Add 6.24.2

Summary:
Potential bugs in the IO uring implementation can cause bad data to be returned in the completion queue. Add some checks in the PosixRandomAccessFile::MultiRead completion handling code to catch such errors and fail the entire MultiRead. Also log some diagnostic messages and stack trace.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/8894

Reviewed By: siying, pdillinger

Differential Revision: D30826982

Pulled By: anand1976

fbshipit-source-id: af91815ac760e095d6cc0466cf8bd5c10167fd15

.circleci/config.yml

@@ -91,6 +91,13 @@ commands:
           command: |
             sudo apt-get update -y && sudo apt-get install -y libgflags-dev
 
+  install-benchmark:
+    steps:
+      - run: # currently doesn't support ubuntu-1604 which doesn't have libbenchmark package, user can still install by building it youself
+          name: Install benchmark
+          command: |
+            sudo apt-get update -y && sudo apt-get install -y libbenchmark-dev
+
   upgrade-cmake:
     steps:
       - run:
@@ -304,7 +311,7 @@ jobs:
       image: ubuntu-1604:202104-01
     resource_class: 2xlarge
     steps:
-      - checkout # check out the code in the project directory
+      - pre-steps
       - install-gflags
       - upgrade-cmake
       - run: (mkdir build && cd build && cmake -DWITH_GFLAGS=1 .. && make V=1 -j20 && ctest -j20) | .circleci/cat_ignore_eagain
@@ -315,9 +322,10 @@ jobs:
       image: ubuntu-2004:202104-01
     resource_class: 2xlarge
     steps:
-      - checkout # check out the code in the project directory
+      - pre-steps
       - install-gflags
-      - run: (mkdir build && cd build && cmake -DWITH_GFLAGS=1 .. && make V=1 -j20 && ctest -j20) | .circleci/cat_ignore_eagain
+      - install-benchmark
+      - run: (mkdir build && cd build && cmake -DWITH_GFLAGS=1 -DWITH_BENCHMARK=1 .. && make V=1 -j20 && ctest -j20 && make microbench) | .circleci/cat_ignore_eagain
       - post-steps
 
   build-linux-unity:
@@ -370,6 +378,17 @@ jobs:
       - run: CC=gcc-10 CXX=g++-10 V=1 SKIP_LINK=1 ROCKSDB_CXX_STANDARD=c++20 make -j16 all | .circleci/cat_ignore_eagain # Linking broken because libgflags compiled with newer ABI
       - 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.
+  build-linux-microbench:
+    machine:
+      image: ubuntu-2004:202010-01
+    resource_class: xlarge
+    steps:
+      - pre-steps
+      - install-benchmark
+      - run: DEBUG_LEVEL=0 make microbench | .circleci/cat_ignore_eagain
+      - post-steps
+
   build-windows:
     executor: windows-2xlarge
     parameters:
@@ -729,12 +748,6 @@ workflows:
           name: "build-windows-vs2017"
           vs_year: "2017"
           cmake_generator: "Visual Studio 15 Win64"
-#  build-windows-vs2015:
-#    jobs:
-#      - build-windows:
-#          name: "build-windows-vs2015"
-#          vs_year: "2015"
-#          cmake_generator: "Visual Studio 14 Win64"
   build-java:
     jobs:
       - build-linux-java
@@ -778,6 +791,9 @@ workflows:
   build-linux-arm:
     jobs:
       - build-linux-arm
+  build-microbench:
+    jobs:
+      - build-linux-microbench
   nightly:
     triggers:
       - schedule:

CMakeLists.txt

@@ -635,6 +635,7 @@ set(SOURCES
         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
@@ -815,9 +816,12 @@ set(SOURCES
         tools/ldb_tool.cc
         tools/sst_dump_tool.cc
         tools/trace_analyzer_tool.cc
-        trace_replay/trace_replay.cc
         trace_replay/block_cache_tracer.cc
         trace_replay/io_tracer.cc
+        trace_replay/trace_record_handler.cc
+        trace_replay/trace_record_result.cc
+        trace_replay/trace_record.cc
+        trace_replay/trace_replay.cc
         util/coding.cc
         util/compaction_job_stats_impl.cc
         util/comparator.cc
@@ -848,6 +852,7 @@ set(SOURCES
         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/debug.cc
         utilities/env_mirror.cc
@@ -856,6 +861,7 @@ set(SOURCES
         utilities/fault_injection_fs.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
@@ -875,6 +881,7 @@ set(SOURCES
         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_tracker.cc
         utilities/transactions/lock/point/point_lock_manager.cc
@@ -933,13 +940,8 @@ if(WIN32)
     port/win/env_win.cc
     port/win/env_default.cc
     port/win/port_win.cc
-    port/win/win_logger.cc)
-  if(NOT MINGW)
-    # Mingw only supports std::thread when using
-    # posix threads.
-    list(APPEND SOURCES
-      port/win/win_thread.cc)
-  endif()
+    port/win/win_logger.cc
+    port/win/win_thread.cc)
 if(WITH_XPRESS)
   list(APPEND SOURCES
     port/win/xpress_win.cc)
@@ -1123,11 +1125,13 @@ if(WITH_TESTS)
     list(APPEND TESTS
         cache/cache_test.cc
         cache/lru_cache_test.cc
+        db/blob/blob_counting_iterator_test.cc
         db/blob/blob_file_addition_test.cc
         db/blob/blob_file_builder_test.cc
         db/blob/blob_file_cache_test.cc
         db/blob/blob_file_garbage_test.cc
         db/blob/blob_file_reader_test.cc
+        db/blob/blob_garbage_meter_test.cc
         db/blob/db_blob_basic_test.cc
         db/blob/db_blob_compaction_test.cc
         db/blob/db_blob_corruption_test.cc
@@ -1424,3 +1428,8 @@ option(WITH_EXAMPLES "build with examples" OFF)
 if(WITH_EXAMPLES)
   add_subdirectory(examples)
 endif()
+
+option(WITH_BENCHMARK "build benchmark tests" OFF)
+if(WITH_BENCHMARK)
+  add_subdirectory(${PROJECT_SOURCE_DIR}/microbench/)
+endif()

HISTORY.md

@@ -1,17 +1,95 @@
 # Rocksdb Change Log
-## Unreleased
+## 6.24.2 (2021-09-16)
+### Bug Fixes
+* Add checks for validity of the IO uring completion queue entries, and fail the BlockBasedTableReader MultiGet sub-batch if there's an invalid completion
+
+## 6.24.1 (2021-08-31)
+### Bug Fixes
+* Fix a race in item ref counting in LRUCache when promoting an item from the SecondaryCache.
+
+## 6.24.0 (2021-08-20)
+### Bug Fixes
+* If the primary's CURRENT file is missing or inaccessible, the secondary instance should not hang repeatedly trying to switch to a new MANIFEST. It should instead return the error code encountered while accessing the file.
+* Restoring backups with BackupEngine is now a logically atomic operation, so that if a restore operation is interrupted, DB::Open on it will fail. Using BackupEngineOptions::sync (default) ensures atomicity even in case of power loss or OS crash.
+* Fixed a race related to the destruction of `ColumnFamilyData` objects. The earlier logic unlocked the DB mutex before destroying the thread-local `SuperVersion` pointers, which could result in a process crash if another thread managed to get a reference to the `ColumnFamilyData` object.
+* Removed a call to `RenameFile()` on a non-existent info log file ("LOG") when opening a new DB. Such a call was guaranteed to fail though did not impact applications since we swallowed the error. Now we also stopped swallowing errors in renaming "LOG" file.
+* Fixed an issue where `OnFlushCompleted` was not called for atomic flush.
+* Fixed a bug affecting the batched `MultiGet` API when used with keys spanning multiple column families and `sorted_input == false`.
+* Fixed a potential incorrect result in opt mode and assertion failures caused by releasing snapshot(s) during compaction.
+* Fixed passing of BlobFileCompletionCallback to Compaction job and Atomic flush job which was default paramter (nullptr). BlobFileCompletitionCallback is internal callback that manages addition of blob files to SSTFileManager.
+* Fixed MultiGet not updating the block_read_count and block_read_byte PerfContext counters
+
+### New Features
+* Made the EventListener extend the Customizable class.
+* EventListeners that have a non-empty Name() and that are registered with the ObjectRegistry can now be serialized to/from the OPTIONS file.
+* Insert warm blocks (data blocks, uncompressed dict blocks, index and filter blocks) in Block cache during flush under option BlockBasedTableOptions.prepopulate_block_cache. Previously it was enabled for only data blocks.
+* BlockBasedTableOptions.prepopulate_block_cache can be dynamically configured using DB::SetOptions.
+* Add CompactionOptionsFIFO.age_for_warm, which allows RocksDB to move old files to warm tier in FIFO compactions. Note that file temperature is still an experimental feature.
+* Add a comment to suggest btrfs user to disable file preallocation by setting `options.allow_fallocate=false`.
+* Fast forward option in Trace replay changed to double type to allow replaying at a lower speed, by settings the value between 0 and 1. This option can be set via `ReplayOptions` in `Replayer::Replay()`, or via `--trace_replay_fast_forward` in db_bench.
+* Add property `LiveSstFilesSizeAtTemperature` to retrieve sst file size at different temperature.
+* Added a stat rocksdb.secondary.cache.hits
+* Added a PerfContext counter secondary_cache_hit_count
+* The integrated BlobDB implementation now supports the tickers `BLOB_DB_BLOB_FILE_BYTES_READ`, `BLOB_DB_GC_NUM_KEYS_RELOCATED`, and `BLOB_DB_GC_BYTES_RELOCATED`, as well as the histograms `BLOB_DB_COMPRESSION_MICROS` and `BLOB_DB_DECOMPRESSION_MICROS`.
+* Added hybrid configuration of Ribbon filter and Bloom filter where some LSM levels use Ribbon for memory space efficiency and some use Bloom for speed. See NewRibbonFilterPolicy. This also changes the default behavior of NewRibbonFilterPolicy to use Bloom for flushes under Leveled and Universal compaction and Ribbon otherwise. The C API function `rocksdb_filterpolicy_create_ribbon` is unchanged but adds new `rocksdb_filterpolicy_create_ribbon_hybrid`.
+
+### Public API change
+* Added APIs to decode and replay trace file via Replayer class. Added `DB::NewDefaultReplayer()` to create a default Replayer instance. Added `TraceReader::Reset()` to restart reading a trace file. Created trace_record.h, trace_record_result.h and utilities/replayer.h files to access the decoded Trace records, replay them, and query the actual operation results.
+* Added Configurable::GetOptionsMap to the public API for use in creating new Customizable classes.
+* Generalized bits_per_key parameters in C API from int to double for greater configurability. Although this is a compatible change for existing C source code, anything depending on C API signatures, such as foreign function interfaces, will need to be updated.
+
+### Performance Improvements
+* Try to avoid updating DBOptions if `SetDBOptions()` does not change any option value.
+
+### Behavior Changes
+* `StringAppendOperator` additionally accepts a string as the delimiter.
+* BackupEngineOptions::sync (default true) now applies to restoring backups in addition to creating backups. This could slow down restores, but ensures they are fully persisted before returning OK. (Consider increasing max_background_operations to improve performance.)
+
+## 6.23.0 (2021-07-16)
+### Behavior Changes
+* Obsolete keys in the bottommost level that were preserved for a snapshot will now be cleaned upon snapshot release in all cases. This form of compaction (snapshot release triggered compaction) previously had an artificial limitation that multiple tombstones needed to be present.
+### Bug Fixes
+* Blob file checksums are now printed in hexadecimal format when using the `manifest_dump` `ldb` command.
+* `GetLiveFilesMetaData()` now populates the `temperature`, `oldest_ancester_time`, and `file_creation_time` fields of its `LiveFileMetaData` results when the information is available. Previously these fields always contained zero indicating unknown.
+* Fix mismatches of OnCompaction{Begin,Completed} in case of DisableManualCompaction().
+* Fix continuous logging of an existing background error on every user write
+* Fix a bug that `Get()` return Status::OK() and an empty value for non-existent key when `read_options.read_tier = kBlockCacheTier`.
+* Fix a bug that stat in `get_context` didn't accumulate to statistics when query is failed.
+* Fixed handling of DBOptions::wal_dir with LoadLatestOptions() or ldb --try_load_options on a copied or moved DB. Previously, when the WAL directory is same as DB directory (default), a copied or moved DB would reference the old path of the DB as the WAL directory, potentially corrupting both copies. Under this change, the wal_dir from DB::GetOptions() or LoadLatestOptions() may now be empty, indicating that the current DB directory is used for WALs. This is also a subtle API change.
+
+### New Features
+* ldb has a new feature, `list_live_files_metadata`, that shows the live SST files, as well as their LSM storage level and the column family they belong to.
+* The new BlobDB implementation now tracks the amount of garbage in each blob file in the MANIFEST.
+* Integrated BlobDB now supports Merge with base values (Put/Delete etc.).
+* RemoteCompaction supports sub-compaction, the job_id in the user interface is changed from `int` to `uint64_t` to support sub-compaction id.
+* Expose statistics option in RemoteCompaction worker.
+
+### Public API change
+* Added APIs to the Customizable class to allow developers to create their own Customizable classes.  Created the utilities/customizable_util.h file to contain helper methods for developing new Customizable classes.
+* Change signature of SecondaryCache::Name().  Make SecondaryCache customizable and add SecondaryCache::CreateFromString method.
+
+## 6.22.0 (2021-06-18)
 ### Behavior Changes
 * Added two additional tickers, MEMTABLE_PAYLOAD_BYTES_AT_FLUSH and MEMTABLE_GARBAGE_BYTES_AT_FLUSH. These stats can be used to estimate the ratio of "garbage" (outdated) bytes in the memtable that are discarded at flush time.
 * Added API comments clarifying safe usage of Disable/EnableManualCompaction and EventListener callbacks for compaction.
-
 ### Bug Fixes
 * fs_posix.cc GetFreeSpace() always report disk space available to root even when running as non-root.  Linux defaults often have disk mounts with 5 to 10 percent of total space reserved only for root.  Out of space could result for non-root users.
 * Subcompactions are now disabled when user-defined timestamps are used, since the subcompaction boundary picking logic is currently not timestamp-aware, which could lead to incorrect results when different subcompactions process keys that only differ by timestamp.
+* Fix an issue that `DeleteFilesInRange()` may cause ongoing compaction reports corruption exception, or ASSERT for debug build. There's no actual data loss or corruption that we find.
+* Fixed confusingly duplicated output in LOG for periodic stats ("DUMPING STATS"), including "Compaction Stats" and "File Read Latency Histogram By Level".
+* Fixed performance bugs in background gathering of block cache entry statistics, that could consume a lot of CPU when there are many column families with a shared block cache.
 
 ### New Features
 * Marked the Ribbon filter and optimize_filters_for_memory features as production-ready, each enabling memory savings for Bloom-like filters. Use `NewRibbonFilterPolicy` in place of `NewBloomFilterPolicy` to use Ribbon filters instead of Bloom, or `ribbonfilter` in place of `bloomfilter` in configuration string.
 * Allow `DBWithTTL` to use `DeleteRange` api just like other DBs. `DeleteRangeCF()` which executes `WriteBatchInternal::DeleteRange()` has been added to the handler in `DBWithTTLImpl::Write()` to implement it.
 * Add BlockBasedTableOptions.prepopulate_block_cache.  If enabled, it prepopulate warm/hot data blocks which are already in memory into block cache at the time of flush. On a flush, the data block that is in memory (in memtables) get flushed to the device. If using Direct IO, additional IO is incurred to read this data back into memory again, which is avoided by enabling this option and it also helps with Distributed FileSystem. More details in include/rocksdb/table.h.
+* Added a `cancel` field to `CompactRangeOptions`, allowing individual in-process manual range compactions to be cancelled.
+
+### New Features
+* Added BlobMetaData to the ColumnFamilyMetaData to return information about blob files
+
+### Public API change
+* Added GetAllColumnFamilyMetaData API to retrieve the ColumnFamilyMetaData about all column families.
 
 ## 6.21.0 (2021-05-21)
 ### Bug Fixes
@@ -23,7 +101,6 @@
 * Handle return code by io_uring_submit_and_wait() and io_uring_wait_cqe().
 * In the IngestExternalFile() API, only try to sync the ingested file if the file is linked and the FileSystem/Env supports reopening a writable file.
 * Fixed a bug that `AdvancedColumnFamilyOptions.max_compaction_bytes` is under-calculated for manual compaction (`CompactRange()`). Manual compaction is split to multiple compactions if the compaction size exceed the `max_compaction_bytes`. The bug creates much larger compaction which size exceed the user setting. On the other hand, larger manual compaction size can increase the subcompaction parallelism, you can tune that by setting `max_compaction_bytes`.
-* Fixed confusingly duplicated output in LOG for periodic stats ("DUMPING STATS"), including "Compaction Stats" and "File Read Latency Histogram By Level".
 
 ### Behavior Changes
 * Due to the fix of false-postive alert of "SST file is ahead of WAL", all the CFs with no SST file (CF empty) will bypass the consistency check. We fixed a false-positive, but introduced a very rare true-negative which will be triggered in the following conditions: A CF with some delete operations in the last a few queries which will result in an empty CF (those are flushed to SST file and a compaction triggered which combines this file and all other SST files and generates an empty CF, or there is another reason to write a manifest entry for this CF after a flush that generates no SST file from an empty CF). The deletion entries are logged in a WAL and this WAL was corrupted, while the CF's log number points to the next WAL (due to the flush). Therefore, the DB can only recover to the point without these trailing deletions and cause the inconsistent DB status.
@@ -36,7 +113,6 @@
 * Add an experimental Remote Compaction feature, which allows the user to run Compaction on a different host or process. The feature is still under development, currently only works on some basic use cases. The interface will be changed without backward/forward compatibility support.
 * RocksDB would validate total entries read in flush, and compare with counter inserted into it. If flush_verify_memtable_count = true (default), flush will fail. Otherwise, only log to info logs.
 * Add `TableProperties::num_filter_entries`, which can be used with `TableProperties::filter_size` to calculate the effective bits per filter entry (unique user key or prefix) for a table file.
-* Added a `cancel` field to `CompactRangeOptions`, allowing individual in-process manual range compactions to be cancelled.
 
 ### Performance Improvements
 * BlockPrefetcher is used by iterators to prefetch data if they anticipate more data to be used in future. It is enabled implicitly by rocksdb. Added change to take in account read pattern if reads are sequential. This would disable prefetching for random reads in MultiGet and iterators as readahead_size is increased exponential doing large prefetches.

Makefile

@@ -344,6 +344,12 @@ ifdef ROCKSDB_VALGRIND_RUN
 	PLATFORM_CCFLAGS += -DROCKSDB_VALGRIND_RUN
 	PLATFORM_CXXFLAGS += -DROCKSDB_VALGRIND_RUN
 endif
+ifdef ROCKSDB_FULL_VALGRIND_RUN
+	# Some tests are slow when run under valgrind and are only run when
+	# explicitly requested via the ROCKSDB_FULL_VALGRIND_RUN compiler flag.
+	PLATFORM_CCFLAGS += -DROCKSDB_VALGRIND_RUN -DROCKSDB_FULL_VALGRIND_RUN
+	PLATFORM_CXXFLAGS += -DROCKSDB_VALGRIND_RUN -DROCKSDB_FULL_VALGRIND_RUN
+endif
 
 ifndef DISABLE_JEMALLOC
 	ifdef JEMALLOC
@@ -488,6 +494,7 @@ VALGRIND_ERROR = 2
 VALGRIND_VER := $(join $(VALGRIND_VER),valgrind)
 
 VALGRIND_OPTS = --error-exitcode=$(VALGRIND_ERROR) --leak-check=full
+# Not yet supported: --show-leak-kinds=definite,possible,reachable --errors-for-leak-kinds=definite,possible,reachable
 
 TEST_OBJECTS = $(patsubst %.cc, $(OBJ_DIR)/%.o, $(TEST_LIB_SOURCES) $(MOCK_LIB_SOURCES)) $(GTEST)
 BENCH_OBJECTS = $(patsubst %.cc, $(OBJ_DIR)/%.o, $(BENCH_LIB_SOURCES))
@@ -516,8 +523,9 @@ endif
 ifdef ASSERT_STATUS_CHECKED
 	# TODO: finish fixing all tests to pass this check
 	TESTS_FAILING_ASC = \
+		c_test \
 		db_test \
-		db_test2 \
+		env_test \
 		range_locking_test \
 		testutil_test \
 
@@ -593,6 +601,8 @@ TEST_LIBS = \
 # TODO: add back forward_iterator_bench, after making it build in all environemnts.
 BENCHMARKS = $(patsubst %.cc, %, $(notdir $(BENCH_MAIN_SOURCES)))
 
+MICROBENCHS = $(patsubst %.cc, %, $(notdir $(MICROBENCH_SOURCES)))
+
 # if user didn't config LIBNAME, set the default
 ifeq ($(LIBNAME),)
   LIBNAME=librocksdb
@@ -731,6 +741,9 @@ test_libs: $(TEST_LIBS)
 
 benchmarks: $(BENCHMARKS)
 
+microbench: $(MICROBENCHS)
+	for t in $(MICROBENCHS); do echo "===== Running benchmark $$t (`date`)"; ./$$t || exit 1; done;
+
 dbg: $(LIBRARY) $(BENCHMARKS) tools $(TESTS)
 
 # creates library and programs
@@ -1044,6 +1057,12 @@ ubsan_crash_test_with_best_efforts_recovery: clean
 	COMPILE_WITH_UBSAN=1 $(MAKE) crash_test_with_best_efforts_recovery
 	$(MAKE) clean
 
+full_valgrind_test:
+	ROCKSDB_FULL_VALGRIND_RUN=1 DISABLE_JEMALLOC=1 $(MAKE) valgrind_check
+
+full_valgrind_test_some:
+	ROCKSDB_FULL_VALGRIND_RUN=1 DISABLE_JEMALLOC=1 $(MAKE) valgrind_check_some
+
 valgrind_test:
 	ROCKSDB_VALGRIND_RUN=1 DISABLE_JEMALLOC=1 $(MAKE) valgrind_check
 
@@ -1162,7 +1181,7 @@ clean-not-downloaded: clean-ext-libraries-bin clean-rocks clean-not-downloaded-r
 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)
+	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(ALL_STATIC_LIBS) $(ALL_SHARED_LIBS) $(MICROBENCHS)
 	rm -rf $(CLEAN_FILES) ios-x86 ios-arm scan_build_report
 	$(FIND) . -name "*.[oda]" -exec rm -f {} \;
 	$(FIND) . -type f -regex ".*\.\(\(gcda\)\|\(gcno\)\)" -exec rm -f {} \;
@@ -1825,6 +1844,9 @@ block_cache_trace_analyzer_test: $(OBJ_DIR)/tools/block_cache_analyzer/block_cac
 defer_test: $(OBJ_DIR)/util/defer_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
+blob_counting_iterator_test: $(OBJ_DIR)/db/blob/blob_counting_iterator_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 blob_file_addition_test: $(OBJ_DIR)/db/blob/blob_file_addition_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
@@ -1840,6 +1862,9 @@ blob_file_garbage_test: $(OBJ_DIR)/db/blob/blob_file_garbage_test.o $(TEST_LIBRA
 blob_file_reader_test: $(OBJ_DIR)/db/blob/blob_file_reader_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
+blob_garbage_meter_test: $(OBJ_DIR)/db/blob/blob_garbage_meter_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 timer_test: $(OBJ_DIR)/util/timer_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
@@ -1870,6 +1895,9 @@ db_write_buffer_manager_test: $(OBJ_DIR)/db/db_write_buffer_manager_test.o $(TES
 clipping_iterator_test: $(OBJ_DIR)/db/compaction/clipping_iterator_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
+ribbon_bench: $(OBJ_DIR)/microbench/ribbon_bench.o $(LIBRARY)
+	$(AM_LINK)
+
 #-------------------------------------------------
 # make install related stuff
 PREFIX ?= /usr/local

TARGETS

@@ -144,6 +144,7 @@ cpp_library(
         "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",
@@ -334,6 +335,9 @@ cpp_library(
         "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/build_version.cc",
         "util/coding.cc",
@@ -367,6 +371,7 @@ cpp_library(
         "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/debug.cc",
@@ -376,6 +381,7 @@ cpp_library(
         "utilities/fault_injection_fs.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",
@@ -395,6 +401,7 @@ cpp_library(
         "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",
@@ -456,6 +463,7 @@ cpp_library(
         "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",
@@ -646,6 +654,9 @@ cpp_library(
         "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/build_version.cc",
         "util/coding.cc",
@@ -679,6 +690,7 @@ cpp_library(
         "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/debug.cc",
@@ -688,6 +700,7 @@ cpp_library(
         "utilities/fault_injection_fs.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",
@@ -707,6 +720,7 @@ cpp_library(
         "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",
@@ -906,6 +920,13 @@ ROCKS_TESTS = [
         [],
         [],
     ],
+    [
+        "blob_counting_iterator_test",
+        "db/blob/blob_counting_iterator_test.cc",
+        "parallel",
+        [],
+        [],
+    ],
     [
         "blob_db_test",
         "utilities/blob_db/blob_db_test.cc",
@@ -948,6 +969,13 @@ ROCKS_TESTS = [
         [],
         [],
     ],
+    [
+        "blob_garbage_meter_test",
+        "db/blob/blob_garbage_meter_test.cc",
+        "parallel",
+        [],
+        [],
+    ],
     [
         "block_based_filter_block_test",
         "table/block_based/block_based_filter_block_test.cc",

build_tools/build_detect_platform

@@ -189,9 +189,6 @@ EOF
                 COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_IOURING_PRESENT"
             fi
         fi
-        if test -z "$USE_FOLLY_DISTRIBUTED_MUTEX"; then
-          USE_FOLLY_DISTRIBUTED_MUTEX=1
-        fi
         # PORT_FILES=port/linux/linux_specific.cc
         ;;
     SunOS)
@@ -596,6 +593,16 @@ EOF
             PLATFORM_CXXFLAGS="$PLATFORM_CXXFLAGS -faligned-new -DHAVE_ALIGNED_NEW"
         fi
     fi
+    if ! test $ROCKSDB_DISABLE_BENCHMARK; then
+        # Test whether google benchmark is available
+        $CXX $PLATFORM_CXXFLAGS -x c++ - -o /dev/null -lbenchmark 2>/dev/null  <<EOF
+          #include <benchmark/benchmark.h>
+          int main() {}
+EOF
+        if [ "$?" = 0 ]; then
+            PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lbenchmark"
+        fi
+    fi
 fi
 
 # TODO(tec): Fix -Wshorten-64-to-32 errors on FreeBSD and enable the warning.

build_tools/fbcode_config_platform007.sh

@@ -167,6 +167,4 @@ VALGRIND_VER="$VALGRIND_BASE/bin/"
 LUA_PATH=
 LUA_LIB=
 
-USE_FOLLY_DISTRIBUTED_MUTEX=1
-
 export CC CXX AR 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_platform009.sh

@@ -168,6 +168,4 @@ VALGRIND_VER="$VALGRIND_BASE/bin/"
 LUA_PATH=
 LUA_LIB=
 
-USE_FOLLY_DISTRIBUTED_MUTEX=1
-
 export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD LUA_PATH LUA_LIB

build_tools/rocksdb-lego-determinator

@@ -1195,6 +1195,40 @@ JAVA_BUILD_TEST_COMMANDS="[
     }
 ]"
 
+#
+# RocksDB fbcode stress/crash test
+#
+FBCODE_STRESS_CRASH_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb Fbcode Stress and Crash Test',
+        'oncall':'$ONCALL',
+        'executeLocal': 'true',
+        'timeout': 86400,
+        'steps': [
+            {
+                'name':'Copy RocksDB code to fbcode repo',
+                'shell':'cd internal_repo_rocksdb/repo && git init && git add * && git commit -a -m \"Make internal_repo_rocksdb/repo a git repo\" && cd ../.. && echo Y | python3 rocks/release_script/release_to_fbcode.py -u internal_repo_rocksdb/repo master || $CONTRUN_NAME=db_stress_fbcode $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+            {
+                'name':'Build RocksDB fbcode stress tests',
+                'shell':'cd $WORKING_DIR; buck build @mode/dbg rocks/tools:rocks_db_stress || $CONTRUN_NAME=db_stress_fbcode $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+            {
+                'name':'Run RocksDB whitebox crash tests',
+                'timeout': 86400,
+                'shell':'cd $WORKING_DIR; mkdir /dev/shm/rocksdb_whitebox_crash_test && TEST_TMPDIR=\$(mktemp -d --tmpdir=/dev/shm/rocksdb_whitebox_crash_test) python3 rocksdb/src/tools/db_crashtest.py --stress_cmd=buck-out/dbg/gen/rocks/tools/rocks_db_stress -secondary_cache_uri=\"$SECONDARY_CACHE_URI\" --env_uri=$ENV_URI -checkpoint_one_in=0 -backup_one_in=0 -cache_size=134217728 -cache_numshardbits=4 $EXTRA_DB_STRESS_ARGS -logtostderr=false whitebox || $CONTRUN_NAME=whitebox_crash_test $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
 
 case $1 in
   unit)
@@ -1332,6 +1366,11 @@ case $1 in
   java_build)
     echo $JAVA_BUILD_TEST_COMMANDS
     ;;
+  fbcode_stress_crash)
+    set -f
+    echo $FBCODE_STRESS_CRASH_TEST_COMMANDS
+    set +f
+    ;;
   *)
     echo "Invalid determinator command"
     exit 1

cache/cache.cc

@@ -10,6 +10,8 @@
 #include "rocksdb/cache.h"
 
 #include "cache/lru_cache.h"
+#include "rocksdb/secondary_cache.h"
+#include "rocksdb/utilities/customizable_util.h"
 #include "rocksdb/utilities/options_type.h"
 #include "util/string_util.h"
 
@@ -34,6 +36,13 @@ static std::unordered_map<std::string, OptionTypeInfo>
 };
 #endif  // ROCKSDB_LITE
 
+Status SecondaryCache::CreateFromString(
+    const ConfigOptions& config_options, const std::string& value,
+    std::shared_ptr<SecondaryCache>* result) {
+  return LoadSharedObject<SecondaryCache>(config_options, value, nullptr,
+                                          result);
+}
+
 Status Cache::CreateFromString(const ConfigOptions& config_options,
                                const std::string& value,
                                std::shared_ptr<Cache>* result) {

cache/cache_bench_tool.cc

@@ -13,11 +13,11 @@
 #include "monitoring/histogram.h"
 #include "port/port.h"
 #include "rocksdb/cache.h"
+#include "rocksdb/convenience.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/secondary_cache.h"
 #include "rocksdb/system_clock.h"
-#include "rocksdb/utilities/object_registry.h"
 #include "table/block_based/cachable_entry.h"
 #include "util/coding.h"
 #include "util/gflags_compat.h"
@@ -233,9 +233,8 @@ class CacheBench {
       LRUCacheOptions opts(FLAGS_cache_size, FLAGS_num_shard_bits, false, 0.5);
 #ifndef ROCKSDB_LITE
       if (!FLAGS_secondary_cache_uri.empty()) {
-        Status s =
-            ObjectRegistry::NewInstance()->NewSharedObject<SecondaryCache>(
-                FLAGS_secondary_cache_uri, &secondary_cache);
+        Status s = SecondaryCache::CreateFromString(
+            ConfigOptions(), FLAGS_secondary_cache_uri, &secondary_cache);
         if (secondary_cache == nullptr) {
           fprintf(
               stderr,

cache/cache_entry_stats.h

@@ -51,7 +51,8 @@ namespace ROCKSDB_NAMESPACE {
 template <class Stats>
 class CacheEntryStatsCollector {
  public:
-  // Gathers stats and saves results into `stats`
+  // Gather and save stats if saved stats are too old. (Use GetStats() to
+  // read saved stats.)
   //
   // Maximum allowed age for a "hit" on saved results is determined by the
   // two interval parameters. Both set to 0 forces a re-scan. For example
@@ -61,10 +62,9 @@ class CacheEntryStatsCollector {
   // Justification: scans can vary wildly in duration, e.g. from 0.02 sec
   // to as much as 20 seconds, so we want to be able to cap the absolute
   // and relative frequency of scans.
-  void GetStats(Stats *stats, int min_interval_seconds,
-                int min_interval_factor) {
+  void CollectStats(int min_interval_seconds, int min_interval_factor) {
     // Waits for any pending reader or writer (collector)
-    std::lock_guard<std::mutex> lock(mutex_);
+    std::lock_guard<std::mutex> lock(working_mutex_);
 
     uint64_t max_age_micros =
         static_cast<uint64_t>(std::max(min_interval_seconds, 0)) * 1000000U;
@@ -79,19 +79,28 @@ class CacheEntryStatsCollector {
     uint64_t start_time_micros = clock_->NowMicros();
     if ((start_time_micros - last_end_time_micros_) > max_age_micros) {
       last_start_time_micros_ = start_time_micros;
-      saved_stats_.BeginCollection(cache_, clock_, start_time_micros);
+      working_stats_.BeginCollection(cache_, clock_, start_time_micros);
 
-      cache_->ApplyToAllEntries(saved_stats_.GetEntryCallback(), {});
+      cache_->ApplyToAllEntries(working_stats_.GetEntryCallback(), {});
       TEST_SYNC_POINT_CALLBACK(
           "CacheEntryStatsCollector::GetStats:AfterApplyToAllEntries", nullptr);
 
       uint64_t end_time_micros = clock_->NowMicros();
       last_end_time_micros_ = end_time_micros;
-      saved_stats_.EndCollection(cache_, clock_, end_time_micros);
+      working_stats_.EndCollection(cache_, clock_, end_time_micros);
     } else {
-      saved_stats_.SkippedCollection();
+      working_stats_.SkippedCollection();
     }
-    // Copy to caller
+
+    // Save so that we don't need to wait for an outstanding collection in
+    // order to make of copy of the last saved stats
+    std::lock_guard<std::mutex> lock2(saved_mutex_);
+    saved_stats_ = working_stats_;
+  }
+
+  // Gets saved stats, regardless of age
+  void GetStats(Stats *stats) {
+    std::lock_guard<std::mutex> lock(saved_mutex_);
     *stats = saved_stats_;
   }
 
@@ -129,6 +138,7 @@ class CacheEntryStatsCollector {
                                  Cache::Priority::HIGH);
         if (!s.ok()) {
           assert(h == nullptr);
+          delete new_ptr;
           return s;
         }
       }
@@ -145,6 +155,7 @@ class CacheEntryStatsCollector {
  private:
   explicit CacheEntryStatsCollector(Cache *cache, SystemClock *clock)
       : saved_stats_(),
+        working_stats_(),
         last_start_time_micros_(0),
         last_end_time_micros_(/*pessimistic*/ 10000000),
         cache_(cache),
@@ -154,10 +165,14 @@ class CacheEntryStatsCollector {
     delete static_cast<CacheEntryStatsCollector *>(value);
   }
 
-  std::mutex mutex_;
+  std::mutex saved_mutex_;
   Stats saved_stats_;
+
+  std::mutex working_mutex_;
+  Stats working_stats_;
   uint64_t last_start_time_micros_;
   uint64_t last_end_time_micros_;
+
   Cache *const cache_;
   SystemClock *const clock_;
 };

cache/clock_cache.cc

@@ -280,7 +280,8 @@ class ClockCacheShard final : public CacheShard {
   Cache::Handle* Lookup(const Slice& key, uint32_t hash,
                         const Cache::CacheItemHelper* /*helper*/,
                         const Cache::CreateCallback& /*create_cb*/,
-                        Cache::Priority /*priority*/, bool /*wait*/) override {
+                        Cache::Priority /*priority*/, bool /*wait*/,
+                        Statistics* /*stats*/) override {
     return Lookup(key, hash);
   }
   bool Release(Cache::Handle* handle, bool /*useful*/,

cache/lru_cache.cc

@@ -13,6 +13,8 @@
 #include <cstdint>
 #include <cstdio>
 
+#include "monitoring/perf_context_imp.h"
+#include "monitoring/statistics.h"
 #include "util/mutexlock.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -358,7 +360,10 @@ Status LRUCacheShard::InsertItem(LRUHandle* e, Cache::Handle** handle,
       if (handle == nullptr) {
         LRU_Insert(e);
       } else {
-        e->Ref();
+        // If caller already holds a ref, no need to take one here
+        if (!e->HasRefs()) {
+          e->Ref();
+        }
         *handle = reinterpret_cast<Cache::Handle*>(e);
       }
     }
@@ -396,22 +401,17 @@ void LRUCacheShard::Promote(LRUHandle* e) {
   if (e->value) {
     Cache::Handle* handle = reinterpret_cast<Cache::Handle*>(e);
     Status s = InsertItem(e, &handle, /*free_handle_on_fail=*/false);
-    if (s.ok()) {
-      // InsertItem would have taken a reference on the item, so decrement it
-      // here as we expect the caller to already hold a reference
-      e->Unref();
-    } else {
+    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 {
     // Since the secondary cache lookup failed, mark the item as not in cache
-    // and charge the cache only for metadata usage, i.e handle, key etc
+    // Don't charge the cache as its only metadata that'll shortly be released
     MutexLock l(&mutex_);
     e->charge = 0;
     e->SetInCache(false);
-    usage_ += e->CalcTotalCharge(metadata_charge_policy_);
   }
 }
 
@@ -419,7 +419,7 @@ Cache::Handle* LRUCacheShard::Lookup(
     const Slice& key, uint32_t hash,
     const ShardedCache::CacheItemHelper* helper,
     const ShardedCache::CreateCallback& create_cb, Cache::Priority priority,
-    bool wait) {
+    bool wait, Statistics* stats) {
   LRUHandle* e = nullptr;
   {
     MutexLock l(&mutex_);
@@ -472,11 +472,18 @@ Cache::Handle* LRUCacheShard::Lookup(
           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->SetIncomplete(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);
       }
     }
   }
@@ -522,7 +529,12 @@ bool LRUCacheShard::Release(Cache::Handle* handle, bool force_erase) {
         last_reference = false;
       }
     }
-    if (last_reference) {
+    // If it was the last reference, and the entry is either not secondary
+    // cache compatible (i.e a dummy entry for accounting), or is secondary
+    // cache compatible and has a non-null value, then decrement the cache
+    // usage. If value is null in the latter case, taht means the lookup
+    // failed and we didn't charge the cache.
+    if (last_reference && (!e->IsSecondaryCacheCompatible() || e->value)) {
       size_t total_charge = e->CalcTotalCharge(metadata_charge_policy_);
       assert(usage_ >= total_charge);
       usage_ -= total_charge;
@@ -553,6 +565,9 @@ Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
     e->SetSecondaryCacheCompatible(true);
     e->info_.helper = helper;
   } else {
+#ifdef __SANITIZE_THREAD__
+    e->is_secondary_cache_compatible_for_tsan = false;
+#endif  // __SANITIZE_THREAD__
     e->info_.deleter = deleter;
   }
   e->charge = charge;
@@ -596,15 +611,12 @@ void LRUCacheShard::Erase(const Slice& key, uint32_t hash) {
 
 bool LRUCacheShard::IsReady(Cache::Handle* handle) {
   LRUHandle* e = reinterpret_cast<LRUHandle*>(handle);
+  MutexLock l(&mutex_);
   bool ready = true;
   if (e->IsPending()) {
     assert(secondary_cache_);
     assert(e->sec_handle);
-    if (e->sec_handle->IsReady()) {
-      Promote(e);
-    } else {
-      ready = false;
-    }
+    ready = e->sec_handle->IsReady();
   }
   return ready;
 }

cache/lru_cache.h

@@ -12,6 +12,7 @@
 #include <string>
 
 #include "cache/sharded_cache.h"
+#include "port/lang.h"
 #include "port/malloc.h"
 #include "port/port.h"
 #include "rocksdb/secondary_cache.h"
@@ -80,8 +81,8 @@ struct LRUHandle {
     IN_HIGH_PRI_POOL = (1 << 2),
     // Whether this entry has had any lookups (hits).
     HAS_HIT = (1 << 3),
-    // Can this be inserted into the tiered cache
-    IS_TIERED_CACHE_COMPATIBLE = (1 << 4),
+    // Can this be inserted into the secondary cache
+    IS_SECONDARY_CACHE_COMPATIBLE = (1 << 4),
     // Is the handle still being read from a lower tier
     IS_PENDING = (1 << 5),
     // Has the item been promoted from a lower tier
@@ -90,6 +91,14 @@ struct LRUHandle {
 
   uint8_t flags;
 
+#ifdef __SANITIZE_THREAD__
+  // TSAN can report a false data race on flags, where one thread is writing
+  // to one of the mutable bits and another thread is reading this immutable
+  // bit. So precisely suppress that TSAN warning, we separate out this bit
+  // during TSAN runs.
+  bool is_secondary_cache_compatible_for_tsan;
+#endif  // __SANITIZE_THREAD__
+
   // Beginning of the key (MUST BE THE LAST FIELD IN THIS STRUCT!)
   char key_data[1];
 
