistry() {
-  STATIC_AVOID_DESTRUCTION(Registry, registry);
-  return registry;
-}
-
-}  // namespace
-
-void RegisterCacheDeleterRole(Cache::DeleterFn fn, CacheEntryRole role) {
-  GetRegistry().Register(fn, role);
-}
-
-UnorderedMap<Cache::DeleterFn, CacheEntryRole> CopyCacheDeleterRoleMap() {
-  return GetRegistry().Copy();
-}
-
 }  // namespace ROCKSDB_NAMESPACE

cache/cache_entry_roles.h

@@ -7,11 +7,8 @@
 
 #include <array>
 #include <cstdint>
-#include <memory>
-#include <type_traits>
 
 #include "rocksdb/cache.h"
-#include "util/hash_containers.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -20,84 +17,4 @@ extern std::array<std::string, kNumCacheEntryRoles>
 extern std::array<std::string, kNumCacheEntryRoles>
     kCacheEntryRoleToHyphenString;
 
-// To associate cache entries with their role, we use a hack on the
-// existing Cache interface. Because the deleter of an entry can authenticate
-// the code origin of an entry, we can elaborate the choice of deleter to
-// also encode role information, without inferring false role information
-// from entries not choosing to encode a role.
-//
-// The rest of this file is for handling mappings between deleters and
-// roles.
-
-// To infer a role from a deleter, the deleter must be registered. This
-// can be done "manually" with this function. This function is thread-safe,
-// and the registration mappings go into private but static storage. (Note
-// that DeleterFn is a function pointer, not std::function. Registrations
-// should not be too many.)
-void RegisterCacheDeleterRole(Cache::DeleterFn fn, CacheEntryRole role);
-
-// Gets a copy of the registered deleter -> role mappings. This is the only
-// function for reading the mappings made with RegisterCacheDeleterRole.
-// Why only this interface for reading?
-// * This function has to be thread safe, which could incur substantial
-// overhead. We should not pay this overhead for every deleter look-up.
-// * This is suitable for preparing for batch operations, like with
-// CacheEntryStatsCollector.
-// * The number of mappings should be sufficiently small (dozens).
-UnorderedMap<Cache::DeleterFn, CacheEntryRole> CopyCacheDeleterRoleMap();
-
-// ************************************************************** //
-// An automatic registration infrastructure. This enables code
-// to simply ask for a deleter associated with a particular type
-// and role, and registration is automatic. In a sense, this is
-// a small dependency injection infrastructure, because linking
-// in new deleter instantiations is essentially sufficient for
-// making stats collection (using CopyCacheDeleterRoleMap) aware
-// of them.
-
-namespace cache_entry_roles_detail {
-
-template <typename T, CacheEntryRole R>
-struct RegisteredDeleter {
-  RegisteredDeleter() { RegisterCacheDeleterRole(Delete, R); }
-
-  // These have global linkage to help ensure compiler optimizations do not
-  // break uniqueness for each <T,R>
-  static void Delete(const Slice& /* key */, void* value) {
-    // Supports T == Something[], unlike delete operator
-    std::default_delete<T>()(
-        static_cast<typename std::remove_extent<T>::type*>(value));
-  }
-};
-
-template <CacheEntryRole R>
-struct RegisteredNoopDeleter {
-  RegisteredNoopDeleter() { RegisterCacheDeleterRole(Delete, R); }
-
-  static void Delete(const Slice& /* key */, void* /* value */) {
-    // Here was `assert(value == nullptr);` but we can also put pointers
-    // to static data in Cache, for testing at least.
-  }
-};
-
-}  // namespace cache_entry_roles_detail
-
-// Get an automatically registered deleter for value type T and role R.
-// Based on C++ semantics, registration is invoked exactly once in a
-// thread-safe way on first call to this function, for each <T, R>.
-template <typename T, CacheEntryRole R>
-Cache::DeleterFn GetCacheEntryDeleterForRole() {
-  static cache_entry_roles_detail::RegisteredDeleter<T, R> reg;
-  return reg.Delete;
-}
-
-// Get an automatically registered no-op deleter (value should be nullptr)
-// and associated with role R. This is used for Cache "reservation" entries
-// such as for WriteBufferManager.
-template <CacheEntryRole R>
-Cache::DeleterFn GetNoopDeleterForRole() {
-  static cache_entry_roles_detail::RegisteredNoopDeleter<R> reg;
-  return reg.Delete;
-}
-
 }  // namespace ROCKSDB_NAMESPACE

cache/cache_entry_stats.h

@@ -10,8 +10,8 @@
 #include <memory>
 #include <mutex>
 
-#include "cache/cache_helpers.h"
 #include "cache/cache_key.h"
+#include "cache/typed_cache.h"
 #include "port/lang.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/status.h"
@@ -111,11 +111,14 @@ class CacheEntryStatsCollector {
   // Gets or creates a shared instance of CacheEntryStatsCollector in the
   // cache itself, and saves into `ptr`. This shared_ptr will hold the
   // entry in cache until all refs are destroyed.
-  static Status GetShared(Cache *cache, SystemClock *clock,
+  static Status GetShared(Cache *raw_cache, SystemClock *clock,
                           std::shared_ptr<CacheEntryStatsCollector> *ptr) {
-    const Slice &cache_key = GetCacheKey();
+    assert(raw_cache);
+    BasicTypedCacheInterface<CacheEntryStatsCollector, CacheEntryRole::kMisc>
+        cache{raw_cache};
 
-    Cache::Handle *h = cache->Lookup(cache_key);
+    const Slice &cache_key = GetCacheKey();
+    auto h = cache.Lookup(cache_key);
     if (h == nullptr) {
       // Not yet in cache, but Cache doesn't provide a built-in way to
       // avoid racing insert. So we double-check under a shared mutex,
@@ -123,15 +126,15 @@ class CacheEntryStatsCollector {
       STATIC_AVOID_DESTRUCTION(std::mutex, static_mutex);
       std::lock_guard<std::mutex> lock(static_mutex);
 
-      h = cache->Lookup(cache_key);
+      h = cache.Lookup(cache_key);
       if (h == nullptr) {
-        auto new_ptr = new CacheEntryStatsCollector(cache, clock);
+        auto new_ptr = new CacheEntryStatsCollector(cache.get(), clock);
         // TODO: non-zero charge causes some tests that count block cache
         // usage to go flaky. Fix the problem somehow so we can use an
         // accurate charge.
         size_t charge = 0;
-        Status s = cache->Insert(cache_key, new_ptr, charge, Deleter, &h,
-                                 Cache::Priority::HIGH);
+        Status s =
+            cache.Insert(cache_key, new_ptr, charge, &h, Cache::Priority::HIGH);
         if (!s.ok()) {
           assert(h == nullptr);
           delete new_ptr;
@@ -140,11 +143,11 @@ class CacheEntryStatsCollector {
       }
     }
     // If we reach here, shared entry is in cache with handle `h`.
-    assert(cache->GetDeleter(h) == Deleter);
+    assert(cache.get()->GetCacheItemHelper(h) == cache.GetBasicHelper());
 
     // Build an aliasing shared_ptr that keeps `ptr` in cache while there
     // are references.
-    *ptr = MakeSharedCacheHandleGuard<CacheEntryStatsCollector>(cache, h);
+    *ptr = cache.SharedGuard(h);
     return Status::OK();
   }
 
@@ -157,10 +160,6 @@ class CacheEntryStatsCollector {
         cache_(cache),
         clock_(clock) {}
 
-  static void Deleter(const Slice &, void *value) {
-    delete static_cast<CacheEntryStatsCollector *>(value);
-  }
-
   static const Slice &GetCacheKey() {
     // For each template instantiation
     static CacheKey ckey = CacheKey::CreateUniqueForProcessLifetime();

cache/cache_helpers.cc

@@ -0,0 +1,40 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#include "cache/cache_helpers.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+void ReleaseCacheHandleCleanup(void* arg1, void* arg2) {
+  Cache* const cache = static_cast<Cache*>(arg1);
+  assert(cache);
+
+  Cache::Handle* const cache_handle = static_cast<Cache::Handle*>(arg2);
+  assert(cache_handle);
+
+  cache->Release(cache_handle);
+}
+
+Status WarmInCache(Cache* cache, const Slice& key, const Slice& saved,
+                   Cache::CreateContext* create_context,
+                   const Cache::CacheItemHelper* helper,
+                   Cache::Priority priority, size_t* out_charge) {
+  assert(helper);
+  assert(helper->create_cb);
+  Cache::ObjectPtr value;
+  size_t charge;
+  Status st = helper->create_cb(saved, create_context,
+                                cache->memory_allocator(), &value, &charge);
+  if (st.ok()) {
+    st =
+        cache->Insert(key, value, helper, charge, /*handle*/ nullptr, priority);
+    if (out_charge) {
+      *out_charge = charge;
+    }
+  }
+  return st;
+}
+
+}  // namespace ROCKSDB_NAMESPACE

cache/cache_helpers.h

@@ -7,7 +7,7 @@
 
 #include <cassert>
 
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/rocksdb_namespace.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -17,22 +17,17 @@ template <typename T>
 T* GetFromCacheHandle(Cache* cache, Cache::Handle* handle) {
   assert(cache);
   assert(handle);
-
   return static_cast<T*>(cache->Value(handle));
 }
 
-// Simple generic deleter for Cache (to be used with Cache::Insert).
-template <typename T>
-void DeleteCacheEntry(const Slice& /* key */, void* value) {
-  delete static_cast<T*>(value);
-}
-
 // Turns a T* into a Slice so it can be used as a key with Cache.
 template <typename T>
-Slice GetSlice(const T* t) {
+Slice GetSliceForKey(const T* t) {
   return Slice(reinterpret_cast<const char*>(t), sizeof(T));
 }
 
+void ReleaseCacheHandleCleanup(void* arg1, void* arg2);
+
 // Generic resource management object for cache handles that releases the handle
 // when destroyed. Has unique ownership of the handle, so copying it is not
 // allowed, while moving it transfers ownership.
@@ -88,7 +83,7 @@ class CacheHandleGuard {
     if (cleanable) {
       if (handle_ != nullptr) {
         assert(cache_);
-        cleanable->RegisterCleanup(&ReleaseCacheHandle, cache_, handle_);
+        cleanable->RegisterCleanup(&ReleaseCacheHandleCleanup, cache_, handle_);
       }
     }
     ResetFields();
@@ -115,16 +110,6 @@ class CacheHandleGuard {
     value_ = nullptr;
   }
 
-  static void ReleaseCacheHandle(void* arg1, void* arg2) {
-    Cache* const cache = static_cast<Cache*>(arg1);
-    assert(cache);
-
-    Cache::Handle* const cache_handle = static_cast<Cache::Handle*>(arg2);
-    assert(cache_handle);
-
-    cache->Release(cache_handle);
-  }
-
  private:
   Cache* cache_ = nullptr;
   Cache::Handle* handle_ = nullptr;
@@ -139,7 +124,16 @@ template <typename T>
 std::shared_ptr<T> MakeSharedCacheHandleGuard(Cache* cache,
                                               Cache::Handle* handle) {
   auto wrapper = std::make_shared<CacheHandleGuard<T>>(cache, handle);
-  return std::shared_ptr<T>(wrapper, static_cast<T*>(cache->Value(handle)));
+  return std::shared_ptr<T>(wrapper, GetFromCacheHandle<T>(cache, handle));
 }
 
+// Given the persistable data (saved) for a block cache entry, parse that
+// into a cache entry object and insert it into the given cache. The charge
+// of the new entry can be returned to the caller through `out_charge`.
+Status WarmInCache(Cache* cache, const Slice& key, const Slice& saved,
+                   Cache::CreateContext* create_context,
+                   const Cache::CacheItemHelper* helper,
+                   Cache::Priority priority = Cache::Priority::LOW,
+                   size_t* out_charge = nullptr);
+
 }  // namespace ROCKSDB_NAMESPACE

cache/cache_key.cc

@@ -8,7 +8,7 @@
 #include <algorithm>
 #include <atomic>
 
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "table/unique_id_impl.h"
 #include "util/hash.h"
 #include "util/math.h"

cache/cache_reservation_manager.cc

@@ -13,7 +13,6 @@
 #include <cstring>
 #include <memory>
 
-#include "cache/cache_entry_roles.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
@@ -41,17 +40,17 @@ CacheReservationManagerImpl<
 template <CacheEntryRole R>
 CacheReservationManagerImpl<R>::CacheReservationManagerImpl(
     std::shared_ptr<Cache> cache, bool delayed_decrease)
-    : delayed_decrease_(delayed_decrease),
+    : cache_(cache),
+      delayed_decrease_(delayed_decrease),
       cache_allocated_size_(0),
       memory_used_(0) {
   assert(cache != nullptr);
-  cache_ = cache;
 }
 
 template <CacheEntryRole R>
 CacheReservationManagerImpl<R>::~CacheReservationManagerImpl() {
   for (auto* handle : dummy_handles_) {
-    cache_->Release(handle, true);
+    cache_.ReleaseAndEraseIfLastRef(handle);
   }
 }
 
@@ -115,8 +114,7 @@ Status CacheReservationManagerImpl<R>::IncreaseCacheReservation(
   Status return_status = Status::OK();
   while (new_mem_used > cache_allocated_size_.load(std::memory_order_relaxed)) {
     Cache::Handle* handle = nullptr;
-    return_status = cache_->Insert(GetNextCacheKey(), nullptr, kSizeDummyEntry,
-                                   GetNoopDeleterForRole<R>(), &handle);
+    return_status = cache_.Insert(GetNextCacheKey(), kSizeDummyEntry, &handle);
 
     if (return_status != Status::OK()) {
       return return_status;
@@ -141,7 +139,7 @@ Status CacheReservationManagerImpl<R>::DecreaseCacheReservation(
          cache_allocated_size_.load(std::memory_order_relaxed)) {
     assert(!dummy_handles_.empty());
     auto* handle = dummy_handles_.back();
-    cache_->Release(handle, true);
+    cache_.ReleaseAndEraseIfLastRef(handle);
     dummy_handles_.pop_back();
     cache_allocated_size_ -= kSizeDummyEntry;
   }
@@ -169,8 +167,9 @@ Slice CacheReservationManagerImpl<R>::GetNextCacheKey() {
 }
 
 template <CacheEntryRole R>
-Cache::DeleterFn CacheReservationManagerImpl<R>::TEST_GetNoopDeleterForRole() {
-  return GetNoopDeleterForRole<R>();
+const Cache::CacheItemHelper*
+CacheReservationManagerImpl<R>::TEST_GetCacheItemHelperForRole() {
+  return CacheInterface::GetHelper();
 }
 
 template class CacheReservationManagerImpl<

cache/cache_reservation_manager.h

@@ -18,7 +18,7 @@
 
 #include "cache/cache_entry_roles.h"
 #include "cache/cache_key.h"
-#include "rocksdb/cache.h"
+#include "cache/typed_cache.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
 #include "util/coding.h"
@@ -197,10 +197,10 @@ class CacheReservationManagerImpl
 
   static constexpr std::size_t GetDummyEntrySize() { return kSizeDummyEntry; }
 
-  // For testing only - it is to help ensure the NoopDeleterForRole<R>
+  // For testing only - it is to help ensure the CacheItemHelperForRole<R>
   // accessed from CacheReservationManagerImpl and the one accessed from the
   // test are from the same translation units
-  static Cache::DeleterFn TEST_GetNoopDeleterForRole();
+  static const Cache::CacheItemHelper *TEST_GetCacheItemHelperForRole();
 
  private:
   static constexpr std::size_t kSizeDummyEntry = 256 * 1024;
@@ -211,7 +211,8 @@ class CacheReservationManagerImpl
   Status IncreaseCacheReservation(std::size_t new_mem_used);
   Status DecreaseCacheReservation(std::size_t new_mem_used);
 
-  std::shared_ptr<Cache> cache_;
+  using CacheInterface = PlaceholderSharedCacheInterface<R>;
+  CacheInterface cache_;
   bool delayed_decrease_;
   std::atomic<std::size_t> cache_allocated_size_;
   std::size_t memory_used_;

cache/cache_test.cc

@@ -16,9 +16,12 @@
 #include <vector>
 
 #include "cache/lru_cache.h"
+#include "cache/typed_cache.h"
 #include "port/stack_trace.h"
+#include "test_util/secondary_cache_test_util.h"
 #include "test_util/testharness.h"
 #include "util/coding.h"
+#include "util/hash_containers.h"
 #include "util/string_util.h"
 
 // HyperClockCache only supports 16-byte keys, so some of the tests
@@ -55,42 +58,43 @@ int DecodeKey32Bits(const Slice& k) {
   return DecodeFixed32(k.data());
 }
 
-void* EncodeValue(uintptr_t v) { return reinterpret_cast<void*>(v); }
+Cache::ObjectPtr EncodeValue(uintptr_t v) {
+  return reinterpret_cast<Cache::ObjectPtr>(v);
+}
 
 int DecodeValue(void* v) {
   return static_cast<int>(reinterpret_cast<uintptr_t>(v));
 }
 
-void DumbDeleter(const Slice& /*key*/, void* /*value*/) {}
+const Cache::CacheItemHelper kDumbHelper{
+    CacheEntryRole::kMisc,
+    [](Cache::ObjectPtr /*value*/, MemoryAllocator* /*alloc*/) {}};
 
-void EraseDeleter1(const Slice& /*key*/, void* value) {
-  Cache* cache = reinterpret_cast<Cache*>(value);
-  cache->Erase("foo");
-}
-
-void EraseDeleter2(const Slice& /*key*/, void* value) {
-  Cache* cache = reinterpret_cast<Cache*>(value);
-  cache->Erase(EncodeKey16Bytes(1234));
-}
-
-const std::string kLRU = "lru";
-const std::string kHyperClock = "hyper_clock";
+const Cache::CacheItemHelper kEraseOnDeleteHelper1{
+    CacheEntryRole::kMisc,
+    [](Cache::ObjectPtr value, MemoryAllocator* /*alloc*/) {
+      Cache* cache = static_cast<Cache*>(value);
+      cache->Erase("foo");
+    }};
 
+const Cache::CacheItemHelper kEraseOnDeleteHelper2{
+    CacheEntryRole::kMisc,
+    [](Cache::ObjectPtr value, MemoryAllocator* /*alloc*/) {
+      Cache* cache = static_cast<Cache*>(value);
+      cache->Erase(EncodeKey16Bytes(1234));
+    }};
 }  // anonymous namespace
 
-class CacheTest : public testing::TestWithParam<std::string> {
+class CacheTest : public testing::Test,
+                  public secondary_cache_test_util::WithCacheTypeParam {
  public:
   static CacheTest* current_;
   static std::string type_;
 
-  static void Deleter(const Slice& key, void* v) {
-    if (type_ == kHyperClock) {
-      current_->deleted_keys_.push_back(DecodeKey16Bytes(key));
-    } else {
-      current_->deleted_keys_.push_back(DecodeKey32Bits(key));
-    }
+  static void Deleter(Cache::ObjectPtr v, MemoryAllocator*) {
     current_->deleted_values_.push_back(DecodeValue(v));
   }
+  static const Cache::CacheItemHelper kHelper;
 
   static const int kCacheSize = 1000;
   static const int kNumShardBits = 4;
@@ -98,13 +102,10 @@ class CacheTest : public testing::TestWithParam<std::string> {
   static const int kCacheSize2 = 100;
   static const int kNumShardBits2 = 2;
 
-  std::vector<int> deleted_keys_;
   std::vector<int> deleted_values_;
   std::shared_ptr<Cache> cache_;
   std::shared_ptr<Cache> cache2_;
 
-  size_t estimated_value_size_ = 1;
-
   CacheTest()
       : cache_(NewCache(kCacheSize, kNumShardBits, false)),
         cache2_(NewCache(kCacheSize2, kNumShardBits2, false)) {
@@ -114,41 +115,6 @@ class CacheTest : public testing::TestWithParam<std::string> {
 
   ~CacheTest() override {}
 
-  std::shared_ptr<Cache> NewCache(size_t capacity) {
-    auto type = GetParam();
-    if (type == kLRU) {
-      return NewLRUCache(capacity);
-    }
-    if (type == kHyperClock) {
-      return HyperClockCacheOptions(
-                 capacity, estimated_value_size_ /*estimated_value_size*/)
-          .MakeSharedCache();
-    }
-    return nullptr;
-  }
-
-  std::shared_ptr<Cache> NewCache(
-      size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-      CacheMetadataChargePolicy charge_policy = kDontChargeCacheMetadata) {
-    auto type = GetParam();
-    if (type == kLRU) {
-      LRUCacheOptions co;
-      co.capacity = capacity;
-      co.num_shard_bits = num_shard_bits;
-      co.strict_capacity_limit = strict_capacity_limit;
-      co.high_pri_pool_ratio = 0;
-      co.metadata_charge_policy = charge_policy;
-      return NewLRUCache(co);
-    }
-    if (type == kHyperClock) {
-      return HyperClockCacheOptions(capacity, 1 /*estimated_value_size*/,
-                                    num_shard_bits, strict_capacity_limit,
-                                    nullptr /*allocator*/, charge_policy)
-          .MakeSharedCache();
-    }
-    return nullptr;
-  }
-
   // These functions encode/decode keys in tests cases that use
   // int keys.
   // Currently, HyperClockCache requires keys to be 16B long, whereas
@@ -182,8 +148,8 @@ class CacheTest : public testing::TestWithParam<std::string> {
 
   void Insert(std::shared_ptr<Cache> cache, int key, int value,
               int charge = 1) {
-    EXPECT_OK(cache->Insert(EncodeKey(key), EncodeValue(value), charge,
-                            &CacheTest::Deleter));
+    EXPECT_OK(cache->Insert(EncodeKey(key), EncodeValue(value), &kHelper,
+                            charge, /*handle*/ nullptr, Cache::Priority::HIGH));
   }
 
   void Erase(std::shared_ptr<Cache> cache, int key) {
@@ -207,6 +173,9 @@ class CacheTest : public testing::TestWithParam<std::string> {
   void Erase2(int key) { Erase(cache2_, key); }
 };
 
+const Cache::CacheItemHelper CacheTest::kHelper{CacheEntryRole::kMisc,
+                                                &CacheTest::Deleter};
+
 CacheTest* CacheTest::current_;
 std::string CacheTest::type_;
 
@@ -236,10 +205,8 @@ TEST_P(CacheTest, UsageTest) {
       key = EncodeKey(i);
     }
     auto kv_size = key.size() + 5;
-    ASSERT_OK(cache->Insert(key, reinterpret_cast<void*>(value), kv_size,
-                            DumbDeleter));
-    ASSERT_OK(precise_cache->Insert(key, reinterpret_cast<void*>(value),
-                                    kv_size, DumbDeleter));
+    ASSERT_OK(cache->Insert(key, value, &kDumbHelper, kv_size));
+    ASSERT_OK(precise_cache->Insert(key, value, &kDumbHelper, kv_size));
     usage += kv_size;
     ASSERT_EQ(usage, cache->GetUsage());
     if (type == kHyperClock) {
@@ -262,10 +229,8 @@ TEST_P(CacheTest, UsageTest) {
     } else {
       key = EncodeKey(static_cast<int>(1000 + i));
     }
-    ASSERT_OK(cache->Insert(key, reinterpret_cast<void*>(value), key.size() + 5,
-                            DumbDeleter));
-    ASSERT_OK(precise_cache->Insert(key, reinterpret_cast<void*>(value),
-                                    key.size() + 5, DumbDeleter));
+    ASSERT_OK(cache->Insert(key, value, &kDumbHelper, key.size() + 5));
+    ASSERT_OK(precise_cache->Insert(key, value, &kDumbHelper, key.size() + 5));
   }
 
   // the usage should be close to the capacity
@@ -320,11 +285,9 @@ TEST_P(CacheTest, PinnedUsageTest) {
     auto kv_size = key.size() + 5;
     Cache::Handle* handle;
     Cache::Handle* handle_in_precise_cache;
-    ASSERT_OK(cache->Insert(key, reinterpret_cast<void*>(value), kv_size,
-                            DumbDeleter, &handle));
+    ASSERT_OK(cache->Insert(key, value, &kDumbHelper, kv_size, &handle));
     assert(handle);
-    ASSERT_OK(precise_cache->Insert(key, reinterpret_cast<void*>(value),
-                                    kv_size, DumbDeleter,
+    ASSERT_OK(precise_cache->Insert(key, value, &kDumbHelper, kv_size,
                                     &handle_in_precise_cache));
     assert(handle_in_precise_cache);
     pinned_usage += kv_size;
@@ -365,10 +328,8 @@ TEST_P(CacheTest, PinnedUsageTest) {
     } else {
       key = EncodeKey(static_cast<int>(1000 + i));
     }
-    ASSERT_OK(cache->Insert(key, reinterpret_cast<void*>(value), key.size() + 5,
-                            DumbDeleter));
-    ASSERT_OK(precise_cache->Insert(key, reinterpret_cast<void*>(value),
-                                    key.size() + 5, DumbDeleter));
+    ASSERT_OK(cache->Insert(key, value, &kDumbHelper, key.size() + 5));
+    ASSERT_OK(precise_cache->Insert(key, value, &kDumbHelper, key.size() + 5));
   }
   ASSERT_EQ(pinned_usage, cache->GetPinnedUsage());
   ASSERT_EQ(precise_cache_pinned_usage, precise_cache->GetPinnedUsage());
@@ -416,8 +377,7 @@ TEST_P(CacheTest, HitAndMiss) {
   ASSERT_EQ(201, Lookup(200));
   ASSERT_EQ(-1, Lookup(300));
 
-  ASSERT_EQ(1U, deleted_keys_.size());
-  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(1U, deleted_values_.size());
   if (GetParam() == kHyperClock) {
     ASSERT_EQ(102, deleted_values_[0]);
   } else {
@@ -438,21 +398,20 @@ TEST_P(CacheTest, InsertSameKey) {
 
 TEST_P(CacheTest, Erase) {
   Erase(200);
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
 
   Insert(100, 101);
   Insert(200, 201);
   Erase(100);
   ASSERT_EQ(-1, Lookup(100));
   ASSERT_EQ(201, Lookup(200));
-  ASSERT_EQ(1U, deleted_keys_.size());
-  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(1U, deleted_values_.size());
   ASSERT_EQ(101, deleted_values_[0]);
 
   Erase(100);
   ASSERT_EQ(-1, Lookup(100));
   ASSERT_EQ(201, Lookup(200));
-  ASSERT_EQ(1U, deleted_keys_.size());
+  ASSERT_EQ(1U, deleted_values_.size());
 }
 
 TEST_P(CacheTest, EntriesArePinned) {
@@ -469,23 +428,21 @@ TEST_P(CacheTest, EntriesArePinned) {
   Insert(100, 102);
   Cache::Handle* h2 = cache_->Lookup(EncodeKey(100));
   ASSERT_EQ(102, DecodeValue(cache_->Value(h2)));
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
   ASSERT_EQ(2U, cache_->GetUsage());
 
   cache_->Release(h1);
-  ASSERT_EQ(1U, deleted_keys_.size());
-  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(1U, deleted_values_.size());
   ASSERT_EQ(101, deleted_values_[0]);
   ASSERT_EQ(1U, cache_->GetUsage());
 
   Erase(100);
   ASSERT_EQ(-1, Lookup(100));
-  ASSERT_EQ(1U, deleted_keys_.size());
+  ASSERT_EQ(1U, deleted_values_.size());
   ASSERT_EQ(1U, cache_->GetUsage());
 
   cache_->Release(h2);
-  ASSERT_EQ(2U, deleted_keys_.size());
-  ASSERT_EQ(100, deleted_keys_[1]);
+  ASSERT_EQ(2U, deleted_values_.size());
   ASSERT_EQ(102, deleted_values_[1]);
   ASSERT_EQ(0U, cache_->GetUsage());
 }
@@ -588,9 +545,9 @@ TEST_P(CacheTest, EvictEmptyCache) {
   // Insert item large than capacity to trigger eviction on empty cache.
   auto cache = NewCache(1, 0, false);
   if (type == kLRU) {
-    ASSERT_OK(cache->Insert("foo", nullptr, 10, DumbDeleter));
+    ASSERT_OK(cache->Insert("foo", nullptr, &kDumbHelper, 10));
   } else {
-    ASSERT_OK(cache->Insert(EncodeKey(1000), nullptr, 10, DumbDeleter));
+    ASSERT_OK(cache->Insert(EncodeKey(1000), nullptr, &kDumbHelper, 10));
   }
 }
 
@@ -601,19 +558,19 @@ TEST_P(CacheTest, EraseFromDeleter) {
   // the cache at that point.
   std::shared_ptr<Cache> cache = NewCache(10, 0, false);
   std::string foo, bar;
-  Cache::DeleterFn erase_deleter;
+  const Cache::CacheItemHelper* erase_helper;
   if (type == kLRU) {
     foo = "foo";
     bar = "bar";
-    erase_deleter = EraseDeleter1;
+    erase_helper = &kEraseOnDeleteHelper1;
   } else {
     foo = EncodeKey(1234);
     bar = EncodeKey(5678);
-    erase_deleter = EraseDeleter2;
+    erase_helper = &kEraseOnDeleteHelper2;
   }
 
-  ASSERT_OK(cache->Insert(foo, nullptr, 1, DumbDeleter));
-  ASSERT_OK(cache->Insert(bar, cache.get(), 1, erase_deleter));
+  ASSERT_OK(cache->Insert(foo, nullptr, &kDumbHelper, 1));
+  ASSERT_OK(cache->Insert(bar, cache.get(), erase_helper, 1));
 
   cache->Erase(bar);
   ASSERT_EQ(nullptr, cache->Lookup(foo));
@@ -675,50 +632,51 @@ TEST_P(CacheTest, NewId) {
   ASSERT_NE(a, b);
 }
 
-class Value {
- public:
-  explicit Value(int v) : v_(v) {}
-
-  int v_;
-};
-
-namespace {
-void deleter(const Slice& /*key*/, void* value) {
-  delete static_cast<Value*>(value);
-}
-}  // namespace
-
 TEST_P(CacheTest, ReleaseAndErase) {
   std::shared_ptr<Cache> cache = NewCache(5, 0, false);
   Cache::Handle* handle;
-  Status s = cache->Insert(EncodeKey(100), EncodeValue(100), 1,
-                           &CacheTest::Deleter, &handle);
+  Status s =
+      cache->Insert(EncodeKey(100), EncodeValue(100), &kHelper, 1, &handle);
   ASSERT_TRUE(s.ok());
   ASSERT_EQ(5U, cache->GetCapacity());
   ASSERT_EQ(1U, cache->GetUsage());
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
   auto erased = cache->Release(handle, true);
   ASSERT_TRUE(erased);
   // This tests that deleter has been called
-  ASSERT_EQ(1U, deleted_keys_.size());
+  ASSERT_EQ(1U, deleted_values_.size());
 }
 
 TEST_P(CacheTest, ReleaseWithoutErase) {
   std::shared_ptr<Cache> cache = NewCache(5, 0, false);
   Cache::Handle* handle;
-  Status s = cache->Insert(EncodeKey(100), EncodeValue(100), 1,
-                           &CacheTest::Deleter, &handle);
+  Status s =
+      cache->Insert(EncodeKey(100), EncodeValue(100), &kHelper, 1, &handle);
   ASSERT_TRUE(s.ok());
   ASSERT_EQ(5U, cache->GetCapacity());
   ASSERT_EQ(1U, cache->GetUsage());
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
   auto erased = cache->Release(handle);
   ASSERT_FALSE(erased);
   // This tests that deleter is not called. When cache has free capacity it is
   // not expected to immediately erase the released items.
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
 }
 
+namespace {
+class Value {
+ public:
+  explicit Value(int v) : v_(v) {}
+
+  int v_;
+
+  static constexpr auto kCacheEntryRole = CacheEntryRole::kMisc;
+};
+
+using SharedCache = BasicTypedSharedCacheInterface<Value>;
+using TypedHandle = SharedCache::TypedHandle;
+}  // namespace
+
 TEST_P(CacheTest, SetCapacity) {
   auto type = GetParam();
   if (type == kHyperClock) {
@@ -731,19 +689,19 @@ TEST_P(CacheTest, SetCapacity) {
   // lets create a cache with capacity 5,
   // then, insert 5 elements, then increase capacity
   // to 10, returned capacity should be 10, usage=5
-  std::shared_ptr<Cache> cache = NewCache(5, 0, false);
-  std::vector<Cache::Handle*> handles(10);
+  SharedCache cache{NewCache(5, 0, false)};
+  std::vector<TypedHandle*> handles(10);
   // Insert 5 entries, but not releasing.
   for (int i = 0; i < 5; i++) {
     std::string key = EncodeKey(i + 1);
-    Status s = cache->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    Status s = cache.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_TRUE(s.ok());
   }
-  ASSERT_EQ(5U, cache->GetCapacity());
-  ASSERT_EQ(5U, cache->GetUsage());
-  cache->SetCapacity(10);
-  ASSERT_EQ(10U, cache->GetCapacity());
-  ASSERT_EQ(5U, cache->GetUsage());
+  ASSERT_EQ(5U, cache.get()->GetCapacity());
+  ASSERT_EQ(5U, cache.get()->GetUsage());
+  cache.get()->SetCapacity(10);
+  ASSERT_EQ(10U, cache.get()->GetCapacity());
+  ASSERT_EQ(5U, cache.get()->GetUsage());
 
   // test2: decrease capacity
   // insert 5 more elements to cache, then release 5,
@@ -751,77 +709,77 @@ TEST_P(CacheTest, SetCapacity) {
   // and usage should be 7
   for (int i = 5; i < 10; i++) {
     std::string key = EncodeKey(i + 1);
-    Status s = cache->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    Status s = cache.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_TRUE(s.ok());
   }
-  ASSERT_EQ(10U, cache->GetCapacity());
-  ASSERT_EQ(10U, cache->GetUsage());
+  ASSERT_EQ(10U, cache.get()->GetCapacity());
+  ASSERT_EQ(10U, cache.get()->GetUsage());
   for (int i = 0; i < 5; i++) {
-    cache->Release(handles[i]);
+    cache.Release(handles[i]);
   }
-  ASSERT_EQ(10U, cache->GetCapacity());
-  ASSERT_EQ(10U, cache->GetUsage());
-  cache->SetCapacity(7);
-  ASSERT_EQ(7, cache->GetCapacity());
-  ASSERT_EQ(7, cache->GetUsage());
+  ASSERT_EQ(10U, cache.get()->GetCapacity());
+  ASSERT_EQ(10U, cache.get()->GetUsage());
+  cache.get()->SetCapacity(7);
+  ASSERT_EQ(7, cache.get()->GetCapacity());
+  ASSERT_EQ(7, cache.get()->GetUsage());
 
   // release remaining 5 to keep valgrind happy
   for (int i = 5; i < 10; i++) {
-    cache->Release(handles[i]);
+    cache.Release(handles[i]);
   }
 
   // Make sure this doesn't crash or upset ASAN/valgrind
-  cache->DisownData();
+  cache.get()->DisownData();
 }
 
 TEST_P(LRUCacheTest, SetStrictCapacityLimit) {
   // test1: set the flag to false. Insert more keys than capacity. See if they
   // all go through.
-  std::shared_ptr<Cache> cache = NewCache(5, 0, false);
-  std::vector<Cache::Handle*> handles(10);
+  SharedCache cache{NewCache(5, 0, false)};
+  std::vector<TypedHandle*> handles(10);
   Status s;
   for (int i = 0; i < 10; i++) {
     std::string key = EncodeKey(i + 1);
-    s = cache->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    s = cache.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_OK(s);
     ASSERT_NE(nullptr, handles[i]);
   }
-  ASSERT_EQ(10, cache->GetUsage());
+  ASSERT_EQ(10, cache.get()->GetUsage());
 
   // test2: set the flag to true. Insert and check if it fails.
   std::string extra_key = EncodeKey(100);
   Value* extra_value = new Value(0);
-  cache->SetStrictCapacityLimit(true);
-  Cache::Handle* handle;
-  s = cache->Insert(extra_key, extra_value, 1, &deleter, &handle);
+  cache.get()->SetStrictCapacityLimit(true);
+  TypedHandle* handle;
+  s = cache.Insert(extra_key, extra_value, 1, &handle);
   ASSERT_TRUE(s.IsMemoryLimit());
   ASSERT_EQ(nullptr, handle);
-  ASSERT_EQ(10, cache->GetUsage());
+  ASSERT_EQ(10, cache.get()->GetUsage());
 
   for (int i = 0; i < 10; i++) {
-    cache->Release(handles[i]);
+    cache.Release(handles[i]);
   }
 
   // test3: init with flag being true.
-  std::shared_ptr<Cache> cache2 = NewCache(5, 0, true);
+  SharedCache cache2{NewCache(5, 0, true)};
   for (int i = 0; i < 5; i++) {
     std::string key = EncodeKey(i + 1);
-    s = cache2->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    s = cache2.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_OK(s);
     ASSERT_NE(nullptr, handles[i]);
   }
-  s = cache2->Insert(extra_key, extra_value, 1, &deleter, &handle);
+  s = cache2.Insert(extra_key, extra_value, 1, &handle);
   ASSERT_TRUE(s.IsMemoryLimit());
   ASSERT_EQ(nullptr, handle);
   // test insert without handle
-  s = cache2->Insert(extra_key, extra_value, 1, &deleter);
+  s = cache2.Insert(extra_key, extra_value, 1);
   // AS if the key have been inserted into cache but get evicted immediately.
   ASSERT_OK(s);
-  ASSERT_EQ(5, cache2->GetUsage());
-  ASSERT_EQ(nullptr, cache2->Lookup(extra_key));
+  ASSERT_EQ(5, cache2.get()->GetUsage());
+  ASSERT_EQ(nullptr, cache2.Lookup(extra_key));
 
   for (int i = 0; i < 5; i++) {
-    cache2->Release(handles[i]);
+    cache2.Release(handles[i]);
   }
 }
 
@@ -829,55 +787,54 @@ TEST_P(CacheTest, OverCapacity) {
   size_t n = 10;
 
   // a LRUCache with n entries and one shard only
-  std::shared_ptr<Cache> cache = NewCache(n, 0, false);
-
-  std::vector<Cache::Handle*> handles(n + 1);
+  SharedCache cache{NewCache(n, 0, false)};
+  std::vector<TypedHandle*> handles(n + 1);
 
   // Insert n+1 entries, but not releasing.
   for (int i = 0; i < static_cast<int>(n + 1); i++) {
     std::string key = EncodeKey(i + 1);
-    Status s = cache->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    Status s = cache.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_TRUE(s.ok());
   }
 
   // Guess what's in the cache now?
   for (int i = 0; i < static_cast<int>(n + 1); i++) {
     std::string key = EncodeKey(i + 1);
-    auto h = cache->Lookup(key);
+    auto h = cache.Lookup(key);
     ASSERT_TRUE(h != nullptr);
-    if (h) cache->Release(h);
+    if (h) cache.Release(h);
   }
 
   // the cache is over capacity since nothing could be evicted
-  ASSERT_EQ(n + 1U, cache->GetUsage());
+  ASSERT_EQ(n + 1U, cache.get()->GetUsage());
   for (int i = 0; i < static_cast<int>(n + 1); i++) {
-    cache->Release(handles[i]);
+    cache.Release(handles[i]);
   }
 
   if (GetParam() == kHyperClock) {
     // Make sure eviction is triggered.
-    ASSERT_OK(cache->Insert(EncodeKey(-1), nullptr, 1, &deleter, &handles[0]));
+    ASSERT_OK(cache.Insert(EncodeKey(-1), nullptr, 1, &handles[0]));
 
     // cache is under capacity now since elements were released
-    ASSERT_GE(n, cache->GetUsage());
+    ASSERT_GE(n, cache.get()->GetUsage());
 
     // clean up
-    cache->Release(handles[0]);
+    cache.Release(handles[0]);
   } else {
     // LRUCache checks for over-capacity in Release.
 
     // cache is exactly at capacity now with minimal eviction
-    ASSERT_EQ(n, cache->GetUsage());
+    ASSERT_EQ(n, cache.get()->GetUsage());
 
     // element 0 is evicted and the rest is there
     // This is consistent with the LRU policy since the element 0
     // was released first
     for (int i = 0; i < static_cast<int>(n + 1); i++) {
       std::string key = EncodeKey(i + 1);
-      auto h = cache->Lookup(key);
+      auto h = cache.Lookup(key);
       if (h) {
         ASSERT_NE(static_cast<size_t>(i), 0U);
-        cache->Release(h);
+        cache.Release(h);
       } else {
         ASSERT_EQ(static_cast<size_t>(i), 0U);
       }
@@ -885,40 +842,15 @@ TEST_P(CacheTest, OverCapacity) {
   }
 }
 
-namespace {
-std::vector<std::pair<int, int>> legacy_callback_state;
-void legacy_callback(void* value, size_t charge) {
-  legacy_callback_state.push_back(
-      {DecodeValue(value), static_cast<int>(charge)});
-}
-};  // namespace
-
-TEST_P(CacheTest, ApplyToAllCacheEntriesTest) {
-  std::vector<std::pair<int, int>> inserted;
-  legacy_callback_state.clear();
-
-  for (int i = 0; i < 10; ++i) {
-    Insert(i, i * 2, i + 1);
-    inserted.push_back({i * 2, i + 1});
-  }
-  cache_->ApplyToAllCacheEntries(legacy_callback, true);
-
-  std::sort(inserted.begin(), inserted.end());
-  std::sort(legacy_callback_state.begin(), legacy_callback_state.end());
-  ASSERT_EQ(inserted.size(), legacy_callback_state.size());
-  for (int i = 0; i < static_cast<int>(inserted.size()); ++i) {
-    EXPECT_EQ(inserted[i], legacy_callback_state[i]);
-  }
-}
-
 TEST_P(CacheTest, ApplyToAllEntriesTest) {
   std::vector<std::string> callback_state;
-  const auto callback = [&](const Slice& key, void* value, size_t charge,
-                            Cache::DeleterFn deleter) {
+  const auto callback = [&](const Slice& key, Cache::ObjectPtr value,
+                            size_t charge,
+                            const Cache::CacheItemHelper* helper) {
     callback_state.push_back(std::to_string(DecodeKey(key)) + "," +
                              std::to_string(DecodeValue(value)) + "," +
                              std::to_string(charge));
-    assert(deleter == &CacheTest::Deleter);
+    assert(helper == &CacheTest::kHelper);
   };
 
   std::vector<std::string> inserted;
@@ -957,8 +889,8 @@ TEST_P(CacheTest, ApplyToAllEntriesDuringResize) {
 
   // For callback
   int special_count = 0;
-  const auto callback = [&](const Slice&, void*, size_t charge,
-                            Cache::DeleterFn) {
+  const auto callback = [&](const Slice&, Cache::ObjectPtr, size_t charge,
+                            const Cache::CacheItemHelper*) {
     if (charge == static_cast<size_t>(kSpecialCharge)) {
       ++special_count;
     }
@@ -1020,13 +952,105 @@ TEST_P(CacheTest, GetChargeAndDeleter) {
   Cache::Handle* h1 = cache_->Lookup(EncodeKey(1));
   ASSERT_EQ(2, DecodeValue(cache_->Value(h1)));
   ASSERT_EQ(1, cache_->GetCharge(h1));
-  ASSERT_EQ(&CacheTest::Deleter, cache_->GetDeleter(h1));
+  ASSERT_EQ(&CacheTest::kHelper, cache_->GetCacheItemHelper(h1));
   cache_->Release(h1);
 }
 
+namespace {
+bool AreTwoCacheKeysOrdered(Cache* cache) {
+  std::vector<std::string> keys;
+  const auto callback = [&](const Slice& key, Cache::ObjectPtr /*value*/,
+                            size_t /*charge*/,
+                            const Cache::CacheItemHelper* /*helper*/) {
+    keys.push_back(key.ToString());
+  };
+  cache->ApplyToAllEntries(callback, /*opts*/ {});
+  EXPECT_EQ(keys.size(), 2U);
+  EXPECT_NE(keys[0], keys[1]);
+  return keys[0] < keys[1];
+}
+}  // namespace
+
+TEST_P(CacheTest, CacheUniqueSeeds) {
+  // kQuasiRandomHashSeed should generate unique seeds (up to 2 billion before
+  // repeating)
+  UnorderedSet<uint32_t> seeds_seen;
+  // Roughly sqrt(number of possible values) for a decent chance at detecting
+  // a random collision if it's possible (shouldn't be)
+  uint16_t kSamples = 20000;
+  seeds_seen.reserve(kSamples);
+
+  // Hash seed should affect ordering of entries in the table, so we should
+  // have extremely high chance of seeing two entries ordered both ways.
+  bool seen_forward_order = false;
+  bool seen_reverse_order = false;
+
+  for (int i = 0; i < kSamples; ++i) {
+    auto cache = NewCache(2, [=](ShardedCacheOptions& opts) {
+      opts.hash_seed = LRUCacheOptions::kQuasiRandomHashSeed;
+      opts.num_shard_bits = 0;
+      opts.metadata_charge_policy = kDontChargeCacheMetadata;
+    });
+    auto val = cache->GetHashSeed();
+    ASSERT_TRUE(seeds_seen.insert(val).second);
+
+    ASSERT_OK(cache->Insert(EncodeKey(1), nullptr, &kHelper, /*charge*/ 1));
+    ASSERT_OK(cache->Insert(EncodeKey(2), nullptr, &kHelper, /*charge*/ 1));
+
+    if (AreTwoCacheKeysOrdered(cache.get())) {
+      seen_forward_order = true;
+    } else {
+      seen_reverse_order = true;
+    }
+  }
+
+  ASSERT_TRUE(seen_forward_order);
+  ASSERT_TRUE(seen_reverse_order);
+}
+
+TEST_P(CacheTest, CacheHostSeed) {
+  // kHostHashSeed should generate a consistent seed within this process
+  // (and other processes on the same host, but not unit testing that).
+  // And we should be able to use that chosen seed as an explicit option
+  // (for debugging).
+  // And we should verify consistent ordering of entries.
+  uint32_t expected_seed = 0;
+  bool expected_order = false;
+  // 10 iterations -> chance of a random seed falsely appearing consistent
+  // should be low, just 1 in 2^9.
+  for (int i = 0; i < 10; ++i) {
+    auto cache = NewCache(2, [=](ShardedCacheOptions& opts) {
+      if (i != 5) {
+        opts.hash_seed = LRUCacheOptions::kHostHashSeed;
+      } else {
+        // Can be used as explicit seed
+        opts.hash_seed = static_cast<int32_t>(expected_seed);
+        ASSERT_GE(opts.hash_seed, 0);
+      }
+      opts.num_shard_bits = 0;
+      opts.metadata_charge_policy = kDontChargeCacheMetadata;
+    });
+    ASSERT_OK(cache->Insert(EncodeKey(1), nullptr, &kHelper, /*charge*/ 1));
+    ASSERT_OK(cache->Insert(EncodeKey(2), nullptr, &kHelper, /*charge*/ 1));
+    uint32_t val = cache->GetHashSeed();
+    bool order = AreTwoCacheKeysOrdered(cache.get());
+    if (i != 0) {
+      ASSERT_EQ(val, expected_seed);
+      ASSERT_EQ(order, expected_order);
+    } else {
+      expected_seed = val;
+      expected_order = order;
+    }
+  }
+  // Printed for reference in case it's needed to reproduce other unit test
+  // failures on another host
+  fprintf(stderr, "kHostHashSeed -> %u\n", (unsigned)expected_seed);
+}
+
 INSTANTIATE_TEST_CASE_P(CacheTestInstance, CacheTest,
-                        testing::Values(kLRU, kHyperClock));
-INSTANTIATE_TEST_CASE_P(CacheTestInstance, LRUCacheTest, testing::Values(kLRU));
+                        secondary_cache_test_util::GetTestingCacheTypes());
+INSTANTIATE_TEST_CASE_P(CacheTestInstance, LRUCacheTest,
+                        testing::Values(secondary_cache_test_util::kLRU));
 
 }  // namespace ROCKSDB_NAMESPACE
 

cache/charged_cache.cc

@@ -11,65 +11,57 @@ namespace ROCKSDB_NAMESPACE {
 
 ChargedCache::ChargedCache(std::shared_ptr<Cache> cache,
                            std::shared_ptr<Cache> block_cache)
-    : cache_(cache),
+    : CacheWrapper(cache),
       cache_res_mgr_(std::make_shared<ConcurrentCacheReservationManager>(
           std::make_shared<
               CacheReservationManagerImpl<CacheEntryRole::kBlobCache>>(
               block_cache))) {}
 
-Status ChargedCache::Insert(const Slice& key, void* value, size_t charge,
-                            DeleterFn deleter, Handle** handle,
-                            Priority priority) {
-  Status s = cache_->Insert(key, value, charge, deleter, handle, priority);
-  if (s.ok()) {
-    // Insert may cause the cache entry eviction if the cache is full. So we
-    // directly call the reservation manager to update the total memory used
-    // in the cache.
-    assert(cache_res_mgr_);
-    cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
-        .PermitUncheckedError();
-  }
-  return s;
-}
-
-Status ChargedCache::Insert(const Slice& key, void* value,
+Status ChargedCache::Insert(const Slice& key, ObjectPtr obj,
                             const CacheItemHelper* helper, size_t charge,
                             Handle** handle, Priority priority) {
-  Status s = cache_->Insert(key, value, helper, charge, handle, priority);
+  Status s = target_->Insert(key, obj, helper, charge, handle, priority);
   if (s.ok()) {
     // Insert may cause the cache entry eviction if the cache is full. So we
     // directly call the reservation manager to update the total memory used
     // in the cache.
     assert(cache_res_mgr_);
-    cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
+    cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
         .PermitUncheckedError();
   }
   return s;
 }
 
-Cache::Handle* ChargedCache::Lookup(const Slice& key, Statistics* stats) {
-  return cache_->Lookup(key, stats);
-}
-
 Cache::Handle* ChargedCache::Lookup(const Slice& key,
                                     const CacheItemHelper* helper,
-                                    const CreateCallback& create_cb,
-                                    Priority priority, bool wait,
-                                    Statistics* stats) {
-  auto handle = cache_->Lookup(key, helper, create_cb, priority, wait, stats);
+                                    CreateContext* create_context,
+                                    Priority priority, Statistics* stats) {
+  auto handle = target_->Lookup(key, helper, create_context, priority, stats);
   // Lookup may promote the KV pair from the secondary cache to the primary
   // cache. So we directly call the reservation manager to update the total
   // memory used in the cache.
-  assert(cache_res_mgr_);
-  cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
-      .PermitUncheckedError();
+  if (helper && helper->create_cb) {
+    assert(cache_res_mgr_);
+    cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
+        .PermitUncheckedError();
+  }
   return handle;
 }
 
+void ChargedCache::WaitAll(AsyncLookupHandle* async_handles, size_t count) {
+  target_->WaitAll(async_handles, count);
+  // In case of any promotions. Although some could finish by return of
+  // StartAsyncLookup, Wait/WaitAll will generally be used, so simpler to
+  // update here.
+  assert(cache_res_mgr_);
+  cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
+      .PermitUncheckedError();
+}
+
 bool ChargedCache::Release(Cache::Handle* handle, bool useful,
                            bool erase_if_last_ref) {
-  size_t memory_used_delta = cache_->GetUsage(handle);
-  bool erased = cache_->Release(handle, useful, erase_if_last_ref);
+  size_t memory_used_delta = target_->GetUsage(handle);
+  bool erased = target_->Release(handle, useful, erase_if_last_ref);
   if (erased) {
     assert(cache_res_mgr_);
     cache_res_mgr_
@@ -80,8 +72,8 @@ bool ChargedCache::Release(Cache::Handle* handle, bool useful,
 }
 
 bool ChargedCache::Release(Cache::Handle* handle, bool erase_if_last_ref) {
-  size_t memory_used_delta = cache_->GetUsage(handle);
-  bool erased = cache_->Release(handle, erase_if_last_ref);
+  size_t memory_used_delta = target_->GetUsage(handle);
+  bool erased = target_->Release(handle, erase_if_last_ref);
   if (erased) {
     assert(cache_res_mgr_);
     cache_res_mgr_
@@ -92,25 +84,25 @@ bool ChargedCache::Release(Cache::Handle* handle, bool erase_if_last_ref) {
 }
 
 void ChargedCache::Erase(const Slice& key) {
-  cache_->Erase(key);
+  target_->Erase(key);
   assert(cache_res_mgr_);
-  cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
+  cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
       .PermitUncheckedError();
 }
 
 void ChargedCache::EraseUnRefEntries() {
-  cache_->EraseUnRefEntries();
+  target_->EraseUnRefEntries();
   assert(cache_res_mgr_);
-  cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
+  cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
       .PermitUncheckedError();
 }
 
 void ChargedCache::SetCapacity(size_t capacity) {
-  cache_->SetCapacity(capacity);
+  target_->SetCapacity(capacity);
   // SetCapacity can result in evictions when the cache capacity is decreased,
   // so we would want to update the cache reservation here as well.
   assert(cache_res_mgr_);
-  cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
+  cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
       .PermitUncheckedError();
 }
 

cache/charged_cache.h

@@ -8,7 +8,7 @@
 #include <string>
 
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -17,22 +17,21 @@ class ConcurrentCacheReservationManager;
 // A cache interface which wraps around another cache and takes care of
 // reserving space in block cache towards a single global memory limit, and
 // forwards all the calls to the underlying cache.
-class ChargedCache : public Cache {
+class ChargedCache : public CacheWrapper {
  public:
   ChargedCache(std::shared_ptr<Cache> cache,
                std::shared_ptr<Cache> block_cache);
-  ~ChargedCache() override = default;
 
-  Status Insert(const Slice& key, void* value, size_t charge, DeleterFn deleter,
-                Handle** handle, Priority priority) override;
-  Status Insert(const Slice& key, void* value, const CacheItemHelper* helper,
+  Status Insert(const Slice& key, ObjectPtr obj, const CacheItemHelper* helper,
                 size_t charge, Handle** handle = nullptr,
                 Priority priority = Priority::LOW) override;
 
-  Cache::Handle* Lookup(const Slice& key, Statistics* stats) override;
   Cache::Handle* Lookup(const Slice& key, const CacheItemHelper* helper,
-                        const CreateCallback& create_cb, Priority priority,
-                        bool wait, Statistics* stats = nullptr) override;
+                        CreateContext* create_context,
+                        Priority priority = Priority::LOW,
+                        Statistics* stats = nullptr) override;
+
+  void WaitAll(AsyncLookupHandle* async_handles, size_t count) override;
 
   bool Release(Cache::Handle* handle, bool useful,
                bool erase_if_last_ref = false) override;
@@ -44,69 +43,9 @@ class ChargedCache : public Cache {
   static const char* kClassName() { return "ChargedCache"; }
   const char* Name() const override { return kClassName(); }
 
-  uint64_t NewId() override { return cache_->NewId(); }
-
   void SetCapacity(size_t capacity) override;
 
-  void SetStrictCapacityLimit(bool strict_capacity_limit) override {
-    cache_->SetStrictCapacityLimit(strict_capacity_limit);
-  }
-
-  bool HasStrictCapacityLimit() const override {
-    return cache_->HasStrictCapacityLimit();
-  }
-
-  void* Value(Cache::Handle* handle) override { return cache_->Value(handle); }
-
-  bool IsReady(Cache::Handle* handle) override {
-    return cache_->IsReady(handle);
-  }
-
-  void Wait(Cache::Handle* handle) override { cache_->Wait(handle); }
-
-  void WaitAll(std::vector<Handle*>& handles) override {
-    cache_->WaitAll(handles);
-  }
-
-  bool Ref(Cache::Handle* handle) override { return cache_->Ref(handle); }
-
-  size_t GetCapacity() const override { return cache_->GetCapacity(); }
-
-  size_t GetUsage() const override { return cache_->GetUsage(); }
-
-  size_t GetUsage(Cache::Handle* handle) const override {
-    return cache_->GetUsage(handle);
-  }
-
-  size_t GetPinnedUsage() const override { return cache_->GetPinnedUsage(); }
-
-  size_t GetCharge(Cache::Handle* handle) const override {
-    return cache_->GetCharge(handle);
-  }
-
-  Cache::DeleterFn GetDeleter(Cache::Handle* handle) const override {
-    return cache_->GetDeleter(handle);
-  }
-
-  void ApplyToAllEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               Cache::DeleterFn deleter)>& callback,
-      const Cache::ApplyToAllEntriesOptions& opts) override {
-    cache_->ApplyToAllEntries(callback, opts);
-  }
-
-  void ApplyToAllCacheEntries(void (*callback)(void* value, size_t charge),
-                              bool thread_safe) override {
-    cache_->ApplyToAllCacheEntries(callback, thread_safe);
-  }
-
-  std::string GetPrintableOptions() const override {
-    return cache_->GetPrintableOptions();
-  }
-
-  void DisownData() override { return cache_->DisownData(); }
-
-  inline Cache* GetCache() const { return cache_.get(); }
+  inline Cache* GetCache() const { return target_.get(); }
 
   inline ConcurrentCacheReservationManager* TEST_GetCacheReservationManager()
       const {
@@ -114,7 +53,6 @@ class ChargedCache : public Cache {
   }
 
  private:
-  std::shared_ptr<Cache> cache_;
   std::shared_ptr<ConcurrentCacheReservationManager> cache_res_mgr_;
 };
 

cache/clock_cache.cc

@@ -9,15 +9,16 @@
 
 #include "cache/clock_cache.h"
 
-#include <cassert>
 #include <functional>
 #include <numeric>
 
 #include "cache/cache_key.h"
+#include "cache/secondary_cache_adapter.h"
 #include "logging/logging.h"
 #include "monitoring/perf_context_imp.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "port/lang.h"
+#include "rocksdb/env.h"
 #include "util/hash.h"
 #include "util/math.h"
 #include "util/random.h"
@@ -50,13 +51,7 @@ inline uint64_t GetInitialCountdown(Cache::Priority priority) {
   }
 }
 
-inline void FreeDataMarkEmpty(ClockHandle& h) {
-  // NOTE: in theory there's more room for parallelism if we copy the handle
-  // data and delay actions like this until after marking the entry as empty,
-  // but performance tests only show a regression by copying the few words
-  // of data.
-  h.FreeData();
-
+inline void MarkEmpty(ClockHandle& h) {
 #ifndef NDEBUG
   // Mark slot as empty, with assertion
   uint64_t meta = h.meta.exchange(0, std::memory_order_release);
@@ -67,6 +62,16 @@ inline void FreeDataMarkEmpty(ClockHandle& h) {
 #endif
 }
 
+inline void FreeDataMarkEmpty(ClockHandle& h, MemoryAllocator* allocator) {
+  // NOTE: in theory there's more room for parallelism if we copy the handle
+  // data and delay actions like this until after marking the entry as empty,
+  // but performance tests only show a regression by copying the few words
+  // of data.
+  h.FreeData(allocator);
+
+  MarkEmpty(h);
+}
+
 inline bool ClockUpdate(ClockHandle& h) {
   uint64_t meta = h.meta.load(std::memory_order_relaxed);
 
@@ -115,24 +120,27 @@ inline bool ClockUpdate(ClockHandle& h) {
 
 }  // namespace
 
-void ClockHandleBasicData::FreeData() const {
-  if (deleter) {
-    UniqueId64x2 unhashed;
-    (*deleter)(
-        ClockCacheShard<HyperClockTable>::ReverseHash(hashed_key, &unhashed),
-        value);
+void ClockHandleBasicData::FreeData(MemoryAllocator* allocator) const {
+  if (helper->del_cb) {
+    helper->del_cb(value, allocator);
   }
 }
 
 HyperClockTable::HyperClockTable(
     size_t capacity, bool /*strict_capacity_limit*/,
-    CacheMetadataChargePolicy metadata_charge_policy, const Opts& opts)
+    CacheMetadataChargePolicy metadata_charge_policy,
+    MemoryAllocator* allocator,
+    const Cache::EvictionCallback* eviction_callback, const uint32_t* hash_seed,
+    const Opts& opts)
     : length_bits_(CalcHashBits(capacity, opts.estimated_value_size,
                                 metadata_charge_policy)),
       length_bits_mask_((size_t{1} << length_bits_) - 1),
       occupancy_limit_(static_cast<size_t>((uint64_t{1} << length_bits_) *
                                            kStrictLoadFactor)),
-      array_(new HandleImpl[size_t{1} << length_bits_]) {
+      array_(new HandleImpl[size_t{1} << length_bits_]),
+      allocator_(allocator),
+      eviction_callback_(*eviction_callback),
+      hash_seed_(*hash_seed) {
   if (metadata_charge_policy ==
       CacheMetadataChargePolicy::kFullChargeCacheMetadata) {
     usage_ += size_t{GetTableSize()} * sizeof(HandleImpl);
@@ -154,7 +162,7 @@ HyperClockTable::~HyperClockTable() {
       case ClockHandle::kStateInvisible:  // rare but possible
       case ClockHandle::kStateVisible:
         assert(GetRefcount(h.meta) == 0);
-        h.FreeData();
+        h.FreeData(allocator_);
 #ifndef NDEBUG
         Rollback(h.hashed_key, &h);
         ReclaimEntryUsage(h.GetTotalCharge());
@@ -365,21 +373,22 @@ inline bool HyperClockTable::ChargeUsageMaybeEvictNonStrict(
   return true;
 }
 