@@ -113,7 +122,11 @@ struct LRUHandle {
   bool InHighPriPool() const { return flags & IN_HIGH_PRI_POOL; }
   bool HasHit() const { return flags & HAS_HIT; }
   bool IsSecondaryCacheCompatible() const {
-    return flags & IS_TIERED_CACHE_COMPATIBLE;
+#ifdef __SANITIZE_THREAD__
+    return is_secondary_cache_compatible_for_tsan;
+#else
+    return flags & IS_SECONDARY_CACHE_COMPATIBLE;
+#endif  // __SANITIZE_THREAD__
   }
   bool IsPending() const { return flags & IS_PENDING; }
   bool IsPromoted() const { return flags & IS_PROMOTED; }
@@ -144,12 +157,15 @@ struct LRUHandle {
 
   void SetHit() { flags |= HAS_HIT; }
 
-  void SetSecondaryCacheCompatible(bool tiered) {
-    if (tiered) {
-      flags |= IS_TIERED_CACHE_COMPATIBLE;
+  void SetSecondaryCacheCompatible(bool compat) {
+    if (compat) {
+      flags |= IS_SECONDARY_CACHE_COMPATIBLE;
     } else {
-      flags &= ~IS_TIERED_CACHE_COMPATIBLE;
+      flags &= ~IS_SECONDARY_CACHE_COMPATIBLE;
     }
+#ifdef __SANITIZE_THREAD__
+    is_secondary_cache_compatible_for_tsan = compat;
+#endif  // __SANITIZE_THREAD__
   }
 
   void SetIncomplete(bool incomp) {
@@ -170,6 +186,11 @@ struct LRUHandle {
 
   void Free() {
     assert(refs == 0);
+#ifdef __SANITIZE_THREAD__
+    // Here we can safely assert they are the same without a data race reported
+    assert(((flags & IS_SECONDARY_CACHE_COMPATIBLE) != 0) ==
+           is_secondary_cache_compatible_for_tsan);
+#endif  // __SANITIZE_THREAD__
     if (!IsSecondaryCacheCompatible() && info_.deleter) {
       (*info_.deleter)(key(), value);
     } else if (IsSecondaryCacheCompatible()) {
@@ -298,10 +319,11 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShard {
   virtual Cache::Handle* Lookup(const Slice& key, uint32_t hash,
                                 const ShardedCache::CacheItemHelper* helper,
                                 const ShardedCache::CreateCallback& create_cb,
-                                ShardedCache::Priority priority,
-                                bool wait) override;
+                                ShardedCache::Priority priority, bool wait,
+                                Statistics* stats) override;
   virtual Cache::Handle* Lookup(const Slice& key, uint32_t hash) override {
-    return Lookup(key, hash, nullptr, nullptr, Cache::Priority::LOW, true);
+    return Lookup(key, hash, nullptr, nullptr, Cache::Priority::LOW, true,
+                  nullptr);
   }
   virtual bool Release(Cache::Handle* handle, bool /*useful*/,
                        bool force_erase) override {

cache/lru_cache_test.cc

@@ -224,7 +224,7 @@ class TestSecondaryCache : public SecondaryCache {
   }
   ~TestSecondaryCache() override { cache_.reset(); }
 
-  std::string Name() override { return "TestSecondaryCache"; }
+  const char* Name() const override { return "TestSecondaryCache"; }
 
   void InjectFailure() { inject_failure_ = true; }
 
@@ -378,7 +378,7 @@ class TestSecondaryCache : public SecondaryCache {
 class DBSecondaryCacheTest : public DBTestBase {
  public:
   DBSecondaryCacheTest()
-      : DBTestBase("/db_secondary_cache_test", /*env_do_fsync=*/true) {
+      : DBTestBase("db_secondary_cache_test", /*env_do_fsync=*/true) {
     fault_fs_.reset(new FaultInjectionTestFS(env_->GetFileSystem()));
     fault_env_.reset(new CompositeEnvWrapper(env_, fault_fs_));
   }
@@ -468,6 +468,7 @@ TEST_F(LRUSecondaryCacheTest, BasicTest) {
       std::make_shared<TestSecondaryCache>(2048);
   opts.secondary_cache = secondary_cache;
   std::shared_ptr<Cache> cache = NewLRUCache(opts);
+  std::shared_ptr<Statistics> stats = CreateDBStatistics();
 
   Random rnd(301);
   std::string str1 = rnd.RandomString(1020);
@@ -476,22 +477,29 @@ TEST_F(LRUSecondaryCacheTest, BasicTest) {
                           str1.length()));
   std::string str2 = rnd.RandomString(1020);
   TestItem* item2 = new TestItem(str2.data(), str2.length());
-  // k2 should be demoted to NVM
+  // k1 should be demoted to NVM
   ASSERT_OK(cache->Insert("k2", item2, &LRUSecondaryCacheTest::helper_,
                           str2.length()));
 
+  get_perf_context()->Reset();
   Cache::Handle* handle;
-  handle = cache->Lookup("k2", &LRUSecondaryCacheTest::helper_,
-                         test_item_creator, Cache::Priority::LOW, true);
+  handle =
+      cache->Lookup("k2", &LRUSecondaryCacheTest::helper_, test_item_creator,
+                    Cache::Priority::LOW, true, stats.get());
   ASSERT_NE(handle, nullptr);
   cache->Release(handle);
   // This lookup should promote k1 and demote k2
-  handle = cache->Lookup("k1", &LRUSecondaryCacheTest::helper_,
-                         test_item_creator, Cache::Priority::LOW, true);
+  handle =
+      cache->Lookup("k1", &LRUSecondaryCacheTest::helper_, test_item_creator,
+                    Cache::Priority::LOW, true, stats.get());
   ASSERT_NE(handle, nullptr);
   cache->Release(handle);
   ASSERT_EQ(secondary_cache->num_inserts(), 2u);
   ASSERT_EQ(secondary_cache->num_lookups(), 1u);
+  ASSERT_EQ(stats->getTickerCount(SECONDARY_CACHE_HITS),
+            secondary_cache->num_lookups());
+  PerfContext perf_ctx = *get_perf_context();
+  ASSERT_EQ(perf_ctx.secondary_cache_hit_count, secondary_cache->num_lookups());
 
   cache.reset();
   secondary_cache.reset();

cache/sharded_cache.cc

@@ -83,10 +83,10 @@ Cache::Handle* ShardedCache::Lookup(const Slice& key,
                                     const CacheItemHelper* helper,
                                     const CreateCallback& create_cb,
                                     Priority priority, bool wait,
-                                    Statistics* /*stats*/) {
+                                    Statistics* stats) {
   uint32_t hash = HashSlice(key);
   return GetShard(Shard(hash))
-      ->Lookup(key, hash, helper, create_cb, priority, wait);
+      ->Lookup(key, hash, helper, create_cb, priority, wait, stats);
 }
 
 bool ShardedCache::IsReady(Handle* handle) {

cache/sharded_cache.h

@@ -34,7 +34,8 @@ class CacheShard {
   virtual Cache::Handle* Lookup(const Slice& key, uint32_t hash,
                                 const Cache::CacheItemHelper* helper,
                                 const Cache::CreateCallback& create_cb,
-                                Cache::Priority priority, bool wait) = 0;
+                                Cache::Priority priority, bool wait,
+                                Statistics* stats) = 0;
   virtual bool Release(Cache::Handle* handle, bool useful,
                        bool force_erase) = 0;
   virtual bool IsReady(Cache::Handle* handle) = 0;

db/blob/blob_counting_iterator.h

@@ -0,0 +1,146 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  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 <cassert>
+
+#include "db/blob/blob_garbage_meter.h"
+#include "rocksdb/rocksdb_namespace.h"
+#include "rocksdb/status.h"
+#include "table/internal_iterator.h"
+#include "test_util/sync_point.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+// An internal iterator that passes each key-value encountered to
+// BlobGarbageMeter as inflow in order to measure the total number and size of
+// blobs in the compaction input on a per-blob file basis.
+class BlobCountingIterator : public InternalIterator {
+ public:
+  BlobCountingIterator(InternalIterator* iter,
+                       BlobGarbageMeter* blob_garbage_meter)
+      : iter_(iter), blob_garbage_meter_(blob_garbage_meter) {
+    assert(iter_);
+    assert(blob_garbage_meter_);
+
+    UpdateAndCountBlobIfNeeded();
+  }
+
+  bool Valid() const override { return iter_->Valid() && status_.ok(); }
+
+  void SeekToFirst() override {
+    iter_->SeekToFirst();
+    UpdateAndCountBlobIfNeeded();
+  }
+
+  void SeekToLast() override {
+    iter_->SeekToLast();
+    UpdateAndCountBlobIfNeeded();
+  }
+
+  void Seek(const Slice& target) override {
+    iter_->Seek(target);
+    UpdateAndCountBlobIfNeeded();
+  }
+
+  void SeekForPrev(const Slice& target) override {
+    iter_->SeekForPrev(target);
+    UpdateAndCountBlobIfNeeded();
+  }
+
+  void Next() override {
+    assert(Valid());
+
+    iter_->Next();
+    UpdateAndCountBlobIfNeeded();
+  }
+
+  bool NextAndGetResult(IterateResult* result) override {
+    assert(Valid());
+
+    const bool res = iter_->NextAndGetResult(result);
+    UpdateAndCountBlobIfNeeded();
+    return res;
+  }
+
+  void Prev() override {
+    assert(Valid());
+
+    iter_->Prev();
+    UpdateAndCountBlobIfNeeded();
+  }
+
+  Slice key() const override {
+    assert(Valid());
+    return iter_->key();
+  }
+
+  Slice user_key() const override {
+    assert(Valid());
+    return iter_->user_key();
+  }
+
+  Slice value() const override {
+    assert(Valid());
+    return iter_->value();
+  }
+
+  Status status() const override { return status_; }
+
+  bool PrepareValue() override {
+    assert(Valid());
+    return iter_->PrepareValue();
+  }
+
+  bool MayBeOutOfLowerBound() override {
+    assert(Valid());
+    return iter_->MayBeOutOfLowerBound();
+  }
+
+  IterBoundCheck UpperBoundCheckResult() override {
+    assert(Valid());
+    return iter_->UpperBoundCheckResult();
+  }
+
+  void SetPinnedItersMgr(PinnedIteratorsManager* pinned_iters_mgr) override {
+    iter_->SetPinnedItersMgr(pinned_iters_mgr);
+  }
+
+  bool IsKeyPinned() const override {
+    assert(Valid());
+    return iter_->IsKeyPinned();
+  }
+
+  bool IsValuePinned() const override {
+    assert(Valid());
+    return iter_->IsValuePinned();
+  }
+
+  Status GetProperty(std::string prop_name, std::string* prop) override {
+    return iter_->GetProperty(prop_name, prop);
+  }
+
+ private:
+  void UpdateAndCountBlobIfNeeded() {
+    assert(!iter_->Valid() || iter_->status().ok());
+
+    if (!iter_->Valid()) {
+      status_ = iter_->status();
+      return;
+    }
+
+    TEST_SYNC_POINT(
+        "BlobCountingIterator::UpdateAndCountBlobIfNeeded:ProcessInFlow");
+
+    status_ = blob_garbage_meter_->ProcessInFlow(key(), value());
+  }
+
+  InternalIterator* iter_;
+  BlobGarbageMeter* blob_garbage_meter_;
+  Status status_;
+};
+
+}  // namespace ROCKSDB_NAMESPACE

db/blob/blob_counting_iterator_test.cc

@@ -0,0 +1,325 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  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 "db/blob/blob_counting_iterator.h"
+
+#include <string>
+#include <vector>
+
+#include "db/blob/blob_garbage_meter.h"
+#include "db/blob/blob_index.h"
+#include "db/blob/blob_log_format.h"
+#include "db/dbformat.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+void CheckInFlow(const BlobGarbageMeter& blob_garbage_meter,
+                 uint64_t blob_file_number, uint64_t count, uint64_t bytes) {
+  const auto& flows = blob_garbage_meter.flows();
+
+  const auto it = flows.find(blob_file_number);
+  if (it == flows.end()) {
+    ASSERT_EQ(count, 0);
+    ASSERT_EQ(bytes, 0);
+    return;
+  }
+
+  const auto& in = it->second.GetInFlow();
+
+  ASSERT_EQ(in.GetCount(), count);
+  ASSERT_EQ(in.GetBytes(), bytes);
+}
+
+TEST(BlobCountingIteratorTest, CountBlobs) {
+  // Note: the input consists of three key-values: two are blob references to
+  // different blob files, while the third one is a plain value.
+  constexpr char user_key0[] = "key0";
+  constexpr char user_key1[] = "key1";
+  constexpr char user_key2[] = "key2";
+
+  const std::vector<std::string> keys{
+      test::KeyStr(user_key0, 1, kTypeBlobIndex),
+      test::KeyStr(user_key1, 2, kTypeBlobIndex),
+      test::KeyStr(user_key2, 3, kTypeValue)};
+
+  constexpr uint64_t first_blob_file_number = 4;
+  constexpr uint64_t first_offset = 1000;
+  constexpr uint64_t first_size = 2000;
+
+  std::string first_blob_index;
+  BlobIndex::EncodeBlob(&first_blob_index, first_blob_file_number, first_offset,
+                        first_size, kNoCompression);
+
+  constexpr uint64_t second_blob_file_number = 6;
+  constexpr uint64_t second_offset = 2000;
+  constexpr uint64_t second_size = 4000;
+
+  std::string second_blob_index;
+  BlobIndex::EncodeBlob(&second_blob_index, second_blob_file_number,
+                        second_offset, second_size, kNoCompression);
+
+  const std::vector<std::string> values{first_blob_index, second_blob_index,
+                                        "raw_value"};
+
+  assert(keys.size() == values.size());
+
+  test::VectorIterator input(keys, values);
+  BlobGarbageMeter blob_garbage_meter;
+
+  BlobCountingIterator blob_counter(&input, &blob_garbage_meter);
+
+  constexpr uint64_t first_expected_bytes =
+      first_size +
+      BlobLogRecord::CalculateAdjustmentForRecordHeader(sizeof(user_key0) - 1);
+  constexpr uint64_t second_expected_bytes =
+      second_size +
+      BlobLogRecord::CalculateAdjustmentForRecordHeader(sizeof(user_key1) - 1);
+
+  // Call SeekToFirst and iterate forward
+  blob_counter.SeekToFirst();
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[0]);
+  ASSERT_EQ(blob_counter.user_key(), user_key0);
+  ASSERT_EQ(blob_counter.value(), values[0]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 1,
+              first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 0, 0);
+
+  blob_counter.Next();
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[1]);
+  ASSERT_EQ(blob_counter.user_key(), user_key1);
+  ASSERT_EQ(blob_counter.value(), values[1]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 1,
+              first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 1,
+              second_expected_bytes);
+
+  blob_counter.Next();
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[2]);
+  ASSERT_EQ(blob_counter.user_key(), user_key2);
+  ASSERT_EQ(blob_counter.value(), values[2]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 1,
+              first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 1,
+              second_expected_bytes);
+
+  blob_counter.Next();
+  ASSERT_FALSE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 1,
+              first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 1,
+              second_expected_bytes);
+
+  // Do it again using NextAndGetResult
+  blob_counter.SeekToFirst();
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[0]);
+  ASSERT_EQ(blob_counter.user_key(), user_key0);
+  ASSERT_EQ(blob_counter.value(), values[0]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 2,
+              2 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 1,
+              second_expected_bytes);
+
+  {
+    IterateResult result;
+    ASSERT_TRUE(blob_counter.NextAndGetResult(&result));
+    ASSERT_EQ(result.key, keys[1]);
+    ASSERT_EQ(blob_counter.user_key(), user_key1);
+    ASSERT_TRUE(blob_counter.Valid());
+    ASSERT_OK(blob_counter.status());
+    ASSERT_EQ(blob_counter.key(), keys[1]);
+    ASSERT_EQ(blob_counter.value(), values[1]);
+    CheckInFlow(blob_garbage_meter, first_blob_file_number, 2,
+                2 * first_expected_bytes);
+    CheckInFlow(blob_garbage_meter, second_blob_file_number, 2,
+                2 * second_expected_bytes);
+  }
+
+  {
+    IterateResult result;
+    ASSERT_TRUE(blob_counter.NextAndGetResult(&result));
+    ASSERT_EQ(result.key, keys[2]);
+    ASSERT_EQ(blob_counter.user_key(), user_key2);
+    ASSERT_TRUE(blob_counter.Valid());
+    ASSERT_OK(blob_counter.status());
+    ASSERT_EQ(blob_counter.key(), keys[2]);
+    ASSERT_EQ(blob_counter.value(), values[2]);
+    CheckInFlow(blob_garbage_meter, first_blob_file_number, 2,
+                2 * first_expected_bytes);
+    CheckInFlow(blob_garbage_meter, second_blob_file_number, 2,
+                2 * second_expected_bytes);
+  }
+
+  {
+    IterateResult result;
+    ASSERT_FALSE(blob_counter.NextAndGetResult(&result));
+    ASSERT_FALSE(blob_counter.Valid());
+    ASSERT_OK(blob_counter.status());
+    CheckInFlow(blob_garbage_meter, first_blob_file_number, 2,
+                2 * first_expected_bytes);
+    CheckInFlow(blob_garbage_meter, second_blob_file_number, 2,
+                2 * second_expected_bytes);
+  }
+
+  // Call SeekToLast and iterate backward
+  blob_counter.SeekToLast();
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[2]);
+  ASSERT_EQ(blob_counter.user_key(), user_key2);
+  ASSERT_EQ(blob_counter.value(), values[2]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 2,
+              2 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 2,
+              2 * second_expected_bytes);
+
+  blob_counter.Prev();
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[1]);
+  ASSERT_EQ(blob_counter.user_key(), user_key1);
+  ASSERT_EQ(blob_counter.value(), values[1]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 2,
+              2 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 3,
+              3 * second_expected_bytes);
+
+  blob_counter.Prev();
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[0]);
+  ASSERT_EQ(blob_counter.user_key(), user_key0);
+  ASSERT_EQ(blob_counter.value(), values[0]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 3,
+              3 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 3,
+              3 * second_expected_bytes);
+
+  blob_counter.Prev();
+  ASSERT_FALSE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 3,
+              3 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 3,
+              3 * second_expected_bytes);
+
+  // Call Seek for all keys (plus one that's greater than all of them)
+  blob_counter.Seek(keys[0]);
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[0]);
+  ASSERT_EQ(blob_counter.user_key(), user_key0);
+  ASSERT_EQ(blob_counter.value(), values[0]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 4,
+              4 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 3,
+              3 * second_expected_bytes);
+
+  blob_counter.Seek(keys[1]);
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[1]);
+  ASSERT_EQ(blob_counter.user_key(), user_key1);
+  ASSERT_EQ(blob_counter.value(), values[1]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 4,
+              4 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 4,
+              4 * second_expected_bytes);
+
+  blob_counter.Seek(keys[2]);
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[2]);
+  ASSERT_EQ(blob_counter.user_key(), user_key2);
+  ASSERT_EQ(blob_counter.value(), values[2]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 4,
+              4 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 4,
+              4 * second_expected_bytes);
+
+  blob_counter.Seek("zzz");
+  ASSERT_FALSE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 4,
+              4 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 4,
+              4 * second_expected_bytes);
+
+  // Call SeekForPrev for all keys (plus one that's less than all of them)
+  blob_counter.SeekForPrev("aaa");
+  ASSERT_FALSE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 4,
+              4 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 4,
+              4 * second_expected_bytes);
+
+  blob_counter.SeekForPrev(keys[0]);
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[0]);
+  ASSERT_EQ(blob_counter.user_key(), user_key0);
+  ASSERT_EQ(blob_counter.value(), values[0]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 5,
+              5 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 4,
+              4 * second_expected_bytes);
+
+  blob_counter.SeekForPrev(keys[1]);
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[1]);
+  ASSERT_EQ(blob_counter.user_key(), user_key1);
+  ASSERT_EQ(blob_counter.value(), values[1]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 5,
+              5 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 5,
+              5 * second_expected_bytes);
+
+  blob_counter.SeekForPrev(keys[2]);
+  ASSERT_TRUE(blob_counter.Valid());
+  ASSERT_OK(blob_counter.status());
+  ASSERT_EQ(blob_counter.key(), keys[2]);
+  ASSERT_EQ(blob_counter.user_key(), user_key2);
+  ASSERT_EQ(blob_counter.value(), values[2]);
+  CheckInFlow(blob_garbage_meter, first_blob_file_number, 5,
+              5 * first_expected_bytes);
+  CheckInFlow(blob_garbage_meter, second_blob_file_number, 5,
+              5 * second_expected_bytes);
+}
+
+TEST(BlobCountingIteratorTest, CorruptBlobIndex) {
+  const std::vector<std::string> keys{
+      test::KeyStr("user_key", 1, kTypeBlobIndex)};
+  const std::vector<std::string> values{"i_am_not_a_blob_index"};
+
+  assert(keys.size() == values.size());
+
+  test::VectorIterator input(keys, values);
+  BlobGarbageMeter blob_garbage_meter;
+
+  BlobCountingIterator blob_counter(&input, &blob_garbage_meter);
+
+  blob_counter.SeekToFirst();
+  ASSERT_FALSE(blob_counter.Valid());
+  ASSERT_NOK(blob_counter.status());
+}
+
+}  // namespace ROCKSDB_NAMESPACE
+
+int main(int argc, char** argv) {
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

db/blob/blob_file_addition.cc

@@ -135,7 +135,8 @@ std::ostream& operator<<(std::ostream& os,
      << " total_blob_count: " << blob_file_addition.GetTotalBlobCount()
      << " total_blob_bytes: " << blob_file_addition.GetTotalBlobBytes()
      << " checksum_method: " << blob_file_addition.GetChecksumMethod()
-     << " checksum_value: " << blob_file_addition.GetChecksumValue();
+     << " checksum_value: "
+     << Slice(blob_file_addition.GetChecksumValue()).ToString(/* hex */ true);
 
   return os;
 }
@@ -146,7 +147,8 @@ JSONWriter& operator<<(JSONWriter& jw,
      << "TotalBlobCount" << blob_file_addition.GetTotalBlobCount()
      << "TotalBlobBytes" << blob_file_addition.GetTotalBlobBytes()
      << "ChecksumMethod" << blob_file_addition.GetChecksumMethod()
-     << "ChecksumValue" << blob_file_addition.GetChecksumValue();
+     << "ChecksumValue"
+     << Slice(blob_file_addition.GetChecksumValue()).ToString(/* hex */ true);
 
   return jw;
 }

db/blob/blob_file_addition_test.cc

@@ -46,7 +46,9 @@ TEST_F(BlobFileAdditionTest, NonEmpty) {
   constexpr uint64_t total_blob_count = 2;
   constexpr uint64_t total_blob_bytes = 123456;
   const std::string checksum_method("SHA1");
-  const std::string checksum_value("bdb7f34a59dfa1592ce7f52e99f98c570c525cbd");
+  const std::string checksum_value(
+      "\xbd\xb7\xf3\x4a\x59\xdf\xa1\x59\x2c\xe7\xf5\x2e\x99\xf9\x8c\x57\x0c\x52"
+      "\x5c\xbd");
 
   BlobFileAddition blob_file_addition(blob_file_number, total_blob_count,
                                       total_blob_bytes, checksum_method,
@@ -113,7 +115,9 @@ TEST_F(BlobFileAdditionTest, DecodeErrors) {
     ASSERT_TRUE(std::strstr(s.getState(), "checksum value"));
   }
 
-  constexpr char checksum_value[] = "bdb7f34a59dfa1592ce7f52e99f98c570c525cbd";
+  constexpr char checksum_value[] =
+      "\xbd\xb7\xf3\x4a\x59\xdf\xa1\x59\x2c\xe7\xf5\x2e\x99\xf9\x8c\x57\x0c\x52"
+      "\x5c\xbd";
   PutLengthPrefixedSlice(&str, checksum_value);
   slice = str;
 
@@ -150,7 +154,7 @@ TEST_F(BlobFileAdditionTest, ForwardCompatibleCustomField) {
   constexpr uint64_t total_blob_count = 9999;
   constexpr uint64_t total_blob_bytes = 100000000;
   const std::string checksum_method("CRC32");
-  const std::string checksum_value("3d87ff57");
+  const std::string checksum_value("\x3d\x87\xff\x57");
 
   BlobFileAddition blob_file_addition(blob_file_number, total_blob_count,
                                       total_blob_bytes, checksum_method,
@@ -178,7 +182,7 @@ TEST_F(BlobFileAdditionTest, ForwardIncompatibleCustomField) {
   constexpr uint64_t total_blob_count = 100;
   constexpr uint64_t total_blob_bytes = 2000000;
   const std::string checksum_method("CRC32B");
-  const std::string checksum_value("6dbdf23a");
+  const std::string checksum_value("\x6d\xbd\xf2\x3a");
 
   BlobFileAddition blob_file_addition(blob_file_number, total_blob_count,
                                       total_blob_bytes, checksum_method,

db/blob/blob_file_builder.cc

@@ -191,7 +191,7 @@ Status BlobFileBuilder::OpenBlobFileIfNeeded() {
       immutable_options_->clock, io_tracer_, statistics,
       immutable_options_->listeners,
       immutable_options_->file_checksum_gen_factory.get(),
-      tmp_set.Contains(FileType::kBlobFile)));
+      tmp_set.Contains(FileType::kBlobFile), false));
 
   constexpr bool do_flush = false;
 
@@ -228,6 +228,7 @@ Status BlobFileBuilder::CompressBlobIfNeeded(
   assert(blob);
   assert(compressed_blob);
   assert(compressed_blob->empty());
+  assert(immutable_options_);
 
   if (blob_compression_type_ == kNoCompression) {
     return Status::OK();
@@ -242,7 +243,16 @@ Status BlobFileBuilder::CompressBlobIfNeeded(
 
   constexpr uint32_t compression_format_version = 2;
 
-  if (!CompressData(*blob, info, compression_format_version, compressed_blob)) {
+  bool success = false;
+
+  {
+    StopWatch stop_watch(immutable_options_->clock, immutable_options_->stats,
+                         BLOB_DB_COMPRESSION_MICROS);
+    success =
+        CompressData(*blob, info, compression_format_version, compressed_blob);
+  }
+
+  if (!success) {
     return Status::Corruption("Error compressing blob");
   }
 

db/blob/blob_file_meta.cc

@@ -8,7 +8,13 @@
 #include <ostream>
 #include <sstream>
 
+#include "db/blob/blob_log_format.h"
+#include "rocksdb/slice.h"
+
 namespace ROCKSDB_NAMESPACE {
+uint64_t SharedBlobFileMetaData::GetBlobFileSize() const {
+  return BlobLogHeader::kSize + total_blob_bytes_ + BlobLogFooter::kSize;
+}
 
 std::string SharedBlobFileMetaData::DebugString() const {
   std::ostringstream oss;
@@ -23,7 +29,8 @@ std::ostream& operator<<(std::ostream& os,
      << " total_blob_count: " << shared_meta.GetTotalBlobCount()
      << " total_blob_bytes: " << shared_meta.GetTotalBlobBytes()
      << " checksum_method: " << shared_meta.GetChecksumMethod()
-     << " checksum_value: " << shared_meta.GetChecksumValue();
+     << " checksum_value: "
+     << Slice(shared_meta.GetChecksumValue()).ToString(/* hex */ true);
 
   return os;
 }

db/blob/blob_file_meta.h

@@ -51,6 +51,7 @@ class SharedBlobFileMetaData {
   SharedBlobFileMetaData(SharedBlobFileMetaData&&) = delete;
   SharedBlobFileMetaData& operator=(SharedBlobFileMetaData&&) = delete;
 
+  uint64_t GetBlobFileSize() const;
   uint64_t GetBlobFileNumber() const { return blob_file_number_; }
   uint64_t GetTotalBlobCount() const { return total_blob_count_; }
   uint64_t GetTotalBlobBytes() const { return total_blob_bytes_; }
@@ -112,6 +113,11 @@ class BlobFileMetaData {
     return shared_meta_;
   }
 
+  uint64_t GetBlobFileSize() const {
+    assert(shared_meta_);
+    return shared_meta_->GetBlobFileSize();
+  }
+
   uint64_t GetBlobFileNumber() const {
     assert(shared_meta_);
     return shared_meta_->GetBlobFileNumber();

db/blob/blob_file_reader.cc

@@ -10,6 +10,7 @@
 
 #include "db/blob/blob_log_format.h"
 #include "file/filename.h"
+#include "monitoring/statistics.h"
 #include "options/cf_options.h"
 #include "rocksdb/file_system.h"
 #include "rocksdb/slice.h"
@@ -17,6 +18,7 @@
 #include "test_util/sync_point.h"
 #include "util/compression.h"
 #include "util/crc32c.h"
+#include "util/stop_watch.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -42,25 +44,28 @@ Status BlobFileReader::Create(
 
   assert(file_reader);
 
+  Statistics* const statistics = immutable_options.stats;
+
   CompressionType compression_type = kNoCompression;
 
   {
-    const Status s =
-        ReadHeader(file_reader.get(), column_family_id, &compression_type);
+    const Status s = ReadHeader(file_reader.get(), column_family_id, statistics,
+                                &compression_type);
     if (!s.ok()) {
       return s;
     }
   }
 
   {
-    const Status s = ReadFooter(file_size, file_reader.get());
+    const Status s = ReadFooter(file_reader.get(), file_size, statistics);
     if (!s.ok()) {
       return s;
     }
   }
 
   blob_file_reader->reset(
-      new BlobFileReader(std::move(file_reader), file_size, compression_type));
+      new BlobFileReader(std::move(file_reader), file_size, compression_type,
+                         immutable_options.clock, statistics));
 
   return Status::OK();
 }
@@ -127,6 +132,7 @@ Status BlobFileReader::OpenFile(
 
 Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
                                   uint32_t column_family_id,
+                                  Statistics* statistics,
                                   CompressionType* compression_type) {
   assert(file_reader);
   assert(compression_type);
@@ -141,8 +147,9 @@ Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
     constexpr uint64_t read_offset = 0;
     constexpr size_t read_size = BlobLogHeader::kSize;
 
-    const Status s = ReadFromFile(file_reader, read_offset, read_size,
-                                  &header_slice, &buf, &aligned_buf);
+    const Status s =
+        ReadFromFile(file_reader, read_offset, read_size, statistics,
+                     &header_slice, &buf, &aligned_buf);
     if (!s.ok()) {
       return s;
     }
@@ -175,8 +182,8 @@ Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
   return Status::OK();
 }
 
-Status BlobFileReader::ReadFooter(uint64_t file_size,
-                                  const RandomAccessFileReader* file_reader) {
+Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
+                                  uint64_t file_size, Statistics* statistics) {
   assert(file_size >= BlobLogHeader::kSize + BlobLogFooter::kSize);
   assert(file_reader);
 
@@ -190,8 +197,9 @@ Status BlobFileReader::ReadFooter(uint64_t file_size,
     const uint64_t read_offset = file_size - BlobLogFooter::kSize;
     constexpr size_t read_size = BlobLogFooter::kSize;
 
-    const Status s = ReadFromFile(file_reader, read_offset, read_size,
-                                  &footer_slice, &buf, &aligned_buf);
+    const Status s =
+        ReadFromFile(file_reader, read_offset, read_size, statistics,
+                     &footer_slice, &buf, &aligned_buf);
     if (!s.ok()) {
       return s;
     }
@@ -220,14 +228,16 @@ Status BlobFileReader::ReadFooter(uint64_t file_size,
 
 Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
                                     uint64_t read_offset, size_t read_size,
-                                    Slice* slice, Buffer* buf,
-                                    AlignedBuf* aligned_buf) {
+                                    Statistics* statistics, Slice* slice,
+                                    Buffer* buf, AlignedBuf* aligned_buf) {
   assert(slice);
   assert(buf);
   assert(aligned_buf);
 
   assert(file_reader);
 
+  RecordTick(statistics, BLOB_DB_BLOB_FILE_BYTES_READ, read_size);
+
   Status s;
 
   if (file_reader->use_direct_io()) {
@@ -256,10 +266,13 @@ Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
 
 BlobFileReader::BlobFileReader(
     std::unique_ptr<RandomAccessFileReader>&& file_reader, uint64_t file_size,
-    CompressionType compression_type)
+    CompressionType compression_type, SystemClock* clock,
+    Statistics* statistics)
     : file_reader_(std::move(file_reader)),
       file_size_(file_size),
-      compression_type_(compression_type) {
+      compression_type_(compression_type),
+      clock_(clock),
+      statistics_(statistics) {
   assert(file_reader_);
 }
 
@@ -304,7 +317,7 @@ Status BlobFileReader::GetBlob(const ReadOptions& read_options,
     TEST_SYNC_POINT("BlobFileReader::GetBlob:ReadFromFile");
 
     const Status s = ReadFromFile(file_reader_.get(), record_offset,
-                                  static_cast<size_t>(record_size),
+                                  static_cast<size_t>(record_size), statistics_,
                                   &record_slice, &buf, &aligned_buf);
     if (!s.ok()) {
       return s;
@@ -324,8 +337,8 @@ Status BlobFileReader::GetBlob(const ReadOptions& read_options,
   const Slice value_slice(record_slice.data() + adjustment, value_size);
 
   {
-    const Status s =
-        UncompressBlobIfNeeded(value_slice, compression_type, value);
+    const Status s = UncompressBlobIfNeeded(value_slice, compression_type,
+                                            clock_, statistics_, value);
     if (!s.ok()) {
       return s;
     }
@@ -382,6 +395,8 @@ Status BlobFileReader::VerifyBlob(const Slice& record_slice,
 
 Status BlobFileReader::UncompressBlobIfNeeded(const Slice& value_slice,
                                               CompressionType compression_type,
+                                              SystemClock* clock,
+                                              Statistics* statistics,
                                               PinnableSlice* value) {
   assert(value);
 
@@ -399,9 +414,14 @@ Status BlobFileReader::UncompressBlobIfNeeded(const Slice& value_slice,
   constexpr uint32_t compression_format_version = 2;
   constexpr MemoryAllocator* allocator = nullptr;
 
-  CacheAllocationPtr output =
-      UncompressData(info, value_slice.data(), value_slice.size(),
-                     &uncompressed_size, compression_format_version, allocator);
+  CacheAllocationPtr output;
+
+  {
+    StopWatch stop_watch(clock, statistics, BLOB_DB_DECOMPRESSION_MICROS);
+    output = UncompressData(info, value_slice.data(), value_slice.size(),
+                            &uncompressed_size, compression_format_version,
+                            allocator);
+  }
 
   TEST_SYNC_POINT_CALLBACK(
       "BlobFileReader::UncompressBlobIfNeeded:TamperWithResult", &output);

db/blob/blob_file_reader.h

@@ -21,6 +21,7 @@ class HistogramImpl;
 struct ReadOptions;
 class Slice;
 class PinnableSlice;
+class Statistics;
 
 class BlobFileReader {
  public:
@@ -44,7 +45,8 @@ class BlobFileReader {
 
  private:
   BlobFileReader(std::unique_ptr<RandomAccessFileReader>&& file_reader,
-                 uint64_t file_size, CompressionType compression_type);
+                 uint64_t file_size, CompressionType compression_type,
+                 SystemClock* clock, Statistics* statistics);
 
   static Status OpenFile(const ImmutableOptions& immutable_options,
                          const FileOptions& file_opts,
@@ -55,17 +57,17 @@ class BlobFileReader {
                          std::unique_ptr<RandomAccessFileReader>* file_reader);
 
   static Status ReadHeader(const RandomAccessFileReader* file_reader,
-                           uint32_t column_family_id,
+                           uint32_t column_family_id, Statistics* statistics,
                            CompressionType* compression_type);
 
-  static Status ReadFooter(uint64_t file_size,
-                           const RandomAccessFileReader* file_reader);
+  static Status ReadFooter(const RandomAccessFileReader* file_reader,
+                           uint64_t file_size, Statistics* statistics);
 
   using Buffer = std::unique_ptr<char[]>;
 
   static Status ReadFromFile(const RandomAccessFileReader* file_reader,
                              uint64_t read_offset, size_t read_size,
-                             Slice* slice, Buffer* buf,
+                             Statistics* statistics, Slice* slice, Buffer* buf,
                              AlignedBuf* aligned_buf);
 
   static Status VerifyBlob(const Slice& record_slice, const Slice& user_key,
@@ -73,6 +75,8 @@ class BlobFileReader {
 
   static Status UncompressBlobIfNeeded(const Slice& value_slice,
                                        CompressionType compression_type,
+                                       SystemClock* clock,
+                                       Statistics* statistics,
                                        PinnableSlice* value);
 
   static void SaveValue(const Slice& src, PinnableSlice* dst);
@@ -80,6 +84,8 @@ class BlobFileReader {
   std::unique_ptr<RandomAccessFileReader> file_reader_;
   uint64_t file_size_;
   CompressionType compression_type_;
+  SystemClock* clock_;
+  Statistics* statistics_;
 };
 
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_garbage_meter.cc

@@ -0,0 +1,100 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  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 "db/blob/blob_garbage_meter.h"
+
+#include "db/blob/blob_index.h"
+#include "db/blob/blob_log_format.h"
+#include "db/dbformat.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+Status BlobGarbageMeter::ProcessInFlow(const Slice& key, const Slice& value) {
+  uint64_t blob_file_number = kInvalidBlobFileNumber;
+  uint64_t bytes = 0;
+
+  const Status s = Parse(key, value, &blob_file_number, &bytes);
+  if (!s.ok()) {
+    return s;
+  }
+
+  if (blob_file_number == kInvalidBlobFileNumber) {
+    return Status::OK();
+  }
+
+  flows_[blob_file_number].AddInFlow(bytes);
+
+  return Status::OK();
+}
+
+Status BlobGarbageMeter::ProcessOutFlow(const Slice& key, const Slice& value) {
+  uint64_t blob_file_number = kInvalidBlobFileNumber;
+  uint64_t bytes = 0;
+
+  const Status s = Parse(key, value, &blob_file_number, &bytes);
+  if (!s.ok()) {
+    return s;
+  }
+
+  if (blob_file_number == kInvalidBlobFileNumber) {
+    return Status::OK();
+  }
+
+  // Note: in order to measure the amount of additional garbage, we only need to
+  // track the outflow for preexisting files, i.e. those that also had inflow.
+  // (Newly written files would only have outflow.)
+  auto it = flows_.find(blob_file_number);
+  if (it == flows_.end()) {
+    return Status::OK();
+  }
+
+  it->second.AddOutFlow(bytes);
+
+  return Status::OK();
+}
+
+Status BlobGarbageMeter::Parse(const Slice& key, const Slice& value,
+                               uint64_t* blob_file_number, uint64_t* bytes) {
+  assert(blob_file_number);
+  assert(*blob_file_number == kInvalidBlobFileNumber);
+  assert(bytes);
+  assert(*bytes == 0);
+
+  ParsedInternalKey ikey;
+
+  {
+    constexpr bool log_err_key = false;
+    const Status s = ParseInternalKey(key, &ikey, log_err_key);
+    if (!s.ok()) {
+      return s;
+    }
+  }
+
+  if (ikey.type != kTypeBlobIndex) {
+    return Status::OK();
+  }
+
+  BlobIndex blob_index;
+
+  {
+    const Status s = blob_index.DecodeFrom(value);
+    if (!s.ok()) {
+      return s;
+    }
+  }
+
+  if (blob_index.IsInlined() || blob_index.HasTTL()) {
+    return Status::Corruption("Unexpected TTL/inlined blob index");
+  }
+
+  *blob_file_number = blob_index.file_number();
+  *bytes =
+      blob_index.size() +
+      BlobLogRecord::CalculateAdjustmentForRecordHeader(ikey.user_key.size());
+
+  return Status::OK();
+}
+
+}  // namespace ROCKSDB_NAMESPACE

db/blob/blob_garbage_meter.h

@@ -0,0 +1,102 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  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 <cassert>
+#include <cstdint>
+#include <unordered_map>
+
+#include "db/blob/blob_constants.h"
+#include "rocksdb/rocksdb_namespace.h"
+#include "rocksdb/status.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+class Slice;
+
+// A class that can be used to compute the amount of additional garbage
+// generated by a compaction. It parses the keys and blob references in the
+// input and output of a compaction, and aggregates the "inflow" and "outflow"
+// on a per-blob file basis. The amount of additional garbage for any given blob
+// file can then be computed by subtracting the outflow from the inflow.
+class BlobGarbageMeter {
+ public:
+  // A class to store the number and total size of blobs on a per-blob file
+  // basis.
+  class BlobStats {
+   public:
+    void Add(uint64_t bytes) {
+      ++count_;
+      bytes_ += bytes;
+    }
+    void Add(uint64_t count, uint64_t bytes) {
+      count_ += count;
+      bytes_ += bytes;
+    }
+
+    uint64_t GetCount() const { return count_; }
+    uint64_t GetBytes() const { return bytes_; }
+
+   private:
+    uint64_t count_ = 0;
+    uint64_t bytes_ = 0;
+  };
+
+  // A class to keep track of the "inflow" and the "outflow" and to compute the
+  // amount of additional garbage for a given blob file.
+  class BlobInOutFlow {
+   public:
+    void AddInFlow(uint64_t bytes) {
+      in_flow_.Add(bytes);
+      assert(IsValid());
+    }
+    void AddOutFlow(uint64_t bytes) {
+      out_flow_.Add(bytes);
+      assert(IsValid());
+    }
+
+    const BlobStats& GetInFlow() const { return in_flow_; }
+    const BlobStats& GetOutFlow() const { return out_flow_; }
+
+    bool IsValid() const {
+      return in_flow_.GetCount() >= out_flow_.GetCount() &&
+             in_flow_.GetBytes() >= out_flow_.GetBytes();
+    }
+    bool HasGarbage() const {
+      assert(IsValid());
+      return in_flow_.GetCount() > out_flow_.GetCount();
+    }
+    uint64_t GetGarbageCount() const {
+      assert(IsValid());
+      assert(HasGarbage());
+      return in_flow_.GetCount() - out_flow_.GetCount();
+    }
+    uint64_t GetGarbageBytes() const {
+      assert(IsValid());
+      assert(HasGarbage());
+      return in_flow_.GetBytes() - out_flow_.GetBytes();
+    }
+
+   private:
+    BlobStats in_flow_;
+    BlobStats out_flow_;
+  };
+
+  Status ProcessInFlow(const Slice& key, const Slice& value);
+  Status ProcessOutFlow(const Slice& key, const Slice& value);
+
+  const std::unordered_map<uint64_t, BlobInOutFlow>& flows() const {
+    return flows_;
+  }
+
+ private:
+  static Status Parse(const Slice& key, const Slice& value,
+                      uint64_t* blob_file_number, uint64_t* bytes);
+
+  std::unordered_map<uint64_t, BlobInOutFlow> flows_;
+};
+
+}  // namespace ROCKSDB_NAMESPACE

db/blob/blob_garbage_meter_test.cc

@@ -0,0 +1,196 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  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 "db/blob/blob_garbage_meter.h"
+
+#include <string>
+#include <vector>
+
+#include "db/blob/blob_index.h"
+#include "db/blob/blob_log_format.h"
+#include "db/dbformat.h"
+#include "test_util/testharness.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+TEST(BlobGarbageMeterTest, MeasureGarbage) {
+  BlobGarbageMeter blob_garbage_meter;
+
+  struct BlobDescriptor {
+    std::string user_key;
+    uint64_t blob_file_number;
+    uint64_t offset;
+    uint64_t size;
+    CompressionType compression_type;
+    bool has_in_flow;
+    bool has_out_flow;
+
+    uint64_t GetExpectedBytes() const {
+      return size +
+             BlobLogRecord::CalculateAdjustmentForRecordHeader(user_key.size());
+    }
+  };
+
+  // Note: blob file 4 has the same inflow and outflow and hence no additional
+  // garbage. Blob file 5 has less outflow than inflow and thus it does have
+  // additional garbage. Blob file 6 is a newly written file (i.e. no inflow,
+  // only outflow) and is thus not tracked by the meter.
+  std::vector<BlobDescriptor> blobs{
+      {"key", 4, 1234, 555, kLZ4Compression, true, true},
+      {"other_key", 4, 6789, 101010, kLZ4Compression, true, true},
+      {"yet_another_key", 5, 22222, 3456, kLZ4Compression, true, true},
+      {"foo_key", 5, 77777, 8888, kLZ4Compression, true, true},
+      {"bar_key", 5, 999999, 1212, kLZ4Compression, true, false},
+      {"baz_key", 5, 1234567, 890, kLZ4Compression, true, false},
+      {"new_key", 6, 7777, 9999, kNoCompression, false, true}};
+
+  for (const auto& blob : blobs) {
+    constexpr SequenceNumber seq = 123;
+    const InternalKey key(blob.user_key, seq, kTypeBlobIndex);
+    const Slice key_slice = key.Encode();
+
+    std::string value;
+    BlobIndex::EncodeBlob(&value, blob.blob_file_number, blob.offset, blob.size,
+                          blob.compression_type);
+    const Slice value_slice(value);
+
+    if (blob.has_in_flow) {
+      ASSERT_OK(blob_garbage_meter.ProcessInFlow(key_slice, value_slice));
+    }
+    if (blob.has_out_flow) {
+      ASSERT_OK(blob_garbage_meter.ProcessOutFlow(key_slice, value_slice));
+    }
+  }
+
+  const auto& flows = blob_garbage_meter.flows();
+  ASSERT_EQ(flows.size(), 2);
+
+  {
+    const auto it = flows.find(4);
+    ASSERT_NE(it, flows.end());
+
+    const auto& flow = it->second;
+
+    constexpr uint64_t expected_count = 2;
+    const uint64_t expected_bytes =
+        blobs[0].GetExpectedBytes() + blobs[1].GetExpectedBytes();
+
+    const auto& in = flow.GetInFlow();
+    ASSERT_EQ(in.GetCount(), expected_count);
+    ASSERT_EQ(in.GetBytes(), expected_bytes);
+
+    const auto& out = flow.GetOutFlow();
+    ASSERT_EQ(out.GetCount(), expected_count);
+    ASSERT_EQ(out.GetBytes(), expected_bytes);
+
+    ASSERT_TRUE(flow.IsValid());
+    ASSERT_FALSE(flow.HasGarbage());
+  }
+
+  {
+    const auto it = flows.find(5);
+    ASSERT_NE(it, flows.end());
+
+    const auto& flow = it->second;
+
+    const auto& in = flow.GetInFlow();
+
+    constexpr uint64_t expected_in_count = 4;
+    const uint64_t expected_in_bytes =
+        blobs[2].GetExpectedBytes() + blobs[3].GetExpectedBytes() +
+        blobs[4].GetExpectedBytes() + blobs[5].GetExpectedBytes();
+
+    ASSERT_EQ(in.GetCount(), expected_in_count);
+    ASSERT_EQ(in.GetBytes(), expected_in_bytes);
+
+    const auto& out = flow.GetOutFlow();
+
+    constexpr uint64_t expected_out_count = 2;
+    const uint64_t expected_out_bytes =
+        blobs[2].GetExpectedBytes() + blobs[3].GetExpectedBytes();
+
+    ASSERT_EQ(out.GetCount(), expected_out_count);
+    ASSERT_EQ(out.GetBytes(), expected_out_bytes);
+
+    ASSERT_TRUE(flow.IsValid());
+    ASSERT_TRUE(flow.HasGarbage());
+    ASSERT_EQ(flow.GetGarbageCount(), expected_in_count - expected_out_count);
+    ASSERT_EQ(flow.GetGarbageBytes(), expected_in_bytes - expected_out_bytes);
+  }
+}
+
+TEST(BlobGarbageMeterTest, PlainValue) {
+  constexpr char user_key[] = "user_key";
+  constexpr SequenceNumber seq = 123;
+
+  const InternalKey key(user_key, seq, kTypeValue);
+  const Slice key_slice = key.Encode();
+
+  constexpr char value[] = "value";
+  const Slice value_slice(value);
+
+  BlobGarbageMeter blob_garbage_meter;
+
+  ASSERT_OK(blob_garbage_meter.ProcessInFlow(key_slice, value_slice));
+  ASSERT_OK(blob_garbage_meter.ProcessOutFlow(key_slice, value_slice));
+  ASSERT_TRUE(blob_garbage_meter.flows().empty());
+}
+
+TEST(BlobGarbageMeterTest, CorruptInternalKey) {
+  constexpr char corrupt_key[] = "i_am_corrupt";
+  const Slice key_slice(corrupt_key);
+
+  constexpr char value[] = "value";
+  const Slice value_slice(value);
+
+  BlobGarbageMeter blob_garbage_meter;
+
+  ASSERT_NOK(blob_garbage_meter.ProcessInFlow(key_slice, value_slice));
+  ASSERT_NOK(blob_garbage_meter.ProcessOutFlow(key_slice, value_slice));
+}
+
+TEST(BlobGarbageMeterTest, CorruptBlobIndex) {
+  constexpr char user_key[] = "user_key";
+  constexpr SequenceNumber seq = 123;
+
+  const InternalKey key(user_key, seq, kTypeBlobIndex);
+  const Slice key_slice = key.Encode();
+
+  constexpr char value[] = "i_am_not_a_blob_index";
+  const Slice value_slice(value);
+
+  BlobGarbageMeter blob_garbage_meter;
+
+  ASSERT_NOK(blob_garbage_meter.ProcessInFlow(key_slice, value_slice));
+  ASSERT_NOK(blob_garbage_meter.ProcessOutFlow(key_slice, value_slice));
+}
+
+TEST(BlobGarbageMeterTest, InlinedTTLBlobIndex) {
+  constexpr char user_key[] = "user_key";
+  constexpr SequenceNumber seq = 123;
+
+  const InternalKey key(user_key, seq, kTypeBlobIndex);
+  const Slice key_slice = key.Encode();
+
+  constexpr uint64_t expiration = 1234567890;
+  constexpr char inlined_value[] = "inlined";
+
+  std::string value;
+  BlobIndex::EncodeInlinedTTL(&value, expiration, inlined_value);
+
+  const Slice value_slice(value);
+
+  BlobGarbageMeter blob_garbage_meter;
+
+  ASSERT_NOK(blob_garbage_meter.ProcessInFlow(key_slice, value_slice));
+  ASSERT_NOK(blob_garbage_meter.ProcessOutFlow(key_slice, value_slice));
+}
+
+}  // namespace ROCKSDB_NAMESPACE
+
+int main(int argc, char** argv) {
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

db/blob/blob_log_format.h

@@ -107,7 +107,8 @@ struct BlobLogRecord {
   // Note that the offset field of BlobIndex actually points to the blob value
   // as opposed to the start of the blob record. The following method can
   // be used to calculate the adjustment needed to read the blob record header.
-  static uint64_t CalculateAdjustmentForRecordHeader(uint64_t key_size) {
+  static constexpr uint64_t CalculateAdjustmentForRecordHeader(
+      uint64_t key_size) {
     return key_size + kHeaderSize;
   }
 

db/blob/blob_log_sequential_reader.h

@@ -10,6 +10,8 @@
 #include "db/blob/blob_log_format.h"
 #include "rocksdb/slice.h"
 
+#define MAX_HEADER_SIZE(a, b, c) (a > b ? (a > c ? a : c) : (b > c ? b : c))
+
 namespace ROCKSDB_NAMESPACE {
 
 class RandomAccessFileReader;
@@ -69,10 +71,13 @@ class BlobLogSequentialReader {
   Statistics* statistics_;
 
   Slice buffer_;
-  char header_buf_[BlobLogRecord::kHeaderSize];
+  char header_buf_[MAX_HEADER_SIZE(BlobLogHeader::kSize, BlobLogFooter::kSize,
+                                   BlobLogRecord::kHeaderSize)];
 
   // which byte to read next
   uint64_t next_byte_;
 };
 
 }  // namespace ROCKSDB_NAMESPACE
+
+#undef MAX_HEADER_SIZE

db/blob/db_blob_basic_test.cc

@@ -16,7 +16,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBBlobBasicTest : public DBTestBase {
  protected:
   DBBlobBasicTest()
-      : DBTestBase("/db_blob_basic_test", /* env_do_fsync */ false) {}
+      : DBTestBase("db_blob_basic_test", /* env_do_fsync */ false) {}
 };
 
 TEST_F(DBBlobBasicTest, GetBlob) {

db/blob/db_blob_compaction_test.cc

@@ -4,6 +4,7 @@
 //  (found in the LICENSE.Apache file in the root directory).
 
 #include "db/blob/blob_index.h"
+#include "db/blob/blob_log_format.h"
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "test_util/sync_point.h"
@@ -14,7 +15,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBBlobCompactionTest : public DBTestBase {
  public:
   explicit DBBlobCompactionTest()
-      : DBTestBase("/db_blob_compaction_test", /*env_do_fsync=*/false) {}
+      : DBTestBase("db_blob_compaction_test", /*env_do_fsync=*/false) {}
 
 #ifndef ROCKSDB_LITE
   const std::vector<InternalStats::CompactionStats>& GetCompactionStats() {
@@ -152,6 +153,31 @@ class AlwaysKeepFilter : public CompactionFilter {
     return CompactionFilter::Decision::kKeep;
   }
 };
+
+class SkipUntilFilter : public CompactionFilter {
+ public:
+  explicit SkipUntilFilter(std::string skip_until)
+      : skip_until_(std::move(skip_until)) {}
+
+  const char* Name() const override {
+    return "rocksdb.compaction.filter.skip.until";
+  }
+
+  CompactionFilter::Decision FilterV2(int /* level */, const Slice& /* key */,
+                                      ValueType /* value_type */,
+                                      const Slice& /* existing_value */,
+                                      std::string* /* new_value */,
+                                      std::string* skip_until) const override {
+    assert(skip_until);
+    *skip_until = skip_until_;
+
+    return CompactionFilter::Decision::kRemoveAndSkipUntil;
+  }
+
+ private:
+  std::string skip_until_;
+};
+
 }  // anonymous namespace
 
 class DBBlobBadCompactionFilterTest
@@ -254,6 +280,49 @@ TEST_F(DBBlobCompactionTest, BlindWriteFilter) {
   Close();
 }
 
+TEST_F(DBBlobCompactionTest, SkipUntilFilter) {
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+
+  std::unique_ptr<CompactionFilter> compaction_filter_guard(
+      new SkipUntilFilter("z"));
+  options.compaction_filter = compaction_filter_guard.get();
+
+  Reopen(options);
+
+  const std::vector<std::string> keys{"a", "b", "c"};
+  const std::vector<std::string> values{"a_value", "b_value", "c_value"};
+  assert(keys.size() == values.size());
+
+  for (size_t i = 0; i < keys.size(); ++i) {
+    ASSERT_OK(Put(keys[i], values[i]));
+  }
+
+  ASSERT_OK(Flush());
+
+  int process_in_flow_called = 0;
+
+  SyncPoint::GetInstance()->SetCallBack(
+      "BlobCountingIterator::UpdateAndCountBlobIfNeeded:ProcessInFlow",
+      [&process_in_flow_called](void* /* arg */) { ++process_in_flow_called; });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), /* begin */ nullptr,
+                              /* end */ nullptr));
+
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+
+  for (const auto& key : keys) {
+    ASSERT_EQ(Get(key), "NOT_FOUND");
+  }
+
+  // Make sure SkipUntil was performed using iteration rather than Seek
+  ASSERT_EQ(process_in_flow_called, keys.size());
+
+  Close();
+}
+
 TEST_P(DBBlobBadCompactionFilterTest, BadDecisionFromCompactionFilter) {
   Options options = GetDefaultOptions();
   options.enable_blob_files = true;
@@ -390,6 +459,138 @@ TEST_F(DBBlobCompactionTest, CompactionFilterReadBlobAndKeep) {
   Close();
 }
 
+TEST_F(DBBlobCompactionTest, TrackGarbage) {
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+
+  Reopen(options);
+
+  // First table+blob file pair: 4 blobs with different keys
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "first_value";
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "second_value";
+  constexpr char third_key[] = "third_key";
+  constexpr char third_value[] = "third_value";
+  constexpr char fourth_key[] = "fourth_key";
+  constexpr char fourth_value[] = "fourth_value";
+
+  ASSERT_OK(Put(first_key, first_value));
+  ASSERT_OK(Put(second_key, second_value));
+  ASSERT_OK(Put(third_key, third_value));
+  ASSERT_OK(Put(fourth_key, fourth_value));
+  ASSERT_OK(Flush());
+
+  // Second table+blob file pair: overwrite 2 existing keys
+  constexpr char new_first_value[] = "new_first_value";
+  constexpr char new_second_value[] = "new_second_value";
+
+  ASSERT_OK(Put(first_key, new_first_value));
+  ASSERT_OK(Put(second_key, new_second_value));
+  ASSERT_OK(Flush());
+
+  // Compact them together. The first blob file should have 2 garbage blobs
+  // corresponding to the 2 overwritten keys.
+  constexpr Slice* begin = nullptr;
+  constexpr Slice* end = nullptr;
+
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), begin, end));
+
+  VersionSet* const versions = dbfull()->TEST_GetVersionSet();
+  assert(versions);
+  assert(versions->GetColumnFamilySet());
+
+  ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
+  assert(cfd);
+
+  Version* const current = cfd->current();
+  assert(current);
+
+  const VersionStorageInfo* const storage_info = current->storage_info();
+  assert(storage_info);
+
+  const auto& blob_files = storage_info->GetBlobFiles();
+  ASSERT_EQ(blob_files.size(), 2);
+
+  {
+    auto it = blob_files.begin();
+    const auto& meta = it->second;
+    assert(meta);
+
+    constexpr uint64_t first_expected_bytes =
+        sizeof(first_value) - 1 +
+        BlobLogRecord::CalculateAdjustmentForRecordHeader(sizeof(first_key) -
+                                                          1);
+    constexpr uint64_t second_expected_bytes =
+        sizeof(second_value) - 1 +
+        BlobLogRecord::CalculateAdjustmentForRecordHeader(sizeof(second_key) -
+                                                          1);
+    constexpr uint64_t third_expected_bytes =
+        sizeof(third_value) - 1 +
+        BlobLogRecord::CalculateAdjustmentForRecordHeader(sizeof(third_key) -
+                                                          1);
+    constexpr uint64_t fourth_expected_bytes =
+        sizeof(fourth_value) - 1 +
+        BlobLogRecord::CalculateAdjustmentForRecordHeader(sizeof(fourth_key) -
+                                                          1);
+
+    ASSERT_EQ(meta->GetTotalBlobCount(), 4);
+    ASSERT_EQ(meta->GetTotalBlobBytes(),
+              first_expected_bytes + second_expected_bytes +
+                  third_expected_bytes + fourth_expected_bytes);
+    ASSERT_EQ(meta->GetGarbageBlobCount(), 2);
+    ASSERT_EQ(meta->GetGarbageBlobBytes(),
+              first_expected_bytes + second_expected_bytes);
+  }
+
+  {
+    auto it = blob_files.rbegin();
+    const auto& meta = it->second;
+    assert(meta);
+
+    constexpr uint64_t new_first_expected_bytes =
+        sizeof(new_first_value) - 1 +
+        BlobLogRecord::CalculateAdjustmentForRecordHeader(sizeof(first_key) -
+                                                          1);
+    constexpr uint64_t new_second_expected_bytes =
+        sizeof(new_second_value) - 1 +
+        BlobLogRecord::CalculateAdjustmentForRecordHeader(sizeof(second_key) -
+                                                          1);
+
+    ASSERT_EQ(meta->GetTotalBlobCount(), 2);
+    ASSERT_EQ(meta->GetTotalBlobBytes(),
+              new_first_expected_bytes + new_second_expected_bytes);
+    ASSERT_EQ(meta->GetGarbageBlobCount(), 0);
+    ASSERT_EQ(meta->GetGarbageBlobBytes(), 0);
+  }
+}
+
+TEST_F(DBBlobCompactionTest, MergeBlobWithBase) {
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  options.disable_auto_compactions = true;
+
+  Reopen(options);
+  ASSERT_OK(Put("Key1", "v1_1"));
+  ASSERT_OK(Put("Key2", "v2_1"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(Merge("Key1", "v1_2"));
+  ASSERT_OK(Merge("Key2", "v2_2"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(Merge("Key1", "v1_3"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), /*begin=*/nullptr,
+                              /*end=*/nullptr));
+  ASSERT_EQ(Get("Key1"), "v1_1,v1_2,v1_3");
+  ASSERT_EQ(Get("Key2"), "v2_1,v2_2");
+  Close();
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/blob/db_blob_corruption_test.cc

@@ -12,7 +12,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBBlobCorruptionTest : public DBTestBase {
  protected:
   DBBlobCorruptionTest()
-      : DBTestBase("/db_blob_corruption_test", /* env_do_fsync */ false) {}
+      : DBTestBase("db_blob_corruption_test", /* env_do_fsync */ false) {}
 
   void Corrupt(FileType filetype, int offset, int bytes_to_corrupt) {
     // Pick file to corrupt

db/blob/db_blob_index_test.cc

@@ -41,8 +41,7 @@ class DBBlobIndexTest : public DBTestBase {
                                        Tier::kImmutableMemtables,
                                        Tier::kL0SstFile, Tier::kLnSstFile};
 
-  DBBlobIndexTest()
-      : DBTestBase("/db_blob_index_test", /*env_do_fsync=*/true) {}
+  DBBlobIndexTest() : DBTestBase("db_blob_index_test", /*env_do_fsync=*/true) {}
 
   ColumnFamilyHandle* cfh() { return dbfull()->DefaultColumnFamily(); }
 

db/builder.cc

@@ -165,7 +165,7 @@ Status BuildTable(
           std::move(file), fname, file_options, ioptions.clock, io_tracer,
           ioptions.stats, ioptions.listeners,
           ioptions.file_checksum_gen_factory.get(),
-          tmp_set.Contains(FileType::kTableFile)));
+          tmp_set.Contains(FileType::kTableFile), false));
 
       builder = NewTableBuilder(tboptions, file_writer.get());
     }
@@ -204,6 +204,8 @@ Status BuildTable(
       const Slice& value = c_iter.value();
       const ParsedInternalKey& ikey = c_iter.ikey();
       // Generate a rolling 64-bit hash of the key and values
+      // Note :
+      // Here "key" integrates 'sequence_number'+'kType'+'user key'.
       s = output_validator.Add(key, value);
       if (!s.ok()) {
         break;

db/c.cc

@@ -88,6 +88,7 @@ using ROCKSDB_NAMESPACE::NewBloomFilterPolicy;
 using ROCKSDB_NAMESPACE::NewCompactOnDeletionCollectorFactory;
 using ROCKSDB_NAMESPACE::NewGenericRateLimiter;
 using ROCKSDB_NAMESPACE::NewLRUCache;
+using ROCKSDB_NAMESPACE::NewRibbonFilterPolicy;
 using ROCKSDB_NAMESPACE::OptimisticTransactionDB;
 using ROCKSDB_NAMESPACE::OptimisticTransactionOptions;
 using ROCKSDB_NAMESPACE::Options;
@@ -3028,6 +3029,11 @@ unsigned char rocksdb_options_get_advise_random_on_open(
   return opt->rep.advise_random_on_open;
 }
 
+void rocksdb_options_set_experimental_mempurge_threshold(rocksdb_options_t* opt,
+                                                         double v) {
+  opt->rep.experimental_mempurge_threshold = v;
+}
+
 void rocksdb_options_set_access_hint_on_compaction_start(
     rocksdb_options_t* opt, int v) {
   switch(v) {
@@ -3834,7 +3840,8 @@ void rocksdb_filterpolicy_destroy(rocksdb_filterpolicy_t* filter) {
   delete filter;
 }
 
-rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom_format(int bits_per_key, bool original_format) {
+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom_format(
+    double bits_per_key, bool original_format) {
   // Make a rocksdb_filterpolicy_t, but override all of its methods so
   // they delegate to a NewBloomFilterPolicy() instead of user
   // supplied C functions.
@@ -3869,14 +3876,63 @@ rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom_format(int bits_per_ke
   return wrapper;
 }
 
-rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom_full(int bits_per_key) {
+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom_full(
+    double bits_per_key) {
   return rocksdb_filterpolicy_create_bloom_format(bits_per_key, false);
 }
 
-rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom(int bits_per_key) {
+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom(double bits_per_key) {
   return rocksdb_filterpolicy_create_bloom_format(bits_per_key, true);
 }
 
+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_ribbon_format(
+    double bloom_equivalent_bits_per_key, int bloom_before_level) {
+  // Make a rocksdb_filterpolicy_t, but override all of its methods so
+  // they delegate to a NewRibbonFilterPolicy() instead of user
+  // supplied C functions.
+  struct Wrapper : public rocksdb_filterpolicy_t {
+    const FilterPolicy* rep_;
+    ~Wrapper() override { delete rep_; }
+    const char* Name() const override { return rep_->Name(); }
+    void CreateFilter(const Slice* keys, int n,
+                      std::string* dst) const override {
+      return rep_->CreateFilter(keys, n, dst);
+    }
+    bool KeyMayMatch(const Slice& key, const Slice& filter) const override {
+      return rep_->KeyMayMatch(key, filter);
+    }
+    ROCKSDB_NAMESPACE::FilterBitsBuilder* GetBuilderWithContext(
+        const ROCKSDB_NAMESPACE::FilterBuildingContext& context)
+        const override {
+      return rep_->GetBuilderWithContext(context);
+    }
+    ROCKSDB_NAMESPACE::FilterBitsReader* GetFilterBitsReader(
+        const Slice& contents) const override {
+      return rep_->GetFilterBitsReader(contents);
+    }
+    static void DoNothing(void*) {}
+  };
+  Wrapper* wrapper = new Wrapper;
+  wrapper->rep_ =
+      NewRibbonFilterPolicy(bloom_equivalent_bits_per_key, bloom_before_level);
+  wrapper->state_ = nullptr;
+  wrapper->delete_filter_ = nullptr;
+  wrapper->destructor_ = &Wrapper::DoNothing;
+  return wrapper;
+}
+
+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_ribbon(
+    double bloom_equivalent_bits_per_key) {
+  return rocksdb_filterpolicy_create_ribbon_format(
+      bloom_equivalent_bits_per_key, /*bloom_before_level = disabled*/ -1);
+}
+
+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_ribbon_hybrid(
+    double bloom_equivalent_bits_per_key, int bloom_before_level) {
+  return rocksdb_filterpolicy_create_ribbon_format(
+      bloom_equivalent_bits_per_key, bloom_before_level);
+}
+
 rocksdb_mergeoperator_t* rocksdb_mergeoperator_create(
     void* state, void (*destructor)(void*),
     char* (*full_merge)(void*, const char* key, size_t key_length,
@@ -4828,6 +4884,27 @@ void rocksdb_optimistictransaction_options_set_set_snapshot(
   opt->rep.set_snapshot = v;
 }
 
+char* rocksdb_optimistictransactiondb_property_value(
+    rocksdb_optimistictransactiondb_t* db, const char* propname) {
+  std::string tmp;
+  if (db->rep->GetProperty(Slice(propname), &tmp)) {
+    // We use strdup() since we expect human readable output.
+    return strdup(tmp.c_str());
+  } else {
+    return nullptr;
+  }
+}
+
+int rocksdb_optimistictransactiondb_property_int(
+    rocksdb_optimistictransactiondb_t* db, const char* propname,
+    uint64_t* out_val) {
+  if (db->rep->GetIntProperty(Slice(propname), out_val)) {
+    return 0;
+  } else {
+    return -1;
+  }
+}
+
 rocksdb_column_family_handle_t* rocksdb_transactiondb_create_column_family(
     rocksdb_transactiondb_t* txn_db,
     const rocksdb_options_t* column_family_options,
@@ -4898,6 +4975,27 @@ void rocksdb_transactiondb_release_snapshot(
   delete snapshot;
 }
 
+char* rocksdb_transactiondb_property_value(rocksdb_transactiondb_t* db,
+                                           const char* propname) {
+  std::string tmp;
+  if (db->rep->GetProperty(Slice(propname), &tmp)) {
+    // We use strdup() since we expect human readable output.
+    return strdup(tmp.c_str());
+  } else {
+    return nullptr;
+  }
+}
+
+int rocksdb_transactiondb_property_int(rocksdb_transactiondb_t* db,
+                                       const char* propname,
+                                       uint64_t* out_val) {
+  if (db->rep->GetIntProperty(Slice(propname), out_val)) {
+    return 0;
+  } else {
+    return -1;
+  }
+}
+
 rocksdb_transaction_t* rocksdb_transaction_begin(
     rocksdb_transactiondb_t* txn_db,
     const rocksdb_writeoptions_t* write_options,
@@ -5300,12 +5398,31 @@ rocksdb_transaction_t* rocksdb_optimistictransaction_begin(
   return old_txn;
 }
 
+// Write batch into OptimisticTransactionDB
+void rocksdb_optimistictransactiondb_write(
+    rocksdb_optimistictransactiondb_t* otxn_db,
+    const rocksdb_writeoptions_t* options, rocksdb_writebatch_t* batch,
+    char** errptr) {
+  SaveError(errptr, otxn_db->rep->Write(options->rep, &batch->rep));
+}
+
 void rocksdb_optimistictransactiondb_close(
     rocksdb_optimistictransactiondb_t* otxn_db) {
   delete otxn_db->rep;
   delete otxn_db;
 }
 
+rocksdb_checkpoint_t* rocksdb_optimistictransactiondb_checkpoint_object_create(
+    rocksdb_optimistictransactiondb_t* otxn_db, char** errptr) {
+  Checkpoint* checkpoint;
+  if (SaveError(errptr, Checkpoint::Create(otxn_db->rep, &checkpoint))) {
+    return nullptr;
+  }
+  rocksdb_checkpoint_t* result = new rocksdb_checkpoint_t;
+  result->rep = checkpoint;
+  return result;
+}
+
 void rocksdb_free(void* ptr) { free(ptr); }
 
 rocksdb_pinnableslice_t* rocksdb_get_pinned(

db/c_test.c

@@ -1043,19 +1043,25 @@ int main(int argc, char** argv) {
   }
 
   StartPhase("filter");
-  for (run = 0; run <= 2; run++) {
-    // First run uses custom filter
-    // Second run uses old block-based bloom filter
-    // Third run uses full bloom filter
+  for (run = 0; run <= 4; run++) {
+    // run=0 uses custom filter
+    // run=1 uses old block-based bloom filter
+    // run=2 run uses full bloom filter
+    // run=3 uses Ribbon
+    // run=4 uses Ribbon-Bloom hybrid configuration
     CheckNoError(err);
     rocksdb_filterpolicy_t* policy;
     if (run == 0) {
       policy = rocksdb_filterpolicy_create(NULL, FilterDestroy, FilterCreate,
                                            FilterKeyMatch, NULL, FilterName);
     } else if (run == 1) {
-      policy = rocksdb_filterpolicy_create_bloom(8);
+      policy = rocksdb_filterpolicy_create_bloom(8.0);
+    } else if (run == 2) {
+      policy = rocksdb_filterpolicy_create_bloom_full(8.0);
+    } else if (run == 3) {
+      policy = rocksdb_filterpolicy_create_ribbon(8.0);
     } else {
-      policy = rocksdb_filterpolicy_create_bloom_full(8);
+      policy = rocksdb_filterpolicy_create_ribbon_hybrid(8.0, 1);
     }
     rocksdb_block_based_options_set_filter_policy(table_options, policy);
 
@@ -1121,9 +1127,12 @@ int main(int argc, char** argv) {
       } else if (run == 1) {
         // Essentially a fingerprint of the block-based Bloom schema
         CheckCondition(hits == 241);
-      } else {
+      } else if (run == 2 || run == 4) {
         // Essentially a fingerprint of full Bloom schema, format_version=5
         CheckCondition(hits == 188);
+      } else {
+        // Essentially a fingerprint of Ribbon schema
+        CheckCondition(hits == 226);
       }
       CheckCondition(
           (keys_to_query - hits) ==

db/column_family.cc

@@ -445,6 +445,9 @@ bool SuperVersion::Unref() {
 
 void SuperVersion::Cleanup() {
   assert(refs.load(std::memory_order_relaxed) == 0);
+  // Since this SuperVersion object is being deleted,
+  // decrement reference to the immutable MemtableList
+  // this SV object was pointing to.
   imm->Unref(&to_delete);
   MemTable* m = mem->Unref();
   if (m != nullptr) {
@@ -454,7 +457,7 @@ void SuperVersion::Cleanup() {
     to_delete.push_back(m);
   }
   current->Unref();
-  cfd->UnrefAndTryDelete(this);
+  cfd->UnrefAndTryDelete();
 }
 
 void SuperVersion::Init(ColumnFamilyData* new_cfd, MemTable* new_mem,
@@ -472,10 +475,10 @@ void SuperVersion::Init(ColumnFamilyData* new_cfd, MemTable* new_mem,
 
 namespace {
 void SuperVersionUnrefHandle(void* ptr) {
-  // UnrefHandle is called when a thread exists or a ThreadLocalPtr gets
-  // destroyed. When former happens, the thread shouldn't see kSVInUse.
-  // When latter happens, we are in ~ColumnFamilyData(), no get should happen as
-  // well.
+  // UnrefHandle is called when a thread exits or a ThreadLocalPtr gets
+  // destroyed. When the former happens, the thread shouldn't see kSVInUse.
+  // When the latter happens, only super_version_ holds a reference
+  // to ColumnFamilyData, so no further queries are possible.
   SuperVersion* sv = static_cast<SuperVersion*>(ptr);
   bool was_last_ref __attribute__((__unused__));
   was_last_ref = sv->Unref();
@@ -502,7 +505,7 @@ ColumnFamilyData::ColumnFamilyData(
     uint32_t id, const std::string& name, Version* _dummy_versions,
     Cache* _table_cache, WriteBufferManager* write_buffer_manager,
     const ColumnFamilyOptions& cf_options, const ImmutableDBOptions& db_options,
-    const FileOptions& file_options, ColumnFamilySet* column_family_set,
+    const FileOptions* file_options, ColumnFamilySet* column_family_set,
     BlockCacheTracer* const block_cache_tracer,
     const std::shared_ptr<IOTracer>& io_tracer,
     const std::string& db_session_id)
@@ -665,7 +668,7 @@ ColumnFamilyData::~ColumnFamilyData() {
   }
 }
 
-bool ColumnFamilyData::UnrefAndTryDelete(SuperVersion* sv_under_cleanup) {
+bool ColumnFamilyData::UnrefAndTryDelete() {
   int old_refs = refs_.fetch_sub(1);
   assert(old_refs > 0);
 
@@ -675,22 +678,17 @@ bool ColumnFamilyData::UnrefAndTryDelete(SuperVersion* sv_under_cleanup) {
     return true;
   }
 
-  // If called under SuperVersion::Cleanup, we should not re-enter Cleanup on
-  // the same SuperVersion. (But while installing a new SuperVersion, this
-  // cfd could be referenced only by two SuperVersions.)
-  if (old_refs == 2 && super_version_ != nullptr &&
-      super_version_ != sv_under_cleanup) {
+  if (old_refs == 2 && super_version_ != nullptr) {
     // Only the super_version_ holds me
     SuperVersion* sv = super_version_;
     super_version_ = nullptr;
-    // Release SuperVersion reference kept in ThreadLocalPtr.
-    // This must be done outside of mutex_ since unref handler can lock mutex.
-    sv->db_mutex->Unlock();
+
+    // Release SuperVersion references kept in ThreadLocalPtr.
     local_sv_.reset();
-    sv->db_mutex->Lock();
 
     if (sv->Unref()) {
-      // May delete this ColumnFamilyData after calling Cleanup()
+      // Note: sv will delete this ColumnFamilyData during Cleanup()
+      assert(sv->cfd == this);
       sv->Cleanup();
       delete sv;
       return true;
@@ -1255,14 +1253,13 @@ bool ColumnFamilyData::ReturnThreadLocalSuperVersion(SuperVersion* sv) {
 void ColumnFamilyData::InstallSuperVersion(
     SuperVersionContext* sv_context, InstrumentedMutex* db_mutex) {
   db_mutex->AssertHeld();
-  return InstallSuperVersion(sv_context, db_mutex, mutable_cf_options_);
+  return InstallSuperVersion(sv_context, mutable_cf_options_);
 }
 
 void ColumnFamilyData::InstallSuperVersion(
-    SuperVersionContext* sv_context, InstrumentedMutex* db_mutex,
+    SuperVersionContext* sv_context,
     const MutableCFOptions& mutable_cf_options) {
   SuperVersion* new_superversion = sv_context->new_superversion.release();
-  new_superversion->db_mutex = db_mutex;
   new_superversion->mutable_cf_options = mutable_cf_options;
   new_superversion->Init(this, mem_, imm_.current(), current_);
   SuperVersion* old_superversion = super_version_;
@@ -1456,14 +1453,14 @@ ColumnFamilySet::ColumnFamilySet(const std::string& dbname,
                                  const std::shared_ptr<IOTracer>& io_tracer,
                                  const std::string& db_session_id)
     : max_column_family_(0),
+      file_options_(file_options),
       dummy_cfd_(new ColumnFamilyData(
           ColumnFamilyData::kDummyColumnFamilyDataId, "", nullptr, nullptr,
-          nullptr, ColumnFamilyOptions(), *db_options, file_options, nullptr,
+          nullptr, ColumnFamilyOptions(), *db_options, &file_options_, nullptr,
           block_cache_tracer, io_tracer, db_session_id)),
       default_cfd_cache_(nullptr),
       db_name_(dbname),
       db_options_(db_options),
-      file_options_(file_options),
       table_cache_(table_cache),
       write_buffer_manager_(_write_buffer_manager),
       write_controller_(_write_controller),
@@ -1535,7 +1532,7 @@ ColumnFamilyData* ColumnFamilySet::CreateColumnFamily(
   assert(column_families_.find(name) == column_families_.end());
   ColumnFamilyData* new_cfd = new ColumnFamilyData(
       id, name, dummy_versions, table_cache_, write_buffer_manager_, options,
-      *db_options_, file_options_, this, block_cache_tracer_, io_tracer_,
+      *db_options_, &file_options_, this, block_cache_tracer_, io_tracer_,
       db_session_id_);
   column_families_.insert({name, id});
   column_family_data_.insert({id, new_cfd});

db/column_family.h

@@ -208,8 +208,6 @@ struct SuperVersion {
   uint64_t version_number;
   WriteStallCondition write_stall_condition;
 
-  InstrumentedMutex* db_mutex;
-
   // should be called outside the mutex
   SuperVersion() = default;
   ~SuperVersion();
@@ -281,8 +279,7 @@ class ColumnFamilyData {
   // UnrefAndTryDelete() decreases the reference count and do free if needed,
   // return true if this is freed else false, UnrefAndTryDelete() can only
   // be called while holding a DB mutex, or during single-threaded recovery.
-  // sv_under_cleanup is only provided when called from SuperVersion::Cleanup.
-  bool UnrefAndTryDelete(SuperVersion* sv_under_cleanup = nullptr);
+  bool UnrefAndTryDelete();
 
   // SetDropped() can only be called under following conditions:
   // 1) Holding a DB mutex,
@@ -453,7 +450,6 @@ class ColumnFamilyData {
   // the clients to allocate SuperVersion outside of mutex.
   // IMPORTANT: Only call this from DBImpl::InstallSuperVersion()
   void InstallSuperVersion(SuperVersionContext* sv_context,
-                           InstrumentedMutex* db_mutex,
                            const MutableCFOptions& mutable_cf_options);
   void InstallSuperVersion(SuperVersionContext* sv_context,
                            InstrumentedMutex* db_mutex);
@@ -520,6 +516,7 @@ class ColumnFamilyData {
   }
 
   ThreadLocalPtr* TEST_GetLocalSV() { return local_sv_.get(); }
+  WriteBufferManager* write_buffer_mgr() { return write_buffer_manager_; }
 
  private:
   friend class ColumnFamilySet;
@@ -529,7 +526,7 @@ class ColumnFamilyData {
                    WriteBufferManager* write_buffer_manager,
                    const ColumnFamilyOptions& options,
                    const ImmutableDBOptions& db_options,
-                   const FileOptions& file_options,
+                   const FileOptions* file_options,
                    ColumnFamilySet* column_family_set,
                    BlockCacheTracer* const block_cache_tracer,
                    const std::shared_ptr<IOTracer>& io_tracer,
@@ -720,6 +717,8 @@ class ColumnFamilySet {
   std::unordered_map<uint32_t, ColumnFamilyData*> column_family_data_;
 
   uint32_t max_column_family_;
+  const FileOptions file_options_;
+
   ColumnFamilyData* dummy_cfd_;
   // We don't hold the refcount here, since default column family always exists
   // We are also not responsible for cleaning up default_cfd_cache_. This is
@@ -729,7 +728,6 @@ class ColumnFamilySet {
 
   const std::string db_name_;
   const ImmutableDBOptions* const db_options_;
-  const FileOptions file_options_;
   Cache* table_cache_;
   WriteBufferManager* write_buffer_manager_;
   WriteController* write_controller_;

db/compaction/compaction.cc

@@ -211,9 +211,9 @@ Compaction::Compaction(
     std::vector<CompactionInputFiles> _inputs, int _output_level,
     uint64_t _target_file_size, uint64_t _max_compaction_bytes,
     uint32_t _output_path_id, CompressionType _compression,
-    CompressionOptions _compression_opts, uint32_t _max_subcompactions,
-    std::vector<FileMetaData*> _grandparents, bool _manual_compaction,
-    double _score, bool _deletion_compaction,
+    CompressionOptions _compression_opts, Temperature _output_temperature,
+    uint32_t _max_subcompactions, std::vector<FileMetaData*> _grandparents,
+    bool _manual_compaction, double _score, bool _deletion_compaction,
     CompactionReason _compaction_reason)
     : input_vstorage_(vstorage),
       start_level_(_inputs[0].level),
@@ -229,6 +229,7 @@ Compaction::Compaction(
       output_path_id_(_output_path_id),
       output_compression_(_compression),
       output_compression_opts_(_compression_opts),
+      output_temperature_(_output_temperature),
       deletion_compaction_(_deletion_compaction),
       inputs_(PopulateWithAtomicBoundaries(vstorage, std::move(_inputs))),
       grandparents_(std::move(_grandparents)),
@@ -237,7 +238,8 @@ Compaction::Compaction(
       is_full_compaction_(IsFullCompaction(vstorage, inputs_)),
       is_manual_compaction_(_manual_compaction),
       is_trivial_move_(false),
-      compaction_reason_(_compaction_reason) {
+      compaction_reason_(_compaction_reason),
+      notify_on_compaction_completion_(false) {
   MarkFilesBeingCompacted(true);
   if (is_manual_compaction_) {
     compaction_reason_ = CompactionReason::kManualCompaction;
@@ -307,6 +309,12 @@ bool Compaction::IsTrivialMove() const {
     return false;
   }
 