-inline HyperClockTable::HandleImpl* HyperClockTable::DetachedInsert(
+inline HyperClockTable::HandleImpl* HyperClockTable::StandaloneInsert(
     const ClockHandleBasicData& proto) {
   // Heap allocated separate from table
   HandleImpl* h = new HandleImpl();
   ClockHandleBasicData* h_alias = h;
   *h_alias = proto;
-  h->SetDetached();
-  // Single reference (detached entries only created if returning a refed
+  h->SetStandalone();
+  // Single reference (standalone entries only created if returning a refed
   // Handle back to user)
   uint64_t meta = uint64_t{ClockHandle::kStateInvisible}
                   << ClockHandle::kStateShift;
   meta |= uint64_t{1} << ClockHandle::kAcquireCounterShift;
   h->meta.store(meta, std::memory_order_release);
-  // Keep track of how much of usage is detached
-  detached_usage_.fetch_add(proto.GetTotalCharge(), std::memory_order_relaxed);
+  // Keep track of how much of usage is standalone
+  standalone_usage_.fetch_add(proto.GetTotalCharge(),
+                              std::memory_order_relaxed);
   return h;
 }
 
@@ -397,7 +406,7 @@ Status HyperClockTable::Insert(const ClockHandleBasicData& proto,
 
   // Usage/capacity handling is somewhat different depending on
   // strict_capacity_limit, but mostly pessimistic.
-  bool use_detached_insert = false;
+  bool use_standalone_insert = false;
   const size_t total_charge = proto.GetTotalCharge();
   if (strict_capacity_limit) {
     Status s = ChargeUsageMaybeEvictStrict(total_charge, capacity,
@@ -415,12 +424,12 @@ Status HyperClockTable::Insert(const ClockHandleBasicData& proto,
       if (handle == nullptr) {
         // Don't insert the entry but still return ok, as if the entry
         // inserted into cache and evicted immediately.
-        proto.FreeData();
+        proto.FreeData(allocator_);
         return Status::OK();
       } else {
-        // Need to track usage of fallback detached insert
+        // Need to track usage of fallback standalone insert
         usage_.fetch_add(total_charge, std::memory_order_relaxed);
-        use_detached_insert = true;
+        use_standalone_insert = true;
       }
     }
   }
@@ -430,7 +439,7 @@ Status HyperClockTable::Insert(const ClockHandleBasicData& proto,
     assert(usage_.load(std::memory_order_relaxed) < SIZE_MAX / 2);
   };
 
-  if (!use_detached_insert) {
+  if (!use_standalone_insert) {
     // Attempt a table insert, but abort if we find an existing entry for the
     // key. If we were to overwrite old entries, we would either
     // * Have to gain ownership over an existing entry to overwrite it, which
@@ -501,8 +510,8 @@ Status HyperClockTable::Insert(const ClockHandleBasicData& proto,
                   std::memory_order_acq_rel);
               // Correct for possible (but rare) overflow
               CorrectNearOverflow(old_meta, h->meta);
-              // Insert detached instead (only if return handle needed)
-              use_detached_insert = true;
+              // Insert standalone instead (only if return handle needed)
+              use_standalone_insert = true;
               return true;
             } else {
               // Mismatch. Pretend we never took the reference
@@ -540,9 +549,9 @@ Status HyperClockTable::Insert(const ClockHandleBasicData& proto,
       // That should be infeasible for roughly n >= 256, so if this assertion
       // fails, that suggests something is going wrong.
       assert(GetTableSize() < 256);
-      use_detached_insert = true;
+      use_standalone_insert = true;
     }
-    if (!use_detached_insert) {
+    if (!use_standalone_insert) {
       // Successfully inserted
       if (handle) {
         *handle = e;
@@ -552,28 +561,56 @@ Status HyperClockTable::Insert(const ClockHandleBasicData& proto,
     // Roll back table insertion
     Rollback(proto.hashed_key, e);
     revert_occupancy_fn();
-    // Maybe fall back on detached insert
+    // Maybe fall back on standalone insert
     if (handle == nullptr) {
       revert_usage_fn();
       // As if unrefed entry immdiately evicted
-      proto.FreeData();
+      proto.FreeData(allocator_);
       return Status::OK();
     }
   }
 
-  // Run detached insert
-  assert(use_detached_insert);
+  // Run standalone insert
+  assert(use_standalone_insert);
 
-  *handle = DetachedInsert(proto);
+  *handle = StandaloneInsert(proto);
 
   // The OkOverwritten status is used to count "redundant" insertions into
   // block cache. This implementation doesn't strictly check for redundant
   // insertions, but we instead are probably interested in how many insertions
-  // didn't go into the table (instead "detached"), which could be redundant
-  // Insert or some other reason (use_detached_insert reasons above).
+  // didn't go into the table (instead "standalone"), which could be redundant
+  // Insert or some other reason (use_standalone_insert reasons above).
   return Status::OkOverwritten();
 }
 
+HyperClockTable::HandleImpl* HyperClockTable::CreateStandalone(
+    ClockHandleBasicData& proto, size_t capacity, bool strict_capacity_limit,
+    bool allow_uncharged) {
+  const size_t total_charge = proto.GetTotalCharge();
+  if (strict_capacity_limit) {
+    Status s = ChargeUsageMaybeEvictStrict(total_charge, capacity,
+                                           /*need_evict_for_occupancy=*/false);
+    if (!s.ok()) {
+      if (allow_uncharged) {
+        proto.total_charge = 0;
+      } else {
+        return nullptr;
+      }
+    }
+  } else {
+    // Case strict_capacity_limit == false
+    bool success =
+        ChargeUsageMaybeEvictNonStrict(total_charge, capacity,
+                                       /*need_evict_for_occupancy=*/false);
+    if (!success) {
+      // Force the issue
+      usage_.fetch_add(total_charge, std::memory_order_relaxed);
+    }
+  }
+
+  return StandaloneInsert(proto);
+}
+
 HyperClockTable::HandleImpl* HyperClockTable::Lookup(
     const UniqueId64x2& hashed_key) {
   size_t probe = 0;
@@ -697,15 +734,15 @@ bool HyperClockTable::Release(HandleImpl* h, bool useful,
         std::memory_order_acquire));
     // Took ownership
     size_t total_charge = h->GetTotalCharge();
-    if (UNLIKELY(h->IsDetached())) {
-      h->FreeData();
-      // Delete detached handle
+    if (UNLIKELY(h->IsStandalone())) {
+      h->FreeData(allocator_);
+      // Delete standalone handle
       delete h;
-      detached_usage_.fetch_sub(total_charge, std::memory_order_relaxed);
+      standalone_usage_.fetch_sub(total_charge, std::memory_order_relaxed);
       usage_.fetch_sub(total_charge, std::memory_order_relaxed);
     } else {
       Rollback(h->hashed_key, h);
-      FreeDataMarkEmpty(*h);
+      FreeDataMarkEmpty(*h, allocator_);
       ReclaimEntryUsage(total_charge);
     }
     return true;
@@ -790,7 +827,7 @@ void HyperClockTable::Erase(const UniqueId64x2& hashed_key) {
                 // Took ownership
                 assert(hashed_key == h->hashed_key);
                 size_t total_charge = h->GetTotalCharge();
-                FreeDataMarkEmpty(*h);
+                FreeDataMarkEmpty(*h, allocator_);
                 ReclaimEntryUsage(total_charge);
                 // We already have a copy of hashed_key in this case, so OK to
                 // delay Rollback until after releasing the entry
@@ -878,7 +915,7 @@ void HyperClockTable::EraseUnRefEntries() {
       // Took ownership
       size_t total_charge = h.GetTotalCharge();
       Rollback(h.hashed_key, &h);
-      FreeDataMarkEmpty(h);
+      FreeDataMarkEmpty(h, allocator_);
       ReclaimEntryUsage(total_charge);
     }
   }
@@ -960,6 +997,9 @@ inline void HyperClockTable::Evict(size_t requested_charge,
   uint64_t max_clock_pointer =
       old_clock_pointer + (ClockHandle::kMaxCountdown << length_bits_);
 
+  // For key reconstructed from hash
+  UniqueId64x2 unhashed;
+
   for (;;) {
     for (size_t i = 0; i < step_size; i++) {
       HandleImpl& h = array_[ModTableSize(Lower32of64(old_clock_pointer + i))];
@@ -968,7 +1008,17 @@ inline void HyperClockTable::Evict(size_t requested_charge,
         Rollback(h.hashed_key, &h);
         *freed_charge += h.GetTotalCharge();
         *freed_count += 1;
-        FreeDataMarkEmpty(h);
+        bool took_ownership = false;
+        if (eviction_callback_) {
+          took_ownership =
+              eviction_callback_(ClockCacheShard<HyperClockTable>::ReverseHash(
+                                     h.GetHash(), &unhashed, hash_seed_),
+                                 reinterpret_cast<Cache::Handle*>(&h));
+        }
+        if (!took_ownership) {
+          h.FreeData(allocator_);
+        }
+        MarkEmpty(h);
       }
     }
 
@@ -990,9 +1040,12 @@ template <class Table>
 ClockCacheShard<Table>::ClockCacheShard(
     size_t capacity, bool strict_capacity_limit,
     CacheMetadataChargePolicy metadata_charge_policy,
+    MemoryAllocator* allocator,
+    const Cache::EvictionCallback* eviction_callback, const uint32_t* hash_seed,
     const typename Table::Opts& opts)
     : CacheShardBase(metadata_charge_policy),
-      table_(capacity, strict_capacity_limit, metadata_charge_policy, opts),
+      table_(capacity, strict_capacity_limit, metadata_charge_policy, allocator,
+             eviction_callback, hash_seed, opts),
       capacity_(capacity),
       strict_capacity_limit_(strict_capacity_limit) {
   // Initial charge metadata should not exceed capacity
@@ -1006,8 +1059,9 @@ void ClockCacheShard<Table>::EraseUnRefEntries() {
 
 template <class Table>
 void ClockCacheShard<Table>::ApplyToSomeEntries(
-    const std::function<void(const Slice& key, void* value, size_t charge,
-                             DeleterFn deleter)>& callback,
+    const std::function<void(const Slice& key, Cache::ObjectPtr value,
+                             size_t charge,
+                             const Cache::CacheItemHelper* helper)>& callback,
     size_t average_entries_per_lock, size_t* state) {
   // The state is essentially going to be the starting hash, which works
   // nicely even if we resize between calls because we use upper-most
@@ -1030,11 +1084,12 @@ void ClockCacheShard<Table>::ApplyToSomeEntries(
     *state = index_end << (sizeof(size_t) * 8u - length_bits);
   }
 
+  auto hash_seed = table_.GetHashSeed();
   table_.ConstApplyToEntriesRange(
-      [callback](const HandleImpl& h) {
+      [callback, hash_seed](const HandleImpl& h) {
         UniqueId64x2 unhashed;
-        callback(ReverseHash(h.hashed_key, &unhashed), h.value,
-                 h.GetTotalCharge(), h.deleter);
+        callback(ReverseHash(h.hashed_key, &unhashed, hash_seed), h.value,
+                 h.GetTotalCharge(), h.helper);
       },
       index_begin, index_end, false);
 }
@@ -1078,9 +1133,9 @@ void ClockCacheShard<Table>::SetStrictCapacityLimit(
 template <class Table>
 Status ClockCacheShard<Table>::Insert(const Slice& key,
                                       const UniqueId64x2& hashed_key,
-                                      void* value, size_t charge,
-                                      Cache::DeleterFn deleter,
-                                      HandleImpl** handle,
+                                      Cache::ObjectPtr value,
+                                      const Cache::CacheItemHelper* helper,
+                                      size_t charge, HandleImpl** handle,
                                       Cache::Priority priority) {
   if (UNLIKELY(key.size() != kCacheKeySize)) {
     return Status::NotSupported("ClockCache only supports key size " +
@@ -1089,12 +1144,31 @@ Status ClockCacheShard<Table>::Insert(const Slice& key,
   ClockHandleBasicData proto;
   proto.hashed_key = hashed_key;
   proto.value = value;
-  proto.deleter = deleter;
+  proto.helper = helper;
   proto.total_charge = charge;
-  Status s = table_.Insert(
-      proto, handle, priority, capacity_.load(std::memory_order_relaxed),
-      strict_capacity_limit_.load(std::memory_order_relaxed));
-  return s;
+  return table_.Insert(proto, handle, priority,
+                       capacity_.load(std::memory_order_relaxed),
+                       strict_capacity_limit_.load(std::memory_order_relaxed));
+}
+
+template <class Table>
+typename ClockCacheShard<Table>::HandleImpl*
+ClockCacheShard<Table>::CreateStandalone(const Slice& key,
+                                         const UniqueId64x2& hashed_key,
+                                         Cache::ObjectPtr obj,
+                                         const Cache::CacheItemHelper* helper,
+                                         size_t charge, bool allow_uncharged) {
+  if (UNLIKELY(key.size() != kCacheKeySize)) {
+    return nullptr;
+  }
+  ClockHandleBasicData proto;
+  proto.hashed_key = hashed_key;
+  proto.value = obj;
+  proto.helper = helper;
+  proto.total_charge = charge;
+  return table_.CreateStandalone(
+      proto, capacity_.load(std::memory_order_relaxed),
+      strict_capacity_limit_.load(std::memory_order_relaxed), allow_uncharged);
 }
 
 template <class Table>
@@ -1155,8 +1229,8 @@ size_t ClockCacheShard<Table>::GetUsage() const {
 }
 
 template <class Table>
-size_t ClockCacheShard<Table>::GetDetachedUsage() const {
-  return table_.GetDetachedUsage();
+size_t ClockCacheShard<Table>::GetStandaloneUsage() const {
+  return table_.GetStandaloneUsage();
 }
 
 template <class Table>
@@ -1190,7 +1264,7 @@ size_t ClockCacheShard<Table>::GetPinnedUsage() const {
       },
       0, table_.GetTableSize(), true);
 
-  return table_pinned_usage + table_.GetDetachedUsage();
+  return table_pinned_usage + table_.GetStandaloneUsage();
 }
 
 template <class Table>
@@ -1211,27 +1285,24 @@ size_t ClockCacheShard<Table>::GetTableAddressCount() const {
 // Explicit instantiation
 template class ClockCacheShard<HyperClockTable>;
 
-HyperClockCache::HyperClockCache(
-    size_t capacity, size_t estimated_value_size, int num_shard_bits,
-    bool strict_capacity_limit,
-    CacheMetadataChargePolicy metadata_charge_policy,
-    std::shared_ptr<MemoryAllocator> memory_allocator)
-    : ShardedCache(capacity, num_shard_bits, strict_capacity_limit,
-                   std::move(memory_allocator)) {
-  assert(estimated_value_size > 0 ||
-         metadata_charge_policy != kDontChargeCacheMetadata);
+HyperClockCache::HyperClockCache(const HyperClockCacheOptions& opts)
+    : ShardedCache(opts) {
+  assert(opts.estimated_entry_charge > 0 ||
+         opts.metadata_charge_policy != kDontChargeCacheMetadata);
   // TODO: should not need to go through two levels of pointer indirection to
   // get to table entries
   size_t per_shard = GetPerShardCapacity();
-  InitShards([=](Shard* cs) {
-    HyperClockTable::Opts opts;
-    opts.estimated_value_size = estimated_value_size;
-    new (cs)
-        Shard(per_shard, strict_capacity_limit, metadata_charge_policy, opts);
+  MemoryAllocator* alloc = this->memory_allocator();
+  InitShards([&](Shard* cs) {
+    HyperClockTable::Opts table_opts;
+    table_opts.estimated_value_size = opts.estimated_entry_charge;
+    new (cs) Shard(per_shard, opts.strict_capacity_limit,
+                   opts.metadata_charge_policy, alloc, &eviction_callback_,
+                   &hash_seed_, table_opts);
   });
 }
 
-void* HyperClockCache::Value(Handle* handle) {
+Cache::ObjectPtr HyperClockCache::Value(Handle* handle) {
   return reinterpret_cast<const HandleImpl*>(handle)->value;
 }
 
@@ -1239,9 +1310,10 @@ size_t HyperClockCache::GetCharge(Handle* handle) const {
   return reinterpret_cast<const HandleImpl*>(handle)->GetTotalCharge();
 }
 
-Cache::DeleterFn HyperClockCache::GetDeleter(Handle* handle) const {
+const Cache::CacheItemHelper* HyperClockCache::GetCacheItemHelper(
+    Handle* handle) const {
   auto h = reinterpret_cast<const HandleImpl*>(handle);
-  return h->deleter;
+  return h->helper;
 }
 
 namespace {
@@ -1256,7 +1328,7 @@ namespace {
 void AddShardEvaluation(const HyperClockCache::Shard& shard,
                         std::vector<double>& predicted_load_factors,
                         size_t& min_recommendation) {
-  size_t usage = shard.GetUsage() - shard.GetDetachedUsage();
+  size_t usage = shard.GetUsage() - shard.GetStandaloneUsage();
   size_t capacity = shard.GetCapacity();
   double usage_ratio = 1.0 * usage / capacity;
 
@@ -1386,19 +1458,25 @@ std::shared_ptr<Cache> NewClockCache(
 }
 
 std::shared_ptr<Cache> HyperClockCacheOptions::MakeSharedCache() const {
-  auto my_num_shard_bits = num_shard_bits;
-  if (my_num_shard_bits >= 20) {
+  // For sanitized options
+  HyperClockCacheOptions opts = *this;
+  if (opts.num_shard_bits >= 20) {
     return nullptr;  // The cache cannot be sharded into too many fine pieces.
   }
-  if (my_num_shard_bits < 0) {
+  if (opts.num_shard_bits < 0) {
     // Use larger shard size to reduce risk of large entries clustering
     // or skewing individual shards.
     constexpr size_t min_shard_size = 32U * 1024U * 1024U;
-    my_num_shard_bits = GetDefaultCacheShardBits(capacity, min_shard_size);
+    opts.num_shard_bits =
+        GetDefaultCacheShardBits(opts.capacity, min_shard_size);
   }
-  return std::make_shared<clock_cache::HyperClockCache>(
-      capacity, estimated_entry_charge, my_num_shard_bits,
-      strict_capacity_limit, metadata_charge_policy, memory_allocator);
+  std::shared_ptr<Cache> cache =
+      std::make_shared<clock_cache::HyperClockCache>(opts);
+  if (opts.secondary_cache) {
+    cache = std::make_shared<CacheWithSecondaryAdapter>(cache,
+                                                        opts.secondary_cache);
+  }
+  return cache;
 }
 
 }  // namespace ROCKSDB_NAMESPACE

cache/clock_cache.h

@@ -145,7 +145,7 @@ class ClockCacheTest;
 //     (erased by user) but can be read by existing references, and ref count
 //     changed by Ref and Release.
 //
-// A special case is "detached" entries, which are heap-allocated handles
+// A special case is "standalone" entries, which are heap-allocated handles
 // not in the table. They are always Invisible and freed on zero refs.
 //
 // State transitions:
@@ -200,8 +200,8 @@ class ClockCacheTest;
 // table occupancy limit has been reached. If strict_capacity_limit=false,
 // we must never fail Insert, and if a Handle* is provided, we have to return
 // a usable Cache handle on success. The solution to this (typically rare)
-// problem is "detached" handles, which are usable by the caller but not
-// actually available for Lookup in the Cache. Detached handles are allocated
+// problem is "standalone" handles, which are usable by the caller but not
+// actually available for Lookup in the Cache. Standalone handles are allocated
 // independently on the heap and specially marked so that they are freed on
 // the heap when their last reference is released.
 //
@@ -305,23 +305,17 @@ constexpr double kLoadFactor = 0.7;
 constexpr double kStrictLoadFactor = 0.84;
 
 struct ClockHandleBasicData {
-  void* value = nullptr;
-  Cache::DeleterFn deleter = nullptr;
+  Cache::ObjectPtr value = nullptr;
+  const Cache::CacheItemHelper* helper = nullptr;
   // A lossless, reversible hash of the fixed-size (16 byte) cache key. This
   // eliminates the need to store a hash separately.
   UniqueId64x2 hashed_key = kNullUniqueId64x2;
   size_t total_charge = 0;
 
-  // For total_charge_and_flags
-  // "Detached" means the handle is allocated separately from hash table.
-  static constexpr uint64_t kFlagDetached = uint64_t{1} << 63;
-  // Extract just the total charge
-  static constexpr uint64_t kTotalChargeMask = kFlagDetached - 1;
-
   inline size_t GetTotalCharge() const { return total_charge; }
 
   // Calls deleter (if non-null) on cache key and value
-  void FreeData() const;
+  void FreeData(MemoryAllocator* allocator) const;
 
   // Required by concept HandleImpl
   const UniqueId64x2& GetHash() const { return hashed_key; }
@@ -398,11 +392,11 @@ class HyperClockTable {
     // TODO: ideally this would be packed into some other data field, such
     // as upper bits of total_charge, but that incurs a measurable performance
     // regression.
-    bool detached = false;
+    bool standalone = false;
 
-    inline bool IsDetached() const { return detached; }
+    inline bool IsStandalone() const { return standalone; }
 
-    inline void SetDetached() { detached = true; }
+    inline void SetStandalone() { standalone = true; }
   };  // struct HandleImpl
 
   struct Opts {
@@ -411,13 +405,19 @@ class HyperClockTable {
 
   HyperClockTable(size_t capacity, bool strict_capacity_limit,
                   CacheMetadataChargePolicy metadata_charge_policy,
-                  const Opts& opts);
+                  MemoryAllocator* allocator,
+                  const Cache::EvictionCallback* eviction_callback,
+                  const uint32_t* hash_seed, const Opts& opts);
   ~HyperClockTable();
 
   Status Insert(const ClockHandleBasicData& proto, HandleImpl** handle,
                 Cache::Priority priority, size_t capacity,
                 bool strict_capacity_limit);
 
+  HandleImpl* CreateStandalone(ClockHandleBasicData& proto, size_t capacity,
+                               bool strict_capacity_limit,
+                               bool allow_uncharged);
+
   HandleImpl* Lookup(const UniqueId64x2& hashed_key);
 
   bool Release(HandleImpl* handle, bool useful, bool erase_if_last_ref);
@@ -444,10 +444,12 @@ class HyperClockTable {
 
   size_t GetUsage() const { return usage_.load(std::memory_order_relaxed); }
 
-  size_t GetDetachedUsage() const {
-    return detached_usage_.load(std::memory_order_relaxed);
+  size_t GetStandaloneUsage() const {
+    return standalone_usage_.load(std::memory_order_relaxed);
   }
 
+  uint32_t GetHashSeed() const { return hash_seed_; }
+
   // Acquire/release N references
   void TEST_RefN(HandleImpl& handle, size_t n);
   void TEST_ReleaseN(HandleImpl* handle, size_t n);
@@ -514,10 +516,12 @@ class HyperClockTable {
                                              size_t capacity,
                                              bool need_evict_for_occupancy);
 
-  // Creates a "detached" handle for returning from an Insert operation that
+  // Creates a "standalone" handle for returning from an Insert operation that
   // cannot be completed by actually inserting into the table.
-  // Updates `detached_usage_` but not `usage_` nor `occupancy_`.
-  inline HandleImpl* DetachedInsert(const ClockHandleBasicData& proto);
+  // Updates `standalone_usage_` but not `usage_` nor `occupancy_`.
+  inline HandleImpl* StandaloneInsert(const ClockHandleBasicData& proto);
+
+  MemoryAllocator* GetAllocator() const { return allocator_; }
 
   // Returns the number of bits used to hash an element in the hash
   // table.
@@ -538,6 +542,15 @@ class HyperClockTable {
   // Array of slots comprising the hash table.
   const std::unique_ptr<HandleImpl[]> array_;
 
+  // From Cache, for deleter
+  MemoryAllocator* const allocator_;
+
+  // A reference to Cache::eviction_callback_
+  const Cache::EvictionCallback& eviction_callback_;
+
+  // A reference to ShardedCacheBase::hash_seed_
+  const uint32_t& hash_seed_;
+
   // We partition the following members into different cache lines
   // to avoid false sharing among Lookup, Release, Erase and Insert
   // operations in ClockCacheShard.
@@ -550,11 +563,11 @@ class HyperClockTable {
   // Number of elements in the table.
   std::atomic<size_t> occupancy_{};
 
-  // Memory usage by entries tracked by the cache (including detached)
+  // Memory usage by entries tracked by the cache (including standalone)
   std::atomic<size_t> usage_{};
 
-  // Part of usage by detached entries (not in table)
-  std::atomic<size_t> detached_usage_{};
+  // Part of usage by standalone entries (not in table)
+  std::atomic<size_t> standalone_usage_{};
 };  // class HyperClockTable
 
 // A single shard of sharded cache.
@@ -563,7 +576,9 @@ class ALIGN_AS(CACHE_LINE_SIZE) ClockCacheShard final : public CacheShardBase {
  public:
   ClockCacheShard(size_t capacity, bool strict_capacity_limit,
                   CacheMetadataChargePolicy metadata_charge_policy,
-                  const typename Table::Opts& opts);
+                  MemoryAllocator* allocator,
+                  const Cache::EvictionCallback* eviction_callback,
+                  const uint32_t* hash_seed, const typename Table::Opts& opts);
 
   // For CacheShard concept
   using HandleImpl = typename Table::HandleImpl;
@@ -573,22 +588,23 @@ class ALIGN_AS(CACHE_LINE_SIZE) ClockCacheShard final : public CacheShardBase {
   static inline uint32_t HashPieceForSharding(HashCref hash) {
     return Upper32of64(hash[0]);
   }
-  static inline HashVal ComputeHash(const Slice& key) {
+  static inline HashVal ComputeHash(const Slice& key, uint32_t seed) {
     assert(key.size() == kCacheKeySize);
     HashVal in;
     HashVal out;
     // NOTE: endian dependence
     // TODO: use GetUnaligned?
     std::memcpy(&in, key.data(), kCacheKeySize);
-    BijectiveHash2x64(in[1], in[0], &out[1], &out[0]);
+    BijectiveHash2x64(in[1], in[0] ^ seed, &out[1], &out[0]);
     return out;
   }
 
   // For reconstructing key from hashed_key. Requires the caller to provide
   // backing storage for the Slice in `unhashed`
   static inline Slice ReverseHash(const UniqueId64x2& hashed,
-                                  UniqueId64x2* unhashed) {
+                                  UniqueId64x2* unhashed, uint32_t seed) {
     BijectiveUnhash2x64(hashed[1], hashed[0], &(*unhashed)[1], &(*unhashed)[0]);
+    (*unhashed)[0] ^= seed;
     // NOTE: endian dependence
     return Slice(reinterpret_cast<const char*>(unhashed), kCacheKeySize);
   }
@@ -600,9 +616,14 @@ class ALIGN_AS(CACHE_LINE_SIZE) ClockCacheShard final : public CacheShardBase {
 
   void SetStrictCapacityLimit(bool strict_capacity_limit);
 
-  Status Insert(const Slice& key, const UniqueId64x2& hashed_key, void* value,
-                size_t charge, Cache::DeleterFn deleter, HandleImpl** handle,
-                Cache::Priority priority);
+  Status Insert(const Slice& key, const UniqueId64x2& hashed_key,
+                Cache::ObjectPtr value, const Cache::CacheItemHelper* helper,
+                size_t charge, HandleImpl** handle, Cache::Priority priority);
+
+  HandleImpl* CreateStandalone(const Slice& key, const UniqueId64x2& hashed_key,
+                               Cache::ObjectPtr obj,
+                               const Cache::CacheItemHelper* helper,
+                               size_t charge, bool allow_uncharged);
 
   HandleImpl* Lookup(const Slice& key, const UniqueId64x2& hashed_key);
 
@@ -618,7 +639,7 @@ class ALIGN_AS(CACHE_LINE_SIZE) ClockCacheShard final : public CacheShardBase {
 
   size_t GetUsage() const;
 
-  size_t GetDetachedUsage() const;
+  size_t GetStandaloneUsage() const;
 
   size_t GetPinnedUsage() const;
 
@@ -629,34 +650,22 @@ class ALIGN_AS(CACHE_LINE_SIZE) ClockCacheShard final : public CacheShardBase {
   size_t GetTableAddressCount() const;
 
   void ApplyToSomeEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
+      const std::function<void(const Slice& key, Cache::ObjectPtr obj,
+                               size_t charge,
+                               const Cache::CacheItemHelper* helper)>& callback,
       size_t average_entries_per_lock, size_t* state);
 
   void EraseUnRefEntries();
 
   std::string GetPrintableOptions() const { return std::string{}; }
 
-  // SecondaryCache not yet supported
-  Status Insert(const Slice& key, const UniqueId64x2& hashed_key, void* value,
-                const Cache::CacheItemHelper* helper, size_t charge,
-                HandleImpl** handle, Cache::Priority priority) {
-    return Insert(key, hashed_key, value, charge, helper->del_cb, handle,
-                  priority);
-  }
-
   HandleImpl* Lookup(const Slice& key, const UniqueId64x2& hashed_key,
                      const Cache::CacheItemHelper* /*helper*/,
-                     const Cache::CreateCallback& /*create_cb*/,
-                     Cache::Priority /*priority*/, bool /*wait*/,
-                     Statistics* /*stats*/) {
+                     Cache::CreateContext* /*create_context*/,
+                     Cache::Priority /*priority*/, Statistics* /*stats*/) {
     return Lookup(key, hashed_key);
   }
 
-  bool IsReady(HandleImpl* /*handle*/) { return true; }
-
-  void Wait(HandleImpl* /*handle*/) {}
-
   // Acquire/release N references
   void TEST_RefN(HandleImpl* handle, size_t n);
   void TEST_ReleaseN(HandleImpl* handle, size_t n);
@@ -679,18 +688,15 @@ class HyperClockCache
  public:
   using Shard = ClockCacheShard<HyperClockTable>;
 
-  HyperClockCache(size_t capacity, size_t estimated_value_size,
-                  int num_shard_bits, bool strict_capacity_limit,
-                  CacheMetadataChargePolicy metadata_charge_policy,
-                  std::shared_ptr<MemoryAllocator> memory_allocator);
+  explicit HyperClockCache(const HyperClockCacheOptions& opts);
 
   const char* Name() const override { return "HyperClockCache"; }
 
-  void* Value(Handle* handle) override;
+  Cache::ObjectPtr Value(Handle* handle) override;
 
   size_t GetCharge(Handle* handle) const override;
 
-  DeleterFn GetDeleter(Handle* handle) const override;
+  const CacheItemHelper* GetCacheItemHelper(Handle* handle) const override;
 
   void ReportProblems(
       const std::shared_ptr<Logger>& /*info_log*/) const override;

cache/compressed_secondary_cache.cc

@@ -9,7 +9,7 @@
 #include <cstdint>
 #include <memory>
 
-#include "memory/memory_allocator.h"
+#include "memory/memory_allocator_impl.h"
 #include "monitoring/perf_context_imp.h"
 #include "util/compression.h"
 #include "util/string_util.h"
@@ -17,30 +17,18 @@
 namespace ROCKSDB_NAMESPACE {
 
 CompressedSecondaryCache::CompressedSecondaryCache(
-    size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-    double high_pri_pool_ratio, double low_pri_pool_ratio,
-    std::shared_ptr<MemoryAllocator> memory_allocator, bool use_adaptive_mutex,
-    CacheMetadataChargePolicy metadata_charge_policy,
-    CompressionType compression_type, uint32_t compress_format_version,
-    bool enable_custom_split_merge)
-    : cache_options_(capacity, num_shard_bits, strict_capacity_limit,
-                     high_pri_pool_ratio, low_pri_pool_ratio, memory_allocator,
-                     use_adaptive_mutex, metadata_charge_policy,
-                     compression_type, compress_format_version,
-                     enable_custom_split_merge) {
-  cache_ =
-      NewLRUCache(capacity, num_shard_bits, strict_capacity_limit,
-                  high_pri_pool_ratio, memory_allocator, use_adaptive_mutex,
-                  metadata_charge_policy, low_pri_pool_ratio);
-}
+    const CompressedSecondaryCacheOptions& opts)
+    : cache_(opts.LRUCacheOptions::MakeSharedCache()), cache_options_(opts) {}
 
 CompressedSecondaryCache::~CompressedSecondaryCache() { cache_.reset(); }
 
 std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
-    const Slice& key, const Cache::CreateCallback& create_cb, bool /*wait*/,
-    bool advise_erase, bool& is_in_sec_cache) {
+    const Slice& key, const Cache::CacheItemHelper* helper,
+    Cache::CreateContext* create_context, bool /*wait*/, bool advise_erase,
+    bool& kept_in_sec_cache) {
+  assert(helper);
   std::unique_ptr<SecondaryCacheResultHandle> handle;
-  is_in_sec_cache = false;
+  kept_in_sec_cache = false;
   Cache::Handle* lru_handle = cache_->Lookup(key);
   if (lru_handle == nullptr) {
     return nullptr;
@@ -64,12 +52,15 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
     ptr = reinterpret_cast<CacheAllocationPtr*>(handle_value);
     handle_value_charge = cache_->GetCharge(lru_handle);
   }
+  MemoryAllocator* allocator = cache_options_.memory_allocator.get();
 
   Status s;
-  void* value{nullptr};
+  Cache::ObjectPtr value{nullptr};
   size_t charge{0};
-  if (cache_options_.compression_type == kNoCompression) {
-    s = create_cb(ptr->get(), handle_value_charge, &value, &charge);
+  if (cache_options_.compression_type == kNoCompression ||
+      cache_options_.do_not_compress_roles.Contains(helper->role)) {
+    s = helper->create_cb(Slice(ptr->get(), handle_value_charge),
+                          create_context, allocator, &value, &charge);
   } else {
     UncompressionContext uncompression_context(cache_options_.compression_type);
     UncompressionInfo uncompression_info(uncompression_context,
@@ -79,14 +70,14 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
     size_t uncompressed_size{0};
     CacheAllocationPtr uncompressed = UncompressData(
         uncompression_info, (char*)ptr->get(), handle_value_charge,
-        &uncompressed_size, cache_options_.compress_format_version,
-        cache_options_.memory_allocator.get());
+        &uncompressed_size, cache_options_.compress_format_version, allocator);
 
     if (!uncompressed) {
       cache_->Release(lru_handle, /*erase_if_last_ref=*/true);
       return nullptr;
     }
-    s = create_cb(uncompressed.get(), uncompressed_size, &value, &charge);
+    s = helper->create_cb(Slice(uncompressed.get(), uncompressed_size),
+                          create_context, allocator, &value, &charge);
   }
 
   if (!s.ok()) {
@@ -98,30 +89,32 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
     cache_->Release(lru_handle, /*erase_if_last_ref=*/true);
     // Insert a dummy handle.
     cache_
-        ->Insert(key, /*value=*/nullptr, /*charge=*/0,
-                 GetDeletionCallback(cache_options_.enable_custom_split_merge))
+        ->Insert(key, /*obj=*/nullptr,
+                 GetHelper(cache_options_.enable_custom_split_merge),
+                 /*charge=*/0)
         .PermitUncheckedError();
   } else {
-    is_in_sec_cache = true;
+    kept_in_sec_cache = true;
     cache_->Release(lru_handle, /*erase_if_last_ref=*/false);
   }
   handle.reset(new CompressedSecondaryCacheResultHandle(value, charge));
   return handle;
 }
 
-Status CompressedSecondaryCache::Insert(const Slice& key, void* value,
+Status CompressedSecondaryCache::Insert(const Slice& key,
+                                        Cache::ObjectPtr value,
                                         const Cache::CacheItemHelper* helper) {
   if (value == nullptr) {
     return Status::InvalidArgument();
   }
 
   Cache::Handle* lru_handle = cache_->Lookup(key);
-  Cache::DeleterFn del_cb =
-      GetDeletionCallback(cache_options_.enable_custom_split_merge);
+  auto internal_helper = GetHelper(cache_options_.enable_custom_split_merge);
   if (lru_handle == nullptr) {
     PERF_COUNTER_ADD(compressed_sec_cache_insert_dummy_count, 1);
     // Insert a dummy handle if the handle is evicted for the first time.
-    return cache_->Insert(key, /*value=*/nullptr, /*charge=*/0, del_cb);
+    return cache_->Insert(key, /*obj=*/nullptr, internal_helper,
+                          /*charge=*/0);
   } else {
     cache_->Release(lru_handle, /*erase_if_last_ref=*/false);
   }
@@ -137,7 +130,8 @@ Status CompressedSecondaryCache::Insert(const Slice& key, void* value,
   Slice val(ptr.get(), size);
 
   std::string compressed_val;
-  if (cache_options_.compression_type != kNoCompression) {
+  if (cache_options_.compression_type != kNoCompression &&
+      !cache_options_.do_not_compress_roles.Contains(helper->role)) {
     PERF_COUNTER_ADD(compressed_sec_cache_uncompressed_bytes, size);
     CompressionOptions compression_opts;
     CompressionContext compression_context(cache_options_.compression_type);
@@ -169,10 +163,10 @@ Status CompressedSecondaryCache::Insert(const Slice& key, void* value,
     size_t charge{0};
     CacheValueChunk* value_chunks_head =
         SplitValueIntoChunks(val, cache_options_.compression_type, charge);
-    return cache_->Insert(key, value_chunks_head, charge, del_cb);
+    return cache_->Insert(key, value_chunks_head, internal_helper, charge);
   } else {
     CacheAllocationPtr* buf = new CacheAllocationPtr(std::move(ptr));
-    return cache_->Insert(key, buf, size, del_cb);
+    return cache_->Insert(key, buf, internal_helper, size);
   }
 }
 
@@ -276,50 +270,35 @@ CacheAllocationPtr CompressedSecondaryCache::MergeChunksIntoValue(
   return ptr;
 }
 
-Cache::DeleterFn CompressedSecondaryCache::GetDeletionCallback(
-    bool enable_custom_split_merge) {
+const Cache::CacheItemHelper* CompressedSecondaryCache::GetHelper(
+    bool enable_custom_split_merge) const {
   if (enable_custom_split_merge) {
-    return [](const Slice& /*key*/, void* obj) {
-      CacheValueChunk* chunks_head = reinterpret_cast<CacheValueChunk*>(obj);
-      while (chunks_head != nullptr) {
-        CacheValueChunk* tmp_chunk = chunks_head;
-        chunks_head = chunks_head->next;
-        tmp_chunk->Free();
-        obj = nullptr;
-      };
-    };
+    static const Cache::CacheItemHelper kHelper{
+        CacheEntryRole::kMisc,
+        [](Cache::ObjectPtr obj, MemoryAllocator* /*alloc*/) {
+          CacheValueChunk* chunks_head = static_cast<CacheValueChunk*>(obj);
+          while (chunks_head != nullptr) {
+            CacheValueChunk* tmp_chunk = chunks_head;
+            chunks_head = chunks_head->next;
+            tmp_chunk->Free();
+            obj = nullptr;
+          };
+        }};
+    return &kHelper;
   } else {
-    return [](const Slice& /*key*/, void* obj) {
-      delete reinterpret_cast<CacheAllocationPtr*>(obj);
-      obj = nullptr;
-    };
+    static const Cache::CacheItemHelper kHelper{
+        CacheEntryRole::kMisc,
+        [](Cache::ObjectPtr obj, MemoryAllocator* /*alloc*/) {
+          delete static_cast<CacheAllocationPtr*>(obj);
+          obj = nullptr;
+        }};
+    return &kHelper;
   }
 }
 
-std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
-    size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-    double high_pri_pool_ratio, double low_pri_pool_ratio,
-    std::shared_ptr<MemoryAllocator> memory_allocator, bool use_adaptive_mutex,
-    CacheMetadataChargePolicy metadata_charge_policy,
-    CompressionType compression_type, uint32_t compress_format_version,
-    bool enable_custom_split_merge) {
-  return std::make_shared<CompressedSecondaryCache>(
-      capacity, num_shard_bits, strict_capacity_limit, high_pri_pool_ratio,
-      low_pri_pool_ratio, memory_allocator, use_adaptive_mutex,
-      metadata_charge_policy, compression_type, compress_format_version,
-      enable_custom_split_merge);
-}
-
-std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
-    const CompressedSecondaryCacheOptions& opts) {
-  // The secondary_cache is disabled for this LRUCache instance.
-  assert(opts.secondary_cache == nullptr);
-  return NewCompressedSecondaryCache(
-      opts.capacity, opts.num_shard_bits, opts.strict_capacity_limit,
-      opts.high_pri_pool_ratio, opts.low_pri_pool_ratio, opts.memory_allocator,
-      opts.use_adaptive_mutex, opts.metadata_charge_policy,
-      opts.compression_type, opts.compress_format_version,
-      opts.enable_custom_split_merge);
+std::shared_ptr<SecondaryCache>
+CompressedSecondaryCacheOptions::MakeSharedSecondaryCache() const {
+  return std::make_shared<CompressedSecondaryCache>(*this);
 }
 
 }  // namespace ROCKSDB_NAMESPACE

cache/compressed_secondary_cache.h

@@ -10,7 +10,7 @@
 #include <memory>
 
 #include "cache/lru_cache.h"
-#include "memory/memory_allocator.h"
+#include "memory/memory_allocator_impl.h"
 #include "rocksdb/secondary_cache.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
@@ -21,7 +21,7 @@ namespace ROCKSDB_NAMESPACE {
 
 class CompressedSecondaryCacheResultHandle : public SecondaryCacheResultHandle {
  public:
-  CompressedSecondaryCacheResultHandle(void* value, size_t size)
+  CompressedSecondaryCacheResultHandle(Cache::ObjectPtr value, size_t size)
       : value_(value), size_(size) {}
   ~CompressedSecondaryCacheResultHandle() override = default;
 
@@ -34,12 +34,12 @@ class CompressedSecondaryCacheResultHandle : public SecondaryCacheResultHandle {
 
   void Wait() override {}
 
-  void* Value() override { return value_; }
+  Cache::ObjectPtr Value() override { return value_; }
 
   size_t Size() override { return size_; }
 
  private:
-  void* value_;
+  Cache::ObjectPtr value_;
   size_t size_;
 };
 
@@ -69,26 +69,19 @@ class CompressedSecondaryCacheResultHandle : public SecondaryCacheResultHandle {
 
 class CompressedSecondaryCache : public SecondaryCache {
  public:
-  CompressedSecondaryCache(
-      size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-      double high_pri_pool_ratio, double low_pri_pool_ratio,
-      std::shared_ptr<MemoryAllocator> memory_allocator = nullptr,
-      bool use_adaptive_mutex = kDefaultToAdaptiveMutex,
-      CacheMetadataChargePolicy metadata_charge_policy =
-          kDefaultCacheMetadataChargePolicy,
-      CompressionType compression_type = CompressionType::kLZ4Compression,
-      uint32_t compress_format_version = 2,
-      bool enable_custom_split_merge = false);
+  explicit CompressedSecondaryCache(
+      const CompressedSecondaryCacheOptions& opts);
   ~CompressedSecondaryCache() override;
 
   const char* Name() const override { return "CompressedSecondaryCache"; }
 
-  Status Insert(const Slice& key, void* value,
+  Status Insert(const Slice& key, Cache::ObjectPtr value,
                 const Cache::CacheItemHelper* helper) override;
 
   std::unique_ptr<SecondaryCacheResultHandle> Lookup(
-      const Slice& key, const Cache::CreateCallback& create_cb, bool /*wait*/,
-      bool advise_erase, bool& is_in_sec_cache) override;
+      const Slice& key, const Cache::CacheItemHelper* helper,
+      Cache::CreateContext* create_context, bool /*wait*/, bool advise_erase,
+      bool& kept_in_sec_cache) override;
 
   bool SupportForceErase() const override { return true; }
 
@@ -103,7 +96,7 @@ class CompressedSecondaryCache : public SecondaryCache {
   std::string GetPrintableOptions() const override;
 
  private:
-  friend class CompressedSecondaryCacheTest;
+  friend class CompressedSecondaryCacheTestBase;
   static constexpr std::array<uint16_t, 8> malloc_bin_sizes_{
       128, 256, 512, 1024, 2048, 4096, 8192, 16384};
 
@@ -129,8 +122,8 @@ class CompressedSecondaryCache : public SecondaryCache {
   CacheAllocationPtr MergeChunksIntoValue(const void* chunks_head,
                                           size_t& charge);
 
-  // An implementation of Cache::DeleterFn.
-  static Cache::DeleterFn GetDeletionCallback(bool enable_custom_split_merge);
+  // TODO: clean up to use cleaner interfaces in typed_cache.h
+  const Cache::CacheItemHelper* GetHelper(bool enable_custom_split_merge) const;
   std::shared_ptr<Cache> cache_;
   CompressedSecondaryCacheOptions cache_options_;
   mutable port::Mutex capacity_mutex_;

cache/compressed_secondary_cache_test.cc

@@ -5,86 +5,43 @@
 
 #include "cache/compressed_secondary_cache.h"
 
+#include <array>
 #include <iterator>
 #include <memory>
 #include <tuple>
 
 #include "memory/jemalloc_nodump_allocator.h"
 #include "rocksdb/convenience.h"
+#include "test_util/secondary_cache_test_util.h"
 #include "test_util/testharness.h"
 #include "test_util/testutil.h"
 
 namespace ROCKSDB_NAMESPACE {
 
-class CompressedSecondaryCacheTest : public testing::Test {
+using secondary_cache_test_util::GetTestingCacheTypes;
+using secondary_cache_test_util::WithCacheType;
+
+// 16 bytes for HCC compatibility
+const std::string key0 = "____    ____key0";
+const std::string key1 = "____    ____key1";
+const std::string key2 = "____    ____key2";
+const std::string key3 = "____    ____key3";
+
+class CompressedSecondaryCacheTestBase : public testing::Test,
+                                         public WithCacheType {
  public:
-  CompressedSecondaryCacheTest() : fail_create_(false) {}
-  ~CompressedSecondaryCacheTest() override = default;
+  CompressedSecondaryCacheTestBase() {}
+  ~CompressedSecondaryCacheTestBase() override = default;
 
  protected:
-  class TestItem {
-   public:
-    TestItem(const char* buf, size_t size) : buf_(new char[size]), size_(size) {
-      memcpy(buf_.get(), buf, size);
-    }
-    ~TestItem() = default;
-
-    char* Buf() { return buf_.get(); }
-    [[nodiscard]] size_t Size() const { return size_; }
-
-   private:
-    std::unique_ptr<char[]> buf_;
-    size_t size_;
-  };
-
-  static size_t SizeCallback(void* obj) {
-    return reinterpret_cast<TestItem*>(obj)->Size();
-  }
-
-  static Status SaveToCallback(void* from_obj, size_t from_offset,
-                               size_t length, void* out) {
-    auto item = reinterpret_cast<TestItem*>(from_obj);
-    const char* buf = item->Buf();
-    EXPECT_EQ(length, item->Size());
-    EXPECT_EQ(from_offset, 0);
-    memcpy(out, buf, length);
-    return Status::OK();
-  }
-
-  static void DeletionCallback(const Slice& /*key*/, void* obj) {
-    delete reinterpret_cast<TestItem*>(obj);
-    obj = nullptr;
-  }
-
-  static Cache::CacheItemHelper helper_;
-
-  static Status SaveToCallbackFail(void* /*obj*/, size_t /*offset*/,
-                                   size_t /*size*/, void* /*out*/) {
-    return Status::NotSupported();
-  }
-
-  static Cache::CacheItemHelper helper_fail_;
-
-  Cache::CreateCallback test_item_creator = [&](const void* buf, size_t size,
-                                                void** out_obj,
-                                                size_t* charge) -> Status {
-    if (fail_create_) {
-      return Status::NotSupported();
-    }
-    *out_obj = reinterpret_cast<void*>(new TestItem((char*)buf, size));
-    *charge = size;
-    return Status::OK();
-  };
-
-  void SetFailCreate(bool fail) { fail_create_ = fail; }
-
   void BasicTestHelper(std::shared_ptr<SecondaryCache> sec_cache,
                        bool sec_cache_is_compressed) {
     get_perf_context()->Reset();
-    bool is_in_sec_cache{true};
+    bool kept_in_sec_cache{true};
     // Lookup an non-existent key.
-    std::unique_ptr<SecondaryCacheResultHandle> handle0 = sec_cache->Lookup(
-        "k0", test_item_creator, true, /*advise_erase=*/true, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle0 =
+        sec_cache->Lookup(key0, GetHelper(), this, true, /*advise_erase=*/true,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle0, nullptr);
 
     Random rnd(301);
@@ -92,25 +49,25 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::string str1(rnd.RandomString(1000));
     TestItem item1(str1.data(), str1.length());
     // A dummy handle is inserted if the item is inserted for the first time.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 1);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
 
-    std::unique_ptr<SecondaryCacheResultHandle> handle1_1 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle1_1 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle1_1, nullptr);
 
     // Insert and Lookup the item k1 for the second time and advise erasing it.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1);
 
-    std::unique_ptr<SecondaryCacheResultHandle> handle1_2 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/true, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle1_2 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/true,
+                          kept_in_sec_cache);
     ASSERT_NE(handle1_2, nullptr);
-    ASSERT_FALSE(is_in_sec_cache);
+    ASSERT_FALSE(kept_in_sec_cache);
     if (sec_cache_is_compressed) {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
                 1000);
@@ -127,22 +84,22 @@ class CompressedSecondaryCacheTest : public testing::Test {
     ASSERT_EQ(memcmp(val1->Buf(), item1.Buf(), item1.Size()), 0);
 
     // Lookup the item k1 again.
-    std::unique_ptr<SecondaryCacheResultHandle> handle1_3 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/true, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle1_3 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/true,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle1_3, nullptr);
 
     // Insert and Lookup the item k2.
     std::string str2(rnd.RandomString(1000));
     TestItem item2(str2.data(), str2.length());
-    ASSERT_OK(sec_cache->Insert("k2", &item2,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 2);
-    std::unique_ptr<SecondaryCacheResultHandle> handle2_1 = sec_cache->Lookup(
-        "k2", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle2_1 =
+        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle2_1, nullptr);
 
-    ASSERT_OK(sec_cache->Insert("k2", &item2,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 2);
     if (sec_cache_is_compressed) {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
@@ -153,8 +110,9 @@ class CompressedSecondaryCacheTest : public testing::Test {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
     }
-    std::unique_ptr<SecondaryCacheResultHandle> handle2_2 = sec_cache->Lookup(
-        "k2", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle2_2 =
+        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_NE(handle2_2, nullptr);
     std::unique_ptr<TestItem> val2 =
         std::unique_ptr<TestItem>(static_cast<TestItem*>(handle2_2->Value()));
@@ -223,28 +181,26 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::string str1(rnd.RandomString(1000));
     TestItem item1(str1.data(), str1.length());
     // Insert a dummy handle.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
     // Insert k1.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
 
     // Insert and Lookup the second item.
     std::string str2(rnd.RandomString(200));
     TestItem item2(str2.data(), str2.length());
     // Insert a dummy handle, k1 is not evicted.
-    ASSERT_OK(sec_cache->Insert("k2", &item2,
-                                &CompressedSecondaryCacheTest::helper_));
-    bool is_in_sec_cache{false};
-    std::unique_ptr<SecondaryCacheResultHandle> handle1 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper()));
+    bool kept_in_sec_cache{false};
+    std::unique_ptr<SecondaryCacheResultHandle> handle1 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle1, nullptr);
 
     // Insert k2 and k1 is evicted.
-    ASSERT_OK(sec_cache->Insert("k2", &item2,
-                                &CompressedSecondaryCacheTest::helper_));
-    std::unique_ptr<SecondaryCacheResultHandle> handle2 = sec_cache->Lookup(
-        "k2", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper()));
+    std::unique_ptr<SecondaryCacheResultHandle> handle2 =
+        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_NE(handle2, nullptr);
     std::unique_ptr<TestItem> val2 =
         std::unique_ptr<TestItem>(static_cast<TestItem*>(handle2->Value()));
@@ -252,27 +208,26 @@ class CompressedSecondaryCacheTest : public testing::Test {
     ASSERT_EQ(memcmp(val2->Buf(), item2.Buf(), item2.Size()), 0);
 
     // Insert k1 again and a dummy handle is inserted.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
 
-    std::unique_ptr<SecondaryCacheResultHandle> handle1_1 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle1_1 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle1_1, nullptr);
 
     // Create Fails.
     SetFailCreate(true);
-    std::unique_ptr<SecondaryCacheResultHandle> handle2_1 = sec_cache->Lookup(
-        "k2", test_item_creator, true, /*advise_erase=*/true, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle2_1 =
+        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/true,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle2_1, nullptr);
 
     // Save Fails.
     std::string str3 = rnd.RandomString(10);
     TestItem item3(str3.data(), str3.length());
     // The Status is OK because a dummy handle is inserted.
-    ASSERT_OK(sec_cache->Insert("k3", &item3,
-                                &CompressedSecondaryCacheTest::helper_fail_));
-    ASSERT_NOK(sec_cache->Insert("k3", &item3,
-                                 &CompressedSecondaryCacheTest::helper_fail_));
+    ASSERT_OK(sec_cache->Insert(key3, &item3, GetHelperFail()));
+    ASSERT_NOK(sec_cache->Insert(key3, &item3, GetHelperFail()));
 
     sec_cache.reset();
   }
@@ -296,28 +251,22 @@ class CompressedSecondaryCacheTest : public testing::Test {
     secondary_cache_opts.enable_custom_split_merge = enable_custom_split_merge;
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
-    LRUCacheOptions lru_cache_opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity =*/1300, /*_num_shard_bits =*/0,
-        /*_strict_capacity_limit =*/false, /*_high_pri_pool_ratio =*/0.5,
-        /*_memory_allocator =*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio =*/0.0);
-    lru_cache_opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(lru_cache_opts);
+        /*_strict_capacity_limit =*/true, secondary_cache);
     std::shared_ptr<Statistics> stats = CreateDBStatistics();
 
     get_perf_context()->Reset();
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1_1 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert(
-        "k1", item1_1, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1_1, GetHelper(), str1.length()));
 
     std::string str2 = rnd.RandomString(1012);
     auto item2_1 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k2", item2_1, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2_1, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 1);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
@@ -326,22 +275,19 @@ class CompressedSecondaryCacheTest : public testing::Test {
     auto item3_1 = new TestItem(str3.data(), str3.length());
     // After this Insert, primary cache contains k3 and secondary cache contains
     // k1's dummy item and k2's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k3", item3_1, &CompressedSecondaryCacheTest::helper_, str3.length()));
+    ASSERT_OK(cache->Insert(key3, item3_1, GetHelper(), str3.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 2);
 
     // After this Insert, primary cache contains k1 and secondary cache contains
     // k1's dummy item, k2's dummy item, and k3's dummy item.
     auto item1_2 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert(
-        "k1", item1_2, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1_2, GetHelper(), str1.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 3);
 
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's item, k2's dummy item, and k3's dummy item.
     auto item2_2 = new TestItem(str2.data(), str2.length());
-    ASSERT_OK(cache->Insert(
-        "k2", item2_2, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2_2, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1);
     if (sec_cache_is_compressed) {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
@@ -356,8 +302,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     // After this Insert, primary cache contains k3 and secondary cache contains
     // k1's item and k2's item.
     auto item3_2 = new TestItem(str3.data(), str3.length());
-    ASSERT_OK(cache->Insert(
-        "k3", item3_2, &CompressedSecondaryCacheTest::helper_, str3.length()));
+    ASSERT_OK(cache->Insert(key3, item3_2, GetHelper(), str3.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 2);
     if (sec_cache_is_compressed) {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
@@ -370,8 +315,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     }
 
     Cache::Handle* handle;
-    handle = cache->Lookup("k3", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key3, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_NE(handle, nullptr);
     auto val3 = static_cast<TestItem*>(cache->Value(handle));
@@ -380,15 +324,13 @@ class CompressedSecondaryCacheTest : public testing::Test {
     cache->Release(handle);
 
     // Lookup an non-existent key.
-    handle = cache->Lookup("k0", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key0, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_EQ(handle, nullptr);
 
     // This Lookup should just insert a dummy handle in the primary cache
     // and the k1 is still in the secondary cache.
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_NE(handle, nullptr);
     ASSERT_EQ(get_perf_context()->block_cache_standalone_handle_count, 1);
@@ -400,8 +342,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     // This Lookup should erase k1 from the secondary cache and insert
     // it into primary cache; then k3 is demoted.
     // k2 and k3 are in secondary cache.
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_NE(handle, nullptr);
     ASSERT_EQ(get_perf_context()->block_cache_standalone_handle_count, 1);
@@ -409,8 +350,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     cache->Release(handle);
 
     // k2 is still in secondary cache.
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_NE(handle, nullptr);
     ASSERT_EQ(get_perf_context()->block_cache_standalone_handle_count, 2);
@@ -418,8 +358,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
 
     // Testing SetCapacity().
     ASSERT_OK(secondary_cache->SetCapacity(0));
-    handle = cache->Lookup("k3", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key3, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_EQ(handle, nullptr);
 
@@ -429,35 +368,30 @@ class CompressedSecondaryCacheTest : public testing::Test {
     ASSERT_EQ(capacity, 7000);
     auto item1_3 = new TestItem(str1.data(), str1.length());
     // After this Insert, primary cache contains k1.
-    ASSERT_OK(cache->Insert(
-        "k1", item1_3, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key1, item1_3, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 3);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 4);
 
     auto item2_3 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k2", item2_3, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key2, item2_3, GetHelper(), str1.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 4);
 
     auto item1_4 = new TestItem(str1.data(), str1.length());
     // After this Insert, primary cache contains k1 and secondary cache contains
     // k1's dummy item and k2's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k1", item1_4, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key1, item1_4, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 5);
 
     auto item2_4 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's real item and k2's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k2", item2_4, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2_4, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 5);
     // This Lookup should just insert a dummy handle in the primary cache
     // and the k1 is still in the secondary cache.
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
 
     ASSERT_NE(handle, nullptr);
@@ -485,31 +419,31 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
 
-    LRUCacheOptions opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity=*/1300, /*_num_shard_bits=*/0,
-        /*_strict_capacity_limit=*/false, /*_high_pri_pool_ratio=*/0.5,
-        /*_memory_allocator=*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio=*/0.0);
-    opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(opts);
+        /*_strict_capacity_limit=*/false, secondary_cache);
 
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1 = std::make_unique<TestItem>(str1.data(), str1.length());
-    ASSERT_NOK(cache->Insert("k1", item1.get(), nullptr, str1.length()));
-    ASSERT_OK(cache->Insert("k1", item1.get(),
-                            &CompressedSecondaryCacheTest::helper_,
-                            str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1.get(), GetHelper(), str1.length()));
     item1.release();  // Appease clang-analyze "potential memory leak"
 
     Cache::Handle* handle;
-    handle = cache->Lookup("k2", nullptr, test_item_creator,
-                           Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, nullptr, this, Cache::Priority::LOW);
     ASSERT_EQ(handle, nullptr);
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, false);
+    handle = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_EQ(handle, nullptr);
 
+    Cache::AsyncLookupHandle ah;
+    ah.key = key2;
+    ah.helper = GetHelper();
+    ah.create_context = this;
+    ah.priority = Cache::Priority::LOW;
+    cache->StartAsyncLookup(ah);
+    cache->Wait(ah);
+    ASSERT_EQ(ah.Result(), nullptr);
+
     cache.reset();
     secondary_cache.reset();
   }
@@ -532,40 +466,29 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
 
-    LRUCacheOptions opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity=*/1300, /*_num_shard_bits=*/0,
-        /*_strict_capacity_limit=*/false, /*_high_pri_pool_ratio=*/0.5,
-        /*_memory_allocator=*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio=*/0.0);
-    opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(opts);
+        /*_strict_capacity_limit=*/true, secondary_cache);
 