+  if (start_level_ == output_level_) {
+    // It doesn't make sense if compaction picker picks files just to trivial
+    // move to the same level.
+    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) &&
@@ -581,6 +589,29 @@ bool Compaction::ShouldFormSubcompactions() const {
   }
 }
 
+bool Compaction::DoesInputReferenceBlobFiles() const {
+  assert(input_version_);
+
+  const VersionStorageInfo* storage_info = input_version_->storage_info();
+  assert(storage_info);
+
+  if (storage_info->GetBlobFiles().empty()) {
+    return false;
+  }
+
+  for (size_t i = 0; i < inputs_.size(); ++i) {
+    for (const FileMetaData* meta : inputs_[i].files) {
+      assert(meta);
+
+      if (meta->oldest_blob_file_number != kInvalidBlobFileNumber) {
+        return true;
+      }
+    }
+  }
+
+  return false;
+}
+
 uint64_t Compaction::MinInputFileOldestAncesterTime() const {
   uint64_t min_oldest_ancester_time = port::kMaxUint64;
   for (const auto& level_files : inputs_) {

db/compaction/compaction.h

@@ -76,7 +76,8 @@ class Compaction {
              std::vector<CompactionInputFiles> inputs, int output_level,
              uint64_t target_file_size, uint64_t max_compaction_bytes,
              uint32_t output_path_id, CompressionType compression,
-             CompressionOptions compression_opts, uint32_t max_subcompactions,
+             CompressionOptions compression_opts,
+             Temperature output_temperature, uint32_t max_subcompactions,
              std::vector<FileMetaData*> grandparents,
              bool manual_compaction = false, double score = -1,
              bool deletion_compaction = false,
@@ -266,6 +267,11 @@ class Compaction {
   // Should this compaction be broken up into smaller ones run in parallel?
   bool ShouldFormSubcompactions() const;
 
+  // Returns true iff at least one input file references a blob file.
+  //
+  // PRE: input version has been set.
+  bool DoesInputReferenceBlobFiles() const;
+
   // test function to validate the functionality of IsBottommostLevel()
   // function -- determines if compaction with inputs and storage is bottommost
   static bool TEST_IsBottommostLevel(
@@ -294,10 +300,22 @@ class Compaction {
 
   uint64_t max_compaction_bytes() const { return max_compaction_bytes_; }
 
+  Temperature output_temperature() const { return output_temperature_; }
+
   uint32_t max_subcompactions() const { return max_subcompactions_; }
 
   uint64_t MinInputFileOldestAncesterTime() const;
 
+  // Called by DBImpl::NotifyOnCompactionCompleted to make sure number of
+  // compaction begin and compaction completion callbacks match.
+  void SetNotifyOnCompactionCompleted() {
+    notify_on_compaction_completion_ = true;
+  }
+
+  bool ShouldNotifyOnCompactionCompleted() const {
+    return notify_on_compaction_completion_;
+  }
+
  private:
   // mark (or clear) all files that are being compacted
   void MarkFilesBeingCompacted(bool mark_as_compacted);
@@ -341,6 +359,7 @@ class Compaction {
   const uint32_t output_path_id_;
   CompressionType output_compression_;
   CompressionOptions output_compression_opts_;
+  Temperature output_temperature_;
   // If true, then the compaction can be done by simply deleting input files.
   const bool deletion_compaction_;
 
@@ -381,6 +400,10 @@ class Compaction {
 
   // Reason for compaction
   CompactionReason compaction_reason_;
+
+  // Notify on compaction completion only if listener was notified on compaction
+  // begin.
+  bool notify_on_compaction_completion_;
 };
 
 // Return sum of sizes of all files in `files`.

db/compaction/compaction_iteration_stats.h

@@ -38,4 +38,6 @@ struct CompactionIterationStats {
   // Blob related statistics
   uint64_t num_blobs_read = 0;
   uint64_t total_blob_bytes_read = 0;
+  uint64_t num_blobs_relocated = 0;
+  uint64_t total_blob_bytes_relocated = 0;
 };

db/compaction/compaction_iterator.cc

@@ -16,22 +16,6 @@
 #include "table/internal_iterator.h"
 #include "test_util/sync_point.h"
 
-#define DEFINITELY_IN_SNAPSHOT(seq, snapshot)                       \
-  ((seq) <= (snapshot) &&                                           \
-   (snapshot_checker_ == nullptr ||                                 \
-    LIKELY(snapshot_checker_->CheckInSnapshot((seq), (snapshot)) == \
-           SnapshotCheckerResult::kInSnapshot)))
-
-#define DEFINITELY_NOT_IN_SNAPSHOT(seq, snapshot)                     \
-  ((seq) > (snapshot) ||                                              \
-   (snapshot_checker_ != nullptr &&                                   \
-    UNLIKELY(snapshot_checker_->CheckInSnapshot((seq), (snapshot)) == \
-             SnapshotCheckerResult::kNotInSnapshot)))
-
-#define IN_EARLIEST_SNAPSHOT(seq) \
-  ((seq) <= earliest_snapshot_ && \
-   (snapshot_checker_ == nullptr || LIKELY(IsInEarliestSnapshot(seq))))
-
 namespace ROCKSDB_NAMESPACE {
 CompactionIterator::CompactionIterator(
     InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
@@ -75,10 +59,8 @@ CompactionIterator::CompactionIterator(
     const std::atomic<bool>* manual_compaction_canceled,
     const std::shared_ptr<Logger> info_log,
     const std::string* full_history_ts_low)
-    : input_(
-          input, cmp,
-          compaction ==
-              nullptr),  // Now only need to count number of entries in flush.
+    : input_(input, cmp,
+             !compaction || compaction->DoesInputReferenceBlobFiles()),
       cmp_(cmp),
       merge_helper_(merge_helper),
       snapshots_(snapshots),
@@ -513,11 +495,11 @@ void CompactionIterator::NextFromInput() {
                         "Unexpected key type %d for compaction output",
                         ikey_.type);
       }
-      assert(current_user_key_snapshot_ == last_snapshot);
-      if (current_user_key_snapshot_ != last_snapshot) {
+      assert(current_user_key_snapshot_ >= last_snapshot);
+      if (current_user_key_snapshot_ < last_snapshot) {
         ROCKS_LOG_FATAL(info_log_,
                         "current_user_key_snapshot_ (%" PRIu64
-                        ") != last_snapshot (%" PRIu64 ")",
+                        ") < last_snapshot (%" PRIu64 ")",
                         current_user_key_snapshot_, last_snapshot);
       }
 
@@ -573,10 +555,20 @@ void CompactionIterator::NextFromInput() {
           ParseInternalKey(input_.key(), &next_ikey, allow_data_in_errors_)
               .ok() &&
           cmp_->Equal(ikey_.user_key, next_ikey.user_key)) {
+#ifndef NDEBUG
+        const Compaction* c =
+            compaction_ ? compaction_->real_compaction() : nullptr;
+#endif
+        TEST_SYNC_POINT_CALLBACK(
+            "CompactionIterator::NextFromInput:SingleDelete:1",
+            const_cast<Compaction*>(c));
+
         // Check whether the next key belongs to the same snapshot as the
         // SingleDelete.
         if (prev_snapshot == 0 ||
-            DEFINITELY_NOT_IN_SNAPSHOT(next_ikey.sequence, prev_snapshot)) {
+            DefinitelyNotInSnapshot(next_ikey.sequence, prev_snapshot)) {
+          TEST_SYNC_POINT_CALLBACK(
+              "CompactionIterator::NextFromInput:SingleDelete:2", nullptr);
           if (next_ikey.type == kTypeSingleDeletion) {
             // We encountered two SingleDeletes in a row.  This could be due to
             // unexpected user input.
@@ -588,8 +580,8 @@ void CompactionIterator::NextFromInput() {
             ++iter_stats_.num_record_drop_obsolete;
             ++iter_stats_.num_single_del_mismatch;
           } else if (has_outputted_key_ ||
-                     DEFINITELY_IN_SNAPSHOT(
-                         ikey_.sequence, earliest_write_conflict_snapshot_)) {
+                     DefinitelyInSnapshot(ikey_.sequence,
+                                          earliest_write_conflict_snapshot_)) {
             // Found a matching value, we can drop the single delete and the
             // value.  It is safe to drop both records since we've already
             // outputted a key in this snapshot, or there is no earlier
@@ -622,6 +614,9 @@ void CompactionIterator::NextFromInput() {
             // Set up the Put to be outputted in the next iteration.
             // (Optimization 3).
             clear_and_output_next_key_ = true;
+            TEST_SYNC_POINT_CALLBACK(
+                "CompactionIterator::NextFromInput:KeepSDForWW",
+                /*arg=*/nullptr);
           }
         } else {
           // We hit the next snapshot without hitting a put, so the iterator
@@ -637,7 +632,8 @@ void CompactionIterator::NextFromInput() {
         // iteration. If the next key is corrupt, we return before the
         // comparison, so the value of has_current_user_key does not matter.
         has_current_user_key_ = false;
-        if (compaction_ != nullptr && IN_EARLIEST_SNAPSHOT(ikey_.sequence) &&
+        if (compaction_ != nullptr &&
+            DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
             compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
                                                        &level_ptrs_)) {
           // Key doesn't exist outside of this range.
@@ -681,7 +677,7 @@ void CompactionIterator::NextFromInput() {
                (ikey_.type == kTypeDeletion ||
                 (ikey_.type == kTypeDeletionWithTimestamp &&
                  cmp_with_history_ts_low_ < 0)) &&
-               IN_EARLIEST_SNAPSHOT(ikey_.sequence) &&
+               DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
                ikeyNotNeededForIncrementalSnapshot() &&
                compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
                                                           &level_ptrs_)) {
@@ -724,6 +720,13 @@ void CompactionIterator::NextFromInput() {
                                  ikey_.user_key, &level_ptrs_));
       ParsedInternalKey next_ikey;
       AdvanceInputIter();
+#ifndef NDEBUG
+      const Compaction* c =
+          compaction_ ? compaction_->real_compaction() : nullptr;
+#endif
+      TEST_SYNC_POINT_CALLBACK(
+          "CompactionIterator::NextFromInput:BottommostDelete:1",
+          const_cast<Compaction*>(c));
       // Skip over all versions of this key that happen to occur in the same
       // snapshot range as the delete.
       //
@@ -736,7 +739,7 @@ void CompactionIterator::NextFromInput() {
                   .ok()) &&
              cmp_->EqualWithoutTimestamp(ikey_.user_key, next_ikey.user_key) &&
              (prev_snapshot == 0 ||
-              DEFINITELY_NOT_IN_SNAPSHOT(next_ikey.sequence, prev_snapshot))) {
+              DefinitelyNotInSnapshot(next_ikey.sequence, prev_snapshot))) {
         AdvanceInputIter();
       }
       // If you find you still need to output a row with this key, we need to output the
@@ -756,13 +759,15 @@ void CompactionIterator::NextFromInput() {
       }
 
       pinned_iters_mgr_.StartPinning();
+      Version* version = compaction_ ? compaction_->input_version() : nullptr;
+
       // We know the merge type entry is not hidden, otherwise we would
       // 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(&input_, range_del_agg_, prev_snapshot,
-                                    bottommost_level_, allow_data_in_errors_);
+      Status s = merge_helper_->MergeUntil(&input_, range_del_agg_,
+                                           prev_snapshot, bottommost_level_,
+                                           allow_data_in_errors_, version);
       merge_out_iter_.SeekToFirst();
 
       if (!s.ok() && !s.IsMergeInProgress()) {
@@ -914,6 +919,9 @@ void CompactionIterator::GarbageCollectBlobIfNeeded() {
     ++iter_stats_.num_blobs_read;
     iter_stats_.total_blob_bytes_read += bytes_read;
 
+    ++iter_stats_.num_blobs_relocated;
+    iter_stats_.total_blob_bytes_relocated += blob_index.size();
+
     value_ = blob_value_;
 
     if (ExtractLargeValueIfNeededImpl()) {
@@ -977,7 +985,8 @@ void CompactionIterator::PrepareOutput() {
     if (valid_ && compaction_ != nullptr &&
         !compaction_->allow_ingest_behind() &&
         ikeyNotNeededForIncrementalSnapshot() && bottommost_level_ &&
-        IN_EARLIEST_SNAPSHOT(ikey_.sequence) && ikey_.type != kTypeMerge) {
+        DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
+        ikey_.type != kTypeMerge) {
       assert(ikey_.type != kTypeDeletion && ikey_.type != kTypeSingleDeletion);
       if (ikey_.type == kTypeDeletion || ikey_.type == kTypeSingleDeletion) {
         ROCKS_LOG_FATAL(info_log_,
@@ -1053,7 +1062,7 @@ inline bool CompactionIterator::ikeyNotNeededForIncrementalSnapshot() {
          (ikey_.sequence < preserve_deletes_seqnum_);
 }
 
-bool CompactionIterator::IsInEarliestSnapshot(SequenceNumber sequence) {
+bool CompactionIterator::IsInCurrentEarliestSnapshot(SequenceNumber sequence) {
   assert(snapshot_checker_ != nullptr);
   bool pre_condition = (earliest_snapshot_ == kMaxSequenceNumber ||
                         (earliest_snapshot_iter_ != snapshots_->end() &&

db/compaction/compaction_iterator.h

@@ -97,6 +97,10 @@ class CompactionIterator {
     virtual double blob_garbage_collection_age_cutoff() const = 0;
 
     virtual Version* input_version() const = 0;
+
+    virtual bool DoesInputReferenceBlobFiles() const = 0;
+
+    virtual const Compaction* real_compaction() const = 0;
   };
 
   class RealCompaction : public CompactionProxy {
@@ -146,6 +150,12 @@ class CompactionIterator {
       return compaction_->input_version();
     }
 
+    bool DoesInputReferenceBlobFiles() const override {
+      return compaction_->DoesInputReferenceBlobFiles();
+    }
+
+    const Compaction* real_compaction() const override { return compaction_; }
+
    private:
     const Compaction* compaction_;
   };
@@ -261,7 +271,13 @@ class CompactionIterator {
                SnapshotCheckerResult::kInSnapshot;
   }
 
-  bool IsInEarliestSnapshot(SequenceNumber sequence);
+  bool IsInCurrentEarliestSnapshot(SequenceNumber sequence);
+
+  bool DefinitelyInSnapshot(SequenceNumber seq, SequenceNumber snapshot);
+
+  bool DefinitelyNotInSnapshot(SequenceNumber seq, SequenceNumber snapshot);
+
+  bool InCurrentEarliestSnapshot(SequenceNumber seq);
 
   // Extract user-defined timestamp from user key if possible and compare it
   // with *full_history_ts_low_ if applicable.
@@ -406,4 +422,27 @@ class CompactionIterator {
             manual_compaction_canceled_->load(std::memory_order_relaxed));
   }
 };
+
+inline bool CompactionIterator::DefinitelyInSnapshot(SequenceNumber seq,
+                                                     SequenceNumber snapshot) {
+  return ((seq) <= (snapshot) &&
+          (snapshot_checker_ == nullptr ||
+           LIKELY(snapshot_checker_->CheckInSnapshot((seq), (snapshot)) ==
+                  SnapshotCheckerResult::kInSnapshot)));
+}
+
+inline bool CompactionIterator::DefinitelyNotInSnapshot(
+    SequenceNumber seq, SequenceNumber snapshot) {
+  return ((seq) > (snapshot) ||
+          (snapshot_checker_ != nullptr &&
+           UNLIKELY(snapshot_checker_->CheckInSnapshot((seq), (snapshot)) ==
+                    SnapshotCheckerResult::kNotInSnapshot)));
+}
+
+inline bool CompactionIterator::InCurrentEarliestSnapshot(SequenceNumber seq) {
+  return ((seq) <= earliest_snapshot_ &&
+          (snapshot_checker_ == nullptr ||
+           LIKELY(IsInCurrentEarliestSnapshot(seq))));
+}
+
 }  // namespace ROCKSDB_NAMESPACE

db/compaction/compaction_iterator_test.cc

@@ -182,6 +182,10 @@ class FakeCompaction : public CompactionIterator::CompactionProxy {
 
   Version* input_version() const override { return nullptr; }
 
+  bool DoesInputReferenceBlobFiles() const override { return false; }
+
+  const Compaction* real_compaction() const override { return nullptr; }
+
   bool key_not_exists_beyond_output_level = false;
 
   bool is_bottommost_level = false;

db/compaction/compaction_job.cc

@@ -20,8 +20,10 @@
 #include <utility>
 #include <vector>
 
+#include "db/blob/blob_counting_iterator.h"
 #include "db/blob/blob_file_addition.h"
 #include "db/blob/blob_file_builder.h"
+#include "db/blob/blob_garbage_meter.h"
 #include "db/builder.h"
 #include "db/compaction/clipping_iterator.h"
 #include "db/db_impl/db_impl.h"
@@ -105,6 +107,8 @@ const char* GetCompactionReasonString(CompactionReason compaction_reason) {
       return "ExternalSstIngestion";
     case CompactionReason::kPeriodicCompaction:
       return "PeriodicCompaction";
+    case CompactionReason::kChangeTemperature:
+      return "ChangeTemperature";
     case CompactionReason::kNumOfReasons:
       // fall through
     default:
@@ -147,6 +151,7 @@ struct CompactionJob::SubcompactionState {
   // State kept for output being generated
   std::vector<Output> outputs;
   std::vector<BlobFileAddition> blob_file_additions;
+  std::unique_ptr<BlobGarbageMeter> blob_garbage_meter;
   std::unique_ptr<WritableFileWriter> outfile;
   std::unique_ptr<TableBuilder> builder;
 
@@ -177,9 +182,17 @@ struct CompactionJob::SubcompactionState {
   uint64_t overlapped_bytes = 0;
   // A flag determine whether the key has been seen in ShouldStopBefore()
   bool seen_key = false;
+  // sub compaction job id, which is used to identify different sub-compaction
+  // within the same compaction job.
+  const uint32_t sub_job_id;
 
-  SubcompactionState(Compaction* c, Slice* _start, Slice* _end, uint64_t size)
-      : compaction(c), start(_start), end(_end), approx_size(size) {
+  SubcompactionState(Compaction* c, Slice* _start, Slice* _end, uint64_t size,
+                     uint32_t _sub_job_id)
+      : compaction(c),
+        start(_start),
+        end(_end),
+        approx_size(size),
+        sub_job_id(_sub_job_id) {
     assert(compaction != nullptr);
   }
 
@@ -229,6 +242,14 @@ struct CompactionJob::SubcompactionState {
 
     return false;
   }
+
+  Status ProcessOutFlowIfNeeded(const Slice& key, const Slice& value) {
+    if (!blob_garbage_meter) {
+      return Status::OK();
+    }
+
+    return blob_garbage_meter->ProcessOutFlow(key, value);
+  }
 };
 
 // Maintains state for the entire compaction
@@ -438,7 +459,8 @@ void CompactionJob::Prepare() {
     for (size_t i = 0; i <= boundaries_.size(); i++) {
       Slice* start = i == 0 ? nullptr : &boundaries_[i - 1];
       Slice* end = i == boundaries_.size() ? nullptr : &boundaries_[i];
-      compact_->sub_compact_states.emplace_back(c, start, end, sizes_[i]);
+      compact_->sub_compact_states.emplace_back(c, start, end, sizes_[i],
+                                                static_cast<uint32_t>(i));
     }
     RecordInHistogram(stats_, NUM_SUBCOMPACTIONS_SCHEDULED,
                       compact_->sub_compact_states.size());
@@ -447,7 +469,8 @@ void CompactionJob::Prepare() {
     constexpr Slice* end = nullptr;
     constexpr uint64_t size = 0;
 
-    compact_->sub_compact_states.emplace_back(c, start, end, size);
+    compact_->sub_compact_states.emplace_back(c, start, end, size,
+                                              /*sub_job_id*/ 0);
   }
 }
 
@@ -858,7 +881,7 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options) {
 
   UpdateCompactionJobStats(stats);
 
-  auto stream = event_logger_->LogToBuffer(log_buffer_);
+  auto stream = event_logger_->LogToBuffer(log_buffer_, 8192);
   stream << "job" << job_id_ << "event"
          << "compaction_finished"
          << "compaction_time_micros" << stats.micros
@@ -962,7 +985,8 @@ void CompactionJob::ProcessKeyValueCompactionWithCompactionService(
       compaction_input.column_family.name.c_str(), job_id_,
       compaction_input.output_level, input_files_oss.str().c_str());
   CompactionServiceJobStatus compaction_status =
-      db_options_.compaction_service->Start(compaction_input_binary, job_id_);
+      db_options_.compaction_service->Start(compaction_input_binary,
+                                            GetCompactionId(sub_compact));
   if (compaction_status != CompactionServiceJobStatus::kSuccess) {
     sub_compact->status =
         Status::Incomplete("CompactionService failed to start compaction job.");
@@ -971,7 +995,7 @@ void CompactionJob::ProcessKeyValueCompactionWithCompactionService(
 
   std::string compaction_result_binary;
   compaction_status = db_options_.compaction_service->WaitForComplete(
-      job_id_, &compaction_result_binary);
+      GetCompactionId(sub_compact), &compaction_result_binary);
 
   CompactionServiceResult compaction_result;
   s = CompactionServiceResult::Read(compaction_result_binary,
@@ -1136,6 +1160,15 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     input = clip.get();
   }
 
+  std::unique_ptr<InternalIterator> blob_counter;
+
+  if (sub_compact->compaction->DoesInputReferenceBlobFiles()) {
+    sub_compact->blob_garbage_meter.reset(new BlobGarbageMeter);
+    blob_counter.reset(
+        new BlobCountingIterator(input, sub_compact->blob_garbage_meter.get()));
+    input = blob_counter.get();
+  }
+
   input->SeekToFirst();
 
   AutoThreadOperationStageUpdater stage_updater(
@@ -1248,6 +1281,11 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       break;
     }
 
+    status = sub_compact->ProcessOutFlowIfNeeded(key, value);
+    if (!status.ok()) {
+      break;
+    }
+
     sub_compact->current_output_file_size =
         sub_compact->builder->EstimatedFileSize();
     const ParsedInternalKey& ikey = c_iter->ikey();
@@ -1331,6 +1369,16 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
 
   RecordTick(stats_, FILTER_OPERATION_TOTAL_TIME,
              c_iter_stats.total_filter_time);
+
+  if (c_iter_stats.num_blobs_relocated > 0) {
+    RecordTick(stats_, BLOB_DB_GC_NUM_KEYS_RELOCATED,
+               c_iter_stats.num_blobs_relocated);
+  }
+  if (c_iter_stats.total_blob_bytes_relocated > 0) {
+    RecordTick(stats_, BLOB_DB_GC_BYTES_RELOCATED,
+               c_iter_stats.total_blob_bytes_relocated);
+  }
+
   RecordDroppedKeys(c_iter_stats, &sub_compact->compaction_job_stats);
   RecordCompactionIOStats();
 
@@ -1415,11 +1463,16 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
 #endif  // ROCKSDB_ASSERT_STATUS_CHECKED
 
   sub_compact->c_iter.reset();
+  blob_counter.reset();
   clip.reset();
   raw_input.reset();
   sub_compact->status = status;
 }
 
+uint64_t CompactionJob::GetCompactionId(SubcompactionState* sub_compact) {
+  return (uint64_t)job_id_ << 32 | sub_compact->sub_job_id;
+}
+
 void CompactionJob::RecordDroppedKeys(
     const CompactionIterationStats& c_iter_stats,
     CompactionJobStats* compaction_job_stats) {
@@ -1784,19 +1837,6 @@ Status CompactionJob::InstallCompactionResults(
   auto* compaction = compact_->compaction;
   assert(compaction);
 
-  // paranoia: verify that the files that we started with
-  // still exist in the current version and in the same original level.
-  // This ensures that a concurrent compaction did not erroneously
-  // pick the same files to compact_.
-  if (!versions_->VerifyCompactionFileConsistency(compaction)) {
-    Compaction::InputLevelSummaryBuffer inputs_summary;
-
-    ROCKS_LOG_ERROR(db_options_.info_log, "[%s] [JOB %d] Compaction %s aborted",
-                    compaction->column_family_data()->GetName().c_str(),
-                    job_id_, compaction->InputLevelSummary(&inputs_summary));
-    return Status::Corruption("Compaction input files inconsistent");
-  }
-
   {
     Compaction::InputLevelSummaryBuffer inputs_summary;
     ROCKS_LOG_INFO(db_options_.info_log,
@@ -1812,6 +1852,8 @@ Status CompactionJob::InstallCompactionResults(
   // Add compaction inputs
   compaction->AddInputDeletions(edit);
 
+  std::unordered_map<uint64_t, BlobGarbageMeter::BlobStats> blob_total_garbage;
+
   for (const auto& sub_compact : compact_->sub_compact_states) {
     for (const auto& out : sub_compact.outputs) {
       edit->AddFile(compaction->output_level(), out.meta);
@@ -1820,6 +1862,29 @@ Status CompactionJob::InstallCompactionResults(
     for (const auto& blob : sub_compact.blob_file_additions) {
       edit->AddBlobFile(blob);
     }
+
+    if (sub_compact.blob_garbage_meter) {
+      const auto& flows = sub_compact.blob_garbage_meter->flows();
+
+      for (const auto& pair : flows) {
+        const uint64_t blob_file_number = pair.first;
+        const BlobGarbageMeter::BlobInOutFlow& flow = pair.second;
+
+        assert(flow.IsValid());
+        if (flow.HasGarbage()) {
+          blob_total_garbage[blob_file_number].Add(flow.GetGarbageCount(),
+                                                   flow.GetGarbageBytes());
+        }
+      }
+    }
+  }
+
+  for (const auto& pair : blob_total_garbage) {
+    const uint64_t blob_file_number = pair.first;
+    const BlobGarbageMeter::BlobStats& stats = pair.second;
+
+    edit->AddBlobFileGarbage(blob_file_number, stats.GetCount(),
+                             stats.GetBytes());
   }
 
   return versions_->LogAndApply(compaction->column_family_data(),
@@ -1874,11 +1939,12 @@ Status CompactionJob::OpenCompactionOutputFile(
 
   // Pass temperature of botommost files to FileSystem.
   FileOptions fo_copy = file_options_;
-  Temperature temperature = Temperature::kUnknown;
-  if (bottommost_level_) {
-    fo_copy.temperature = temperature =
+  Temperature temperature = sub_compact->compaction->output_temperature();
+  if (temperature == Temperature::kUnknown && bottommost_level_) {
+    temperature =
         sub_compact->compaction->mutable_cf_options()->bottommost_temperature;
   }
+  fo_copy.temperature = temperature;
 
   Status s;
   IOStatus io_s = NewWritableFile(fs_.get(), fname, &writable_file, fo_copy);
@@ -1948,7 +2014,7 @@ Status CompactionJob::OpenCompactionOutputFile(
       std::move(writable_file), fname, file_options_, db_options_.clock,
       io_tracer_, db_options_.stats, listeners,
       db_options_.file_checksum_gen_factory.get(),
-      tmp_set.Contains(FileType::kTableFile)));
+      tmp_set.Contains(FileType::kTableFile), false));
 
   TableBuilderOptions tboptions(
       *cfd->ioptions(), *(sub_compact->compaction->mutable_cf_options()),
@@ -2179,8 +2245,8 @@ Status CompactionServiceCompactionJob::Run() {
   Slice end = compaction_input_.end;
   compact_->sub_compact_states.emplace_back(
       c, compaction_input_.has_begin ? &begin : nullptr,
-      compaction_input_.has_end ? &end : nullptr,
-      compaction_input_.approx_size);
+      compaction_input_.has_end ? &end : nullptr, compaction_input_.approx_size,
+      /*sub_job_id*/ 0);
 
   log_buffer_->FlushBufferToLog();
   LogCompaction();
@@ -2219,7 +2285,7 @@ Status CompactionServiceCompactionJob::Run() {
     status = io_s;
   }
   if (status.ok()) {
-    // TODO: Add verify_table() and VerifyCompactionFileConsistency()
+    // TODO: Add verify_table()
   }
 
   // Finish up all book-keeping to unify the subcompaction results

db/compaction/compaction_job.h

@@ -168,7 +168,7 @@ class CompactionJob {
   void UpdateCompactionInputStatsHelper(
       int* num_files, uint64_t* bytes_read, int input_level);
 
-  int job_id_;
+  uint32_t job_id_;
 
   CompactionJobStats* compaction_job_stats_;
 
@@ -219,6 +219,8 @@ class CompactionJob {
   std::string full_history_ts_low_;
   BlobFileCompletionCallback* blob_callback_;
 
+  uint64_t GetCompactionId(SubcompactionState* sub_compact);
+
   // Get table file name in where it's outputting to, which should also be in
   // `output_directory_`.
   virtual std::string GetTableFileName(uint64_t file_number);

db/compaction/compaction_job_test.cc

@@ -334,8 +334,8 @@ class CompactionJobTestBase : public testing::Test {
         cfd->current()->storage_info(), *cfd->ioptions(),
         *cfd->GetLatestMutableCFOptions(), mutable_db_options_,
         compaction_input_files, output_level, 1024 * 1024, 10 * 1024 * 1024, 0,
-        kNoCompression, cfd->GetLatestMutableCFOptions()->compression_opts, 0,
-        {}, true);
+        kNoCompression, cfd->GetLatestMutableCFOptions()->compression_opts,
+        Temperature::kUnknown, 0, {}, true);
     compaction.SetInputVersion(cfd->current());
 
     LogBuffer log_buffer(InfoLogLevel::INFO_LEVEL, db_options_.info_log.get());

db/compaction/compaction_picker.cc

@@ -358,7 +358,7 @@ Compaction* CompactionPicker::CompactFiles(
       output_level, compact_options.output_file_size_limit,
       mutable_cf_options.max_compaction_bytes, output_path_id, compression_type,
       GetCompressionOptions(mutable_cf_options, vstorage, output_level),
-      compact_options.max_subcompactions,
+      Temperature::kUnknown, compact_options.max_subcompactions,
       /* grandparents */ {}, true);
   RegisterCompaction(c);
   return c;
@@ -634,7 +634,8 @@ Compaction* CompactionPicker::CompactRange(
         GetCompressionType(ioptions_, vstorage, mutable_cf_options,
                            output_level, 1),
         GetCompressionOptions(mutable_cf_options, vstorage, output_level),
-        compact_range_options.max_subcompactions, /* grandparents */ {},
+        Temperature::kUnknown, compact_range_options.max_subcompactions,
+        /* grandparents */ {},
         /* is manual */ true);
     RegisterCompaction(c);
     vstorage->ComputeCompactionScore(ioptions_, mutable_cf_options);
@@ -812,7 +813,8 @@ Compaction* CompactionPicker::CompactRange(
       GetCompressionType(ioptions_, vstorage, mutable_cf_options, output_level,
                          vstorage->base_level()),
       GetCompressionOptions(mutable_cf_options, vstorage, output_level),
-      compact_range_options.max_subcompactions, std::move(grandparents),
+      Temperature::kUnknown, compact_range_options.max_subcompactions,
+      std::move(grandparents),
       /* is manual compaction */ true);
 
   TEST_SYNC_POINT_CALLBACK("CompactionPicker::CompactRange:Return", compaction);

db/compaction/compaction_picker_fifo.cc

@@ -111,7 +111,7 @@ Compaction* FIFOCompactionPicker::PickTTLCompaction(
   Compaction* c = new Compaction(
       vstorage, ioptions_, mutable_cf_options, mutable_db_options,
       std::move(inputs), 0, 0, 0, 0, kNoCompression,
-      mutable_cf_options.compression_opts,
+      mutable_cf_options.compression_opts, Temperature::kUnknown,
       /* max_subcompactions */ 0, {}, /* is manual */ false,
       vstorage->CompactionScore(0),
       /* is deletion compaction */ true, CompactionReason::kFIFOTtl);
@@ -154,7 +154,8 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
             {comp_inputs}, 0, 16 * 1024 * 1024 /* output file size limit */,
             0 /* max compaction bytes, not applicable */,
             0 /* output path ID */, mutable_cf_options.compression,
-            mutable_cf_options.compression_opts, 0 /* max_subcompactions */, {},
+            mutable_cf_options.compression_opts, Temperature::kUnknown,
+            0 /* max_subcompactions */, {},
             /* is manual */ false, vstorage->CompactionScore(0),
             /* is deletion compaction */ false,
             CompactionReason::kFIFOReduceNumFiles);
@@ -203,13 +204,119 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
   Compaction* c = new Compaction(
       vstorage, ioptions_, mutable_cf_options, mutable_db_options,
       std::move(inputs), 0, 0, 0, 0, kNoCompression,
-      mutable_cf_options.compression_opts,
+      mutable_cf_options.compression_opts, Temperature::kUnknown,
       /* max_subcompactions */ 0, {}, /* is manual */ false,
       vstorage->CompactionScore(0),
       /* is deletion compaction */ true, CompactionReason::kFIFOMaxSize);
   return c;
 }
 
+Compaction* FIFOCompactionPicker::PickCompactionToWarm(
+    const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+    const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
+    LogBuffer* log_buffer) {
+  if (mutable_cf_options.compaction_options_fifo.age_for_warm == 0) {
+    return nullptr;
+  }
+
+  const int kLevel0 = 0;
+  const std::vector<FileMetaData*>& level_files = vstorage->LevelFiles(kLevel0);
+
+  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());
+    return nullptr;
+  }
+  const uint64_t current_time = static_cast<uint64_t>(_current_time);
+
+  if (!level0_compactions_in_progress_.empty()) {
+    ROCKS_LOG_BUFFER(
+        log_buffer,
+        "[%s] FIFO compaction: Already executing compaction. Parallel "
+        "compactions are not supported",
+        cf_name.c_str());
+    return nullptr;
+  }
+
+  std::vector<CompactionInputFiles> inputs;
+  inputs.emplace_back();
+  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 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.
+        return nullptr;
+      }
+      uint64_t oldest_ancester_time = f->TryGetOldestAncesterTime();
+      if (oldest_ancester_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.
+        break;
+      }
+      if (prev_file != nullptr) {
+        compaction_size += prev_file->fd.GetFileSize();
+        if (compaction_size > mutable_cf_options.max_compaction_bytes) {
+          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.
+        break;
+      } else {
+        assert(prev_file == nullptr);
+      }
+    }
+  }
+
+  if (inputs[0].files.empty()) {
+    return nullptr;
+  }
+
+  Compaction* c = new Compaction(
+      vstorage, ioptions_, mutable_cf_options, mutable_db_options,
+      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,
+      /* max_subcompactions */ 0, {}, /* is manual */ false,
+      vstorage->CompactionScore(0),
+      /* is deletion compaction */ false, CompactionReason::kChangeTemperature);
+  return c;
+}
+
 Compaction* FIFOCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
@@ -225,6 +332,10 @@ Compaction* FIFOCompactionPicker::PickCompaction(
     c = PickSizeCompaction(cf_name, mutable_cf_options, mutable_db_options,
                            vstorage, log_buffer);
   }
+  if (c == nullptr) {
+    c = PickCompactionToWarm(cf_name, mutable_cf_options, mutable_db_options,
+                             vstorage, log_buffer);
+  }
   RegisterCompaction(c);
   return c;
 }

db/compaction/compaction_picker_fifo.h

@@ -52,6 +52,12 @@ class FIFOCompactionPicker : public CompactionPicker {
                                  const MutableDBOptions& mutable_db_options,
                                  VersionStorageInfo* version,
                                  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);
 };
 }  // namespace ROCKSDB_NAMESPACE
 #endif  // !ROCKSDB_LITE

db/compaction/compaction_picker_level.cc

@@ -336,6 +336,7 @@ Compaction* LevelCompactionBuilder::GetCompaction() {
       GetCompressionType(ioptions_, vstorage_, mutable_cf_options_,
                          output_level_, vstorage_->base_level()),
       GetCompressionOptions(mutable_cf_options_, vstorage_, output_level_),
+      Temperature::kUnknown,
       /* max_subcompactions */ 0, std::move(grandparents_), is_manual_,
       start_level_score_, false /* deletion_compaction */, compaction_reason_);
 

db/compaction/compaction_picker_test.cc

@@ -98,7 +98,9 @@ class CompactionPickerTest : public testing::Test {
   void Add(int level, uint32_t file_number, const char* smallest,
            const char* largest, uint64_t file_size = 1, uint32_t path_id = 0,
            SequenceNumber smallest_seq = 100, SequenceNumber largest_seq = 100,
-           size_t compensated_file_size = 0, bool marked_for_compact = false) {
+           size_t compensated_file_size = 0, bool marked_for_compact = false,
+           Temperature temperature = Temperature::kUnknown,
+           uint64_t oldest_ancestor_time = kUnknownOldestAncesterTime) {
     VersionStorageInfo* vstorage;
     if (temp_vstorage_) {
       vstorage = temp_vstorage_.get();
@@ -115,6 +117,8 @@ class CompactionPickerTest : public testing::Test {
         kUnknownFileChecksum, kUnknownFileChecksumFuncName);
     f->compensated_file_size =
         (compensated_file_size != 0) ? compensated_file_size : file_size;
+    f->temperature = temperature;
+    f->oldest_ancester_time = oldest_ancestor_time;
     vstorage->AddFile(level, f);
     files_.emplace_back(f);
     file_map_.insert({file_number, {f, level}});
@@ -757,6 +761,245 @@ TEST_F(CompactionPickerTest, NeedsCompactionFIFO) {
               vstorage_->CompactionScore(0) >= 1);
   }
 }
+
+TEST_F(CompactionPickerTest, FIFOToWarm1) {
+  NewVersionStorage(1, kCompactionStyleFIFO);
+  const uint64_t kFileSize = 100000;
+  const uint64_t kMaxSize = kFileSize * 100000;
+  uint64_t kWarmThreshold = 2000;
+
+  fifo_options_.max_table_files_size = kMaxSize;
+  fifo_options_.age_for_warm = kWarmThreshold;
+  mutable_cf_options_.compaction_options_fifo = fifo_options_;
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
+  FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+  int64_t current_time = 0;
+  ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
+  uint64_t threshold_time =
+      static_cast<uint64_t>(current_time) - kWarmThreshold;
+  Add(0, 6U, "240", "290", 2 * kFileSize, 0, 2900, 3000, 0, true,
+      Temperature::kUnknown, static_cast<uint64_t>(current_time) - 100);
+  Add(0, 5U, "240", "290", 2 * kFileSize, 0, 2700, 2800, 0, true,
+      Temperature::kUnknown, threshold_time + 100);
+  Add(0, 4U, "260", "300", 1 * kFileSize, 0, 2500, 2600, 0, true,
+      Temperature::kUnknown, threshold_time - 2000);
+  Add(0, 3U, "200", "300", 4 * kFileSize, 0, 2300, 2400, 0, true,
+      Temperature::kUnknown, threshold_time - 3000);
+  UpdateVersionStorageInfo();
+
+  ASSERT_EQ(fifo_compaction_picker.NeedsCompaction(vstorage_.get()), true);
+  std::unique_ptr<Compaction> compaction(fifo_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  ASSERT_EQ(3U, compaction->input(0, 0)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, FIFOToWarm2) {
+  NewVersionStorage(1, kCompactionStyleFIFO);
+  const uint64_t kFileSize = 100000;
+  const uint64_t kMaxSize = kFileSize * 100000;
+  uint64_t kWarmThreshold = 2000;
+
+  fifo_options_.max_table_files_size = kMaxSize;
+  fifo_options_.age_for_warm = kWarmThreshold;
+  mutable_cf_options_.compaction_options_fifo = fifo_options_;
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
+  FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+  int64_t current_time = 0;
+  ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
+  uint64_t threshold_time =
+      static_cast<uint64_t>(current_time) - kWarmThreshold;
+  Add(0, 6U, "240", "290", 2 * kFileSize, 0, 2900, 3000, 0, true,
+      Temperature::kUnknown, static_cast<uint64_t>(current_time) - 100);
+  Add(0, 5U, "240", "290", 2 * kFileSize, 0, 2700, 2800, 0, true,
+      Temperature::kUnknown, threshold_time + 100);
+  Add(0, 4U, "260", "300", 1 * kFileSize, 0, 2500, 2600, 0, true,
+      Temperature::kUnknown, threshold_time - 2000);
+  Add(0, 3U, "200", "300", 4 * kFileSize, 0, 2300, 2400, 0, true,
+      Temperature::kUnknown, threshold_time - 3000);
+  Add(0, 2U, "200", "300", 4 * kFileSize, 0, 2100, 2200, 0, true,
+      Temperature::kUnknown, threshold_time - 4000);
+  UpdateVersionStorageInfo();
+
+  ASSERT_EQ(fifo_compaction_picker.NeedsCompaction(vstorage_.get()), true);
+  std::unique_ptr<Compaction> compaction(fifo_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(2U, compaction->num_input_files(0));
+  ASSERT_EQ(2U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(3U, compaction->input(0, 1)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, FIFOToWarmMaxSize) {
+  NewVersionStorage(1, kCompactionStyleFIFO);
+  const uint64_t kFileSize = 100000;
+  const uint64_t kMaxSize = kFileSize * 100000;
+  uint64_t kWarmThreshold = 2000;
+
+  fifo_options_.max_table_files_size = kMaxSize;
+  fifo_options_.age_for_warm = kWarmThreshold;
+  mutable_cf_options_.compaction_options_fifo = fifo_options_;
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_compaction_bytes = kFileSize * 9;
+  FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+  int64_t current_time = 0;
+  ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
+  uint64_t threshold_time =
+      static_cast<uint64_t>(current_time) - kWarmThreshold;
+  Add(0, 6U, "240", "290", 2 * kFileSize, 0, 2900, 3000, 0, true,
+      Temperature::kUnknown, static_cast<uint64_t>(current_time) - 100);
+  Add(0, 5U, "240", "290", 2 * kFileSize, 0, 2700, 2800, 0, true,
+      Temperature::kUnknown, threshold_time + 100);
+  Add(0, 4U, "260", "300", 1 * kFileSize, 0, 2500, 2600, 0, true,
+      Temperature::kUnknown, threshold_time - 2000);
+  Add(0, 3U, "200", "300", 4 * kFileSize, 0, 2300, 2400, 0, true,
+      Temperature::kUnknown, threshold_time - 3000);
+  Add(0, 2U, "200", "300", 4 * kFileSize, 0, 2100, 2200, 0, true,
+      Temperature::kUnknown, threshold_time - 4000);
+  Add(0, 1U, "200", "300", 4 * kFileSize, 0, 2000, 2100, 0, true,
+      Temperature::kUnknown, threshold_time - 5000);
+  UpdateVersionStorageInfo();
+
+  ASSERT_EQ(fifo_compaction_picker.NeedsCompaction(vstorage_.get()), true);
+  std::unique_ptr<Compaction> compaction(fifo_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(2U, compaction->num_input_files(0));
+  ASSERT_EQ(1U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(2U, compaction->input(0, 1)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, FIFOToWarmWithExistingWarm) {
+  NewVersionStorage(1, kCompactionStyleFIFO);
+  const uint64_t kFileSize = 100000;
+  const uint64_t kMaxSize = kFileSize * 100000;
+  uint64_t kWarmThreshold = 2000;
+
+  fifo_options_.max_table_files_size = kMaxSize;
+  fifo_options_.age_for_warm = kWarmThreshold;
+  mutable_cf_options_.compaction_options_fifo = fifo_options_;
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
+  FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+  int64_t current_time = 0;
+  ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
+  uint64_t threshold_time =
+      static_cast<uint64_t>(current_time) - kWarmThreshold;
+  Add(0, 6U, "240", "290", 2 * kFileSize, 0, 2900, 3000, 0, true,
+      Temperature::kUnknown, static_cast<uint64_t>(current_time) - 100);
+  Add(0, 5U, "240", "290", 2 * kFileSize, 0, 2700, 2800, 0, true,
+      Temperature::kUnknown, threshold_time + 100);
+  Add(0, 4U, "260", "300", 1 * kFileSize, 0, 2500, 2600, 0, true,
+      Temperature::kUnknown, threshold_time - 2000);
+  Add(0, 3U, "200", "300", 4 * kFileSize, 0, 2300, 2400, 0, true,
+      Temperature::kUnknown, threshold_time - 3000);
+  Add(0, 2U, "200", "300", 4 * kFileSize, 0, 2100, 2200, 0, true,
+      Temperature::kUnknown, threshold_time - 4000);
+  Add(0, 1U, "200", "300", 4 * kFileSize, 0, 2000, 2100, 0, true,
+      Temperature::kWarm, threshold_time - 5000);
+  UpdateVersionStorageInfo();
+
+  ASSERT_EQ(fifo_compaction_picker.NeedsCompaction(vstorage_.get()), true);
+  std::unique_ptr<Compaction> compaction(fifo_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(2U, compaction->num_input_files(0));
+  ASSERT_EQ(2U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(3U, compaction->input(0, 1)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, FIFOToWarmWithOngoing) {
+  NewVersionStorage(1, kCompactionStyleFIFO);
+  const uint64_t kFileSize = 100000;
+  const uint64_t kMaxSize = kFileSize * 100000;
+  uint64_t kWarmThreshold = 2000;
+
+  fifo_options_.max_table_files_size = kMaxSize;
+  fifo_options_.age_for_warm = kWarmThreshold;
+  mutable_cf_options_.compaction_options_fifo = fifo_options_;
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
+  FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+  int64_t current_time = 0;
+  ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
+  uint64_t threshold_time =
+      static_cast<uint64_t>(current_time) - kWarmThreshold;
+  Add(0, 6U, "240", "290", 2 * kFileSize, 0, 2900, 3000, 0, true,
+      Temperature::kUnknown, static_cast<uint64_t>(current_time) - 100);
+  Add(0, 5U, "240", "290", 2 * kFileSize, 0, 2700, 2800, 0, true,
+      Temperature::kUnknown, threshold_time + 100);
+  Add(0, 4U, "260", "300", 1 * kFileSize, 0, 2500, 2600, 0, true,
+      Temperature::kUnknown, threshold_time - 2000);
+  Add(0, 3U, "200", "300", 4 * kFileSize, 0, 2300, 2400, 0, true,
+      Temperature::kUnknown, threshold_time - 3000);
+  Add(0, 2U, "200", "300", 4 * kFileSize, 0, 2100, 2200, 0, true,
+      Temperature::kUnknown, threshold_time - 4000);
+  Add(0, 1U, "200", "300", 4 * kFileSize, 0, 2000, 2100, 0, true,
+      Temperature::kWarm, threshold_time - 5000);
+  file_map_[2].first->being_compacted = true;
+  UpdateVersionStorageInfo();
+
+  ASSERT_EQ(fifo_compaction_picker.NeedsCompaction(vstorage_.get()), true);
+  std::unique_ptr<Compaction> compaction(fifo_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  // Stop if a file is being compacted
+  ASSERT_TRUE(compaction.get() == nullptr);
+}
+
+TEST_F(CompactionPickerTest, FIFOToWarmWithHotBetweenWarms) {
+  NewVersionStorage(1, kCompactionStyleFIFO);
+  const uint64_t kFileSize = 100000;
+  const uint64_t kMaxSize = kFileSize * 100000;
+  uint64_t kWarmThreshold = 2000;
+
+  fifo_options_.max_table_files_size = kMaxSize;
+  fifo_options_.age_for_warm = kWarmThreshold;
+  mutable_cf_options_.compaction_options_fifo = fifo_options_;
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
+  FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+  int64_t current_time = 0;
+  ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
+  uint64_t threshold_time =
+      static_cast<uint64_t>(current_time) - kWarmThreshold;
+  Add(0, 6U, "240", "290", 2 * kFileSize, 0, 2900, 3000, 0, true,
+      Temperature::kUnknown, static_cast<uint64_t>(current_time) - 100);
+  Add(0, 5U, "240", "290", 2 * kFileSize, 0, 2700, 2800, 0, true,
+      Temperature::kUnknown, threshold_time + 100);
+  Add(0, 4U, "260", "300", 1 * kFileSize, 0, 2500, 2600, 0, true,
+      Temperature::kUnknown, threshold_time - 2000);
+  Add(0, 3U, "200", "300", 4 * kFileSize, 0, 2300, 2400, 0, true,
+      Temperature::kWarm, threshold_time - 3000);
+  Add(0, 2U, "200", "300", 4 * kFileSize, 0, 2100, 2200, 0, true,
+      Temperature::kUnknown, threshold_time - 4000);
+  Add(0, 1U, "200", "300", 4 * kFileSize, 0, 2000, 2100, 0, true,
+      Temperature::kWarm, threshold_time - 5000);
+  UpdateVersionStorageInfo();
+
+  ASSERT_EQ(fifo_compaction_picker.NeedsCompaction(vstorage_.get()), true);
+  std::unique_ptr<Compaction> compaction(fifo_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  // Stop if a file is being compacted
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  ASSERT_EQ(2U, compaction->input(0, 0)->fd.GetNumber());
+}
+
 #endif  // ROCKSDB_LITE
 
 TEST_F(CompactionPickerTest, CompactionPriMinOverlapping1) {

db/compaction/compaction_picker_universal.cc

@@ -728,6 +728,7 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSortedRuns(
                          1, enable_compression),
       GetCompressionOptions(mutable_cf_options_, vstorage_, start_level,
                             enable_compression),
+      Temperature::kUnknown,
       /* max_subcompactions */ 0, /* grandparents */ {}, /* is manual */ false,
       score_, false /* deletion_compaction */, compaction_reason);
 }
@@ -955,6 +956,7 @@ Compaction* UniversalCompactionBuilder::PickDeleteTriggeredCompaction() {
       GetCompressionType(ioptions_, vstorage_, mutable_cf_options_,
                          output_level, 1),
       GetCompressionOptions(mutable_cf_options_, vstorage_, output_level),
+      Temperature::kUnknown,
       /* max_subcompactions */ 0, /* grandparents */ {}, /* is manual */ false,
       score_, false /* deletion_compaction */,
       CompactionReason::kFilesMarkedForCompaction);
@@ -1029,6 +1031,7 @@ Compaction* UniversalCompactionBuilder::PickCompactionToOldest(
                          output_level, 1, true /* enable_compression */),
       GetCompressionOptions(mutable_cf_options_, vstorage_, output_level,
                             true /* enable_compression */),
+      Temperature::kUnknown,
       /* max_subcompactions */ 0, /* grandparents */ {}, /* is manual */ false,
       score_, false /* deletion_compaction */, compaction_reason);
 }

db/compaction/compaction_service_test.cc

@@ -12,16 +12,16 @@ namespace ROCKSDB_NAMESPACE {
 
 class MyTestCompactionService : public CompactionService {
  public:
-  MyTestCompactionService(const std::string& db_path,
-                          std::shared_ptr<FileSystem> fs, Options& options)
-      : db_path_(db_path), fs_(fs), options_(options) {}
+  MyTestCompactionService(const std::string& db_path, Options& options,
+                          std::shared_ptr<Statistics> statistics = nullptr)
+      : db_path_(db_path), options_(options), statistics_(statistics) {}
 
   static const char* kClassName() { return "MyTestCompactionService"; }
 
   const char* Name() const override { return kClassName(); }
 
   CompactionServiceJobStatus Start(const std::string& compaction_service_input,
-                                   int job_id) override {
+                                   uint64_t job_id) override {
     InstrumentedMutexLock l(&mutex_);
     jobs_.emplace(job_id, compaction_service_input);
     CompactionServiceJobStatus s = CompactionServiceJobStatus::kSuccess;
@@ -30,7 +30,7 @@ class MyTestCompactionService : public CompactionService {
   }
 
   CompactionServiceJobStatus WaitForComplete(
-      int job_id, std::string* compaction_service_result) override {
+      uint64_t job_id, std::string* compaction_service_result) override {
     std::string compaction_input;
     {
       InstrumentedMutexLock l(&mutex_);
@@ -54,6 +54,7 @@ class MyTestCompactionService : public CompactionService {
     options_override.prefix_extractor = options_.prefix_extractor;
     options_override.table_factory = options_.table_factory;
     options_override.sst_partitioner_factory = options_.sst_partitioner_factory;
+    options_override.statistics = statistics_;
 
     Status s = DB::OpenAndCompact(
         db_path_, db_path_ + "/" + ROCKSDB_NAMESPACE::ToString(job_id),
@@ -73,10 +74,10 @@ class MyTestCompactionService : public CompactionService {
  private:
   InstrumentedMutex mutex_;
   std::atomic_int compaction_num_{0};
-  std::map<int, std::string> jobs_;
+  std::map<uint64_t, std::string> jobs_;
   const std::string db_path_;
-  std::shared_ptr<FileSystem> fs_;
   Options options_;
+  std::shared_ptr<Statistics> statistics_;
 };
 
 class CompactionServiceTest : public DBTestBase {
@@ -123,8 +124,10 @@ class CompactionServiceTest : public DBTestBase {
 TEST_F(CompactionServiceTest, BasicCompactions) {
   Options options = CurrentOptions();
   options.env = env_;
+  options.statistics = CreateDBStatistics();
+  std::shared_ptr<Statistics> compactor_statistics = CreateDBStatistics();
   options.compaction_service = std::make_shared<MyTestCompactionService>(
-      dbname_, env_->GetFileSystem(), options);
+      dbname_, options, compactor_statistics);
 
   DestroyAndReopen(options);
 
@@ -158,6 +161,12 @@ TEST_F(CompactionServiceTest, BasicCompactions) {
       dynamic_cast<MyTestCompactionService*>(options.compaction_service.get());
   ASSERT_GE(my_cs->GetCompactionNum(), 1);
 
+  // make sure the compaction statistics is only recorded on remote side
+  ASSERT_GE(
+      compactor_statistics->getTickerCount(COMPACTION_KEY_DROP_NEWER_ENTRY), 1);
+  ASSERT_EQ(options.statistics->getTickerCount(COMPACTION_KEY_DROP_NEWER_ENTRY),
+            0);
+
   // Test failed compaction
   SyncPoint::GetInstance()->SetCallBack(
       "DBImplSecondary::CompactWithoutInstallation::End", [&](void* status) {
@@ -195,8 +204,8 @@ TEST_F(CompactionServiceTest, ManualCompaction) {
   Options options = CurrentOptions();
   options.env = env_;
   options.disable_auto_compactions = true;
-  options.compaction_service = std::make_shared<MyTestCompactionService>(
-      dbname_, env_->GetFileSystem(), options);
+  options.compaction_service =
+      std::make_shared<MyTestCompactionService>(dbname_, options);
   DestroyAndReopen(options);
   GenerateTestData();
 
@@ -236,8 +245,8 @@ TEST_F(CompactionServiceTest, FailedToStart) {
   Options options = CurrentOptions();
   options.env = env_;
   options.disable_auto_compactions = true;
-  options.compaction_service = std::make_shared<MyTestCompactionService>(
-      dbname_, env_->GetFileSystem(), options);
+  options.compaction_service =
+      std::make_shared<MyTestCompactionService>(dbname_, options);
   DestroyAndReopen(options);
   GenerateTestData();
 
@@ -261,8 +270,8 @@ TEST_F(CompactionServiceTest, InvalidResult) {
   Options options = CurrentOptions();
   options.env = env_;
   options.disable_auto_compactions = true;
-  options.compaction_service = std::make_shared<MyTestCompactionService>(
-      dbname_, env_->GetFileSystem(), options);
+  options.compaction_service =
+      std::make_shared<MyTestCompactionService>(dbname_, options);
   DestroyAndReopen(options);
   GenerateTestData();
 
@@ -282,22 +291,31 @@ TEST_F(CompactionServiceTest, InvalidResult) {
   ASSERT_FALSE(s.ok());
 }
 
-// TODO: support sub-compaction
-TEST_F(CompactionServiceTest, DISABLED_SubCompaction) {
+TEST_F(CompactionServiceTest, SubCompaction) {
   Options options = CurrentOptions();
   options.env = env_;
   options.max_subcompactions = 10;
   options.target_file_size_base = 1 << 10;  // 1KB
   options.disable_auto_compactions = true;
-  options.compaction_service = std::make_shared<MyTestCompactionService>(
-      dbname_, env_->GetFileSystem(), options);
+  options.compaction_service =
+      std::make_shared<MyTestCompactionService>(dbname_, options);
 
   DestroyAndReopen(options);
   GenerateTestData();
+  VerifyTestData();
+
+  auto my_cs =
+      dynamic_cast<MyTestCompactionService*>(options.compaction_service.get());
+  int compaction_num_before = my_cs->GetCompactionNum();
 
   auto cro = CompactRangeOptions();
   cro.max_subcompactions = 10;
-  db_->CompactRange(cro, nullptr, nullptr);
+  Status s = db_->CompactRange(cro, nullptr, nullptr);
+  ASSERT_OK(s);
+  VerifyTestData();
+  int compaction_num = my_cs->GetCompactionNum() - compaction_num_before;
+  // make sure there's sub-compaction by checking the compaction number
+  ASSERT_GE(compaction_num, 2);
 }
 
 class PartialDeleteCompactionFilter : public CompactionFilter {
@@ -319,10 +337,11 @@ class PartialDeleteCompactionFilter : public CompactionFilter {
 TEST_F(CompactionServiceTest, CompactionFilter) {
   Options options = CurrentOptions();
   options.env = env_;
-  auto delete_comp_filter = PartialDeleteCompactionFilter();
-  options.compaction_filter = &delete_comp_filter;
-  options.compaction_service = std::make_shared<MyTestCompactionService>(
-      dbname_, env_->GetFileSystem(), options);
+  std::unique_ptr<CompactionFilter> delete_comp_filter(
+      new PartialDeleteCompactionFilter());
+  options.compaction_filter = delete_comp_filter.get();
+  options.compaction_service =
+      std::make_shared<MyTestCompactionService>(dbname_, options);
 
   DestroyAndReopen(options);
 
@@ -364,8 +383,8 @@ TEST_F(CompactionServiceTest, CompactionFilter) {
 TEST_F(CompactionServiceTest, Snapshot) {
   Options options = CurrentOptions();
   options.env = env_;
-  options.compaction_service = std::make_shared<MyTestCompactionService>(
-      dbname_, env_->GetFileSystem(), options);
+  options.compaction_service =
+      std::make_shared<MyTestCompactionService>(dbname_, options);
 
   DestroyAndReopen(options);
 
@@ -391,8 +410,8 @@ TEST_F(CompactionServiceTest, ConcurrentCompaction) {
   Options options = CurrentOptions();
   options.level0_file_num_compaction_trigger = 100;
   options.env = env_;
-  options.compaction_service = std::make_shared<MyTestCompactionService>(
-      dbname_, env_->GetFileSystem(), options);
+  options.compaction_service =
+      std::make_shared<MyTestCompactionService>(dbname_, options);
   options.max_background_jobs = 20;
 
   DestroyAndReopen(options);

db/db_basic_test.cc

@@ -31,7 +31,7 @@ namespace ROCKSDB_NAMESPACE {
 
 class DBBasicTest : public DBTestBase {
  public:
-  DBBasicTest() : DBTestBase("/db_basic_test", /*env_do_fsync=*/false) {}
+  DBBasicTest() : DBTestBase("db_basic_test", /*env_do_fsync=*/false) {}
 };
 
 TEST_F(DBBasicTest, OpenWhenOpen) {
@@ -1362,6 +1362,28 @@ TEST_P(DBMultiGetTestWithParam, MultiGetMultiCFSnapshot) {
   }
 }
 
+TEST_P(DBMultiGetTestWithParam, MultiGetMultiCFUnsorted) {
+  Options options = CurrentOptions();
+  CreateAndReopenWithCF({"one", "two"}, options);
+
+  ASSERT_OK(Put(1, "foo", "bar"));
+  ASSERT_OK(Put(2, "baz", "xyz"));
+  ASSERT_OK(Put(1, "abc", "def"));
+
+  // Note: keys for the same CF do not form a consecutive range
+  std::vector<int> cfs{1, 2, 1};
+  std::vector<std::string> keys{"foo", "baz", "abc"};
+  std::vector<std::string> values;
+
+  values =
+      MultiGet(cfs, keys, /* snapshot */ nullptr, /* batched */ GetParam());
+
+  ASSERT_EQ(values.size(), 3);
+  ASSERT_EQ(values[0], "bar");
+  ASSERT_EQ(values[1], "xyz");
+  ASSERT_EQ(values[2], "def");
+}
+
 INSTANTIATE_TEST_CASE_P(DBMultiGetTestWithParam, DBMultiGetTestWithParam,
                         testing::Bool());
 
@@ -2576,7 +2598,7 @@ class DBBasicTestTrackWal : public DBTestBase,
                             public testing::WithParamInterface<bool> {
  public:
   DBBasicTestTrackWal()
-      : DBTestBase("/db_basic_test_track_wal", /*env_do_fsync=*/false) {}
+      : DBTestBase("db_basic_test_track_wal", /*env_do_fsync=*/false) {}
 
   int CountWalFiles() {
     VectorLogPtr log_files;
@@ -2734,6 +2756,11 @@ class DBBasicTestMultiGet : public DBTestBase {
         EXPECT_OK(dbfull()->Flush(FlushOptions(), handles_[cf]));
       }
     }
+    // Clear compressed cache, which is always pre-populated
+    if (compressed_cache_) {
+      compressed_cache_->SetCapacity(0);
+      compressed_cache_->SetCapacity(1048576);
+    }
   }
 
   bool CheckValue(int i, const std::string& value) {

db/db_block_cache_test.cc

@@ -7,15 +7,20 @@
 // 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.
 #include <cstdlib>
+#include <functional>
 #include <memory>
 
 #include "cache/cache_entry_roles.h"
 #include "cache/lru_cache.h"
+#include "db/column_family.h"
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
+#include "rocksdb/statistics.h"
 #include "rocksdb/table.h"
 #include "util/compression.h"
+#include "util/defer.h"
 #include "util/random.h"
+#include "utilities/fault_injection_fs.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -38,7 +43,7 @@ class DBBlockCacheTest : public DBTestBase {
   const size_t kValueSize = 100;
 
   DBBlockCacheTest()
-      : DBTestBase("/db_block_cache_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_block_cache_test", /*env_do_fsync=*/true) {}
 
   BlockBasedTableOptions GetTableOptions() {
     BlockBasedTableOptions table_options;
@@ -152,13 +157,15 @@ class DBBlockCacheTest : public DBTestBase {
   }
 
 #ifndef ROCKSDB_LITE
-  const std::array<size_t, kNumCacheEntryRoles>& GetCacheEntryRoleCountsBg() {
+  const std::array<size_t, kNumCacheEntryRoles> GetCacheEntryRoleCountsBg() {
     // Verify in cache entry role stats
     ColumnFamilyHandleImpl* cfh =
         static_cast<ColumnFamilyHandleImpl*>(dbfull()->DefaultColumnFamily());
     InternalStats* internal_stats_ptr = cfh->cfd()->internal_stats();
-    return internal_stats_ptr->TEST_GetCacheEntryRoleStats(/*foreground=*/false)
-        .entry_counts;
+    InternalStats::CacheEntryRoleStats stats;
+    internal_stats_ptr->TEST_GetCacheEntryRoleStats(&stats,
+                                                    /*foreground=*/false);
+    return stats.entry_counts;
   }
 #endif  // ROCKSDB_LITE
 };
@@ -170,7 +177,13 @@ TEST_F(DBBlockCacheTest, IteratorBlockCacheUsage) {
   auto options = GetOptions(table_options);
   InitTable(options);
 
-  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  LRUCacheOptions co;
+  co.capacity = 0;
+  co.num_shard_bits = 0;
+  co.strict_capacity_limit = false;
+  // Needed not to count entry stats collector
+  co.metadata_charge_policy = kDontChargeCacheMetadata;
+  std::shared_ptr<Cache> cache = NewLRUCache(co);
   table_options.block_cache = cache;
   options.table_factory.reset(NewBlockBasedTableFactory(table_options));
   Reopen(options);
@@ -194,7 +207,13 @@ TEST_F(DBBlockCacheTest, TestWithoutCompressedBlockCache) {
   auto options = GetOptions(table_options);
   InitTable(options);
 
-  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  LRUCacheOptions co;
+  co.capacity = 0;
+  co.num_shard_bits = 0;
+  co.strict_capacity_limit = false;
+  // Needed not to count entry stats collector
+  co.metadata_charge_policy = kDontChargeCacheMetadata;
+  std::shared_ptr<Cache> cache = NewLRUCache(co);
   table_options.block_cache = cache;
   options.table_factory.reset(NewBlockBasedTableFactory(table_options));
   Reopen(options);
@@ -265,7 +284,13 @@ TEST_F(DBBlockCacheTest, TestWithCompressedBlockCache) {
 
   ReadOptions read_options;
   std::shared_ptr<Cache> compressed_cache = NewLRUCache(1 << 25, 0, false);
-  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  LRUCacheOptions co;
+  co.capacity = 0;
+  co.num_shard_bits = 0;
+  co.strict_capacity_limit = false;
+  // Needed not to count entry stats collector
+  co.metadata_charge_policy = kDontChargeCacheMetadata;
+  std::shared_ptr<Cache> cache = NewLRUCache(co);
   table_options.block_cache = cache;
   table_options.no_block_cache = false;
   table_options.block_cache_compressed = compressed_cache;
@@ -299,8 +324,8 @@ TEST_F(DBBlockCacheTest, TestWithCompressedBlockCache) {
   // Load last key block.
   ASSERT_EQ("Result incomplete: Insert failed due to LRU cache being full.",
             Get(ToString(kNumBlocks - 1)));
-  // Failure won't record the miss counter.
-  CheckCacheCounters(options, 0, 0, 0, 1);
+  // Failure will also record the miss counter.
+  CheckCacheCounters(options, 1, 0, 0, 1);
   CheckCompressedCacheCounters(options, 1, 0, 1, 0);
 
   // Clear strict capacity limit flag. This time we shall hit compressed block
@@ -476,12 +501,98 @@ TEST_F(DBBlockCacheTest, WarmCacheWithDataBlocksDuringFlush) {
     ASSERT_OK(Put(ToString(i), value));
     ASSERT_OK(Flush());
     ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_DATA_ADD));
-
     ASSERT_EQ(value, Get(ToString(i)));
     ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_DATA_MISS));
     ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_DATA_HIT));
   }
 }
+
+// This test cache data, index and filter blocks during flush.
+TEST_F(DBBlockCacheTest, WarmCacheWithBlocksDuringFlush) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
+
+  BlockBasedTableOptions table_options;
+  table_options.block_cache = NewLRUCache(1 << 25, 0, false);
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.prepopulate_block_cache =
+      BlockBasedTableOptions::PrepopulateBlockCache::kFlushOnly;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  DestroyAndReopen(options);
+
+  std::string value(kValueSize, 'a');
+  for (size_t i = 1; i <= kNumBlocks; i++) {
+    ASSERT_OK(Put(ToString(i), value));
+    ASSERT_OK(Flush());
+    ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_DATA_ADD));
+    ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_INDEX_ADD));
+    ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_FILTER_ADD));
+
+    ASSERT_EQ(value, Get(ToString(i)));
+
+    ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_DATA_MISS));
+    ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_DATA_HIT));
+
+    ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_INDEX_MISS));
+    ASSERT_EQ(i * 3, options.statistics->getTickerCount(BLOCK_CACHE_INDEX_HIT));
+
+    ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_FILTER_MISS));
+    ASSERT_EQ(i * 2,
+              options.statistics->getTickerCount(BLOCK_CACHE_FILTER_HIT));
+  }
+}
+
+TEST_F(DBBlockCacheTest, DynamicallyWarmCacheDuringFlush) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
+
+  BlockBasedTableOptions table_options;
+  table_options.block_cache = NewLRUCache(1 << 25, 0, false);
+  table_options.cache_index_and_filter_blocks = false;
+  table_options.prepopulate_block_cache =
+      BlockBasedTableOptions::PrepopulateBlockCache::kFlushOnly;
+
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  DestroyAndReopen(options);
+
+  std::string value(kValueSize, 'a');
+
+  for (size_t i = 1; i <= 5; i++) {
+    ASSERT_OK(Put(ToString(i), value));
+    ASSERT_OK(Flush());
+    ASSERT_EQ(1,
+              options.statistics->getAndResetTickerCount(BLOCK_CACHE_DATA_ADD));
+
+    ASSERT_EQ(value, Get(ToString(i)));
+    ASSERT_EQ(0,
+              options.statistics->getAndResetTickerCount(BLOCK_CACHE_DATA_ADD));
+    ASSERT_EQ(
+        0, options.statistics->getAndResetTickerCount(BLOCK_CACHE_DATA_MISS));
+    ASSERT_EQ(1,
+              options.statistics->getAndResetTickerCount(BLOCK_CACHE_DATA_HIT));
+  }
+
+  ASSERT_OK(dbfull()->SetOptions(
+      {{"block_based_table_factory", "{prepopulate_block_cache=kDisable;}"}}));
+
+  for (size_t i = 6; i <= kNumBlocks; i++) {
+    ASSERT_OK(Put(ToString(i), value));
+    ASSERT_OK(Flush());
+    ASSERT_EQ(0,
+              options.statistics->getAndResetTickerCount(BLOCK_CACHE_DATA_ADD));
+
+    ASSERT_EQ(value, Get(ToString(i)));
+    ASSERT_EQ(1,
+              options.statistics->getAndResetTickerCount(BLOCK_CACHE_DATA_ADD));
+    ASSERT_EQ(
+        1, options.statistics->getAndResetTickerCount(BLOCK_CACHE_DATA_MISS));
+    ASSERT_EQ(0,
+              options.statistics->getAndResetTickerCount(BLOCK_CACHE_DATA_HIT));
+  }
+}
 #endif
 
 namespace {
@@ -944,10 +1055,15 @@ TEST_F(DBBlockCacheTest, CacheCompressionDict) {
 }
 
 static void ClearCache(Cache* cache) {
+  auto roles = CopyCacheDeleterRoleMap();
   std::deque<std::string> keys;
   Cache::ApplyToAllEntriesOptions opts;
   auto callback = [&](const Slice& key, void* /*value*/, size_t /*charge*/,
-                      Cache::DeleterFn /*deleter*/) {
+                      Cache::DeleterFn deleter) {
+    if (roles.find(deleter) == roles.end()) {
+      // Keep the stats collector
+      return;
+    }
     keys.push_back(key.ToString());
   };
   cache->ApplyToAllEntries(callback, opts);
@@ -1126,6 +1242,9 @@ TEST_F(DBBlockCacheTest, CacheEntryRoleStats) {
                               &h, Cache::Priority::HIGH));
       ASSERT_GT(cache->GetUsage(), cache->GetCapacity());
       expected = {};
+      // For CacheEntryStatsCollector
+      expected[static_cast<size_t>(CacheEntryRole::kMisc)] = 1;
+      // For Fill-it-up
       expected[static_cast<size_t>(CacheEntryRole::kMisc)]++;
       // Still able to hit on saved stats
       EXPECT_EQ(prev_expected, GetCacheEntryRoleCountsBg());
@@ -1134,6 +1253,48 @@ TEST_F(DBBlockCacheTest, CacheEntryRoleStats) {
       EXPECT_EQ(expected, GetCacheEntryRoleCountsBg());
 
       cache->Release(h);
+
+      // Now we test that the DB mutex is not held during scans, for the ways
+      // we know how to (possibly) trigger them. Without a better good way to
+      // check this, we simply inject an acquire & release of the DB mutex
+      // deep in the stat collection code. If we were already holding the
+      // mutex, that is UB that would at least be found by TSAN.
+      int scan_count = 0;
+      SyncPoint::GetInstance()->SetCallBack(
+          "CacheEntryStatsCollector::GetStats:AfterApplyToAllEntries",
+          [this, &scan_count](void*) {
+            dbfull()->TEST_LockMutex();
+            dbfull()->TEST_UnlockMutex();
+            ++scan_count;
+          });
+      SyncPoint::GetInstance()->EnableProcessing();
+
+      // Different things that might trigger a scan, with mock sleeps to
+      // force a miss.
+      env_->MockSleepForSeconds(10000);
+      dbfull()->DumpStats();
+      ASSERT_EQ(scan_count, 1);
+
+      env_->MockSleepForSeconds(10000);
+      ASSERT_TRUE(
+          db_->GetMapProperty(DB::Properties::kBlockCacheEntryStats, &values));
+      ASSERT_EQ(scan_count, 2);
+
+      env_->MockSleepForSeconds(10000);
+      std::string value_str;
+      ASSERT_TRUE(
+          db_->GetProperty(DB::Properties::kBlockCacheEntryStats, &value_str));
+      ASSERT_EQ(scan_count, 3);
+
+      env_->MockSleepForSeconds(10000);
+      ASSERT_TRUE(db_->GetProperty(DB::Properties::kCFStats, &value_str));
+      // To match historical speed, querying this property no longer triggers
+      // a scan, even if results are old. But periodic dump stats should keep
+      // things reasonably updated.
+      ASSERT_EQ(scan_count, /*unchanged*/ 3);
+
+      SyncPoint::GetInstance()->DisableProcessing();
+      SyncPoint::GetInstance()->ClearAllCallBacks();
     }
     EXPECT_GE(iterations_tested, 1);
   }
@@ -1141,13 +1302,238 @@ TEST_F(DBBlockCacheTest, CacheEntryRoleStats) {
 
 #endif  // ROCKSDB_LITE
 
+class DBBlockCacheKeyTest
+    : public DBTestBase,
+      public testing::WithParamInterface<std::tuple<bool, bool>> {
+ public:
+  DBBlockCacheKeyTest()
+      : DBTestBase("db_block_cache_test", /*env_do_fsync=*/false) {}
+
+  void SetUp() override {
+    use_compressed_cache_ = std::get<0>(GetParam());
+    exclude_file_numbers_ = std::get<1>(GetParam());
+  }
+
+  bool use_compressed_cache_;
+  bool exclude_file_numbers_;
+};
+
+// Disable LinkFile so that we can physically copy a DB using Checkpoint.
+// Disable file GetUniqueId to enable stable cache keys.
+class StableCacheKeyTestFS : public FaultInjectionTestFS {
+ public:
+  explicit StableCacheKeyTestFS(const std::shared_ptr<FileSystem>& base)
+      : FaultInjectionTestFS(base) {
+    SetFailGetUniqueId(true);
+  }
+
+  virtual ~StableCacheKeyTestFS() override {}
+
+  IOStatus LinkFile(const std::string&, const std::string&, const IOOptions&,
+                    IODebugContext*) override {
+    return IOStatus::NotSupported("Disabled");
+  }
+};
+
+TEST_P(DBBlockCacheKeyTest, StableCacheKeys) {
+  std::shared_ptr<StableCacheKeyTestFS> test_fs{
+      new StableCacheKeyTestFS(env_->GetFileSystem())};
+  std::unique_ptr<CompositeEnvWrapper> test_env{
+      new CompositeEnvWrapper(env_, test_fs)};
+
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
+  options.env = test_env.get();
+
+  BlockBasedTableOptions table_options;
+
+  int key_count = 0;
+  uint64_t expected_stat = 0;
+
+  std::function<void()> verify_stats;
+  if (use_compressed_cache_) {
+    if (!Snappy_Supported()) {
+      ROCKSDB_GTEST_SKIP("Compressed cache test requires snappy support");
+      return;
+    }
+    options.compression = CompressionType::kSnappyCompression;
+    table_options.no_block_cache = true;
+    table_options.block_cache_compressed = NewLRUCache(1 << 25, 0, false);
+    verify_stats = [&options, &expected_stat] {
+      // One for ordinary SST file and one for external SST file
+      ASSERT_EQ(expected_stat,
+                options.statistics->getTickerCount(BLOCK_CACHE_COMPRESSED_ADD));
+    };
+  } else {
+    table_options.cache_index_and_filter_blocks = true;
+    table_options.block_cache = NewLRUCache(1 << 25, 0, false);
+    verify_stats = [&options, &expected_stat] {
+      ASSERT_EQ(expected_stat,
+                options.statistics->getTickerCount(BLOCK_CACHE_DATA_ADD));
+      ASSERT_EQ(expected_stat,
+                options.statistics->getTickerCount(BLOCK_CACHE_INDEX_ADD));
+      ASSERT_EQ(expected_stat,
+                options.statistics->getTickerCount(BLOCK_CACHE_FILTER_ADD));
+    };
+  }
+
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"koko"}, options);
+
+  if (exclude_file_numbers_) {
+    // Simulate something like old behavior without file numbers in properties.
+    // This is a "control" side of the test that also ensures safely degraded
+    // behavior on old files.
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+        "PropertyBlockBuilder::AddTableProperty:Start", [&](void* arg) {
+          TableProperties* props = reinterpret_cast<TableProperties*>(arg);
+          props->orig_file_number = 0;
+        });
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+  }
+
+  std::function<void()> perform_gets = [&key_count, &expected_stat, this]() {
+    if (exclude_file_numbers_) {
+      // No cache key reuse should happen, because we can't rely on current
+      // file number being stable
+      expected_stat += key_count;
+    } else {
+      // Cache keys should be stable
+      expected_stat = key_count;
+    }
+    for (int i = 0; i < key_count; ++i) {
+      ASSERT_EQ(Get(1, Key(i)), "abc");
+    }
+  };
+
+  // Ordinary SST files with same session id
+  const std::string something_compressible(500U, 'x');
+  for (int i = 0; i < 2; ++i) {
+    ASSERT_OK(Put(1, Key(key_count), "abc"));
+    ASSERT_OK(Put(1, Key(key_count) + "a", something_compressible));
+    ASSERT_OK(Flush(1));
+    ++key_count;
+  }
+
+#ifndef ROCKSDB_LITE
+  // Save an export of those ordinary SST files for later
+  std::string export_files_dir = dbname_ + "/exported";
+  ExportImportFilesMetaData* metadata_ptr_ = nullptr;
+  Checkpoint* checkpoint;
+  ASSERT_OK(Checkpoint::Create(db_, &checkpoint));
+  ASSERT_OK(checkpoint->ExportColumnFamily(handles_[1], export_files_dir,
+                                           &metadata_ptr_));
+  ASSERT_NE(metadata_ptr_, nullptr);
+  delete checkpoint;
+  checkpoint = nullptr;
+
+  // External SST files with same session id
+  SstFileWriter sst_file_writer(EnvOptions(), options);
+  std::vector<std::string> external;
+  for (int i = 0; i < 2; ++i) {
+    std::string f = dbname_ + "/external" + ToString(i) + ".sst";
+    external.push_back(f);
+    ASSERT_OK(sst_file_writer.Open(f));
+    ASSERT_OK(sst_file_writer.Put(Key(key_count), "abc"));
+    ASSERT_OK(
+        sst_file_writer.Put(Key(key_count) + "a", something_compressible));
+    ++key_count;
+    ExternalSstFileInfo external_info;
+    ASSERT_OK(sst_file_writer.Finish(&external_info));
+    IngestExternalFileOptions ingest_opts;
+    ASSERT_OK(db_->IngestExternalFile(handles_[1], {f}, ingest_opts));
+  }
+
+  if (exclude_file_numbers_) {
+    // FIXME(peterd): figure out where these extra two ADDs are coming from
+    options.statistics->recordTick(BLOCK_CACHE_INDEX_ADD,
+                                   uint64_t{0} - uint64_t{2});
+    options.statistics->recordTick(BLOCK_CACHE_FILTER_ADD,
+                                   uint64_t{0} - uint64_t{2});
+    options.statistics->recordTick(BLOCK_CACHE_COMPRESSED_ADD,
+                                   uint64_t{0} - uint64_t{2});
+  }
+#endif
+
+  perform_gets();
+  verify_stats();
+
+  // Make sure we can cache hit after re-open
+  ReopenWithColumnFamilies({"default", "koko"}, options);
+
+  perform_gets();
+  verify_stats();
+
+  // Make sure we can cache hit even on a full copy of the DB. Using
+  // StableCacheKeyTestFS, Checkpoint will resort to full copy not hard link.
+  // (Checkpoint  not available in LITE mode to test this.)
+#ifndef ROCKSDB_LITE
+  auto db_copy_name = dbname_ + "-copy";
+  ASSERT_OK(Checkpoint::Create(db_, &checkpoint));
+  ASSERT_OK(checkpoint->CreateCheckpoint(db_copy_name));
+  delete checkpoint;
+
+  Close();
+  Destroy(options);
+
+  // Switch to the DB copy
+  SaveAndRestore<std::string> save_dbname(&dbname_, db_copy_name);
+  ReopenWithColumnFamilies({"default", "koko"}, options);
+
+  perform_gets();
+  verify_stats();
+
+  // And ensure that re-importing + ingesting the same files into a
+  // different DB uses same cache keys
+  DestroyAndReopen(options);
+
+  ColumnFamilyHandle* cfh = nullptr;
+  ASSERT_OK(db_->CreateColumnFamilyWithImport(ColumnFamilyOptions(), "yoyo",
+                                              ImportColumnFamilyOptions(),
+                                              *metadata_ptr_, &cfh));
+  ASSERT_NE(cfh, nullptr);
+  delete cfh;
+  cfh = nullptr;
+  delete metadata_ptr_;
+  metadata_ptr_ = nullptr;
+
+  DestroyDB(export_files_dir, options);
+
+  ReopenWithColumnFamilies({"default", "yoyo"}, options);
+
+  IngestExternalFileOptions ingest_opts;
+  ASSERT_OK(db_->IngestExternalFile(handles_[1], {external}, ingest_opts));
+
+  if (exclude_file_numbers_) {
+    // FIXME(peterd): figure out where these extra two ADDs are coming from
+    options.statistics->recordTick(BLOCK_CACHE_INDEX_ADD,
+                                   uint64_t{0} - uint64_t{2});
+    options.statistics->recordTick(BLOCK_CACHE_FILTER_ADD,
+                                   uint64_t{0} - uint64_t{2});
+  }
+
+  perform_gets();
+  verify_stats();
+#endif  // !ROCKSDB_LITE
+
+  Close();
+  Destroy(options);
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+INSTANTIATE_TEST_CASE_P(DBBlockCacheKeyTest, DBBlockCacheKeyTest,
+                        ::testing::Combine(::testing::Bool(),
+                                           ::testing::Bool()));
+
 class DBBlockCachePinningTest
     : public DBTestBase,
       public testing::WithParamInterface<
           std::tuple<bool, PinningTier, PinningTier, PinningTier>> {
  public:
   DBBlockCachePinningTest()
-      : DBTestBase("/db_block_cache_test", /*env_do_fsync=*/false) {}
+      : DBTestBase("db_block_cache_test", /*env_do_fsync=*/false) {}
 
   void SetUp() override {
     partition_index_and_filters_ = std::get<0>(GetParam());

db/db_bloom_filter_test.cc

@@ -28,7 +28,7 @@ using BFP = BloomFilterPolicy;
 class DBBloomFilterTest : public DBTestBase {
  public:
   DBBloomFilterTest()
-      : DBTestBase("/db_bloom_filter_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_bloom_filter_test", /*env_do_fsync=*/true) {}
 };
 
 class DBBloomFilterTestWithParam : public DBTestBase,
@@ -42,7 +42,7 @@ class DBBloomFilterTestWithParam : public DBTestBase,
 
  public:
   DBBloomFilterTestWithParam()
-      : DBTestBase("/db_bloom_filter_tests", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_bloom_filter_tests", /*env_do_fsync=*/true) {}
 