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert("k1", item1,
-                            &CompressedSecondaryCacheTest::helper_fail_,
-                            str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1, GetHelperFail(), str1.length()));
 
     std::string str2 = rnd.RandomString(1002);
     auto item2 = new TestItem(str2.data(), str2.length());
     // k1 should be demoted to the secondary cache.
-    ASSERT_OK(cache->Insert("k2", item2,
-                            &CompressedSecondaryCacheTest::helper_fail_,
-                            str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2, GetHelperFail(), str2.length()));
 
     Cache::Handle* handle;
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_fail_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, GetHelperFail(), this, Cache::Priority::LOW);
     ASSERT_NE(handle, nullptr);
     cache->Release(handle);
     // This lookup should fail, since k1 demotion would have failed.
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_fail_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key1, GetHelperFail(), this, Cache::Priority::LOW);
     ASSERT_EQ(handle, nullptr);
     // Since k1 was not promoted, k2 should still be in cache.
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_fail_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, GetHelperFail(), this, Cache::Priority::LOW);
     ASSERT_NE(handle, nullptr);
     cache->Release(handle);
 
@@ -591,39 +514,30 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
 
-    LRUCacheOptions opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity=*/1300, /*_num_shard_bits=*/0,
-        /*_strict_capacity_limit=*/false, /*_high_pri_pool_ratio=*/0.5,
-        /*_memory_allocator=*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio=*/0.0);
-    opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(opts);
+        /*_strict_capacity_limit=*/true, secondary_cache);
 
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert("k1", item1, &CompressedSecondaryCacheTest::helper_,
-                            str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1, GetHelper(), str1.length()));
 
     std::string str2 = rnd.RandomString(1002);
     auto item2 = new TestItem(str2.data(), str2.length());
     // k1 should be demoted to the secondary cache.
-    ASSERT_OK(cache->Insert("k2", item2, &CompressedSecondaryCacheTest::helper_,
-                            str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2, GetHelper(), str2.length()));
 
     Cache::Handle* handle;
     SetFailCreate(true);
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle, nullptr);
     cache->Release(handle);
     // This lookup should fail, since k1 creation would have failed
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_EQ(handle, nullptr);
     // Since k1 didn't get promoted, k2 should still be in cache
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle, nullptr);
     cache->Release(handle);
 
@@ -649,43 +563,34 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
 
-    LRUCacheOptions opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity=*/1300, /*_num_shard_bits=*/0,
-        /*_strict_capacity_limit=*/false, /*_high_pri_pool_ratio=*/0.5,
-        /*_memory_allocator=*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio=*/0.0);
-    opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(opts);
+        /*_strict_capacity_limit=*/false, secondary_cache);
 
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1_1 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert(
-        "k1", item1_1, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1_1, GetHelper(), str1.length()));
 
     std::string str2 = rnd.RandomString(1002);
     std::string str2_clone{str2};
     auto item2 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's dummy item.
-    ASSERT_OK(cache->Insert("k2", item2, &CompressedSecondaryCacheTest::helper_,
-                            str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2, GetHelper(), str2.length()));
 
     // After this Insert, primary cache contains k1 and secondary cache contains
     // k1's dummy item and k2's dummy item.
     auto item1_2 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert(
-        "k1", item1_2, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1_2, GetHelper(), str1.length()));
 
     auto item2_2 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's item and k2's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k2", item2_2, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2_2, GetHelper(), str2.length()));
 
     Cache::Handle* handle2;
-    handle2 = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                            test_item_creator, Cache::Priority::LOW, true);
+    handle2 = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle2, nullptr);
     cache->Release(handle2);
 
@@ -693,14 +598,12 @@ class CompressedSecondaryCacheTest : public testing::Test {
     // strict_capacity_limit is true, but the lookup should still succeed.
     // A k1's dummy item is inserted into primary cache.
     Cache::Handle* handle1;
-    handle1 = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                            test_item_creator, Cache::Priority::LOW, true);
+    handle1 = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle1, nullptr);
     cache->Release(handle1);
 
     // Since k1 didn't get inserted, k2 should still be in cache
-    handle2 = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                            test_item_creator, Cache::Priority::LOW, true);
+    handle2 = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle2, nullptr);
     cache->Release(handle2);
 
@@ -723,8 +626,9 @@ class CompressedSecondaryCacheTest : public testing::Test {
 
     using CacheValueChunk = CompressedSecondaryCache::CacheValueChunk;
     std::unique_ptr<CompressedSecondaryCache> sec_cache =
-        std::make_unique<CompressedSecondaryCache>(1000, 0, true, 0.5, 0.0,
-                                                   allocator);
+        std::make_unique<CompressedSecondaryCache>(
+            CompressedSecondaryCacheOptions(1000, 0, true, 0.5, 0.0,
+                                            allocator));
     Random rnd(301);
     // 8500 = 8169 + 233 + 98, so there should be 3 chunks after split.
     size_t str_size{8500};
@@ -741,7 +645,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     current_chunk = current_chunk->next;
     ASSERT_EQ(current_chunk->size, 98);
 
-    sec_cache->GetDeletionCallback(true)("dummy", chunks_head);
+    sec_cache->GetHelper(true)->del_cb(chunks_head, /*alloc*/ nullptr);
   }
 
   void MergeChunksIntoValueTest() {
@@ -775,7 +679,8 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::string str = str1 + str2 + str3;
 
     std::unique_ptr<CompressedSecondaryCache> sec_cache =
-        std::make_unique<CompressedSecondaryCache>(1000, 0, true, 0.5, 0.0);
+        std::make_unique<CompressedSecondaryCache>(
+            CompressedSecondaryCacheOptions(1000, 0, true, 0.5, 0.0));
     size_t charge{0};
     CacheAllocationPtr value =
         sec_cache->MergeChunksIntoValue(chunks_head, charge);
@@ -805,8 +710,9 @@ class CompressedSecondaryCacheTest : public testing::Test {
 
     using CacheValueChunk = CompressedSecondaryCache::CacheValueChunk;
     std::unique_ptr<CompressedSecondaryCache> sec_cache =
-        std::make_unique<CompressedSecondaryCache>(1000, 0, true, 0.5, 0.0,
-                                                   allocator);
+        std::make_unique<CompressedSecondaryCache>(
+            CompressedSecondaryCacheOptions(1000, 0, true, 0.5, 0.0,
+                                            allocator));
     Random rnd(301);
     // 8500 = 8169 + 233 + 98, so there should be 3 chunks after split.
     size_t str_size{8500};
@@ -822,31 +728,29 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::string value_str{value.get(), charge};
     ASSERT_EQ(strcmp(value_str.data(), str.data()), 0);
 
-    sec_cache->GetDeletionCallback(true)("dummy", chunks_head);
+    sec_cache->GetHelper(true)->del_cb(chunks_head, /*alloc*/ nullptr);
   }
-
- private:
-  bool fail_create_;
 };
 
-Cache::CacheItemHelper CompressedSecondaryCacheTest::helper_(
-    CompressedSecondaryCacheTest::SizeCallback,
-    CompressedSecondaryCacheTest::SaveToCallback,
-    CompressedSecondaryCacheTest::DeletionCallback);
+class CompressedSecondaryCacheTest
+    : public CompressedSecondaryCacheTestBase,
+      public testing::WithParamInterface<std::string> {
+  const std::string& Type() override { return GetParam(); }
+};
 
-Cache::CacheItemHelper CompressedSecondaryCacheTest::helper_fail_(
-    CompressedSecondaryCacheTest::SizeCallback,
-    CompressedSecondaryCacheTest::SaveToCallbackFail,
-    CompressedSecondaryCacheTest::DeletionCallback);
+INSTANTIATE_TEST_CASE_P(CompressedSecondaryCacheTest,
+                        CompressedSecondaryCacheTest, GetTestingCacheTypes());
 
 class CompressedSecCacheTestWithCompressAndAllocatorParam
-    : public CompressedSecondaryCacheTest,
-      public ::testing::WithParamInterface<std::tuple<bool, bool>> {
+    : public CompressedSecondaryCacheTestBase,
+      public ::testing::WithParamInterface<
+          std::tuple<bool, bool, std::string>> {
  public:
   CompressedSecCacheTestWithCompressAndAllocatorParam() {
     sec_cache_is_compressed_ = std::get<0>(GetParam());
     use_jemalloc_ = std::get<1>(GetParam());
   }
+  const std::string& Type() override { return std::get<2>(GetParam()); }
   bool sec_cache_is_compressed_;
   bool use_jemalloc_;
 };
@@ -857,20 +761,20 @@ TEST_P(CompressedSecCacheTestWithCompressAndAllocatorParam, BasicTes) {
 
 INSTANTIATE_TEST_CASE_P(CompressedSecCacheTests,
                         CompressedSecCacheTestWithCompressAndAllocatorParam,
-                        ::testing::Combine(testing::Bool(), testing::Bool()));
+                        ::testing::Combine(testing::Bool(), testing::Bool(),
+                                           GetTestingCacheTypes()));
 
 class CompressedSecondaryCacheTestWithCompressionParam
-    : public CompressedSecondaryCacheTest,
-      public ::testing::WithParamInterface<bool> {
+    : public CompressedSecondaryCacheTestBase,
+      public ::testing::WithParamInterface<std::tuple<bool, std::string>> {
  public:
   CompressedSecondaryCacheTestWithCompressionParam() {
-    sec_cache_is_compressed_ = GetParam();
+    sec_cache_is_compressed_ = std::get<0>(GetParam());
   }
+  const std::string& Type() override { return std::get<1>(GetParam()); }
   bool sec_cache_is_compressed_;
 };
 
-#ifndef ROCKSDB_LITE
-
 TEST_P(CompressedSecondaryCacheTestWithCompressionParam, BasicTestFromString) {
   std::shared_ptr<SecondaryCache> sec_cache{nullptr};
   std::string sec_cache_uri;
@@ -934,7 +838,6 @@ TEST_P(CompressedSecondaryCacheTestWithCompressionParam,
   BasicTestHelper(sec_cache, sec_cache_is_compressed_);
 }
 
-#endif  // ROCKSDB_LITE
 
 TEST_P(CompressedSecondaryCacheTestWithCompressionParam, FailsTest) {
   FailsTest(sec_cache_is_compressed_);
@@ -960,18 +863,92 @@ TEST_P(CompressedSecondaryCacheTestWithCompressionParam,
   IntegrationFullCapacityTest(sec_cache_is_compressed_);
 }
 
+TEST_P(CompressedSecondaryCacheTestWithCompressionParam, EntryRoles) {
+  CompressedSecondaryCacheOptions opts;
+  opts.capacity = 2048;
+  opts.num_shard_bits = 0;
+
+  if (sec_cache_is_compressed_) {
+    if (!LZ4_Supported()) {
+      ROCKSDB_GTEST_SKIP("This test requires LZ4 support.");
+      return;
+    }
+  } else {
+    opts.compression_type = CompressionType::kNoCompression;
+  }
+
+  // Select a random subset to include, for fast test
+  Random& r = *Random::GetTLSInstance();
+  CacheEntryRoleSet do_not_compress;
+  for (uint32_t i = 0; i < kNumCacheEntryRoles; ++i) {
+    // A few included on average, but decent chance of zero
+    if (r.OneIn(5)) {
+      do_not_compress.Add(static_cast<CacheEntryRole>(i));
+    }
+  }
+  opts.do_not_compress_roles = do_not_compress;
+
+  std::shared_ptr<SecondaryCache> sec_cache = NewCompressedSecondaryCache(opts);
+
+  // Fixed seed to ensure consistent compressibility (doesn't compress)
+  std::string junk(Random(301).RandomString(1000));
+
+  for (uint32_t i = 0; i < kNumCacheEntryRoles; ++i) {
+    CacheEntryRole role = static_cast<CacheEntryRole>(i);
+
+    // Uniquify `junk`
+    junk[0] = static_cast<char>(i);
+    TestItem item{junk.data(), junk.length()};
+    Slice ith_key = Slice(junk.data(), 16);
+
+    get_perf_context()->Reset();
+    ASSERT_OK(sec_cache->Insert(ith_key, &item, GetHelper(role)));
+    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 1U);
+
+    ASSERT_OK(sec_cache->Insert(ith_key, &item, GetHelper(role)));
+    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1U);
+
+    bool kept_in_sec_cache{true};
+    std::unique_ptr<SecondaryCacheResultHandle> handle =
+        sec_cache->Lookup(ith_key, GetHelper(role), this, true,
+                          /*advise_erase=*/true, kept_in_sec_cache);
+    ASSERT_NE(handle, nullptr);
+
+    // Lookup returns the right data
+    std::unique_ptr<TestItem> val =
+        std::unique_ptr<TestItem>(static_cast<TestItem*>(handle->Value()));
+    ASSERT_NE(val, nullptr);
+    ASSERT_EQ(memcmp(val->Buf(), item.Buf(), item.Size()), 0);
+
+    bool compressed =
+        sec_cache_is_compressed_ && !do_not_compress.Contains(role);
+    if (compressed) {
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
+                1000);
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes,
+                1007);
+    } else {
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
+    }
+  }
+}
+
 INSTANTIATE_TEST_CASE_P(CompressedSecCacheTests,
                         CompressedSecondaryCacheTestWithCompressionParam,
-                        testing::Bool());
+                        testing::Combine(testing::Bool(),
+                                         GetTestingCacheTypes()));
 
 class CompressedSecCacheTestWithCompressAndSplitParam
-    : public CompressedSecondaryCacheTest,
-      public ::testing::WithParamInterface<std::tuple<bool, bool>> {
+    : public CompressedSecondaryCacheTestBase,
+      public ::testing::WithParamInterface<
+          std::tuple<bool, bool, std::string>> {
  public:
   CompressedSecCacheTestWithCompressAndSplitParam() {
     sec_cache_is_compressed_ = std::get<0>(GetParam());
     enable_custom_split_merge_ = std::get<1>(GetParam());
   }
+  const std::string& Type() override { return std::get<2>(GetParam()); }
   bool sec_cache_is_compressed_;
   bool enable_custom_split_merge_;
 };
@@ -982,17 +959,18 @@ TEST_P(CompressedSecCacheTestWithCompressAndSplitParam, BasicIntegrationTest) {
 
 INSTANTIATE_TEST_CASE_P(CompressedSecCacheTests,
                         CompressedSecCacheTestWithCompressAndSplitParam,
-                        ::testing::Combine(testing::Bool(), testing::Bool()));
+                        ::testing::Combine(testing::Bool(), testing::Bool(),
+                                           GetTestingCacheTypes()));
 
-TEST_F(CompressedSecondaryCacheTest, SplitValueIntoChunksTest) {
+TEST_P(CompressedSecondaryCacheTest, SplitValueIntoChunksTest) {
   SplitValueIntoChunksTest();
 }
 
-TEST_F(CompressedSecondaryCacheTest, MergeChunksIntoValueTest) {
+TEST_P(CompressedSecondaryCacheTest, MergeChunksIntoValueTest) {
   MergeChunksIntoValueTest();
 }
 
-TEST_F(CompressedSecondaryCacheTest, SplictValueAndMergeChunksTest) {
+TEST_P(CompressedSecondaryCacheTest, SplictValueAndMergeChunksTest) {
   SplictValueAndMergeChunksTest();
 }
 

cache/lru_cache.cc

@@ -14,28 +14,29 @@
 #include <cstdio>
 #include <cstdlib>
 
+#include "cache/secondary_cache_adapter.h"
 #include "monitoring/perf_context_imp.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "port/lang.h"
 #include "util/distributed_mutex.h"
 
 namespace ROCKSDB_NAMESPACE {
 namespace lru_cache {
 
-// A distinct pointer value for marking "dummy" cache entries
-void* const kDummyValueMarker = const_cast<char*>("kDummyValueMarker");
-
-LRUHandleTable::LRUHandleTable(int max_upper_hash_bits)
+LRUHandleTable::LRUHandleTable(int max_upper_hash_bits,
+                               MemoryAllocator* allocator)
     : length_bits_(/* historical starting size*/ 4),
       list_(new LRUHandle* [size_t{1} << length_bits_] {}),
       elems_(0),
-      max_length_bits_(max_upper_hash_bits) {}
+      max_length_bits_(max_upper_hash_bits),
+      allocator_(allocator) {}
 
 LRUHandleTable::~LRUHandleTable() {
+  auto alloc = allocator_;
   ApplyToEntriesRange(
-      [](LRUHandle* h) {
+      [alloc](LRUHandle* h) {
         if (!h->HasRefs()) {
-          h->Free();
+          h->Free(alloc);
         }
       },
       0, size_t{1} << length_bits_);
@@ -95,7 +96,7 @@ void LRUHandleTable::Resize() {
   std::unique_ptr<LRUHandle* []> new_list {
     new LRUHandle* [size_t{1} << new_length_bits] {}
   };
-  uint32_t count = 0;
+  [[maybe_unused]] uint32_t count = 0;
   for (uint32_t i = 0; i < old_length; i++) {
     LRUHandle* h = list_[i];
     while (h != nullptr) {
@@ -118,7 +119,8 @@ LRUCacheShard::LRUCacheShard(size_t capacity, bool strict_capacity_limit,
                              double low_pri_pool_ratio, bool use_adaptive_mutex,
                              CacheMetadataChargePolicy metadata_charge_policy,
                              int max_upper_hash_bits,
-                             SecondaryCache* secondary_cache)
+                             MemoryAllocator* allocator,
+                             const Cache::EvictionCallback* eviction_callback)
     : CacheShardBase(metadata_charge_policy),
       capacity_(0),
       high_pri_pool_usage_(0),
@@ -128,11 +130,11 @@ LRUCacheShard::LRUCacheShard(size_t capacity, bool strict_capacity_limit,
       high_pri_pool_capacity_(0),
       low_pri_pool_ratio_(low_pri_pool_ratio),
       low_pri_pool_capacity_(0),
-      table_(max_upper_hash_bits),
+      table_(max_upper_hash_bits, allocator),
       usage_(0),
       lru_usage_(0),
       mutex_(use_adaptive_mutex),
-      secondary_cache_(secondary_cache) {
+      eviction_callback_(*eviction_callback) {
   // Make empty circular linked list.
   lru_.next = &lru_;
   lru_.prev = &lru_;
@@ -159,13 +161,14 @@ void LRUCacheShard::EraseUnRefEntries() {
   }
 
   for (auto entry : last_reference_list) {
-    entry->Free();
+    entry->Free(table_.GetAllocator());
   }
 }
 
 void LRUCacheShard::ApplyToSomeEntries(
-    const std::function<void(const Slice& key, void* value, size_t charge,
-                             DeleterFn deleter)>& callback,
+    const std::function<void(const Slice& key, Cache::ObjectPtr value,
+                             size_t charge,
+                             const Cache::CacheItemHelper* helper)>& callback,
     size_t average_entries_per_lock, size_t* state) {
   // The state is essentially going to be the starting hash, which works
   // nicely even if we resize between calls because we use upper-most
@@ -192,11 +195,8 @@ void LRUCacheShard::ApplyToSomeEntries(
   table_.ApplyToEntriesRange(
       [callback,
        metadata_charge_policy = metadata_charge_policy_](LRUHandle* h) {
-        DeleterFn deleter = h->IsSecondaryCacheCompatible()
-                                ? h->info_.helper->del_cb
-                                : h->info_.deleter;
         callback(h->key(), h->value, h->GetCharge(metadata_charge_policy),
-                 deleter);
+                 h->helper);
       },
       index_begin, index_end);
 }
@@ -334,16 +334,19 @@ void LRUCacheShard::EvictFromLRU(size_t charge,
   }
 }
 
-void LRUCacheShard::TryInsertIntoSecondaryCache(
-    autovector<LRUHandle*> evicted_handles) {
-  for (auto entry : evicted_handles) {
-    if (secondary_cache_ && entry->IsSecondaryCacheCompatible() &&
-        !entry->IsInSecondaryCache()) {
-      secondary_cache_->Insert(entry->key(), entry->value, entry->info_.helper)
-          .PermitUncheckedError();
+void LRUCacheShard::NotifyEvicted(
+    const autovector<LRUHandle*>& evicted_handles) {
+  MemoryAllocator* alloc = table_.GetAllocator();
+  for (LRUHandle* entry : evicted_handles) {
+    if (eviction_callback_ &&
+        eviction_callback_(entry->key(),
+                           reinterpret_cast<Cache::Handle*>(entry))) {
+      // Callback took ownership of obj; just free handle
+      free(entry);
+    } else {
+      // Free the entries here outside of mutex for performance reasons.
+      entry->Free(alloc);
     }
-    // Free the entries here outside of mutex for performance reasons.
-    entry->Free();
   }
 }
 
@@ -357,7 +360,7 @@ void LRUCacheShard::SetCapacity(size_t capacity) {
     EvictFromLRU(0, &last_reference_list);
   }
 
-  TryInsertIntoSecondaryCache(last_reference_list);
+  NotifyEvicted(last_reference_list);
 }
 
 void LRUCacheShard::SetStrictCapacityLimit(bool strict_capacity_limit) {
@@ -365,8 +368,7 @@ void LRUCacheShard::SetStrictCapacityLimit(bool strict_capacity_limit) {
   strict_capacity_limit_ = strict_capacity_limit;
 }
 
-Status LRUCacheShard::InsertItem(LRUHandle* e, LRUHandle** handle,
-                                 bool free_handle_on_fail) {
+Status LRUCacheShard::InsertItem(LRUHandle* e, LRUHandle** handle) {
   Status s = Status::OK();
   autovector<LRUHandle*> last_reference_list;
 
@@ -385,10 +387,9 @@ Status LRUCacheShard::InsertItem(LRUHandle* e, LRUHandle** handle,
         // into cache and get evicted immediately.
         last_reference_list.push_back(e);
       } else {
-        if (free_handle_on_fail) {
-          free(e);
-          *handle = nullptr;
-        }
+        free(e);
+        e = nullptr;
+        *handle = nullptr;
         s = Status::MemoryLimit("Insert failed due to LRU cache being full.");
       }
     } else {
@@ -420,192 +421,27 @@ Status LRUCacheShard::InsertItem(LRUHandle* e, LRUHandle** handle,
     }
   }
 
-  TryInsertIntoSecondaryCache(last_reference_list);
+  NotifyEvicted(last_reference_list);
 
   return s;
 }
 
-void LRUCacheShard::Promote(LRUHandle* e) {
-  SecondaryCacheResultHandle* secondary_handle = e->sec_handle;
-
-  assert(secondary_handle->IsReady());
-  // e is not thread-shared here; OK to modify "immutable" fields as well as
-  // "mutable" (normally requiring mutex)
-  e->SetIsPending(false);
-  e->value = secondary_handle->Value();
-  assert(e->total_charge == 0);
-  size_t value_size = secondary_handle->Size();
-  delete secondary_handle;
-
-  if (e->value) {
-    e->CalcTotalCharge(value_size, metadata_charge_policy_);
-    Status s;
-    if (e->IsStandalone()) {
-      assert(secondary_cache_ && secondary_cache_->SupportForceErase());
-
-      // Insert a dummy handle and return a standalone handle to caller.
-      // Charge the standalone handle.
-      autovector<LRUHandle*> last_reference_list;
-      bool free_standalone_handle{false};
-      {
-        DMutexLock l(mutex_);
-
-        // Free the space following strict LRU policy until enough space
-        // is freed or the lru list is empty.
-        EvictFromLRU(e->total_charge, &last_reference_list);
-
-        if ((usage_ + e->total_charge) > capacity_ && strict_capacity_limit_) {
-          free_standalone_handle = true;
-        } else {
-          usage_ += e->total_charge;
-        }
-      }
-
-      TryInsertIntoSecondaryCache(last_reference_list);
-      if (free_standalone_handle) {
-        e->Unref();
-        e->Free();
-        e = nullptr;
-      } else {
-        PERF_COUNTER_ADD(block_cache_standalone_handle_count, 1);
-      }
-
-      // Insert a dummy handle into the primary cache. This dummy handle is
-      // not IsSecondaryCacheCompatible().
-      // FIXME? This should not overwrite an existing non-dummy entry in the
-      // rare case that one exists
-      Cache::Priority priority =
-          e->IsHighPri() ? Cache::Priority::HIGH : Cache::Priority::LOW;
-      s = Insert(e->key(), e->hash, kDummyValueMarker, /*charge=*/0,
-                 /*deleter=*/nullptr, /*helper=*/nullptr, /*handle=*/nullptr,
-                 priority);
-    } else {
-      e->SetInCache(true);
-      LRUHandle* handle = e;
-      // This InsertItem() could fail if the cache is over capacity and
-      // strict_capacity_limit_ is true. In such a case, we don't want
-      // InsertItem() to free the handle, since the item is already in memory
-      // and the caller will most likely just read it from disk if we erase it
-      // here.
-      s = InsertItem(e, &handle, /*free_handle_on_fail=*/false);
-      if (s.ok()) {
-        PERF_COUNTER_ADD(block_cache_real_handle_count, 1);
-      }
-    }
-
-    if (!s.ok()) {
-      // Item is in memory, but not accounted against the cache capacity.
-      // When the handle is released, the item should get deleted.
-      assert(!e->InCache());
-    }
-  } else {
-    // Secondary cache lookup failed. The caller will take care of detecting
-    // this and eventually releasing e.
-    assert(!e->value);
-    assert(!e->InCache());
-  }
-}
-
 LRUHandle* LRUCacheShard::Lookup(const Slice& key, uint32_t hash,
-                                 const Cache::CacheItemHelper* helper,
-                                 const Cache::CreateCallback& create_cb,
-                                 Cache::Priority priority, bool wait,
-                                 Statistics* stats) {
-  LRUHandle* e = nullptr;
-  bool found_dummy_entry{false};
-  {
-    DMutexLock l(mutex_);
-    e = table_.Lookup(key, hash);
-    if (e != nullptr) {
-      assert(e->InCache());
-      if (e->value == kDummyValueMarker) {
-        // For a dummy handle, if it was retrieved from secondary cache,
-        // it may still exist in secondary cache.
-        // If the handle exists in secondary cache, the value should be
-        // erased from sec cache and be inserted into primary cache.
-        found_dummy_entry = true;
-        // Let the dummy entry be overwritten
-        e = nullptr;
-      } else {
-        if (!e->HasRefs()) {
-          // The entry is in LRU since it's in hash and has no external
-          // references.
-          LRU_Remove(e);
-        }
-        e->Ref();
-        e->SetHit();
-      }
-    }
-  }
-
-  // If handle table lookup failed or the handle is a dummy one, allocate
-  // a handle outside the mutex if we re going to lookup in the secondary cache.
-  //
-  // When a block is firstly Lookup from CompressedSecondaryCache, we just
-  // insert a dummy block into the primary cache (charging the actual size of
-  // the block) and don't erase the block from CompressedSecondaryCache. A
-  // standalone handle is returned to the caller. Only if the block is hit
-  // again, we erase it from CompressedSecondaryCache and add it into the
-  // primary cache.
-  if (!e && secondary_cache_ && helper && helper->saveto_cb) {
-    // For objects from the secondary cache, we expect the caller to provide
-    // a way to create/delete the primary cache object. The only case where
-    // a deleter would not be required is for dummy entries inserted for
-    // accounting purposes, which we won't demote to the secondary cache
-    // anyway.
-    assert(create_cb && helper->del_cb);
-    bool is_in_sec_cache{false};
-    std::unique_ptr<SecondaryCacheResultHandle> secondary_handle =
-        secondary_cache_->Lookup(key, create_cb, wait, found_dummy_entry,
-                                 is_in_sec_cache);
-    if (secondary_handle != nullptr) {
-      e = static_cast<LRUHandle*>(malloc(sizeof(LRUHandle) - 1 + key.size()));
-
-      e->m_flags = 0;
-      e->im_flags = 0;
-      e->SetSecondaryCacheCompatible(true);
-      e->info_.helper = helper;
-      e->key_length = key.size();
-      e->hash = hash;
-      e->refs = 0;
-      e->next = e->prev = nullptr;
-      e->SetPriority(priority);
-      memcpy(e->key_data, key.data(), key.size());
-      e->value = nullptr;
-      e->sec_handle = secondary_handle.release();
-      e->total_charge = 0;
-      e->Ref();
-      e->SetIsInSecondaryCache(is_in_sec_cache);
-      e->SetIsStandalone(secondary_cache_->SupportForceErase() &&
-                         !found_dummy_entry);
-
-      if (wait) {
-        Promote(e);
-        if (e) {
-          if (!e->value) {
-            // The secondary cache returned a handle, but the lookup failed.
-            e->Unref();
-            e->Free();
-            e = nullptr;
-          } else {
-            PERF_COUNTER_ADD(secondary_cache_hit_count, 1);
-            RecordTick(stats, SECONDARY_CACHE_HITS);
-          }
-        }
-      } else {
-        // If wait is false, we always return a handle and let the caller
-        // release the handle after checking for success or failure.
-        e->SetIsPending(true);
-        // This may be slightly inaccurate, if the lookup eventually fails.
-        // But the probability is very low.
-        PERF_COUNTER_ADD(secondary_cache_hit_count, 1);
-        RecordTick(stats, SECONDARY_CACHE_HITS);
-      }
-    } else {
-      // Caller will most likely overwrite the dummy entry with an Insert
-      // after this Lookup fails
-      assert(e == nullptr);
+                                 const Cache::CacheItemHelper* /*helper*/,
+                                 Cache::CreateContext* /*create_context*/,
+                                 Cache::Priority /*priority*/,
+                                 Statistics* /*stats*/) {
+  DMutexLock l(mutex_);
+  LRUHandle* e = table_.Lookup(key, hash);
+  if (e != nullptr) {
+    assert(e->InCache());
+    if (!e->HasRefs()) {
+      // The entry is in LRU since it's in hash and has no external
+      // references.
+      LRU_Remove(e);
     }
+    e->Ref();
+    e->SetHit();
   }
   return e;
 }
@@ -614,8 +450,6 @@ bool LRUCacheShard::Ref(LRUHandle* e) {
   DMutexLock l(mutex_);
   // To create another reference - entry must be already externally referenced.
   assert(e->HasRefs());
-  // Pending handles are not for sharing
-  assert(!e->IsPending());
   e->Ref();
   return true;
 }
@@ -639,14 +473,13 @@ bool LRUCacheShard::Release(LRUHandle* e, bool /*useful*/,
   if (e == nullptr) {
     return false;
   }
-  bool last_reference = false;
-  // Must Wait or WaitAll first on pending handles. Otherwise, would leak
-  // a secondary cache handle.
-  assert(!e->IsPending());
+  bool must_free;
+  bool was_in_cache;
   {
     DMutexLock l(mutex_);
-    last_reference = e->Unref();
-    if (last_reference && e->InCache()) {
+    must_free = e->Unref();
+    was_in_cache = e->InCache();
+    if (must_free && was_in_cache) {
       // The item is still in cache, and nobody else holds a reference to it.
       if (usage_ > capacity_ || erase_if_last_ref) {
         // The LRU list must be empty since the cache is full.
@@ -657,28 +490,39 @@ bool LRUCacheShard::Release(LRUHandle* e, bool /*useful*/,
       } else {
         // Put the item back on the LRU list, and don't free it.
         LRU_Insert(e);
-        last_reference = false;
+        must_free = false;
       }
     }
-    // If it was the last reference, then decrement the cache usage.
-    if (last_reference) {
+    // If about to be freed, then decrement the cache usage.
+    if (must_free) {
       assert(usage_ >= e->total_charge);
       usage_ -= e->total_charge;
     }
   }
 
   // Free the entry here outside of mutex for performance reasons.
-  if (last_reference) {
-    e->Free();
+  if (must_free) {
+    // Only call eviction callback if we're sure no one requested erasure
+    // FIXME: disabled because of test churn
+    if (false && was_in_cache && !erase_if_last_ref && eviction_callback_ &&
+        eviction_callback_(e->key(), reinterpret_cast<Cache::Handle*>(e))) {
+      // Callback took ownership of obj; just free handle
+      free(e);
+    } else {
+      e->Free(table_.GetAllocator());
+    }
   }
-  return last_reference;
+  return must_free;
 }
 
-Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
-                             size_t charge,
-                             void (*deleter)(const Slice& key, void* value),
-                             const Cache::CacheItemHelper* helper,
-                             LRUHandle** handle, Cache::Priority priority) {
+LRUHandle* LRUCacheShard::CreateHandle(const Slice& key, uint32_t hash,
+                                       Cache::ObjectPtr value,
+                                       const Cache::CacheItemHelper* helper,
+                                       size_t charge) {
+  assert(helper);
+  // value == nullptr is reserved for indicating failure in SecondaryCache
+  assert(!(helper->IsSecondaryCacheCompatible() && value == nullptr));
+
   // Allocate the memory here outside of the mutex.
   // If the cache is full, we'll have to release it.
   // It shouldn't happen very often though.
@@ -688,27 +532,58 @@ Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
   e->value = value;
   e->m_flags = 0;
   e->im_flags = 0;
-  if (helper) {
-    // Use only one of the two parameters
-    assert(deleter == nullptr);
-    // value == nullptr is reserved for indicating failure for when secondary
-    // cache compatible
-    assert(value != nullptr);
-    e->SetSecondaryCacheCompatible(true);
-    e->info_.helper = helper;
-  } else {
-    e->info_.deleter = deleter;
-  }
+  e->helper = helper;
   e->key_length = key.size();
   e->hash = hash;
   e->refs = 0;
   e->next = e->prev = nullptr;
-  e->SetInCache(true);
-  e->SetPriority(priority);
   memcpy(e->key_data, key.data(), key.size());
   e->CalcTotalCharge(charge, metadata_charge_policy_);
 
-  return InsertItem(e, handle, /* free_handle_on_fail */ true);
+  return e;
+}
+
+Status LRUCacheShard::Insert(const Slice& key, uint32_t hash,
+                             Cache::ObjectPtr value,
+                             const Cache::CacheItemHelper* helper,
+                             size_t charge, LRUHandle** handle,
+                             Cache::Priority priority) {
+  LRUHandle* e = CreateHandle(key, hash, value, helper, charge);
+  e->SetPriority(priority);
+  e->SetInCache(true);
+  return InsertItem(e, handle);
+}
+
+LRUHandle* LRUCacheShard::CreateStandalone(const Slice& key, uint32_t hash,
+                                           Cache::ObjectPtr value,
+                                           const Cache::CacheItemHelper* helper,
+                                           size_t charge,
+                                           bool allow_uncharged) {
+  LRUHandle* e = CreateHandle(key, hash, value, helper, charge);
+  e->SetIsStandalone(true);
+  e->Ref();
+
+  autovector<LRUHandle*> last_reference_list;
+
+  {
+    DMutexLock l(mutex_);
+
+    EvictFromLRU(e->total_charge, &last_reference_list);
+
+    if (strict_capacity_limit_ && (usage_ + e->total_charge) > capacity_) {
+      if (allow_uncharged) {
+        e->total_charge = 0;
+      } else {
+        free(e);
+        e = nullptr;
+      }
+    } else {
+      usage_ += e->total_charge;
+    }
+  }
+
+  NotifyEvicted(last_reference_list);
+  return e;
 }
 
 void LRUCacheShard::Erase(const Slice& key, uint32_t hash) {
@@ -733,20 +608,10 @@ void LRUCacheShard::Erase(const Slice& key, uint32_t hash) {
   // Free the entry here outside of mutex for performance reasons.
   // last_reference will only be true if e != nullptr.
   if (last_reference) {
-    e->Free();
+    e->Free(table_.GetAllocator());
   }
 }
 
-bool LRUCacheShard::IsReady(LRUHandle* e) {
-  bool ready = true;
-  if (e->IsPending()) {
-    assert(secondary_cache_);
-    assert(e->sec_handle);
-    ready = e->sec_handle->IsReady();
-  }
-  return ready;
-}
-
 size_t LRUCacheShard::GetUsage() const {
   DMutexLock l(mutex_);
   return usage_;
@@ -781,30 +646,20 @@ void LRUCacheShard::AppendPrintableOptions(std::string& str) const {
   str.append(buffer);
 }
 
-LRUCache::LRUCache(size_t capacity, int num_shard_bits,
-                   bool strict_capacity_limit, double high_pri_pool_ratio,
-                   double low_pri_pool_ratio,
-                   std::shared_ptr<MemoryAllocator> allocator,
-                   bool use_adaptive_mutex,
-                   CacheMetadataChargePolicy metadata_charge_policy,
-                   std::shared_ptr<SecondaryCache> _secondary_cache)
-    : ShardedCache(capacity, num_shard_bits, strict_capacity_limit,
-                   std::move(allocator)),
-      secondary_cache_(std::move(_secondary_cache)) {
+LRUCache::LRUCache(const LRUCacheOptions& opts) : ShardedCache(opts) {
   size_t per_shard = GetPerShardCapacity();
-  SecondaryCache* secondary_cache = secondary_cache_.get();
-  InitShards([=](LRUCacheShard* cs) {
-    new (cs) LRUCacheShard(
-        per_shard, strict_capacity_limit, high_pri_pool_ratio,
-        low_pri_pool_ratio, use_adaptive_mutex, metadata_charge_policy,
-        /* max_upper_hash_bits */ 32 - num_shard_bits, secondary_cache);
+  MemoryAllocator* alloc = memory_allocator();
+  InitShards([&](LRUCacheShard* cs) {
+    new (cs) LRUCacheShard(per_shard, opts.strict_capacity_limit,
+                           opts.high_pri_pool_ratio, opts.low_pri_pool_ratio,
+                           opts.use_adaptive_mutex, opts.metadata_charge_policy,
+                           /* max_upper_hash_bits */ 32 - opts.num_shard_bits,
+                           alloc, &eviction_callback_);
   });
 }
 
-void* LRUCache::Value(Handle* handle) {
+Cache::ObjectPtr LRUCache::Value(Handle* handle) {
   auto h = reinterpret_cast<const LRUHandle*>(handle);
-  assert(!h->IsPending() || h->value == nullptr);
-  assert(h->value != kDummyValueMarker);
   return h->value;
 }
 
@@ -813,13 +668,10 @@ size_t LRUCache::GetCharge(Handle* handle) const {
       GetShard(0).metadata_charge_policy_);
 }
 
-Cache::DeleterFn LRUCache::GetDeleter(Handle* handle) const {
+const Cache::CacheItemHelper* LRUCache::GetCacheItemHelper(
+    Handle* handle) const {
   auto h = reinterpret_cast<const LRUHandle*>(handle);
-  if (h->IsSecondaryCacheCompatible()) {
-    return h->info_.helper->del_cb;
-  } else {
-    return h->info_.deleter;
-  }
+  return h->helper;
 }
 
 size_t LRUCache::TEST_GetLRUSize() {
@@ -830,51 +682,9 @@ double LRUCache::GetHighPriPoolRatio() {
   return GetShard(0).GetHighPriPoolRatio();
 }
 
-void LRUCache::WaitAll(std::vector<Handle*>& handles) {
-  if (secondary_cache_) {
-    std::vector<SecondaryCacheResultHandle*> sec_handles;
-    sec_handles.reserve(handles.size());
-    for (Handle* handle : handles) {
-      if (!handle) {
-        continue;
-      }
-      LRUHandle* lru_handle = reinterpret_cast<LRUHandle*>(handle);
-      if (!lru_handle->IsPending()) {
-        continue;
-      }
-      sec_handles.emplace_back(lru_handle->sec_handle);
-    }
-    secondary_cache_->WaitAll(sec_handles);
-    for (Handle* handle : handles) {
-      if (!handle) {
-        continue;
-      }
-      LRUHandle* lru_handle = reinterpret_cast<LRUHandle*>(handle);
-      if (!lru_handle->IsPending()) {
-        continue;
-      }
-      GetShard(lru_handle->hash).Promote(lru_handle);
-    }
-  }
-}
-
-void LRUCache::AppendPrintableOptions(std::string& str) const {
-  ShardedCache::AppendPrintableOptions(str);  // options from shard
-  if (secondary_cache_) {
-    str.append("  secondary_cache:\n");
-    str.append(secondary_cache_->GetPrintableOptions());
-  }
-}
-
 }  // namespace lru_cache
 
-std::shared_ptr<Cache> NewLRUCache(
-    size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-    double high_pri_pool_ratio,
-    std::shared_ptr<MemoryAllocator> memory_allocator, bool use_adaptive_mutex,
-    CacheMetadataChargePolicy metadata_charge_policy,
-    const std::shared_ptr<SecondaryCache>& secondary_cache,
-    double low_pri_pool_ratio) {
+std::shared_ptr<Cache> LRUCacheOptions::MakeSharedCache() const {
   if (num_shard_bits >= 20) {
     return nullptr;  // The cache cannot be sharded into too many fine pieces.
   }
@@ -890,32 +700,24 @@ std::shared_ptr<Cache> NewLRUCache(
     // Invalid high_pri_pool_ratio and low_pri_pool_ratio combination
     return nullptr;
   }
-  if (num_shard_bits < 0) {
-    num_shard_bits = GetDefaultCacheShardBits(capacity);
+  // For sanitized options
+  LRUCacheOptions opts = *this;
+  if (opts.num_shard_bits < 0) {
+    opts.num_shard_bits = GetDefaultCacheShardBits(capacity);
   }
-  return std::make_shared<LRUCache>(
-      capacity, num_shard_bits, strict_capacity_limit, high_pri_pool_ratio,
-      low_pri_pool_ratio, std::move(memory_allocator), use_adaptive_mutex,
-      metadata_charge_policy, secondary_cache);
+  std::shared_ptr<Cache> cache = std::make_shared<LRUCache>(opts);
+  if (secondary_cache) {
+    cache = std::make_shared<CacheWithSecondaryAdapter>(cache, secondary_cache);
+  }
+  return cache;
 }
 
-std::shared_ptr<Cache> NewLRUCache(const LRUCacheOptions& cache_opts) {
-  return NewLRUCache(cache_opts.capacity, cache_opts.num_shard_bits,
-                     cache_opts.strict_capacity_limit,
-                     cache_opts.high_pri_pool_ratio,
-                     cache_opts.memory_allocator, cache_opts.use_adaptive_mutex,
-                     cache_opts.metadata_charge_policy,
-                     cache_opts.secondary_cache, cache_opts.low_pri_pool_ratio);
-}
-
-std::shared_ptr<Cache> NewLRUCache(
-    size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-    double high_pri_pool_ratio,
-    std::shared_ptr<MemoryAllocator> memory_allocator, bool use_adaptive_mutex,
-    CacheMetadataChargePolicy metadata_charge_policy,
-    double low_pri_pool_ratio) {
-  return NewLRUCache(capacity, num_shard_bits, strict_capacity_limit,
-                     high_pri_pool_ratio, memory_allocator, use_adaptive_mutex,
-                     metadata_charge_policy, nullptr, low_pri_pool_ratio);
+std::shared_ptr<GeneralCache> LRUCacheOptions::MakeSharedGeneralCache() const {
+  if (secondary_cache) {
+    // Not allowed for a GeneralCache
+    return nullptr;
+  }
+  // Works while GeneralCache is an alias for Cache
+  return MakeSharedCache();
 }
 }  // namespace ROCKSDB_NAMESPACE

cache/lru_cache.h

@@ -13,9 +13,9 @@
 
 #include "cache/sharded_cache.h"
 #include "port/lang.h"
+#include "port/likely.h"
 #include "port/malloc.h"
 #include "port/port.h"
-#include "rocksdb/secondary_cache.h"
 #include "util/autovector.h"
 #include "util/distributed_mutex.h"
 
@@ -48,19 +48,9 @@ namespace lru_cache {
 // While refs > 0, public properties like value and deleter must not change.
 
 struct LRUHandle {
-  void* value;
-  union Info {
-    Info() {}
-    ~Info() {}
-    Cache::DeleterFn deleter;
-    const Cache::CacheItemHelper* helper;
-  } info_;
-  // An entry is not added to the LRUHandleTable until the secondary cache
-  // lookup is complete, so its safe to have this union.
-  union {
-    LRUHandle* next_hash;
-    SecondaryCacheResultHandle* sec_handle;
-  };
+  Cache::ObjectPtr value;
+  const Cache::CacheItemHelper* helper;
+  LRUHandle* next_hash;
   LRUHandle* next;
   LRUHandle* prev;
   size_t total_charge;  // TODO(opt): Only allow uint32_t?
@@ -93,14 +83,8 @@ struct LRUHandle {
     IM_IS_HIGH_PRI = (1 << 0),
     // Whether this entry is low priority entry.
     IM_IS_LOW_PRI = (1 << 1),
-    // Can this be inserted into the secondary cache.
-    IM_IS_SECONDARY_CACHE_COMPATIBLE = (1 << 2),
-    // Is the handle still being read from a lower tier.
-    IM_IS_PENDING = (1 << 3),
-    // Whether this handle is still in a lower tier
-    IM_IS_IN_SECONDARY_CACHE = (1 << 4),
     // Marks result handles that should not be inserted into cache
-    IM_IS_STANDALONE = (1 << 5),
+    IM_IS_STANDALONE = (1 << 2),
   };
 
   // Beginning of the key (MUST BE THE LAST FIELD IN THIS STRUCT!)
@@ -130,13 +114,6 @@ struct LRUHandle {
   bool IsLowPri() const { return im_flags & IM_IS_LOW_PRI; }
   bool InLowPriPool() const { return m_flags & M_IN_LOW_PRI_POOL; }
   bool HasHit() const { return m_flags & M_HAS_HIT; }
-  bool IsSecondaryCacheCompatible() const {
-    return im_flags & IM_IS_SECONDARY_CACHE_COMPATIBLE;
-  }
-  bool IsPending() const { return im_flags & IM_IS_PENDING; }
-  bool IsInSecondaryCache() const {
-    return im_flags & IM_IS_IN_SECONDARY_CACHE;
-  }
   bool IsStandalone() const { return im_flags & IM_IS_STANDALONE; }
 
   void SetInCache(bool in_cache) {
@@ -178,30 +155,6 @@ struct LRUHandle {
 
   void SetHit() { m_flags |= M_HAS_HIT; }
 
-  void SetSecondaryCacheCompatible(bool compat) {
-    if (compat) {
-      im_flags |= IM_IS_SECONDARY_CACHE_COMPATIBLE;
-    } else {
-      im_flags &= ~IM_IS_SECONDARY_CACHE_COMPATIBLE;
-    }
-  }
-
-  void SetIsPending(bool pending) {
-    if (pending) {
-      im_flags |= IM_IS_PENDING;
-    } else {
-      im_flags &= ~IM_IS_PENDING;
-    }
-  }
-
-  void SetIsInSecondaryCache(bool is_in_secondary_cache) {
-    if (is_in_secondary_cache) {
-      im_flags |= IM_IS_IN_SECONDARY_CACHE;
-    } else {
-      im_flags &= ~IM_IS_IN_SECONDARY_CACHE;
-    }
-  }
-
   void SetIsStandalone(bool is_standalone) {
     if (is_standalone) {
       im_flags |= IM_IS_STANDALONE;
@@ -210,22 +163,11 @@ struct LRUHandle {
     }
   }
 
-  void Free() {
+  void Free(MemoryAllocator* allocator) {
     assert(refs == 0);
-
-    if (!IsSecondaryCacheCompatible() && info_.deleter) {
-      (*info_.deleter)(key(), value);
-    } else if (IsSecondaryCacheCompatible()) {
-      if (IsPending()) {
-        assert(sec_handle != nullptr);
-        SecondaryCacheResultHandle* tmp_sec_handle = sec_handle;
-        tmp_sec_handle->Wait();
-        value = tmp_sec_handle->Value();
-        delete tmp_sec_handle;
-      }
-      if (value) {
-        (*info_.helper->del_cb)(key(), value);
-      }
+    assert(helper);
+    if (helper->del_cb) {
+      helper->del_cb(value, allocator);
     }
 
     free(this);
@@ -267,7 +209,7 @@ struct LRUHandle {
 // 4.4.3's builtin hashtable.
 class LRUHandleTable {
  public:
-  explicit LRUHandleTable(int max_upper_hash_bits);
+  explicit LRUHandleTable(int max_upper_hash_bits, MemoryAllocator* allocator);
   ~LRUHandleTable();
 
   LRUHandle* Lookup(const Slice& key, uint32_t hash);
@@ -291,6 +233,8 @@ class LRUHandleTable {
 
   size_t GetOccupancyCount() const { return elems_; }
 
+  MemoryAllocator* GetAllocator() const { return allocator_; }
+
  private:
   // Return a pointer to slot that points to a cache entry that
   // matches key/hash.  If there is no such cache entry, return a
@@ -312,16 +256,22 @@ class LRUHandleTable {
 
   // Set from max_upper_hash_bits (see constructor).
   const int max_length_bits_;
+
+  // From Cache, needed for delete
+  MemoryAllocator* const allocator_;
 };
 
 // A single shard of sharded cache.
 class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
  public:
+  // NOTE: the eviction_callback ptr is saved, as is it assumed to be kept
+  // alive in Cache.
   LRUCacheShard(size_t capacity, bool strict_capacity_limit,
                 double high_pri_pool_ratio, double low_pri_pool_ratio,
                 bool use_adaptive_mutex,
                 CacheMetadataChargePolicy metadata_charge_policy,
-                int max_upper_hash_bits, SecondaryCache* secondary_cache);
+                int max_upper_hash_bits, MemoryAllocator* allocator,
+                const Cache::EvictionCallback* eviction_callback);
 
  public:  // Type definitions expected as parameter to ShardedCache
   using HandleImpl = LRUHandle;
@@ -329,8 +279,8 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   using HashCref = uint32_t;
 
  public:  // Function definitions expected as parameter to ShardedCache
-  static inline HashVal ComputeHash(const Slice& key) {
-    return Lower32of64(GetSliceNPHash64(key));
+  static inline HashVal ComputeHash(const Slice& key, uint32_t seed) {
+    return Lower32of64(GetSliceNPHash64(key, seed));
   }
 
   // Separate from constructor so caller can easily make an array of LRUCache
@@ -348,29 +298,21 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   void SetLowPriorityPoolRatio(double low_pri_pool_ratio);
 
   // Like Cache methods, but with an extra "hash" parameter.
-  inline Status Insert(const Slice& key, uint32_t hash, void* value,
-                       size_t charge, Cache::DeleterFn deleter,
-                       LRUHandle** handle, Cache::Priority priority) {
-    return Insert(key, hash, value, charge, deleter, nullptr, handle, priority);
-  }
-  inline Status Insert(const Slice& key, uint32_t hash, void* value,
-                       const Cache::CacheItemHelper* helper, size_t charge,
-                       LRUHandle** handle, Cache::Priority priority) {
-    assert(helper);
-    return Insert(key, hash, value, charge, nullptr, helper, handle, priority);
-  }
-  // If helper_cb is null, the values of the following arguments don't matter.
+  Status Insert(const Slice& key, uint32_t hash, Cache::ObjectPtr value,
+                const Cache::CacheItemHelper* helper, size_t charge,
+                LRUHandle** handle, Cache::Priority priority);
+
+  LRUHandle* CreateStandalone(const Slice& key, uint32_t hash,
+                              Cache::ObjectPtr obj,
+                              const Cache::CacheItemHelper* helper,
+                              size_t charge, bool allow_uncharged);
+
   LRUHandle* Lookup(const Slice& key, uint32_t hash,
                     const Cache::CacheItemHelper* helper,
-                    const Cache::CreateCallback& create_cb,
-                    Cache::Priority priority, bool wait, Statistics* stats);
-  inline LRUHandle* Lookup(const Slice& key, uint32_t hash) {
-    return Lookup(key, hash, nullptr, nullptr, Cache::Priority::LOW, true,
-                  nullptr);
-  }
+                    Cache::CreateContext* create_context,
+                    Cache::Priority priority, Statistics* stats);
+
   bool Release(LRUHandle* handle, bool useful, bool erase_if_last_ref);
-  bool IsReady(LRUHandle* /*handle*/);
-  void Wait(LRUHandle* /*handle*/) {}
   bool Ref(LRUHandle* handle);
   void Erase(const Slice& key, uint32_t hash);
 
@@ -384,8 +326,9 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   size_t GetTableAddressCount() const;
 
   void ApplyToSomeEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
+      const std::function<void(const Slice& key, Cache::ObjectPtr value,
+                               size_t charge,
+                               const Cache::CacheItemHelper* helper)>& callback,
       size_t average_entries_per_lock, size_t* state);
 
   void EraseUnRefEntries();
@@ -409,23 +352,10 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
  private:
   friend class LRUCache;
   // Insert an item into the hash table and, if handle is null, insert into
-  // the LRU list. Older items are evicted as necessary. If the cache is full
-  // and free_handle_on_fail is true, the item is deleted and handle is set to
-  // nullptr.
-  Status InsertItem(LRUHandle* item, LRUHandle** handle,
-                    bool free_handle_on_fail);
-  Status Insert(const Slice& key, uint32_t hash, void* value, size_t charge,
-                DeleterFn deleter, const Cache::CacheItemHelper* helper,
-                LRUHandle** handle, Cache::Priority priority);
-  // Promote an item looked up from the secondary cache to the LRU cache.
-  // The item may be still in the secondary cache.
-  // It is only inserted into the hash table and not the LRU list, and only
-  // if the cache is not at full capacity, as is the case during Insert.  The
-  // caller should hold a reference on the LRUHandle. When the caller releases
-  // the last reference, the item is added to the LRU list.
-  // The item is promoted to the high pri or low pri pool as specified by the
-  // caller in Lookup.
-  void Promote(LRUHandle* e);
+  // the LRU list. Older items are evicted as necessary. Frees `item` on
+  // non-OK status.
+  Status InsertItem(LRUHandle* item, LRUHandle** handle);
+
   void LRU_Remove(LRUHandle* e);
   void LRU_Insert(LRUHandle* e);
 
@@ -439,8 +369,11 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   // holding the mutex_.
   void EvictFromLRU(size_t charge, autovector<LRUHandle*>* deleted);
 
-  // Try to insert the evicted handles into the secondary cache.
-  void TryInsertIntoSecondaryCache(autovector<LRUHandle*> evicted_handles);
+  void NotifyEvicted(const autovector<LRUHandle*>& evicted_handles);
+
+  LRUHandle* CreateHandle(const Slice& key, uint32_t hash,
+                          Cache::ObjectPtr value,
+                          const Cache::CacheItemHelper* helper, size_t charge);
 
   // Initialized before use.
   size_t capacity_;
@@ -503,8 +436,8 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   // don't mind mutex_ invoking the non-const actions.
   mutable DMutex mutex_;
 
-  // Owned by LRUCache
-  SecondaryCache* secondary_cache_;
+  // A reference to Cache::eviction_callback_
+  const Cache::EvictionCallback& eviction_callback_;
 };
 
 class LRUCache
@@ -513,28 +446,16 @@ class LRUCache
 #endif
     : public ShardedCache<LRUCacheShard> {
  public:
-  LRUCache(size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-           double high_pri_pool_ratio, double low_pri_pool_ratio,
-           std::shared_ptr<MemoryAllocator> memory_allocator = nullptr,
-           bool use_adaptive_mutex = kDefaultToAdaptiveMutex,
-           CacheMetadataChargePolicy metadata_charge_policy =
-               kDontChargeCacheMetadata,
-           std::shared_ptr<SecondaryCache> secondary_cache = nullptr);
+  explicit LRUCache(const LRUCacheOptions& opts);
   const char* Name() const override { return "LRUCache"; }
-  void* Value(Handle* handle) override;
+  ObjectPtr Value(Handle* handle) override;
   size_t GetCharge(Handle* handle) const override;
-  DeleterFn GetDeleter(Handle* handle) const override;
-  void WaitAll(std::vector<Handle*>& handles) override;
+  const CacheItemHelper* GetCacheItemHelper(Handle* handle) const override;
 
   // Retrieves number of elements in LRU, for unit test purpose only.
   size_t TEST_GetLRUSize();
   // Retrieves high pri pool ratio.
   double GetHighPriPoolRatio();
-
-  void AppendPrintableOptions(std::string& str) const override;
-
- private:
-  std::shared_ptr<SecondaryCache> secondary_cache_;
 };
 
 }  // namespace lru_cache

cache/secondary_cache.cc

@@ -11,20 +11,32 @@ namespace ROCKSDB_NAMESPACE {
 
 namespace {
 
-size_t SliceSize(void* obj) { return static_cast<Slice*>(obj)->size(); }
+void NoopDelete(Cache::ObjectPtr, MemoryAllocator*) {}
 
-Status SliceSaveTo(void* from_obj, size_t from_offset, size_t length,
-                   void* out) {
+size_t SliceSize(Cache::ObjectPtr obj) {
+  return static_cast<Slice*>(obj)->size();
+}
+
+Status SliceSaveTo(Cache::ObjectPtr from_obj, size_t from_offset, size_t length,
+                   char* out) {
   const Slice& slice = *static_cast<Slice*>(from_obj);
   std::memcpy(out, slice.data() + from_offset, length);
   return Status::OK();
 }
 
+Status FailCreate(const Slice&, Cache::CreateContext*, MemoryAllocator*,
+                  Cache::ObjectPtr*, size_t*) {
+  return Status::NotSupported("Only for dumping data into SecondaryCache");
+}
+
 }  // namespace
 
 Status SecondaryCache::InsertSaved(const Slice& key, const Slice& saved) {
+  static Cache::CacheItemHelper helper_no_secondary{CacheEntryRole::kMisc,
+                                                    &NoopDelete};
   static Cache::CacheItemHelper helper{
-      &SliceSize, &SliceSaveTo, GetNoopDeleterForRole<CacheEntryRole::kMisc>()};
+      CacheEntryRole::kMisc, &NoopDelete, &SliceSize,
+      &SliceSaveTo,          &FailCreate, &helper_no_secondary};
   // NOTE: depends on Insert() being synchronous, not keeping pointer `&saved`
   return Insert(key, const_cast<Slice*>(&saved), &helper);
 }

cache/secondary_cache_adapter.cc

@@ -0,0 +1,295 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#include "cache/secondary_cache_adapter.h"
+
+#include "monitoring/perf_context_imp.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+namespace {
+// A distinct pointer value for marking "dummy" cache entries
+struct Dummy {
+  char val[7] = "kDummy";
+};
+const Dummy kDummy{};
+Cache::ObjectPtr const kDummyObj = const_cast<Dummy*>(&kDummy);
+}  // namespace
+
+CacheWithSecondaryAdapter::CacheWithSecondaryAdapter(
+    std::shared_ptr<Cache> target,
+    std::shared_ptr<SecondaryCache> secondary_cache)
+    : CacheWrapper(std::move(target)),
+      secondary_cache_(std::move(secondary_cache)) {
+  target_->SetEvictionCallback([this](const Slice& key, Handle* handle) {
+    return EvictionHandler(key, handle);
+  });
+}
+
+CacheWithSecondaryAdapter::~CacheWithSecondaryAdapter() {
+  // `*this` will be destroyed before `*target_`, so we have to prevent
+  // use after free
+  target_->SetEvictionCallback({});
+}
+
+bool CacheWithSecondaryAdapter::EvictionHandler(const Slice& key,
+                                                Handle* handle) {
+  auto helper = GetCacheItemHelper(handle);
+  if (helper->IsSecondaryCacheCompatible()) {
+    auto obj = target_->Value(handle);
+    // Ignore dummy entry
+    if (obj != kDummyObj) {
+      // Spill into secondary cache.
+      secondary_cache_->Insert(key, obj, helper).PermitUncheckedError();
+    }
+  }
+  // Never takes ownership of obj
+  return false;
+}
+
+bool CacheWithSecondaryAdapter::ProcessDummyResult(Cache::Handle** handle,
+                                                   bool erase) {
+  if (*handle && target_->Value(*handle) == kDummyObj) {
+    target_->Release(*handle, erase);
+    *handle = nullptr;
+    return true;
+  } else {
+    return false;
+  }
+}
+
+void CacheWithSecondaryAdapter::CleanupCacheObject(
+    ObjectPtr obj, const CacheItemHelper* helper) {
+  if (helper->del_cb) {
+    helper->del_cb(obj, memory_allocator());
+  }
+}
+
+Cache::Handle* CacheWithSecondaryAdapter::Promote(
+    std::unique_ptr<SecondaryCacheResultHandle>&& secondary_handle,
+    const Slice& key, const CacheItemHelper* helper, Priority priority,
+    Statistics* stats, bool found_dummy_entry, bool kept_in_sec_cache) {
+  assert(secondary_handle->IsReady());
+
+  ObjectPtr obj = secondary_handle->Value();
+  if (!obj) {
+    // Nothing found.
+    return nullptr;
+  }
+  // Found something.
+  switch (helper->role) {
+    case CacheEntryRole::kFilterBlock:
+      RecordTick(stats, SECONDARY_CACHE_FILTER_HITS);
+      break;
+    case CacheEntryRole::kIndexBlock:
+      RecordTick(stats, SECONDARY_CACHE_INDEX_HITS);
+      break;
+    case CacheEntryRole::kDataBlock:
+      RecordTick(stats, SECONDARY_CACHE_DATA_HITS);
+      break;
+    default:
+      break;
+  }
+  PERF_COUNTER_ADD(secondary_cache_hit_count, 1);
+  RecordTick(stats, SECONDARY_CACHE_HITS);
+
+  // Note: SecondaryCache::Size() is really charge (from the CreateCallback)
+  size_t charge = secondary_handle->Size();
+  Handle* result = nullptr;
+  // Insert into primary cache, possibly as a standalone+dummy entries.
+  if (secondary_cache_->SupportForceErase() && !found_dummy_entry) {
+    // Create standalone and insert dummy
+    // Allow standalone to be created even if cache is full, to avoid
+    // reading the entry from storage.
+    result =
+        CreateStandalone(key, obj, helper, charge, /*allow_uncharged*/ true);
+    assert(result);
+    PERF_COUNTER_ADD(block_cache_standalone_handle_count, 1);
+
+    // Insert dummy to record recent use
+    // TODO: try to avoid case where inserting this dummy could overwrite a
+    // regular entry
+    Status s = Insert(key, kDummyObj, &kNoopCacheItemHelper, /*charge=*/0,
+                      /*handle=*/nullptr, priority);
+    s.PermitUncheckedError();
+    // Nothing to do or clean up on dummy insertion failure
+  } else {
+    // Insert regular entry into primary cache.
+    // Don't allow it to spill into secondary cache again if it was kept there.
+    Status s = Insert(
+        key, obj, kept_in_sec_cache ? helper->without_secondary_compat : helper,
+        charge, &result, priority);
+    if (s.ok()) {
+      assert(result);
+      PERF_COUNTER_ADD(block_cache_real_handle_count, 1);
+    } else {
+      // Create standalone result instead, even if cache is full, to avoid
+      // reading the entry from storage.
+      result =
+          CreateStandalone(key, obj, helper, charge, /*allow_uncharged*/ true);
+      assert(result);
+      PERF_COUNTER_ADD(block_cache_standalone_handle_count, 1);
+    }
+  }
+  return result;
+}
+
+Cache::Handle* CacheWithSecondaryAdapter::Lookup(const Slice& key,
+                                                 const CacheItemHelper* helper,
+                                                 CreateContext* create_context,
+                                                 Priority priority,
+                                                 Statistics* stats) {
+  // NOTE: we could just StartAsyncLookup() and Wait(), but this should be a bit
+  // more efficient
+  Handle* result =
+      target_->Lookup(key, helper, create_context, priority, stats);
+  bool secondary_compatible = helper && helper->IsSecondaryCacheCompatible();
+  bool found_dummy_entry =
+      ProcessDummyResult(&result, /*erase=*/secondary_compatible);
+  if (!result && secondary_compatible) {
+    // Try our secondary cache
+    bool kept_in_sec_cache = false;
+    std::unique_ptr<SecondaryCacheResultHandle> secondary_handle =
+        secondary_cache_->Lookup(key, helper, create_context, /*wait*/ true,
+                                 found_dummy_entry, /*out*/ kept_in_sec_cache);
+    if (secondary_handle) {
+      result = Promote(std::move(secondary_handle), key, helper, priority,
+                       stats, found_dummy_entry, kept_in_sec_cache);
+    }
+  }
+  return result;
+}
+
+Cache::ObjectPtr CacheWithSecondaryAdapter::Value(Handle* handle) {
+  ObjectPtr v = target_->Value(handle);
+  // TODO with stacked secondaries: might fail in EvictionHandler
+  assert(v != kDummyObj);
+  return v;
+}
+
+void CacheWithSecondaryAdapter::StartAsyncLookupOnMySecondary(
+    AsyncLookupHandle& async_handle) {
+  assert(!async_handle.IsPending());
+  assert(async_handle.result_handle == nullptr);
+
+  std::unique_ptr<SecondaryCacheResultHandle> secondary_handle =
+      secondary_cache_->Lookup(async_handle.key, async_handle.helper,
+                               async_handle.create_context, /*wait*/ false,
+                               async_handle.found_dummy_entry,
+                               /*out*/ async_handle.kept_in_sec_cache);
+  if (secondary_handle) {
+    // TODO with stacked secondaries: Check & process if already ready?
+    async_handle.pending_handle = secondary_handle.release();
+    async_handle.pending_cache = secondary_cache_.get();
+  }
+}
+
+void CacheWithSecondaryAdapter::StartAsyncLookup(
+    AsyncLookupHandle& async_handle) {
+  target_->StartAsyncLookup(async_handle);
+  if (!async_handle.IsPending()) {
+    bool secondary_compatible =
+        async_handle.helper &&
+        async_handle.helper->IsSecondaryCacheCompatible();
+    async_handle.found_dummy_entry |= ProcessDummyResult(
+        &async_handle.result_handle, /*erase=*/secondary_compatible);
+
+    if (async_handle.Result() == nullptr && secondary_compatible) {
+      // Not found and not pending on another secondary cache
+      StartAsyncLookupOnMySecondary(async_handle);
+    }
+  }
+}
+
+void CacheWithSecondaryAdapter::WaitAll(AsyncLookupHandle* async_handles,
+                                        size_t count) {
+  if (count == 0) {
+    // Nothing to do
+    return;
+  }
+  // Requests that are pending on *my* secondary cache, at the start of this
+  // function
+  std::vector<AsyncLookupHandle*> my_pending;
+  // Requests that are pending on an "inner" secondary cache (managed somewhere
+  // under target_), as of the start of this function
+  std::vector<AsyncLookupHandle*> inner_pending;
+
+  // Initial accounting of pending handles, excluding those already handled
+  // by "outer" secondary caches. (See cur->pending_cache = nullptr.)
+  for (size_t i = 0; i < count; ++i) {
+    AsyncLookupHandle* cur = async_handles + i;
+    if (cur->pending_cache) {
+      assert(cur->IsPending());
+      assert(cur->helper);
+      assert(cur->helper->IsSecondaryCacheCompatible());
+      if (cur->pending_cache == secondary_cache_.get()) {
+        my_pending.push_back(cur);
+        // Mark as "to be handled by this caller"
+        cur->pending_cache = nullptr;
+      } else {
+        // Remember as potentially needing a lookup in my secondary
+        inner_pending.push_back(cur);
+      }
+    }
+  }
+
+  // Wait on inner-most cache lookups first
+  // TODO with stacked secondaries: because we are not using proper
+  // async/await constructs here yet, there is a false synchronization point
+  // here where all the results at one level are needed before initiating
+  // any lookups at the next level. Probably not a big deal, but worth noting.
+  if (!inner_pending.empty()) {
+    target_->WaitAll(async_handles, count);
+  }
+
+  // For those that failed to find something, convert to lookup in my
+  // secondary cache.
+  for (AsyncLookupHandle* cur : inner_pending) {
+    if (cur->Result() == nullptr) {
+      // Not found, try my secondary
+      StartAsyncLookupOnMySecondary(*cur);
+      if (cur->IsPending()) {
+        assert(cur->pending_cache == secondary_cache_.get());
+        my_pending.push_back(cur);
+        // Mark as "to be handled by this caller"
+        cur->pending_cache = nullptr;
+      }
+    }
+  }
+
+  // Wait on all lookups on my secondary cache
+  {
+    std::vector<SecondaryCacheResultHandle*> my_secondary_handles;
+    for (AsyncLookupHandle* cur : my_pending) {
+      my_secondary_handles.push_back(cur->pending_handle);
+    }
+    secondary_cache_->WaitAll(my_secondary_handles);
+  }
+
+  // Process results
+  for (AsyncLookupHandle* cur : my_pending) {
+    std::unique_ptr<SecondaryCacheResultHandle> secondary_handle(
+        cur->pending_handle);
+    cur->pending_handle = nullptr;
+    cur->result_handle = Promote(
+        std::move(secondary_handle), cur->key, cur->helper, cur->priority,
+        cur->stats, cur->found_dummy_entry, cur->kept_in_sec_cache);
+    assert(cur->pending_cache == nullptr);
+  }
+}
+
+std::string CacheWithSecondaryAdapter::GetPrintableOptions() const {
+  std::string str = target_->GetPrintableOptions();
+  str.append("  secondary_cache:\n");
+  str.append(secondary_cache_->GetPrintableOptions());
+  return str;
+}
+
+const char* CacheWithSecondaryAdapter::Name() const {
+  // To the user, at least for now, configure the underlying cache with
+  // a secondary cache. So we pretend to be that cache
+  return target_->Name();
+}
+}  // namespace ROCKSDB_NAMESPACE

cache/secondary_cache_adapter.h

@@ -0,0 +1,52 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#pragma once
+
+#include "rocksdb/secondary_cache.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+class CacheWithSecondaryAdapter : public CacheWrapper {
+ public:
+  explicit CacheWithSecondaryAdapter(
+      std::shared_ptr<Cache> target,
+      std::shared_ptr<SecondaryCache> secondary_cache);
+
+  ~CacheWithSecondaryAdapter() override;
+
+  Handle* Lookup(const Slice& key, const CacheItemHelper* helper,
+                 CreateContext* create_context,
+                 Priority priority = Priority::LOW,
+                 Statistics* stats = nullptr) override;
+
+  ObjectPtr Value(Handle* handle) override;
+
+  void StartAsyncLookup(AsyncLookupHandle& async_handle) override;
+
+  void WaitAll(AsyncLookupHandle* async_handles, size_t count) override;
+
+  std::string GetPrintableOptions() const override;
+
+  const char* Name() const override;
+
+ private:
+  bool EvictionHandler(const Slice& key, Handle* handle);
+
+  void StartAsyncLookupOnMySecondary(AsyncLookupHandle& async_handle);
+
+  Handle* Promote(
+      std::unique_ptr<SecondaryCacheResultHandle>&& secondary_handle,
+      const Slice& key, const CacheItemHelper* helper, Priority priority,
+      Statistics* stats, bool found_dummy_entry, bool kept_in_sec_cache);
+
+  bool ProcessDummyResult(Cache::Handle** handle, bool erase);
+
+  void CleanupCacheObject(ObjectPtr obj, const CacheItemHelper* helper);
+
+  std::shared_ptr<SecondaryCache> secondary_cache_;
+};
+
+}  // namespace ROCKSDB_NAMESPACE

cache/sharded_cache.cc

@@ -13,20 +13,57 @@
 #include <cstdint>
 #include <memory>
 
+#include "env/unique_id_gen.h"
+#include "rocksdb/env.h"
 #include "util/hash.h"
 #include "util/math.h"
 #include "util/mutexlock.h"
 
 namespace ROCKSDB_NAMESPACE {
+namespace {
+// The generated seeds must fit in 31 bits so that
+// ShardedCacheOptions::hash_seed can be set to it explicitly, for
+// diagnostic/debugging purposes.
+constexpr uint32_t kSeedMask = 0x7fffffff;
+uint32_t DetermineSeed(int32_t hash_seed_option) {
+  if (hash_seed_option >= 0) {
+    // User-specified exact seed
+    return static_cast<uint32_t>(hash_seed_option);
+  }
+  static SemiStructuredUniqueIdGen gen;
+  if (hash_seed_option == ShardedCacheOptions::kHostHashSeed) {
+    std::string hostname;
+    Status s = Env::Default()->GetHostNameString(&hostname);
+    if (s.ok()) {
+      return GetSliceHash(hostname) & kSeedMask;
+    } else {
+      // Fall back on something stable within the process.
+      return static_cast<uint32_t>(gen.GetBaseUpper()) & kSeedMask;
+    }
+  } else {
+    // for kQuasiRandomHashSeed and fallback
+    uint32_t val = gen.GenerateNext<uint32_t>() & kSeedMask;
+    // Perform some 31-bit bijective transformations so that we get
+    // quasirandom, not just incrementing. (An incrementing seed from a
+    // random starting point would be fine, but hard to describe in a name.)
+    // See https://en.wikipedia.org/wiki/Quasirandom and using a murmur-like
+    // transformation here for our bijection in the lower 31 bits.
+    // See https://en.wikipedia.org/wiki/MurmurHash
+    val *= /*31-bit prime*/ 1150630961;
+    val ^= (val & kSeedMask) >> 17;
+    val *= /*31-bit prime*/ 1320603883;
+    return val & kSeedMask;
+  }
+}
+}  // namespace
 
-ShardedCacheBase::ShardedCacheBase(size_t capacity, int num_shard_bits,
-                                   bool strict_capacity_limit,
-                                   std::shared_ptr<MemoryAllocator> allocator)
-    : Cache(std::move(allocator)),
+ShardedCacheBase::ShardedCacheBase(const ShardedCacheOptions& opts)
+    : Cache(opts.memory_allocator),
       last_id_(1),
-      shard_mask_((uint32_t{1} << num_shard_bits) - 1),
-      strict_capacity_limit_(strict_capacity_limit),
-      capacity_(capacity) {}
+      shard_mask_((uint32_t{1} << opts.num_shard_bits) - 1),
+      hash_seed_(DetermineSeed(opts.hash_seed)),
+      strict_capacity_limit_(opts.strict_capacity_limit),
+      capacity_(opts.capacity) {}
 
 size_t ShardedCacheBase::ComputePerShardCapacity(size_t capacity) const {
   uint32_t num_shards = GetNumShards();

cache/sharded_cache.h

@@ -15,7 +15,7 @@
 
 #include "port/lang.h"
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "util/hash.h"
 #include "util/mutexlock.h"
 
@@ -34,8 +34,8 @@ class CacheShardBase {
   std::string GetPrintableOptions() const { return ""; }
   using HashVal = uint64_t;
   using HashCref = uint64_t;
-  static inline HashVal ComputeHash(const Slice& key) {
-    return GetSliceNPHash64(key);
+  static inline HashVal ComputeHash(const Slice& key, uint32_t seed) {
+    return GetSliceNPHash64(key, seed);
   }
   static inline uint32_t HashPieceForSharding(HashCref hash) {
     return Lower32of64(hash);
@@ -49,21 +49,19 @@ class CacheShardBase {
     HashCref GetHash() const;
     ...
   };
-  Status Insert(const Slice& key, HashCref hash, void* value, size_t charge,
-                DeleterFn deleter, HandleImpl** handle,
-                Cache::Priority priority) = 0;
-  Status Insert(const Slice& key, HashCref hash, void* value,
+  Status Insert(const Slice& key, HashCref hash, Cache::ObjectPtr value,
                 const Cache::CacheItemHelper* helper, size_t charge,
-                HandleImpl** handle, Cache::Priority priority) = 0;
-  HandleImpl* Lookup(const Slice& key, HashCref hash) = 0;
+                HandleImpl** handle, Cache::Priority priority,
+                bool standalone) = 0;
+  Handle* CreateStandalone(const Slice& key, HashCref hash, ObjectPtr obj,
+                           const CacheItemHelper* helper,
+                           size_t charge, bool allow_uncharged) = 0;
   HandleImpl* Lookup(const Slice& key, HashCref hash,
                         const Cache::CacheItemHelper* helper,
-                        const Cache::CreateCallback& create_cb,
-                        Cache::Priority priority, bool wait,
+                        Cache::CreateContext* create_context,
+                        Cache::Priority priority,
                         Statistics* stats) = 0;
   bool Release(HandleImpl* handle, bool useful, bool erase_if_last_ref) = 0;
-  bool IsReady(HandleImpl* handle) = 0;
-  void Wait(HandleImpl* handle) = 0;
   bool Ref(HandleImpl* handle) = 0;
   void Erase(const Slice& key, HashCref hash) = 0;
   void SetCapacity(size_t capacity) = 0;
@@ -77,8 +75,9 @@ class CacheShardBase {
   // *state == 0 and implementation sets *state = SIZE_MAX to indicate
   // completion.
   void ApplyToSomeEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
+      const std::function<void(const Slice& key, ObjectPtr value,
+                               size_t charge,
+                               const Cache::CacheItemHelper* helper)>& callback,
       size_t average_entries_per_lock, size_t* state) = 0;
   void EraseUnRefEntries() = 0;
   */
@@ -90,9 +89,7 @@ class CacheShardBase {
 // Portions of ShardedCache that do not depend on the template parameter
 class ShardedCacheBase : public Cache {
  public:
-  ShardedCacheBase(size_t capacity, int num_shard_bits,
-                   bool strict_capacity_limit,
-                   std::shared_ptr<MemoryAllocator> memory_allocator);
+  explicit ShardedCacheBase(const ShardedCacheOptions& opts);
   virtual ~ShardedCacheBase() = default;
 
   int GetNumShardBits() const;
@@ -107,6 +104,8 @@ class ShardedCacheBase : public Cache {
   size_t GetUsage(Handle* handle) const override;
   std::string GetPrintableOptions() const override;
 
+  uint32_t GetHashSeed() const override { return hash_seed_; }
+
  protected:  // fns
   virtual void AppendPrintableOptions(std::string& str) const = 0;
   size_t GetPerShardCapacity() const;
@@ -115,6 +114,7 @@ class ShardedCacheBase : public Cache {
  protected:                        // data
   std::atomic<uint64_t> last_id_;  // For NewId
   const uint32_t shard_mask_;
+  const uint32_t hash_seed_;
 
   // Dynamic configuration parameters, guarded by config_mutex_
   bool strict_capacity_limit_;
@@ -135,10 +135,8 @@ class ShardedCache : public ShardedCacheBase {
   using HashCref = typename CacheShard::HashCref;
   using HandleImpl = typename CacheShard::HandleImpl;
 
-  ShardedCache(size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-               std::shared_ptr<MemoryAllocator> allocator)
-      : ShardedCacheBase(capacity, num_shard_bits, strict_capacity_limit,
-                         allocator),
+  explicit ShardedCache(const ShardedCacheOptions& opts)
+      : ShardedCacheBase(opts),
         shards_(reinterpret_cast<CacheShard*>(port::cacheline_aligned_alloc(
             sizeof(CacheShard) * GetNumShards()))),
         destroy_shards_in_dtor_(false) {}
@@ -172,41 +170,38 @@ class ShardedCache : public ShardedCacheBase {
         [s_c_l](CacheShard* cs) { cs->SetStrictCapacityLimit(s_c_l); });
   }
 
-  Status Insert(const Slice& key, void* value, size_t charge, DeleterFn deleter,
-                Handle** handle, Priority priority) override {
-    HashVal hash = CacheShard::ComputeHash(key);
-    auto h_out = reinterpret_cast<HandleImpl**>(handle);
-    return GetShard(hash).Insert(key, hash, value, charge, deleter, h_out,
-                                 priority);
-  }
-  Status Insert(const Slice& key, void* value, const CacheItemHelper* helper,
+  Status Insert(const Slice& key, ObjectPtr obj, const CacheItemHelper* helper,
                 size_t charge, Handle** handle = nullptr,
                 Priority priority = Priority::LOW) override {
-    if (!helper) {
-      return Status::InvalidArgument();
-    }
-    HashVal hash = CacheShard::ComputeHash(key);
+    assert(helper);
+    HashVal hash = CacheShard::ComputeHash(key, hash_seed_);
     auto h_out = reinterpret_cast<HandleImpl**>(handle);
-    return GetShard(hash).Insert(key, hash, value, helper, charge, h_out,
+    return GetShard(hash).Insert(key, hash, obj, helper, charge, h_out,
                                  priority);
   }
 
-  Handle* Lookup(const Slice& key, Statistics* /*stats*/) override {
-    HashVal hash = CacheShard::ComputeHash(key);
-    HandleImpl* result = GetShard(hash).Lookup(key, hash);
+  Handle* CreateStandalone(const Slice& key, ObjectPtr obj,
+                           const CacheItemHelper* helper, size_t charge,
+                           bool allow_uncharged) override {
+    assert(helper);
+    HashVal hash = CacheShard::ComputeHash(key, hash_seed_);
+    HandleImpl* result = GetShard(hash).CreateStandalone(
+        key, hash, obj, helper, charge, allow_uncharged);
     return reinterpret_cast<Handle*>(result);
   }
-  Handle* Lookup(const Slice& key, const CacheItemHelper* helper,
-                 const CreateCallback& create_cb, Priority priority, bool wait,
+
+  Handle* Lookup(const Slice& key, const CacheItemHelper* helper = nullptr,
+                 CreateContext* create_context = nullptr,
+                 Priority priority = Priority::LOW,
                  Statistics* stats = nullptr) override {
-    HashVal hash = CacheShard::ComputeHash(key);
-    HandleImpl* result = GetShard(hash).Lookup(key, hash, helper, create_cb,
-                                               priority, wait, stats);
+    HashVal hash = CacheShard::ComputeHash(key, hash_seed_);
+    HandleImpl* result = GetShard(hash).Lookup(key, hash, helper,
+                                               create_context, priority, stats);
     return reinterpret_cast<Handle*>(result);
   }
 
   void Erase(const Slice& key) override {
-    HashVal hash = CacheShard::ComputeHash(key);
+    HashVal hash = CacheShard::ComputeHash(key, hash_seed_);
     GetShard(hash).Erase(key, hash);
   }
 
@@ -215,14 +210,6 @@ class ShardedCache : public ShardedCacheBase {
     auto h = reinterpret_cast<HandleImpl*>(handle);
     return GetShard(h->GetHash()).Release(h, useful, erase_if_last_ref);
   }
-  bool IsReady(Handle* handle) override {
-    auto h = reinterpret_cast<HandleImpl*>(handle);
-    return GetShard(h->GetHash()).IsReady(h);
-  }
-  void Wait(Handle* handle) override {
-    auto h = reinterpret_cast<HandleImpl*>(handle);
-    GetShard(h->GetHash()).Wait(h);
-  }
   bool Ref(Handle* handle) override {
     auto h = reinterpret_cast<HandleImpl*>(handle);
     return GetShard(h->GetHash()).Ref(h);
@@ -238,14 +225,14 @@ class ShardedCache : public ShardedCacheBase {
     return SumOverShards2(&CacheShard::GetPinnedUsage);
   }
   size_t GetOccupancyCount() const override {
-    return SumOverShards2(&CacheShard::GetPinnedUsage);
+    return SumOverShards2(&CacheShard::GetOccupancyCount);
   }
   size_t GetTableAddressCount() const override {
     return SumOverShards2(&CacheShard::GetTableAddressCount);
   }
   void ApplyToAllEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
+      const std::function<void(const Slice& key, ObjectPtr value, size_t charge,
+                               const CacheItemHelper* helper)>& callback,
       const ApplyToAllEntriesOptions& opts) override {
     uint32_t num_shards = GetNumShards();
     // Iterate over part of each shard, rotating between shards, to

cache/typed_cache.h

@@ -0,0 +1,375 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+// APIs for accessing Cache in a type-safe and convenient way. Cache is kept
+// at a low, thin level of abstraction so that different implementations can
+// be plugged in, but these wrappers provide clean, convenient access to the
+// most common operations.
+//
+// A number of template classes are needed for sharing common structure. The
+// key classes are these:
+//
+// * PlaceholderCacheInterface - Used for making cache reservations, with
+// entries that have a charge but no value.
+// * BasicTypedCacheInterface<TValue> - Used for primary cache storage of
+// objects of type TValue.
+// * FullTypedCacheHelper<TValue, TCreateContext> - Used for secondary cache
+// compatible storage of objects of type TValue.
+// * For each of these, there's a "Shared" version
+// (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache,
+// rather than assuming external ownership by holding only a raw `Cache*`.
+
+#pragma once
+
+#include <algorithm>
+#include <cstdint>
+#include <memory>
+#include <type_traits>
+
+#include "cache/cache_helpers.h"
+#include "rocksdb/advanced_cache.h"
+#include "rocksdb/advanced_options.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+// For future consideration:
+// * Pass in value to Insert with std::unique_ptr& to simplify ownership
+//   transfer logic in callers
+// * Make key type a template parameter (e.g. useful for table cache)
+// * Closer integration with CacheHandleGuard (opt-in, so not always
+//   paying the extra overhead)
+
+#define CACHE_TYPE_DEFS()                     \
+  using Priority = Cache::Priority;           \
+  using Handle = Cache::Handle;               \
+  using ObjectPtr = Cache::ObjectPtr;         \
+  using CreateContext = Cache::CreateContext; \
+  using CacheItemHelper = Cache::CacheItemHelper /* caller ; */
+
+template <typename CachePtr>
+class BaseCacheInterface {
+ public:
+  CACHE_TYPE_DEFS();
+
+  /*implicit*/ BaseCacheInterface(CachePtr cache) : cache_(std::move(cache)) {}
+
+  inline void Release(Handle* handle) { cache_->Release(handle); }
+
+  inline void ReleaseAndEraseIfLastRef(Handle* handle) {
+    cache_->Release(handle, /*erase_if_last_ref*/ true);
+  }
+
+  inline void RegisterReleaseAsCleanup(Handle* handle, Cleanable& cleanable) {
+    cleanable.RegisterCleanup(&ReleaseCacheHandleCleanup, get(), handle);
+  }
+
+  inline Cache* get() const { return &*cache_; }
+
+  explicit inline operator bool() const noexcept { return cache_ != nullptr; }
+
+ protected:
+  CachePtr cache_;
+};
+
+// PlaceholderCacheInterface - Used for making cache reservations, with
+// entries that have a charge but no value. CacheEntryRole is required as
+// a template parameter.
+template <CacheEntryRole kRole, typename CachePtr = Cache*>
+class PlaceholderCacheInterface : public BaseCacheInterface<CachePtr> {
+ public:
+  CACHE_TYPE_DEFS();
+  using BaseCacheInterface<CachePtr>::BaseCacheInterface;
+
+  inline Status Insert(const Slice& key, size_t charge, Handle** handle) {
+    return this->cache_->Insert(key, /*value=*/nullptr, GetHelper(), charge,
+                                handle);
+  }
+
+  static const Cache::CacheItemHelper* GetHelper() {
+    static const Cache::CacheItemHelper kHelper{kRole};
+    return &kHelper;
+  }
+};
+
+template <CacheEntryRole kRole>
+using PlaceholderSharedCacheInterface =
+    PlaceholderCacheInterface<kRole, std::shared_ptr<Cache>>;
+
+template <class TValue>
+class BasicTypedCacheHelperFns {
+ public:
+  CACHE_TYPE_DEFS();
+  // E.g. char* for char[]
+  using TValuePtr = std::remove_extent_t<TValue>*;
+
+ protected:
+  inline static ObjectPtr UpCastValue(TValuePtr value) { return value; }
+  inline static TValuePtr DownCastValue(ObjectPtr value) {
+    return static_cast<TValuePtr>(value);
+  }
+
+  static void Delete(ObjectPtr value, MemoryAllocator* allocator) {
+    // FIXME: Currently, no callers actually allocate the ObjectPtr objects
+    // using the custom allocator, just subobjects that keep a reference to
+    // the allocator themselves (with CacheAllocationPtr).
+    if (/*DISABLED*/ false && allocator) {
+      if constexpr (std::is_destructible_v<TValue>) {
+        DownCastValue(value)->~TValue();
+      }
+      allocator->Deallocate(value);
+    } else {
+      // Like delete but properly handles TValue=char[] etc.
+      std::default_delete<TValue>{}(DownCastValue(value));
+    }
+  }
+};
+
+// In its own class to try to minimize the number of distinct CacheItemHelper
+// instances (e.g. don't vary by CachePtr)
+template <class TValue, CacheEntryRole kRole>
+class BasicTypedCacheHelper : public BasicTypedCacheHelperFns<TValue> {
+ public:
+  static const Cache::CacheItemHelper* GetBasicHelper() {
+    static const Cache::CacheItemHelper kHelper{kRole,
+                                                &BasicTypedCacheHelper::Delete};
+    return &kHelper;
+  }
+};
+
+// BasicTypedCacheInterface - Used for primary cache storage of objects of
+// type TValue, which can be cleaned up with std::default_delete<TValue>. The
+// role is provided by TValue::kCacheEntryRole or given in an optional
+// template parameter.
+template <class TValue, CacheEntryRole kRole = TValue::kCacheEntryRole,
+          typename CachePtr = Cache*>
+class BasicTypedCacheInterface : public BaseCacheInterface<CachePtr>,
+                                 public BasicTypedCacheHelper<TValue, kRole> {
+ public:
+  CACHE_TYPE_DEFS();
+  using typename BasicTypedCacheHelperFns<TValue>::TValuePtr;
+  struct TypedHandle : public Handle {};
+  using BasicTypedCacheHelper<TValue, kRole>::GetBasicHelper;
+  // ctor
+  using BaseCacheInterface<CachePtr>::BaseCacheInterface;
+  struct TypedAsyncLookupHandle : public Cache::AsyncLookupHandle {
+    TypedHandle* Result() {
+      return reinterpret_cast<TypedHandle*>(Cache::AsyncLookupHandle::Result());
+    }
+  };
+
+  inline Status Insert(const Slice& key, TValuePtr value, size_t charge,
+                       TypedHandle** handle = nullptr,
+                       Priority priority = Priority::LOW) {
+    auto untyped_handle = reinterpret_cast<Handle**>(handle);
+    return this->cache_->Insert(
+        key, BasicTypedCacheHelperFns<TValue>::UpCastValue(value),
+        GetBasicHelper(), charge, untyped_handle, priority);
+  }
+
+  inline TypedHandle* Lookup(const Slice& key, Statistics* stats = nullptr) {
+    return reinterpret_cast<TypedHandle*>(
+        this->cache_->BasicLookup(key, stats));
+  }
+
+  inline void StartAsyncLookup(TypedAsyncLookupHandle& async_handle) {
+    assert(async_handle.helper == nullptr);
+    this->cache_->StartAsyncLookup(async_handle);
+  }
+
+  inline CacheHandleGuard<TValue> Guard(TypedHandle* handle) {
+    if (handle) {
+      return CacheHandleGuard<TValue>(&*this->cache_, handle);
+    } else {
+      return {};
+    }
+  }
+
+  inline std::shared_ptr<TValue> SharedGuard(TypedHandle* handle) {
+    if (handle) {
+      return MakeSharedCacheHandleGuard<TValue>(&*this->cache_, handle);
+    } else {
+      return {};
+    }
+  }
+
+  inline TValuePtr Value(TypedHandle* handle) {
+    return BasicTypedCacheHelperFns<TValue>::DownCastValue(
+        this->cache_->Value(handle));
+  }
+};
+
+// BasicTypedSharedCacheInterface - Like BasicTypedCacheInterface but with a
+// shared_ptr<Cache> for keeping Cache alive.
+template <class TValue, CacheEntryRole kRole = TValue::kCacheEntryRole>
+using BasicTypedSharedCacheInterface =
+    BasicTypedCacheInterface<TValue, kRole, std::shared_ptr<Cache>>;
+
+// TValue must implement ContentSlice() and ~TValue
+// TCreateContext must implement Create(std::unique_ptr<TValue>*, ...)
+template <class TValue, class TCreateContext>
+class FullTypedCacheHelperFns : public BasicTypedCacheHelperFns<TValue> {
+ public:
+  CACHE_TYPE_DEFS();
+
+ protected:
+  using typename BasicTypedCacheHelperFns<TValue>::TValuePtr;
+  using BasicTypedCacheHelperFns<TValue>::DownCastValue;
+  using BasicTypedCacheHelperFns<TValue>::UpCastValue;
+
+  static size_t Size(ObjectPtr v) {
+    TValuePtr value = DownCastValue(v);
+    auto slice = value->ContentSlice();
+    return slice.size();
+  }
+
+  static Status SaveTo(ObjectPtr v, size_t from_offset, size_t length,
+                       char* out) {
+    TValuePtr value = DownCastValue(v);
+    auto slice = value->ContentSlice();
+    assert(from_offset < slice.size());
+    assert(from_offset + length <= slice.size());
+    std::copy_n(slice.data() + from_offset, length, out);
+    return Status::OK();
+  }
+
+  static Status Create(const Slice& data, CreateContext* context,
+                       MemoryAllocator* allocator, ObjectPtr* out_obj,
+                       size_t* out_charge) {
+    std::unique_ptr<TValue> value = nullptr;
+    if constexpr (sizeof(TCreateContext) > 0) {
+      TCreateContext* tcontext = static_cast<TCreateContext*>(context);
+      tcontext->Create(&value, out_charge, data, allocator);
+    } else {
+      TCreateContext::Create(&value, out_charge, data, allocator);
+    }
+    *out_obj = UpCastValue(value.release());
+    return Status::OK();
+  }
+};
+
+// In its own class to try to minimize the number of distinct CacheItemHelper
+// instances (e.g. don't vary by CachePtr)
+template <class TValue, class TCreateContext, CacheEntryRole kRole>
+class FullTypedCacheHelper
+    : public FullTypedCacheHelperFns<TValue, TCreateContext> {
+ public:
+  static const Cache::CacheItemHelper* GetFullHelper() {
+    static const Cache::CacheItemHelper kHelper{
+        kRole,
+        &FullTypedCacheHelper::Delete,
+        &FullTypedCacheHelper::Size,
+        &FullTypedCacheHelper::SaveTo,
+        &FullTypedCacheHelper::Create,
+        BasicTypedCacheHelper<TValue, kRole>::GetBasicHelper()};
+    return &kHelper;
+  }
+};
+
+// FullTypedCacheHelper - Used for secondary cache compatible storage of
+// objects of type TValue. In addition to BasicTypedCacheInterface constraints,
+// we require TValue::ContentSlice() to return persistable data. This
+// simplifies usage for the normal case of simple secondary cache compatibility
+// (can give you a Slice to the data already in memory). In addition to
+// TCreateContext performing the role of Cache::CreateContext, it is also
+// expected to provide a function Create(std::unique_ptr<TValue>* value,
+// size_t* out_charge, const Slice& data, MemoryAllocator* allocator) for
+// creating new TValue.
+template <class TValue, class TCreateContext,
+          CacheEntryRole kRole = TValue::kCacheEntryRole,
+          typename CachePtr = Cache*>
+class FullTypedCacheInterface
+    : public BasicTypedCacheInterface<TValue, kRole, CachePtr>,
+      public FullTypedCacheHelper<TValue, TCreateContext, kRole> {
+ public:
+  CACHE_TYPE_DEFS();
+  using typename BasicTypedCacheInterface<TValue, kRole, CachePtr>::TypedHandle;
+  using typename BasicTypedCacheInterface<TValue, kRole,
+                                          CachePtr>::TypedAsyncLookupHandle;
+  using typename BasicTypedCacheHelperFns<TValue>::TValuePtr;
+  using BasicTypedCacheHelper<TValue, kRole>::GetBasicHelper;
+  using FullTypedCacheHelper<TValue, TCreateContext, kRole>::GetFullHelper;
+  using BasicTypedCacheHelperFns<TValue>::UpCastValue;
+  using BasicTypedCacheHelperFns<TValue>::DownCastValue;
+  // ctor
+  using BasicTypedCacheInterface<TValue, kRole,
+                                 CachePtr>::BasicTypedCacheInterface;
+
+  // Insert with SecondaryCache compatibility (subject to CacheTier).
+  // (Basic Insert() also inherited.)
+  inline Status InsertFull(
+      const Slice& key, TValuePtr value, size_t charge,
+      TypedHandle** handle = nullptr, Priority priority = Priority::LOW,
+      CacheTier lowest_used_cache_tier = CacheTier::kNonVolatileBlockTier) {
+    auto untyped_handle = reinterpret_cast<Handle**>(handle);
+    auto helper = lowest_used_cache_tier == CacheTier::kNonVolatileBlockTier
+                      ? GetFullHelper()
+                      : GetBasicHelper();
+    return this->cache_->Insert(key, UpCastValue(value), helper, charge,
+                                untyped_handle, priority);
+  }
+
+  // Like SecondaryCache::InsertSaved, with SecondaryCache compatibility
+  // (subject to CacheTier).
+  inline Status InsertSaved(
+      const Slice& key, const Slice& data, TCreateContext* create_context,
+      Priority priority = Priority::LOW,
+      CacheTier lowest_used_cache_tier = CacheTier::kNonVolatileBlockTier,
+      size_t* out_charge = nullptr) {
+    ObjectPtr value;
+    size_t charge;
+    Status st = GetFullHelper()->create_cb(data, create_context,
+                                           this->cache_->memory_allocator(),
+                                           &value, &charge);
+    if (out_charge) {
+      *out_charge = charge;
+    }
+    if (st.ok()) {
+      st = InsertFull(key, DownCastValue(value), charge, nullptr /*handle*/,
+                      priority, lowest_used_cache_tier);
+    } else {
+      GetFullHelper()->del_cb(value, this->cache_->memory_allocator());
+    }
+    return st;
+  }
+
+  // Lookup with SecondaryCache support (subject to CacheTier).
+  // (Basic Lookup() also inherited.)
+  inline TypedHandle* LookupFull(
+      const Slice& key, TCreateContext* create_context = nullptr,
+      Priority priority = Priority::LOW, Statistics* stats = nullptr,
+      CacheTier lowest_used_cache_tier = CacheTier::kNonVolatileBlockTier) {
+    if (lowest_used_cache_tier == CacheTier::kNonVolatileBlockTier) {
+      return reinterpret_cast<TypedHandle*>(this->cache_->Lookup(
+          key, GetFullHelper(), create_context, priority, stats));
+    } else {
+      return BasicTypedCacheInterface<TValue, kRole, CachePtr>::Lookup(key,
+                                                                       stats);
+    }
+  }
+
+  inline void StartAsyncLookupFull(
+      TypedAsyncLookupHandle& async_handle,
+      CacheTier lowest_used_cache_tier = CacheTier::kNonVolatileBlockTier) {
+    if (lowest_used_cache_tier == CacheTier::kNonVolatileBlockTier) {
+      async_handle.helper = GetFullHelper();
+      this->cache_->StartAsyncLookup(async_handle);
+    } else {
+      BasicTypedCacheInterface<TValue, kRole, CachePtr>::StartAsyncLookup(
+          async_handle);
+    }
+  }
+};
+
+// FullTypedSharedCacheInterface - Like FullTypedCacheInterface but with a
+// shared_ptr<Cache> for keeping Cache alive.
+template <class TValue, class TCreateContext,
+          CacheEntryRole kRole = TValue::kCacheEntryRole>
+using FullTypedSharedCacheInterface =
+    FullTypedCacheInterface<TValue, TCreateContext, kRole,
+                            std::shared_ptr<Cache>>;
+
+#undef CACHE_TYPE_DEFS
+
+}  // namespace ROCKSDB_NAMESPACE

coverage/coverage_test.sh

@@ -12,7 +12,7 @@ fi
 ROOT=".."
 # Fetch right version of gcov
 if [ -d /mnt/gvfs/third-party -a -z "$CXX" ]; then
-  source $ROOT/build_tools/fbcode_config_platform009.sh
+  source $ROOT/build_tools/fbcode_config_platform010.sh
   GCOV=$GCC_BASE/bin/gcov
 else
   GCOV=$(which gcov)

db/arena_wrapped_db_iter.cc

@@ -47,6 +47,9 @@ void ArenaWrappedDBIter::Init(
   read_options_ = read_options;
   allow_refresh_ = allow_refresh;
   memtable_range_tombstone_iter_ = nullptr;
+  if (!env->GetFileSystem()->use_async_io()) {
+    read_options_.async_io = false;
+  }
 }
 
 Status ArenaWrappedDBIter::Refresh() {

db/blob/blob_contents.cc

@@ -13,12 +13,6 @@
 
 namespace ROCKSDB_NAMESPACE {
 
-std::unique_ptr<BlobContents> BlobContents::Create(
-    CacheAllocationPtr&& allocation, size_t size) {
-  return std::unique_ptr<BlobContents>(
-      new BlobContents(std::move(allocation), size));
-}
-
 size_t BlobContents::ApproximateMemoryUsage() const {
   size_t usage = 0;
 
@@ -45,46 +39,4 @@ size_t BlobContents::ApproximateMemoryUsage() const {
   return usage;
 }
 
-size_t BlobContents::SizeCallback(void* obj) {
-  assert(obj);
-
-  return static_cast<const BlobContents*>(obj)->size();
-}
-
-Status BlobContents::SaveToCallback(void* from_obj, size_t from_offset,
-                                    size_t length, void* out) {
-  assert(from_obj);
-
-  const BlobContents* buf = static_cast<const BlobContents*>(from_obj);
-  assert(buf->size() >= from_offset + length);
-
-  memcpy(out, buf->data().data() + from_offset, length);
-
-  return Status::OK();
-}
-
-Cache::CacheItemHelper* BlobContents::GetCacheItemHelper() {
-  static Cache::CacheItemHelper cache_helper(
-      &SizeCallback, &SaveToCallback,
-      GetCacheEntryDeleterForRole<BlobContents, CacheEntryRole::kBlobValue>());
-
-  return &cache_helper;
-}
-
-Status BlobContents::CreateCallback(CacheAllocationPtr&& allocation,
-                                    const void* buf, size_t size,
-                                    void** out_obj, size_t* charge) {
-  assert(allocation);
-
-  memcpy(allocation.get(), buf, size);
-
-  std::unique_ptr<BlobContents> obj = Create(std::move(allocation), size);
-  BlobContents* const contents = obj.release();
-
-  *out_obj = contents;
-  *charge = contents->ApproximateMemoryUsage();
-
-  return Status::OK();
-}
-
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_contents.h

@@ -7,8 +7,8 @@
 
 #include <memory>
 
-#include "memory/memory_allocator.h"
-#include "rocksdb/cache.h"
+#include "memory/memory_allocator_impl.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/rocksdb_namespace.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
@@ -18,8 +18,8 @@ namespace ROCKSDB_NAMESPACE {
 // A class representing a single uncompressed value read from a blob file.
 class BlobContents {
  public:
-  static std::unique_ptr<BlobContents> Create(CacheAllocationPtr&& allocation,
-                                              size_t size);
+  BlobContents(CacheAllocationPtr&& allocation, size_t size)
+      : allocation_(std::move(allocation)), data_(allocation_.get(), size) {}
 
   BlobContents(const BlobContents&) = delete;
   BlobContents& operator=(const BlobContents&) = delete;
@@ -34,23 +34,26 @@ class BlobContents {
 
   size_t ApproximateMemoryUsage() const;
 
-  // Callbacks for secondary cache
-  static size_t SizeCallback(void* obj);
-
-  static Status SaveToCallback(void* from_obj, size_t from_offset,
-                               size_t length, void* out);
-
-  static Cache::CacheItemHelper* GetCacheItemHelper();
-
-  static Status CreateCallback(CacheAllocationPtr&& allocation, const void* buf,
-                               size_t size, void** out_obj, size_t* charge);
+  // For TypedCacheInterface
+  const Slice& ContentSlice() const { return data_; }
+  static constexpr CacheEntryRole kCacheEntryRole = CacheEntryRole::kBlobValue;
 
  private:
-  BlobContents(CacheAllocationPtr&& allocation, size_t size)
-      : allocation_(std::move(allocation)), data_(allocation_.get(), size) {}
-
   CacheAllocationPtr allocation_;
   Slice data_;
 };
 
+class BlobContentsCreator : public Cache::CreateContext {
+ public:
+  static void Create(std::unique_ptr<BlobContents>* out, size_t* out_charge,
+                     const Slice& contents, MemoryAllocator* alloc) {
+    auto raw = new BlobContents(AllocateAndCopyBlock(contents, alloc),
+                                contents.size());
+    out->reset(raw);
+    if (out_charge) {
+      *out_charge = raw->ApproximateMemoryUsage();
+    }
+  }
+};
+
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_counting_iterator.h

@@ -123,6 +123,10 @@ class BlobCountingIterator : public InternalIterator {
     return iter_->GetProperty(prop_name, prop);
   }
 
+  bool IsDeleteRangeSentinelKey() const override {
+    return iter_->IsDeleteRangeSentinelKey();
+  }
+
  private:
   void UpdateAndCountBlobIfNeeded() {
     assert(!iter_->Valid() || iter_->status().ok());

db/blob/blob_file_builder.cc

@@ -13,6 +13,7 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/blob_log_format.h"
 #include "db/blob/blob_log_writer.h"
+#include "db/blob/blob_source.h"
 #include "db/event_helpers.h"
 #include "db/version_set.h"
 #include "file/filename.h"
@@ -258,6 +259,7 @@ Status BlobFileBuilder::CompressBlobIfNeeded(
     return Status::OK();
   }
 
+  // TODO: allow user CompressionOptions, including max_compressed_bytes_per_kb
   CompressionOptions opts;
   CompressionContext context(blob_compression_type_);
   constexpr uint64_t sample_for_compression = 0;
@@ -393,7 +395,7 @@ Status BlobFileBuilder::PutBlobIntoCacheIfNeeded(const Slice& blob,
                                                  uint64_t blob_offset) const {
   Status s = Status::OK();
 
-  auto blob_cache = immutable_options_->blob_cache;
+  BlobSource::SharedCacheInterface blob_cache{immutable_options_->blob_cache};
   auto statistics = immutable_options_->statistics.get();
   bool warm_cache =
       prepopulate_blob_cache_ == PrepopulateBlobCache::kFlushOnly &&
@@ -407,34 +409,12 @@ Status BlobFileBuilder::PutBlobIntoCacheIfNeeded(const Slice& blob,
 
     const Cache::Priority priority = Cache::Priority::BOTTOM;
 
-    // Objects to be put into the cache have to be heap-allocated and
-    // self-contained, i.e. own their contents. The Cache has to be able to
-    // take unique ownership of them.
-    CacheAllocationPtr allocation =
-        AllocateBlock(blob.size(), blob_cache->memory_allocator());
-    memcpy(allocation.get(), blob.data(), blob.size());
-    std::unique_ptr<BlobContents> buf =
-        BlobContents::Create(std::move(allocation), blob.size());
-
-    Cache::CacheItemHelper* const cache_item_helper =
-        BlobContents::GetCacheItemHelper();
-    assert(cache_item_helper);
-
-    if (immutable_options_->lowest_used_cache_tier ==
-        CacheTier::kNonVolatileBlockTier) {
-      s = blob_cache->Insert(key, buf.get(), cache_item_helper,
-                             buf->ApproximateMemoryUsage(),
-                             nullptr /* cache_handle */, priority);
-    } else {
-      s = blob_cache->Insert(key, buf.get(), buf->ApproximateMemoryUsage(),
-                             cache_item_helper->del_cb,
-                             nullptr /* cache_handle */, priority);
-    }
+    s = blob_cache.InsertSaved(key, blob, nullptr /*context*/, priority,
+                               immutable_options_->lowest_used_cache_tier);
 
     if (s.ok()) {
       RecordTick(statistics, BLOB_DB_CACHE_ADD);
-      RecordTick(statistics, BLOB_DB_CACHE_BYTES_WRITE, buf->size());
-      buf.release();
+      RecordTick(statistics, BLOB_DB_CACHE_BYTES_WRITE, blob.size());
     } else {
       RecordTick(statistics, BLOB_DB_CACHE_ADD_FAILURES);
     }

db/blob/blob_file_cache.cc

@@ -37,18 +37,18 @@ BlobFileCache::BlobFileCache(Cache* cache,
 }
 
 Status BlobFileCache::GetBlobFileReader(
-    uint64_t blob_file_number,
+    const ReadOptions& read_options, uint64_t blob_file_number,
     CacheHandleGuard<BlobFileReader>* blob_file_reader) {
   assert(blob_file_reader);
   assert(blob_file_reader->IsEmpty());
 
-  const Slice key = GetSlice(&blob_file_number);
+  const Slice key = GetSliceForKey(&blob_file_number);
 
   assert(cache_);
 
-  Cache::Handle* handle = cache_->Lookup(key);
+  TypedHandle* handle = cache_.Lookup(key);
   if (handle) {
-    *blob_file_reader = CacheHandleGuard<BlobFileReader>(cache_, handle);
+    *blob_file_reader = cache_.Guard(handle);
     return Status::OK();
   }
 
@@ -57,9 +57,9 @@ Status BlobFileCache::GetBlobFileReader(
   // Check again while holding mutex
   MutexLock lock(mutex_.get(key));
 
-  handle = cache_->Lookup(key);
+  handle = cache_.Lookup(key);
   if (handle) {
-    *blob_file_reader = CacheHandleGuard<BlobFileReader>(cache_, handle);
+    *blob_file_reader = cache_.Guard(handle);
     return Status::OK();
   }
 
@@ -73,7 +73,7 @@ Status BlobFileCache::GetBlobFileReader(
   {
     assert(file_options_);
     const Status s = BlobFileReader::Create(
-        *immutable_options_, *file_options_, column_family_id_,
+        *immutable_options_, read_options, *file_options_, column_family_id_,
         blob_file_read_hist_, blob_file_number, io_tracer_, &reader);
     if (!s.ok()) {
       RecordTick(statistics, NO_FILE_ERRORS);
@@ -84,8 +84,7 @@ Status BlobFileCache::GetBlobFileReader(
   {
     constexpr size_t charge = 1;
 
-    const Status s = cache_->Insert(key, reader.get(), charge,
-                                    &DeleteCacheEntry<BlobFileReader>, &handle);
+    const Status s = cache_.Insert(key, reader.get(), charge, &handle);
     if (!s.ok()) {
       RecordTick(statistics, NO_FILE_ERRORS);
       return s;
@@ -94,7 +93,7 @@ Status BlobFileCache::GetBlobFileReader(
 
   reader.release();
 
-  *blob_file_reader = CacheHandleGuard<BlobFileReader>(cache_, handle);
+  *blob_file_reader = cache_.Guard(handle);
 
   return Status::OK();
 }

db/blob/blob_file_cache.h

@@ -7,7 +7,8 @@
 
 #include <cinttypes>
 
-#include "cache/cache_helpers.h"
+#include "cache/typed_cache.h"
+#include "db/blob/blob_file_reader.h"
 #include "rocksdb/rocksdb_namespace.h"
 #include "util/mutexlock.h"
 
@@ -18,7 +19,6 @@ struct ImmutableOptions;
 struct FileOptions;
 class HistogramImpl;
 class Status;
-class BlobFileReader;
 class Slice;
 class IOTracer;
 
@@ -32,11 +32,15 @@ class BlobFileCache {
   BlobFileCache(const BlobFileCache&) = delete;
   BlobFileCache& operator=(const BlobFileCache&) = delete;
 
-  Status GetBlobFileReader(uint64_t blob_file_number,
+  Status GetBlobFileReader(const ReadOptions& read_options,
+                           uint64_t blob_file_number,
                            CacheHandleGuard<BlobFileReader>* blob_file_reader);
 
  private:
-  Cache* cache_;
+  using CacheInterface =
+      BasicTypedCacheInterface<BlobFileReader, CacheEntryRole::kMisc>;
+  using TypedHandle = CacheInterface::TypedHandle;
+  CacheInterface cache_;
   // Note: mutex_ below is used to guard against multiple threads racing to open
   // the same file.
   Striped<port::Mutex, Slice> mutex_;

db/blob/blob_file_cache_test.cc

@@ -118,7 +118,9 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader) {
   // First try: reader should be opened and put in cache
   CacheHandleGuard<BlobFileReader> first;
 
-  ASSERT_OK(blob_file_cache.GetBlobFileReader(blob_file_number, &first));
+  const ReadOptions read_options;
+  ASSERT_OK(blob_file_cache.GetBlobFileReader(read_options, blob_file_number,
+                                              &first));
   ASSERT_NE(first.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 0);
@@ -126,7 +128,8 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader) {
   // Second try: reader should be served from cache
   CacheHandleGuard<BlobFileReader> second;
 
-  ASSERT_OK(blob_file_cache.GetBlobFileReader(blob_file_number, &second));
+  ASSERT_OK(blob_file_cache.GetBlobFileReader(read_options, blob_file_number,
+                                              &second));
   ASSERT_NE(second.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 0);
@@ -163,19 +166,21 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader_Race) {
   CacheHandleGuard<BlobFileReader> first;
   CacheHandleGuard<BlobFileReader> second;
 
+  const ReadOptions read_options;
   SyncPoint::GetInstance()->SetCallBack(
       "BlobFileCache::GetBlobFileReader:DoubleCheck", [&](void* /* arg */) {
         // Disabling sync points to prevent infinite recursion
         SyncPoint::GetInstance()->DisableProcessing();
-
-        ASSERT_OK(blob_file_cache.GetBlobFileReader(blob_file_number, &second));
+        ASSERT_OK(blob_file_cache.GetBlobFileReader(read_options,
+                                                    blob_file_number, &second));
         ASSERT_NE(second.GetValue(), nullptr);
         ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
         ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 0);
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
-  ASSERT_OK(blob_file_cache.GetBlobFileReader(blob_file_number, &first));
+  ASSERT_OK(blob_file_cache.GetBlobFileReader(read_options, blob_file_number,
+                                              &first));
   ASSERT_NE(first.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 0);
@@ -213,8 +218,10 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader_IOError) {
 
   CacheHandleGuard<BlobFileReader> reader;
 
+  const ReadOptions read_options;
   ASSERT_TRUE(
-      blob_file_cache.GetBlobFileReader(blob_file_number, &reader).IsIOError());
+      blob_file_cache.GetBlobFileReader(read_options, blob_file_number, &reader)
+          .IsIOError());
   ASSERT_EQ(reader.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 1);
@@ -253,8 +260,10 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader_CacheFull) {
   // strict_capacity_limit is set
   CacheHandleGuard<BlobFileReader> reader;
 
-  ASSERT_TRUE(blob_file_cache.GetBlobFileReader(blob_file_number, &reader)
-                  .IsMemoryLimit());
+  const ReadOptions read_options;
+  ASSERT_TRUE(
+      blob_file_cache.GetBlobFileReader(read_options, blob_file_number, &reader)
+          .IsMemoryLimit());
   ASSERT_EQ(reader.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 1);

db/blob/blob_file_completion_callback.h

@@ -23,32 +23,19 @@ class BlobFileCompletionCallback {
       const std::vector<std::shared_ptr<EventListener>>& listeners,
       const std::string& dbname)
       : event_logger_(event_logger), listeners_(listeners), dbname_(dbname) {
-#ifndef ROCKSDB_LITE
     sst_file_manager_ = sst_file_manager;
     mutex_ = mutex;
     error_handler_ = error_handler;
-#else
-    (void)sst_file_manager;
-    (void)mutex;
-    (void)error_handler;
-#endif  // ROCKSDB_LITE
   }
 
   void OnBlobFileCreationStarted(const std::string& file_name,
                                  const std::string& column_family_name,
                                  int job_id,
                                  BlobFileCreationReason creation_reason) {
-#ifndef ROCKSDB_LITE
     // Notify the listeners.
     EventHelpers::NotifyBlobFileCreationStarted(listeners_, dbname_,
                                                 column_family_name, file_name,
                                                 job_id, creation_reason);
-#else
-    (void)file_name;
-    (void)column_family_name;
-    (void)job_id;
-    (void)creation_reason;
-#endif
   }
 
   Status OnBlobFileCompleted(const std::string& file_name,
@@ -61,7 +48,6 @@ class BlobFileCompletionCallback {
                              uint64_t blob_count, uint64_t blob_bytes) {
     Status s;
 
-#ifndef ROCKSDB_LITE
     auto sfm = static_cast<SstFileManagerImpl*>(sst_file_manager_);
     if (sfm) {
       // Report new blob files to SstFileManagerImpl
@@ -74,7 +60,6 @@ class BlobFileCompletionCallback {
         error_handler_->SetBGError(s, BackgroundErrorReason::kFlush);
       }
     }
-#endif  // !ROCKSDB_LITE
 
     // Notify the listeners.
     EventHelpers::LogAndNotifyBlobFileCreationFinished(
@@ -89,11 +74,9 @@ class BlobFileCompletionCallback {
   }
 
  private:
-#ifndef ROCKSDB_LITE
   SstFileManager* sst_file_manager_;
   InstrumentedMutex* mutex_;
   ErrorHandler* error_handler_;
-#endif  // ROCKSDB_LITE
   EventLogger* event_logger_;
   std::vector<std::shared_ptr<EventListener>> listeners_;
   std::string dbname_;

db/blob/blob_file_reader.cc

@@ -12,7 +12,7 @@
 #include "db/blob/blob_log_format.h"
 #include "file/file_prefetch_buffer.h"
 #include "file/filename.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "options/cf_options.h"
 #include "rocksdb/file_system.h"
 #include "rocksdb/slice.h"
@@ -26,9 +26,10 @@
 namespace ROCKSDB_NAMESPACE {
 
 Status BlobFileReader::Create(
-    const ImmutableOptions& immutable_options, const FileOptions& file_options,
-    uint32_t column_family_id, HistogramImpl* blob_file_read_hist,
-    uint64_t blob_file_number, const std::shared_ptr<IOTracer>& io_tracer,
+    const ImmutableOptions& immutable_options, const ReadOptions& read_options,
+    const FileOptions& file_options, uint32_t column_family_id,
+    HistogramImpl* blob_file_read_hist, uint64_t blob_file_number,
+    const std::shared_ptr<IOTracer>& io_tracer,
     std::unique_ptr<BlobFileReader>* blob_file_reader) {
   assert(blob_file_reader);
   assert(!*blob_file_reader);
@@ -52,15 +53,17 @@ Status BlobFileReader::Create(
   CompressionType compression_type = kNoCompression;
 
   {
-    const Status s = ReadHeader(file_reader.get(), column_family_id, statistics,
-                                &compression_type);
+    const Status s =
+        ReadHeader(file_reader.get(), read_options, column_family_id,
+                   statistics, &compression_type);
     if (!s.ok()) {
       return s;
     }
   }
 
   {
-    const Status s = ReadFooter(file_reader.get(), file_size, statistics);
+    const Status s =
+        ReadFooter(file_reader.get(), read_options, file_size, statistics);
     if (!s.ok()) {
       return s;
     }
@@ -134,6 +137,7 @@ Status BlobFileReader::OpenFile(
 }
 
 Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
+                                  const ReadOptions& read_options,
                                   uint32_t column_family_id,
                                   Statistics* statistics,
                                   CompressionType* compression_type) {
@@ -151,9 +155,10 @@ Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
     constexpr size_t read_size = BlobLogHeader::kSize;
 
     // TODO: rate limit reading headers from blob files.
-    const Status s = ReadFromFile(file_reader, read_offset, read_size,
-                                  statistics, &header_slice, &buf, &aligned_buf,
-                                  Env::IO_TOTAL /* rate_limiter_priority */);
+    const Status s =
+        ReadFromFile(file_reader, read_options, read_offset, read_size,
+                     statistics, &header_slice, &buf, &aligned_buf,
+                     Env::IO_TOTAL /* rate_limiter_priority */);
     if (!s.ok()) {
       return s;
     }
@@ -187,6 +192,7 @@ Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
 }
 
 Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
+                                  const ReadOptions& read_options,
                                   uint64_t file_size, Statistics* statistics) {
   assert(file_size >= BlobLogHeader::kSize + BlobLogFooter::kSize);
   assert(file_reader);
@@ -202,9 +208,10 @@ Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
     constexpr size_t read_size = BlobLogFooter::kSize;
 
     // TODO: rate limit reading footers from blob files.
-    const Status s = ReadFromFile(file_reader, read_offset, read_size,
-                                  statistics, &footer_slice, &buf, &aligned_buf,
-                                  Env::IO_TOTAL /* rate_limiter_priority */);
+    const Status s =
+        ReadFromFile(file_reader, read_options, read_offset, read_size,
+                     statistics, &footer_slice, &buf, &aligned_buf,
+                     Env::IO_TOTAL /* rate_limiter_priority */);
     if (!s.ok()) {
       return s;
     }
@@ -232,6 +239,7 @@ Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
 }
 
 Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
+                                    const ReadOptions& read_options,
                                     uint64_t read_offset, size_t read_size,
                                     Statistics* statistics, Slice* slice,
                                     Buffer* buf, AlignedBuf* aligned_buf,
@@ -246,17 +254,23 @@ Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
 
   Status s;
 
+  IOOptions io_options;
+  s = file_reader->PrepareIOOptions(read_options, io_options);
+  if (!s.ok()) {
+    return s;
+  }
+
   if (file_reader->use_direct_io()) {
     constexpr char* scratch = nullptr;
 
-    s = file_reader->Read(IOOptions(), read_offset, read_size, slice, scratch,
+    s = file_reader->Read(io_options, read_offset, read_size, slice, scratch,
                           aligned_buf, rate_limiter_priority);
   } else {
     buf->reset(new char[read_size]);
     constexpr AlignedBuf* aligned_scratch = nullptr;
 