   ~DBBloomFilterTestWithParam() override {}
 
@@ -527,7 +527,7 @@ TEST_P(DBBloomFilterTestWithParam, BloomFilter) {
   } while (ChangeCompactOptions());
 }
 
-#ifndef ROCKSDB_VALGRIND_RUN
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 INSTANTIATE_TEST_CASE_P(
     FormatDef, DBBloomFilterTestDefFormatVersion,
     ::testing::Values(
@@ -551,7 +551,7 @@ INSTANTIATE_TEST_CASE_P(
                         test::kLatestFormatVersion),
         std::make_tuple(BFP::kAutoBloom, true, test::kLatestFormatVersion),
         std::make_tuple(BFP::kAutoBloom, false, test::kLatestFormatVersion)));
-#endif  // ROCKSDB_VALGRIND_RUN
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 TEST_F(DBBloomFilterTest, BloomFilterRate) {
   while (ChangeFilterOptions()) {
@@ -1094,7 +1094,7 @@ class DBBloomFilterTestVaryPrefixAndFormatVer
 
  public:
   DBBloomFilterTestVaryPrefixAndFormatVer()
-      : DBTestBase("/db_bloom_filter_tests", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_bloom_filter_tests", /*env_do_fsync=*/true) {}
 
   ~DBBloomFilterTestVaryPrefixAndFormatVer() override {}
 

db/db_compaction_filter_test.cc

@@ -22,7 +22,7 @@ static std::string NEW_VALUE = "NewValue";
 class DBTestCompactionFilter : public DBTestBase {
  public:
   DBTestCompactionFilter()
-      : DBTestBase("/db_compaction_filter_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_compaction_filter_test", /*env_do_fsync=*/true) {}
 };
 
 // Param variant of DBTestBase::ChangeCompactOptions
@@ -46,7 +46,7 @@ class DBTestCompactionFilterWithCompactParam
   }
 };
 
-#ifndef ROCKSDB_VALGRIND_RUN
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 INSTANTIATE_TEST_CASE_P(
     CompactionFilterWithOption, DBTestCompactionFilterWithCompactParam,
     ::testing::Values(DBTestBase::OptionConfig::kDefault,
@@ -55,11 +55,11 @@ INSTANTIATE_TEST_CASE_P(
                       DBTestBase::OptionConfig::kLevelSubcompactions,
                       DBTestBase::OptionConfig::kUniversalSubcompactions));
 #else
-// Run fewer cases in valgrind
+// Run fewer cases in non-full valgrind to save time.
 INSTANTIATE_TEST_CASE_P(CompactionFilterWithOption,
                         DBTestCompactionFilterWithCompactParam,
                         ::testing::Values(DBTestBase::OptionConfig::kDefault));
-#endif  // ROCKSDB_VALGRIND_RUN
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 class KeepFilter : public CompactionFilter {
  public:

db/db_compaction_test.cc

@@ -32,7 +32,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBCompactionTest : public DBTestBase {
  public:
   DBCompactionTest()
-      : DBTestBase("/db_compaction_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_compaction_test", /*env_do_fsync=*/true) {}
 };
 
 class DBCompactionTestWithParam
@@ -40,7 +40,7 @@ class DBCompactionTestWithParam
       public testing::WithParamInterface<std::tuple<uint32_t, bool>> {
  public:
   DBCompactionTestWithParam()
-      : DBTestBase("/db_compaction_test", /*env_do_fsync=*/true) {
+      : DBTestBase("db_compaction_test", /*env_do_fsync=*/true) {
     max_subcompactions_ = std::get<0>(GetParam());
     exclusive_manual_compaction_ = std::get<1>(GetParam());
   }
@@ -58,7 +58,7 @@ class DBCompactionTestWithBottommostParam
       public testing::WithParamInterface<BottommostLevelCompaction> {
  public:
   DBCompactionTestWithBottommostParam()
-      : DBTestBase("/db_compaction_test", /*env_do_fsync=*/true) {
+      : DBTestBase("db_compaction_test", /*env_do_fsync=*/true) {
     bottommost_level_compaction_ = GetParam();
   }
 
@@ -301,7 +301,7 @@ const SstFileMetaData* PickFileRandomly(
 }
 }  // anonymous namespace
 
-#ifndef ROCKSDB_VALGRIND_RUN
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 // All the TEST_P tests run once with sub_compactions disabled (i.e.
 // options.max_subcompactions = 1) and once with it enabled
 TEST_P(DBCompactionTestWithParam, CompactionDeletionTrigger) {
@@ -366,7 +366,7 @@ TEST_P(DBCompactionTestWithParam, CompactionDeletionTrigger) {
     }
   }
 }
-#endif  // ROCKSDB_VALGRIND_RUN
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 TEST_P(DBCompactionTestWithParam, CompactionsPreserveDeletes) {
   //  For each options type we test following
@@ -2332,7 +2332,7 @@ TEST_P(DBCompactionTestWithParam, LevelCompactionCFPathUse) {
   option_vector.emplace_back(DBOptions(options), cf_opt1);
   CreateColumnFamilies({"one"},option_vector[1]);
 
-  // Configura CF2 specific paths.
+  // Configure CF2 specific paths.
   cf_opt2.cf_paths.emplace_back(dbname_ + "cf2", 500 * 1024);
   cf_opt2.cf_paths.emplace_back(dbname_ + "cf2_2", 4 * 1024 * 1024);
   cf_opt2.cf_paths.emplace_back(dbname_ + "cf2_3", 1024 * 1024 * 1024);
@@ -2387,13 +2387,16 @@ TEST_P(DBCompactionTestWithParam, LevelCompactionCFPathUse) {
   // Check that default column family uses db_paths.
   // And Column family "one" uses cf_paths.
 
-  // First three 110KB files are not going to second path.
-  // After that, (100K, 200K)
+  // The compaction in level0 outputs the sst files in level1.
+  // The first path cannot hold level1's data(400KB+400KB > 500KB),
+  // so every compaction move a sst file to second path. Please
+  // refer to LevelCompactionBuilder::GetPathId.
   for (int num = 0; num < 3; num++) {
     generate_file();
   }
+  check_sstfilecount(0, 1);
+  check_sstfilecount(1, 2);
 
-  // Another 110KB triggers a compaction to 400K file to fill up first path
   generate_file();
   check_sstfilecount(1, 3);
 
@@ -3600,6 +3603,41 @@ TEST_F(DBCompactionTest, CompactFilesOverlapInL0Bug) {
   ASSERT_EQ("new_val", Get(Key(0)));
 }
 
+TEST_F(DBCompactionTest, DeleteFilesInRangeConflictWithCompaction) {
+  Options options = CurrentOptions();
+  DestroyAndReopen(options);
+  const Snapshot* snapshot = nullptr;
+  const int kMaxKey = 10;
+
+  for (int i = 0; i < kMaxKey; i++) {
+    ASSERT_OK(Put(Key(i), Key(i)));
+    ASSERT_OK(Delete(Key(i)));
+    if (!snapshot) {
+      snapshot = db_->GetSnapshot();
+    }
+  }
+  ASSERT_OK(Flush());
+  MoveFilesToLevel(1);
+  ASSERT_OK(Put(Key(kMaxKey), Key(kMaxKey)));
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  // test DeleteFilesInRange() deletes the files already picked for compaction
+  SyncPoint::GetInstance()->LoadDependency(
+      {{"VersionSet::LogAndApply:WriteManifestStart",
+        "BackgroundCallCompaction:0"},
+       {"DBImpl::BackgroundCompaction:Finish",
+        "VersionSet::LogAndApply:WriteManifestDone"}});
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  // release snapshot which mark bottommost file for compaction
+  db_->ReleaseSnapshot(snapshot);
+  std::string begin_string = Key(0);
+  std::string end_string = Key(kMaxKey + 1);
+  Slice begin(begin_string);
+  Slice end(end_string);
+  ASSERT_OK(DeleteFilesInRange(db_, db_->DefaultColumnFamily(), &begin, &end));
+  SyncPoint::GetInstance()->DisableProcessing();
+}
+
 TEST_F(DBCompactionTest, CompactBottomLevelFilesWithDeletions) {
   // bottom-level files may contain deletions due to snapshots protecting the
   // deleted keys. Once the snapshot is released, we should see files with many
@@ -4970,7 +5008,7 @@ class CompactionPriTest : public DBTestBase,
                           public testing::WithParamInterface<uint32_t> {
  public:
   CompactionPriTest()
-      : DBTestBase("/compaction_pri_test", /*env_do_fsync=*/true) {
+      : DBTestBase("compaction_pri_test", /*env_do_fsync=*/true) {
     compaction_pri_ = GetParam();
   }
 
@@ -6711,6 +6749,69 @@ TEST_F(DBCompactionTest, CompactionWithChecksumHandoffManifest2) {
   Destroy(options);
 }
 
+TEST_F(DBCompactionTest, FIFOWarm) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleFIFO;
+  options.num_levels = 1;
+  options.max_open_files = -1;
+  options.level0_file_num_compaction_trigger = 2;
+  options.create_if_missing = true;
+  CompactionOptionsFIFO fifo_options;
+  fifo_options.age_for_warm = 1000;
+  fifo_options.max_table_files_size = 100000000;
+  options.compaction_options_fifo = fifo_options;
+  env_->SetMockSleep();
+  Reopen(options);
+
+  int total_warm = 0;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "NewWritableFile::FileOptions.temperature", [&](void* arg) {
+        Temperature temperature = *(static_cast<Temperature*>(arg));
+        if (temperature == Temperature::kWarm) {
+          total_warm++;
+        }
+      });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  // The file system does not support checksum handoff. The check
+  // will be ignored.
+  ASSERT_OK(Put(Key(0), "value1"));
+  env_->MockSleepForSeconds(800);
+  ASSERT_OK(Put(Key(2), "value2"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(Put(Key(0), "value1"));
+  env_->MockSleepForSeconds(800);
+  ASSERT_OK(Put(Key(2), "value2"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(Put(Key(0), "value1"));
+  env_->MockSleepForSeconds(800);
+  ASSERT_OK(Put(Key(2), "value2"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+  ASSERT_OK(Put(Key(0), "value1"));
+  env_->MockSleepForSeconds(800);
+  ASSERT_OK(Put(Key(2), "value2"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+
+  ColumnFamilyMetaData metadata;
+  db_->GetColumnFamilyMetaData(&metadata);
+  ASSERT_EQ(4, metadata.file_count);
+  ASSERT_EQ(Temperature::kUnknown, metadata.levels[0].files[0].temperature);
+  ASSERT_EQ(Temperature::kUnknown, metadata.levels[0].files[1].temperature);
+  ASSERT_EQ(Temperature::kWarm, metadata.levels[0].files[2].temperature);
+  ASSERT_EQ(Temperature::kWarm, metadata.levels[0].files[3].temperature);
+  ASSERT_EQ(2, total_warm);
+
+  Destroy(options);
+}
+
 #endif  // !defined(ROCKSDB_LITE)
 
 }  // namespace ROCKSDB_NAMESPACE

db/db_dynamic_level_test.cc

@@ -22,7 +22,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBTestDynamicLevel : public DBTestBase {
  public:
   DBTestDynamicLevel()
-      : DBTestBase("/db_dynamic_level_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_dynamic_level_test", /*env_do_fsync=*/true) {}
 };
 
 TEST_F(DBTestDynamicLevel, DynamicLevelMaxBytesBase) {

db/db_encryption_test.cc

@@ -17,7 +17,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBEncryptionTest : public DBTestBase {
  public:
   DBEncryptionTest()
-      : DBTestBase("/db_encryption_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_encryption_test", /*env_do_fsync=*/true) {}
   Env* GetTargetEnv() {
     if (encrypted_env_ != nullptr) {
       return (static_cast<EnvWrapper*>(encrypted_env_))->target();

db/db_filesnapshot.cc

@@ -127,7 +127,22 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
       bg_cv_.Wait();
     }
   }
-  return wal_manager_.GetSortedWalFiles(files);
+
+  // Disable deletion in order to avoid the case where a file is deleted in
+  // the middle of the process so IO error is returned.
+  Status s = DisableFileDeletions();
+  bool file_deletion_supported = !s.IsNotSupported();
+  if (s.ok() || !file_deletion_supported) {
+    s = wal_manager_.GetSortedWalFiles(files);
+    if (file_deletion_supported) {
+      Status s2 = EnableFileDeletions(false);
+      if (!s2.ok() && s.ok()) {
+        s = s2;
+      }
+    }
+  }
+
+  return s;
 }
 
 Status DBImpl::GetCurrentWalFile(std::unique_ptr<LogFile>* current_log_file) {

db/db_flush_test.cc

@@ -8,6 +8,7 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include <atomic>
+#include <limits>
 
 #include "db/db_impl/db_impl.h"
 #include "db/db_test_util.h"
@@ -24,9 +25,13 @@
 
 namespace ROCKSDB_NAMESPACE {
 
+// This is a static filter used for filtering
+// kvs during the compaction process.
+static std::string NEW_VALUE = "NewValue";
+
 class DBFlushTest : public DBTestBase {
  public:
-  DBFlushTest() : DBTestBase("/db_flush_test", /*env_do_fsync=*/true) {}
+  DBFlushTest() : DBTestBase("db_flush_test", /*env_do_fsync=*/true) {}
 };
 
 class DBFlushDirectIOTest : public DBFlushTest,
@@ -658,6 +663,717 @@ TEST_F(DBFlushTest, StatisticsGarbageRangeDeletes) {
   Close();
 }
 
+#ifndef ROCKSDB_LITE
+// This simple Listener can only handle one flush at a time.
+class TestFlushListener : public EventListener {
+ public:
+  TestFlushListener(Env* env, DBFlushTest* test)
+      : slowdown_count(0), stop_count(0), db_closed(), env_(env), test_(test) {
+    db_closed = false;
+  }
+
+  ~TestFlushListener() override {
+    prev_fc_info_.status.PermitUncheckedError();  // Ignore the status
+  }
+  void OnTableFileCreated(const TableFileCreationInfo& info) override {
+    // remember the info for later checking the FlushJobInfo.
+    prev_fc_info_ = info;
+    ASSERT_GT(info.db_name.size(), 0U);
+    ASSERT_GT(info.cf_name.size(), 0U);
+    ASSERT_GT(info.file_path.size(), 0U);
+    ASSERT_GT(info.job_id, 0);
+    ASSERT_GT(info.table_properties.data_size, 0U);
+    ASSERT_GT(info.table_properties.raw_key_size, 0U);
+    ASSERT_GT(info.table_properties.raw_value_size, 0U);
+    ASSERT_GT(info.table_properties.num_data_blocks, 0U);
+    ASSERT_GT(info.table_properties.num_entries, 0U);
+    ASSERT_EQ(info.file_checksum, kUnknownFileChecksum);
+    ASSERT_EQ(info.file_checksum_func_name, kUnknownFileChecksumFuncName);
+  }
+
+  void OnFlushCompleted(DB* db, const FlushJobInfo& info) override {
+    flushed_dbs_.push_back(db);
+    flushed_column_family_names_.push_back(info.cf_name);
+    if (info.triggered_writes_slowdown) {
+      slowdown_count++;
+    }
+    if (info.triggered_writes_stop) {
+      stop_count++;
+    }
+    // verify whether the previously created file matches the flushed file.
+    ASSERT_EQ(prev_fc_info_.db_name, db->GetName());
+    ASSERT_EQ(prev_fc_info_.cf_name, info.cf_name);
+    ASSERT_EQ(prev_fc_info_.job_id, info.job_id);
+    ASSERT_EQ(prev_fc_info_.file_path, info.file_path);
+    ASSERT_EQ(TableFileNameToNumber(info.file_path), info.file_number);
+
+    // Note: the following chunk relies on the notification pertaining to the
+    // database pointed to by DBTestBase::db_, and is thus bypassed when
+    // that assumption does not hold (see the test case MultiDBMultiListeners
+    // below).
+    ASSERT_TRUE(test_);
+    if (db == test_->db_) {
+      std::vector<std::vector<FileMetaData>> files_by_level;
+      test_->dbfull()->TEST_GetFilesMetaData(db->DefaultColumnFamily(),
+                                             &files_by_level);
+
+      ASSERT_FALSE(files_by_level.empty());
+      auto it = std::find_if(files_by_level[0].begin(), files_by_level[0].end(),
+                             [&](const FileMetaData& meta) {
+                               return meta.fd.GetNumber() == info.file_number;
+                             });
+      ASSERT_NE(it, files_by_level[0].end());
+      ASSERT_EQ(info.oldest_blob_file_number, it->oldest_blob_file_number);
+    }
+
+    ASSERT_EQ(db->GetEnv()->GetThreadID(), info.thread_id);
+    ASSERT_GT(info.thread_id, 0U);
+  }
+
+  std::vector<std::string> flushed_column_family_names_;
+  std::vector<DB*> flushed_dbs_;
+  int slowdown_count;
+  int stop_count;
+  bool db_closing;
+  std::atomic_bool db_closed;
+  TableFileCreationInfo prev_fc_info_;
+
+ protected:
+  Env* env_;
+  DBFlushTest* test_;
+};
+#endif  // !ROCKSDB_LITE
+
+TEST_F(DBFlushTest, MemPurgeBasic) {
+  Options options = CurrentOptions();
+
+  // The following options are used to enforce several values that
+  // may already exist as default values to make this test resilient
+  // to default value updates in the future.
+  options.statistics = CreateDBStatistics();
+
+  // Record all statistics.
+  options.statistics->set_stats_level(StatsLevel::kAll);
+
+  // create the DB if it's not already present
+  options.create_if_missing = true;
+
+  // Useful for now as we are trying to compare uncompressed data savings on
+  // flush().
+  options.compression = kNoCompression;
+
+  // Prevent memtable in place updates. Should already be disabled
+  // (from Wiki:
+  //  In place updates can be enabled by toggling on the bool
+  //  inplace_update_support flag. However, this flag is by default set to
+  //  false
+  //  because this thread-safe in-place update support is not compatible
+  //  with concurrent memtable writes. Note that the bool
+  //  allow_concurrent_memtable_write is set to true by default )
+  options.inplace_update_support = false;
+  options.allow_concurrent_memtable_write = true;
+
+  // Enforce size of a single MemTable to 64MB (64MB = 67108864 bytes).
+  options.write_buffer_size = 1 << 20;
+  // Activate the MemPurge prototype.
+  options.experimental_mempurge_threshold = 1.0;
+#ifndef ROCKSDB_LITE
+  TestFlushListener* listener = new TestFlushListener(options.env, this);
+  options.listeners.emplace_back(listener);
+#endif  // !ROCKSDB_LITE
+  ASSERT_OK(TryReopen(options));
+  uint32_t mempurge_count = 0;
+  uint32_t sst_count = 0;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::FlushJob:MemPurgeSuccessful",
+      [&](void* /*arg*/) { mempurge_count++; });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::FlushJob:SSTFileCreated", [&](void* /*arg*/) { sst_count++; });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  std::string KEY1 = "IamKey1";
+  std::string KEY2 = "IamKey2";
+  std::string KEY3 = "IamKey3";
+  std::string KEY4 = "IamKey4";
+  std::string KEY5 = "IamKey5";
+  std::string KEY6 = "IamKey6";
+  std::string KEY7 = "IamKey7";
+  std::string KEY8 = "IamKey8";
+  std::string KEY9 = "IamKey9";
+  std::string RNDKEY1, RNDKEY2, RNDKEY3;
+  const std::string NOT_FOUND = "NOT_FOUND";
+
+  // Heavy overwrite workload,
+  // more than would fit in maximum allowed memtables.
+  Random rnd(719);
+  const size_t NUM_REPEAT = 100;
+  const size_t RAND_KEYS_LENGTH = 57;
+  const size_t RAND_VALUES_LENGTH = 10240;
+  std::string p_v1, p_v2, p_v3, p_v4, p_v5, p_v6, p_v7, p_v8, p_v9, p_rv1,
+      p_rv2, p_rv3;
+
+  // Insert a very first set of keys that will be
+  // mempurged at least once.
+  p_v1 = rnd.RandomString(RAND_VALUES_LENGTH);
+  p_v2 = rnd.RandomString(RAND_VALUES_LENGTH);
+  p_v3 = rnd.RandomString(RAND_VALUES_LENGTH);
+  p_v4 = rnd.RandomString(RAND_VALUES_LENGTH);
+  ASSERT_OK(Put(KEY1, p_v1));
+  ASSERT_OK(Put(KEY2, p_v2));
+  ASSERT_OK(Put(KEY3, p_v3));
+  ASSERT_OK(Put(KEY4, p_v4));
+  ASSERT_EQ(Get(KEY1), p_v1);
+  ASSERT_EQ(Get(KEY2), p_v2);
+  ASSERT_EQ(Get(KEY3), p_v3);
+  ASSERT_EQ(Get(KEY4), p_v4);
+
+  // Insertion of of K-V pairs, multiple times (overwrites).
+  for (size_t i = 0; i < NUM_REPEAT; i++) {
+    // Create value strings of arbitrary length RAND_VALUES_LENGTH bytes.
+    p_v5 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v6 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v7 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v8 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v9 = rnd.RandomString(RAND_VALUES_LENGTH);
+
+    ASSERT_OK(Put(KEY5, p_v5));
+    ASSERT_OK(Put(KEY6, p_v6));
+    ASSERT_OK(Put(KEY7, p_v7));
+    ASSERT_OK(Put(KEY8, p_v8));
+    ASSERT_OK(Put(KEY9, p_v9));
+
+    ASSERT_EQ(Get(KEY1), p_v1);
+    ASSERT_EQ(Get(KEY2), p_v2);
+    ASSERT_EQ(Get(KEY3), p_v3);
+    ASSERT_EQ(Get(KEY4), p_v4);
+    ASSERT_EQ(Get(KEY5), p_v5);
+    ASSERT_EQ(Get(KEY6), p_v6);
+    ASSERT_EQ(Get(KEY7), p_v7);
+    ASSERT_EQ(Get(KEY8), p_v8);
+    ASSERT_EQ(Get(KEY9), p_v9);
+  }
+
+  // Check that there was at least one mempurge
+  const uint32_t EXPECTED_MIN_MEMPURGE_COUNT = 1;
+  // Check that there was no SST files created during flush.
+  const uint32_t EXPECTED_SST_COUNT = 0;
+
+  EXPECT_GE(mempurge_count, EXPECTED_MIN_MEMPURGE_COUNT);
+  EXPECT_EQ(sst_count, EXPECTED_SST_COUNT);
+
+  const uint32_t mempurge_count_record = mempurge_count;
+
+  // Insertion of of K-V pairs, no overwrites.
+  for (size_t i = 0; i < NUM_REPEAT; i++) {
+    // Create value strings of arbitrary length RAND_VALUES_LENGTH bytes.
+    RNDKEY1 = rnd.RandomString(RAND_KEYS_LENGTH);
+    RNDKEY2 = rnd.RandomString(RAND_KEYS_LENGTH);
+    RNDKEY3 = rnd.RandomString(RAND_KEYS_LENGTH);
+    p_rv1 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_rv2 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_rv3 = rnd.RandomString(RAND_VALUES_LENGTH);
+
+    ASSERT_OK(Put(RNDKEY1, p_rv1));
+    ASSERT_OK(Put(RNDKEY2, p_rv2));
+    ASSERT_OK(Put(RNDKEY3, p_rv3));
+
+    ASSERT_EQ(Get(KEY1), p_v1);
+    ASSERT_EQ(Get(KEY2), p_v2);
+    ASSERT_EQ(Get(KEY3), p_v3);
+    ASSERT_EQ(Get(KEY4), p_v4);
+    ASSERT_EQ(Get(KEY5), p_v5);
+    ASSERT_EQ(Get(KEY6), p_v6);
+    ASSERT_EQ(Get(KEY7), p_v7);
+    ASSERT_EQ(Get(KEY8), p_v8);
+    ASSERT_EQ(Get(KEY9), p_v9);
+    ASSERT_EQ(Get(RNDKEY1), p_rv1);
+    ASSERT_EQ(Get(RNDKEY2), p_rv2);
+    ASSERT_EQ(Get(RNDKEY3), p_rv3);
+  }
+
+  // Assert that at least one flush to storage has been performed
+  ASSERT_GT(sst_count, EXPECTED_SST_COUNT);
+  // (which will consequently increase the number of mempurges recorded too).
+  ASSERT_GE(mempurge_count, mempurge_count_record);
+
+  // Assert that there is no data corruption, even with
+  // a flush to storage.
+  ASSERT_EQ(Get(KEY1), p_v1);
+  ASSERT_EQ(Get(KEY2), p_v2);
+  ASSERT_EQ(Get(KEY3), p_v3);
+  ASSERT_EQ(Get(KEY4), p_v4);
+  ASSERT_EQ(Get(KEY5), p_v5);
+  ASSERT_EQ(Get(KEY6), p_v6);
+  ASSERT_EQ(Get(KEY7), p_v7);
+  ASSERT_EQ(Get(KEY8), p_v8);
+  ASSERT_EQ(Get(KEY9), p_v9);
+  ASSERT_EQ(Get(RNDKEY1), p_rv1);
+  ASSERT_EQ(Get(RNDKEY2), p_rv2);
+  ASSERT_EQ(Get(RNDKEY3), p_rv3);
+
+  Close();
+}
+
+TEST_F(DBFlushTest, MemPurgeDeleteAndDeleteRange) {
+  Options options = CurrentOptions();
+
+  options.statistics = CreateDBStatistics();
+  options.statistics->set_stats_level(StatsLevel::kAll);
+  options.create_if_missing = true;
+  options.compression = kNoCompression;
+  options.inplace_update_support = false;
+  options.allow_concurrent_memtable_write = true;
+#ifndef ROCKSDB_LITE
+  TestFlushListener* listener = new TestFlushListener(options.env, this);
+  options.listeners.emplace_back(listener);
+#endif  // !ROCKSDB_LITE
+  // Enforce size of a single MemTable to 64MB (64MB = 67108864 bytes).
+  options.write_buffer_size = 1 << 20;
+  // Activate the MemPurge prototype.
+  options.experimental_mempurge_threshold = 1.0;
+
+  ASSERT_OK(TryReopen(options));
+
+  uint32_t mempurge_count = 0;
+  uint32_t sst_count = 0;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::FlushJob:MemPurgeSuccessful",
+      [&](void* /*arg*/) { mempurge_count++; });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::FlushJob:SSTFileCreated", [&](void* /*arg*/) { sst_count++; });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  std::string KEY1 = "ThisIsKey1";
+  std::string KEY2 = "ThisIsKey2";
+  std::string KEY3 = "ThisIsKey3";
+  std::string KEY4 = "ThisIsKey4";
+  std::string KEY5 = "ThisIsKey5";
+  const std::string NOT_FOUND = "NOT_FOUND";
+
+  Random rnd(117);
+  const size_t NUM_REPEAT = 100;
+  const size_t RAND_VALUES_LENGTH = 10240;
+
+  std::string key, value, p_v1, p_v2, p_v3, p_v3b, p_v4, p_v5;
+  int count = 0;
+  const int EXPECTED_COUNT_FORLOOP = 3;
+  const int EXPECTED_COUNT_END = 4;
+
+  ReadOptions ropt;
+  ropt.pin_data = true;
+  ropt.total_order_seek = true;
+  Iterator* iter = nullptr;
+
+  // Insertion of of K-V pairs, multiple times.
+  // Also insert DeleteRange
+  for (size_t i = 0; i < NUM_REPEAT; i++) {
+    // Create value strings of arbitrary length RAND_VALUES_LENGTH bytes.
+    p_v1 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v2 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v3 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v3b = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v4 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v5 = rnd.RandomString(RAND_VALUES_LENGTH);
+    ASSERT_OK(Put(KEY1, p_v1));
+    ASSERT_OK(Put(KEY2, p_v2));
+    ASSERT_OK(Put(KEY3, p_v3));
+    ASSERT_OK(Put(KEY4, p_v4));
+    ASSERT_OK(Put(KEY5, p_v5));
+    ASSERT_OK(Delete(KEY2));
+    ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), KEY2,
+                               KEY4));
+    ASSERT_OK(Put(KEY3, p_v3b));
+    ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), KEY1,
+                               KEY3));
+    ASSERT_OK(Delete(KEY1));
+
+    ASSERT_EQ(Get(KEY1), NOT_FOUND);
+    ASSERT_EQ(Get(KEY2), NOT_FOUND);
+    ASSERT_EQ(Get(KEY3), p_v3b);
+    ASSERT_EQ(Get(KEY4), p_v4);
+    ASSERT_EQ(Get(KEY5), p_v5);
+
+    iter = db_->NewIterator(ropt);
+    iter->SeekToFirst();
+    count = 0;
+    for (; iter->Valid(); iter->Next()) {
+      ASSERT_OK(iter->status());
+      key = (iter->key()).ToString(false);
+      value = (iter->value()).ToString(false);
+      if (key.compare(KEY3) == 0)
+        ASSERT_EQ(value, p_v3b);
+      else if (key.compare(KEY4) == 0)
+        ASSERT_EQ(value, p_v4);
+      else if (key.compare(KEY5) == 0)
+        ASSERT_EQ(value, p_v5);
+      else
+        ASSERT_EQ(value, NOT_FOUND);
+      count++;
+    }
+
+    // Expected count here is 3: KEY3, KEY4, KEY5.
+    ASSERT_EQ(count, EXPECTED_COUNT_FORLOOP);
+    if (iter) {
+      delete iter;
+    }
+  }
+
+  // Check that there was at least one mempurge
+  const uint32_t EXPECTED_MIN_MEMPURGE_COUNT = 1;
+  // Check that there was no SST files created during flush.
+  const uint32_t EXPECTED_SST_COUNT = 0;
+
+  EXPECT_GE(mempurge_count, EXPECTED_MIN_MEMPURGE_COUNT);
+  EXPECT_EQ(sst_count, EXPECTED_SST_COUNT);
+
+  // Additional test for the iterator+memPurge.
+  ASSERT_OK(Put(KEY2, p_v2));
+  iter = db_->NewIterator(ropt);
+  iter->SeekToFirst();
+  ASSERT_OK(Put(KEY4, p_v4));
+  count = 0;
+  for (; iter->Valid(); iter->Next()) {
+    ASSERT_OK(iter->status());
+    key = (iter->key()).ToString(false);
+    value = (iter->value()).ToString(false);
+    if (key.compare(KEY2) == 0)
+      ASSERT_EQ(value, p_v2);
+    else if (key.compare(KEY3) == 0)
+      ASSERT_EQ(value, p_v3b);
+    else if (key.compare(KEY4) == 0)
+      ASSERT_EQ(value, p_v4);
+    else if (key.compare(KEY5) == 0)
+      ASSERT_EQ(value, p_v5);
+    else
+      ASSERT_EQ(value, NOT_FOUND);
+    count++;
+  }
+
+  // Expected count here is 4: KEY2, KEY3, KEY4, KEY5.
+  ASSERT_EQ(count, EXPECTED_COUNT_END);
+  if (iter) delete iter;
+
+  Close();
+}
+
+// Create a Compaction Fitler that will be invoked
+// at flush time and will update the value of a KV pair
+// if the key string is "lower" than the filter_key_ string.
+class ConditionalUpdateFilter : public CompactionFilter {
+ public:
+  explicit ConditionalUpdateFilter(const std::string* filtered_key)
+      : filtered_key_(filtered_key) {}
+  bool Filter(int /*level*/, const Slice& key, const Slice& /*value*/,
+              std::string* new_value, bool* value_changed) const override {
+    // If key<filtered_key_, update the value of the KV-pair.
+    if (key.compare(*filtered_key_) < 0) {
+      assert(new_value != nullptr);
+      *new_value = NEW_VALUE;
+      *value_changed = true;
+    }
+    return false /*do not remove this KV-pair*/;
+  }
+
+  const char* Name() const override { return "ConditionalUpdateFilter"; }
+
+ private:
+  const std::string* filtered_key_;
+};
+
+class ConditionalUpdateFilterFactory : public CompactionFilterFactory {
+ public:
+  explicit ConditionalUpdateFilterFactory(const Slice& filtered_key)
+      : filtered_key_(filtered_key.ToString()) {}
+
+  std::unique_ptr<CompactionFilter> CreateCompactionFilter(
+      const CompactionFilter::Context& /*context*/) override {
+    return std::unique_ptr<CompactionFilter>(
+        new ConditionalUpdateFilter(&filtered_key_));
+  }
+
+  const char* Name() const override { return "ConditionalUpdateFilterFactory"; }
+
+  bool ShouldFilterTableFileCreation(
+      TableFileCreationReason reason) const override {
+    // This compaction filter will be invoked
+    // at flush time (and therefore at MemPurge time).
+    return (reason == TableFileCreationReason::kFlush);
+  }
+
+ private:
+  std::string filtered_key_;
+};
+
+TEST_F(DBFlushTest, MemPurgeAndCompactionFilter) {
+  Options options = CurrentOptions();
+
+  std::string KEY1 = "ThisIsKey1";
+  std::string KEY2 = "ThisIsKey2";
+  std::string KEY3 = "ThisIsKey3";
+  std::string KEY4 = "ThisIsKey4";
+  std::string KEY5 = "ThisIsKey5";
+  std::string KEY6 = "ThisIsKey6";
+  std::string KEY7 = "ThisIsKey7";
+  std::string KEY8 = "ThisIsKey8";
+  std::string KEY9 = "ThisIsKey9";
+  const std::string NOT_FOUND = "NOT_FOUND";
+
+  options.statistics = CreateDBStatistics();
+  options.statistics->set_stats_level(StatsLevel::kAll);
+  options.create_if_missing = true;
+  options.compression = kNoCompression;
+  options.inplace_update_support = false;
+  options.allow_concurrent_memtable_write = true;
+#ifndef ROCKSDB_LITE
+  TestFlushListener* listener = new TestFlushListener(options.env, this);
+  options.listeners.emplace_back(listener);
+#endif  // !ROCKSDB_LITE
+  // Create a ConditionalUpdate compaction filter
+  // that will update all the values of the KV pairs
+  // where the keys are "lower" than KEY4.
+  options.compaction_filter_factory =
+      std::make_shared<ConditionalUpdateFilterFactory>(KEY4);
+
+  // Enforce size of a single MemTable to 64MB (64MB = 67108864 bytes).
+  options.write_buffer_size = 1 << 20;
+  // Activate the MemPurge prototype.
+  options.experimental_mempurge_threshold = 1.0;
+
+  ASSERT_OK(TryReopen(options));
+
+  uint32_t mempurge_count = 0;
+  uint32_t sst_count = 0;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::FlushJob:MemPurgeSuccessful",
+      [&](void* /*arg*/) { mempurge_count++; });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::FlushJob:SSTFileCreated", [&](void* /*arg*/) { sst_count++; });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  Random rnd(53);
+  const size_t NUM_REPEAT = 1000;
+  const size_t RAND_VALUES_LENGTH = 10240;
+  std::string p_v1, p_v2, p_v3, p_v4, p_v5, p_v6, p_v7, p_v8, p_v9;
+
+  p_v1 = rnd.RandomString(RAND_VALUES_LENGTH);
+  p_v2 = rnd.RandomString(RAND_VALUES_LENGTH);
+  p_v3 = rnd.RandomString(RAND_VALUES_LENGTH);
+  p_v4 = rnd.RandomString(RAND_VALUES_LENGTH);
+  p_v5 = rnd.RandomString(RAND_VALUES_LENGTH);
+  ASSERT_OK(Put(KEY1, p_v1));
+  ASSERT_OK(Put(KEY2, p_v2));
+  ASSERT_OK(Put(KEY3, p_v3));
+  ASSERT_OK(Put(KEY4, p_v4));
+  ASSERT_OK(Put(KEY5, p_v5));
+  ASSERT_OK(Delete(KEY1));
+
+  // Insertion of of K-V pairs, multiple times.
+  for (size_t i = 0; i < NUM_REPEAT; i++) {
+    // Create value strings of arbitrary
+    // length RAND_VALUES_LENGTH bytes.
+    p_v6 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v7 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v8 = rnd.RandomString(RAND_VALUES_LENGTH);
+    p_v9 = rnd.RandomString(RAND_VALUES_LENGTH);
+    ASSERT_OK(Put(KEY6, p_v6));
+    ASSERT_OK(Put(KEY7, p_v7));
+    ASSERT_OK(Put(KEY8, p_v8));
+    ASSERT_OK(Put(KEY9, p_v9));
+
+    ASSERT_OK(Delete(KEY7));
+  }
+
+  // Check that there was at least one mempurge
+  const uint32_t EXPECTED_MIN_MEMPURGE_COUNT = 1;
+  // Check that there was no SST files created during flush.
+  const uint32_t EXPECTED_SST_COUNT = 0;
+
+  EXPECT_GE(mempurge_count, EXPECTED_MIN_MEMPURGE_COUNT);
+  EXPECT_EQ(sst_count, EXPECTED_SST_COUNT);
+
+  // Verify that the ConditionalUpdateCompactionFilter
+  // updated the values of KEY2 and KEY3, and not KEY4 and KEY5.
+  ASSERT_EQ(Get(KEY1), NOT_FOUND);
+  ASSERT_EQ(Get(KEY2), NEW_VALUE);
+  ASSERT_EQ(Get(KEY3), NEW_VALUE);
+  ASSERT_EQ(Get(KEY4), p_v4);
+  ASSERT_EQ(Get(KEY5), p_v5);
+}
+
+TEST_F(DBFlushTest, MemPurgeWALSupport) {
+  Options options = CurrentOptions();
+
+  options.statistics = CreateDBStatistics();
+  options.statistics->set_stats_level(StatsLevel::kAll);
+  options.create_if_missing = true;
+  options.compression = kNoCompression;
+  options.inplace_update_support = false;
+  options.allow_concurrent_memtable_write = true;
+
+  // Enforce size of a single MemTable to 128KB.
+  options.write_buffer_size = 128 << 10;
+  // Activate the MemPurge prototype.
+  options.experimental_mempurge_threshold = 1.0;
+
+  ASSERT_OK(TryReopen(options));
+
+  const size_t KVSIZE = 10;
+
+  do {
+    CreateAndReopenWithCF({"pikachu"}, options);
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_OK(Put(1, "baz", "v5"));
+
+    ReopenWithColumnFamilies({"default", "pikachu"}, options);
+    ASSERT_EQ("v1", Get(1, "foo"));
+
+    ASSERT_EQ("v1", Get(1, "foo"));
+    ASSERT_EQ("v5", Get(1, "baz"));
+    ASSERT_OK(Put(0, "bar", "v2"));
+    ASSERT_OK(Put(1, "bar", "v2"));
+    ASSERT_OK(Put(1, "foo", "v3"));
+    uint32_t mempurge_count = 0;
+    uint32_t sst_count = 0;
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+        "DBImpl::FlushJob:MemPurgeSuccessful",
+        [&](void* /*arg*/) { mempurge_count++; });
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+        "DBImpl::FlushJob:SSTFileCreated", [&](void* /*arg*/) { sst_count++; });
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+    std::vector<std::string> keys;
+    for (size_t k = 0; k < KVSIZE; k++) {
+      keys.push_back("IamKey" + std::to_string(k));
+    }
+
+    std::string RNDKEY, RNDVALUE;
+    const std::string NOT_FOUND = "NOT_FOUND";
+
+    // Heavy overwrite workload,
+    // more than would fit in maximum allowed memtables.
+    Random rnd(719);
+    const size_t NUM_REPEAT = 100;
+    const size_t RAND_KEY_LENGTH = 4096;
+    const size_t RAND_VALUES_LENGTH = 1024;
+    std::vector<std::string> values_default(KVSIZE), values_pikachu(KVSIZE);
+
+    // Insert a very first set of keys that will be
+    // mempurged at least once.
+    for (size_t k = 0; k < KVSIZE / 2; k++) {
+      values_default[k] = rnd.RandomString(RAND_VALUES_LENGTH);
+      values_pikachu[k] = rnd.RandomString(RAND_VALUES_LENGTH);
+    }
+
+    // Insert keys[0:KVSIZE/2] to
+    // both 'default' and 'pikachu' CFs.
+    for (size_t k = 0; k < KVSIZE / 2; k++) {
+      ASSERT_OK(Put(0, keys[k], values_default[k]));
+      ASSERT_OK(Put(1, keys[k], values_pikachu[k]));
+    }
+
+    // Check that the insertion was seamless.
+    for (size_t k = 0; k < KVSIZE / 2; k++) {
+      ASSERT_EQ(Get(0, keys[k]), values_default[k]);
+      ASSERT_EQ(Get(1, keys[k]), values_pikachu[k]);
+    }
+
+    // Insertion of of K-V pairs, multiple times (overwrites)
+    // into 'default' CF. Will trigger mempurge.
+    for (size_t j = 0; j < NUM_REPEAT; j++) {
+      // Create value strings of arbitrary length RAND_VALUES_LENGTH bytes.
+      for (size_t k = KVSIZE / 2; k < KVSIZE; k++) {
+        values_default[k] = rnd.RandomString(RAND_VALUES_LENGTH);
+      }
+
+      // Insert K-V into default CF.
+      for (size_t k = KVSIZE / 2; k < KVSIZE; k++) {
+        ASSERT_OK(Put(0, keys[k], values_default[k]));
+      }
+
+      // Check key validity, for all keys, both in
+      // default and pikachu CFs.
+      for (size_t k = 0; k < KVSIZE; k++) {
+        ASSERT_EQ(Get(0, keys[k]), values_default[k]);
+      }
+      // Note that at this point, only keys[0:KVSIZE/2]
+      // have been inserted into Pikachu.
+      for (size_t k = 0; k < KVSIZE / 2; k++) {
+        ASSERT_EQ(Get(1, keys[k]), values_pikachu[k]);
+      }
+    }
+
+    // Insertion of of K-V pairs, multiple times (overwrites)
+    // into 'pikachu' CF. Will trigger mempurge.
+    // Check that we keep the older logs for 'default' imm().
+    for (size_t j = 0; j < NUM_REPEAT; j++) {
+      // Create value strings of arbitrary length RAND_VALUES_LENGTH bytes.
+      for (size_t k = KVSIZE / 2; k < KVSIZE; k++) {
+        values_pikachu[k] = rnd.RandomString(RAND_VALUES_LENGTH);
+      }
+
+      // Insert K-V into pikachu CF.
+      for (size_t k = KVSIZE / 2; k < KVSIZE; k++) {
+        ASSERT_OK(Put(1, keys[k], values_pikachu[k]));
+      }
+
+      // Check key validity, for all keys,
+      // both in default and pikachu.
+      for (size_t k = 0; k < KVSIZE; k++) {
+        ASSERT_EQ(Get(0, keys[k]), values_default[k]);
+        ASSERT_EQ(Get(1, keys[k]), values_pikachu[k]);
+      }
+    }
+
+    // Check that there was at least one mempurge
+    const uint32_t EXPECTED_MIN_MEMPURGE_COUNT = 1;
+    // Check that there was no SST files created during flush.
+    const uint32_t EXPECTED_SST_COUNT = 0;
+
+    EXPECT_GE(mempurge_count, EXPECTED_MIN_MEMPURGE_COUNT);
+    if (options.experimental_mempurge_threshold ==
+        std::numeric_limits<double>::max()) {
+      EXPECT_EQ(sst_count, EXPECTED_SST_COUNT);
+    }
+
+    ReopenWithColumnFamilies({"default", "pikachu"}, options);
+    // Check that there was no data corruption anywhere,
+    // not in 'default' nor in 'Pikachu' CFs.
+    ASSERT_EQ("v3", Get(1, "foo"));
+    ASSERT_OK(Put(1, "foo", "v4"));
+    ASSERT_EQ("v4", Get(1, "foo"));
+    ASSERT_EQ("v2", Get(1, "bar"));
+    ASSERT_EQ("v5", Get(1, "baz"));
+    // Check keys in 'Default' and 'Pikachu'.
+    // keys[0:KVSIZE/2] were for sure contained
+    // in the imm() at Reopen/recovery time.
+    for (size_t k = 0; k < KVSIZE; k++) {
+      ASSERT_EQ(Get(0, keys[k]), values_default[k]);
+      ASSERT_EQ(Get(1, keys[k]), values_pikachu[k]);
+    }
+    // Insertion of random K-V pairs to trigger
+    // a flush in the Pikachu CF.
+    for (size_t j = 0; j < NUM_REPEAT; j++) {
+      RNDKEY = rnd.RandomString(RAND_KEY_LENGTH);
+      RNDVALUE = rnd.RandomString(RAND_VALUES_LENGTH);
+      ASSERT_OK(Put(1, RNDKEY, RNDVALUE));
+    }
+    // ASsert than there was at least one flush to storage.
+    EXPECT_GT(sst_count, EXPECTED_SST_COUNT);
+    ReopenWithColumnFamilies({"default", "pikachu"}, options);
+    ASSERT_EQ("v4", Get(1, "foo"));
+    ASSERT_EQ("v2", Get(1, "bar"));
+    ASSERT_EQ("v5", Get(1, "baz"));
+    // Since values in default are held in mutable mem()
+    // and imm(), check if the flush in pikachu didn't
+    // affect these values.
+    for (size_t k = 0; k < KVSIZE; k++) {
+      ASSERT_EQ(Get(0, keys[k]), values_default[k]);
+      ASSERT_EQ(Get(1, keys[k]), values_pikachu[k]);
+    }
+    ASSERT_EQ(Get(1, RNDKEY), RNDVALUE);
+  } while (ChangeWalOptions());
+}
+
 TEST_P(DBFlushDirectIOTest, DirectIO) {
   Options options;
   options.create_if_missing = true;