-    s = file_reader->Read(IOOptions(), read_offset, read_size, slice,
-                          buf->get(), aligned_scratch, rate_limiter_priority);
+    s = file_reader->Read(io_options, read_offset, read_size, slice, buf->get(),
+                          aligned_scratch, rate_limiter_priority);
   }
 
   if (!s.ok()) {
@@ -324,8 +338,13 @@ Status BlobFileReader::GetBlob(
     Status s;
     constexpr bool for_compaction = true;
 
+    IOOptions io_options;
+    s = file_reader_->PrepareIOOptions(read_options, io_options);
+    if (!s.ok()) {
+      return s;
+    }
     prefetched = prefetch_buffer->TryReadFromCache(
-        IOOptions(), file_reader_.get(), record_offset,
+        io_options, file_reader_.get(), record_offset,
         static_cast<size_t>(record_size), &record_slice, &s,
         read_options.rate_limiter_priority, for_compaction);
     if (!s.ok()) {
@@ -338,10 +357,10 @@ Status BlobFileReader::GetBlob(
     PERF_COUNTER_ADD(blob_read_count, 1);
     PERF_COUNTER_ADD(blob_read_byte, record_size);
     PERF_TIMER_GUARD(blob_read_time);
-    const Status s = ReadFromFile(file_reader_.get(), record_offset,
-                                  static_cast<size_t>(record_size), statistics_,
-                                  &record_slice, &buf, &aligned_buf,
-                                  read_options.rate_limiter_priority);
+    const Status s = ReadFromFile(
+        file_reader_.get(), read_options, record_offset,
+        static_cast<size_t>(record_size), statistics_, &record_slice, &buf,
+        &aligned_buf, read_options.rate_limiter_priority);
     if (!s.ok()) {
       return s;
     }
@@ -569,12 +588,7 @@ Status BlobFileReader::UncompressBlobIfNeeded(
   assert(result);
 
   if (compression_type == kNoCompression) {
-    CacheAllocationPtr allocation =
-        AllocateBlock(value_slice.size(), allocator);
-    memcpy(allocation.get(), value_slice.data(), value_slice.size());
-
-    *result = BlobContents::Create(std::move(allocation), value_slice.size());
-
+    BlobContentsCreator::Create(result, nullptr, value_slice, allocator);
     return Status::OK();
   }
 
@@ -602,7 +616,7 @@ Status BlobFileReader::UncompressBlobIfNeeded(
     return Status::Corruption("Unable to uncompress blob");
   }
 
-  *result = BlobContents::Create(std::move(output), uncompressed_size);
+  result->reset(new BlobContents(std::move(output), uncompressed_size));
 
   return Status::OK();
 }

db/blob/blob_file_reader.h

@@ -29,6 +29,7 @@ class Statistics;
 class BlobFileReader {
  public:
   static Status Create(const ImmutableOptions& immutable_options,
+                       const ReadOptions& read_options,
                        const FileOptions& file_options,
                        uint32_t column_family_id,
                        HistogramImpl* blob_file_read_hist,
@@ -74,15 +75,18 @@ class BlobFileReader {
                          std::unique_ptr<RandomAccessFileReader>* file_reader);
 
   static Status ReadHeader(const RandomAccessFileReader* file_reader,
+                           const ReadOptions& read_options,
                            uint32_t column_family_id, Statistics* statistics,
                            CompressionType* compression_type);
 
   static Status ReadFooter(const RandomAccessFileReader* file_reader,
-                           uint64_t file_size, Statistics* statistics);
+                           const ReadOptions& read_options, uint64_t file_size,
+                           Statistics* statistics);
 
   using Buffer = std::unique_ptr<char[]>;
 
   static Status ReadFromFile(const RandomAccessFileReader* file_reader,
+                             const ReadOptions& read_options,
                              uint64_t read_offset, size_t read_size,
                              Statistics* statistics, Slice* slice, Buffer* buf,
                              AlignedBuf* aligned_buf,

db/blob/blob_file_reader_test.cc

@@ -172,12 +172,12 @@ TEST_F(BlobFileReaderTest, CreateReaderAndGetBlob) {
 
   std::unique_ptr<BlobFileReader> reader;
 
+  ReadOptions read_options;
   ASSERT_OK(BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader));
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader));
 
   // Make sure the blob can be retrieved with and without checksum verification
-  ReadOptions read_options;
   read_options.verify_checksums = false;
 
   constexpr FilePrefetchBuffer* prefetch_buffer = nullptr;
@@ -479,11 +479,11 @@ TEST_F(BlobFileReaderTest, Malformed) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     column_family_id, blob_file_read_hist,
-                                     blob_file_number, nullptr /*IOTracer*/,
-                                     &reader)
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), column_family_id,
+                                     blob_file_read_hist, blob_file_number,
+                                     nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
 }
 
@@ -513,11 +513,11 @@ TEST_F(BlobFileReaderTest, TTL) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     column_family_id, blob_file_read_hist,
-                                     blob_file_number, nullptr /*IOTracer*/,
-                                     &reader)
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), column_family_id,
+                                     blob_file_read_hist, blob_file_number,
+                                     nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
 }
 
@@ -552,11 +552,11 @@ TEST_F(BlobFileReaderTest, ExpirationRangeInHeader) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     column_family_id, blob_file_read_hist,
-                                     blob_file_number, nullptr /*IOTracer*/,
-                                     &reader)
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), column_family_id,
+                                     blob_file_read_hist, blob_file_number,
+                                     nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
 }
 
@@ -591,11 +591,11 @@ TEST_F(BlobFileReaderTest, ExpirationRangeInFooter) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     column_family_id, blob_file_read_hist,
-                                     blob_file_number, nullptr /*IOTracer*/,
-                                     &reader)
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), column_family_id,
+                                     blob_file_read_hist, blob_file_number,
+                                     nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
 }
 
@@ -629,9 +629,9 @@ TEST_F(BlobFileReaderTest, IncorrectColumnFamily) {
   std::unique_ptr<BlobFileReader> reader;
 
   constexpr uint32_t incorrect_column_family_id = 2;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     incorrect_column_family_id,
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), incorrect_column_family_id,
                                      blob_file_read_hist, blob_file_number,
                                      nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
@@ -664,10 +664,10 @@ TEST_F(BlobFileReaderTest, BlobCRCError) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  const ReadOptions read_options;
   ASSERT_OK(BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader));
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader));
 
   SyncPoint::GetInstance()->SetCallBack(
       "BlobFileReader::VerifyBlob:CheckBlobCRC", [](void* arg) {
@@ -728,13 +728,12 @@ TEST_F(BlobFileReaderTest, Compression) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  ReadOptions read_options;
   ASSERT_OK(BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader));
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader));
 
   // Make sure the blob can be retrieved with and without checksum verification
-  ReadOptions read_options;
   read_options.verify_checksums = false;
 
   constexpr FilePrefetchBuffer* prefetch_buffer = nullptr;
@@ -803,10 +802,10 @@ TEST_F(BlobFileReaderTest, UncompressionError) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  const ReadOptions read_options;
   ASSERT_OK(BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader));
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader));
 
   SyncPoint::GetInstance()->SetCallBack(
       "BlobFileReader::UncompressBlobIfNeeded:TamperWithResult", [](void* arg) {
@@ -895,10 +894,10 @@ TEST_P(BlobFileReaderIOErrorTest, IOError) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  const ReadOptions read_options;
   const Status s = BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader);
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader);
 
   const bool fail_during_create =
       (sync_point_ != "BlobFileReader::GetBlob:ReadFromFile");
@@ -983,10 +982,10 @@ TEST_P(BlobFileReaderDecodingErrorTest, DecodingError) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  const ReadOptions read_options;
   const Status s = BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader);
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader);
 
   const bool fail_during_create =
       sync_point_ != "BlobFileReader::GetBlob:TamperWithResult";

db/blob/blob_log_sequential_reader.cc

@@ -7,7 +7,7 @@
 #include "db/blob/blob_log_sequential_reader.h"
 
 #include "file/random_access_file_reader.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "util/stop_watch.h"
 
 namespace ROCKSDB_NAMESPACE {

db/blob/blob_log_writer.cc

@@ -10,7 +10,7 @@
 
 #include "db/blob/blob_log_format.h"
 #include "file/writable_file_writer.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "rocksdb/system_clock.h"
 #include "test_util/sync_point.h"
 #include "util/coding.h"

db/blob/blob_source.cc

@@ -13,7 +13,7 @@
 #include "db/blob/blob_contents.h"
 #include "db/blob/blob_file_reader.h"
 #include "db/blob/blob_log_format.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "options/cf_options.h"
 #include "table/get_context.h"
 #include "table/multiget_context.h"
@@ -30,16 +30,14 @@ BlobSource::BlobSource(const ImmutableOptions* immutable_options,
       blob_file_cache_(blob_file_cache),
       blob_cache_(immutable_options->blob_cache),
       lowest_used_cache_tier_(immutable_options->lowest_used_cache_tier) {
-#ifndef ROCKSDB_LITE
   auto bbto =
       immutable_options->table_factory->GetOptions<BlockBasedTableOptions>();
   if (bbto &&
       bbto->cache_usage_options.options_overrides.at(CacheEntryRole::kBlobCache)
               .charged == CacheEntryRoleOptions::Decision::kEnabled) {
-    blob_cache_ = std::make_shared<ChargedCache>(immutable_options->blob_cache,
-                                                 bbto->block_cache);
+    blob_cache_ = SharedCacheInterface{std::make_shared<ChargedCache>(
+        immutable_options->blob_cache, bbto->block_cache)};
   }
-#endif  // ROCKSDB_LITE
 }
 
 BlobSource::~BlobSource() = default;
@@ -82,9 +80,8 @@ Status BlobSource::PutBlobIntoCache(
   assert(cached_blob);
   assert(cached_blob->IsEmpty());
 
-  Cache::Handle* cache_handle = nullptr;
+  TypedHandle* cache_handle = nullptr;
   const Status s = InsertEntryIntoCache(cache_key, blob->get(),
-                                        (*blob)->ApproximateMemoryUsage(),
                                         &cache_handle, Cache::Priority::BOTTOM);
   if (s.ok()) {
     blob->release();
@@ -106,26 +103,10 @@ Status BlobSource::PutBlobIntoCache(
   return s;
 }
 
-Cache::Handle* BlobSource::GetEntryFromCache(const Slice& key) const {
-  Cache::Handle* cache_handle = nullptr;
-
-  if (lowest_used_cache_tier_ == CacheTier::kNonVolatileBlockTier) {
-    Cache::CreateCallback create_cb =
-        [allocator = blob_cache_->memory_allocator()](
-            const void* buf, size_t size, void** out_obj,
-            size_t* charge) -> Status {
-      return BlobContents::CreateCallback(AllocateBlock(size, allocator), buf,
-                                          size, out_obj, charge);
-    };
-
-    cache_handle = blob_cache_->Lookup(key, BlobContents::GetCacheItemHelper(),
-                                       create_cb, Cache::Priority::BOTTOM,
-                                       true /* wait_for_cache */, statistics_);
-  } else {
-    cache_handle = blob_cache_->Lookup(key, statistics_);
-  }
-
-  return cache_handle;
+BlobSource::TypedHandle* BlobSource::GetEntryFromCache(const Slice& key) const {
+  return blob_cache_.LookupFull(key, nullptr /* context */,
+                                Cache::Priority::BOTTOM, statistics_,
+                                lowest_used_cache_tier_);
 }
 
 void BlobSource::PinCachedBlob(CacheHandleGuard<BlobContents>* cached_blob,
@@ -166,24 +147,11 @@ void BlobSource::PinOwnedBlob(std::unique_ptr<BlobContents>* owned_blob,
 }
 
 Status BlobSource::InsertEntryIntoCache(const Slice& key, BlobContents* value,
-                                        size_t charge,
-                                        Cache::Handle** cache_handle,
+                                        TypedHandle** cache_handle,
                                         Cache::Priority priority) const {
-  Status s;
-
-  Cache::CacheItemHelper* const cache_item_helper =
-      BlobContents::GetCacheItemHelper();
-  assert(cache_item_helper);
-
-  if (lowest_used_cache_tier_ == CacheTier::kNonVolatileBlockTier) {
-    s = blob_cache_->Insert(key, value, cache_item_helper, charge, cache_handle,
-                            priority);
-  } else {
-    s = blob_cache_->Insert(key, value, charge, cache_item_helper->del_cb,
-                            cache_handle, priority);
-  }
-
-  return s;
+  return blob_cache_.InsertFull(key, value, value->ApproximateMemoryUsage(),
+                                cache_handle, priority,
+                                lowest_used_cache_tier_);
 }
 
 Status BlobSource::GetBlob(const ReadOptions& read_options,
@@ -241,7 +209,8 @@ Status BlobSource::GetBlob(const ReadOptions& read_options,
 
   {
     CacheHandleGuard<BlobFileReader> blob_file_reader;
-    s = blob_file_cache_->GetBlobFileReader(file_number, &blob_file_reader);
+    s = blob_file_cache_->GetBlobFileReader(read_options, file_number,
+                                            &blob_file_reader);
     if (!s.ok()) {
       return s;
     }
@@ -252,9 +221,10 @@ Status BlobSource::GetBlob(const ReadOptions& read_options,
       return Status::Corruption("Compression type mismatch when reading blob");
     }
 
-    MemoryAllocator* const allocator = (blob_cache_ && read_options.fill_cache)
-                                           ? blob_cache_->memory_allocator()
-                                           : nullptr;
+    MemoryAllocator* const allocator =
+        (blob_cache_ && read_options.fill_cache)
+            ? blob_cache_.get()->memory_allocator()
+            : nullptr;
 
     uint64_t read_size = 0;
     s = blob_file_reader.GetValue()->GetBlob(
@@ -403,8 +373,8 @@ void BlobSource::MultiGetBlobFromOneFile(const ReadOptions& read_options,
     }
 
     CacheHandleGuard<BlobFileReader> blob_file_reader;
-    Status s =
-        blob_file_cache_->GetBlobFileReader(file_number, &blob_file_reader);
+    Status s = blob_file_cache_->GetBlobFileReader(read_options, file_number,
+                                                   &blob_file_reader);
     if (!s.ok()) {
       for (size_t i = 0; i < _blob_reqs.size(); ++i) {
         BlobReadRequest* const req = _blob_reqs[i].first;
@@ -418,9 +388,10 @@ void BlobSource::MultiGetBlobFromOneFile(const ReadOptions& read_options,
 
     assert(blob_file_reader.GetValue());
 
-    MemoryAllocator* const allocator = (blob_cache_ && read_options.fill_cache)
-                                           ? blob_cache_->memory_allocator()
-                                           : nullptr;
+    MemoryAllocator* const allocator =
+        (blob_cache_ && read_options.fill_cache)
+            ? blob_cache_.get()->memory_allocator()
+            : nullptr;
 
     blob_file_reader.GetValue()->MultiGetBlob(read_options, allocator,
                                               _blob_reqs, &_bytes_read);

db/blob/blob_source.h

@@ -8,8 +8,9 @@
 #include <cinttypes>
 #include <memory>
 
-#include "cache/cache_helpers.h"
 #include "cache/cache_key.h"
+#include "cache/typed_cache.h"
+#include "db/blob/blob_contents.h"
 #include "db/blob/blob_file_cache.h"
 #include "db/blob/blob_read_request.h"
 #include "rocksdb/cache.h"
@@ -23,7 +24,6 @@ struct ImmutableOptions;
 class Status;
 class FilePrefetchBuffer;
 class Slice;
-class BlobContents;
 
 // BlobSource is a class that provides universal access to blobs, regardless of
 // whether they are in the blob cache, secondary cache, or (remote) storage.
@@ -95,9 +95,9 @@ class BlobSource {
                                uint64_t* bytes_read);
 
   inline Status GetBlobFileReader(
-      uint64_t blob_file_number,
+      const ReadOptions& read_options, uint64_t blob_file_number,
       CacheHandleGuard<BlobFileReader>* blob_file_reader) {
-    return blob_file_cache_->GetBlobFileReader(blob_file_number,
+    return blob_file_cache_->GetBlobFileReader(read_options, blob_file_number,
                                                blob_file_reader);
   }
 
@@ -106,6 +106,14 @@ class BlobSource {
   bool TEST_BlobInCache(uint64_t file_number, uint64_t file_size,
                         uint64_t offset, size_t* charge = nullptr) const;
 
+  // For TypedSharedCacheInterface
+  void Create(BlobContents** out, const char* buf, size_t size,
+              MemoryAllocator* alloc);
+
+  using SharedCacheInterface =
+      FullTypedSharedCacheInterface<BlobContents, BlobContentsCreator>;
+  using TypedHandle = SharedCacheInterface::TypedHandle;
+
  private:
   Status GetBlobFromCache(const Slice& cache_key,
                           CacheHandleGuard<BlobContents>* cached_blob) const;
@@ -120,10 +128,10 @@ class BlobSource {
   static void PinOwnedBlob(std::unique_ptr<BlobContents>* owned_blob,
                            PinnableSlice* value);
 
-  Cache::Handle* GetEntryFromCache(const Slice& key) const;
+  TypedHandle* GetEntryFromCache(const Slice& key) const;
 
   Status InsertEntryIntoCache(const Slice& key, BlobContents* value,
-                              size_t charge, Cache::Handle** cache_handle,
+                              TypedHandle** cache_handle,
                               Cache::Priority priority) const;
 
   inline CacheKey GetCacheKey(uint64_t file_number, uint64_t /*file_size*/,
@@ -141,7 +149,7 @@ class BlobSource {
   BlobFileCache* blob_file_cache_;
 
   // A cache to store uncompressed blobs.
-  std::shared_ptr<Cache> blob_cache_;
+  mutable SharedCacheInterface blob_cache_;
 
   // The control option of how the cache tiers will be used. Currently rocksdb
   // support block/blob cache (volatile tier) and secondary cache (this tier

db/blob/blob_source_test.cc

@@ -517,7 +517,8 @@ TEST_F(BlobSourceTest, GetCompressedBlobs) {
                   compression, blob_offsets, blob_sizes);
 
     CacheHandleGuard<BlobFileReader> blob_file_reader;
-    ASSERT_OK(blob_source.GetBlobFileReader(file_number, &blob_file_reader));
+    ASSERT_OK(blob_source.GetBlobFileReader(read_options, file_number,
+                                            &blob_file_reader));
     ASSERT_NE(blob_file_reader.GetValue(), nullptr);
 
     const uint64_t file_size = blob_file_reader.GetValue()->GetFileSize();
@@ -1139,26 +1140,18 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
                          blob_file_cache.get());
 
   CacheHandleGuard<BlobFileReader> file_reader;
-  ASSERT_OK(blob_source.GetBlobFileReader(file_number, &file_reader));
+  ReadOptions read_options;
+  ASSERT_OK(
+      blob_source.GetBlobFileReader(read_options, file_number, &file_reader));
   ASSERT_NE(file_reader.GetValue(), nullptr);
   const uint64_t file_size = file_reader.GetValue()->GetFileSize();
   ASSERT_EQ(file_reader.GetValue()->GetCompressionType(), kNoCompression);
 
-  ReadOptions read_options;
   read_options.verify_checksums = true;
 
   auto blob_cache = options_.blob_cache;
   auto secondary_cache = lru_cache_opts_.secondary_cache;
 
-  Cache::CreateCallback create_cb = [](const void* buf, size_t size,
-                                       void** out_obj,
-                                       size_t* charge) -> Status {
-    CacheAllocationPtr allocation(new char[size]);
-
-    return BlobContents::CreateCallback(std::move(allocation), buf, size,
-                                        out_obj, charge);
-  };
-
   {
     // GetBlob
     std::vector<PinnableSlice> values(keys.size());
@@ -1219,15 +1212,16 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
     {
       CacheKey cache_key = base_cache_key.WithOffset(blob_offsets[0]);
       const Slice key0 = cache_key.AsSlice();
-      auto handle0 = blob_cache->Lookup(key0, statistics);
+      auto handle0 = blob_cache->BasicLookup(key0, statistics);
       ASSERT_EQ(handle0, nullptr);
 
       // key0's item should be in the secondary cache.
-      bool is_in_sec_cache = false;
-      auto sec_handle0 =
-          secondary_cache->Lookup(key0, create_cb, true,
-                                  /*advise_erase=*/true, is_in_sec_cache);
-      ASSERT_FALSE(is_in_sec_cache);
+      bool kept_in_sec_cache = false;
+      auto sec_handle0 = secondary_cache->Lookup(
+          key0, BlobSource::SharedCacheInterface::GetFullHelper(),
+          /*context*/ nullptr, true,
+          /*advise_erase=*/true, kept_in_sec_cache);
+      ASSERT_FALSE(kept_in_sec_cache);
       ASSERT_NE(sec_handle0, nullptr);
       ASSERT_TRUE(sec_handle0->IsReady());
       auto value = static_cast<BlobContents*>(sec_handle0->Value());
@@ -1246,15 +1240,16 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
     {
       CacheKey cache_key = base_cache_key.WithOffset(blob_offsets[1]);
       const Slice key1 = cache_key.AsSlice();
-      auto handle1 = blob_cache->Lookup(key1, statistics);
+      auto handle1 = blob_cache->BasicLookup(key1, statistics);
       ASSERT_NE(handle1, nullptr);
       blob_cache->Release(handle1);
 
-      bool is_in_sec_cache = false;
-      auto sec_handle1 =
-          secondary_cache->Lookup(key1, create_cb, true,
-                                  /*advise_erase=*/true, is_in_sec_cache);
-      ASSERT_FALSE(is_in_sec_cache);
+      bool kept_in_sec_cache = false;
+      auto sec_handle1 = secondary_cache->Lookup(
+          key1, BlobSource::SharedCacheInterface::GetFullHelper(),
+          /*context*/ nullptr, true,
+          /*advise_erase=*/true, kept_in_sec_cache);
+      ASSERT_FALSE(kept_in_sec_cache);
       ASSERT_EQ(sec_handle1, nullptr);
 
       ASSERT_TRUE(blob_source.TEST_BlobInCache(file_number, file_size,
@@ -1276,7 +1271,7 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
       // key0 should be in the primary cache.
       CacheKey cache_key0 = base_cache_key.WithOffset(blob_offsets[0]);
       const Slice key0 = cache_key0.AsSlice();
-      auto handle0 = blob_cache->Lookup(key0, statistics);
+      auto handle0 = blob_cache->BasicLookup(key0, statistics);
       ASSERT_NE(handle0, nullptr);
       auto value = static_cast<BlobContents*>(blob_cache->Value(handle0));
       ASSERT_NE(value, nullptr);
@@ -1286,12 +1281,12 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
       // key1 is not in the primary cache and is in the secondary cache.
       CacheKey cache_key1 = base_cache_key.WithOffset(blob_offsets[1]);
       const Slice key1 = cache_key1.AsSlice();
-      auto handle1 = blob_cache->Lookup(key1, statistics);
+      auto handle1 = blob_cache->BasicLookup(key1, statistics);
       ASSERT_EQ(handle1, nullptr);
 
       // erase key0 from the primary cache.
       blob_cache->Erase(key0);
-      handle0 = blob_cache->Lookup(key0, statistics);
+      handle0 = blob_cache->BasicLookup(key0, statistics);
       ASSERT_EQ(handle0, nullptr);
 
       // key1 promotion should succeed due to the primary cache being empty. we
@@ -1307,7 +1302,7 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
       // in the secondary cache. So, the primary cache's Lookup() without
       // secondary cache support cannot see it. (NOTE: The dummy handle used
       // to be a leaky abstraction but not anymore.)
-      handle1 = blob_cache->Lookup(key1, statistics);
+      handle1 = blob_cache->BasicLookup(key1, statistics);
       ASSERT_EQ(handle1, nullptr);
 
       // But after another access, it is promoted to primary cache
@@ -1315,7 +1310,7 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
                                                blob_offsets[1]));
 
       // And Lookup() can find it (without secondary cache support)
-      handle1 = blob_cache->Lookup(key1, statistics);
+      handle1 = blob_cache->BasicLookup(key1, statistics);
       ASSERT_NE(handle1, nullptr);
       ASSERT_NE(blob_cache->Value(handle1), nullptr);
       blob_cache->Release(handle1);
@@ -1379,7 +1374,7 @@ class BlobSourceCacheReservationTest : public DBTestBase {
 
   static constexpr std::size_t kSizeDummyEntry = CacheReservationManagerImpl<
       CacheEntryRole::kBlobCache>::GetDummyEntrySize();
-  static constexpr std::size_t kCacheCapacity = 1 * kSizeDummyEntry;
+  static constexpr std::size_t kCacheCapacity = 2 * kSizeDummyEntry;
   static constexpr int kNumShardBits = 0;  // 2^0 shard
 
   static constexpr uint32_t kColumnFamilyId = 1;
@@ -1398,7 +1393,6 @@ class BlobSourceCacheReservationTest : public DBTestBase {
   std::string db_session_id_;
 };
 
-#ifndef ROCKSDB_LITE
 TEST_F(BlobSourceCacheReservationTest, SimpleCacheReservation) {
   options_.cf_paths.emplace_back(
       test::PerThreadDBPath(
@@ -1513,11 +1507,10 @@ TEST_F(BlobSourceCacheReservationTest, SimpleCacheReservation) {
   }
 }
 
-TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservationOnFullCache) {
+TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservation) {
   options_.cf_paths.emplace_back(
       test::PerThreadDBPath(
-          env_,
-          "BlobSourceCacheReservationTest_IncreaseCacheReservationOnFullCache"),
+          env_, "BlobSourceCacheReservationTest_IncreaseCacheReservation"),
       0);
 
   GenerateKeysAndBlobs();
@@ -1525,7 +1518,7 @@ TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservationOnFullCache) {
   DestroyAndReopen(options_);
 
   ImmutableOptions immutable_options(options_);
-  constexpr size_t blob_size = kSizeDummyEntry / (kNumBlobs / 2);
+  constexpr size_t blob_size = 24 << 10;  // 24KB
   for (size_t i = 0; i < kNumBlobs; ++i) {
     blob_file_size_ -= blobs_[i].size();  // old blob size
     blob_strs_[i].resize(blob_size, '@');
@@ -1583,11 +1576,6 @@ TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservationOnFullCache) {
 
     std::vector<PinnableSlice> values(keys_.size());
 
-    // Since we resized each blob to be kSizeDummyEntry / (num_blobs / 2), we
-    // can't fit all the blobs in the cache at the same time, which means we
-    // should observe cache evictions once we reach the cache's capacity.
-    // Due to the overhead of the cache and the BlobContents objects, as well as
-    // jemalloc bin sizes, this happens after inserting seven blobs.
     uint64_t blob_bytes = 0;
     for (size_t i = 0; i < kNumBlobs; ++i) {
       ASSERT_OK(blob_source.GetBlob(
@@ -1598,22 +1586,21 @@ TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservationOnFullCache) {
       // Release cache handle
       values[i].Reset();
 
-      if (i < kNumBlobs / 2 - 1) {
-        size_t charge = 0;
-        ASSERT_TRUE(blob_source.TEST_BlobInCache(
-            kBlobFileNumber, blob_file_size_, blob_offsets[i], &charge));
+      size_t charge = 0;
+      ASSERT_TRUE(blob_source.TEST_BlobInCache(kBlobFileNumber, blob_file_size_,
+                                               blob_offsets[i], &charge));
 
-        blob_bytes += charge;
-      }
+      blob_bytes += charge;
 
-      ASSERT_EQ(cache_res_mgr->GetTotalReservedCacheSize(), kSizeDummyEntry);
+      ASSERT_EQ(cache_res_mgr->GetTotalReservedCacheSize(),
+                (blob_bytes <= kSizeDummyEntry) ? kSizeDummyEntry
+                                                : (2 * kSizeDummyEntry));
       ASSERT_EQ(cache_res_mgr->GetTotalMemoryUsed(), blob_bytes);
       ASSERT_EQ(cache_res_mgr->GetTotalMemoryUsed(),
                 options_.blob_cache->GetUsage());
     }
   }
 }
-#endif  // ROCKSDB_LITE
 
 }  // namespace ROCKSDB_NAMESPACE
 

db/blob/db_blob_basic_test.cc

@@ -11,6 +11,7 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/blob_log_format.h"
 #include "db/db_test_util.h"
+#include "db/db_with_timestamp_test_util.h"
 #include "port/stack_trace.h"
 #include "test_util/sync_point.h"
 #include "utilities/fault_injection_env.h"
@@ -584,7 +585,6 @@ TEST_F(DBBlobBasicTest, MultiGetBlobsFromCache) {
   }
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobBasicTest, MultiGetWithDirectIO) {
   Options options = GetDefaultOptions();
 
@@ -773,7 +773,6 @@ TEST_F(DBBlobBasicTest, MultiGetWithDirectIO) {
     ASSERT_EQ(values[2], second_blob);
   }
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_F(DBBlobBasicTest, MultiGetBlobsFromMultipleFiles) {
   Options options = GetDefaultOptions();
@@ -1062,7 +1061,6 @@ TEST_F(DBBlobBasicTest, GetBlob_IndexWithInvalidFileNumber) {
                   .IsCorruption());
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobBasicTest, GenerateIOTracing) {
   Options options = GetDefaultOptions();
   options.enable_blob_files = true;
@@ -1117,7 +1115,6 @@ TEST_F(DBBlobBasicTest, GenerateIOTracing) {
     ASSERT_GT(blob_files_op_count, 2);
   }
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_F(DBBlobBasicTest, BestEffortsRecovery_MissingNewestBlobFile) {
   Options options = GetDefaultOptions();
@@ -1219,7 +1216,6 @@ TEST_F(DBBlobBasicTest, MultiGetMergeBlobWithPut) {
   ASSERT_EQ(values[2], "v2_0");
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobBasicTest, Properties) {
   Options options = GetDefaultOptions();
   options.enable_blob_files = true;
@@ -1382,7 +1378,6 @@ TEST_F(DBBlobBasicTest, PropertiesMultiVersion) {
                  BlobLogRecord::CalculateAdjustmentForRecordHeader(key_size) +
                  blob_size + BlobLogFooter::kSize));
 }
-#endif  // !ROCKSDB_LITE
 
 class DBBlobBasicIOErrorTest : public DBBlobBasicTest,
                                public testing::WithParamInterface<std::string> {
@@ -1632,7 +1627,6 @@ TEST_F(DBBlobBasicTest, WarmCacheWithBlobsDuringFlush) {
             options.statistics->getTickerCount(BLOB_DB_CACHE_ADD));
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobBasicTest, DynamicallyWarmCacheDuringFlush) {
   Options options = GetDefaultOptions();
 
@@ -1700,7 +1694,6 @@ TEST_F(DBBlobBasicTest, DynamicallyWarmCacheDuringFlush) {
                               /*end=*/nullptr));
   EXPECT_EQ(0, options.statistics->getTickerCount(BLOB_DB_CACHE_ADD));
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_F(DBBlobBasicTest, WarmCacheWithBlobsSecondary) {
   CompressedSecondaryCacheOptions secondary_cache_opts;
@@ -1779,6 +1772,461 @@ TEST_F(DBBlobBasicTest, WarmCacheWithBlobsSecondary) {
             1);
 }
 
+TEST_F(DBBlobBasicTest, GetEntityBlob) {
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+
+  Reopen(options);
+
+  constexpr char key[] = "key";
+  constexpr char blob_value[] = "blob_value";
+
+  constexpr char other_key[] = "other_key";
+  constexpr char other_blob_value[] = "other_blob_value";
+
+  ASSERT_OK(Put(key, blob_value));
+  ASSERT_OK(Put(other_key, other_blob_value));
+
+  ASSERT_OK(Flush());
+
+  WideColumns expected_columns{{kDefaultWideColumnName, blob_value}};
+  WideColumns other_expected_columns{
+      {kDefaultWideColumnName, other_blob_value}};
+
+  {
+    PinnableWideColumns result;
+    ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(), key,
+                             &result));
+    ASSERT_EQ(result.columns(), expected_columns);
+  }
+
+  {
+    PinnableWideColumns result;
+    ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                             other_key, &result));
+
+    ASSERT_EQ(result.columns(), other_expected_columns);
+  }
+
+  {
+    constexpr size_t num_keys = 2;
+
+    std::array<Slice, num_keys> keys{{key, other_key}};
+    std::array<PinnableWideColumns, num_keys> results;
+    std::array<Status, num_keys> statuses;
+
+    db_->MultiGetEntity(ReadOptions(), db_->DefaultColumnFamily(), num_keys,
+                        &keys[0], &results[0], &statuses[0]);
+
+    ASSERT_OK(statuses[0]);
+    ASSERT_EQ(results[0].columns(), expected_columns);
+
+    ASSERT_OK(statuses[1]);
+    ASSERT_EQ(results[1].columns(), other_expected_columns);
+  }
+}
+
+class DBBlobWithTimestampTest : public DBBasicTestWithTimestampBase {
+ protected:
+  DBBlobWithTimestampTest()
+      : DBBasicTestWithTimestampBase("db_blob_with_timestamp_test") {}
+};
+
+TEST_F(DBBlobWithTimestampTest, GetBlob) {
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+  WriteOptions write_opts;
+  const std::string ts = Timestamp(1, 0);
+  constexpr char key[] = "key";
+  constexpr char blob_value[] = "blob_value";
+
+  ASSERT_OK(db_->Put(write_opts, key, ts, blob_value));
+
+  ASSERT_OK(Flush());
+
+  const std::string read_ts = Timestamp(2, 0);
+  Slice read_ts_slice(read_ts);
+  ReadOptions read_opts;
+  read_opts.timestamp = &read_ts_slice;
+  std::string value;
+  ASSERT_OK(db_->Get(read_opts, key, &value));
+  ASSERT_EQ(value, blob_value);
+}
+
+TEST_F(DBBlobWithTimestampTest, MultiGetBlobs) {
+  constexpr size_t min_blob_size = 6;
+
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+  options.min_blob_size = min_blob_size;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  // Put then retrieve three key-values. The first value is below the size limit
+  // and is thus stored inline; the other two are stored separately as blobs.
+  constexpr size_t num_keys = 3;
+
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "short";
+  static_assert(sizeof(first_value) - 1 < min_blob_size,
+                "first_value too long to be inlined");
+
+  DestroyAndReopen(options);
+  WriteOptions write_opts;
+  const std::string ts = Timestamp(1, 0);
+  ASSERT_OK(db_->Put(write_opts, first_key, ts, first_value));
+
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "long_value";
+  static_assert(sizeof(second_value) - 1 >= min_blob_size,
+                "second_value too short to be stored as blob");
+
+  ASSERT_OK(db_->Put(write_opts, second_key, ts, second_value));
+
+  constexpr char third_key[] = "third_key";
+  constexpr char third_value[] = "other_long_value";
+  static_assert(sizeof(third_value) - 1 >= min_blob_size,
+                "third_value too short to be stored as blob");
+
+  ASSERT_OK(db_->Put(write_opts, third_key, ts, third_value));
+
+  ASSERT_OK(Flush());
+
+  ReadOptions read_options;
+  const std::string read_ts = Timestamp(2, 0);
+  Slice read_ts_slice(read_ts);
+  read_options.timestamp = &read_ts_slice;
+  std::array<Slice, num_keys> keys{{first_key, second_key, third_key}};
+
+  {
+    std::array<PinnableSlice, num_keys> values;
+    std::array<Status, num_keys> statuses;
+
+    db_->MultiGet(read_options, db_->DefaultColumnFamily(), num_keys, &keys[0],
+                  &values[0], &statuses[0]);
+
+    ASSERT_OK(statuses[0]);
+    ASSERT_EQ(values[0], first_value);
+
+    ASSERT_OK(statuses[1]);
+    ASSERT_EQ(values[1], second_value);
+
+    ASSERT_OK(statuses[2]);
+    ASSERT_EQ(values[2], third_value);
+  }
+}
+
+TEST_F(DBBlobWithTimestampTest, GetMergeBlobWithPut) {
+  Options options = GetDefaultOptions();
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  WriteOptions write_opts;
+  const std::string ts = Timestamp(1, 0);
+  ASSERT_OK(db_->Put(write_opts, "Key1", ts, "v1"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key1", ts, "v2"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key1", ts, "v3"));
+  ASSERT_OK(Flush());
+
+  std::string value;
+  const std::string read_ts = Timestamp(2, 0);
+  Slice read_ts_slice(read_ts);
+  ReadOptions read_opts;
+  read_opts.timestamp = &read_ts_slice;
+  ASSERT_OK(db_->Get(read_opts, "Key1", &value));
+  ASSERT_EQ(value, "v1,v2,v3");
+}
+
+TEST_F(DBBlobWithTimestampTest, MultiGetMergeBlobWithPut) {
+  constexpr size_t num_keys = 3;
+
+  Options options = GetDefaultOptions();
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  WriteOptions write_opts;
+  const std::string ts = Timestamp(1, 0);
+
+  ASSERT_OK(db_->Put(write_opts, "Key0", ts, "v0_0"));
+  ASSERT_OK(db_->Put(write_opts, "Key1", ts, "v1_0"));
+  ASSERT_OK(db_->Put(write_opts, "Key2", ts, "v2_0"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key0", ts, "v0_1"));
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key1", ts, "v1_1"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key0", ts, "v0_2"));
+  ASSERT_OK(Flush());
+
+  const std::string read_ts = Timestamp(2, 0);
+  Slice read_ts_slice(read_ts);
+  ReadOptions read_opts;
+  read_opts.timestamp = &read_ts_slice;
+  std::array<Slice, num_keys> keys{{"Key0", "Key1", "Key2"}};
+  std::array<PinnableSlice, num_keys> values;
+  std::array<Status, num_keys> statuses;
+
+  db_->MultiGet(read_opts, db_->DefaultColumnFamily(), num_keys, &keys[0],
+                &values[0], &statuses[0]);
+
+  ASSERT_OK(statuses[0]);
+  ASSERT_EQ(values[0], "v0_0,v0_1,v0_2");
+
+  ASSERT_OK(statuses[1]);
+  ASSERT_EQ(values[1], "v1_0,v1_1");
+
+  ASSERT_OK(statuses[2]);
+  ASSERT_EQ(values[2], "v2_0");
+}
+
+TEST_F(DBBlobWithTimestampTest, IterateBlobs) {
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  int num_blobs = 5;
+  std::vector<std::string> keys;
+  std::vector<std::string> blobs;
+
+  WriteOptions write_opts;
+  std::vector<std::string> write_timestamps = {Timestamp(1, 0),
+                                               Timestamp(2, 0)};
+
+  // For each key in ["key0", ... "keyi", ...], write two versions:
+  // Timestamp(1, 0), "blobi0"
+  // Timestamp(2, 0), "blobi1"
+  for (int i = 0; i < num_blobs; i++) {
+    keys.push_back("key" + std::to_string(i));
+    blobs.push_back("blob" + std::to_string(i));
+    for (size_t j = 0; j < write_timestamps.size(); j++) {
+      ASSERT_OK(db_->Put(write_opts, keys[i], write_timestamps[j],
+                         blobs[i] + std::to_string(j)));
+    }
+  }
+  ASSERT_OK(Flush());
+
+  ReadOptions read_options;
+  std::vector<std::string> read_timestamps = {Timestamp(0, 0), Timestamp(3, 0)};
+  Slice ts_upper_bound(read_timestamps[1]);
+  read_options.timestamp = &ts_upper_bound;
+
+  auto check_iter_entry =
+      [](const Iterator* iter, const std::string& expected_key,
+         const std::string& expected_ts, const std::string& expected_value,
+         bool key_is_internal = true) {
+        ASSERT_OK(iter->status());
+        if (key_is_internal) {
+          std::string expected_ukey_and_ts;
+          expected_ukey_and_ts.assign(expected_key.data(), expected_key.size());
+          expected_ukey_and_ts.append(expected_ts.data(), expected_ts.size());
+
+          ParsedInternalKey parsed_ikey;
+          ASSERT_OK(ParseInternalKey(iter->key(), &parsed_ikey,
+                                     true /* log_err_key */));
+          ASSERT_EQ(parsed_ikey.user_key, expected_ukey_and_ts);
+        } else {
+          ASSERT_EQ(iter->key(), expected_key);
+        }
+        ASSERT_EQ(iter->timestamp(), expected_ts);
+        ASSERT_EQ(iter->value(), expected_value);
+      };
+
+  // Forward iterating one version of each key, get in this order:
+  // [("key0", Timestamp(2, 0), "blob01"),
+  //  ("key1", Timestamp(2, 0), "blob11")...]
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToFirst();
+    for (int i = 0; i < num_blobs; i++) {
+      check_iter_entry(iter.get(), keys[i], write_timestamps[1],
+                       blobs[i] + std::to_string(1), /*key_is_internal*/ false);
+      iter->Next();
+    }
+  }
+
+  // Forward iteration, then reverse to backward.
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToFirst();
+    for (int i = 0; i < num_blobs * 2 - 1; i++) {
+      if (i < num_blobs) {
+        check_iter_entry(iter.get(), keys[i], write_timestamps[1],
+                         blobs[i] + std::to_string(1),
+                         /*key_is_internal*/ false);
+        if (i != num_blobs - 1) {
+          iter->Next();
+        }
+      } else {
+        if (i != num_blobs) {
+          check_iter_entry(iter.get(), keys[num_blobs * 2 - 1 - i],
+                           write_timestamps[1],
+                           blobs[num_blobs * 2 - 1 - i] + std::to_string(1),
+                           /*key_is_internal*/ false);
+        }
+        iter->Prev();
+      }
+    }
+  }
+
+  // Backward iterating one versions of each key, get in this order:
+  // [("key4", Timestamp(2, 0), "blob41"),
+  //  ("key3", Timestamp(2, 0), "blob31")...]
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToLast();
+    for (int i = 0; i < num_blobs; i++) {
+      check_iter_entry(iter.get(), keys[num_blobs - 1 - i], write_timestamps[1],
+                       blobs[num_blobs - 1 - i] + std::to_string(1),
+                       /*key_is_internal*/ false);
+      iter->Prev();
+    }
+  }
+
+  // Backward iteration, then reverse to forward.
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToLast();
+    for (int i = 0; i < num_blobs * 2 - 1; i++) {
+      if (i < num_blobs) {
+        check_iter_entry(iter.get(), keys[num_blobs - 1 - i],
+                         write_timestamps[1],
+                         blobs[num_blobs - 1 - i] + std::to_string(1),
+                         /*key_is_internal*/ false);
+        if (i != num_blobs - 1) {
+          iter->Prev();
+        }
+      } else {
+        if (i != num_blobs) {
+          check_iter_entry(iter.get(), keys[i - num_blobs], write_timestamps[1],
+                           blobs[i - num_blobs] + std::to_string(1),
+                           /*key_is_internal*/ false);
+        }
+        iter->Next();
+      }
+    }
+  }
+
+  Slice ts_lower_bound(read_timestamps[0]);
+  read_options.iter_start_ts = &ts_lower_bound;
+  // Forward iterating multiple versions of the same key, get in this order:
+  // [("key0", Timestamp(2, 0), "blob01"),
+  //  ("key0", Timestamp(1, 0), "blob00"),
+  //  ("key1", Timestamp(2, 0), "blob11")...]
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToFirst();
+    for (int i = 0; i < num_blobs; i++) {
+      for (size_t j = write_timestamps.size(); j > 0; --j) {
+        check_iter_entry(iter.get(), keys[i], write_timestamps[j - 1],
+                         blobs[i] + std::to_string(j - 1));
+        iter->Next();
+      }
+    }
+  }
+
+  // Backward iterating multiple versions of the same key, get in this order:
+  // [("key4", Timestamp(1, 0), "blob00"),
+  //  ("key4", Timestamp(2, 0), "blob01"),
+  //  ("key3", Timestamp(1, 0), "blob10")...]
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToLast();
+    for (int i = num_blobs; i > 0; i--) {
+      for (size_t j = 0; j < write_timestamps.size(); j++) {
+        check_iter_entry(iter.get(), keys[i - 1], write_timestamps[j],
+                         blobs[i - 1] + std::to_string(j));
+        iter->Prev();
+      }
+    }
+  }
+
+  int upper_bound_idx = num_blobs - 2;
+  int lower_bound_idx = 1;
+  Slice upper_bound_slice(keys[upper_bound_idx]);
+  Slice lower_bound_slice(keys[lower_bound_idx]);
+  read_options.iterate_upper_bound = &upper_bound_slice;
+  read_options.iterate_lower_bound = &lower_bound_slice;
+
+  // Forward iteration with upper and lower bound.
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToFirst();
+    for (int i = lower_bound_idx; i < upper_bound_idx; i++) {
+      for (size_t j = write_timestamps.size(); j > 0; --j) {
+        check_iter_entry(iter.get(), keys[i], write_timestamps[j - 1],
+                         blobs[i] + std::to_string(j - 1));
+        iter->Next();
+      }
+    }
+  }
+
+  // Backward iteration with upper and lower bound.
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToLast();
+    for (int i = upper_bound_idx; i > lower_bound_idx; i--) {
+      for (size_t j = 0; j < write_timestamps.size(); j++) {
+        check_iter_entry(iter.get(), keys[i - 1], write_timestamps[j],
+                         blobs[i - 1] + std::to_string(j));
+        iter->Prev();
+      }
+    }
+  }
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/blob/db_blob_compaction_test.cc

@@ -16,7 +16,6 @@ class DBBlobCompactionTest : public DBTestBase {
   explicit DBBlobCompactionTest()
       : DBTestBase("db_blob_compaction_test", /*env_do_fsync=*/false) {}
 
-#ifndef ROCKSDB_LITE
   const std::vector<InternalStats::CompactionStats>& GetCompactionStats() {
     VersionSet* const versions = dbfull()->GetVersionSet();
     assert(versions);
@@ -30,7 +29,6 @@ class DBBlobCompactionTest : public DBTestBase {
 
     return internal_stats->TEST_GetCompactionStats();
   }
-#endif  // ROCKSDB_LITE
 };
 
 namespace {
@@ -250,7 +248,6 @@ TEST_F(DBBlobCompactionTest, FilterByKeyLength) {
   ASSERT_OK(db_->Get(ReadOptions(), long_key, &value));
   ASSERT_EQ("value", value);
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -258,7 +255,6 @@ TEST_F(DBBlobCompactionTest, FilterByKeyLength) {
   // this involves neither reading nor writing blobs
   ASSERT_EQ(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_EQ(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }
@@ -299,7 +295,6 @@ TEST_F(DBBlobCompactionTest, FilterByValueLength) {
     ASSERT_EQ(long_value, value);
   }
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -307,12 +302,10 @@ TEST_F(DBBlobCompactionTest, FilterByValueLength) {
   // this involves reading but not writing blobs
   ASSERT_GT(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_EQ(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobCompactionTest, BlobCompactWithStartingLevel) {
   Options options = GetDefaultOptions();
 
@@ -388,7 +381,6 @@ TEST_F(DBBlobCompactionTest, BlobCompactWithStartingLevel) {
 
   Close();
 }
-#endif
 
 TEST_F(DBBlobCompactionTest, BlindWriteFilter) {
   Options options = GetDefaultOptions();
@@ -413,7 +405,6 @@ TEST_F(DBBlobCompactionTest, BlindWriteFilter) {
     ASSERT_EQ(new_blob_value, Get(key));
   }
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -421,7 +412,6 @@ TEST_F(DBBlobCompactionTest, BlindWriteFilter) {
   // this involves writing but not reading blobs
   ASSERT_EQ(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_GT(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }
@@ -540,7 +530,6 @@ TEST_F(DBBlobCompactionTest, CompactionFilter) {
     ASSERT_EQ(kv.second + std::string(padding), Get(kv.first));
   }
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -548,7 +537,6 @@ TEST_F(DBBlobCompactionTest, CompactionFilter) {
   // this involves reading and writing blobs
   ASSERT_GT(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_GT(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }
@@ -606,7 +594,6 @@ TEST_F(DBBlobCompactionTest, CompactionFilterReadBlobAndKeep) {
                               /*end=*/nullptr));
   ASSERT_EQ(blob_files, GetBlobFileNumbers());
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -614,7 +601,6 @@ TEST_F(DBBlobCompactionTest, CompactionFilterReadBlobAndKeep) {
   // this involves reading but not writing blobs
   ASSERT_GT(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_EQ(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }

db/blob/db_blob_corruption_test.cc

@@ -34,7 +34,6 @@ class DBBlobCorruptionTest : public DBTestBase {
   }
 };
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobCorruptionTest, VerifyWholeBlobFileChecksum) {
   Options options = GetDefaultOptions();
   options.enable_blob_files = true;
@@ -71,7 +70,6 @@ TEST_F(DBBlobCorruptionTest, VerifyWholeBlobFileChecksum) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }
-#endif  // !ROCKSDB_LITE
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/blob/db_blob_index_test.cc

@@ -131,9 +131,7 @@ class DBBlobIndexTest : public DBTestBase {
         ASSERT_OK(Flush());
         ASSERT_OK(
             dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
-#ifndef ROCKSDB_LITE
         ASSERT_EQ("0,1", FilesPerLevel());
-#endif  // !ROCKSDB_LITE
         break;
     }
   }
@@ -459,7 +457,6 @@ TEST_F(DBBlobIndexTest, Iterate) {
     verify(15, Status::kOk, get_value(16, 0), get_value(14, 0),
            create_blob_iterator, check_is_blob(false));
 
-#ifndef ROCKSDB_LITE
     // Iterator with blob support and using seek.
     ASSERT_OK(dbfull()->SetOptions(
         cfh(), {{"max_sequential_skip_in_iterations", "0"}}));
@@ -484,7 +481,6 @@ TEST_F(DBBlobIndexTest, Iterate) {
            create_blob_iterator, check_is_blob(false));
     verify(15, Status::kOk, get_value(16, 0), get_value(14, 0),
            create_blob_iterator, check_is_blob(false));
-#endif  // !ROCKSDB_LITE
 
     for (auto* snapshot : snapshots) {
       dbfull()->ReleaseSnapshot(snapshot);
@@ -584,12 +580,10 @@ TEST_F(DBBlobIndexTest, IntegratedBlobIterate) {
   Status expected_status;
   verify(1, expected_status, expected_value);
 
-#ifndef ROCKSDB_LITE
   // Test DBIter::FindValueForCurrentKeyUsingSeek flow.
   ASSERT_OK(dbfull()->SetOptions(cfh(),
                                  {{"max_sequential_skip_in_iterations", "0"}}));
   verify(1, expected_status, expected_value);
-#endif  // !ROCKSDB_LITE
 }
 
 }  // namespace ROCKSDB_NAMESPACE

db/builder.cc

@@ -56,8 +56,8 @@ TableBuilder* NewTableBuilder(const TableBuilderOptions& tboptions,
 Status BuildTable(
     const std::string& dbname, VersionSet* versions,
     const ImmutableDBOptions& db_options, const TableBuilderOptions& tboptions,
-    const FileOptions& file_options, TableCache* table_cache,
-    InternalIterator* iter,
+    const FileOptions& file_options, const ReadOptions& read_options,
+    TableCache* table_cache, InternalIterator* iter,
     std::vector<std::unique_ptr<FragmentedRangeTombstoneIterator>>
         range_del_iters,
     FileMetaData* meta, std::vector<BlobFileAddition>* blob_file_additions,
@@ -71,8 +71,9 @@ Status BuildTable(
     int job_id, const Env::IOPriority io_priority,
     TableProperties* table_properties, Env::WriteLifeTimeHint write_hint,
     const std::string* full_history_ts_low,
-    BlobFileCompletionCallback* blob_callback, uint64_t* num_input_entries,
-    uint64_t* memtable_payload_bytes, uint64_t* memtable_garbage_bytes) {
+    BlobFileCompletionCallback* blob_callback, Version* version,
+    uint64_t* num_input_entries, uint64_t* memtable_payload_bytes,
+    uint64_t* memtable_garbage_bytes) {
   assert((tboptions.column_family_id ==
           TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
          tboptions.column_family_name.empty());
@@ -106,11 +107,9 @@ Status BuildTable(
   std::vector<std::string> blob_file_paths;
   std::string file_checksum = kUnknownFileChecksum;
   std::string file_checksum_func_name = kUnknownFileChecksumFuncName;
-#ifndef ROCKSDB_LITE
   EventHelpers::NotifyTableFileCreationStarted(ioptions.listeners, dbname,
                                                tboptions.column_family_name,
                                                fname, job_id, tboptions.reason);
-#endif  // !ROCKSDB_LITE
   Env* env = db_options.env;
   assert(env);
   FileSystem* fs = db_options.fs.get();
@@ -175,10 +174,10 @@ Status BuildTable(
       builder = NewTableBuilder(tboptions, file_writer.get());
     }
 
+    auto ucmp = tboptions.internal_comparator.user_comparator();
     MergeHelper merge(
-        env, tboptions.internal_comparator.user_comparator(),
-        ioptions.merge_operator.get(), compaction_filter.get(), ioptions.logger,
-        true /* internal key corruption is not ok */,
+        env, ucmp, ioptions.merge_operator.get(), compaction_filter.get(),
+        ioptions.logger, true /* internal key corruption is not ok */,
         snapshots.empty() ? 0 : snapshots.back(), snapshot_checker);
 
     std::unique_ptr<BlobFileBuilder> blob_file_builder(
@@ -196,9 +195,8 @@ Status BuildTable(
 
     const std::atomic<bool> kManualCompactionCanceledFalse{false};
     CompactionIterator c_iter(
-        iter, tboptions.internal_comparator.user_comparator(), &merge,
-        kMaxSequenceNumber, &snapshots, earliest_write_conflict_snapshot,
-        job_snapshot, snapshot_checker, env,
+        iter, ucmp, &merge, kMaxSequenceNumber, &snapshots,
+        earliest_write_conflict_snapshot, job_snapshot, snapshot_checker, env,
         ShouldReportDetailedTime(env, ioptions.stats),
         true /* internal key corruption is not ok */, range_del_agg.get(),
         blob_file_builder.get(), ioptions.allow_data_in_errors,
@@ -241,14 +239,28 @@ Status BuildTable(
 
     if (s.ok()) {
       auto range_del_it = range_del_agg->NewIterator();
+      Slice last_tombstone_start_user_key{};
       for (range_del_it->SeekToFirst(); range_del_it->Valid();
            range_del_it->Next()) {
         auto tombstone = range_del_it->Tombstone();
         auto kv = tombstone.Serialize();
         builder->Add(kv.first.Encode(), kv.second);
-        meta->UpdateBoundariesForRange(kv.first, tombstone.SerializeEndKey(),
-                                       tombstone.seq_,
+        InternalKey tombstone_end = tombstone.SerializeEndKey();
+        meta->UpdateBoundariesForRange(kv.first, tombstone_end, tombstone.seq_,
                                        tboptions.internal_comparator);
+        if (version) {
+          if (last_tombstone_start_user_key.empty() ||
+              ucmp->CompareWithoutTimestamp(last_tombstone_start_user_key,
+                                            range_del_it->start_key()) < 0) {
+            SizeApproximationOptions approx_opts;
+            approx_opts.files_size_error_margin = 0.1;
+            meta->compensated_range_deletion_size += versions->ApproximateSize(
+                approx_opts, read_options, version, kv.first.Encode(),
+                tombstone_end.Encode(), 0 /* start_level */, -1 /* end_level */,
+                TableReaderCaller::kFlush);
+          }
+          last_tombstone_start_user_key = range_del_it->start_key();
+        }
       }
     }
 
@@ -279,6 +291,7 @@ Status BuildTable(
     if (s.ok() && !empty) {
       uint64_t file_size = builder->FileSize();
       meta->fd.file_size = file_size;
+      meta->tail_size = builder->GetTailSize();
       meta->marked_for_compaction = builder->NeedCompact();
       assert(meta->fd.GetFileSize() > 0);
       tp = builder
@@ -357,7 +370,6 @@ Status BuildTable(
       // here because this is a special case after we finish the table building.
       // No matter whether use_direct_io_for_flush_and_compaction is true,
       // the goal is to cache it here for further user reads.
-      ReadOptions read_options;
       std::unique_ptr<InternalIterator> it(table_cache->NewIterator(
           read_options, file_options, tboptions.internal_comparator, *meta,
           nullptr /* range_del_agg */, mutable_cf_options.prefix_extractor,
@@ -369,7 +381,8 @@ Status BuildTable(
           MaxFileSizeForL0MetaPin(mutable_cf_options),
           /*smallest_compaction_key=*/nullptr,
           /*largest_compaction_key*/ nullptr,
-          /*allow_unprepared_value*/ false));
+          /*allow_unprepared_value*/ false,
+          mutable_cf_options.block_protection_bytes_per_key));
       s = it->status();
       if (s.ok() && paranoid_file_checks) {
         OutputValidator file_validator(tboptions.internal_comparator,

db/builder.h

@@ -13,6 +13,7 @@
 #include "db/range_tombstone_fragmenter.h"
 #include "db/seqno_to_time_mapping.h"
 #include "db/table_properties_collector.h"
+#include "db/version_set.h"
 #include "logging/event_logger.h"
 #include "options/cf_options.h"
 #include "rocksdb/comparator.h"
@@ -52,8 +53,8 @@ TableBuilder* NewTableBuilder(const TableBuilderOptions& tboptions,
 extern Status BuildTable(
     const std::string& dbname, VersionSet* versions,
     const ImmutableDBOptions& db_options, const TableBuilderOptions& tboptions,
-    const FileOptions& file_options, TableCache* table_cache,
-    InternalIterator* iter,
+    const FileOptions& file_options, const ReadOptions& read_options,
+    TableCache* table_cache, InternalIterator* iter,
     std::vector<std::unique_ptr<FragmentedRangeTombstoneIterator>>
         range_del_iters,
     FileMetaData* meta, std::vector<BlobFileAddition>* blob_file_additions,
@@ -70,7 +71,7 @@ extern Status BuildTable(
     Env::WriteLifeTimeHint write_hint = Env::WLTH_NOT_SET,
     const std::string* full_history_ts_low = nullptr,
     BlobFileCompletionCallback* blob_callback = nullptr,
-    uint64_t* num_input_entries = nullptr,
+    Version* version = nullptr, uint64_t* num_input_entries = nullptr,
     uint64_t* memtable_payload_bytes = nullptr,
     uint64_t* memtable_garbage_bytes = nullptr);
 

db/c.cc

@@ -7,8 +7,6 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
-#ifndef ROCKSDB_LITE
-
 #include "rocksdb/c.h"
 
 #include <cstdlib>
@@ -17,7 +15,7 @@
 #include <vector>
 
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/comparator.h"
 #include "rocksdb/convenience.h"
@@ -69,6 +67,7 @@ using ROCKSDB_NAMESPACE::CompactionOptionsFIFO;
 using ROCKSDB_NAMESPACE::CompactRangeOptions;
 using ROCKSDB_NAMESPACE::Comparator;
 using ROCKSDB_NAMESPACE::CompressionType;
+using ROCKSDB_NAMESPACE::ConfigOptions;
 using ROCKSDB_NAMESPACE::CuckooTableOptions;
 using ROCKSDB_NAMESPACE::DB;
 using ROCKSDB_NAMESPACE::DBOptions;
@@ -78,6 +77,7 @@ using ROCKSDB_NAMESPACE::EnvOptions;
 using ROCKSDB_NAMESPACE::FileLock;
 using ROCKSDB_NAMESPACE::FilterPolicy;
 using ROCKSDB_NAMESPACE::FlushOptions;
+using ROCKSDB_NAMESPACE::HyperClockCacheOptions;
 using ROCKSDB_NAMESPACE::InfoLogLevel;
 using ROCKSDB_NAMESPACE::IngestExternalFileOptions;
 using ROCKSDB_NAMESPACE::Iterator;
@@ -208,6 +208,9 @@ struct rocksdb_logger_t {
 struct rocksdb_lru_cache_options_t {
   LRUCacheOptions rep;
 };
+struct rocksdb_hyper_clock_cache_options_t {
+  HyperClockCacheOptions rep;
+};
 struct rocksdb_memory_allocator_t {
   std::shared_ptr<MemoryAllocator> rep;
 };
@@ -1805,6 +1808,17 @@ void rocksdb_flush_cf(rocksdb_t* db, const rocksdb_flushoptions_t* options,
   SaveError(errptr, db->rep->Flush(options->rep, column_family->rep));
 }
 
+void rocksdb_flush_cfs(rocksdb_t* db, const rocksdb_flushoptions_t* options,
+                       rocksdb_column_family_handle_t** column_families,
+                       int num_column_families, char** errptr) {
+  std::vector<ColumnFamilyHandle*> column_family_handles;
+  for (int i = 0; i < num_column_families; i++) {
+    column_family_handles.push_back(column_families[i]->rep);
+  }
+
+  SaveError(errptr, db->rep->Flush(options->rep, column_family_handles));
+}
+
 void rocksdb_flush_wal(rocksdb_t* db, unsigned char sync, char** errptr) {
   SaveError(errptr, db->rep->FlushWAL(sync));
 }
@@ -2498,8 +2512,12 @@ void rocksdb_load_latest_options(
     rocksdb_options_t*** list_column_family_options, char** errptr) {
   DBOptions db_opt;
   std::vector<ColumnFamilyDescriptor> cf_descs;
-  Status s = LoadLatestOptions(std::string(db_path), env->rep, &db_opt,
-                               &cf_descs, ignore_unknown_options, &cache->rep);
+  ConfigOptions config_opts;
+  config_opts.ignore_unknown_options = ignore_unknown_options;
+  config_opts.input_strings_escaped = true;
+  config_opts.env = env->rep;
+  Status s = LoadLatestOptions(config_opts, std::string(db_path), &db_opt,
+                               &cf_descs, &cache->rep);
   if (s.ok()) {
     char** cf_names = (char**)malloc(cf_descs.size() * sizeof(char*));
     rocksdb_options_t** cf_options = (rocksdb_options_t**)malloc(
@@ -2588,6 +2606,12 @@ void rocksdb_block_based_options_set_partition_filters(
   options->rep.partition_filters = partition_filters;
 }
 
+void rocksdb_block_based_options_set_optimize_filters_for_memory(
+    rocksdb_block_based_table_options_t* options,
+    unsigned char optimize_filters_for_memory) {
+  options->rep.optimize_filters_for_memory = optimize_filters_for_memory;
+}
+
 void rocksdb_block_based_options_set_use_delta_encoding(
     rocksdb_block_based_table_options_t* options,
     unsigned char use_delta_encoding) {
@@ -2614,14 +2638,6 @@ void rocksdb_block_based_options_set_block_cache(
   }
 }
 
-void rocksdb_block_based_options_set_block_cache_compressed(
-    rocksdb_block_based_table_options_t* options,
-    rocksdb_cache_t* block_cache_compressed) {
-  if (block_cache_compressed) {
-    options->rep.block_cache_compressed = block_cache_compressed->rep;
-  }
-}
-
 void rocksdb_block_based_options_set_whole_key_filtering(
     rocksdb_block_based_table_options_t* options, unsigned char v) {
   options->rep.whole_key_filtering = v;
@@ -4048,6 +4064,8 @@ uint64_t rocksdb_perfcontext_metric(rocksdb_perfcontext_t* context,
       return rep->blob_decompress_time;
     case rocksdb_internal_range_del_reseek_count:
       return rep->internal_range_del_reseek_count;
+    case rocksdb_block_read_cpu_time:
+      return rep->block_read_cpu_time;
     default:
       break;
   }
@@ -4443,6 +4461,15 @@ rocksdb_readoptions_get_io_timeout(rocksdb_readoptions_t* opt) {
   return opt->rep.io_timeout.count();
 }
 
+void rocksdb_readoptions_set_async_io(rocksdb_readoptions_t* opt,
+                                      unsigned char v) {
+  opt->rep.async_io = v;
+}
+
+unsigned char rocksdb_readoptions_get_async_io(rocksdb_readoptions_t* opt) {
+  return opt->rep.async_io;
+}
+
 void rocksdb_readoptions_set_timestamp(rocksdb_readoptions_t* opt,
                                        const char* ts, size_t tslen) {
   if (ts == nullptr) {
@@ -4654,12 +4681,59 @@ rocksdb_cache_t* rocksdb_cache_create_lru_with_strict_capacity_limit(
 }
 
 rocksdb_cache_t* rocksdb_cache_create_lru_opts(
-    rocksdb_lru_cache_options_t* opt) {
+    const rocksdb_lru_cache_options_t* opt) {
   rocksdb_cache_t* c = new rocksdb_cache_t;
   c->rep = NewLRUCache(opt->rep);
   return c;
 }
 
+rocksdb_hyper_clock_cache_options_t* rocksdb_hyper_clock_cache_options_create(
+    size_t capacity, size_t estimated_entry_charge) {
+  return new rocksdb_hyper_clock_cache_options_t{
+      HyperClockCacheOptions(capacity, estimated_entry_charge)};
+}
+
+void rocksdb_hyper_clock_cache_options_destroy(
+    rocksdb_hyper_clock_cache_options_t* opt) {
+  delete opt;
+}
+
+void rocksdb_hyper_clock_cache_options_set_capacity(
+    rocksdb_hyper_clock_cache_options_t* opts, size_t capacity) {
+  opts->rep.capacity = capacity;
+}
+
+void rocksdb_hyper_clock_cache_options_set_estimated_entry_charge(
+    rocksdb_hyper_clock_cache_options_t* opts, size_t estimated_entry_charge) {
+  opts->rep.estimated_entry_charge = estimated_entry_charge;
+}
+
+void rocksdb_hyper_clock_cache_options_set_num_shard_bits(
+    rocksdb_hyper_clock_cache_options_t* opts, int num_shard_bits) {
+  opts->rep.num_shard_bits = num_shard_bits;
+}
+
+void rocksdb_hyper_clock_cache_options_set_memory_allocator(
+    rocksdb_hyper_clock_cache_options_t* opts,
+    rocksdb_memory_allocator_t* memory_allocator) {
+  opts->rep.memory_allocator = memory_allocator->rep;
+}
+
+rocksdb_cache_t* rocksdb_cache_create_hyper_clock(
+    size_t capacity, size_t estimated_entry_charge) {
+  HyperClockCacheOptions opts(capacity, estimated_entry_charge);
+  rocksdb_cache_t* c = new rocksdb_cache_t;
+  c->rep = opts.MakeSharedCache();
+  return c;
+}
+
+rocksdb_cache_t* rocksdb_cache_create_hyper_clock_opts(
+    const rocksdb_hyper_clock_cache_options_t* opts) {
+  rocksdb_cache_t* c = new rocksdb_cache_t;
+  c->rep = opts->rep.MakeSharedCache();
+  return c;
+}
+
 void rocksdb_cache_destroy(rocksdb_cache_t* cache) { delete cache; }
 
 void rocksdb_cache_disown_data(rocksdb_cache_t* cache) {
@@ -4670,18 +4744,26 @@ void rocksdb_cache_set_capacity(rocksdb_cache_t* cache, size_t capacity) {
   cache->rep->SetCapacity(capacity);
 }
 
-size_t rocksdb_cache_get_capacity(rocksdb_cache_t* cache) {
+size_t rocksdb_cache_get_capacity(const rocksdb_cache_t* cache) {
   return cache->rep->GetCapacity();
 }
 
-size_t rocksdb_cache_get_usage(rocksdb_cache_t* cache) {
+size_t rocksdb_cache_get_usage(const rocksdb_cache_t* cache) {
   return cache->rep->GetUsage();
 }
 
-size_t rocksdb_cache_get_pinned_usage(rocksdb_cache_t* cache) {
+size_t rocksdb_cache_get_pinned_usage(const rocksdb_cache_t* cache) {
   return cache->rep->GetPinnedUsage();
 }
 
+size_t rocksdb_cache_get_table_address_count(const rocksdb_cache_t* cache) {
+  return cache->rep->GetTableAddressCount();
+}
+
+size_t rocksdb_cache_get_occupancy_count(const rocksdb_cache_t* cache) {
+  return cache->rep->GetOccupancyCount();
+}
+
 rocksdb_dbpath_t* rocksdb_dbpath_create(const char* path,
                                         uint64_t target_size) {
   rocksdb_dbpath_t* result = new rocksdb_dbpath_t;
@@ -4900,6 +4982,12 @@ void rocksdb_ingestexternalfileoptions_set_ingest_behind(
   opt->rep.ingest_behind = ingest_behind;
 }
 
+void rocksdb_ingestexternalfileoptions_set_fail_if_not_bottommost_level(
+    rocksdb_ingestexternalfileoptions_t* opt,
+    unsigned char fail_if_not_bottommost_level) {
+  opt->rep.fail_if_not_bottommost_level = fail_if_not_bottommost_level;
+}
+
 void rocksdb_ingestexternalfileoptions_destroy(
     rocksdb_ingestexternalfileoptions_t* opt) {
   delete opt;
@@ -5061,6 +5149,16 @@ rocksdb_fifo_compaction_options_t* rocksdb_fifo_compaction_options_create() {
   return result;
 }
 
+void rocksdb_fifo_compaction_options_set_allow_compaction(
+    rocksdb_fifo_compaction_options_t* fifo_opts, unsigned char allow_compaction) {
+  fifo_opts->rep.allow_compaction = allow_compaction;
+}
+
+unsigned char rocksdb_fifo_compaction_options_get_allow_compaction(
+    rocksdb_fifo_compaction_options_t* fifo_opts) {
+  return fifo_opts->rep.allow_compaction;
+}
+
 void rocksdb_fifo_compaction_options_set_max_table_files_size(
     rocksdb_fifo_compaction_options_t* fifo_opts, uint64_t size) {
   fifo_opts->rep.max_table_files_size = size;
@@ -5765,6 +5863,35 @@ void rocksdb_transaction_multi_get(rocksdb_transaction_t* txn,
   }
 }
 
+void rocksdb_transaction_multi_get_for_update(
+    rocksdb_transaction_t* txn, const rocksdb_readoptions_t* options,
+    size_t num_keys, const char* const* keys_list,
+    const size_t* keys_list_sizes, char** values_list,
+    size_t* values_list_sizes, char** errs) {
+  std::vector<Slice> keys(num_keys);
+  for (size_t i = 0; i < num_keys; i++) {
+    keys[i] = Slice(keys_list[i], keys_list_sizes[i]);
+  }
+  std::vector<std::string> values(num_keys);
+  std::vector<Status> statuses =
+      txn->rep->MultiGetForUpdate(options->rep, keys, &values);
+  for (size_t i = 0; i < num_keys; i++) {
+    if (statuses[i].ok()) {
+      values_list[i] = CopyString(values[i]);
+      values_list_sizes[i] = values[i].size();
+      errs[i] = nullptr;
+    } else {
+      values_list[i] = nullptr;
+      values_list_sizes[i] = 0;
+      if (!statuses[i].IsNotFound()) {
+        errs[i] = strdup(statuses[i].ToString().c_str());
+      } else {
+        errs[i] = nullptr;
+      }
+    }
+  }
+}
+
 void rocksdb_transaction_multi_get_cf(
     rocksdb_transaction_t* txn, const rocksdb_readoptions_t* options,
     const rocksdb_column_family_handle_t* const* column_families,
@@ -5797,6 +5924,38 @@ void rocksdb_transaction_multi_get_cf(
   }
 }
 
+void rocksdb_transaction_multi_get_for_update_cf(
+    rocksdb_transaction_t* txn, const rocksdb_readoptions_t* options,
+    const rocksdb_column_family_handle_t* const* column_families,
+    size_t num_keys, const char* const* keys_list,
+    const size_t* keys_list_sizes, char** values_list,
+    size_t* values_list_sizes, char** errs) {
+  std::vector<Slice> keys(num_keys);
+  std::vector<ColumnFamilyHandle*> cfs(num_keys);
+  for (size_t i = 0; i < num_keys; i++) {
+    keys[i] = Slice(keys_list[i], keys_list_sizes[i]);
+    cfs[i] = column_families[i]->rep;
+  }
+  std::vector<std::string> values(num_keys);
+  std::vector<Status> statuses =
+      txn->rep->MultiGetForUpdate(options->rep, cfs, keys, &values);
+  for (size_t i = 0; i < num_keys; i++) {
+    if (statuses[i].ok()) {
+      values_list[i] = CopyString(values[i]);
+      values_list_sizes[i] = values[i].size();
+      errs[i] = nullptr;
+    } else {
+      values_list[i] = nullptr;
+      values_list_sizes[i] = 0;
+      if (!statuses[i].IsNotFound()) {
+        errs[i] = strdup(statuses[i].ToString().c_str());
+      } else {
+        errs[i] = nullptr;
+      }
+    }
+  }
+}
+
 // Read a key outside a transaction
 char* rocksdb_transactiondb_get(rocksdb_transactiondb_t* txn_db,
                                 const rocksdb_readoptions_t* options,
@@ -6098,6 +6257,18 @@ void rocksdb_transactiondb_flush_cf(
   SaveError(errptr, txn_db->rep->Flush(options->rep, column_family->rep));
 }
 
+void rocksdb_transactiondb_flush_cfs(
+    rocksdb_transactiondb_t* txn_db, const rocksdb_flushoptions_t* options,
+    rocksdb_column_family_handle_t** column_families, int num_column_families,
+    char** errptr) {
+  std::vector<ColumnFamilyHandle*> column_family_handles;
+  for (int i = 0; i < num_column_families; i++) {
+    column_family_handles.push_back(column_families[i]->rep);
+  }
+
+  SaveError(errptr, txn_db->rep->Flush(options->rep, column_family_handles));
+}
+
 rocksdb_checkpoint_t* rocksdb_transactiondb_checkpoint_object_create(
     rocksdb_transactiondb_t* txn_db, char** errptr) {
   Checkpoint* checkpoint;
@@ -6386,5 +6557,3 @@ void rocksdb_enable_manual_compaction(rocksdb_t* db) {
 }
 
 }  // end extern "C"
-
-#endif  // !ROCKSDB_LITE

db/c_test.c

@@ -5,8 +5,6 @@
 
 #include <stdio.h>
 
-#ifndef ROCKSDB_LITE  // Lite does not support C API
-
 #include <assert.h>
 #include <stddef.h>
 #include <stdlib.h>
@@ -490,6 +488,19 @@ static void CheckTxnPinGetCF(rocksdb_transaction_t* txn,
   rocksdb_pinnableslice_destroy(p);
 }
 
+static void CheckTxnGetForUpdate(rocksdb_transaction_t* txn,
+                                 const rocksdb_readoptions_t* options,
+                                 const char* key, const char* expected) {
+  char* err = NULL;
+  size_t val_len;
+  char* val;
+  val = rocksdb_transaction_get_for_update(txn, options, key, strlen(key),
+                                           &val_len, true, &err);
+  CheckNoError(err);
+  CheckEqual(expected, val, val_len);
+  Free(&val);
+}
+
 static void CheckTxnDBGet(rocksdb_transactiondb_t* txn_db,
                           const rocksdb_readoptions_t* options, const char* key,
                           const char* expected) {
@@ -517,6 +528,20 @@ static void CheckTxnDBGetCF(rocksdb_transactiondb_t* txn_db,
   Free(&val);
 }
 
+static void CheckTxnGetForUpdateCF(
+    rocksdb_transaction_t* txn, const rocksdb_readoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    const char* expected) {
+  char* err = NULL;
+  size_t val_len;
+  char* val;
+  val = rocksdb_transaction_get_for_update_cf(
+      txn, options, column_family, key, strlen(key), &val_len, true, &err);
+  CheckNoError(err);
+  CheckEqual(expected, val, val_len);
+  Free(&val);
+}
+
 static void CheckTxnDBPinGet(rocksdb_transactiondb_t* txn_db,
                              const rocksdb_readoptions_t* options,
                              const char* key, const char* expected) {
@@ -2572,6 +2597,9 @@ int main(int argc, char** argv) {
     rocksdb_readoptions_set_io_timeout(ro, 400);
     CheckCondition(400 == rocksdb_readoptions_get_io_timeout(ro));
 
+    rocksdb_readoptions_set_async_io(ro, 1);
+    CheckCondition(1 == rocksdb_readoptions_get_async_io(ro));
+
     rocksdb_readoptions_destroy(ro);
   }
 
@@ -3203,6 +3231,120 @@ int main(int argc, char** argv) {
     rocksdb_transactiondb_options_destroy(txn_db_options);
   }
 
+  StartPhase("transactions_multi_get_for_update");
+  {
+    // open a TransactionDB
+    txn_db_options = rocksdb_transactiondb_options_create();
+    rocksdb_transactiondb_options_set_transaction_lock_timeout(txn_db_options,
+                                                               0);
+    txn_options = rocksdb_transaction_options_create();
+    rocksdb_options_set_create_if_missing(options, 1);
+    txn_db = rocksdb_transactiondb_open(options, txn_db_options, dbname, &err);
+    CheckNoError(err);
+
+    rocksdb_transactiondb_put(txn_db, woptions, "foo", 3, "hey", 3, &err);
+    CheckNoError(err);
+    rocksdb_transactiondb_put(txn_db, woptions, "bar", 3, "hello", 5, &err);
+    CheckNoError(err);
+
+    // begin transactions
+    txn = rocksdb_transaction_begin(txn_db, woptions, txn_options, NULL);
+    rocksdb_transaction_t* txn2 =
+        rocksdb_transaction_begin(txn_db, woptions, txn_options, NULL);
+
+    // multi get
+    {
+      const char* keys[2] = {"foo", "bar"};
+      const size_t keys_sizes[2] = {3, 3};
+      char* vals[2];
+      size_t vals_sizes[2];
+      char* errs[2];
+      const char* expected[2] = {"hey", "hello"};
+      rocksdb_transaction_multi_get_for_update(
+          txn, roptions, 2, keys, keys_sizes, vals, vals_sizes, errs);
+      CheckMultiGetValues(2, vals, vals_sizes, errs, expected);
+    }
+
+    char* conflict_err = NULL;
+    size_t val_len;
+    rocksdb_transaction_get_for_update(txn2, roptions, "foo", 3, &val_len, true,
+                                       &conflict_err);
+    // get-for-update conflict
+    CheckCondition(conflict_err != NULL);
+    Free(&conflict_err);
+
+    // commit
+    rocksdb_transaction_commit(txn, &err);
+    CheckNoError(err);
+
+    // should work after first tx is commited
+    CheckTxnGetForUpdate(txn2, roptions, "foo", "hey");
+
+    // commit the second one
+    rocksdb_transaction_commit(txn2, &err);
+    CheckNoError(err);
+
+    // destroy txns
+    rocksdb_transaction_destroy(txn);
+    rocksdb_transaction_destroy(txn2);
+
+    // same for column families
+
+    rocksdb_column_family_handle_t* cfh;
+    cfh = rocksdb_transactiondb_create_column_family(txn_db, options,
+                                                     "txn_db_cf", &err);
+    CheckNoError(err);
+
+    rocksdb_transactiondb_put_cf(txn_db, woptions, cfh, "cf_foo", 6, "cf_hello",
+                                 8, &err);
+    CheckNoError(err);
+    rocksdb_transactiondb_put_cf(txn_db, woptions, cfh, "cf_bar", 6, "cf_hey",
+                                 6, &err);
+    CheckNoError(err);
+
+    txn = rocksdb_transaction_begin(txn_db, woptions, txn_options, NULL);
+    txn2 = rocksdb_transaction_begin(txn_db, woptions, txn_options, NULL);
+
+    {
+      const rocksdb_column_family_handle_t* get_handles[2] = {cfh, cfh};
+      const char* keys[2] = {"cf_foo", "cf_bar"};
+      const size_t keys_sizes[2] = {6, 6};
+      char* vals[2];
+      size_t vals_sizes[2];
+      char* errs[2];
+      const char* expected[2] = {"cf_hello", "cf_hey"};
+      rocksdb_transaction_multi_get_for_update_cf(txn, roptions, get_handles, 2,
+                                                  keys, keys_sizes, vals,
+                                                  vals_sizes, errs);
+      CheckMultiGetValues(2, vals, vals_sizes, errs, expected);
+    }
+
+    char* conflict_err_cf = NULL;
+    size_t val_len_cf;
+    rocksdb_transaction_get_for_update_cf(txn2, roptions, cfh, "cf_foo", 6,
+                                          &val_len_cf, true, &conflict_err_cf);
+    CheckCondition(conflict_err_cf != NULL);
+    Free(&conflict_err_cf);
+
+    rocksdb_transaction_commit(txn, &err);
+    CheckNoError(err);
+
+    CheckTxnGetForUpdateCF(txn2, roptions, cfh, "cf_foo", "cf_hello");
+
+    rocksdb_transaction_commit(txn2, &err);
+    CheckNoError(err);
+
+    // close and destroy
+    rocksdb_column_family_handle_destroy(cfh);
+    rocksdb_transaction_destroy(txn);
+    rocksdb_transaction_destroy(txn2);
+    rocksdb_transactiondb_close(txn_db);
+    rocksdb_destroy_db(options, dbname, &err);
+    CheckNoError(err);
+    rocksdb_transaction_options_destroy(txn_options);
+    rocksdb_transactiondb_options_destroy(txn_db_options);
+  }
+
   StartPhase("optimistic_transactions");
   {
     rocksdb_options_t* db_options = rocksdb_options_create();
@@ -3465,12 +3607,3 @@ int main(int argc, char** argv) {
   fprintf(stderr, "PASS\n");
   return 0;
 }
-
-#else
-
-int main(void) {
-  fprintf(stderr, "SKIPPED\n");
-  return 0;
-}
-
-#endif  // !ROCKSDB_LITE

db/column_family.cc

@@ -53,11 +53,9 @@ ColumnFamilyHandleImpl::ColumnFamilyHandleImpl(
 
 ColumnFamilyHandleImpl::~ColumnFamilyHandleImpl() {
   if (cfd_ != nullptr) {
-#ifndef ROCKSDB_LITE
     for (auto& listener : cfd_->ioptions()->listeners) {
       listener->OnColumnFamilyHandleDeletionStarted(this);
     }
-#endif  // ROCKSDB_LITE
     // Job id == 0 means that this is not our background process, but rather
     // user thread
     // Need to hold some shared pointers owned by the initial_cf_options
@@ -88,15 +86,10 @@ const std::string& ColumnFamilyHandleImpl::GetName() const {
 }
 
 Status ColumnFamilyHandleImpl::GetDescriptor(ColumnFamilyDescriptor* desc) {
-#ifndef ROCKSDB_LITE
   // accessing mutable cf-options requires db mutex.
   InstrumentedMutexLock l(mutex_);
   *desc = ColumnFamilyDescriptor(cfd()->GetName(), cfd()->GetLatestCFOptions());
   return Status::OK();
-#else
-  (void)desc;
-  return Status::NotSupported();
-#endif  // !ROCKSDB_LITE
 }
 
 const Comparator* ColumnFamilyHandleImpl::GetComparator() const {
@@ -347,7 +340,6 @@ ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
         result.hard_pending_compaction_bytes_limit;
   }
 
-#ifndef ROCKSDB_LITE
   // When the DB is stopped, it's possible that there are some .trash files that
   // were not deleted yet, when we open the DB we will find these .trash files
   // and schedule them to be deleted (or delete immediately if SstFileManager
@@ -359,7 +351,6 @@ ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
                                       result.cf_paths[i].path)
         .PermitUncheckedError();
   }
-#endif
 
   if (result.cf_paths.empty()) {
     result.cf_paths = db_options.db_paths;
@@ -557,7 +548,6 @@ ColumnFamilyData::ColumnFamilyData(
       next_(nullptr),
       prev_(nullptr),
       log_number_(0),
-      flush_reason_(FlushReason::kOthers),
       column_family_set_(column_family_set),
       queued_for_flush_(false),
       queued_for_compaction_(false),
@@ -565,7 +555,8 @@ ColumnFamilyData::ColumnFamilyData(
       allow_2pc_(db_options.allow_2pc),
       last_memtable_id_(0),
       db_paths_registered_(false),
-      mempurge_used_(false) {
+      mempurge_used_(false),
+      next_epoch_number_(1) {
   if (id_ != kDummyColumnFamilyDataId) {
     // TODO(cc): RegisterDbPaths can be expensive, considering moving it
     // outside of this constructor which might be called with db mutex held.
@@ -602,7 +593,6 @@ ColumnFamilyData::ColumnFamilyData(
     if (ioptions_.compaction_style == kCompactionStyleLevel) {
       compaction_picker_.reset(
           new LevelCompactionPicker(ioptions_, &internal_comparator_));
-#ifndef ROCKSDB_LITE
     } else if (ioptions_.compaction_style == kCompactionStyleUniversal) {
       compaction_picker_.reset(
           new UniversalCompactionPicker(ioptions_, &internal_comparator_));
@@ -616,7 +606,6 @@ ColumnFamilyData::ColumnFamilyData(
                      "Column family %s does not use any background compaction. "
                      "Compactions can only be done via CompactFiles\n",
                      GetName().c_str());
-#endif  // !ROCKSDB_LITE
     } else {
       ROCKS_LOG_ERROR(ioptions_.logger,
                       "Unable to recognize the specified compaction style %d. "
@@ -880,7 +869,7 @@ int GetL0ThresholdSpeedupCompaction(int level0_file_num_compaction_trigger,
 }
 }  // anonymous namespace
 
-std::pair<WriteStallCondition, ColumnFamilyData::WriteStallCause>
+std::pair<WriteStallCondition, WriteStallCause>
 ColumnFamilyData::GetWriteStallConditionAndCause(
     int num_unflushed_memtables, int num_l0_files,
     uint64_t num_compaction_needed_bytes,
@@ -953,7 +942,8 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
       internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_STOPS, 1);
       if (compaction_picker_->IsLevel0CompactionInProgress()) {
         internal_stats_->AddCFStats(
-            InternalStats::LOCKED_L0_FILE_COUNT_LIMIT_STOPS, 1);
+            InternalStats::L0_FILE_COUNT_LIMIT_STOPS_WITH_ONGOING_COMPACTION,
+            1);
       }
       ROCKS_LOG_WARN(ioptions_.logger,
                      "[%s] Stopping writes because we have %d level-0 files",
@@ -974,7 +964,7 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
           SetupDelay(write_controller, compaction_needed_bytes,
                      prev_compaction_needed_bytes_, was_stopped,
                      mutable_cf_options.disable_auto_compactions);
-      internal_stats_->AddCFStats(InternalStats::MEMTABLE_LIMIT_SLOWDOWNS, 1);
+      internal_stats_->AddCFStats(InternalStats::MEMTABLE_LIMIT_DELAYS, 1);
       ROCKS_LOG_WARN(
           ioptions_.logger,
           "[%s] Stalling writes because we have %d immutable memtables "
@@ -992,11 +982,11 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
           SetupDelay(write_controller, compaction_needed_bytes,
                      prev_compaction_needed_bytes_, was_stopped || near_stop,
                      mutable_cf_options.disable_auto_compactions);
-      internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_SLOWDOWNS,
-                                  1);
+      internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_DELAYS, 1);
       if (compaction_picker_->IsLevel0CompactionInProgress()) {
         internal_stats_->AddCFStats(
-            InternalStats::LOCKED_L0_FILE_COUNT_LIMIT_SLOWDOWNS, 1);
+            InternalStats::L0_FILE_COUNT_LIMIT_DELAYS_WITH_ONGOING_COMPACTION,
+            1);
       }
       ROCKS_LOG_WARN(ioptions_.logger,
                      "[%s] Stalling writes because we have %d level-0 files "
@@ -1022,7 +1012,7 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
                      prev_compaction_needed_bytes_, was_stopped || near_stop,
                      mutable_cf_options.disable_auto_compactions);
       internal_stats_->AddCFStats(
-          InternalStats::PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS, 1);
+          InternalStats::PENDING_COMPACTION_BYTES_LIMIT_DELAYS, 1);
       ROCKS_LOG_WARN(
           ioptions_.logger,
           "[%s] Stalling writes because of estimated pending compaction "
@@ -1128,12 +1118,9 @@ bool ColumnFamilyData::NeedsCompaction() const {
 Compaction* ColumnFamilyData::PickCompaction(
     const MutableCFOptions& mutable_options,
     const MutableDBOptions& mutable_db_options, LogBuffer* log_buffer) {
-  SequenceNumber earliest_mem_seqno =
-      std::min(mem_->GetEarliestSequenceNumber(),
-               imm_.current()->GetEarliestSequenceNumber(false));
   auto* result = compaction_picker_->PickCompaction(
       GetName(), mutable_options, mutable_db_options, current_->storage_info(),
-      log_buffer, earliest_mem_seqno);
+      log_buffer);
   if (result != nullptr) {
     result->SetInputVersion(current_);
   }
@@ -1154,6 +1141,7 @@ Status ColumnFamilyData::RangesOverlapWithMemtables(
   *overlap = false;
   // Create an InternalIterator over all unflushed memtables
   Arena arena;
+  // TODO: plumb Env::IOActivity
   ReadOptions read_opts;
   read_opts.total_order_seek = true;
   MergeIteratorBuilder merge_iter_builder(&internal_comparator_, &arena);
@@ -1212,17 +1200,15 @@ Compaction* ColumnFamilyData::CompactRange(
     const InternalKey* begin, const InternalKey* end,
     InternalKey** compaction_end, bool* conflict,
     uint64_t max_file_num_to_ignore, const std::string& trim_ts) {
-  SequenceNumber earliest_mem_seqno =
-      std::min(mem_->GetEarliestSequenceNumber(),
-               imm_.current()->GetEarliestSequenceNumber(false));
   auto* result = compaction_picker_->CompactRange(
       GetName(), mutable_cf_options, mutable_db_options,
       current_->storage_info(), input_level, output_level,
       compact_range_options, begin, end, compaction_end, conflict,
-      max_file_num_to_ignore, trim_ts, earliest_mem_seqno);
+      max_file_num_to_ignore, trim_ts);
   if (result != nullptr) {
     result->SetInputVersion(current_);
   }
+  TEST_SYNC_POINT("ColumnFamilyData::CompactRange:Return");
   return result;
 }
 
@@ -1442,10 +1428,40 @@ Status ColumnFamilyData::ValidateOptions(
         "Memtable per key-value checksum protection only supports 0, 1, 2, 4 "
         "or 8 bytes per key.");
   }
+  if (std::find(supported.begin(), supported.end(),
+                cf_options.block_protection_bytes_per_key) == supported.end()) {
+    return Status::NotSupported(
+        "Block per key-value checksum protection only supports 0, 1, 2, 4 "
+        "or 8 bytes per key.");
+  }
+
+  if (!cf_options.compaction_options_fifo.file_temperature_age_thresholds
+           .empty()) {
+    if (cf_options.compaction_style != kCompactionStyleFIFO) {
+      return Status::NotSupported(
+          "Option file_temperature_age_thresholds only supports FIFO "
+          "compaction.");
+    } else if (cf_options.num_levels > 1) {
+      return Status::NotSupported(
+          "Option file_temperature_age_thresholds is only supported when "
+          "num_levels = 1.");
+    } else {
+      const auto& ages =
+          cf_options.compaction_options_fifo.file_temperature_age_thresholds;
+      assert(ages.size() >= 1);
+      // check that age is sorted
+      for (size_t i = 0; i < ages.size() - 1; ++i) {
+        if (ages[i].age >= ages[i + 1].age) {
+          return Status::NotSupported(
+              "Option file_temperature_age_thresholds requires elements to be "
+              "sorted in increasing order with respect to `age` field.");
+        }
+      }
+    }
+  }
   return s;
 }
 
-#ifndef ROCKSDB_LITE
 Status ColumnFamilyData::SetOptions(
     const DBOptions& db_opts,
     const std::unordered_map<std::string, std::string>& options_map) {
@@ -1464,7 +1480,6 @@ Status ColumnFamilyData::SetOptions(
   }
   return s;
 }
-#endif  // ROCKSDB_LITE
 
 // REQUIRES: DB mutex held
 Env::WriteLifeTimeHint ColumnFamilyData::CalculateSSTWriteHint(int level) {
@@ -1523,6 +1538,13 @@ FSDirectory* ColumnFamilyData::GetDataDir(size_t path_id) const {
   return data_dirs_[path_id].get();
 }
 
+void ColumnFamilyData::RecoverEpochNumbers() {
+  assert(current_);
+  auto* vstorage = current_->storage_info();
+  assert(vstorage);
+  vstorage->RecoverEpochNumbers(this);
+}
+
 ColumnFamilySet::ColumnFamilySet(const std::string& dbname,
                                  const ImmutableDBOptions* db_options,
                                  const FileOptions& file_options,

db/column_family.h

@@ -310,10 +310,6 @@ class ColumnFamilyData {
   void SetLogNumber(uint64_t log_number) { log_number_ = log_number; }
   uint64_t GetLogNumber() const { return log_number_; }
 
-  void SetFlushReason(FlushReason flush_reason) {
-    flush_reason_ = flush_reason;
-  }
-  FlushReason GetFlushReason() const { return flush_reason_; }
   // thread-safe
   const FileOptions* soptions() const;
   const ImmutableOptions* ioptions() const { return &ioptions_; }
@@ -339,12 +335,10 @@ class ColumnFamilyData {
   // Validate CF options against DB options
   static Status ValidateOptions(const DBOptions& db_options,
                                 const ColumnFamilyOptions& cf_options);
-#ifndef ROCKSDB_LITE
   // REQUIRES: DB mutex held
   Status SetOptions(
       const DBOptions& db_options,
       const std::unordered_map<std::string, std::string>& options_map);
-#endif  // ROCKSDB_LITE
 
   InternalStats* internal_stats() { return internal_stats_.get(); }
 
@@ -468,12 +462,6 @@ class ColumnFamilyData {
   bool queued_for_flush() { return queued_for_flush_; }
   bool queued_for_compaction() { return queued_for_compaction_; }
 
-  enum class WriteStallCause {
-    kNone,
-    kMemtableLimit,
-    kL0FileCountLimit,
-    kPendingCompactionBytes,
-  };
   static std::pair<WriteStallCondition, WriteStallCause>
   GetWriteStallConditionAndCause(
       int num_unflushed_memtables, int num_l0_files,
@@ -533,6 +521,24 @@ class ColumnFamilyData {
   void SetMempurgeUsed() { mempurge_used_ = true; }
   bool GetMempurgeUsed() { return mempurge_used_; }
 
+  // Allocate and return a new epoch number
+  uint64_t NewEpochNumber() { return next_epoch_number_.fetch_add(1); }
+
+  // Get the next epoch number to be assigned
+  uint64_t GetNextEpochNumber() const { return next_epoch_number_.load(); }
+
+  // Set the next epoch number to be assigned
+  void SetNextEpochNumber(uint64_t next_epoch_number) {
+    next_epoch_number_.store(next_epoch_number);
+  }
+
+  // Reset the next epoch number to be assigned
+  void ResetNextEpochNumber() { next_epoch_number_.store(1); }
+
+  // Recover the next epoch number of this CF and epoch number
+  // of its files (if missing)
+  void RecoverEpochNumbers();
+
  private:
   friend class ColumnFamilySet;
   ColumnFamilyData(uint32_t id, const std::string& name,
@@ -598,8 +604,6 @@ class ColumnFamilyData {
   // recovered from
   uint64_t log_number_;
 
-  std::atomic<FlushReason> flush_reason_;
-
   // An object that keeps all the compaction stats
   // and picks the next compaction
   std::unique_ptr<CompactionPicker> compaction_picker_;
@@ -634,6 +638,8 @@ class ColumnFamilyData {
   // a Version associated with this CFD
   std::shared_ptr<CacheReservationManager> file_metadata_cache_res_mgr_;
   bool mempurge_used_;
+
+  std::atomic<uint64_t> next_epoch_number_;
 };
 
 // ColumnFamilySet has interesting thread-safety requirements

db/column_family_test.cc

@@ -71,11 +71,7 @@ class ColumnFamilyTestBase : public testing::Test {
     for (auto h : handles_) {
       ColumnFamilyDescriptor cfdescriptor;
       Status s = h->GetDescriptor(&cfdescriptor);
-#ifdef ROCKSDB_LITE
-      EXPECT_TRUE(s.IsNotSupported());
-#else
       EXPECT_OK(s);
-#endif  // ROCKSDB_LITE
       column_families.push_back(cfdescriptor);
     }
     ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
@@ -197,12 +193,10 @@ class ColumnFamilyTestBase : public testing::Test {
                                &db_);
   }
 
-#ifndef ROCKSDB_LITE  // ReadOnlyDB is not supported
   void AssertOpenReadOnly(std::vector<std::string> cf,
                           std::vector<ColumnFamilyOptions> options = {}) {
     ASSERT_OK(OpenReadOnly(cf, options));
   }
-#endif  // !ROCKSDB_LITE
 
   void Open(std::vector<std::string> cf,
             std::vector<ColumnFamilyOptions> options = {}) {
@@ -224,27 +218,16 @@ class ColumnFamilyTestBase : public testing::Test {
   }
 
   bool IsDbWriteStopped() {
-#ifndef ROCKSDB_LITE
     uint64_t v;
     EXPECT_TRUE(dbfull()->GetIntProperty("rocksdb.is-write-stopped", &v));
     return (v == 1);
-#else
-    return dbfull()->TEST_write_controler().IsStopped();
-#endif  // !ROCKSDB_LITE
   }
 
   uint64_t GetDbDelayedWriteRate() {
-#ifndef ROCKSDB_LITE
     uint64_t v;
     EXPECT_TRUE(
         dbfull()->GetIntProperty("rocksdb.actual-delayed-write-rate", &v));
     return v;
-#else
-    if (!dbfull()->TEST_write_controler().NeedsDelay()) {
-      return 0;
-    }
-    return dbfull()->TEST_write_controler().delayed_write_rate();
-#endif  // !ROCKSDB_LITE
   }
 
   void Destroy(const std::vector<ColumnFamilyDescriptor>& column_families =
@@ -267,7 +250,6 @@ class ColumnFamilyTestBase : public testing::Test {
           db_->CreateColumnFamily(current_cf_opt, cfs[i], &handles_[cfi]));
       names_[cfi] = cfs[i];
 
-#ifndef ROCKSDB_LITE  // RocksDBLite does not support GetDescriptor
       // Verify the CF options of the returned CF handle.
       ColumnFamilyDescriptor desc;
       ASSERT_OK(handles_[cfi]->GetDescriptor(&desc));
@@ -276,7 +258,6 @@ class ColumnFamilyTestBase : public testing::Test {
       ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(
           ConfigOptions(), desc.options,
           SanitizeOptions(dbfull()->immutable_db_options(), current_cf_opt)));
-#endif  // !ROCKSDB_LITE
       cfi++;
     }
   }
@@ -325,7 +306,6 @@ class ColumnFamilyTestBase : public testing::Test {
     ASSERT_OK(db_->FlushWAL(/*sync=*/false));
   }
 
-#ifndef ROCKSDB_LITE  // TEST functions in DB are not supported in lite
   void WaitForFlush(int cf) {
     ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[cf]));
   }
@@ -339,7 +319,6 @@ class ColumnFamilyTestBase : public testing::Test {
   void AssertMaxTotalInMemoryState(uint64_t value) {
     ASSERT_EQ(value, MaxTotalInMemoryState());
   }
-#endif  // !ROCKSDB_LITE
 
   Status Put(int cf, const std::string& key, const std::string& value) {
     return db_->Put(WriteOptions(), handles_[cf], Slice(key), Slice(value));
@@ -377,7 +356,6 @@ class ColumnFamilyTestBase : public testing::Test {
                        "rocksdb.num-files-at-level" + std::to_string(level));
   }
 
-#ifndef ROCKSDB_LITE
   // Return spread of files per level
   std::string FilesPerLevel(int cf) {
     std::string result;
@@ -394,31 +372,19 @@ class ColumnFamilyTestBase : public testing::Test {
     result.resize(last_non_zero_offset);
     return result;
   }
-#endif
 
   void AssertFilesPerLevel(const std::string& value, int cf) {
-#ifndef ROCKSDB_LITE
     ASSERT_EQ(value, FilesPerLevel(cf));
-#else
-    (void)value;
-    (void)cf;
-#endif
   }
 
-#ifndef ROCKSDB_LITE  // GetLiveFilesMetaData is not supported
   int CountLiveFiles() {
     std::vector<LiveFileMetaData> metadata;
     db_->GetLiveFilesMetaData(&metadata);
     return static_cast<int>(metadata.size());
   }
-#endif  // !ROCKSDB_LITE
 
   void AssertCountLiveFiles(int expected_value) {
-#ifndef ROCKSDB_LITE
     ASSERT_EQ(expected_value, CountLiveFiles());
-#else
-    (void)expected_value;
-#endif
   }
 
   // Do n memtable flushes, each of which produces an sstable
@@ -432,7 +398,6 @@ class ColumnFamilyTestBase : public testing::Test {
     }
   }
 
-#ifndef ROCKSDB_LITE  // GetSortedWalFiles is not supported
   int CountLiveLogFiles() {
     int micros_wait_for_log_deletion = 20000;
     env_->SleepForMicroseconds(micros_wait_for_log_deletion);
@@ -461,25 +426,18 @@ class ColumnFamilyTestBase : public testing::Test {
     return ret;
     return 0;
   }
-#endif  // !ROCKSDB_LITE
 
   void AssertCountLiveLogFiles(int value) {
-#ifndef ROCKSDB_LITE  // GetSortedWalFiles is not supported
     ASSERT_EQ(value, CountLiveLogFiles());
-#else
-    (void)value;
-#endif  // !ROCKSDB_LITE
   }
 
   void AssertNumberOfImmutableMemtables(std::vector<int> num_per_cf) {
     assert(num_per_cf.size() == handles_.size());
 
-#ifndef ROCKSDB_LITE  // GetProperty is not supported in lite
     for (size_t i = 0; i < num_per_cf.size(); ++i) {
       ASSERT_EQ(num_per_cf[i], GetProperty(static_cast<int>(i),
                                            "rocksdb.num-immutable-mem-table"));
     }
-#endif  // !ROCKSDB_LITE
   }
 
   void CopyFile(const std::string& source, const std::string& destination,
@@ -575,7 +533,6 @@ TEST_P(ColumnFamilyTest, DontReuseColumnFamilyID) {
   }
 }
 
-#ifndef ROCKSDB_LITE
 TEST_P(ColumnFamilyTest, CreateCFRaceWithGetAggProperty) {
   Open();
 
@@ -598,7 +555,6 @@ TEST_P(ColumnFamilyTest, CreateCFRaceWithGetAggProperty) {
 
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
 }
-#endif  // !ROCKSDB_LITE
 
 class FlushEmptyCFTestWithParam
     : public ColumnFamilyTestBase,
@@ -942,7 +898,6 @@ TEST_P(ColumnFamilyTest, IgnoreRecoveredLog) {
   }
 }
 
-#ifndef ROCKSDB_LITE  // TEST functions used are not supported
 TEST_P(ColumnFamilyTest, FlushTest) {
   Open();
   CreateColumnFamiliesAndReopen({"one", "two"});
@@ -1057,7 +1012,6 @@ TEST_P(ColumnFamilyTest, LogDeletionTest) {
   AssertCountLiveLogFiles(4);
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_P(ColumnFamilyTest, CrashAfterFlush) {
   std::unique_ptr<FaultInjectionTestEnv> fault_env(
@@ -1097,7 +1051,6 @@ TEST_P(ColumnFamilyTest, OpenNonexistentColumnFamily) {
   ASSERT_TRUE(TryOpen({"default", "dne"}).IsInvalidArgument());
 }
 
-#ifndef ROCKSDB_LITE  // WaitForFlush() is not supported
 // Makes sure that obsolete log files get deleted
 TEST_P(ColumnFamilyTest, DifferentWriteBufferSizes) {
   // disable flushing stale column families
@@ -1205,14 +1158,12 @@ TEST_P(ColumnFamilyTest, DifferentWriteBufferSizes) {
   AssertCountLiveLogFiles(7);
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
 // The test is commented out because we want to test that snapshot is
 // not created for memtables not supported it, but There isn't a memtable
 // that doesn't support snapshot right now. If we have one later, we can
 // re-enable the test.
 //
-// #ifndef ROCKSDB_LITE  // Cuckoo is not supported in lite
 //   TEST_P(ColumnFamilyTest, MemtableNotSupportSnapshot) {
 //   db_options_.allow_concurrent_memtable_write = false;
 //   Open();
@@ -1232,7 +1183,6 @@ TEST_P(ColumnFamilyTest, DifferentWriteBufferSizes) {
 //   {second}); auto* s3 = dbfull()->GetSnapshot(); ASSERT_TRUE(s3 == nullptr);
 //   Close();
 // }
-// #endif  // !ROCKSDB_LITE
 
 class TestComparator : public Comparator {
   int Compare(const ROCKSDB_NAMESPACE::Slice& /*a*/,
@@ -1299,7 +1249,6 @@ TEST_P(ColumnFamilyTest, DifferentMergeOperators) {
   Close();
 }
 
-#ifndef ROCKSDB_LITE  // WaitForFlush() is not supported
 TEST_P(ColumnFamilyTest, DifferentCompactionStyles) {
   Open();
   CreateColumnFamilies({"one", "two"});
@@ -1367,9 +1316,7 @@ TEST_P(ColumnFamilyTest, DifferentCompactionStyles) {
 
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE
 // Sync points not supported in RocksDB Lite
 
 TEST_P(ColumnFamilyTest, MultipleManualCompactions) {
@@ -2033,9 +1980,7 @@ TEST_P(ColumnFamilyTest, SameCFAutomaticManualCompactions) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  // Tailing iterator not supported
 namespace {
 std::string IterStatus(Iterator* iter) {
   std::string result;
@@ -2093,9 +2038,7 @@ TEST_P(ColumnFamilyTest, NewIteratorsTest) {
     Destroy();
   }
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  // ReadOnlyDB is not supported
 TEST_P(ColumnFamilyTest, ReadOnlyDBTest) {
   Open();
   CreateColumnFamiliesAndReopen({"one", "two", "three", "four"});
@@ -2144,9 +2087,7 @@ TEST_P(ColumnFamilyTest, ReadOnlyDBTest) {
   s = OpenReadOnly({"one", "four"});
   ASSERT_TRUE(!s.ok());
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  //  WaitForFlush() is not supported in lite
 TEST_P(ColumnFamilyTest, DontRollEmptyLogs) {
   Open();
   CreateColumnFamiliesAndReopen({"one", "two", "three", "four"});
@@ -2168,9 +2109,7 @@ TEST_P(ColumnFamilyTest, DontRollEmptyLogs) {
   ASSERT_EQ(static_cast<size_t>(total_new_writable_files), handles_.size() + 1);
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  //  WaitForCompaction() is not supported in lite
 TEST_P(ColumnFamilyTest, FlushStaleColumnFamilies) {
   Open();
   CreateColumnFamilies({"one", "two"});
@@ -2217,7 +2156,6 @@ TEST_P(ColumnFamilyTest, FlushStaleColumnFamilies) {
   ASSERT_EQ(0, dbfull()->TEST_total_log_size());
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_P(ColumnFamilyTest, CreateMissingColumnFamilies) {
   Status s = TryOpen({"one", "two"});
@@ -2457,8 +2395,6 @@ TEST_P(ColumnFamilyTest, FlushAndDropRaceCondition) {
   Destroy();
 }
 
-#ifndef ROCKSDB_LITE
-// skipped as persisting options is not supported in ROCKSDB_LITE
 namespace {
 std::atomic<int> test_stage(0);
 std::atomic<bool> ordered_by_writethread(false);
@@ -2540,7 +2476,6 @@ TEST_P(ColumnFamilyTest, CreateAndDropRace) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_P(ColumnFamilyTest, WriteStallSingleColumnFamily) {
   const uint64_t kBaseRate = 800000u;
@@ -2950,7 +2885,6 @@ TEST_P(ColumnFamilyTest, CreateDropAndDestroy) {
   ASSERT_OK(db_->DestroyColumnFamilyHandle(cfh));
 }
 
-#ifndef ROCKSDB_LITE
 TEST_P(ColumnFamilyTest, CreateDropAndDestroyWithoutFileDeletion) {
   ColumnFamilyHandle* cfh;
   Open();
@@ -3005,9 +2939,7 @@ TEST_P(ColumnFamilyTest, FlushCloseWALFiles) {
   db_options_.env = env_;
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  // WaitForFlush() is not supported
 TEST_P(ColumnFamilyTest, IteratorCloseWALFile1) {
   SpecialEnv env(Env::Default());
   db_options_.env = &env;
@@ -3114,9 +3046,7 @@ TEST_P(ColumnFamilyTest, IteratorCloseWALFile2) {
   db_options_.env = env_;
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  // TEST functions are not supported in lite
 TEST_P(ColumnFamilyTest, ForwardIteratorCloseWALFile) {
   SpecialEnv env(Env::Default());
   // Allow both of flush and purge job to schedule.
@@ -3192,7 +3122,6 @@ TEST_P(ColumnFamilyTest, ForwardIteratorCloseWALFile) {
   db_options_.env = env_;
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
 // Disable on windows because SyncWAL requires env->IsSyncThreadSafe()
 // to return true which is not so in unbuffered mode.

db/compact_files_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-#ifndef ROCKSDB_LITE
 
 #include <mutex>
 #include <string>
@@ -121,7 +120,9 @@ TEST_F(CompactFilesTest, L0ConflictsFiles) {
 TEST_F(CompactFilesTest, MultipleLevel) {
   Options options;
   options.create_if_missing = true;
-  options.level_compaction_dynamic_level_bytes = true;
+  // Otherwise background compaction can happen to
+  // drain unnecessary level
+  options.level_compaction_dynamic_level_bytes = false;
   options.num_levels = 6;
   // Add listener
   FlushedFileCollector* collector = new FlushedFileCollector();
@@ -182,7 +183,6 @@ TEST_F(CompactFilesTest, MultipleLevel) {
   for (int invalid_output_level = 0; invalid_output_level < 5;
        invalid_output_level++) {
     s = db->CompactFiles(CompactionOptions(), files, invalid_output_level);
-    std::cout << s.ToString() << std::endl;
     ASSERT_TRUE(s.IsInvalidArgument());
   }
 
@@ -490,13 +490,3 @@ int main(int argc, char** argv) {
   return RUN_ALL_TESTS();
 }
 
-#else
-#include <stdio.h>
-
-int main(int /*argc*/, char** /*argv*/) {
-  fprintf(stderr,
-          "SKIPPED as DBImpl::CompactFiles is not supported in ROCKSDB_LITE\n");
-  return 0;
-}
-
-#endif  // !ROCKSDB_LITE

db/compaction/clipping_iterator.h

@@ -188,6 +188,11 @@ class ClippingIterator : public InternalIterator {
     return iter_->GetProperty(prop_name, prop);
   }
 
+  bool IsDeleteRangeSentinelKey() const override {
+    assert(valid_);
+    return iter_->IsDeleteRangeSentinelKey();
+  }
+
  private:
   void UpdateValid() {
     assert(!iter_->Valid() || iter_->status().ok());

db/compaction/compaction.cc

@@ -238,12 +238,19 @@ Compaction::Compaction(
       inputs_(PopulateWithAtomicBoundaries(vstorage, std::move(_inputs))),
       grandparents_(std::move(_grandparents)),
       score_(_score),
-      bottommost_level_(IsBottommostLevel(output_level_, vstorage, inputs_)),
+      bottommost_level_(
+          // For simplicity, we don't support the concept of "bottommost level"
+          // with
+          // `CompactionReason::kExternalSstIngestion` and
+          // `CompactionReason::kRefitLevel`
+          (_compaction_reason == CompactionReason::kExternalSstIngestion ||
+           _compaction_reason == CompactionReason::kRefitLevel)
+              ? false
+              : IsBottommostLevel(output_level_, vstorage, inputs_)),
       is_full_compaction_(IsFullCompaction(vstorage, inputs_)),
       is_manual_compaction_(_manual_compaction),
       trim_ts_(_trim_ts),
       is_trivial_move_(false),
-
       compaction_reason_(_compaction_reason),
       notify_on_compaction_completion_(false),
       enable_blob_garbage_collection_(
@@ -258,8 +265,15 @@ Compaction::Compaction(
                   _blob_garbage_collection_age_cutoff > 1
               ? mutable_cf_options()->blob_garbage_collection_age_cutoff
               : _blob_garbage_collection_age_cutoff),
-      penultimate_level_(EvaluatePenultimateLevel(
-          vstorage, immutable_options_, start_level_, output_level_)) {
+      penultimate_level_(
+          // For simplicity, we don't support the concept of "penultimate level"
+          // with `CompactionReason::kExternalSstIngestion` and
+          // `CompactionReason::kRefitLevel`
+          _compaction_reason == CompactionReason::kExternalSstIngestion ||
+                  _compaction_reason == CompactionReason::kRefitLevel
+              ? Compaction::kInvalidLevel
+              : EvaluatePenultimateLevel(vstorage, immutable_options_,
+                                         start_level_, output_level_)) {
   MarkFilesBeingCompacted(true);
   if (is_manual_compaction_) {
     compaction_reason_ = CompactionReason::kManualCompaction;
@@ -332,6 +346,7 @@ void Compaction::PopulatePenultimateLevelOutputRange() {
   // the case that the penultimate level is empty).
   if (immutable_options_.compaction_style == kCompactionStyleUniversal) {
     exclude_level = kInvalidLevel;
+    penultimate_output_range_type_ = PenultimateOutputRangeType::kFullRange;
     std::set<uint64_t> penultimate_inputs;
     for (const auto& input_lvl : inputs_) {
       if (input_lvl.level == penultimate_level_) {
@@ -345,7 +360,8 @@ void Compaction::PopulatePenultimateLevelOutputRange() {
       if (penultimate_inputs.find(file->fd.GetNumber()) ==
           penultimate_inputs.end()) {
         exclude_level = number_levels_ - 1;
-        penultimate_output_range_type_ = PenultimateOutputRangeType::kFullRange;
+        penultimate_output_range_type_ =
+            PenultimateOutputRangeType::kNonLastRange;
         break;
       }
     }
@@ -354,35 +370,6 @@ void Compaction::PopulatePenultimateLevelOutputRange() {
   GetBoundaryKeys(input_vstorage_, inputs_,
                   &penultimate_level_smallest_user_key_,
                   &penultimate_level_largest_user_key_, exclude_level);
-
-  // If there's a case that the penultimate level output range is overlapping
-  // with the existing files, disable the penultimate level output by setting
-  // the range to empty. One example is the range delete could have overlap
-  // boundary with the next file. (which is actually a false overlap)
-  // TODO: Exclude such false overlap, so it won't disable the penultimate
-  //  output.
-  std::set<uint64_t> penultimate_inputs;
-  for (const auto& input_lvl : inputs_) {
-    if (input_lvl.level == penultimate_level_) {
-      for (const auto& file : input_lvl.files) {
-        penultimate_inputs.emplace(file->fd.GetNumber());
-      }
-    }
-  }
-
-  auto penultimate_files = input_vstorage_->LevelFiles(penultimate_level_);
-  for (const auto& file : penultimate_files) {
-    if (penultimate_inputs.find(file->fd.GetNumber()) ==
-            penultimate_inputs.end() &&
-        OverlapPenultimateLevelOutputRange(file->smallest.user_key(),
-                                           file->largest.user_key())) {
-      // basically disable the penultimate range output. which should be rare
-      // or a false overlap caused by range del
-      penultimate_level_smallest_user_key_ = "";
-      penultimate_level_largest_user_key_ = "";
-      penultimate_output_range_type_ = PenultimateOutputRangeType::kDisabled;
-    }
-  }
 }
 
 Compaction::~Compaction() {
@@ -478,6 +465,11 @@ bool Compaction::IsTrivialMove() const {
     return false;
   }
 
+  if (compaction_reason_ == CompactionReason::kChangeTemperature) {
+    // Changing temperature usually requires rewriting the file.
+    return false;
+  }
+
   // Used in universal compaction, where trivial move can be done if the
   // input files are non overlapping
   if ((mutable_cf_options_.compaction_options_universal.allow_trivial_move) &&
@@ -807,6 +799,16 @@ uint64_t Compaction::MinInputFileOldestAncesterTime(
   return min_oldest_ancester_time;
 }
 
+uint64_t Compaction::MinInputFileEpochNumber() const {
+  uint64_t min_epoch_number = std::numeric_limits<uint64_t>::max();
+  for (const auto& inputs_per_level : inputs_) {
+    for (const auto& file : inputs_per_level.files) {
+      min_epoch_number = std::min(min_epoch_number, file->epoch_number);
+    }
+  }
+  return min_epoch_number;
+}
+
 int Compaction::EvaluatePenultimateLevel(
     const VersionStorageInfo* vstorage,
     const ImmutableOptions& immutable_options, const int start_level,

db/compaction/compaction.h

@@ -378,6 +378,9 @@ class Compaction {
   // This is used to filter out some input files' ancester's time range.
   uint64_t MinInputFileOldestAncesterTime(const InternalKey* start,
                                           const InternalKey* end) const;
+  // Return the minimum epoch number among
+  // input files' associated with this compaction
+  uint64_t MinInputFileEpochNumber() const;
 
   // Called by DBImpl::NotifyOnCompactionCompleted to make sure number of
   // compaction begin and compaction completion callbacks match.
@@ -543,13 +546,16 @@ struct PerKeyPlacementContext {
   const Slice value;
   const SequenceNumber seq_num;
 
-  bool output_to_penultimate_level;
+  bool& output_to_penultimate_level;
 
   PerKeyPlacementContext(int _level, Slice _key, Slice _value,
-                         SequenceNumber _seq_num)
-      : level(_level), key(_key), value(_value), seq_num(_seq_num) {
-    output_to_penultimate_level = false;
-  }
+                         SequenceNumber _seq_num,
+                         bool& _output_to_penultimate_level)
+      : level(_level),
+        key(_key),
+        value(_value),
+        seq_num(_seq_num),
+        output_to_penultimate_level(_output_to_penultimate_level) {}
 };
 #endif /* !NDEBUG */
 

db/compaction/compaction_iterator.cc

@@ -13,6 +13,7 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/prefetch_buffer_collection.h"
 #include "db/snapshot_checker.h"
+#include "db/wide/wide_column_serialization.h"
 #include "logging/logging.h"
 #include "port/likely.h"
 #include "rocksdb/listener.h"
@@ -124,6 +125,9 @@ CompactionIterator::CompactionIterator(
          timestamp_size_ == full_history_ts_low_->size());
 #endif
   input_.SetPinnedItersMgr(&pinned_iters_mgr_);
+  // The default `merge_until_status_` does not need to be checked since it is
+  // overwritten as soon as `MergeUntil()` is called
+  merge_until_status_.PermitUncheckedError();
   TEST_SYNC_POINT_CALLBACK("CompactionIterator:AfterInit", compaction_.get());
 }
 