db/db_impl/db_impl.cc

@@ -63,6 +63,7 @@
 #include "memtable/hash_linklist_rep.h"
 #include "memtable/hash_skiplist_rep.h"
 #include "monitoring/in_memory_stats_history.h"
+#include "monitoring/instrumented_mutex.h"
 #include "monitoring/iostats_context_imp.h"
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/persistent_stats_history.h"
@@ -93,14 +94,17 @@
 #include "table/table_builder.h"
 #include "table/two_level_iterator.h"
 #include "test_util/sync_point.h"
+#include "trace_replay/trace_replay.h"
 #include "util/autovector.h"
 #include "util/cast_util.h"
 #include "util/coding.h"
 #include "util/compression.h"
 #include "util/crc32c.h"
+#include "util/defer.h"
 #include "util/mutexlock.h"
 #include "util/stop_watch.h"
 #include "util/string_util.h"
+#include "utilities/trace/replayer_impl.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -548,12 +552,45 @@ Status DBImpl::CloseHelper() {
   flush_scheduler_.Clear();
   trim_history_scheduler_.Clear();
 
+  // For now, simply trigger a manual flush at close time
+  // on all the column families.
+  // TODO(bjlemaire): Check if this is needed. Also, in the
+  // future we can contemplate doing a more fine-grained
+  // flushing by first checking if there is a need for
+  // flushing (but need to implement something
+  // else than imm()->IsFlushPending() because the output
+  // memtables added to imm() dont trigger flushes).
+  if (immutable_db_options_.experimental_mempurge_threshold > 0.0) {
+    Status flush_ret;
+    mutex_.Unlock();
+    for (ColumnFamilyData* cf : *versions_->GetColumnFamilySet()) {
+      if (immutable_db_options_.atomic_flush) {
+        flush_ret = AtomicFlushMemTables({cf}, FlushOptions(),
+                                         FlushReason::kManualFlush);
+        if (!flush_ret.ok()) {
+          ROCKS_LOG_INFO(
+              immutable_db_options_.info_log,
+              "Atomic flush memtables failed upon closing (mempurge).");
+        }
+      } else {
+        flush_ret =
+            FlushMemTable(cf, FlushOptions(), FlushReason::kManualFlush);
+        if (!flush_ret.ok()) {
+          ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                         "Flush memtables failed upon closing (mempurge).");
+        }
+      }
+    }
+    mutex_.Lock();
+  }
+
   while (!flush_queue_.empty()) {
     const FlushRequest& flush_req = PopFirstFromFlushQueue();
     for (const auto& iter : flush_req) {
       iter.first->UnrefAndTryDelete();
     }
   }
+
   while (!compaction_queue_.empty()) {
     auto cfd = PopFirstFromCompactionQueue();
     cfd->UnrefAndTryDelete();
@@ -606,7 +643,7 @@ Status DBImpl::CloseHelper() {
       ROCKS_LOG_WARN(
           immutable_db_options_.info_log,
           "Unable to Sync WAL file %s with error -- %s",
-          LogFileName(immutable_db_options_.wal_dir, log_number).c_str(),
+          LogFileName(immutable_db_options_.GetWalDir(), log_number).c_str(),
           s.ToString().c_str());
       // Retain the first error
       if (ret.ok()) {
@@ -700,7 +737,8 @@ void DBImpl::MaybeIgnoreError(Status* s) const {
 const Status DBImpl::CreateArchivalDirectory() {
   if (immutable_db_options_.WAL_ttl_seconds > 0 ||
       immutable_db_options_.WAL_size_limit_MB > 0) {
-    std::string archivalPath = ArchivalDirectory(immutable_db_options_.wal_dir);
+    std::string archivalPath =
+        ArchivalDirectory(immutable_db_options_.GetWalDir());
     return env_->CreateDirIfMissing(archivalPath);
   }
   return Status::OK();
@@ -911,18 +949,31 @@ void DBImpl::DumpStats() {
   if (shutdown_initiated_) {
     return;
   }
+
   TEST_SYNC_POINT("DBImpl::DumpStats:StartRunning");
   {
     InstrumentedMutexLock l(&mutex_);
+    for (auto cfd : *versions_->GetColumnFamilySet()) {
+      if (cfd->initialized()) {
+        // Release DB mutex for gathering cache entry stats. Pass over all
+        // column families for this first so that other stats are dumped
+        // near-atomically.
+        InstrumentedMutexUnlock u(&mutex_);
+        cfd->internal_stats()->CollectCacheEntryStats(/*foreground=*/false);
+      }
+    }
+
     const std::string* property = &DB::Properties::kDBStats;
     const DBPropertyInfo* property_info = GetPropertyInfo(*property);
     assert(property_info != nullptr);
+    assert(!property_info->need_out_of_mutex);
     default_cf_internal_stats_->GetStringProperty(*property_info, *property,
                                                   &stats);
 
     property = &DB::Properties::kCFStatsNoFileHistogram;
     property_info = GetPropertyInfo(*property);
     assert(property_info != nullptr);
+    assert(!property_info->need_out_of_mutex);
     for (auto cfd : *versions_->GetColumnFamilySet()) {
       if (cfd->initialized()) {
         cfd->internal_stats()->GetStringProperty(*property_info, *property,
@@ -933,6 +984,7 @@ void DBImpl::DumpStats() {
     property = &DB::Properties::kCFFileHistogram;
     property_info = GetPropertyInfo(*property);
     assert(property_info != nullptr);
+    assert(!property_info->need_out_of_mutex);
     for (auto cfd : *versions_->GetColumnFamilySet()) {
       if (cfd->initialized()) {
         cfd->internal_stats()->GetStringProperty(*property_info, *property,
@@ -1088,9 +1140,19 @@ Status DBImpl::SetDBOptions(
     InstrumentedMutexLock l(&mutex_);
     s = GetMutableDBOptionsFromStrings(mutable_db_options_, options_map,
                                        &new_options);
+
     if (new_options.bytes_per_sync == 0) {
       new_options.bytes_per_sync = 1024 * 1024;
     }
+
+    if (MutableDBOptionsAreEqual(mutable_db_options_, new_options)) {
+      ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                     "SetDBOptions(), input option value is not changed, "
+                     "skipping updating.");
+      persist_options_status.PermitUncheckedError();
+      return s;
+    }
+
     DBOptions new_db_options =
         BuildDBOptions(immutable_db_options_, new_options);
     if (s.ok()) {
@@ -1736,6 +1798,8 @@ Status DBImpl::GetImpl(const ReadOptions& read_options, const Slice& key,
   }
   // If timestamp is used, we use read callback to ensure <key,t,s> is returned
   // only if t <= read_opts.timestamp and s <= snapshot.
+  // HACK: temporarily overwrite input struct field but restore
+  SaveAndRestore<ReadCallback*> restore_callback(&get_impl_options.callback);
   if (ts_sz > 0) {
     assert(!get_impl_options
                 .callback);  // timestamp with callback is not supported
@@ -2228,20 +2292,18 @@ void DBImpl::MultiGet(const ReadOptions& read_options, const size_t num_keys,
       multiget_cf_data;
   size_t cf_start = 0;
   ColumnFamilyHandle* cf = sorted_keys[0]->column_family;
+
   for (size_t i = 0; i < num_keys; ++i) {
     KeyContext* key_ctx = sorted_keys[i];
     if (key_ctx->column_family != cf) {
-      multiget_cf_data.emplace_back(
-          MultiGetColumnFamilyData(cf, cf_start, i - cf_start, nullptr));
+      multiget_cf_data.emplace_back(cf, cf_start, i - cf_start, nullptr);
       cf_start = i;
       cf = key_ctx->column_family;
     }
   }
-  {
-    // multiget_cf_data.emplace_back(
-    // MultiGetColumnFamilyData(cf, cf_start, num_keys - cf_start, nullptr));
-    multiget_cf_data.emplace_back(cf, cf_start, num_keys - cf_start, nullptr);
-  }
+
+  multiget_cf_data.emplace_back(cf, cf_start, num_keys - cf_start, nullptr);
+
   std::function<MultiGetColumnFamilyData*(
       autovector<MultiGetColumnFamilyData,
                  MultiGetContext::MAX_BATCH_SIZE>::iterator&)>
@@ -2301,7 +2363,7 @@ struct CompareKeyContext {
         static_cast<ColumnFamilyHandleImpl*>(lhs->column_family);
     uint32_t cfd_id1 = cfh->cfd()->GetID();
     const Comparator* comparator = cfh->cfd()->user_comparator();
-    cfh = static_cast<ColumnFamilyHandleImpl*>(lhs->column_family);
+    cfh = static_cast<ColumnFamilyHandleImpl*>(rhs->column_family);
     uint32_t cfd_id2 = cfh->cfd()->GetID();
 
     if (cfd_id1 < cfd_id2) {
@@ -2325,39 +2387,24 @@ struct CompareKeyContext {
 void DBImpl::PrepareMultiGetKeys(
     size_t num_keys, bool sorted_input,
     autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* sorted_keys) {
-#ifndef NDEBUG
   if (sorted_input) {
-    for (size_t index = 0; index < sorted_keys->size(); ++index) {
-      if (index > 0) {
-        KeyContext* lhs = (*sorted_keys)[index - 1];
-        KeyContext* rhs = (*sorted_keys)[index];
-        ColumnFamilyHandleImpl* cfh =
-            static_cast_with_check<ColumnFamilyHandleImpl>(lhs->column_family);
-        uint32_t cfd_id1 = cfh->cfd()->GetID();
-        const Comparator* comparator = cfh->cfd()->user_comparator();
-        cfh =
-            static_cast_with_check<ColumnFamilyHandleImpl>(lhs->column_family);
-        uint32_t cfd_id2 = cfh->cfd()->GetID();
+#ifndef NDEBUG
+    CompareKeyContext key_context_less;
 
-        assert(cfd_id1 <= cfd_id2);
-        if (cfd_id1 < cfd_id2) {
-          continue;
-        }
+    for (size_t index = 1; index < sorted_keys->size(); ++index) {
+      const KeyContext* const lhs = (*sorted_keys)[index - 1];
+      const KeyContext* const rhs = (*sorted_keys)[index];
 
-        // Both keys are from the same column family
-        int cmp = comparator->CompareWithoutTimestamp(
-            *(lhs->key), /*a_has_ts=*/false, *(rhs->key), /*b_has_ts=*/false);
-        assert(cmp <= 0);
-      }
-      index++;
+      // lhs should be <= rhs, or in other words, rhs should NOT be < lhs
+      assert(!key_context_less(rhs, lhs));
     }
-  }
 #endif
-  if (!sorted_input) {
-    CompareKeyContext sort_comparator;
-    std::sort(sorted_keys->begin(), sorted_keys->begin() + num_keys,
-              sort_comparator);
+
+    return;
   }
+
+  std::sort(sorted_keys->begin(), sorted_keys->begin() + num_keys,
+            CompareKeyContext());
 }
 
 void DBImpl::MultiGet(const ReadOptions& read_options,
@@ -3231,16 +3278,21 @@ bool DBImpl::GetProperty(ColumnFamilyHandle* column_family,
     }
     return ret_value;
   } else if (property_info->handle_string) {
-    InstrumentedMutexLock l(&mutex_);
-    return cfd->internal_stats()->GetStringProperty(*property_info, property,
-                                                    value);
-  } else if (property_info->handle_string_dbimpl) {
-    std::string tmp_value;
-    bool ret_value = (this->*(property_info->handle_string_dbimpl))(&tmp_value);
-    if (ret_value) {
-      *value = tmp_value;
+    if (property_info->need_out_of_mutex) {
+      return cfd->internal_stats()->GetStringProperty(*property_info, property,
+                                                      value);
+    } else {
+      InstrumentedMutexLock l(&mutex_);
+      return cfd->internal_stats()->GetStringProperty(*property_info, property,
+                                                      value);
+    }
+  } else if (property_info->handle_string_dbimpl) {
+    if (property_info->need_out_of_mutex) {
+      return (this->*(property_info->handle_string_dbimpl))(value);
+    } else {
+      InstrumentedMutexLock l(&mutex_);
+      return (this->*(property_info->handle_string_dbimpl))(value);
     }
-    return ret_value;
   }
   // Shouldn't reach here since exactly one of handle_string and handle_int
   // should be non-nullptr.
@@ -3258,9 +3310,14 @@ bool DBImpl::GetMapProperty(ColumnFamilyHandle* column_family,
   if (property_info == nullptr) {
     return false;
   } else if (property_info->handle_map) {
-    InstrumentedMutexLock l(&mutex_);
-    return cfd->internal_stats()->GetMapProperty(*property_info, property,
-                                                 value);
+    if (property_info->need_out_of_mutex) {
+      return cfd->internal_stats()->GetMapProperty(*property_info, property,
+                                                   value);
+    } else {
+      InstrumentedMutexLock l(&mutex_);
+      return cfd->internal_stats()->GetMapProperty(*property_info, property,
+                                                   value);
+    }
   }
   // If we reach this point it means that handle_map is not provided for the
   // requested property
@@ -3709,6 +3766,8 @@ Status DBImpl::DeleteFilesInRanges(ColumnFamilyHandle* column_family,
           deleted_files.insert(level_file);
           level_file->being_compacted = true;
         }
+        vstorage->ComputeCompactionScore(*cfd->ioptions(),
+                                         *cfd->GetLatestMutableCFOptions());
       }
     }
     if (edit.GetDeletedFiles().empty()) {
@@ -3772,6 +3831,17 @@ void DBImpl::GetColumnFamilyMetaData(ColumnFamilyHandle* column_family,
   ReturnAndCleanupSuperVersion(cfd, sv);
 }
 
+void DBImpl::GetAllColumnFamilyMetaData(
+    std::vector<ColumnFamilyMetaData>* metadata) {
+  InstrumentedMutexLock l(&mutex_);
+  for (auto cfd : *(versions_->GetColumnFamilySet())) {
+    {
+      metadata->emplace_back();
+      cfd->current()->GetColumnFamilyMetaData(&metadata->back());
+    }
+  }
+}
+
 #endif  // ROCKSDB_LITE
 
 Status DBImpl::CheckConsistency() {
@@ -3875,10 +3945,10 @@ Status DBImpl::GetDbSessionId(std::string& session_id) const {
   return Status::OK();
 }
 
-void DBImpl::SetDbSessionId() {
+std::string DBImpl::GenerateDbSessionId(Env* env) {
   // GenerateUniqueId() generates an identifier that has a negligible
   // probability of being duplicated, ~128 bits of entropy
-  std::string uuid = env_->GenerateUniqueId();
+  std::string uuid = env->GenerateUniqueId();
 
   // Hash and reformat that down to a more compact format, 20 characters
   // in base-36 ([0-9A-Z]), which is ~103 bits of entropy, which is enough
@@ -3889,15 +3959,21 @@ void DBImpl::SetDbSessionId() {
   // * Visually distinct from DB id format
   uint64_t a = NPHash64(uuid.data(), uuid.size(), 1234U);
   uint64_t b = NPHash64(uuid.data(), uuid.size(), 5678U);
-  db_session_id_.resize(20);
+  std::string db_session_id;
+  db_session_id.resize(20);
   static const char* const base36 = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ";
   size_t i = 0;
   for (; i < 10U; ++i, a /= 36U) {
-    db_session_id_[i] = base36[a % 36];
+    db_session_id[i] = base36[a % 36];
   }
   for (; i < 20U; ++i, b /= 36U) {
-    db_session_id_[i] = base36[b % 36];
+    db_session_id[i] = base36[b % 36];
   }
+  return db_session_id;
+}
+
+void DBImpl::SetDbSessionId() {
+  db_session_id_ = GenerateDbSessionId(env_);
   TEST_SYNC_POINT_CALLBACK("DBImpl::SetDbSessionId", &db_session_id_);
 }
 
@@ -3967,7 +4043,7 @@ Status DestroyDB(const std::string& dbname, const Options& options,
   ImmutableDBOptions soptions(SanitizeOptions(dbname, options));
   Env* env = soptions.env;
   std::vector<std::string> filenames;
-  bool wal_in_db_path = IsWalDirSameAsDBPath(&soptions);
+  bool wal_in_db_path = soptions.IsWalDirSameAsDBPath();
 
   // Reset the logger because it holds a handle to the
   // log file and prevents cleanup and directory removal
@@ -4033,7 +4109,7 @@ Status DestroyDB(const std::string& dbname, const Options& options,
     std::vector<std::string> walDirFiles;
     std::string archivedir = ArchivalDirectory(dbname);
     bool wal_dir_exists = false;
-    if (dbname != soptions.wal_dir) {
+    if (!soptions.IsWalDirSameAsDBPath(dbname)) {
       wal_dir_exists = env->GetChildren(soptions.wal_dir, &walDirFiles).ok();
       archivedir = ArchivalDirectory(soptions.wal_dir);
     }
@@ -4123,6 +4199,8 @@ Status DBImpl::WriteOptionsFile(bool need_mutex_lock,
 
   TEST_SYNC_POINT("DBImpl::WriteOptionsFile:1");
   TEST_SYNC_POINT("DBImpl::WriteOptionsFile:2");
+  TEST_SYNC_POINT_CALLBACK("DBImpl::WriteOptionsFile:PersistOptions",
+                           &db_options);
 
   std::string file_name =
       TempOptionsFileName(GetName(), versions_->NewFileNumber());
@@ -4289,9 +4367,7 @@ SequenceNumber DBImpl::GetEarliestMemTableSequenceNumber(SuperVersion* sv,
 
   return earliest_seq;
 }
-#endif  // ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE
 Status DBImpl::GetLatestSequenceForKey(SuperVersion* sv, const Slice& key,
                                        bool cache_only,
                                        SequenceNumber lower_bound_seq,
@@ -5033,11 +5109,19 @@ Status DBImpl::EndTrace() {
     s = tracer_->Close();
     tracer_.reset();
   } else {
-    return Status::IOError("No trace file to close");
+    s = Status::IOError("No trace file to close");
   }
   return s;
 }
 
+Status DBImpl::NewDefaultReplayer(
+    const std::vector<ColumnFamilyHandle*>& handles,
+    std::unique_ptr<TraceReader>&& reader,
+    std::unique_ptr<Replayer>* replayer) {
+  replayer->reset(new ReplayerImpl(this, handles, std::move(reader)));
+  return Status::OK();
+}
+
 Status DBImpl::StartBlockCacheTrace(
     const TraceOptions& trace_options,
     std::unique_ptr<TraceWriter>&& trace_writer) {

db/db_impl/db_impl.h

@@ -20,6 +20,7 @@
 #include <vector>
 
 #include "db/column_family.h"
+#include "db/compaction/compaction_iterator.h"
 #include "db/compaction/compaction_job.h"
 #include "db/dbformat.h"
 #include "db/error_handler.h"
@@ -50,9 +51,15 @@
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/status.h"
+#ifndef ROCKSDB_LITE
 #include "rocksdb/trace_reader_writer.h"
+#endif  // ROCKSDB_LITE
 #include "rocksdb/transaction_log.h"
+#ifndef ROCKSDB_LITE
+#include "rocksdb/utilities/replayer.h"
+#endif  // ROCKSDB_LITE
 #include "rocksdb/write_buffer_manager.h"
+#include "table/merging_iterator.h"
 #include "table/scoped_arena_iterator.h"
 #include "util/autovector.h"
 #include "util/hash.h"
@@ -400,12 +407,13 @@ class DBImpl : public DB {
       FileChecksumList* checksum_list) override;
 
   // Obtains the meta data of the specified column family of the DB.
-  // Status::NotFound() will be returned if the current DB does not have
-  // any column family match the specified name.
   // TODO(yhchiang): output parameter is placed in the end in this codebase.
   virtual void GetColumnFamilyMetaData(ColumnFamilyHandle* column_family,
                                        ColumnFamilyMetaData* metadata) override;
 
+  void GetAllColumnFamilyMetaData(
+      std::vector<ColumnFamilyMetaData>* metadata) override;
+
   Status SuggestCompactRange(ColumnFamilyHandle* column_family,
                              const Slice* begin, const Slice* end) override;
 
@@ -461,6 +469,12 @@ class DBImpl : public DB {
   using DB::EndTrace;
   virtual Status EndTrace() override;
 
+  using DB::NewDefaultReplayer;
+  virtual Status NewDefaultReplayer(
+      const std::vector<ColumnFamilyHandle*>& handles,
+      std::unique_ptr<TraceReader>&& reader,
+      std::unique_ptr<Replayer>* replayer) override;
+
   using DB::StartBlockCacheTrace;
   Status StartBlockCacheTrace(
       const TraceOptions& options,
@@ -962,7 +976,7 @@ class DBImpl : public DB {
 
   // Return the maximum overlapping data (in bytes) at next level for any
   // file at a level >= 1.
-  int64_t TEST_MaxNextLevelOverlappingBytes(
+  uint64_t TEST_MaxNextLevelOverlappingBytes(
       ColumnFamilyHandle* column_family = nullptr);
 
   // Return the current manifest file no.
@@ -974,8 +988,10 @@ class DBImpl : public DB {
   // get total level0 file size. Only for testing.
   uint64_t TEST_GetLevel0TotalSize();
 
-  void TEST_GetFilesMetaData(ColumnFamilyHandle* column_family,
-                             std::vector<std::vector<FileMetaData>>* metadata);
+  void TEST_GetFilesMetaData(
+      ColumnFamilyHandle* column_family,
+      std::vector<std::vector<FileMetaData>>* metadata,
+      std::vector<std::shared_ptr<BlobFileMetaData>>* blob_metadata = nullptr);
 
   void TEST_LockMutex();
 
@@ -1098,6 +1114,8 @@ class DBImpl : public DB {
     State state_;
   };
 
+  static std::string GenerateDbSessionId(Env* env);
+
  protected:
   const std::string dbname_;
   std::string db_id_;
@@ -1184,11 +1202,6 @@ class DBImpl : public DB {
                               ColumnFamilyHandle* column_family,
                               const Slice* begin, const Slice* end);
 
-  Status GetApproximateSizesInternal(const SizeApproximationOptions& options,
-                                     ColumnFamilyHandle* column_family,
-                                     const Range* range, int n,
-                                     uint64_t* sizes);
-
   // The following two functions can only be called when:
   // 1. WriteThread::Writer::EnterUnbatched() is used.
   // 2. db_mutex is NOT held

db/db_impl/db_impl_compaction_flush.cc

@@ -7,6 +7,7 @@
 // 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.
 #include <cinttypes>
+#include <deque>
 
 #include "db/builder.h"
 #include "db/db_impl/db_impl.h"
@@ -151,7 +152,6 @@ Status DBImpl::FlushMemTableToOutputFile(
   assert(cfd);
   assert(cfd->imm()->NumNotFlushed() != 0);
   assert(cfd->imm()->IsFlushPending());
-
   FlushJob flush_job(
       dbname_, cfd, immutable_db_options_, mutable_cf_options,
       port::kMaxUint64 /* memtable_id */, file_options_for_compaction_,
@@ -199,7 +199,7 @@ Status DBImpl::FlushMemTableToOutputFile(
     need_cancel = true;
   }
   TEST_SYNC_POINT("DBImpl::FlushMemTableToOutputFile:AfterPickMemtables");
-
+  bool switched_to_mempurge = false;
   // Within flush_job.Run, rocksdb may call event listener to notify
   // file creation and deletion.
   //
@@ -207,7 +207,8 @@ Status DBImpl::FlushMemTableToOutputFile(
   // and EventListener callback will be called when the db_mutex
   // is unlocked by the current thread.
   if (s.ok()) {
-    s = flush_job.Run(&logs_with_prep_tracker_, &file_meta);
+    s = flush_job.Run(&logs_with_prep_tracker_, &file_meta,
+                      &switched_to_mempurge);
     need_cancel = false;
   }
 
@@ -283,7 +284,9 @@ Status DBImpl::FlushMemTableToOutputFile(
     // from never needing it or ignoring the flush job status
     io_s.PermitUncheckedError();
   }
-  if (s.ok()) {
+  // If flush ran smoothly and no mempurge happened
+  // install new SST file path.
+  if (s.ok() && (!switched_to_mempurge)) {
 #ifndef ROCKSDB_LITE
     // may temporarily unlock and lock the mutex.
     NotifyOnFlushCompleted(cfd, mutable_cf_options,
@@ -408,10 +411,13 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
         stats_, &event_logger_, mutable_cf_options.report_bg_io_stats,
         false /* sync_output_directory */, false /* write_manifest */,
         thread_pri, io_tracer_, db_id_, db_session_id_,
-        cfd->GetFullHistoryTsLow()));
+        cfd->GetFullHistoryTsLow(), &blob_callback_));
   }
 
   std::vector<FileMetaData> file_meta(num_cfs);
+  // Use of deque<bool> because vector<bool>
+  // is specific and doesn't allow &v[i].
+  std::deque<bool> switched_to_mempurge(num_cfs, false);
   Status s;
   IOStatus log_io_s = IOStatus::OK();
   assert(num_cfs == static_cast<int>(jobs.size()));
@@ -461,10 +467,13 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
   }
 
   if (s.ok()) {
+    assert(switched_to_mempurge.size() ==
+           static_cast<long unsigned int>(num_cfs));
     // TODO (yanqin): parallelize jobs with threads.
     for (int i = 1; i != num_cfs; ++i) {
       exec_status[i].second =
-          jobs[i]->Run(&logs_with_prep_tracker_, &file_meta[i]);
+          jobs[i]->Run(&logs_with_prep_tracker_, &file_meta[i],
+                       &(switched_to_mempurge.at(i)));
       exec_status[i].first = true;
       io_status[i] = jobs[i]->io_status();
     }
@@ -476,8 +485,9 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     }
     assert(exec_status.size() > 0);
     assert(!file_meta.empty());
-    exec_status[0].second =
-        jobs[0]->Run(&logs_with_prep_tracker_, &file_meta[0]);
+    exec_status[0].second = jobs[0]->Run(
+        &logs_with_prep_tracker_, file_meta.data() /* &file_meta[0] */,
+        switched_to_mempurge.empty() ? nullptr : &(switched_to_mempurge.at(0)));
     exec_status[0].first = true;
     io_status[0] = jobs[0]->io_status();
 
@@ -591,6 +601,8 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     autovector<const autovector<MemTable*>*> mems_list;
     autovector<const MutableCFOptions*> mutable_cf_options_list;
     autovector<FileMetaData*> tmp_file_meta;
+    autovector<std::list<std::unique_ptr<FlushJobInfo>>*>
+        committed_flush_jobs_info;
     for (int i = 0; i != num_cfs; ++i) {
       const auto& mems = jobs[i]->GetMemTables();
       if (!cfds[i]->IsDropped() && !mems.empty()) {
@@ -598,13 +610,18 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
         mems_list.emplace_back(&mems);
         mutable_cf_options_list.emplace_back(&all_mutable_cf_options[i]);
         tmp_file_meta.emplace_back(&file_meta[i]);
+#ifndef ROCKSDB_LITE
+        committed_flush_jobs_info.emplace_back(
+            jobs[i]->GetCommittedFlushJobsInfo());
+#endif  //! ROCKSDB_LITE
       }
     }
 
     s = InstallMemtableAtomicFlushResults(
         nullptr /* imm_lists */, tmp_cfds, mutable_cf_options_list, mems_list,
         versions_.get(), &logs_with_prep_tracker_, &mutex_, tmp_file_meta,
-        &job_context->memtables_to_free, directories_.GetDbDir(), log_buffer);
+        committed_flush_jobs_info, &job_context->memtables_to_free,
+        directories_.GetDbDir(), log_buffer);
   }
 
   if (s.ok()) {
@@ -650,6 +667,11 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
         immutable_db_options_.sst_file_manager.get());
     assert(all_mutable_cf_options.size() == static_cast<size_t>(num_cfs));
     for (int i = 0; s.ok() && i != num_cfs; ++i) {
+      // If mempurge happened instead of Flush,
+      // no NotifyOnFlushCompleted call (no SST file created).
+      if (switched_to_mempurge[i]) {
+        continue;
+      }
       if (cfds[i]->IsDropped()) {
         continue;
       }
@@ -1258,7 +1280,7 @@ Status DBImpl::CompactFilesImpl(
       c->mutable_cf_options()->report_bg_io_stats, dbname_,
       &compaction_job_stats, Env::Priority::USER, io_tracer_,
       &manual_compaction_paused_, nullptr, db_id_, db_session_id_,
-      c->column_family_data()->GetFullHistoryTsLow());
+      c->column_family_data()->GetFullHistoryTsLow(), &blob_callback_);
 
   // Creating a compaction influences the compaction score because the score
   // takes running compactions into account (by skipping files that are already
@@ -1395,6 +1417,8 @@ void DBImpl::NotifyOnCompactionBegin(ColumnFamilyData* cfd, Compaction* c,
       manual_compaction_paused_.load(std::memory_order_acquire) > 0) {
     return;
   }
+
+  c->SetNotifyOnCompactionCompleted();
   Version* current = cfd->current();
   current->Ref();
   // release lock while notifying events
@@ -1430,10 +1454,8 @@ void DBImpl::NotifyOnCompactionCompleted(
   if (shutting_down_.load(std::memory_order_acquire)) {
     return;
   }
-  // TODO: Should disabling manual compaction squash compaction completed
-  //   notifications that aren't the result of a shutdown?
-  if (c->is_manual_compaction() &&
-      manual_compaction_paused_.load(std::memory_order_acquire) > 0) {
+
+  if (c->ShouldNotifyOnCompactionCompleted() == false) {
     return;
   }
 
@@ -2398,6 +2420,17 @@ void DBImpl::SchedulePendingFlush(const FlushRequest& flush_req,
     assert(flush_req.size() == 1);
     ColumnFamilyData* cfd = flush_req[0].first;
     assert(cfd);
+    // Note: SchedulePendingFlush is always preceded
+    // with an imm()->FlushRequested() call. However,
+    // we want to make this code snipper more resilient to
+    // future changes. Therefore, we add the following if
+    // statement - note that calling it twice (or more)
+    // doesn't break anything.
+    if (immutable_db_options_.experimental_mempurge_threshold > 0.0) {
+      // If imm() contains silent memtables,
+      // requesting a flush will mark the imm_needed as true.
+      cfd->imm()->FlushRequested();
+    }
     if (!cfd->queued_for_flush() && cfd->imm()->IsFlushPending()) {
       cfd->Ref();
       cfd->set_queued_for_flush(true);
@@ -2539,6 +2572,11 @@ Status DBImpl::BackgroundFlush(bool* made_progress, JobContext* job_context,
 
     for (const auto& iter : flush_req) {
       ColumnFamilyData* cfd = iter.first;
+      if (immutable_db_options_.experimental_mempurge_threshold > 0.0) {
+        // If imm() contains silent memtables,
+        // requesting a flush will mark the imm_needed as true.
+        cfd->imm()->FlushRequested();
+      }
       if (cfd->IsDropped() || !cfd->imm()->IsFlushPending()) {
         // can't flush this CF, try next one
         column_families_not_to_flush.push_back(cfd);
@@ -2770,9 +2808,11 @@ void DBImpl::BackgroundCallCompaction(PrepickedCompaction* prepicked_compaction,
 
     if (prepicked_compaction != nullptr &&
         prepicked_compaction->task_token != nullptr) {
-      // Releasing task tokens affects the DB state, so must be done before we
-      // potentially signal the DB close process to proceed below.
-      prepicked_compaction->task_token->ReleaseOnce();
+      // Releasing task tokens affects (and asserts on) the DB state, so
+      // must be done before we potentially signal the DB close process to
+      // proceed below.
+      prepicked_compaction->task_token.reset();
+      ;
     }
 
     if (made_progress ||
@@ -3153,7 +3193,8 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
         &compaction_job_stats, thread_pri, io_tracer_,
         is_manual ? &manual_compaction_paused_ : nullptr,
         is_manual ? manual_compaction->canceled : nullptr, db_id_,
-        db_session_id_, c->column_family_data()->GetFullHistoryTsLow());
+        db_session_id_, c->column_family_data()->GetFullHistoryTsLow(),
+        &blob_callback_);
     compaction_job.Prepare();
 
     NotifyOnCompactionBegin(c->column_family_data(), c.get(), status,
@@ -3452,7 +3493,7 @@ void DBImpl::InstallSuperVersionAndScheduleWork(
   if (UNLIKELY(sv_context->new_superversion == nullptr)) {
     sv_context->NewSuperVersion();
   }
-  cfd->InstallSuperVersion(sv_context, &mutex_, mutable_cf_options);
+  cfd->InstallSuperVersion(sv_context, mutable_cf_options);
 
   // There may be a small data race here. The snapshot tricking bottommost
   // compaction may already be released here. But assuming there will always be

db/db_impl/db_impl_debug.cc

@@ -43,7 +43,7 @@ bool DBImpl::TEST_WALBufferIsEmpty(bool lock) {
   return res;
 }
 
-int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes(
+uint64_t DBImpl::TEST_MaxNextLevelOverlappingBytes(
     ColumnFamilyHandle* column_family) {
   ColumnFamilyData* cfd;
   if (column_family == nullptr) {
@@ -58,7 +58,8 @@ int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes(
 
 void DBImpl::TEST_GetFilesMetaData(
     ColumnFamilyHandle* column_family,
-    std::vector<std::vector<FileMetaData>>* metadata) {
+    std::vector<std::vector<FileMetaData>>* metadata,
+    std::vector<std::shared_ptr<BlobFileMetaData>>* blob_metadata) {
   auto cfh = static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
   auto cfd = cfh->cfd();
   InstrumentedMutexLock l(&mutex_);
@@ -72,6 +73,12 @@ void DBImpl::TEST_GetFilesMetaData(
       (*metadata)[level].push_back(*f);
     }
   }
+  if (blob_metadata != nullptr) {
+    blob_metadata->clear();
+    for (const auto& blob : cfd->current()->storage_info()->GetBlobFiles()) {
+      blob_metadata->push_back(blob.second);
+    }
+  }
 }
 
 uint64_t DBImpl::TEST_Current_Manifest_FileNo() {

db/db_impl/db_impl_files.cc

@@ -38,20 +38,26 @@ uint64_t DBImpl::MinObsoleteSstNumberToKeep() {
 }
 
 Status DBImpl::DisableFileDeletions() {
-  InstrumentedMutexLock l(&mutex_);
-  return DisableFileDeletionsWithLock();
+  Status s;
+  int my_disable_delete_obsolete_files;
+  {
+    InstrumentedMutexLock l(&mutex_);
+    s = DisableFileDeletionsWithLock();
+    my_disable_delete_obsolete_files = disable_delete_obsolete_files_;
+  }
+  if (my_disable_delete_obsolete_files == 1) {
+    ROCKS_LOG_INFO(immutable_db_options_.info_log, "File Deletions Disabled");
+  } else {
+    ROCKS_LOG_WARN(immutable_db_options_.info_log,
+                   "File Deletions Disabled, but already disabled. Counter: %d",
+                   my_disable_delete_obsolete_files);
+  }
+  return s;
 }
 
 Status DBImpl::DisableFileDeletionsWithLock() {
   mutex_.AssertHeld();
   ++disable_delete_obsolete_files_;
-  if (disable_delete_obsolete_files_ == 1) {
-    ROCKS_LOG_INFO(immutable_db_options_.info_log, "File Deletions Disabled");
-  } else {
-    ROCKS_LOG_WARN(immutable_db_options_.info_log,
-                   "File Deletions Disabled, but already disabled. Counter: %d",
-                   disable_delete_obsolete_files_);
-  }
   return Status::OK();
 }
 
@@ -215,7 +221,8 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,
     }
 
     // Add log files in wal_dir
-    if (immutable_db_options_.wal_dir != dbname_) {
+
+    if (!immutable_db_options_.IsWalDirSameAsDBPath(dbname_)) {
       std::vector<std::string> log_files;
       Status s = env_->GetChildren(immutable_db_options_.wal_dir, &log_files);
       s.PermitUncheckedError();  // TODO: What should we do on error?
@@ -395,10 +402,10 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
                                  blob_file.GetPath());
   }
 