@@ -178,6 +182,20 @@ void CompactionIterator::Next() {
       ikey_.user_key = current_key_.GetUserKey();
       validity_info_.SetValid(ValidContext::kMerge1);
     } else {
+      if (merge_until_status_.IsMergeInProgress()) {
+        // `Status::MergeInProgress()` tells us that the previous `MergeUntil()`
+        // produced only merge operands. Those merge operands were accessed and
+        // written out using `merge_out_iter_`. Since `merge_out_iter_` is
+        // exhausted at this point, all merge operands have been written out.
+        //
+        // Still, there may be a base value (PUT, DELETE, SINGLEDEL, etc.) that
+        // needs to be written out. Normally, `CompactionIterator` would skip it
+        // on the basis that it has already output something in the same
+        // snapshot stripe. To prevent this, we reset `has_current_user_key_` to
+        // trick the future iteration from finding out the snapshot stripe is
+        // unchanged.
+        has_current_user_key_ = false;
+      }
       // We consumed all pinned merge operands, release pinned iterators
       pinned_iters_mgr_.ReleasePinnedData();
       // MergeHelper moves the iterator to the first record after the merged
@@ -204,39 +222,47 @@ void CompactionIterator::Next() {
 
 bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
                                               Slice* skip_until) {
-  // TODO: support compaction filter for wide-column entities
-  if (!compaction_filter_ ||
-      (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex)) {
+  if (!compaction_filter_) {
     return true;
   }
-  bool error = false;
-  // If the user has specified a compaction filter and the sequence
-  // number is greater than any external snapshot, then invoke the
-  // filter. If the return value of the compaction filter is true,
-  // replace the entry with a deletion marker.
-  CompactionFilter::Decision filter = CompactionFilter::Decision::kUndetermined;
-  compaction_filter_value_.clear();
-  compaction_filter_skip_until_.Clear();
+
+  if (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex &&
+      ikey_.type != kTypeWideColumnEntity) {
+    return true;
+  }
+
+  CompactionFilter::Decision decision =
+      CompactionFilter::Decision::kUndetermined;
   CompactionFilter::ValueType value_type =
       ikey_.type == kTypeValue ? CompactionFilter::ValueType::kValue
-                               : CompactionFilter::ValueType::kBlobIndex;
+      : ikey_.type == kTypeBlobIndex
+          ? CompactionFilter::ValueType::kBlobIndex
+          : CompactionFilter::ValueType::kWideColumnEntity;
+
   // Hack: pass internal key to BlobIndexCompactionFilter since it needs
   // to get sequence number.
   assert(compaction_filter_);
-  Slice& filter_key =
-      (ikey_.type == kTypeValue ||
+  const Slice& filter_key =
+      (ikey_.type != kTypeBlobIndex ||
        !compaction_filter_->IsStackedBlobDbInternalCompactionFilter())
           ? ikey_.user_key
           : key_;
+
+  compaction_filter_value_.clear();
+  compaction_filter_skip_until_.Clear();
+
+  std::vector<std::pair<std::string, std::string>> new_columns;
+
   {
     StopWatchNano timer(clock_, report_detailed_time_);
-    if (kTypeBlobIndex == ikey_.type) {
-      filter = compaction_filter_->FilterBlobByKey(
+
+    if (ikey_.type == kTypeBlobIndex) {
+      decision = compaction_filter_->FilterBlobByKey(
           level_, filter_key, &compaction_filter_value_,
           compaction_filter_skip_until_.rep());
-      if (CompactionFilter::Decision::kUndetermined == filter &&
+      if (decision == CompactionFilter::Decision::kUndetermined &&
           !compaction_filter_->IsStackedBlobDbInternalCompactionFilter()) {
-        if (compaction_ == nullptr) {
+        if (!compaction_) {
           status_ =
               Status::Corruption("Unexpected blob index outside of compaction");
           validity_info_.Invalidate();
@@ -282,33 +308,61 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
         value_type = CompactionFilter::ValueType::kValue;
       }
     }
-    if (CompactionFilter::Decision::kUndetermined == filter) {
-      filter = compaction_filter_->FilterV2(
-          level_, filter_key, value_type,
-          blob_value_.empty() ? value_ : blob_value_, &compaction_filter_value_,
+
+    if (decision == CompactionFilter::Decision::kUndetermined) {
+      const Slice* existing_val = nullptr;
+      const WideColumns* existing_col = nullptr;
+
+      WideColumns existing_columns;
+
+      if (ikey_.type != kTypeWideColumnEntity) {
+        if (!blob_value_.empty()) {
+          existing_val = &blob_value_;
+        } else {
+          existing_val = &value_;
+        }
+      } else {
+        Slice value_copy = value_;
+        const Status s =
+            WideColumnSerialization::Deserialize(value_copy, existing_columns);
+
+        if (!s.ok()) {
+          status_ = s;
+          validity_info_.Invalidate();
+          return false;
+        }
+
+        existing_col = &existing_columns;
+      }
+
+      decision = compaction_filter_->FilterV3(
+          level_, filter_key, value_type, existing_val, existing_col,
+          &compaction_filter_value_, &new_columns,
           compaction_filter_skip_until_.rep());
     }
+
     iter_stats_.total_filter_time +=
         env_ != nullptr && report_detailed_time_ ? timer.ElapsedNanos() : 0;
   }
 
-  if (CompactionFilter::Decision::kUndetermined == filter) {
-    // Should not reach here, since FilterV2 should never return kUndetermined.
-    status_ =
-        Status::NotSupported("FilterV2() should never return kUndetermined");
+  if (decision == CompactionFilter::Decision::kUndetermined) {
+    // Should not reach here, since FilterV2/FilterV3 should never return
+    // kUndetermined.
+    status_ = Status::NotSupported(
+        "FilterV2/FilterV3 should never return kUndetermined");
     validity_info_.Invalidate();
     return false;
   }
 
-  if (filter == CompactionFilter::Decision::kRemoveAndSkipUntil &&
+  if (decision == CompactionFilter::Decision::kRemoveAndSkipUntil &&
       cmp_->Compare(*compaction_filter_skip_until_.rep(), ikey_.user_key) <=
           0) {
     // Can't skip to a key smaller than the current one.
-    // Keep the key as per FilterV2 documentation.
-    filter = CompactionFilter::Decision::kKeep;
+    // Keep the key as per FilterV2/FilterV3 documentation.
+    decision = CompactionFilter::Decision::kKeep;
   }
 
-  if (filter == CompactionFilter::Decision::kRemove) {
+  if (decision == CompactionFilter::Decision::kRemove) {
     // convert the current key to a delete; key_ is pointing into
     // current_key_ at this point, so updating current_key_ updates key()
     ikey_.type = kTypeDeletion;
@@ -316,7 +370,7 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
     // no value associated with delete
     value_.clear();
     iter_stats_.num_record_drop_user++;
-  } else if (filter == CompactionFilter::Decision::kPurge) {
+  } else if (decision == CompactionFilter::Decision::kPurge) {
     // convert the current key to a single delete; key_ is pointing into
     // current_key_ at this point, so updating current_key_ updates key()
     ikey_.type = kTypeSingleDeletion;
@@ -324,19 +378,19 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
     // no value associated with single delete
     value_.clear();
     iter_stats_.num_record_drop_user++;
-  } else if (filter == CompactionFilter::Decision::kChangeValue) {
-    if (ikey_.type == kTypeBlobIndex) {
-      // value transfer from blob file to inlined data
+  } else if (decision == CompactionFilter::Decision::kChangeValue) {
+    if (ikey_.type != kTypeValue) {
       ikey_.type = kTypeValue;
-      current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+      current_key_.UpdateInternalKey(ikey_.sequence, kTypeValue);
     }
+
     value_ = compaction_filter_value_;
-  } else if (filter == CompactionFilter::Decision::kRemoveAndSkipUntil) {
+  } else if (decision == CompactionFilter::Decision::kRemoveAndSkipUntil) {
     *need_skip = true;
     compaction_filter_skip_until_.ConvertFromUserKey(kMaxSequenceNumber,
                                                      kValueTypeForSeek);
     *skip_until = compaction_filter_skip_until_.Encode();
-  } else if (filter == CompactionFilter::Decision::kChangeBlobIndex) {
+  } else if (decision == CompactionFilter::Decision::kChangeBlobIndex) {
     // Only the StackableDB-based BlobDB impl's compaction filter should return
     // kChangeBlobIndex. Decision about rewriting blob and changing blob index
     // in the integrated BlobDB impl is made in subsequent call to
@@ -348,23 +402,56 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
       validity_info_.Invalidate();
       return false;
     }
-    if (ikey_.type == kTypeValue) {
-      // value transfer from inlined data to blob file
+
+    if (ikey_.type != kTypeBlobIndex) {
       ikey_.type = kTypeBlobIndex;
-      current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+      current_key_.UpdateInternalKey(ikey_.sequence, kTypeBlobIndex);
     }
+
     value_ = compaction_filter_value_;
-  } else if (filter == CompactionFilter::Decision::kIOError) {
+  } else if (decision == CompactionFilter::Decision::kIOError) {
     if (!compaction_filter_->IsStackedBlobDbInternalCompactionFilter()) {
       status_ = Status::NotSupported(
           "CompactionFilter for integrated BlobDB should not return kIOError");
       validity_info_.Invalidate();
       return false;
     }
+
     status_ = Status::IOError("Failed to access blob during compaction filter");
-    error = true;
+    validity_info_.Invalidate();
+    return false;
+  } else if (decision == CompactionFilter::Decision::kChangeWideColumnEntity) {
+    WideColumns sorted_columns;
+
+    sorted_columns.reserve(new_columns.size());
+    for (const auto& column : new_columns) {
+      sorted_columns.emplace_back(column.first, column.second);
+    }
+
+    std::sort(sorted_columns.begin(), sorted_columns.end(),
+              [](const WideColumn& lhs, const WideColumn& rhs) {
+                return lhs.name().compare(rhs.name()) < 0;
+              });
+
+    {
+      const Status s = WideColumnSerialization::Serialize(
+          sorted_columns, compaction_filter_value_);
+      if (!s.ok()) {
+        status_ = s;
+        validity_info_.Invalidate();
+        return false;
+      }
+    }
+
+    if (ikey_.type != kTypeWideColumnEntity) {
+      ikey_.type = kTypeWideColumnEntity;
+      current_key_.UpdateInternalKey(ikey_.sequence, kTypeWideColumnEntity);
+    }
+
+    value_ = compaction_filter_value_;
   }
-  return !error;
+
+  return true;
 }
 
 void CompactionIterator::NextFromInput() {
@@ -377,6 +464,7 @@ void CompactionIterator::NextFromInput() {
     value_ = input_.value();
     blob_value_.Reset();
     iter_stats_.num_input_records++;
+    is_range_del_ = input_.IsDeleteRangeSentinelKey();
 
     Status pik_status = ParseInternalKey(key_, &ikey_, allow_data_in_errors_);
     if (!pik_status.ok()) {
@@ -396,7 +484,10 @@ void CompactionIterator::NextFromInput() {
       break;
     }
     TEST_SYNC_POINT_CALLBACK("CompactionIterator:ProcessKV", &ikey_);
-
+    if (is_range_del_) {
+      validity_info_.SetValid(kRangeDeletion);
+      break;
+    }
     // Update input statistics
     if (ikey_.type == kTypeDeletion || ikey_.type == kTypeSingleDeletion ||
         ikey_.type == kTypeDeletionWithTimestamp) {
@@ -618,6 +709,14 @@ void CompactionIterator::NextFromInput() {
 
       ParsedInternalKey next_ikey;
       AdvanceInputIter();
+      while (input_.Valid() && input_.IsDeleteRangeSentinelKey() &&
+             ParseInternalKey(input_.key(), &next_ikey, allow_data_in_errors_)
+                 .ok() &&
+             cmp_->EqualWithoutTimestamp(ikey_.user_key, next_ikey.user_key)) {
+        // skip range tombstone start keys with the same user key
+        // since they are not "real" point keys.
+        AdvanceInputIter();
+      }
 
       // Check whether the next key exists, is not corrupt, and is the same key
       // as the single delete.
@@ -625,6 +724,7 @@ void CompactionIterator::NextFromInput() {
           ParseInternalKey(input_.key(), &next_ikey, allow_data_in_errors_)
               .ok() &&
           cmp_->EqualWithoutTimestamp(ikey_.user_key, next_ikey.user_key)) {
+        assert(!input_.IsDeleteRangeSentinelKey());
 #ifndef NDEBUG
         const Compaction* c =
             compaction_ ? compaction_->real_compaction() : nullptr;
@@ -849,12 +949,14 @@ void CompactionIterator::NextFromInput() {
       // Note that a deletion marker of type kTypeDeletionWithTimestamp will be
       // considered to have a different user key unless the timestamp is older
       // than *full_history_ts_low_.
+      //
+      // Range tombstone start keys are skipped as they are not "real" keys.
       while (!IsPausingManualCompaction() && !IsShuttingDown() &&
              input_.Valid() &&
              (ParseInternalKey(input_.key(), &next_ikey, allow_data_in_errors_)
                   .ok()) &&
              cmp_->EqualWithoutTimestamp(ikey_.user_key, next_ikey.user_key) &&
-             (prev_snapshot == 0 ||
+             (prev_snapshot == 0 || input_.IsDeleteRangeSentinelKey() ||
               DefinitelyNotInSnapshot(next_ikey.sequence, prev_snapshot))) {
         AdvanceInputIter();
       }
@@ -880,14 +982,15 @@ void CompactionIterator::NextFromInput() {
       // have hit (A)
       // We encapsulate the merge related state machine in a different
       // object to minimize change to the existing flow.
-      Status s = merge_helper_->MergeUntil(
+      merge_until_status_ = merge_helper_->MergeUntil(
           &input_, range_del_agg_, prev_snapshot, bottommost_level_,
           allow_data_in_errors_, blob_fetcher_.get(), full_history_ts_low_,
           prefetch_buffers_.get(), &iter_stats_);
       merge_out_iter_.SeekToFirst();
 
-      if (!s.ok() && !s.IsMergeInProgress()) {
-        status_ = s;
+      if (!merge_until_status_.ok() &&
+          !merge_until_status_.IsMergeInProgress()) {
+        status_ = merge_until_status_;
         return;
       } else if (merge_out_iter_.Valid()) {
         // NOTE: key, value, and ikey_ refer to old entries.
@@ -1098,17 +1201,7 @@ void CompactionIterator::GarbageCollectBlobIfNeeded() {
 
 void CompactionIterator::DecideOutputLevel() {
   assert(compaction_->SupportsPerKeyPlacement());
-#ifndef NDEBUG
-  // Could be overridden by unittest
-  PerKeyPlacementContext context(level_, ikey_.user_key, value_,
-                                 ikey_.sequence);
-  TEST_SYNC_POINT_CALLBACK("CompactionIterator::PrepareOutput.context",
-                           &context);
-  output_to_penultimate_level_ = context.output_to_penultimate_level;
-#else
   output_to_penultimate_level_ = false;
-#endif  // NDEBUG
-
   // if the key is newer than the cutoff sequence or within the earliest
   // snapshot, it should output to the penultimate level.
   if (ikey_.sequence > preclude_last_level_min_seqno_ ||
@@ -1116,6 +1209,17 @@ void CompactionIterator::DecideOutputLevel() {
     output_to_penultimate_level_ = true;
   }
 
+#ifndef NDEBUG
+  // Could be overridden by unittest
+  PerKeyPlacementContext context(level_, ikey_.user_key, value_, ikey_.sequence,
+                                 output_to_penultimate_level_);
+  TEST_SYNC_POINT_CALLBACK("CompactionIterator::PrepareOutput.context",
+                           &context);
+  if (ikey_.sequence > earliest_snapshot_) {
+    output_to_penultimate_level_ = true;
+  }
+#endif  // NDEBUG
+
   if (output_to_penultimate_level_) {
     // If it's decided to output to the penultimate level, but unsafe to do so,
     // still output to the last level. For example, moving the data from a lower
@@ -1147,10 +1251,12 @@ void CompactionIterator::DecideOutputLevel() {
 
 void CompactionIterator::PrepareOutput() {
   if (Valid()) {
-    if (ikey_.type == kTypeValue) {
-      ExtractLargeValueIfNeeded();
-    } else if (ikey_.type == kTypeBlobIndex) {
-      GarbageCollectBlobIfNeeded();
+    if (LIKELY(!is_range_del_)) {
+      if (ikey_.type == kTypeValue) {
+        ExtractLargeValueIfNeeded();
+      } else if (ikey_.type == kTypeBlobIndex) {
+        GarbageCollectBlobIfNeeded();
+      }
     }
 
     if (compaction_ != nullptr && compaction_->SupportsPerKeyPlacement()) {
@@ -1173,7 +1279,7 @@ void CompactionIterator::PrepareOutput() {
         DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
         ikey_.type != kTypeMerge && current_key_committed_ &&
         !output_to_penultimate_level_ &&
-        ikey_.sequence < preserve_time_min_seqno_) {
+        ikey_.sequence < preserve_time_min_seqno_ && !is_range_del_) {
       if (ikey_.type == kTypeDeletion ||
           (ikey_.type == kTypeSingleDeletion && timestamp_size_ == 0)) {
         ROCKS_LOG_FATAL(
@@ -1306,6 +1412,7 @@ std::unique_ptr<BlobFetcher> CompactionIterator::CreateBlobFetcherIfNeeded(
   }
 
   ReadOptions read_options;
+  read_options.io_activity = Env::IOActivity::kCompaction;
   read_options.fill_cache = false;
 
   return std::unique_ptr<BlobFetcher>(new BlobFetcher(version, read_options));

db/compaction/compaction_iterator.h

@@ -63,6 +63,10 @@ class SequenceIterWrapper : public InternalIterator {
   void SeekToLast() override { assert(false); }
 
   uint64_t num_itered() const { return num_itered_; }
+  bool IsDeleteRangeSentinelKey() const override {
+    assert(Valid());
+    return inner_iter_->IsDeleteRangeSentinelKey();
+  }
 
  private:
   InternalKeyComparator icmp_;
@@ -242,7 +246,12 @@ class CompactionIterator {
   const Status& status() const { return status_; }
   const ParsedInternalKey& ikey() const { return ikey_; }
   inline bool Valid() const { return validity_info_.IsValid(); }
-  const Slice& user_key() const { return current_user_key_; }
+  const Slice& user_key() const {
+    if (UNLIKELY(is_range_del_)) {
+      return ikey_.user_key;
+    }
+    return current_user_key_;
+  }
   const CompactionIterationStats& iter_stats() const { return iter_stats_; }
   uint64_t num_input_entry_scanned() const { return input_.num_itered(); }
   // If the current key should be placed on penultimate level, only valid if
@@ -252,6 +261,8 @@ class CompactionIterator {
   }
   Status InputStatus() const { return input_.status(); }
 
+  bool IsDeleteRangeSentinelKey() const { return is_range_del_; }
+
  private:
   // Processes the input stream to find the next output
   void NextFromInput();
@@ -385,6 +396,7 @@ class CompactionIterator {
     kKeepSD = 8,
     kKeepDel = 9,
     kNewUserKey = 10,
+    kRangeDeletion = 11,
   };
 
   struct ValidityInfo {
@@ -432,6 +444,7 @@ class CompactionIterator {
   bool clear_and_output_next_key_ = false;
 
   MergeOutputIterator merge_out_iter_;
+  Status merge_until_status_;
   // PinnedIteratorsManager used to pin input_ Iterator blocks while reading
   // merge operands and then releasing them after consuming them.
   PinnedIteratorsManager pinned_iters_mgr_;
@@ -492,6 +505,10 @@ class CompactionIterator {
     // This is a best-effort facility, so memory_order_relaxed is sufficient.
     return manual_compaction_canceled_.load(std::memory_order_relaxed);
   }
+
+  // Stores whether the current compaction iterator output
+  // is a range tombstone start key.
+  bool is_range_del_{false};
 };
 
 inline bool CompactionIterator::DefinitelyInSnapshot(SequenceNumber seq,

db/compaction/compaction_job.cc

@@ -99,6 +99,8 @@ const char* GetCompactionReasonString(CompactionReason compaction_reason) {
       return "ForcedBlobGC";
     case CompactionReason::kRoundRobinTtl:
       return "RoundRobinTtl";
+    case CompactionReason::kRefitLevel:
+      return "RefitLevel";
     case CompactionReason::kNumOfReasons:
       // fall through
     default:
@@ -190,8 +192,8 @@ CompactionJob::CompactionJob(
   assert(log_buffer_ != nullptr);
 
   const auto* cfd = compact_->compaction->column_family_data();
-  ThreadStatusUtil::SetColumnFamily(cfd, cfd->ioptions()->env,
-                                    db_options_.enable_thread_tracking);
+  ThreadStatusUtil::SetEnableTracking(db_options_.enable_thread_tracking);
+  ThreadStatusUtil::SetColumnFamily(cfd);
   ThreadStatusUtil::SetThreadOperation(ThreadStatus::OP_COMPACTION);
   ReportStartedCompaction(compaction);
 }
@@ -202,10 +204,6 @@ CompactionJob::~CompactionJob() {
 }
 
 void CompactionJob::ReportStartedCompaction(Compaction* compaction) {
-  const auto* cfd = compact_->compaction->column_family_data();
-  ThreadStatusUtil::SetColumnFamily(cfd, cfd->ioptions()->env,
-                                    db_options_.enable_thread_tracking);
-
   ThreadStatusUtil::SetThreadOperationProperty(ThreadStatus::COMPACTION_JOB_ID,
                                                job_id_);
 
@@ -289,12 +287,14 @@ void CompactionJob::Prepare() {
                c->immutable_options()->preclude_last_level_data_seconds);
 
   if (preserve_time_duration > 0) {
+    const ReadOptions read_options(Env::IOActivity::kCompaction);
     // setup seqno_time_mapping_
     seqno_time_mapping_.SetMaxTimeDuration(preserve_time_duration);
     for (const auto& each_level : *c->inputs()) {
       for (const auto& fmd : each_level.files) {
         std::shared_ptr<const TableProperties> tp;
-        Status s = cfd->current()->GetTableProperties(&tp, fmd, nullptr);
+        Status s =
+            cfd->current()->GetTableProperties(read_options, &tp, fmd, nullptr);
         if (s.ok()) {
           seqno_time_mapping_.Add(tp->seqno_to_time_mapping)
               .PermitUncheckedError();
@@ -470,7 +470,7 @@ void CompactionJob::GenSubcompactionBoundaries() {
   // overlap with N-1 other ranges. Since we requested a relatively large number
   // (128) of ranges from each input files, even N range overlapping would
   // cause relatively small inaccuracy.
-
+  const ReadOptions read_options(Env::IOActivity::kCompaction);
   auto* c = compact_->compaction;
   if (c->max_subcompactions() <= 1 &&
       !(c->immutable_options()->compaction_pri == kRoundRobin &&
@@ -504,7 +504,9 @@ void CompactionJob::GenSubcompactionBoundaries() {
         FileMetaData* f = flevel->files[i].file_metadata;
         std::vector<TableReader::Anchor> my_anchors;
         Status s = cfd->table_cache()->ApproximateKeyAnchors(
-            ReadOptions(), icomp, *f, my_anchors);
+            read_options, icomp, *f,
+            c->mutable_cf_options()->block_protection_bytes_per_key,
+            my_anchors);
         if (!s.ok() || my_anchors.empty()) {
           my_anchors.emplace_back(f->largest.user_key(), f->fd.GetFileSize());
         }
@@ -720,11 +722,12 @@ Status CompactionJob::Run() {
         // use_direct_io_for_flush_and_compaction is true, we will regard this
         // verification as user reads since the goal is to cache it here for
         // further user reads
-        ReadOptions read_options;
+        const ReadOptions verify_table_read_options(
+            Env::IOActivity::kCompaction);
         InternalIterator* iter = cfd->table_cache()->NewIterator(
-            read_options, file_options_, cfd->internal_comparator(),
-            files_output[file_idx]->meta, /*range_del_agg=*/nullptr,
-            prefix_extractor,
+            verify_table_read_options, file_options_,
+            cfd->internal_comparator(), files_output[file_idx]->meta,
+            /*range_del_agg=*/nullptr, prefix_extractor,
             /*table_reader_ptr=*/nullptr,
             cfd->internal_stats()->GetFileReadHist(
                 compact_->compaction->output_level()),
@@ -734,7 +737,9 @@ Status CompactionJob::Run() {
                 *compact_->compaction->mutable_cf_options()),
             /*smallest_compaction_key=*/nullptr,
             /*largest_compaction_key=*/nullptr,
-            /*allow_unprepared_value=*/false);
+            /*allow_unprepared_value=*/false,
+            compact_->compaction->mutable_cf_options()
+                ->block_protection_bytes_per_key);
         auto s = iter->status();
 
         if (s.ok() && paranoid_file_checks_) {
@@ -976,7 +981,6 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options) {
 
 void CompactionJob::NotifyOnSubcompactionBegin(
     SubcompactionState* sub_compact) {
-#ifndef ROCKSDB_LITE
   Compaction* c = compact_->compaction;
 
   if (db_options_.listeners.empty()) {
@@ -1002,14 +1006,10 @@ void CompactionJob::NotifyOnSubcompactionBegin(
   }
   info.status.PermitUncheckedError();
 
-#else
-  (void)sub_compact;
-#endif  // ROCKSDB_LITE
 }
 
 void CompactionJob::NotifyOnSubcompactionCompleted(
     SubcompactionState* sub_compact) {
-#ifndef ROCKSDB_LITE
 
   if (db_options_.listeners.empty()) {
     return;
@@ -1030,16 +1030,11 @@ void CompactionJob::NotifyOnSubcompactionCompleted(
   for (const auto& listener : db_options_.listeners) {
     listener->OnSubcompactionCompleted(info);
   }
-#else
-  (void)sub_compact;
-#endif  // ROCKSDB_LITE
 }
 
 void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   assert(sub_compact);
   assert(sub_compact->compaction);
-
-#ifndef ROCKSDB_LITE
   if (db_options_.compaction_service) {
     CompactionServiceJobStatus comp_status =
         ProcessKeyValueCompactionWithCompactionService(sub_compact);
@@ -1050,7 +1045,6 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     // fallback to local compaction
     assert(comp_status == CompactionServiceJobStatus::kUseLocal);
   }
-#endif  // !ROCKSDB_LITE
 
   uint64_t prev_cpu_micros = db_options_.clock->CPUMicros();
 
@@ -1091,6 +1085,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   read_options.verify_checksums = true;
   read_options.fill_cache = false;
   read_options.rate_limiter_priority = GetRateLimiterPriority();
+  read_options.io_activity = Env::IOActivity::kCompaction;
   // Compaction iterators shouldn't be confined to a single prefix.
   // Compactions use Seek() for
   // (a) concurrent compactions,
@@ -1126,6 +1121,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   IterKey end_ikey;
   Slice start_slice;
   Slice end_slice;
+  Slice start_user_key{};
+  Slice end_user_key{};
 
   static constexpr char kMaxTs[] =
       "\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff";
@@ -1148,6 +1145,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
                                    &ts_slice);
     }
     start_slice = start_ikey.GetInternalKey();
+    start_user_key = start_ikey.GetUserKey();
   }
   if (end.has_value()) {
     end_ikey.SetInternalKey(end.value(), kMaxSequenceNumber, kValueTypeForSeek);
@@ -1156,6 +1154,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
                                  &ts_slice);
     }
     end_slice = end_ikey.GetInternalKey();
+    end_user_key = end_ikey.GetUserKey();
   }
 
   std::unique_ptr<InternalIterator> clip;
@@ -1271,11 +1270,15 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       [this, sub_compact](CompactionOutputs& outputs) {
         return this->OpenCompactionOutputFile(sub_compact, outputs);
       };
+
   const CompactionFileCloseFunc close_file_func =
-      [this, sub_compact](CompactionOutputs& outputs, const Status& status,
-                          const Slice& next_table_min_key) {
-        return this->FinishCompactionOutputFile(status, sub_compact, outputs,
-                                                next_table_min_key);
+      [this, sub_compact, start_user_key, end_user_key](
+          CompactionOutputs& outputs, const Status& status,
+          const Slice& next_table_min_key) {
+        return this->FinishCompactionOutputFile(
+            status, sub_compact, outputs, next_table_min_key,
+            sub_compact->start.has_value() ? &start_user_key : nullptr,
+            sub_compact->end.has_value() ? &end_user_key : nullptr);
       };
 
   Status status;
@@ -1286,7 +1289,6 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   while (status.ok() && !cfd->IsDropped() && c_iter->Valid()) {
     // Invariant: c_iter.status() is guaranteed to be OK if c_iter->Valid()
     // returns true.
-
     assert(!end.has_value() || cfd->user_comparator()->Compare(
                                    c_iter->user_key(), end.value()) < 0);
 
@@ -1466,7 +1468,8 @@ void CompactionJob::RecordDroppedKeys(
 
 Status CompactionJob::FinishCompactionOutputFile(
     const Status& input_status, SubcompactionState* sub_compact,
-    CompactionOutputs& outputs, const Slice& next_table_min_key) {
+    CompactionOutputs& outputs, const Slice& next_table_min_key,
+    const Slice* comp_start_user_key, const Slice* comp_end_user_key) {
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_COMPACTION_SYNC_FILE);
   assert(sub_compact != nullptr);
@@ -1496,12 +1499,10 @@ Status CompactionJob::FinishCompactionOutputFile(
     // output_to_penultimate_level compaction here, as it's only used to decide
     // if range dels could be dropped.
     if (outputs.HasRangeDel()) {
-      s = outputs.AddRangeDels(
-          sub_compact->start.has_value() ? &(sub_compact->start.value())
-                                         : nullptr,
-          sub_compact->end.has_value() ? &(sub_compact->end.value()) : nullptr,
-          range_del_out_stats, bottommost_level_, cfd->internal_comparator(),
-          earliest_snapshot, next_table_min_key, full_history_ts_low_);
+      s = outputs.AddRangeDels(comp_start_user_key, comp_end_user_key,
+                               range_del_out_stats, bottommost_level_,
+                               cfd->internal_comparator(), earliest_snapshot,
+                               next_table_min_key, full_history_ts_low_);
     }
     RecordDroppedKeys(range_del_out_stats, &sub_compact->compaction_job_stats);
     TEST_SYNC_POINT("CompactionJob::FinishCompactionOutputFile1");
@@ -1613,7 +1614,6 @@ Status CompactionJob::FinishCompactionOutputFile(
       TableFileCreationReason::kCompaction, status_for_listener, file_checksum,
       file_checksum_func_name);
 
-#ifndef ROCKSDB_LITE
   // Report new file to SstFileManagerImpl
   auto sfm =
       static_cast<SstFileManagerImpl*>(db_options_.sst_file_manager.get());
@@ -1632,7 +1632,6 @@ Status CompactionJob::FinishCompactionOutputFile(
       db_error_handler_->SetBGError(s, BackgroundErrorReason::kCompaction);
     }
   }
-#endif
 
   outputs.ResetBuilder();
   return s;
@@ -1644,6 +1643,7 @@ Status CompactionJob::InstallCompactionResults(
 
   db_mutex_->AssertHeld();
 
+  const ReadOptions read_options(Env::IOActivity::kCompaction);
   auto* compaction = compact_->compaction;
   assert(compaction);
 
@@ -1721,8 +1721,8 @@ Status CompactionJob::InstallCompactionResults(
   }
 
   return versions_->LogAndApply(compaction->column_family_data(),
-                                mutable_cf_options, edit, db_mutex_,
-                                db_directory_);
+                                mutable_cf_options, read_options, edit,
+                                db_mutex_, db_directory_);
 }
 
 void CompactionJob::RecordCompactionIOStats() {
@@ -1757,11 +1757,9 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
   std::string fname = GetTableFileName(file_number);
   // Fire events.
   ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
-#ifndef ROCKSDB_LITE
   EventHelpers::NotifyTableFileCreationStarted(
       cfd->ioptions()->listeners, dbname_, cfd->GetName(), fname, job_id_,
       TableFileCreationReason::kCompaction);
-#endif  // !ROCKSDB_LITE
   // Make the output file
   std::unique_ptr<FSWritableFile> writable_file;
 #ifndef NDEBUG
@@ -1834,12 +1832,14 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
   }
 
   // Initialize a SubcompactionState::Output and add it to sub_compact->outputs
+  uint64_t epoch_number = sub_compact->compaction->MinInputFileEpochNumber();
   {
     FileMetaData meta;
     meta.fd = FileDescriptor(file_number,
                              sub_compact->compaction->output_path_id(), 0);
     meta.oldest_ancester_time = oldest_ancester_time;
     meta.file_creation_time = current_time;
+    meta.epoch_number = epoch_number;
     meta.temperature = temperature;
     assert(!db_id_.empty());
     assert(!db_session_id_.empty());
@@ -1896,7 +1896,6 @@ void CompactionJob::CleanupCompaction() {
   compact_ = nullptr;
 }
 
-#ifndef ROCKSDB_LITE
 namespace {
 void CopyPrefix(const Slice& src, size_t prefix_length, std::string* dst) {
   assert(prefix_length > 0);
@@ -1905,7 +1904,6 @@ void CopyPrefix(const Slice& src, size_t prefix_length, std::string* dst) {
 }
 }  // namespace
 
-#endif  // !ROCKSDB_LITE
 
 void CompactionJob::UpdateCompactionStats() {
   assert(compact_);
@@ -1952,7 +1950,6 @@ void CompactionJob::UpdateCompactionInputStatsHelper(int* num_files,
 
 void CompactionJob::UpdateCompactionJobStats(
     const InternalStats::CompactionStats& stats) const {
-#ifndef ROCKSDB_LITE
   compaction_job_stats_->elapsed_micros = stats.micros;
 
   // input information
@@ -1979,9 +1976,6 @@ void CompactionJob::UpdateCompactionJobStats(
     CopyPrefix(compact_->LargestUserKey(), CompactionJobStats::kMaxPrefixLength,
                &compaction_job_stats_->largest_output_key_prefix);
   }
-#else
-  (void)stats;
-#endif  // !ROCKSDB_LITE
 }
 
 void CompactionJob::LogCompaction() {

db/compaction/compaction_job.h

@@ -256,7 +256,9 @@ class CompactionJob {
   Status FinishCompactionOutputFile(const Status& input_status,
                                     SubcompactionState* sub_compact,
                                     CompactionOutputs& outputs,
-                                    const Slice& next_table_min_key);
+                                    const Slice& next_table_min_key,
+                                    const Slice* comp_start_user_key,
+                                    const Slice* comp_end_user_key);
   Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options);
   Status OpenCompactionOutputFile(SubcompactionState* sub_compact,
                                   CompactionOutputs& outputs);
@@ -402,6 +404,7 @@ struct CompactionServiceOutputFile {
   std::string largest_internal_key;
   uint64_t oldest_ancester_time;
   uint64_t file_creation_time;
+  uint64_t epoch_number;
   uint64_t paranoid_hash;
   bool marked_for_compaction;
   UniqueId64x2 unique_id;
@@ -411,8 +414,8 @@ struct CompactionServiceOutputFile {
       const std::string& name, SequenceNumber smallest, SequenceNumber largest,
       std::string _smallest_internal_key, std::string _largest_internal_key,
       uint64_t _oldest_ancester_time, uint64_t _file_creation_time,
-      uint64_t _paranoid_hash, bool _marked_for_compaction,
-      UniqueId64x2 _unique_id)
+      uint64_t _epoch_number, uint64_t _paranoid_hash,
+      bool _marked_for_compaction, UniqueId64x2 _unique_id)
       : file_name(name),
         smallest_seqno(smallest),
         largest_seqno(largest),
@@ -420,6 +423,7 @@ struct CompactionServiceOutputFile {
         largest_internal_key(std::move(_largest_internal_key)),
         oldest_ancester_time(_oldest_ancester_time),
         file_creation_time(_file_creation_time),
+        epoch_number(_epoch_number),
         paranoid_hash(_paranoid_hash),
         marked_for_compaction(_marked_for_compaction),
         unique_id(std::move(_unique_id)) {}

db/compaction/compaction_job_stats_test.cc

@@ -24,7 +24,7 @@
 #include "db/write_batch_internal.h"
 #include "env/mock_env.h"
 #include "file/filename.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "monitoring/thread_status_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/cache.h"
@@ -54,12 +54,11 @@
 #include "util/compression.h"
 #include "util/hash.h"
 #include "util/mutexlock.h"
-#include "util/rate_limiter.h"
+#include "util/rate_limiter_impl.h"
 #include "util/string_util.h"
 #include "utilities/merge_operators.h"
 
 #if !defined(IOS_CROSS_COMPILE)
-#ifndef ROCKSDB_LITE
 namespace ROCKSDB_NAMESPACE {
 
 static std::string RandomString(Random* rnd, int len, double ratio) {
@@ -959,15 +958,6 @@ int main(int argc, char** argv) {
   return RUN_ALL_TESTS();
 }
 
-#else
-#include <stdio.h>
-
-int main(int /*argc*/, char** /*argv*/) {
-  fprintf(stderr, "SKIPPED, not supported in ROCKSDB_LITE\n");
-  return 0;
-}
-
-#endif  // !ROCKSDB_LITE
 
 #else
 

db/compaction/compaction_job_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-#ifndef ROCKSDB_LITE
 
 #include "db/compaction/compaction_job.h"
 
@@ -374,22 +373,25 @@ class CompactionJobTestBase : public testing::Test {
     } else if (table_type_ == TableTypeForTest::kMockTable) {
       file_size = 10;
       EXPECT_OK(mock_table_factory_->CreateMockTable(
-          env_, GenerateFileName(file_number), std::move(contents)));
+          env_, GenerateFileName(file_number), contents));
     } else {
       assert(false);
     }
 
     VersionEdit edit;
-    edit.AddFile(level, file_number, 0, file_size, smallest_key, largest_key,
-                 smallest_seqno, largest_seqno, false, Temperature::kUnknown,
-                 oldest_blob_file_number, kUnknownOldestAncesterTime,
-                 kUnknownFileCreationTime, kUnknownFileChecksum,
-                 kUnknownFileChecksumFuncName, kNullUniqueId64x2);
+    edit.AddFile(
+        level, file_number, 0, file_size, smallest_key, largest_key,
+        smallest_seqno, largest_seqno, false, Temperature::kUnknown,
+        oldest_blob_file_number, kUnknownOldestAncesterTime,
+        kUnknownFileCreationTime,
+        versions_->GetColumnFamilySet()->GetDefault()->NewEpochNumber(),
+        kUnknownFileChecksum, kUnknownFileChecksumFuncName, kNullUniqueId64x2,
+        0, 0);
 
     mutex_.Lock();
-    EXPECT_OK(
-        versions_->LogAndApply(versions_->GetColumnFamilySet()->GetDefault(),
-                               mutable_cf_options_, &edit, &mutex_, nullptr));
+    EXPECT_OK(versions_->LogAndApply(
+        versions_->GetColumnFamilySet()->GetDefault(), mutable_cf_options_,
+        read_options_, &edit, &mutex_, nullptr));
     mutex_.Unlock();
   }
 
@@ -452,7 +454,8 @@ class CompactionJobTestBase : public testing::Test {
       Status s = cf_options_.table_factory->NewTableReader(
           read_opts,
           TableReaderOptions(*cfd->ioptions(), nullptr, FileOptions(),
-                             cfd_->internal_comparator()),
+                             cfd_->internal_comparator(),
+                             0 /* block_protection_bytes_per_key */),
           std::move(freader), file_size, &table_reader, false);
       ASSERT_OK(s);
       assert(table_reader);
@@ -725,6 +728,7 @@ class CompactionJobTestBase : public testing::Test {
   ColumnFamilyOptions cf_options_;
   MutableCFOptions mutable_cf_options_;
   MutableDBOptions mutable_db_options_;
+  const ReadOptions read_options_;
   std::shared_ptr<Cache> table_cache_;
   WriteController write_controller_;
   WriteBufferManager write_buffer_manager_;
@@ -1655,7 +1659,7 @@ TEST_F(CompactionJobTest, ResultSerialization) {
         rnd.RandomBinaryString(rnd.Uniform(kStrMaxLen)),
         rnd.RandomBinaryString(rnd.Uniform(kStrMaxLen)),
         rnd64.Uniform(UINT64_MAX), rnd64.Uniform(UINT64_MAX),
-        rnd64.Uniform(UINT64_MAX), rnd.OneIn(2), id);
+        rnd64.Uniform(UINT64_MAX), rnd64.Uniform(UINT64_MAX), rnd.OneIn(2), id);
   }
   result.output_level = rnd.Uniform(10);
   result.output_path = rnd.RandomString(rnd.Uniform(kStrMaxLen));
@@ -2438,14 +2442,3 @@ int main(int argc, char** argv) {
   RegisterCustomObjects(argc, argv);
   return RUN_ALL_TESTS();
 }
-
-#else
-#include <stdio.h>
-
-int main(int /*argc*/, char** /*argv*/) {
-  fprintf(stderr,
-          "SKIPPED as CompactionJobStats is not supported in ROCKSDB_LITE\n");
-  return 0;
-}
-
-#endif  // ROCKSDB_LITE

db/compaction/compaction_outputs.cc

@@ -43,6 +43,7 @@ Status CompactionOutputs::Finish(const Status& intput_status,
   const uint64_t current_bytes = builder_->FileSize();
   if (s.ok()) {
     meta->fd.file_size = current_bytes;
+    meta->tail_size = builder_->GetTailSize();
     meta->marked_for_compaction = builder_->NeedCompact();
   }
   current_output().finished = true;
@@ -76,6 +77,46 @@ IOStatus CompactionOutputs::WriterSyncClose(const Status& input_status,
   return io_s;
 }
 
+bool CompactionOutputs::UpdateFilesToCutForTTLStates(
+    const Slice& internal_key) {
+  if (!files_to_cut_for_ttl_.empty()) {
+    const InternalKeyComparator* icmp =
+        &compaction_->column_family_data()->internal_comparator();
+    if (cur_files_to_cut_for_ttl_ != -1) {
+      // Previous key is inside the range of a file
+      if (icmp->Compare(internal_key,
+                        files_to_cut_for_ttl_[cur_files_to_cut_for_ttl_]
+                            ->largest.Encode()) > 0) {
+        next_files_to_cut_for_ttl_ = cur_files_to_cut_for_ttl_ + 1;
+        cur_files_to_cut_for_ttl_ = -1;
+        return true;
+      }
+    } else {
+      // Look for the key position
+      while (next_files_to_cut_for_ttl_ <
+             static_cast<int>(files_to_cut_for_ttl_.size())) {
+        if (icmp->Compare(internal_key,
+                          files_to_cut_for_ttl_[next_files_to_cut_for_ttl_]
+                              ->smallest.Encode()) >= 0) {
+          if (icmp->Compare(internal_key,
+                            files_to_cut_for_ttl_[next_files_to_cut_for_ttl_]
+                                ->largest.Encode()) <= 0) {
+            // With in the current file
+            cur_files_to_cut_for_ttl_ = next_files_to_cut_for_ttl_;
+            return true;
+          }
+          // Beyond the current file
+          next_files_to_cut_for_ttl_++;
+        } else {
+          // Still fall into the gap
+          break;
+        }
+      }
+    }
+  }
+  return false;
+}
+
 size_t CompactionOutputs::UpdateGrandparentBoundaryInfo(
     const Slice& internal_key) {
   size_t curr_key_boundary_switched_num = 0;
@@ -185,18 +226,39 @@ uint64_t CompactionOutputs::GetCurrentKeyGrandparentOverlappedBytes(
 
 bool CompactionOutputs::ShouldStopBefore(const CompactionIterator& c_iter) {
   assert(c_iter.Valid());
-
-  // always update grandparent information like overlapped file number, size
-  // etc.
   const Slice& internal_key = c_iter.key();
+#ifndef NDEBUG
+  bool should_stop = false;
+  std::pair<bool*, const Slice> p{&should_stop, internal_key};
+  TEST_SYNC_POINT_CALLBACK(
+      "CompactionOutputs::ShouldStopBefore::manual_decision", (void*)&p);
+  if (should_stop) {
+    return true;
+  }
+#endif  // NDEBUG
   const uint64_t previous_overlapped_bytes = grandparent_overlapped_bytes_;
-  size_t num_grandparent_boundaries_crossed =
-      UpdateGrandparentBoundaryInfo(internal_key);
+  const InternalKeyComparator* icmp =
+      &compaction_->column_family_data()->internal_comparator();
+  size_t num_grandparent_boundaries_crossed = 0;
+  bool should_stop_for_ttl = false;
+  // Always update grandparent information like overlapped file number, size
+  // etc., and TTL states.
+  // If compaction_->output_level() == 0, there is no need to update grandparent
+  // info, and that `grandparent` should be empty.
+  if (compaction_->output_level() > 0) {
+    num_grandparent_boundaries_crossed =
+        UpdateGrandparentBoundaryInfo(internal_key);
+    should_stop_for_ttl = UpdateFilesToCutForTTLStates(internal_key);
+  }
 
   if (!HasBuilder()) {
     return false;
   }
 
+  if (should_stop_for_ttl) {
+    return true;
+  }
+
   // If there's user defined partitioner, check that first
   if (partitioner_ && partitioner_->ShouldPartition(PartitionerRequest(
                           last_key_for_partitioner_, c_iter.user_key(),
@@ -214,9 +276,6 @@ bool CompactionOutputs::ShouldStopBefore(const CompactionIterator& c_iter) {
     return true;
   }
 
-  const InternalKeyComparator* icmp =
-      &compaction_->column_family_data()->internal_comparator();
-
   // Check if it needs to split for RoundRobin
   // Invalid local_output_split_key indicates that we do not need to split
   if (local_output_split_key_ != nullptr && !is_split_) {
@@ -290,41 +349,6 @@ bool CompactionOutputs::ShouldStopBefore(const CompactionIterator& c_iter) {
     }
   }
 
-  // check ttl file boundaries if there's any
-  if (!files_to_cut_for_ttl_.empty()) {
-    if (cur_files_to_cut_for_ttl_ != -1) {
-      // Previous key is inside the range of a file
-      if (icmp->Compare(internal_key,
-                        files_to_cut_for_ttl_[cur_files_to_cut_for_ttl_]
-                            ->largest.Encode()) > 0) {
-        next_files_to_cut_for_ttl_ = cur_files_to_cut_for_ttl_ + 1;
-        cur_files_to_cut_for_ttl_ = -1;
-        return true;
-      }
-    } else {
-      // Look for the key position
-      while (next_files_to_cut_for_ttl_ <
-             static_cast<int>(files_to_cut_for_ttl_.size())) {
-        if (icmp->Compare(internal_key,
-                          files_to_cut_for_ttl_[next_files_to_cut_for_ttl_]
-                              ->smallest.Encode()) >= 0) {
-          if (icmp->Compare(internal_key,
-                            files_to_cut_for_ttl_[next_files_to_cut_for_ttl_]
-                                ->largest.Encode()) <= 0) {
-            // With in the current file
-            cur_files_to_cut_for_ttl_ = next_files_to_cut_for_ttl_;
-            return true;
-          }
-          // Beyond the current file
-          next_files_to_cut_for_ttl_++;
-        } else {
-          // Still fall into the gap
-          break;
-        }
-      }
-    }
-  }
-
   return false;
 }
 
@@ -333,8 +357,14 @@ Status CompactionOutputs::AddToOutput(
     const CompactionFileOpenFunc& open_file_func,
     const CompactionFileCloseFunc& close_file_func) {
   Status s;
+  bool is_range_del = c_iter.IsDeleteRangeSentinelKey();
+  if (is_range_del && compaction_->bottommost_level()) {
+    // We don't consider range tombstone for bottommost level since:
+    // 1. there is no grandparent and hence no overlap to consider
+    // 2. range tombstone may be dropped at bottommost level.
+    return s;
+  }
   const Slice& key = c_iter.key();
-
   if (ShouldStopBefore(c_iter) && HasBuilder()) {
     s = close_file_func(*this, c_iter.InputStatus(), key);
     if (!s.ok()) {
@@ -344,6 +374,13 @@ Status CompactionOutputs::AddToOutput(
     grandparent_boundary_switched_num_ = 0;
     grandparent_overlapped_bytes_ =
         GetCurrentKeyGrandparentOverlappedBytes(key);
+    if (UNLIKELY(is_range_del)) {
+      // lower bound for this new output file, this is needed as the lower bound
+      // does not come from the smallest point key in this case.
+      range_tombstone_lower_bound_.DecodeFrom(key);
+    } else {
+      range_tombstone_lower_bound_.Clear();
+    }
   }
 
   // Open output file if necessary
@@ -354,6 +391,17 @@ Status CompactionOutputs::AddToOutput(
     }
   }
 
+  // c_iter may emit range deletion keys, so update `last_key_for_partitioner_`
+  // here before returning below when `is_range_del` is true
+  if (partitioner_) {
+    last_key_for_partitioner_.assign(c_iter.user_key().data_,
+                                     c_iter.user_key().size_);
+  }
+
+  if (UNLIKELY(is_range_del)) {
+    return s;
+  }
+
   assert(builder_ != nullptr);
   const Slice& value = c_iter.value();
   s = current_output().validator.Add(key, value);
@@ -377,106 +425,173 @@ Status CompactionOutputs::AddToOutput(
   s = current_output().meta.UpdateBoundaries(key, value, ikey.sequence,
                                              ikey.type);
 
-  if (partitioner_) {
-    last_key_for_partitioner_.assign(c_iter.user_key().data_,
-                                     c_iter.user_key().size_);
-  }
-
   return s;
 }
 
+namespace {
+void SetMaxSeqAndTs(InternalKey& internal_key, const Slice& user_key,
+                    const size_t ts_sz) {
+  if (ts_sz) {
+    static constexpr char kTsMax[] = "\xff\xff\xff\xff\xff\xff\xff\xff\xff";
+    if (ts_sz <= strlen(kTsMax)) {
+      internal_key = InternalKey(user_key, kMaxSequenceNumber,
+                                 kTypeRangeDeletion, Slice(kTsMax, ts_sz));
+    } else {
+      internal_key =
+          InternalKey(user_key, kMaxSequenceNumber, kTypeRangeDeletion,
+                      std::string(ts_sz, '\xff'));
+    }
+  } else {
+    internal_key.Set(user_key, kMaxSequenceNumber, kTypeRangeDeletion);
+  }
+}
+}  // namespace
+
 Status CompactionOutputs::AddRangeDels(
     const Slice* comp_start_user_key, const Slice* comp_end_user_key,
     CompactionIterationStats& range_del_out_stats, bool bottommost_level,
     const InternalKeyComparator& icmp, SequenceNumber earliest_snapshot,
     const Slice& next_table_min_key, const std::string& full_history_ts_low) {
+  // The following example does not happen since
+  // CompactionOutput::ShouldStopBefore() always return false for the first
+  // point key. But we should consider removing this dependency. Suppose for the
+  // first compaction output file,
+  //  - next_table_min_key.user_key == comp_start_user_key
+  //  - no point key is in the output file
+  //  - there is a range tombstone @seqno to be added that covers
+  //  comp_start_user_key
+  // Then meta.smallest will be set to comp_start_user_key@seqno
+  // and meta.largest will be set to comp_start_user_key@kMaxSequenceNumber
+  // which violates the assumption that meta.smallest should be <= meta.largest.
   assert(HasRangeDel());
   FileMetaData& meta = current_output().meta;
   const Comparator* ucmp = icmp.user_comparator();
-
+  InternalKey lower_bound_buf, upper_bound_buf;
   Slice lower_bound_guard, upper_bound_guard;
   std::string smallest_user_key;
   const Slice *lower_bound, *upper_bound;
-  bool lower_bound_from_sub_compact = false;
 
+  // We first determine the internal key lower_bound and upper_bound for
+  // this output file. All and only range tombstones that overlap with
+  // [lower_bound, upper_bound] should be added to this file. File
+  // boundaries (meta.smallest/largest) should be updated accordingly when
+  // extended by range tombstones.
   size_t output_size = outputs_.size();
   if (output_size == 1) {
-    // For the first output table, include range tombstones before the min
-    // key but after the subcompaction boundary.
-    lower_bound = comp_start_user_key;
-    lower_bound_from_sub_compact = true;
-  } else if (meta.smallest.size() > 0) {
+    // This is the first file in the subcompaction.
+    //
+    // When outputting a range tombstone that spans a subcompaction boundary,
+    // the files on either side of that boundary need to include that
+    // boundary's user key. Otherwise, the spanning range tombstone would lose
+    // coverage.
+    //
+    // To achieve this while preventing files from overlapping in internal key
+    // (an LSM invariant violation), we allow the earlier file to include the
+    // boundary user key up to `kMaxSequenceNumber,kTypeRangeDeletion`. The
+    // later file can begin at the boundary user key at the newest key version
+    // it contains. At this point that version number is unknown since we have
+    // not processed the range tombstones yet, so permit any version. Same story
+    // applies to timestamp, and a non-nullptr `comp_start_user_key` should have
+    // `kMaxTs` here, which similarly permits any timestamp.
+    if (comp_start_user_key) {
+      lower_bound_buf.Set(*comp_start_user_key, kMaxSequenceNumber,
+                          kTypeRangeDeletion);
+      lower_bound_guard = lower_bound_buf.Encode();
+      lower_bound = &lower_bound_guard;
+    } else {
+      lower_bound = nullptr;
+    }
+  } else {
     // For subsequent output tables, only include range tombstones from min
     // key onwards since the previous file was extended to contain range
     // tombstones falling before min key.
-    smallest_user_key = meta.smallest.user_key().ToString(false /*hex*/);
-    lower_bound_guard = Slice(smallest_user_key);
-    lower_bound = &lower_bound_guard;
-  } else {
-    lower_bound = nullptr;
-  }
-  if (!next_table_min_key.empty()) {
-    // This may be the last file in the subcompaction in some cases, so we
-    // need to compare the end key of subcompaction with the next file start
-    // key. When the end key is chosen by the subcompaction, we know that
-    // it must be the biggest key in output file. Therefore, it is safe to
-    // use the smaller key as the upper bound of the output file, to ensure
-    // that there is no overlapping between different output files.
-    upper_bound_guard = ExtractUserKey(next_table_min_key);
-    if (comp_end_user_key != nullptr &&
-        ucmp->CompareWithoutTimestamp(upper_bound_guard, *comp_end_user_key) >=
-            0) {
-      upper_bound = comp_end_user_key;
+    if (range_tombstone_lower_bound_.size() > 0) {
+      assert(meta.smallest.size() == 0 ||
+             icmp.Compare(range_tombstone_lower_bound_, meta.smallest) < 0);
+      lower_bound_guard = range_tombstone_lower_bound_.Encode();
     } else {
-      upper_bound = &upper_bound_guard;
+      assert(meta.smallest.size() > 0);
+      lower_bound_guard = meta.smallest.Encode();
     }
-  } else {
-    // This is the last file in the subcompaction, so extend until the
-    // subcompaction ends.
-    upper_bound = comp_end_user_key;
-  }
-  bool has_overlapping_endpoints;
-  if (upper_bound != nullptr && meta.largest.size() > 0) {
-    has_overlapping_endpoints = ucmp->CompareWithoutTimestamp(
-                                    meta.largest.user_key(), *upper_bound) == 0;
-  } else {
-    has_overlapping_endpoints = false;
+    lower_bound = &lower_bound_guard;
   }
 
+  const size_t ts_sz = ucmp->timestamp_size();
+  if (next_table_min_key.empty()) {
+    // Last file of the subcompaction.
+    if (comp_end_user_key) {
+      upper_bound_buf.Set(*comp_end_user_key, kMaxSequenceNumber,
+                          kTypeRangeDeletion);
+      upper_bound_guard = upper_bound_buf.Encode();
+      upper_bound = &upper_bound_guard;
+    } else {
+      upper_bound = nullptr;
+    }
+  } else {
+    // There is another file coming whose coverage will begin at
+    // `next_table_min_key`. The current file needs to extend range tombstone
+    // coverage through its own keys (through `meta.largest`) and through user
+    // keys preceding `next_table_min_key`'s user key.
+    ParsedInternalKey next_table_min_key_parsed;
+    ParseInternalKey(next_table_min_key, &next_table_min_key_parsed,
+                     false /* log_err_key */)
+        .PermitUncheckedError();
+    assert(next_table_min_key_parsed.sequence < kMaxSequenceNumber);
+    assert(meta.largest.size() == 0 ||
+           icmp.Compare(meta.largest.Encode(), next_table_min_key) < 0);
+    assert(!lower_bound || icmp.Compare(*lower_bound, next_table_min_key) <= 0);
+    if (meta.largest.size() > 0 &&
+        ucmp->EqualWithoutTimestamp(meta.largest.user_key(),
+                                    next_table_min_key_parsed.user_key)) {
+      // Caution: this assumes meta.largest.Encode() lives longer than
+      // upper_bound, which is only true if meta.largest is never updated.
+      // This just happens to be the case here since meta.largest serves
+      // as the upper_bound.
+      upper_bound_guard = meta.largest.Encode();
+    } else {
+      SetMaxSeqAndTs(upper_bound_buf, next_table_min_key_parsed.user_key,
+                     ts_sz);
+      upper_bound_guard = upper_bound_buf.Encode();
+    }
+    upper_bound = &upper_bound_guard;
+  }
+  if (lower_bound && upper_bound &&
+      icmp.Compare(*lower_bound, *upper_bound) > 0) {
+    assert(meta.smallest.size() == 0 &&
+           ucmp->EqualWithoutTimestamp(ExtractUserKey(*lower_bound),
+                                       ExtractUserKey(*upper_bound)));
+    // This can only happen when lower_bound have the same user key as
+    // next_table_min_key and that there is no point key in the current
+    // compaction output file.
+    return Status::OK();
+  }
   // The end key of the subcompaction must be bigger or equal to the upper
   // bound. If the end of subcompaction is null or the upper bound is null,
   // it means that this file is the last file in the compaction. So there
   // will be no overlapping between this file and others.
   assert(comp_end_user_key == nullptr || upper_bound == nullptr ||
-         ucmp->CompareWithoutTimestamp(*upper_bound, *comp_end_user_key) <= 0);
-  auto it = range_del_agg_->NewIterator(lower_bound, upper_bound,
-                                        has_overlapping_endpoints);
-  // Position the range tombstone output iterator. There may be tombstone
-  // fragments that are entirely out of range, so make sure that we do not
-  // include those.
-  if (lower_bound != nullptr) {
-    it->Seek(*lower_bound);
-  } else {
-    it->SeekToFirst();
-  }
-  for (; it->Valid(); it->Next()) {
+         ucmp->CompareWithoutTimestamp(ExtractUserKey(*upper_bound),
+                                       *comp_end_user_key) <= 0);
+  auto it = range_del_agg_->NewIterator(lower_bound, upper_bound);
+  Slice last_tombstone_start_user_key{};
+  bool reached_lower_bound = false;
+  const ReadOptions read_options(Env::IOActivity::kCompaction);
+  for (it->SeekToFirst(); it->Valid(); it->Next()) {
     auto tombstone = it->Tombstone();
-    if (upper_bound != nullptr) {
-      int cmp =
-          ucmp->CompareWithoutTimestamp(*upper_bound, tombstone.start_key_);
-      if ((has_overlapping_endpoints && cmp < 0) ||
-          (!has_overlapping_endpoints && cmp <= 0)) {
-        // Tombstones starting after upper_bound only need to be included in
-        // the next table. If the current SST ends before upper_bound, i.e.,
-        // `has_overlapping_endpoints == false`, we can also skip over range
-        // tombstones that start exactly at upper_bound. Such range
-        // tombstones will be included in the next file and are not relevant
-        // to the point keys or endpoints of the current file.
-        break;
-      }
+    auto kv = tombstone.Serialize();
+    InternalKey tombstone_end = tombstone.SerializeEndKey();
+    // TODO: the underlying iterator should support clamping the bounds.
+    // tombstone_end.Encode is of form user_key@kMaxSeqno
+    // if it is equal to lower_bound, there is no need to include
+    // such range tombstone.
+    if (!reached_lower_bound && lower_bound &&
+        icmp.Compare(tombstone_end.Encode(), *lower_bound) <= 0) {
+      continue;
     }
+    assert(!lower_bound ||
+           icmp.Compare(*lower_bound, tombstone_end.Encode()) <= 0);
+    reached_lower_bound = true;
 
-    const size_t ts_sz = ucmp->timestamp_size();
     // Garbage collection for range tombstones.
     // If user-defined timestamp is enabled, range tombstones are dropped if
     // they are at bottommost_level, below full_history_ts_low and not visible
@@ -495,96 +610,119 @@ Status CompactionOutputs::AddRangeDels(
       continue;
     }
 
-    auto kv = tombstone.Serialize();
     assert(lower_bound == nullptr ||
-           ucmp->CompareWithoutTimestamp(*lower_bound, kv.second) < 0);
+           ucmp->CompareWithoutTimestamp(ExtractUserKey(*lower_bound),
+                                         kv.second) < 0);
+    InternalKey tombstone_start = kv.first;
+    if (lower_bound &&
+        ucmp->CompareWithoutTimestamp(tombstone_start.user_key(),
+                                      ExtractUserKey(*lower_bound)) < 0) {
+      // This just updates the non-timestamp portion of `tombstone_start`'s user
+      // key. Ideally there would be a simpler API usage
+      ParsedInternalKey tombstone_start_parsed;
+      ParseInternalKey(tombstone_start.Encode(), &tombstone_start_parsed,
+                       false /* log_err_key */)
+          .PermitUncheckedError();
+      // timestamp should be from where sequence number is from, which is from
+      // tombstone in this case
+      std::string ts =
+          tombstone_start_parsed.GetTimestamp(ucmp->timestamp_size())
+              .ToString();
+      tombstone_start_parsed.user_key = ExtractUserKey(*lower_bound);
+      tombstone_start.SetFrom(tombstone_start_parsed, ts);
+    }
+    if (upper_bound != nullptr &&
+        icmp.Compare(*upper_bound, tombstone_start.Encode()) < 0) {
+      break;
+    }
+    // Here we show that *only* range tombstones that overlap with
+    // [lower_bound, upper_bound] are added to the current file, and
+    // sanity checking invariants that should hold:
+    // - [tombstone_start, tombstone_end] overlaps with [lower_bound,
+    // upper_bound]
+    // - meta.smallest <= meta.largest
+    // Corresponding assertions are made, the proof is broken is any of them
+    // fails.
+    // TODO: show that *all* range tombstones that overlap with
+    //  [lower_bound, upper_bound] are added.
+    // TODO: some invariant about boundaries are correctly updated.
+    //
+    // Note that `tombstone_start` is updated in the if condition above, we use
+    // tombstone_start to refer to its initial value, i.e.,
+    // it->Tombstone().first, and use tombstone_start* to refer to its value
+    // after the update.
+    //
+    // To show [lower_bound, upper_bound] overlaps with [tombstone_start,
+    // tombstone_end]:
+    // lower_bound <= upper_bound from the if condition right after all
+    // bounds are initialized. We assume each tombstone fragment has
+    // start_key.user_key < end_key.user_key, so
+    // tombstone_start < tombstone_end by
+    // FragmentedTombstoneIterator::Tombstone(). So these two ranges are both
+    // non-emtpy. The flag `reached_lower_bound` and the if logic before it
+    // ensures lower_bound <= tombstone_end. tombstone_start is only updated
+    // if it has a smaller user_key than lower_bound user_key, so
+    // tombstone_start <= tombstone_start*. The above if condition implies
+    // tombstone_start* <= upper_bound. So we have
+    // tombstone_start <= upper_bound and lower_bound <= tombstone_end
+    // and the two ranges overlap.
+    //
+    // To show meta.smallest <= meta.largest:
+    // From the implementation of UpdateBoundariesForRange(), it suffices to
+    // prove that when it is first called in this function, its parameters
+    // satisfy `start <= end`, where start = max(tombstone_start*, lower_bound)
+    // and end = min(tombstone_end, upper_bound). From the above proof we have
+    // lower_bound <= tombstone_end and lower_bound <= upper_bound. We only need
+    // to show that tombstone_start* <= min(tombstone_end, upper_bound).
+    // Note that tombstone_start*.user_key = max(tombstone_start.user_key,
+    // lower_bound.user_key). Assuming tombstone_end always has
+    // kMaxSequenceNumber and lower_bound.seqno < kMaxSequenceNumber.
+    // Since lower_bound <= tombstone_end and lower_bound.seqno <
+    // tombstone_end.seqno (in absolute number order, not internal key order),
+    // lower_bound.user_key < tombstone_end.user_key.
+    // Since lower_bound.user_key < tombstone_end.user_key and
+    // tombstone_start.user_key < tombstone_end.user_key, tombstone_start* <
+    // tombstone_end. Since tombstone_start* <= upper_bound from the above proof
+    // and tombstone_start* < tombstone_end, tombstone_start* <=
+    // min(tombstone_end, upper_bound), so the two ranges overlap.
+
     // Range tombstone is not supported by output validator yet.
     builder_->Add(kv.first.Encode(), kv.second);
-    InternalKey smallest_candidate = std::move(kv.first);
-    if (lower_bound != nullptr &&
-        ucmp->CompareWithoutTimestamp(smallest_candidate.user_key(),
-                                      *lower_bound) <= 0) {
-      // Pretend the smallest key has the same user key as lower_bound
-      // (the max key in the previous table or subcompaction) in order for
-      // files to appear key-space partitioned.
-      //
-      // When lower_bound is chosen by a subcompaction, we know that
-      // subcompactions over smaller keys cannot contain any keys at
-      // lower_bound. We also know that smaller subcompactions exist,
-      // because otherwise the subcompaction woud be unbounded on the left.
-      // As a result, we know that no other files on the output level will
-      // contain actual keys at lower_bound (an output file may have a
-      // largest key of lower_bound@kMaxSequenceNumber, but this only
-      // indicates a large range tombstone was truncated). Therefore, it is
-      // safe to use the tombstone's sequence number, to ensure that keys at
-      // lower_bound at lower levels are covered by truncated tombstones.
-      //
-      // If lower_bound was chosen by the smallest data key in the file,
-      // choose lowest seqnum so this file's smallest internal key comes
-      // after the previous file's largest. The fake seqnum is OK because
-      // the read path's file-picking code only considers user key.
-      if (lower_bound_from_sub_compact) {
-        if (ts_sz) {
-          assert(tombstone.ts_.size() == ts_sz);
-          smallest_candidate = InternalKey(*lower_bound, tombstone.seq_,
-                                           kTypeRangeDeletion, tombstone.ts_);
-        } else {
-          smallest_candidate =
-              InternalKey(*lower_bound, tombstone.seq_, kTypeRangeDeletion);
-        }
-      } else {
-        smallest_candidate = InternalKey(*lower_bound, 0, kTypeRangeDeletion);
-      }
+    if (lower_bound &&
+        icmp.Compare(tombstone_start.Encode(), *lower_bound) < 0) {
+      tombstone_start.DecodeFrom(*lower_bound);
     }
-    InternalKey largest_candidate = tombstone.SerializeEndKey();
-    if (upper_bound != nullptr &&
-        ucmp->CompareWithoutTimestamp(*upper_bound,
-                                      largest_candidate.user_key()) <= 0) {
-      // Pretend the largest key has the same user key as upper_bound (the
-      // min key in the following table or subcompaction) in order for files
-      // to appear key-space partitioned.
-      //
-      // Choose highest seqnum so this file's largest internal key comes
-      // before the next file's/subcompaction's smallest. The fake seqnum is
-      // OK because the read path's file-picking code only considers the
-      // user key portion.
-      //
-      // Note Seek() also creates InternalKey with (user_key,
-      // kMaxSequenceNumber), but with kTypeDeletion (0x7) instead of
-      // kTypeRangeDeletion (0xF), so the range tombstone comes before the
-      // Seek() key in InternalKey's ordering. So Seek() will look in the
-      // next file for the user key
-      if (ts_sz) {
-        static constexpr char kTsMax[] = "\xff\xff\xff\xff\xff\xff\xff\xff\xff";
-        if (ts_sz <= strlen(kTsMax)) {
-          largest_candidate =
-              InternalKey(*upper_bound, kMaxSequenceNumber, kTypeRangeDeletion,
-                          Slice(kTsMax, ts_sz));
-        } else {
-          largest_candidate =
-              InternalKey(*upper_bound, kMaxSequenceNumber, kTypeRangeDeletion,
-                          std::string(ts_sz, '\xff'));
-        }
-      } else {
-        largest_candidate =
-            InternalKey(*upper_bound, kMaxSequenceNumber, kTypeRangeDeletion);
-      }
+    if (upper_bound && icmp.Compare(*upper_bound, tombstone_end.Encode()) < 0) {
+      tombstone_end.DecodeFrom(*upper_bound);
     }
-#ifndef NDEBUG
-    SequenceNumber smallest_ikey_seqnum = kMaxSequenceNumber;
-    if (meta.smallest.size() > 0) {
-      smallest_ikey_seqnum = GetInternalKeySeqno(meta.smallest.Encode());
-    }
-#endif
-    meta.UpdateBoundariesForRange(smallest_candidate, largest_candidate,
+    assert(icmp.Compare(tombstone_start, tombstone_end) <= 0);
+    meta.UpdateBoundariesForRange(tombstone_start, tombstone_end,
                                   tombstone.seq_, icmp);
-    // The smallest key in a file is used for range tombstone truncation, so
-    // it cannot have a seqnum of 0 (unless the smallest data key in a file
-    // has a seqnum of 0). Otherwise, the truncated tombstone may expose
-    // deleted keys at lower levels.
-    assert(smallest_ikey_seqnum == 0 ||
-           ExtractInternalKeyFooter(meta.smallest.Encode()) !=
-               PackSequenceAndType(0, kTypeRangeDeletion));
+    if (!bottommost_level) {
+      bool start_user_key_changed =
+          last_tombstone_start_user_key.empty() ||
+          ucmp->CompareWithoutTimestamp(last_tombstone_start_user_key,
+                                        it->start_key()) < 0;
+      last_tombstone_start_user_key = it->start_key();
+      if (start_user_key_changed) {
+        // If tombstone_start >= tombstone_end, then either no key range is
+        // covered, or that they have the same user key. If they have the same
+        // user key, then the internal key range should only be within this
+        // level, and no keys from older levels is covered.
+        if (ucmp->CompareWithoutTimestamp(tombstone_start.user_key(),
+                                          tombstone_end.user_key()) < 0) {
+          SizeApproximationOptions approx_opts;
+          approx_opts.files_size_error_margin = 0.1;
+          auto approximate_covered_size =
+              compaction_->input_version()->version_set()->ApproximateSize(
+                  approx_opts, read_options, compaction_->input_version(),
+                  tombstone_start.Encode(), tombstone_end.Encode(),
+                  compaction_->output_level() + 1 /* start_level */,
+                  -1 /* end_level */, kCompaction);
+          meta.compensated_range_deletion_size += approximate_covered_size;
+        }
+      }
+    }
   }
   return Status::OK();
 }

db/compaction/compaction_outputs.h

@@ -167,9 +167,15 @@ class CompactionOutputs {
     current_output_file_size_ = 0;
   }
 
-  // Add range-dels from the aggregator to the current output file
+  // Add range deletions from the range_del_agg_ to the current output file.
+  // Input parameters, `range_tombstone_lower_bound_` and current output's
+  // metadata determine the bounds on range deletions to add. Updates output
+  // file metadata boundary if extended by range tombstones.
+  //
   // @param comp_start_user_key and comp_end_user_key include timestamp if
-  // user-defined timestamp is enabled.
+  // user-defined timestamp is enabled. Their timestamp should be max timestamp.
+  // @param next_table_min_key internal key lower bound for the next compaction
+  // output.
   // @param full_history_ts_low used for range tombstone garbage collection.
   Status AddRangeDels(const Slice* comp_start_user_key,
                       const Slice* comp_end_user_key,
@@ -221,6 +227,13 @@ class CompactionOutputs {
     }
   }
 
+  // Updates states related to file cutting for TTL.
+  // Returns a boolean value indicating whether the current
+  // compaction output file should be cut before `internal_key`.
+  //
+  // @param internal_key the current key to be added to output.
+  bool UpdateFilesToCutForTTLStates(const Slice& internal_key);
+
   // update tracked grandparents information like grandparent index, if it's
   // in the gap between 2 grandparent files, accumulated grandparent files size
   // etc.
@@ -307,6 +320,7 @@ class CompactionOutputs {
   std::unique_ptr<SstPartitioner> partitioner_;
 
   // A flag determines if this subcompaction has been split by the cursor
+  // for RoundRobin compaction
   bool is_split_ = false;
 
   // We also maintain the output split key for each subcompaction to avoid
@@ -338,6 +352,10 @@ class CompactionOutputs {
   // for the current output file, how many file boundaries has it crossed,
   // basically number of files overlapped * 2
   size_t grandparent_boundary_switched_num_ = 0;
+
+  // The smallest key of the current output file, this is set when current
+  // output file's smallest key is a range tombstone start key.
+  InternalKey range_tombstone_lower_bound_;
 };
 
 // helper struct to concatenate the last level and penultimate level outputs

db/compaction/compaction_picker.cc

@@ -20,7 +20,7 @@
 #include "file/filename.h"
 #include "logging/log_buffer.h"
 #include "logging/logging.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "test_util/sync_point.h"
 #include "util/random.h"
 #include "util/string_util.h"
@@ -31,27 +31,15 @@ bool FindIntraL0Compaction(const std::vector<FileMetaData*>& level_files,
                            size_t min_files_to_compact,
                            uint64_t max_compact_bytes_per_del_file,
                            uint64_t max_compaction_bytes,
-                           CompactionInputFiles* comp_inputs,
-                           const SequenceNumber earliest_mem_seqno) {
-  // Do not pick ingested file when there is at least one memtable not flushed
-  // which of seqno is overlap with the sst.
+                           CompactionInputFiles* comp_inputs) {
   TEST_SYNC_POINT("FindIntraL0Compaction");
+
   size_t start = 0;
-  for (; start < level_files.size(); start++) {
-    if (level_files[start]->being_compacted) {
-      return false;
-    }
-    // If there is no data in memtable, the earliest sequence number would the
-    // largest sequence number in last memtable.
-    // Because all files are sorted in descending order by largest_seqno, so we
-    // only need to check the first one.
-    if (level_files[start]->fd.largest_seqno <= earliest_mem_seqno) {
-      break;
-    }
-  }
-  if (start >= level_files.size()) {
+
+  if (level_files.size() == 0 || level_files[start]->being_compacted) {
     return false;
   }
+
   size_t compact_bytes = static_cast<size_t>(level_files[start]->fd.file_size);
   size_t compact_bytes_per_del_file = std::numeric_limits<size_t>::max();
   // Compaction range will be [start, limit).
@@ -613,8 +601,7 @@ Compaction* CompactionPicker::CompactRange(
     int input_level, int output_level,
     const CompactRangeOptions& compact_range_options, const InternalKey* begin,
     const InternalKey* end, InternalKey** compaction_end, bool* manual_conflict,
-    uint64_t max_file_num_to_ignore, const std::string& trim_ts,
-    const SequenceNumber /*earliest_mem_seqno*/) {
+    uint64_t max_file_num_to_ignore, const std::string& trim_ts) {
   // CompactionPickerFIFO has its own implementation of compact range
   assert(ioptions_.compaction_style != kCompactionStyleFIFO);
 
@@ -890,7 +877,6 @@ Compaction* CompactionPicker::CompactRange(
   return compaction;
 }
 
-#ifndef ROCKSDB_LITE
 namespace {
 // Test whether two files have overlapping key-ranges.
 bool HaveOverlappingKeyRanges(const Comparator* c, const SstFileMetaData& a,
@@ -919,8 +905,7 @@ bool HaveOverlappingKeyRanges(const Comparator* c, const SstFileMetaData& a,
 
 Status CompactionPicker::SanitizeCompactionInputFilesForAllLevels(
     std::unordered_set<uint64_t>* input_files,
-    const ColumnFamilyMetaData& cf_meta, const int output_level,
-    const SequenceNumber earliest_mem_seqno) const {
+    const ColumnFamilyMetaData& cf_meta, const int output_level) const {
   auto& levels = cf_meta.levels;
   auto comparator = icmp_->user_comparator();
 
@@ -997,12 +982,6 @@ Status CompactionPicker::SanitizeCompactionInputFilesForAllLevels(
                                current_files[f].name +
                                " is currently being compacted.");
       }
-      if (output_level == 0 &&
-          current_files[f].largest_seqno > earliest_mem_seqno) {
-        return Status::Aborted(
-            "Necessary compaction input file " + current_files[f].name +
-            " has overlapping seqnos with earliest memtable seqnos.");
-      }
 
       input_files->insert(TableFileNameToNumber(current_files[f].name));
     }
@@ -1060,14 +1039,12 @@ Status CompactionPicker::SanitizeCompactionInputFilesForAllLevels(
         "A running compaction is writing to the same output level in an "
         "overlapping key range");
   }
-
   return Status::OK();
 }
 
 Status CompactionPicker::SanitizeCompactionInputFiles(
     std::unordered_set<uint64_t>* input_files,
-    const ColumnFamilyMetaData& cf_meta, const int output_level,
-    const SequenceNumber earliest_mem_seqno) const {
+    const ColumnFamilyMetaData& cf_meta, const int output_level) const {
   assert(static_cast<int>(cf_meta.levels.size()) - 1 ==
          cf_meta.levels[cf_meta.levels.size() - 1].level);
   if (output_level >= static_cast<int>(cf_meta.levels.size())) {
@@ -1093,8 +1070,8 @@ Status CompactionPicker::SanitizeCompactionInputFiles(
         "A compaction must contain at least one file.");
   }
 
-  Status s = SanitizeCompactionInputFilesForAllLevels(
-      input_files, cf_meta, output_level, earliest_mem_seqno);
+  Status s = SanitizeCompactionInputFilesForAllLevels(input_files, cf_meta,
+                                                      output_level);
 
   if (!s.ok()) {
     return s;
@@ -1138,7 +1115,6 @@ Status CompactionPicker::SanitizeCompactionInputFiles(
 
   return Status::OK();
 }
-#endif  // !ROCKSDB_LITE
 
 void CompactionPicker::RegisterCompaction(Compaction* c) {
   if (c == nullptr) {
@@ -1148,7 +1124,11 @@ void CompactionPicker::RegisterCompaction(Compaction* c) {
          c->output_level() == 0 ||
          !FilesRangeOverlapWithCompaction(*c->inputs(), c->output_level(),
                                           c->GetPenultimateLevel()));
-  if (c->start_level() == 0 ||
+  // CompactionReason::kExternalSstIngestion's start level is just a placeholder
+  // number without actual meaning as file ingestion technically does not have
+  // an input level like other compactions
+  if ((c->start_level() == 0 &&
+       c->compaction_reason() != CompactionReason::kExternalSstIngestion) ||
       ioptions_.compaction_style == kCompactionStyleUniversal) {
     level0_compactions_in_progress_.insert(c);
   }

db/compaction/compaction_picker.h

@@ -52,23 +52,15 @@ class CompactionPicker {
 
   // Pick level and inputs for a new compaction.
   //
-  // `earliest_mem_seqno` is the earliest seqno of unflushed memtables.
-  // It is needed to compare with compaction input SST files' largest seqnos
-  // in order to exclude those of seqnos potentially overlap with memtables'
-  // seqnos when doing compaction to L0. This will avoid creating a SST files in
-  // L0 newer than a unflushed memtable. Such SST file can exist in the first
-  // place when it's ingested or resulted from compaction involving files
-  // ingested.
-  //
   // Returns nullptr if there is no compaction to be done.
   // Otherwise returns a pointer to a heap-allocated object that
   // describes the compaction.  Caller should delete the result.
-  virtual Compaction* PickCompaction(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-      LogBuffer* log_buffer, const SequenceNumber earliest_mem_seqno) = 0;
+  virtual Compaction* PickCompaction(const std::string& cf_name,
+                                     const MutableCFOptions& mutable_cf_options,
+                                     const MutableDBOptions& mutable_db_options,
+                                     VersionStorageInfo* vstorage,
+                                     LogBuffer* log_buffer) = 0;
 
-  // `earliest_mem_seqno`: see PickCompaction() API
   // Return a compaction object for compacting the range [begin,end] in
   // the specified level.  Returns nullptr if there is nothing in that
   // level that overlaps the specified range.  Caller should delete
@@ -87,8 +79,7 @@ class CompactionPicker {
       const CompactRangeOptions& compact_range_options,
       const InternalKey* begin, const InternalKey* end,
       InternalKey** compaction_end, bool* manual_conflict,
-      uint64_t max_file_num_to_ignore, const std::string& trim_ts,
-      const SequenceNumber earliest_mem_seqno);
+      uint64_t max_file_num_to_ignore, const std::string& trim_ts);
 
   // The maximum allowed output level.  Default value is NumberLevels() - 1.
   virtual int MaxOutputLevel() const { return NumberLevels() - 1; }
@@ -102,18 +93,9 @@ class CompactionPicker {
 // into a valid one by adding more files, the function will return a
 // non-ok status with specific reason.
 //
-// Cases of returning non-ok status include but not limited to:
-// - When output_level == 0 and input_files contains sst files
-// of largest seqno greater than `earliest_mem_seqno`. This will
-// avoid creating a SST files in L0 newer than a unflushed memtable.
-// Such SST file can exist in the first place when it's ingested or
-// resulted from compaction involving files ingested.
-#ifndef ROCKSDB_LITE
-  Status SanitizeCompactionInputFiles(
-      std::unordered_set<uint64_t>* input_files,
-      const ColumnFamilyMetaData& cf_meta, const int output_level,
-      const SequenceNumber earliest_mem_seqno) const;
-#endif  // ROCKSDB_LITE
+  Status SanitizeCompactionInputFiles(std::unordered_set<uint64_t>* input_files,
+                                      const ColumnFamilyMetaData& cf_meta,
+                                      const int output_level) const;
 
   // Free up the files that participated in a compaction
   //
@@ -245,12 +227,9 @@ class CompactionPicker {
 
 // A helper function to SanitizeCompactionInputFiles() that
 // sanitizes "input_files" by adding necessary files.
-#ifndef ROCKSDB_LITE
   virtual Status SanitizeCompactionInputFilesForAllLevels(
       std::unordered_set<uint64_t>* input_files,
-      const ColumnFamilyMetaData& cf_meta, const int output_level,
-      const SequenceNumber earliest_mem_seqno) const;
-#endif  // ROCKSDB_LITE
+      const ColumnFamilyMetaData& cf_meta, const int output_level) const;
 
   // Keeps track of all compactions that are running on Level0.
   // Protected by DB mutex
@@ -263,7 +242,6 @@ class CompactionPicker {
   const InternalKeyComparator* const icmp_;
 };
 
-#ifndef ROCKSDB_LITE
 // A dummy compaction that never triggers any automatic
 // compaction.
 class NullCompactionPicker : public CompactionPicker {
@@ -274,27 +252,27 @@ class NullCompactionPicker : public CompactionPicker {
   virtual ~NullCompactionPicker() {}
 
   // Always return "nullptr"
-  Compaction* PickCompaction(
-      const std::string& /*cf_name*/,
-      const MutableCFOptions& /*mutable_cf_options*/,
-      const MutableDBOptions& /*mutable_db_options*/,
-      VersionStorageInfo* /*vstorage*/, LogBuffer* /* log_buffer */,
-      const SequenceNumber /* earliest_mem_seqno */) override {
+  Compaction* PickCompaction(const std::string& /*cf_name*/,
+                             const MutableCFOptions& /*mutable_cf_options*/,
+                             const MutableDBOptions& /*mutable_db_options*/,
+                             VersionStorageInfo* /*vstorage*/,
+                             LogBuffer* /* log_buffer */) override {
     return nullptr;
   }
 
   // Always return "nullptr"
-  Compaction* CompactRange(
-      const std::string& /*cf_name*/,
-      const MutableCFOptions& /*mutable_cf_options*/,
-      const MutableDBOptions& /*mutable_db_options*/,
-      VersionStorageInfo* /*vstorage*/, int /*input_level*/,
-      int /*output_level*/,
-      const CompactRangeOptions& /*compact_range_options*/,
-      const InternalKey* /*begin*/, const InternalKey* /*end*/,
-      InternalKey** /*compaction_end*/, bool* /*manual_conflict*/,
-      uint64_t /*max_file_num_to_ignore*/, const std::string& /*trim_ts*/,
-      const SequenceNumber /* earliest_mem_seqno */) override {
+  Compaction* CompactRange(const std::string& /*cf_name*/,
+                           const MutableCFOptions& /*mutable_cf_options*/,
+                           const MutableDBOptions& /*mutable_db_options*/,
+                           VersionStorageInfo* /*vstorage*/,
+                           int /*input_level*/, int /*output_level*/,
+                           const CompactRangeOptions& /*compact_range_options*/,
+                           const InternalKey* /*begin*/,
+                           const InternalKey* /*end*/,
+                           InternalKey** /*compaction_end*/,
+                           bool* /*manual_conflict*/,
+                           uint64_t /*max_file_num_to_ignore*/,
+                           const std::string& /*trim_ts*/) override {
     return nullptr;
   }
 
@@ -304,7 +282,6 @@ class NullCompactionPicker : public CompactionPicker {
     return false;
   }
 };
-#endif  // !ROCKSDB_LITE
 
 // Attempts to find an intra L0 compaction conforming to the given parameters.
 //
@@ -321,14 +298,12 @@ class NullCompactionPicker : public CompactionPicker {
 //                                        initialized with corresponding input
 //                                        files. Cannot be nullptr.
 //
-// @param earliest_mem_seqno              See PickCompaction() API
 // @return                                true iff compaction was found.
 bool FindIntraL0Compaction(const std::vector<FileMetaData*>& level_files,
                            size_t min_files_to_compact,
                            uint64_t max_compact_bytes_per_del_file,
                            uint64_t max_compaction_bytes,
-                           CompactionInputFiles* comp_inputs,
-                           const SequenceNumber earliest_mem_seqno);
+                           CompactionInputFiles* comp_inputs);
 
 CompressionType GetCompressionType(const VersionStorageInfo* vstorage,
                                    const MutableCFOptions& mutable_cf_options,

db/compaction/compaction_picker_fifo.cc

@@ -8,7 +8,6 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/compaction/compaction_picker_fifo.h"
-#ifndef ROCKSDB_LITE
 
 #include <cinttypes>
 #include <string>
@@ -17,6 +16,7 @@
 #include "db/column_family.h"
 #include "logging/log_buffer.h"
 #include "logging/logging.h"
+#include "options/options_helper.h"
 #include "util/string_util.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -139,7 +139,7 @@ Compaction* FIFOCompactionPicker::PickTTLCompaction(
 Compaction* FIFOCompactionPicker::PickSizeCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer, const SequenceNumber earliest_mem_seqno) {
+    LogBuffer* log_buffer) {
   // compute the total size and identify the last non-empty level
   int last_level = 0;
   uint64_t total_size = 0;
@@ -176,8 +176,7 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
                   .level0_file_num_compaction_trigger /* min_files_to_compact */
               ,
               max_compact_bytes_per_del_file,
-              mutable_cf_options.max_compaction_bytes, &comp_inputs,
-              earliest_mem_seqno)) {
+              mutable_cf_options.max_compaction_bytes, &comp_inputs)) {
         Compaction* c = new Compaction(
             vstorage, ioptions_, mutable_cf_options, mutable_db_options,
             {comp_inputs}, 0, 16 * 1024 * 1024 /* output file size limit */,
@@ -233,7 +232,8 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
         break;
       }
     }
-  } else {
+  } else if (total_size >
+             mutable_cf_options.compaction_options_fifo.max_table_files_size) {
     // If the last level is non-L0, we actually don't know which file is
     // logically the oldest since the file creation time only represents
     // when this file was compacted to this level, which is independent
@@ -248,15 +248,27 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
       inputs[0].files.push_back(f);
       char tmp_fsize[16];
       AppendHumanBytes(f->fd.GetFileSize(), tmp_fsize, sizeof(tmp_fsize));
-      ROCKS_LOG_BUFFER(log_buffer,
-                       "[%s] FIFO compaction: picking file %" PRIu64
-                       " with size %s for deletion",
-                       cf_name.c_str(), f->fd.GetNumber(), tmp_fsize);
+      ROCKS_LOG_BUFFER(
+          log_buffer,
+          "[%s] FIFO compaction: picking file %" PRIu64
+          " with size %s for deletion under total size %" PRIu64
+          " vs max table files size %" PRIu64,
+          cf_name.c_str(), f->fd.GetNumber(), tmp_fsize, total_size,
+          mutable_cf_options.compaction_options_fifo.max_table_files_size);
+
       if (total_size <=
           mutable_cf_options.compaction_options_fifo.max_table_files_size) {
         break;
       }
     }
+  } else {
+    ROCKS_LOG_BUFFER(
+        log_buffer,
+        "[%s] FIFO compaction: nothing to do. Total size %" PRIu64
+        ", max size %" PRIu64 "\n",
+        cf_name.c_str(), total_size,
+        mutable_cf_options.compaction_options_fifo.max_table_files_size);
+    return nullptr;
   }
 
   Compaction* c = new Compaction(
@@ -273,36 +285,38 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
   return c;
 }
 
-Compaction* FIFOCompactionPicker::PickCompactionToWarm(
+Compaction* FIFOCompactionPicker::PickTemperatureChangeCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer, const SequenceNumber earliest_mem_seqno) {
-  TEST_SYNC_POINT("PickCompactionToWarm");
-  if (mutable_cf_options.compaction_options_fifo.age_for_warm == 0) {
+    LogBuffer* log_buffer) {
+  const std::vector<FileTemperatureAge>& ages =
+      mutable_cf_options.compaction_options_fifo
+          .file_temperature_age_thresholds;
+  if (ages.empty()) {
     return nullptr;
   }
 
-  // PickCompactionToWarm is only triggered if there is no non-L0 files.
-  for (int level = 1; level < vstorage->num_levels(); ++level) {
-    if (GetTotalFilesSize(vstorage->LevelFiles(level)) > 0) {
-      return nullptr;
-    }
+  // Does not apply to multi-level FIFO.
+  if (vstorage->num_levels() > 1) {
+    return nullptr;
   }
 
   const int kLevel0 = 0;
   const std::vector<FileMetaData*>& level_files = vstorage->LevelFiles(kLevel0);
+  if (level_files.empty()) {
+    return nullptr;
+  }
 
   int64_t _current_time;
   auto status = ioptions_.clock->GetCurrentTime(&_current_time);
   if (!status.ok()) {
-    ROCKS_LOG_BUFFER(log_buffer,
-                     "[%s] FIFO compaction: Couldn't get current time: %s. "
-                     "Not doing compactions based on warm threshold. ",
-                     cf_name.c_str(), status.ToString().c_str());
+    ROCKS_LOG_BUFFER(
+        log_buffer,
+        "[%s] FIFO compaction: Couldn't get current time: %s. "
+        "Not doing compactions based on file temperature-age threshold. ",
+        cf_name.c_str(), status.ToString().c_str());
     return nullptr;
   }
-  TEST_SYNC_POINT_CALLBACK("PickCompactionToWarm::BeforeGetCurrentTime",
-                           &_current_time);
   const uint64_t current_time = static_cast<uint64_t>(_current_time);
 
   if (!level0_compactions_in_progress_.empty()) {
@@ -319,57 +333,77 @@ Compaction* FIFOCompactionPicker::PickCompactionToWarm(
   inputs[0].level = 0;
 
   // avoid underflow
-  if (current_time > mutable_cf_options.compaction_options_fifo.age_for_warm) {
-    uint64_t create_time_threshold =
-        current_time - mutable_cf_options.compaction_options_fifo.age_for_warm;
+  uint64_t min_age = ages[0].age;
+  // kLastTemperature means target temperature is to be determined.
+  Temperature compaction_target_temp = Temperature::kLastTemperature;
+  if (current_time > min_age) {
+    uint64_t create_time_threshold = current_time - min_age;
     uint64_t compaction_size = 0;
-    // We will ideally identify a file qualifying for warm tier by knowing
-    // the timestamp for the youngest entry in the file. However, right now
-    // we don't have the information. We infer it by looking at timestamp
-    // of the next file's (which is just younger) oldest entry's timestamp.
-    FileMetaData* prev_file = nullptr;
-    for (auto ritr = level_files.rbegin(); ritr != level_files.rend(); ++ritr) {
-      FileMetaData* f = *ritr;
-      assert(f);
-      if (f->being_compacted) {
-        // Right now this probably won't happen as we never try to schedule
-        // two compactions in parallel, so here we just simply don't schedule
-        // anything.
+    // We will ideally identify a file qualifying for temperature change by
+    // knowing the timestamp for the youngest entry in the file. However, right
+    // now we don't have the information. We infer it by looking at timestamp of
+    // the previous file's (which is just younger) oldest entry's timestamp.
+    Temperature cur_target_temp;
+    // avoid index underflow
+    assert(level_files.size() >= 1);
+    for (size_t index = level_files.size() - 1; index >= 1; --index) {
+      // Try to add cur_file to compaction inputs.
+      FileMetaData* cur_file = level_files[index];
+      // prev_file is just younger than cur_file
+      FileMetaData* prev_file = level_files[index - 1];
+      if (cur_file->being_compacted) {
+        // Should not happen since we check for
+        // `level0_compactions_in_progress_` above. Here we simply just don't
+        // schedule anything.
         return nullptr;
       }
-      uint64_t oldest_ancester_time = f->TryGetOldestAncesterTime();
-      if (oldest_ancester_time == kUnknownOldestAncesterTime) {
+      uint64_t oldest_ancestor_time = prev_file->TryGetOldestAncesterTime();
+      if (oldest_ancestor_time == kUnknownOldestAncesterTime) {
         // Older files might not have enough information. It is possible to
         // handle these files by looking at newer files, but maintaining the
         // logic isn't worth it.
         break;
       }
-      if (oldest_ancester_time > create_time_threshold) {
-        // The previous file (which has slightly older data) doesn't qualify
-        // for warm tier.
+      if (oldest_ancestor_time > create_time_threshold) {
+        // cur_file is too fresh
         break;
       }
-      if (prev_file != nullptr &&
-          prev_file->fd.largest_seqno <= earliest_mem_seqno) {
-        compaction_size += prev_file->fd.GetFileSize();
-        if (compaction_size > mutable_cf_options.max_compaction_bytes) {
+      cur_target_temp = ages[0].temperature;
+      for (size_t i = 1; i < ages.size(); ++i) {
+        if (current_time >= ages[i].age &&
+            oldest_ancestor_time <= current_time - ages[i].age) {
+          cur_target_temp = ages[i].temperature;
+        }
+      }
+      if (cur_file->temperature == cur_target_temp) {
+        if (inputs[0].empty()) {
+          continue;
+        } else {
           break;
         }
-        inputs[0].files.push_back(prev_file);
-        ROCKS_LOG_BUFFER(log_buffer,
-                         "[%s] FIFO compaction: picking file %" PRIu64
-                         " with next file's oldest time %" PRIu64 " for warm",
-                         cf_name.c_str(), prev_file->fd.GetNumber(),
-                         oldest_ancester_time);
       }
-      if (f->temperature == Temperature::kUnknown ||
-          f->temperature == Temperature::kHot) {
-        prev_file = f;
-      } else if (!inputs[0].files.empty()) {
-        // A warm file newer than files picked.
+
+      // cur_file needs to change temperature
+      if (compaction_target_temp == Temperature::kLastTemperature) {
+        assert(inputs[0].empty());
+        compaction_target_temp = cur_target_temp;
+      } else if (cur_target_temp != compaction_target_temp) {
+        assert(!inputs[0].empty());
+        break;
+      }
+      if (inputs[0].empty() || compaction_size + cur_file->fd.GetFileSize() <=
+                                   mutable_cf_options.max_compaction_bytes) {
+        inputs[0].files.push_back(cur_file);
+        compaction_size += cur_file->fd.GetFileSize();
+        ROCKS_LOG_BUFFER(
+            log_buffer,
+            "[%s] FIFO compaction: picking file %" PRIu64
+            " with next file's oldest time %" PRIu64 " for temperature %s.",
+            cf_name.c_str(), cur_file->fd.GetNumber(), oldest_ancestor_time,
+            temperature_to_string[cur_target_temp].c_str());
+      }
+      if (compaction_size > mutable_cf_options.max_compaction_bytes) {
         break;
-      } else {
-        assert(prev_file == nullptr);
       }
     }
   }
@@ -383,7 +417,7 @@ Compaction* FIFOCompactionPicker::PickCompactionToWarm(
       std::move(inputs), 0, 0 /* output file size limit */,
       0 /* max compaction bytes, not applicable */, 0 /* output path ID */,
       mutable_cf_options.compression, mutable_cf_options.compression_opts,
-      Temperature::kWarm,
+      compaction_target_temp,
       /* max_subcompactions */ 0, {}, /* is manual */ false, /* trim_ts */ "",
       vstorage->CompactionScore(0),
       /* is deletion compaction */ false, /* l0_files_might_overlap */ true,
@@ -394,7 +428,7 @@ Compaction* FIFOCompactionPicker::PickCompactionToWarm(
 Compaction* FIFOCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer, const SequenceNumber earliest_mem_seqno) {
+    LogBuffer* log_buffer) {
   Compaction* c = nullptr;
   if (mutable_cf_options.ttl > 0) {
     c = PickTTLCompaction(cf_name, mutable_cf_options, mutable_db_options,
@@ -402,11 +436,11 @@ Compaction* FIFOCompactionPicker::PickCompaction(
   }
   if (c == nullptr) {
     c = PickSizeCompaction(cf_name, mutable_cf_options, mutable_db_options,
-                           vstorage, log_buffer, earliest_mem_seqno);
+                           vstorage, log_buffer);
   }
   if (c == nullptr) {
-    c = PickCompactionToWarm(cf_name, mutable_cf_options, mutable_db_options,
-                             vstorage, log_buffer, earliest_mem_seqno);
+    c = PickTemperatureChangeCompaction(
+        cf_name, mutable_cf_options, mutable_db_options, vstorage, log_buffer);
   }
   RegisterCompaction(c);
   return c;
@@ -419,8 +453,7 @@ Compaction* FIFOCompactionPicker::CompactRange(
     const CompactRangeOptions& /*compact_range_options*/,
     const InternalKey* /*begin*/, const InternalKey* /*end*/,
     InternalKey** compaction_end, bool* /*manual_conflict*/,
-    uint64_t /*max_file_num_to_ignore*/, const std::string& /*trim_ts*/,
-    const SequenceNumber earliest_mem_seqno) {
+    uint64_t /*max_file_num_to_ignore*/, const std::string& /*trim_ts*/) {
 #ifdef NDEBUG
   (void)input_level;
   (void)output_level;
@@ -429,12 +462,10 @@ Compaction* FIFOCompactionPicker::CompactRange(
   assert(output_level == 0);
   *compaction_end = nullptr;
   LogBuffer log_buffer(InfoLogLevel::INFO_LEVEL, ioptions_.logger);
-  Compaction* c =
-      PickCompaction(cf_name, mutable_cf_options, mutable_db_options, vstorage,
-                     &log_buffer, earliest_mem_seqno);
+  Compaction* c = PickCompaction(cf_name, mutable_cf_options,
+                                 mutable_db_options, vstorage, &log_buffer);
   log_buffer.FlushBufferToLog();
   return c;
 }
 
 }  // namespace ROCKSDB_NAMESPACE
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_picker_fifo.h

@@ -8,7 +8,6 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #pragma once
-#ifndef ROCKSDB_LITE
 
 #include "db/compaction/compaction_picker.h"
 
@@ -19,15 +18,12 @@ class FIFOCompactionPicker : public CompactionPicker {
                        const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
 
-  virtual Compaction* PickCompaction(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      const MutableDBOptions& mutable_db_options, VersionStorageInfo* version,
-      LogBuffer* log_buffer, const SequenceNumber earliest_mem_seqno) override;
+  virtual Compaction* PickCompaction(const std::string& cf_name,
+                                     const MutableCFOptions& mutable_cf_options,
+                                     const MutableDBOptions& mutable_db_options,
+                                     VersionStorageInfo* version,
+                                     LogBuffer* log_buffer) override;
 
-  // `earliest_mem_seqno`: see PickCompaction() API for more. In FIFO's
-  // implementation of CompactRange(), different from others, we will not return
-  // `nullptr` right away when intput files of compaction to L0 has seqnos
-  // potentially overlapping with memtable's but exlucde those files.
   virtual Compaction* CompactRange(
       const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
       const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
@@ -35,8 +31,7 @@ class FIFOCompactionPicker : public CompactionPicker {
       const CompactRangeOptions& compact_range_options,
       const InternalKey* begin, const InternalKey* end,
       InternalKey** compaction_end, bool* manual_conflict,
-      uint64_t max_file_num_to_ignore, const std::string& trim_ts,
-      const SequenceNumber earliest_mem_seqno) override;
+      uint64_t max_file_num_to_ignore, const std::string& trim_ts) override;
 
   // The maximum allowed output level.  Always returns 0.
   virtual int MaxOutputLevel() const override { return 0; }
@@ -55,15 +50,11 @@ class FIFOCompactionPicker : public CompactionPicker {
                                  const MutableCFOptions& mutable_cf_options,
                                  const MutableDBOptions& mutable_db_options,
                                  VersionStorageInfo* version,
-                                 LogBuffer* log_buffer,
-                                 SequenceNumber earliest_mem_seqno);
+                                 LogBuffer* log_buffer);
 
-  Compaction* PickCompactionToWarm(const std::string& cf_name,
-                                   const MutableCFOptions& mutable_cf_options,
-                                   const MutableDBOptions& mutable_db_options,
-                                   VersionStorageInfo* version,
-                                   LogBuffer* log_buffer,
-                                   const SequenceNumber earliest_mem_seqno);
+  Compaction* PickTemperatureChangeCompaction(
+      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
+      LogBuffer* log_buffer);
 };
 }  // namespace ROCKSDB_NAMESPACE
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_picker_level.cc

@@ -50,7 +50,6 @@ class LevelCompactionBuilder {
  public:
   LevelCompactionBuilder(const std::string& cf_name,
                          VersionStorageInfo* vstorage,
-                         const SequenceNumber earliest_mem_seqno,
                          CompactionPicker* compaction_picker,
                          LogBuffer* log_buffer,
                          const MutableCFOptions& mutable_cf_options,
@@ -58,7 +57,6 @@ class LevelCompactionBuilder {
                          const MutableDBOptions& mutable_db_options)
       : cf_name_(cf_name),
         vstorage_(vstorage),
-        earliest_mem_seqno_(earliest_mem_seqno),
         compaction_picker_(compaction_picker),
         log_buffer_(log_buffer),
         mutable_cf_options_(mutable_cf_options),
@@ -85,7 +83,7 @@ class LevelCompactionBuilder {
 
   Compaction* GetCompaction();
 
-  // For the specfied level, pick a file that we want to compact.
+  // From `start_level_`, pick files to compact to `output_level_`.
   // Returns false if there is no file to compact.
   // If it returns true, inputs->files.size() will be exactly one for
   // all compaction priorities except round-robin. For round-robin,
@@ -109,8 +107,9 @@ class LevelCompactionBuilder {
   bool PickIntraL0Compaction();
 
   // Return true if TrivialMove is extended. `start_index` is the index of
-  // the intiial file picked, which should already be in `start_level_inputs_`.
-  bool TryExtendNonL0TrivialMove(int start_index);
+  // the initial file picked, which should already be in `start_level_inputs_`.
+  bool TryExtendNonL0TrivialMove(int start_index,
+                                 bool only_expand_right = false);
 
   // Picks a file from level_files to compact.
   // level_files is a vector of (level, file metadata) in ascending order of
@@ -122,7 +121,6 @@ class LevelCompactionBuilder {
 
   const std::string& cf_name_;
   VersionStorageInfo* vstorage_;
-  const SequenceNumber earliest_mem_seqno_;
   CompactionPicker* compaction_picker_;
   LogBuffer* log_buffer_;
   int start_level_ = -1;
@@ -196,7 +194,10 @@ void LevelCompactionBuilder::SetupInitialFiles() {
       }
       output_level_ =
           (start_level_ == 0) ? vstorage_->base_level() : start_level_ + 1;
-      if (PickFileToCompact()) {
+      bool picked_file_to_compact = PickFileToCompact();
+      TEST_SYNC_POINT_CALLBACK("PostPickFileToCompact",
+                               &picked_file_to_compact);
+      if (picked_file_to_compact) {
         // found the compaction!
         if (start_level_ == 0) {
           // L0 score = `num L0 files` / `level0_file_num_compaction_trigger`
@@ -355,7 +356,8 @@ void LevelCompactionBuilder::SetupOtherFilesWithRoundRobinExpansion() {
   vstorage_->GetOverlappingInputs(output_level_, &smallest, &largest,
                                   &output_level_inputs.files);
   if (output_level_inputs.empty()) {
-    if (TryExtendNonL0TrivialMove((int)start_index)) {
+    if (TryExtendNonL0TrivialMove((int)start_index,
+                                  true /* only_expand_right */)) {
       return;
     }
   }
@@ -447,21 +449,21 @@ bool LevelCompactionBuilder::SetupOtherInputsIfNeeded() {
       compaction_inputs_.push_back(output_level_inputs_);
     }
 
+    // In some edge cases we could pick a compaction that will be compacting
+    // a key range that overlap with another running compaction, and both
+    // of them have the same output level. This could happen if
+    // (1) we are running a non-exclusive manual compaction
+    // (2) AddFile ingest a new file into the LSM tree
+    // We need to disallow this from happening.
+    if (compaction_picker_->FilesRangeOverlapWithCompaction(
+            compaction_inputs_, output_level_,
+            Compaction::EvaluatePenultimateLevel(
+                vstorage_, ioptions_, start_level_, output_level_))) {
+      // This compaction output could potentially conflict with the output
+      // of a currently running compaction, we cannot run it.
+      return false;
+    }
     if (!is_l0_trivial_move_) {
-      // In some edge cases we could pick a compaction that will be compacting
-      // a key range that overlap with another running compaction, and both
-      // of them have the same output level. This could happen if
-      // (1) we are running a non-exclusive manual compaction
-      // (2) AddFile ingest a new file into the LSM tree
-      // We need to disallow this from happening.
-      if (compaction_picker_->FilesRangeOverlapWithCompaction(
-              compaction_inputs_, output_level_,
-              Compaction::EvaluatePenultimateLevel(
-                  vstorage_, ioptions_, start_level_, output_level_))) {
-        // This compaction output could potentially conflict with the output
-        // of a currently running compaction, we cannot run it.
-        return false;
-      }
       compaction_picker_->GetGrandparents(vstorage_, start_level_inputs_,
                                           output_level_inputs_, &grandparents_);
     }
@@ -501,6 +503,16 @@ Compaction* LevelCompactionBuilder::PickCompaction() {
 }
 
 Compaction* LevelCompactionBuilder::GetCompaction() {
+  // TryPickL0TrivialMove() does not apply to the case when compacting L0 to an
+  // empty output level. So L0 files is picked in PickFileToCompact() by
+  // compaction score. We may still be able to do trivial move when this file
+  // does not overlap with other L0s. This happens when
+  // compaction_inputs_[0].size() == 1 since SetupOtherL0FilesIfNeeded() did not
+  // pull in more L0s.
+  assert(!compaction_inputs_.empty());
+  bool l0_files_might_overlap =
+      start_level_ == 0 && !is_l0_trivial_move_ &&
+      (compaction_inputs_.size() > 1 || compaction_inputs_[0].size() > 1);
   auto c = new Compaction(
       vstorage_, ioptions_, mutable_cf_options_, mutable_db_options_,
       std::move(compaction_inputs_), output_level_,
@@ -515,8 +527,7 @@ Compaction* LevelCompactionBuilder::GetCompaction() {
       Temperature::kUnknown,
       /* max_subcompactions */ 0, std::move(grandparents_), is_manual_,
       /* trim_ts */ "", start_level_score_, false /* deletion_compaction */,
-      /* l0_files_might_overlap */ start_level_ == 0 && !is_l0_trivial_move_,
-      compaction_reason_);
+      l0_files_might_overlap, compaction_reason_);
 
   // If it's level 0 compaction, make sure we don't execute any other level 0
   // compactions in parallel
@@ -653,7 +664,8 @@ bool LevelCompactionBuilder::TryPickL0TrivialMove() {
   return false;
 }
 
-bool LevelCompactionBuilder::TryExtendNonL0TrivialMove(int start_index) {
+bool LevelCompactionBuilder::TryExtendNonL0TrivialMove(int start_index,
+                                                       bool only_expand_right) {
   if (start_level_inputs_.size() == 1 &&
       (ioptions_.db_paths.empty() || ioptions_.db_paths.size() == 1) &&
       (mutable_cf_options_.compression_per_level.empty())) {
@@ -670,6 +682,7 @@ bool LevelCompactionBuilder::TryExtendNonL0TrivialMove(int start_index) {
     size_t total_size = initial_file->fd.GetFileSize();
     CompactionInputFiles output_level_inputs;
     output_level_inputs.level = output_level_;
+    // Expand towards right
     for (int i = start_index + 1;
          i < static_cast<int>(level_files.size()) &&
          start_level_inputs_.size() < kMaxMultiTrivialMove;
@@ -702,6 +715,37 @@ bool LevelCompactionBuilder::TryExtendNonL0TrivialMove(int start_index) {
       }
       start_level_inputs_.files.push_back(next_file);
     }
+    // Expand towards left
+    if (!only_expand_right) {
+      for (int i = start_index - 1;
+           i >= 0 && start_level_inputs_.size() < kMaxMultiTrivialMove; i--) {
+        FileMetaData* next_file = level_files[i];
+        if (next_file->being_compacted) {
+          break;
+        }
+        vstorage_->GetOverlappingInputs(output_level_, &(next_file->smallest),
+                                        &(initial_file->largest),
+                                        &output_level_inputs.files);
+        if (!output_level_inputs.empty()) {
+          break;
+        }
+        if (i > 0 && compaction_picker_->icmp()
+                             ->user_comparator()
+                             ->CompareWithoutTimestamp(
+                                 next_file->smallest.user_key(),
+                                 level_files[i - 1]->largest.user_key()) == 0) {
+          // Not a clean up after adding the next file. Skip.
+          break;
+        }
+        total_size += next_file->fd.GetFileSize();
+        if (total_size > mutable_cf_options_.max_compaction_bytes) {
+          break;
+        }
+        // keep `files` sorted in increasing order by key range
+        start_level_inputs_.files.insert(start_level_inputs_.files.begin(),
+                                         next_file);
+      }
+    }
     return start_level_inputs_.size() > 1;
   }
   return false;
@@ -785,7 +829,10 @@ bool LevelCompactionBuilder::PickFileToCompact() {
     vstorage_->GetOverlappingInputs(output_level_, &smallest, &largest,
                                     &output_level_inputs.files);
     if (output_level_inputs.empty()) {
-      if (TryExtendNonL0TrivialMove(index)) {
+      if (start_level_ > 0 &&
+          TryExtendNonL0TrivialMove(index,
+                                    ioptions_.compaction_pri ==
+                                        kRoundRobin /* only_expand_right */)) {
         break;
       }
     } else {
@@ -825,16 +872,16 @@ bool LevelCompactionBuilder::PickIntraL0Compaction() {
   return FindIntraL0Compaction(level_files, kMinFilesForIntraL0Compaction,
                                std::numeric_limits<uint64_t>::max(),
                                mutable_cf_options_.max_compaction_bytes,
-                               &start_level_inputs_, earliest_mem_seqno_);
+                               &start_level_inputs_);
 }
 }  // namespace
 
 Compaction* LevelCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer, const SequenceNumber earliest_mem_seqno) {
-  LevelCompactionBuilder builder(cf_name, vstorage, earliest_mem_seqno, this,
-                                 log_buffer, mutable_cf_options, ioptions_,
+    LogBuffer* log_buffer) {
+  LevelCompactionBuilder builder(cf_name, vstorage, this, log_buffer,
+                                 mutable_cf_options, ioptions_,
                                  mutable_db_options);
   return builder.PickCompaction();
 }

db/compaction/compaction_picker_level.h

@@ -20,10 +20,11 @@ class LevelCompactionPicker : public CompactionPicker {
   LevelCompactionPicker(const ImmutableOptions& ioptions,
                         const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
-  virtual Compaction* PickCompaction(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-      LogBuffer* log_buffer, const SequenceNumber earliest_mem_seqno) override;
+  virtual Compaction* PickCompaction(const std::string& cf_name,
+                                     const MutableCFOptions& mutable_cf_options,
+                                     const MutableDBOptions& mutable_db_options,
+                                     VersionStorageInfo* vstorage,
+                                     LogBuffer* log_buffer) override;
 
   virtual bool NeedsCompaction(
       const VersionStorageInfo* vstorage) const override;

db/compaction/compaction_picker_universal.cc

@@ -8,7 +8,6 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/compaction/compaction_picker_universal.h"
-#ifndef ROCKSDB_LITE
 
 #include <cinttypes>
 #include <limits>
@@ -20,7 +19,7 @@
 #include "file/filename.h"
 #include "logging/log_buffer.h"
 #include "logging/logging.h"
-#include "monitoring/statistics.h"
+#include "monitoring/statistics_impl.h"
 #include "test_util/sync_point.h"
 #include "util/random.h"
 #include "util/string_util.h"
@@ -293,7 +292,7 @@ bool UniversalCompactionPicker::NeedsCompaction(
 Compaction* UniversalCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer, const SequenceNumber /* earliest_mem_seqno */) {
+    LogBuffer* log_buffer) {
   UniversalCompactionBuilder builder(ioptions_, icmp_, cf_name,
                                      mutable_cf_options, mutable_db_options,
                                      vstorage, this, log_buffer);
@@ -400,6 +399,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
   if (!vstorage_->FilesMarkedForPeriodicCompaction().empty()) {
     // Always need to do a full compaction for periodic compaction.
     c = PickPeriodicCompaction();
+    TEST_SYNC_POINT_CALLBACK("PostPickPeriodicCompaction", c);
   }
 
   // Check for size amplification.
@@ -408,6 +408,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
           static_cast<size_t>(
               mutable_cf_options_.level0_file_num_compaction_trigger)) {
     if ((c = PickCompactionToReduceSizeAmp()) != nullptr) {
+      TEST_SYNC_POINT("PickCompactionToReduceSizeAmpReturnNonnullptr");
       ROCKS_LOG_BUFFER(log_buffer_, "[%s] Universal: compacting for size amp\n",
                        cf_name_.c_str());
     } else {
@@ -417,6 +418,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
           mutable_cf_options_.compaction_options_universal.size_ratio;
 
       if ((c = PickCompactionToReduceSortedRuns(ratio, UINT_MAX)) != nullptr) {
+        TEST_SYNC_POINT("PickCompactionToReduceSortedRunsReturnNonnullptr");
         ROCKS_LOG_BUFFER(log_buffer_,
                          "[%s] Universal: compacting for size ratio\n",
                          cf_name_.c_str());
@@ -457,6 +459,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
 
   if (c == nullptr) {
     if ((c = PickDeleteTriggeredCompaction()) != nullptr) {
+      TEST_SYNC_POINT("PickDeleteTriggeredCompactionReturnNonnullptr");
       ROCKS_LOG_BUFFER(log_buffer_,
                        "[%s] Universal: delete triggered compaction\n",
                        cf_name_.c_str());
@@ -1447,4 +1450,3 @@ uint64_t UniversalCompactionBuilder::GetMaxOverlappingBytes() const {
 }
 }  // namespace ROCKSDB_NAMESPACE
 
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_picker_universal.h

@@ -8,7 +8,6 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #pragma once
-#ifndef ROCKSDB_LITE
 
 #include "db/compaction/compaction_picker.h"
 
@@ -18,14 +17,14 @@ class UniversalCompactionPicker : public CompactionPicker {
   UniversalCompactionPicker(const ImmutableOptions& ioptions,
                             const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
-  virtual Compaction* PickCompaction(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-      LogBuffer* log_buffer, const SequenceNumber earliest_mem_seqno) override;
+  virtual Compaction* PickCompaction(const std::string& cf_name,
+                                     const MutableCFOptions& mutable_cf_options,
+                                     const MutableDBOptions& mutable_db_options,
+                                     VersionStorageInfo* vstorage,
+                                     LogBuffer* log_buffer) override;
   virtual int MaxOutputLevel() const override { return NumberLevels() - 1; }
 
   virtual bool NeedsCompaction(
       const VersionStorageInfo* vstorage) const override;
 };
 }  // namespace ROCKSDB_NAMESPACE
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_service_job.cc

@@ -16,7 +16,6 @@
 #include "options/options_helper.h"
 #include "rocksdb/utilities/options_type.h"
 
-#ifndef ROCKSDB_LITE
 namespace ROCKSDB_NAMESPACE {
 class SubcompactionState;
 
@@ -190,6 +189,7 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
     meta.largest.DecodeFrom(file.largest_internal_key);
     meta.oldest_ancester_time = file.oldest_ancester_time;
     meta.file_creation_time = file.file_creation_time;
+    meta.epoch_number = file.epoch_number;
     meta.marked_for_compaction = file.marked_for_compaction;
     meta.unique_id = file.unique_id;
 
@@ -333,8 +333,9 @@ Status CompactionServiceCompactionJob::Run() {
         MakeTableFileName(meta.fd.GetNumber()), meta.fd.smallest_seqno,
         meta.fd.largest_seqno, meta.smallest.Encode().ToString(),
         meta.largest.Encode().ToString(), meta.oldest_ancester_time,
-        meta.file_creation_time, output_file.validator.GetHash(),
-        meta.marked_for_compaction, meta.unique_id);
+        meta.file_creation_time, meta.epoch_number,
+        output_file.validator.GetHash(), meta.marked_for_compaction,
+        meta.unique_id);
   }
   InternalStats::CompactionStatsFull compaction_stats;
   sub_compact->AggregateCompactionStats(compaction_stats);
@@ -489,6 +490,10 @@ static std::unordered_map<std::string, OptionTypeInfo>
          {offsetof(struct CompactionServiceOutputFile, file_creation_time),
           OptionType::kUInt64T, OptionVerificationType::kNormal,
           OptionTypeFlags::kNone}},
+        {"epoch_number",
+         {offsetof(struct CompactionServiceOutputFile, epoch_number),
+          OptionType::kUInt64T, OptionVerificationType::kNormal,
+          OptionTypeFlags::kNone}},
         {"paranoid_hash",
          {offsetof(struct CompactionServiceOutputFile, paranoid_hash),
           OptionType::kUInt64T, OptionVerificationType::kNormal,
@@ -826,4 +831,3 @@ bool CompactionServiceInput::TEST_Equals(CompactionServiceInput* other,
 #endif  // NDEBUG
 }  // namespace ROCKSDB_NAMESPACE
 
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_service_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-#ifndef ROCKSDB_LITE
 
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
@@ -929,7 +928,7 @@ TEST_F(CompactionServiceTest, TablePropertiesCollector) {
     }
     ASSERT_OK(Flush());
   }
-  ASSERT_OK(dbfull()->TEST_WaitForCompact(true));
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
 
   ASSERT_OK(db_->GetPropertiesOfAllTables(&fname_to_props));
 
@@ -953,14 +952,3 @@ int main(int argc, char** argv) {
   RegisterCustomObjects(argc, argv);
   return RUN_ALL_TESTS();
 }
-
-#else
-#include <stdio.h>
-
-int main(int /*argc*/, char** /*argv*/) {
-  fprintf(stderr,
-          "SKIPPED as CompactionService is not supported in ROCKSDB_LITE\n");
-  return 0;
-}
-
-#endif  // ROCKSDB_LITE

db/compaction/sst_partitioner.cc

@@ -15,11 +15,9 @@
 namespace ROCKSDB_NAMESPACE {
 static std::unordered_map<std::string, OptionTypeInfo>
     sst_fixed_prefix_type_info = {
-#ifndef ROCKSDB_LITE
         {"length",
          {0, OptionType::kSizeT, OptionVerificationType::kNormal,
           OptionTypeFlags::kNone}},
-#endif  // ROCKSDB_LITE
 };
 
 SstPartitionerFixedPrefixFactory::SstPartitionerFixedPrefixFactory(size_t len)
@@ -58,7 +56,6 @@ std::shared_ptr<SstPartitionerFactory> NewSstPartitionerFixedPrefixFactory(
   return std::make_shared<SstPartitionerFixedPrefixFactory>(prefix_len);
 }
 
-#ifndef ROCKSDB_LITE
 namespace {
 static int RegisterSstPartitionerFactories(ObjectLibrary& library,
                                            const std::string& /*arg*/) {
@@ -73,18 +70,14 @@ static int RegisterSstPartitionerFactories(ObjectLibrary& library,
   return 1;
 }
 }  // namespace
-#endif  // ROCKSDB_LITE
 
 Status SstPartitionerFactory::CreateFromString(
     const ConfigOptions& options, const std::string& value,
     std::shared_ptr<SstPartitionerFactory>* result) {
-#ifndef ROCKSDB_LITE
   static std::once_flag once;
   std::call_once(once, [&]() {
     RegisterSstPartitionerFactories(*(ObjectLibrary::Default().get()), "");
   });
-#endif  // ROCKSDB_LITE
-  return LoadSharedObject<SstPartitionerFactory>(options, value, nullptr,
-                                                 result);
+  return LoadSharedObject<SstPartitionerFactory>(options, value, result);
 }
 }  // namespace ROCKSDB_NAMESPACE

db/compaction/subcompaction_state.h

@@ -84,6 +84,11 @@ class SubcompactionState {
   // Assign range dels aggregator, for each range_del, it can only be assigned
   // to one output level, for per_key_placement, it's going to be the
   // penultimate level.
+  // TODO: This does not work for per_key_placement + user-defined timestamp +
+  //  DeleteRange() combo. If user-defined timestamp is enabled,
+  //  it is possible for a range tombstone to belong to bottommost level (
+  //  seqno < earliest snapshot) without being dropped (garbage collection
+  //  for user-defined timestamp).
   void AssignRangeDelAggregator(
       std::unique_ptr<CompactionRangeDelAggregator>&& range_del_agg) {
     if (compaction->SupportsPerKeyPlacement()) {
@@ -99,7 +104,6 @@ class SubcompactionState {
     penultimate_level_outputs_.RemoveLastEmptyOutput();
   }
 
-#ifndef ROCKSDB_LITE
   void BuildSubcompactionJobInfo(
       SubcompactionJobInfo& subcompaction_job_info) const {
     const Compaction* c = compaction;
@@ -113,7 +117,6 @@ class SubcompactionState {
     subcompaction_job_info.output_level = c->output_level();
     subcompaction_job_info.stats = compaction_job_stats;
   }
-#endif  // !ROCKSDB_LITE
 
   SubcompactionState() = delete;
   SubcompactionState(const SubcompactionState&) = delete;
@@ -196,8 +199,11 @@ class SubcompactionState {
                               const CompactionFileCloseFunc& close_file_func) {
     // Call FinishCompactionOutputFile() even if status is not ok: it needs to
     // close the output file.
+    // CloseOutput() may open new compaction output files.
+    is_current_penultimate_level_ = true;
     Status s = penultimate_level_outputs_.CloseOutput(
         curr_status, open_file_func, close_file_func);
+    is_current_penultimate_level_ = false;
     s = compaction_outputs_.CloseOutput(s, open_file_func, close_file_func);
     return s;
   }