+  auto wal_dir = immutable_db_options_.GetWalDir();
   for (auto file_num : state.log_delete_files) {
     if (file_num > 0) {
-      candidate_files.emplace_back(LogFileName(file_num),
-                                   immutable_db_options_.wal_dir);
+      candidate_files.emplace_back(LogFileName(file_num), wal_dir);
     }
   }
   for (const auto& filename : state.manifest_delete_files) {
@@ -517,12 +524,6 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
         break;
       case kOptionsFile:
         keep = (number >= optsfile_num2);
-        TEST_SYNC_POINT_CALLBACK(
-            "DBImpl::PurgeObsoleteFiles:CheckOptionsFiles:1",
-            reinterpret_cast<void*>(&number));
-        TEST_SYNC_POINT_CALLBACK(
-            "DBImpl::PurgeObsoleteFiles:CheckOptionsFiles:2",
-            reinterpret_cast<void*>(&keep));
         break;
       case kCurrentFile:
       case kDBLockFile:
@@ -547,8 +548,7 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
       fname = BlobFileName(candidate_file.file_path, number);
       dir_to_sync = candidate_file.file_path;
     } else {
-      dir_to_sync =
-          (type == kWalFile) ? immutable_db_options_.wal_dir : dbname_;
+      dir_to_sync = (type == kWalFile) ? wal_dir : dbname_;
       fname = dir_to_sync +
               ((!dir_to_sync.empty() && dir_to_sync.back() == '/') ||
                        (!to_delete.empty() && to_delete.front() == '/')

db/db_impl/db_impl_open.cc

@@ -111,16 +111,28 @@ DBOptions SanitizeOptions(const std::string& dbname, const DBOptions& src,
     result.recycle_log_file_num = 0;
   }
 
-  if (result.wal_dir.empty()) {
+  if (result.db_paths.size() == 0) {
+    result.db_paths.emplace_back(dbname, std::numeric_limits<uint64_t>::max());
+  } else if (result.wal_dir.empty()) {
     // Use dbname as default
     result.wal_dir = dbname;
   }
-  if (result.wal_dir.back() == '/') {
-    result.wal_dir = result.wal_dir.substr(0, result.wal_dir.size() - 1);
+  if (!result.wal_dir.empty()) {
+    // If there is a wal_dir already set, check to see if the wal_dir is the
+    // same as the dbname AND the same as the db_path[0] (which must exist from
+    // a few lines ago). If the wal_dir matches both of these values, then clear
+    // the wal_dir value, which will make wal_dir == dbname.  Most likely this
+    // condition was the result of reading an old options file where we forced
+    // wal_dir to be set (to dbname).
+    auto npath = NormalizePath(dbname + "/");
+    if (npath == NormalizePath(result.wal_dir + "/") &&
+        npath == NormalizePath(result.db_paths[0].path + "/")) {
+      result.wal_dir.clear();
+    }
   }
 
-  if (result.db_paths.size() == 0) {
-    result.db_paths.emplace_back(dbname, std::numeric_limits<uint64_t>::max());
+  if (!result.wal_dir.empty() && result.wal_dir.back() == '/') {
+    result.wal_dir = result.wal_dir.substr(0, result.wal_dir.size() - 1);
   }
 
   if (result.use_direct_reads && result.compaction_readahead_size == 0) {
@@ -141,7 +153,7 @@ DBOptions SanitizeOptions(const std::string& dbname, const DBOptions& src,
 
 #ifndef ROCKSDB_LITE
   ImmutableDBOptions immutable_db_options(result);
-  if (!IsWalDirSameAsDBPath(&immutable_db_options)) {
+  if (!immutable_db_options.IsWalDirSameAsDBPath()) {
     // Either the WAL dir and db_paths[0]/db_name are not the same, or we
     // cannot tell for sure. In either case, assume they're different and
     // explicitly cleanup the trash log files (bypass DeleteScheduler)
@@ -149,13 +161,14 @@ DBOptions SanitizeOptions(const std::string& dbname, const DBOptions& src,
     // DeleteScheduler::CleanupDirectory on the same dir later, it will be
     // safe
     std::vector<std::string> filenames;
-    Status s = result.env->GetChildren(result.wal_dir, &filenames);
+    auto wal_dir = immutable_db_options.GetWalDir();
+    Status s = result.env->GetChildren(wal_dir, &filenames);
     s.PermitUncheckedError();  //**TODO: What to do on error?
     for (std::string& filename : filenames) {
       if (filename.find(".log.trash", filename.length() -
                                           std::string(".log.trash").length()) !=
           std::string::npos) {
-        std::string trash_file = result.wal_dir + "/" + filename;
+        std::string trash_file = wal_dir + "/" + filename;
         result.env->DeleteFile(trash_file).PermitUncheckedError();
       }
     }
@@ -300,7 +313,8 @@ Status DBImpl::NewDB(std::vector<std::string>* new_filenames) {
     std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
         std::move(file), manifest, file_options, immutable_db_options_.clock,
         io_tracer_, nullptr /* stats */, immutable_db_options_.listeners,
-        nullptr, tmp_set.Contains(FileType::kDescriptorFile)));
+        nullptr, tmp_set.Contains(FileType::kDescriptorFile),
+        tmp_set.Contains(FileType::kDescriptorFile)));
     log::Writer log(std::move(file_writer), 0, false);
     std::string record;
     new_db.EncodeTo(&record);
@@ -542,12 +556,12 @@ Status DBImpl::Recover(
     // Note that prev_log_number() is no longer used, but we pay
     // attention to it in case we are recovering a database
     // produced by an older version of rocksdb.
+    auto wal_dir = immutable_db_options_.GetWalDir();
     if (!immutable_db_options_.best_efforts_recovery) {
-      s = env_->GetChildren(immutable_db_options_.wal_dir, &files_in_wal_dir);
+      s = env_->GetChildren(wal_dir, &files_in_wal_dir);
     }
     if (s.IsNotFound()) {
-      return Status::InvalidArgument("wal_dir not found",
-                                     immutable_db_options_.wal_dir);
+      return Status::InvalidArgument("wal_dir not found", wal_dir);
     } else if (!s.ok()) {
       return s;
     }
@@ -563,8 +577,7 @@ Status DBImpl::Recover(
               "existing log file: ",
               file);
         } else {
-          wal_files[number] =
-              LogFileName(immutable_db_options_.wal_dir, number);
+          wal_files[number] = LogFileName(wal_dir, number);
         }
       }
     }
@@ -644,7 +657,7 @@ Status DBImpl::Recover(
     if (s.ok()) {
       const std::string normalized_dbname = NormalizePath(dbname_);
       const std::string normalized_wal_dir =
-          NormalizePath(immutable_db_options_.wal_dir);
+          NormalizePath(immutable_db_options_.GetWalDir());
       if (immutable_db_options_.best_efforts_recovery) {
         filenames = std::move(files_in_dbname);
       } else if (normalized_dbname == normalized_wal_dir) {
@@ -849,7 +862,8 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& wal_numbers,
     // update the file number allocation counter in VersionSet.
     versions_->MarkFileNumberUsed(wal_number);
     // Open the log file
-    std::string fname = LogFileName(immutable_db_options_.wal_dir, wal_number);
+    std::string fname =
+        LogFileName(immutable_db_options_.GetWalDir(), wal_number);
 
     ROCKS_LOG_INFO(immutable_db_options_.info_log,
                    "Recovering log #%" PRIu64 " mode %d", wal_number,
@@ -1274,7 +1288,8 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& wal_numbers,
 Status DBImpl::GetLogSizeAndMaybeTruncate(uint64_t wal_number, bool truncate,
                                           LogFileNumberSize* log_ptr) {
   LogFileNumberSize log(wal_number);
-  std::string fname = LogFileName(immutable_db_options_.wal_dir, wal_number);
+  std::string fname =
+      LogFileName(immutable_db_options_.GetWalDir(), wal_number);
   Status s;
   // This gets the appear size of the wals, not including preallocated space.
   s = env_->GetFileSize(fname, &log.size);
@@ -1419,8 +1434,10 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
                       meta.fd.GetFileSize(), s.ToString().c_str());
       mutex_.Lock();
 
-      io_s.PermitUncheckedError();  // TODO(AR) is this correct, or should we
-                                    // return io_s if not ok()?
+      // TODO(AR) is this ok?
+      if (!io_s.ok() && s.ok()) {
+        s = io_s;
+      }
     }
   }
   ReleaseFileNumberFromPendingOutputs(pending_outputs_inserted_elem);
@@ -1514,15 +1531,14 @@ IOStatus DBImpl::CreateWAL(uint64_t log_file_num, uint64_t recycle_log_number,
       BuildDBOptions(immutable_db_options_, mutable_db_options_);
   FileOptions opt_file_options =
       fs_->OptimizeForLogWrite(file_options_, db_options);
-  std::string log_fname =
-      LogFileName(immutable_db_options_.wal_dir, log_file_num);
+  std::string wal_dir = immutable_db_options_.GetWalDir();
+  std::string log_fname = LogFileName(wal_dir, log_file_num);
 
   if (recycle_log_number) {
     ROCKS_LOG_INFO(immutable_db_options_.info_log,
                    "reusing log %" PRIu64 " from recycle list\n",
                    recycle_log_number);
-    std::string old_log_fname =
-        LogFileName(immutable_db_options_.wal_dir, recycle_log_number);
+    std::string old_log_fname = LogFileName(wal_dir, recycle_log_number);
     TEST_SYNC_POINT("DBImpl::CreateWAL:BeforeReuseWritableFile1");
     TEST_SYNC_POINT("DBImpl::CreateWAL:BeforeReuseWritableFile2");
     io_s = fs_->ReuseWritableFile(log_fname, old_log_fname, opt_file_options,
@@ -1540,7 +1556,8 @@ IOStatus DBImpl::CreateWAL(uint64_t log_file_num, uint64_t recycle_log_number,
     std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
         std::move(lfile), log_fname, opt_file_options,
         immutable_db_options_.clock, io_tracer_, nullptr /* stats */, listeners,
-        nullptr, tmp_set.Contains(FileType::kWalFile)));
+        nullptr, tmp_set.Contains(FileType::kWalFile),
+        tmp_set.Contains(FileType::kWalFile)));
     *new_log = new log::Writer(std::move(file_writer), log_file_num,
                                immutable_db_options_.recycle_log_file_num > 0,
                                immutable_db_options_.manual_wal_flush);
@@ -1572,7 +1589,7 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
   }
 
   DBImpl* impl = new DBImpl(db_options, dbname, seq_per_batch, batch_per_txn);
-  s = impl->env_->CreateDirIfMissing(impl->immutable_db_options_.wal_dir);
+  s = impl->env_->CreateDirIfMissing(impl->immutable_db_options_.GetWalDir());
   if (s.ok()) {
     std::vector<std::string> paths;
     for (auto& db_path : impl->immutable_db_options_.db_paths) {
@@ -1604,7 +1621,7 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
     return s;
   }
 
-  impl->wal_in_db_path_ = IsWalDirSameAsDBPath(&impl->immutable_db_options_);
+  impl->wal_in_db_path_ = impl->immutable_db_options_.IsWalDirSameAsDBPath();
 
   impl->mutex_.Lock();
   // Handles create_if_missing, error_if_exists
@@ -1671,6 +1688,7 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
 
       impl->DeleteObsoleteFiles();
       s = impl->directories_.GetDbDir()->Fsync(IOOptions(), nullptr);
+      TEST_SYNC_POINT("DBImpl::Open:AfterDeleteFilesAndSyncDir");
     }
     if (s.ok()) {
       // In WritePrepared there could be gap in sequence numbers. This breaks

db/db_impl/db_impl_secondary.cc

@@ -41,6 +41,9 @@ Status DBImplSecondary::Recover(
           ->Recover(column_families, &manifest_reader_, &manifest_reporter_,
                     &manifest_reader_status_);
   if (!s.ok()) {
+    if (manifest_reader_status_) {
+      manifest_reader_status_->PermitUncheckedError();
+    }
     return s;
   }
   if (immutable_db_options_.paranoid_checks && s.ok()) {
@@ -97,10 +100,10 @@ Status DBImplSecondary::FindNewLogNumbers(std::vector<uint64_t>* logs) {
   assert(logs != nullptr);
   std::vector<std::string> filenames;
   Status s;
-  s = env_->GetChildren(immutable_db_options_.wal_dir, &filenames);
+  s = env_->GetChildren(immutable_db_options_.GetWalDir(), &filenames);
   if (s.IsNotFound()) {
     return Status::InvalidArgument("Failed to open wal_dir",
-                                   immutable_db_options_.wal_dir);
+                                   immutable_db_options_.GetWalDir());
   } else if (!s.ok()) {
     return s;
   }
@@ -140,7 +143,8 @@ Status DBImplSecondary::MaybeInitLogReader(
     // initialize log reader from log_number
     // TODO: min_log_number_to_keep_2pc check needed?
     // Open the log file
-    std::string fname = LogFileName(immutable_db_options_.wal_dir, log_number);
+    std::string fname =
+        LogFileName(immutable_db_options_.GetWalDir(), log_number);
     ROCKS_LOG_INFO(immutable_db_options_.info_log,
                    "Recovering log #%" PRIu64 " mode %d", log_number,
                    static_cast<int>(immutable_db_options_.wal_recovery_mode));
@@ -627,7 +631,7 @@ Status DB::OpenAsSecondary(
       &impl->write_controller_, impl->io_tracer_));
   impl->column_family_memtables_.reset(
       new ColumnFamilyMemTablesImpl(impl->versions_->GetColumnFamilySet()));
-  impl->wal_in_db_path_ = IsWalDirSameAsDBPath(&impl->immutable_db_options_);
+  impl->wal_in_db_path_ = impl->immutable_db_options_.IsWalDirSameAsDBPath();
 
   impl->mutex_.Lock();
   s = impl->Recover(column_families, true, false, false);
@@ -767,6 +771,7 @@ Status DB::OpenAndCompact(
   compaction_input.db_options.env = override_options.env;
   compaction_input.db_options.file_checksum_gen_factory =
       override_options.file_checksum_gen_factory;
+  compaction_input.db_options.statistics = override_options.statistics;
   compaction_input.column_family.options.comparator =
       override_options.comparator;
   compaction_input.column_family.options.merge_operator =

db/db_impl/db_impl_secondary.h

@@ -278,6 +278,16 @@ class DBImplSecondary : public DBImpl {
       return AddColumnFamilyId(column_family_id);
     }
 
+    Status MarkBeginPrepare(bool) override { return Status::OK(); }
+
+    Status MarkEndPrepare(const Slice&) override { return Status::OK(); }
+
+    Status MarkRollback(const Slice&) override { return Status::OK(); }
+
+    Status MarkCommit(const Slice&) override { return Status::OK(); }
+
+    Status MarkNoop(bool) override { return Status::OK(); }
+
     const std::unordered_set<uint32_t>& column_families() const {
       return column_family_ids_;
     }

db/db_impl/db_impl_write.cc

@@ -220,7 +220,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
 
   bool need_log_sync = write_options.sync;
   bool need_log_dir_sync = need_log_sync && !log_dir_synced_;
-  if (!two_write_queues_ || !disable_memtable) {
+  assert(!two_write_queues_ || !disable_memtable);
+  {
     // With concurrent writes we do preprocess only in the write thread that
     // also does write to memtable to avoid sync issue on shared data structure
     // with the other thread
@@ -251,6 +252,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
       write_thread_.EnterAsBatchGroupLeader(&w, &write_group);
 
   IOStatus io_s;
+  Status pre_release_cb_status;
   if (status.ok()) {
     // Rules for when we can update the memtable concurrently
     // 1. supported by memtable
@@ -361,7 +363,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
               writer->sequence, disable_memtable, writer->log_used, index++,
               pre_release_callback_cnt);
           if (!ws.ok()) {
-            status = ws;
+            status = pre_release_cb_status = ws;
             break;
           }
         }
@@ -414,10 +416,13 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
 
   if (!w.CallbackFailed()) {
     if (!io_s.ok()) {
+      assert(pre_release_cb_status.ok());
       IOStatusCheck(io_s);
     } else {
-      WriteStatusCheck(status);
+      WriteStatusCheck(pre_release_cb_status);
     }
+  } else {
+    assert(io_s.ok() && pre_release_cb_status.ok());
   }
 
   if (need_log_sync) {
@@ -1647,12 +1652,7 @@ Status DBImpl::TrimMemtableHistory(WriteContext* context) {
   for (auto& cfd : cfds) {
     autovector<MemTable*> to_delete;
     bool trimmed = cfd->imm()->TrimHistory(
-        &to_delete, cfd->mem()->ApproximateMemoryUsage());
-    if (!to_delete.empty()) {
-      for (auto m : to_delete) {
-        delete m;
-      }
-    }
+        &context->memtables_to_free_, cfd->mem()->ApproximateMemoryUsage());
     if (trimmed) {
       context->superversion_context.NewSuperVersion();
       assert(context->superversion_context.new_superversion.get() != nullptr);
@@ -1911,6 +1911,8 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
       for (auto cf : empty_cfs) {
         if (cf->IsEmpty()) {
           cf->SetLogNumber(logfile_number_);
+          // MEMPURGE: No need to change this, because new adds
+          // should still receive new sequence numbers.
           cf->mem()->SetCreationSeq(versions_->LastSequence());
         }  // cf may become non-empty.
       }
@@ -1933,11 +1935,13 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   }
 
   cfd->mem()->SetNextLogNumber(logfile_number_);
+  assert(new_mem != nullptr);
   cfd->imm()->Add(cfd->mem(), &context->memtables_to_free_);
   new_mem->Ref();
   cfd->SetMemtable(new_mem);
   InstallSuperVersionAndScheduleWork(cfd, &context->superversion_context,
                                      mutable_cf_options);
+
 #ifndef ROCKSDB_LITE
   mutex_.Unlock();
   // Notify client that memtable is sealed, now that we have successfully

db/db_info_dumper.cc

@@ -109,31 +109,30 @@ void DumpDBFileSummary(const ImmutableDBOptions& options,
   }
 
   // Get wal file in wal_dir
-  if (dbname.compare(options.wal_dir) != 0) {
-    if (!env->GetChildren(options.wal_dir, &files).ok()) {
-      Error(options.info_log,
-          "Error when reading %s dir\n",
-          options.wal_dir.c_str());
+  const auto& wal_dir = options.GetWalDir(dbname);
+  if (!options.IsWalDirSameAsDBPath(dbname)) {
+    if (!env->GetChildren(wal_dir, &files).ok()) {
+      Error(options.info_log, "Error when reading %s dir\n", wal_dir.c_str());
       return;
     }
     wal_info.clear();
     for (const std::string& file : files) {
       if (ParseFileName(file, &number, &type)) {
         if (type == kWalFile) {
-          if (env->GetFileSize(options.wal_dir + "/" + file, &file_size).ok()) {
+          if (env->GetFileSize(wal_dir + "/" + file, &file_size).ok()) {
             wal_info.append(file)
                 .append(" size: ")
                 .append(std::to_string(file_size))
                 .append(" ; ");
           } else {
             Error(options.info_log, "Error when reading LOG file %s/%s\n",
-                  options.wal_dir.c_str(), file.c_str());
+                  wal_dir.c_str(), file.c_str());
           }
         }
       }
     }
   }
-  Header(options.info_log, "Write Ahead Log file in %s: %s\n",
-         options.wal_dir.c_str(), wal_info.c_str());
+  Header(options.info_log, "Write Ahead Log file in %s: %s\n", wal_dir.c_str(),
+         wal_info.c_str());
 }
 }  // namespace ROCKSDB_NAMESPACE

db/db_inplace_update_test.cc

@@ -14,7 +14,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBTestInPlaceUpdate : public DBTestBase {
  public:
   DBTestInPlaceUpdate()
-      : DBTestBase("/db_inplace_update_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_inplace_update_test", /*env_do_fsync=*/true) {}
 };
 
 TEST_F(DBTestInPlaceUpdate, InPlaceUpdate) {

db/db_io_failure_test.cc

@@ -15,8 +15,7 @@ namespace ROCKSDB_NAMESPACE {
 
 class DBIOFailureTest : public DBTestBase {
  public:
-  DBIOFailureTest()
-      : DBTestBase("/db_io_failure_test", /*env_do_fsync=*/true) {}
+  DBIOFailureTest() : DBTestBase("db_io_failure_test", /*env_do_fsync=*/true) {}
 };
 
 #ifndef ROCKSDB_LITE

db/db_iterator_test.cc

@@ -35,7 +35,7 @@ class DummyReadCallback : public ReadCallback {
 class DBIteratorTest : public DBTestBase,
                        public testing::WithParamInterface<bool> {
  public:
-  DBIteratorTest() : DBTestBase("/db_iterator_test", /*env_do_fsync=*/true) {}
+  DBIteratorTest() : DBTestBase("db_iterator_test", /*env_do_fsync=*/true) {}
 
   Iterator* NewIterator(const ReadOptions& read_options,
                         ColumnFamilyHandle* column_family = nullptr) {
@@ -1536,9 +1536,11 @@ class DBIteratorTestForPinnedData : public DBIteratorTest {
 }
 };
 
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 TEST_P(DBIteratorTestForPinnedData, PinnedDataIteratorRandomizedNormal) {
   PinnedDataIteratorRandomized(TestConfig::NORMAL);
 }
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 TEST_P(DBIteratorTestForPinnedData, PinnedDataIteratorRandomizedCLoseAndOpen) {
   PinnedDataIteratorRandomized(TestConfig::CLOSE_AND_OPEN);

db/db_kv_checksum_test.cc

@@ -29,7 +29,7 @@ class DbKvChecksumTest
       public ::testing::WithParamInterface<std::tuple<WriteBatchOpType, char>> {
  public:
   DbKvChecksumTest()
-      : DBTestBase("/db_kv_checksum_test", /*env_do_fsync=*/false) {
+      : DBTestBase("db_kv_checksum_test", /*env_do_fsync=*/false) {
     op_type_ = std::get<0>(GetParam());
     corrupt_byte_addend_ = std::get<1>(GetParam());
   }

db/db_log_iter_test.cc

@@ -21,7 +21,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBTestXactLogIterator : public DBTestBase {
  public:
   DBTestXactLogIterator()
-      : DBTestBase("/db_log_iter_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_log_iter_test", /*env_do_fsync=*/true) {}
 
   std::unique_ptr<TransactionLogIterator> OpenTransactionLogIter(
       const SequenceNumber seq) {
@@ -185,31 +185,38 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorCorruptedLog) {
   do {
     Options options = OptionsForLogIterTest();
     DestroyAndReopen(options);
+
     for (int i = 0; i < 1024; i++) {
       ASSERT_OK(Put("key" + ToString(i), DummyString(10)));
     }
-    ASSERT_OK(dbfull()->Flush(FlushOptions()));
-    ASSERT_OK(dbfull()->FlushWAL(false));
+
+    ASSERT_OK(Flush());
+    ASSERT_OK(db_->FlushWAL(false));
+
     // Corrupt this log to create a gap
-    ROCKSDB_NAMESPACE::VectorLogPtr wal_files;
-    ASSERT_OK(dbfull()->GetSortedWalFiles(wal_files));
+    ASSERT_OK(db_->DisableFileDeletions());
+
+    VectorLogPtr wal_files;
+    ASSERT_OK(db_->GetSortedWalFiles(wal_files));
+    ASSERT_FALSE(wal_files.empty());
+
     const auto logfile_path = dbname_ + "/" + wal_files.front()->PathName();
-    if (mem_env_) {
-      mem_env_->Truncate(logfile_path, wal_files.front()->SizeFileBytes() / 2);
-    } else {
-      ASSERT_EQ(0, truncate(logfile_path.c_str(),
-                   wal_files.front()->SizeFileBytes() / 2));
-    }
+    ASSERT_OK(test::TruncateFile(env_, logfile_path,
+                                 wal_files.front()->SizeFileBytes() / 2));
+
+    ASSERT_OK(db_->EnableFileDeletions());
 
     // Insert a new entry to a new log file
     ASSERT_OK(Put("key1025", DummyString(10)));
-    ASSERT_OK(dbfull()->FlushWAL(false));
+    ASSERT_OK(db_->FlushWAL(false));
+
     // Try to read from the beginning. Should stop before the gap and read less
     // than 1025 entries
     auto iter = OpenTransactionLogIter(0);
-    int count;
+    int count = 0;
     SequenceNumber last_sequence_read = ReadRecords(iter, count, false);
     ASSERT_LT(last_sequence_read, 1025U);
+
     // Try to read past the gap, should be able to seek to key1025
     auto iter2 = OpenTransactionLogIter(last_sequence_read + 1);
     ExpectRecords(1, iter2);

db/db_memtable_test.cc

@@ -17,7 +17,7 @@ namespace ROCKSDB_NAMESPACE {
 
 class DBMemTableTest : public DBTestBase {
  public:
-  DBMemTableTest() : DBTestBase("/db_memtable_test", /*env_do_fsync=*/true) {}
+  DBMemTableTest() : DBTestBase("db_memtable_test", /*env_do_fsync=*/true) {}
 };
 
 class MockMemTableRep : public MemTableRep {

db/db_merge_operand_test.cc

@@ -44,7 +44,7 @@ class LimitedStringAppendMergeOp : public StringAppendTESTOperator {
 class DBMergeOperandTest : public DBTestBase {
  public:
   DBMergeOperandTest()
-      : DBTestBase("/db_merge_operand_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_merge_operand_test", /*env_do_fsync=*/true) {}
 };
 
 TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {

db/db_merge_operator_test.cc

@@ -37,7 +37,7 @@ class TestReadCallback : public ReadCallback {
 class DBMergeOperatorTest : public DBTestBase {
  public:
   DBMergeOperatorTest()
-      : DBTestBase("/db_merge_operator_test", /*env_do_fsync=*/false) {}
+      : DBTestBase("db_merge_operator_test", /*env_do_fsync=*/false) {}
 
   std::string GetWithReadCallback(SnapshotChecker* snapshot_checker,
                                   const Slice& key,

db/db_options_test.cc

@@ -27,7 +27,7 @@ namespace ROCKSDB_NAMESPACE {
 
 class DBOptionsTest : public DBTestBase {
  public:
-  DBOptionsTest() : DBTestBase("/db_options_test", /*env_do_fsync=*/true) {}
+  DBOptionsTest() : DBTestBase("db_options_test", /*env_do_fsync=*/true) {}
 
 #ifndef ROCKSDB_LITE
   std::unordered_map<std::string, std::string> GetMutableDBOptionsMap(
@@ -98,6 +98,66 @@ TEST_F(DBOptionsTest, ImmutableTrackAndVerifyWalsInManifest) {
 // RocksDB lite don't support dynamic options.
 #ifndef ROCKSDB_LITE
 
+TEST_F(DBOptionsTest, AvoidUpdatingOptions) {
+  Options options;
+  options.env = env_;
+  options.max_background_jobs = 4;
+  options.delayed_write_rate = 1024;
+
+  Reopen(options);
+
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  bool is_changed_stats = false;
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::WriteOptionsFile:PersistOptions", [&](void* /*arg*/) {
+        ASSERT_FALSE(is_changed_stats);  // should only save options file once
+        is_changed_stats = true;
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  // helper function to check the status and reset after each check
+  auto is_changed = [&] {
+    bool ret = is_changed_stats;
+    is_changed_stats = false;
+    return ret;
+  };
+
+  // without changing the value, but it's sanitized to a different value
+  ASSERT_OK(dbfull()->SetDBOptions({{"bytes_per_sync", "0"}}));
+  ASSERT_TRUE(is_changed());
+
+  // without changing the value
+  ASSERT_OK(dbfull()->SetDBOptions({{"max_background_jobs", "4"}}));
+  ASSERT_FALSE(is_changed());
+
+  // changing the value
+  ASSERT_OK(dbfull()->SetDBOptions({{"bytes_per_sync", "123"}}));
+  ASSERT_TRUE(is_changed());
+
+  // update again
+  ASSERT_OK(dbfull()->SetDBOptions({{"bytes_per_sync", "123"}}));
+  ASSERT_FALSE(is_changed());
+
+  // without changing a default value
+  ASSERT_OK(dbfull()->SetDBOptions({{"strict_bytes_per_sync", "false"}}));
+  ASSERT_FALSE(is_changed());
+
+  // now change
+  ASSERT_OK(dbfull()->SetDBOptions({{"strict_bytes_per_sync", "true"}}));
+  ASSERT_TRUE(is_changed());
+
+  // multiple values without change
+  ASSERT_OK(dbfull()->SetDBOptions(
+      {{"max_total_wal_size", "0"}, {"stats_dump_period_sec", "600"}}));
+  ASSERT_FALSE(is_changed());
+
+  // multiple values with change
+  ASSERT_OK(dbfull()->SetDBOptions(
+      {{"max_open_files", "100"}, {"stats_dump_period_sec", "600"}}));
+  ASSERT_TRUE(is_changed());
+}
+
 TEST_F(DBOptionsTest, GetLatestDBOptions) {
   // GetOptions should be able to get latest option changed by SetOptions.
   Options options;

db/db_properties_test.cc

@@ -27,7 +27,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBPropertiesTest : public DBTestBase {
  public:
   DBPropertiesTest()
-      : DBTestBase("/db_properties_test", /*env_do_fsync=*/false) {}
+      : DBTestBase("db_properties_test", /*env_do_fsync=*/false) {}
 };
 
 #ifndef ROCKSDB_LITE

db/db_range_del_test.cc

@@ -14,7 +14,7 @@ namespace ROCKSDB_NAMESPACE {
 
 class DBRangeDelTest : public DBTestBase {
  public:
-  DBRangeDelTest() : DBTestBase("/db_range_del_test", /*env_do_fsync=*/false) {}
+  DBRangeDelTest() : DBTestBase("db_range_del_test", /*env_do_fsync=*/false) {}
 
   std::string GetNumericStr(int key) {
     uint64_t uint64_key = static_cast<uint64_t>(key);
@@ -660,6 +660,16 @@ TEST_F(DBRangeDelTest, TableEvictedDuringScan) {
   ASSERT_EQ(kNum, expected);
   delete iter;
   db_->ReleaseSnapshot(snapshot);
+
+  // Also test proper cache handling in GetRangeTombstoneIterator,
+  // via TablesRangeTombstoneSummary. (This once triggered memory leak
+  // report with ASAN.)
+  opts.max_open_files = 1;
+  Reopen(opts);
+
+  std::string str;
+  ASSERT_OK(dbfull()->TablesRangeTombstoneSummary(db_->DefaultColumnFamily(),
+                                                  100, &str));
 }
 
 TEST_F(DBRangeDelTest, GetCoveredKeyFromMutableMemtable) {

db/db_secondary_test.cc

@@ -10,6 +10,7 @@
 #include "db/db_impl/db_impl_secondary.h"
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
+#include "rocksdb/utilities/transaction_db.h"
 #include "test_util/sync_point.h"
 #include "utilities/fault_injection_env.h"
 
@@ -19,7 +20,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBSecondaryTest : public DBTestBase {
  public:
   DBSecondaryTest()
-      : DBTestBase("/db_secondary_test", /*env_do_fsync=*/true),
+      : DBTestBase("db_secondary_test", /*env_do_fsync=*/true),
         secondary_path_(),
         handles_secondary_(),
         db_secondary_(nullptr) {
@@ -114,6 +115,18 @@ void DBSecondaryTest::CheckFileTypeCounts(const std::string& dir,
   ASSERT_EQ(expected_manifest, manifest_cnt);
 }
 
+TEST_F(DBSecondaryTest, NonExistingDb) {
+  Destroy(last_options_);
+
+  Options options = GetDefaultOptions();
+  options.env = env_;
+  options.max_open_files = -1;
+  const std::string dbname = "/doesnt/exist";
+  Status s =
+      DB::OpenAsSecondary(options, dbname, secondary_path_, &db_secondary_);
+  ASSERT_TRUE(s.IsIOError());
+}
+
 TEST_F(DBSecondaryTest, ReopenAsSecondary) {
   Options options;
   options.env = env_;
@@ -598,17 +611,19 @@ TEST_F(DBSecondaryTest, SwitchToNewManifestDuringOpen) {
   SyncPoint::GetInstance()->LoadDependency(
       {{"ReactiveVersionSet::MaybeSwitchManifest:AfterGetCurrentManifestPath:0",
         "VersionSet::ProcessManifestWrites:BeforeNewManifest"},
-       {"VersionSet::ProcessManifestWrites:AfterNewManifest",
+       {"DBImpl::Open:AfterDeleteFilesAndSyncDir",
         "ReactiveVersionSet::MaybeSwitchManifest:AfterGetCurrentManifestPath:"
         "1"}});
   SyncPoint::GetInstance()->EnableProcessing();
 
-  // Make sure db calls RecoverLogFiles so as to trigger a manifest write,
-  // which causes the db to switch to a new MANIFEST upon start.
   port::Thread ro_db_thread([&]() {
     Options options1;
     options1.env = env_;
     options1.max_open_files = -1;
+    Status s = TryOpenSecondary(options1);
+    ASSERT_TRUE(s.IsTryAgain());
+
+    // Try again
     OpenSecondary(options1);
     CloseSecondary();
   });
@@ -1115,6 +1130,39 @@ TEST_F(DBSecondaryTest, InconsistencyDuringCatchUp) {
   Status s = db_secondary_->TryCatchUpWithPrimary();
   ASSERT_TRUE(s.IsCorruption());
 }
+
+TEST_F(DBSecondaryTest, OpenWithTransactionDB) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+
+  // Destroy the DB to recreate as a TransactionDB.
+  Close();
+  Destroy(options, true);
+
+  // Create a TransactionDB.
+  TransactionDB* txn_db = nullptr;
+  TransactionDBOptions txn_db_opts;
+  ASSERT_OK(TransactionDB::Open(options, txn_db_opts, dbname_, &txn_db));
+  ASSERT_NE(txn_db, nullptr);
+  db_ = txn_db;
+
+  std::vector<std::string> cfs = {"new_CF"};
+  CreateColumnFamilies(cfs, options);
+  ASSERT_EQ(handles_.size(), 1);
+
+  WriteOptions wopts;
+  TransactionOptions txn_opts;
+  Transaction* txn1 = txn_db->BeginTransaction(wopts, txn_opts, nullptr);
+  ASSERT_NE(txn1, nullptr);
+  ASSERT_OK(txn1->Put(handles_[0], "k1", "v1"));
+  ASSERT_OK(txn1->Commit());
+  delete txn1;
+
+  options = CurrentOptions();
+  options.max_open_files = -1;
+  ASSERT_OK(TryOpenSecondary(options));
+}
+
 #endif  //! ROCKSDB_LITE
 
 }  // namespace ROCKSDB_NAMESPACE

db/db_sst_test.cc

@@ -18,7 +18,7 @@ namespace ROCKSDB_NAMESPACE {
 
 class DBSSTTest : public DBTestBase {
  public:
-  DBSSTTest() : DBTestBase("/db_sst_test", /*env_do_fsync=*/true) {}
+  DBSSTTest() : DBTestBase("db_sst_test", /*env_do_fsync=*/true) {}
 };
 
 #ifndef ROCKSDB_LITE
@@ -757,7 +757,7 @@ class DBWALTestWithParam
       public testing::WithParamInterface<std::tuple<std::string, bool>> {
  public:
   explicit DBWALTestWithParam()
-      : DBTestBase("/db_wal_test_with_params", /*env_do_fsync=*/true) {
+      : DBTestBase("db_wal_test_with_params", /*env_do_fsync=*/true) {
     wal_dir_ = std::get<0>(GetParam());
     wal_dir_same_as_dbname_ = std::get<1>(GetParam());
   }
@@ -1569,6 +1569,96 @@ TEST_F(DBSSTTest, GetTotalSstFilesSizeVersionsFilesShared) {
   ASSERT_EQ(total_sst_files_size, 0);
 }
 
+// This test if blob files are recorded by SST File Manager when Compaction job
+// creates/delete them and in case of AtomicFlush.
+TEST_F(DBSSTTest, DBWithSFMForBlobFilesAtomicFlush) {
+  std::shared_ptr<SstFileManager> sst_file_manager(NewSstFileManager(env_));
+  auto sfm = static_cast<SstFileManagerImpl*>(sst_file_manager.get());
+  Options options = CurrentOptions();
+  options.sst_file_manager = sst_file_manager;
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+  options.disable_auto_compactions = true;
+  options.enable_blob_garbage_collection = true;
+  options.blob_garbage_collection_age_cutoff = 0.5;
+  options.atomic_flush = true;
+
+  int files_added = 0;
+  int files_deleted = 0;
+  int files_scheduled_to_delete = 0;
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "SstFileManagerImpl::OnAddFile", [&](void* arg) {
+        const std::string* const file_path =
+            static_cast<const std::string*>(arg);
+        if (EndsWith(*file_path, ".blob")) {
+          files_added++;
+        }
+      });
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "SstFileManagerImpl::OnDeleteFile", [&](void* arg) {
+        const std::string* const file_path =
+            static_cast<const std::string*>(arg);
+        if (EndsWith(*file_path, ".blob")) {
+          files_deleted++;
+        }
+      });
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "SstFileManagerImpl::ScheduleFileDeletion", [&](void* arg) {
+        assert(arg);
+        const std::string* const file_path =
+            static_cast<const std::string*>(arg);
+        if (EndsWith(*file_path, ".blob")) {
+          ++files_scheduled_to_delete;
+        }
+      });
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  DestroyAndReopen(options);
+  Random rnd(301);
+
+  ASSERT_OK(Put("key_1", "value_1"));
+  ASSERT_OK(Put("key_2", "value_2"));
+  ASSERT_OK(Put("key_3", "value_3"));
+  ASSERT_OK(Put("key_4", "value_4"));
+  ASSERT_OK(Flush());
+
+  // Overwrite will create the garbage data.
+  ASSERT_OK(Put("key_3", "new_value_3"));
+  ASSERT_OK(Put("key_4", "new_value_4"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(Put("Key5", "blob_value5"));
+  ASSERT_OK(Put("Key6", "blob_value6"));
+  ASSERT_OK(Flush());
+
+  ASSERT_EQ(files_added, 3);
+  ASSERT_EQ(files_deleted, 0);
+  ASSERT_EQ(files_scheduled_to_delete, 0);
+  files_added = 0;
+
+  constexpr Slice* begin = nullptr;
+  constexpr Slice* end = nullptr;
+  // Compaction job will create a new file and delete the older files.
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), begin, end));
+
+  ASSERT_EQ(files_added, 1);
+  ASSERT_EQ(files_deleted, 1);
+  ASSERT_EQ(files_scheduled_to_delete, 1);
+
+  Close();
+  ASSERT_OK(DestroyDB(dbname_, options));
+  sfm->WaitForEmptyTrash();
+  ASSERT_EQ(files_deleted, 4);
+  ASSERT_EQ(files_scheduled_to_delete, 4);
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+}
+
 #endif  // ROCKSDB_LITE
 
 }  // namespace ROCKSDB_NAMESPACE

db/db_statistics_test.cc

@@ -16,7 +16,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBStatisticsTest : public DBTestBase {
  public:
   DBStatisticsTest()
-      : DBTestBase("/db_statistics_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_statistics_test", /*env_do_fsync=*/true) {}
 };
 
 TEST_F(DBStatisticsTest, CompressionStatsTest) {

db/db_table_properties_test.cc

@@ -52,7 +52,7 @@ class DBTablePropertiesTest : public DBTestBase,
                               public testing::WithParamInterface<std::string> {
  public:
   DBTablePropertiesTest()
-      : DBTestBase("/db_table_properties_test", /*env_do_fsync=*/false) {}
+      : DBTestBase("db_table_properties_test", /*env_do_fsync=*/false) {}
   TablePropertiesCollection TestGetPropertiesOfTablesInRange(
       std::vector<Range> ranges, std::size_t* num_properties = nullptr,
       std::size_t* num_files = nullptr);
@@ -281,7 +281,7 @@ class DBTableHostnamePropertyTest
       public ::testing::WithParamInterface<std::tuple<int, std::string>> {
  public:
   DBTableHostnamePropertyTest()
-      : DBTestBase("/db_table_hostname_property_test",
+      : DBTestBase("db_table_hostname_property_test",
                    /*env_do_fsync=*/false) {}
 };
 

db/db_tailing_iter_test.cc

@@ -21,7 +21,7 @@ namespace ROCKSDB_NAMESPACE {
 class DBTestTailingIterator : public DBTestBase {
  public:
   DBTestTailingIterator()
-      : DBTestBase("/db_tailing_iterator_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("db_tailing_iterator_test", /*env_do_fsync=*/true) {}
 };
 
 TEST_F(DBTestTailingIterator, TailingIteratorSingle) {

db/db_test.cc

@@ -25,6 +25,7 @@
 
 #include "cache/lru_cache.h"
 #include "db/blob/blob_index.h"
+#include "db/blob/blob_log_format.h"
 #include "db/db_impl/db_impl.h"
 #include "db/db_test_util.h"
 #include "db/dbformat.h"
@@ -74,7 +75,7 @@ namespace ROCKSDB_NAMESPACE {
 // If fsync needs to be covered in a test, put it in other places.
 class DBTest : public DBTestBase {
  public:
-  DBTest() : DBTestBase("/db_test", /*env_do_fsync=*/false) {}
+  DBTest() : DBTestBase("db_test", /*env_do_fsync=*/false) {}
 };
 
 class DBTestWithParam
@@ -1027,10 +1028,10 @@ TEST_F(DBTest, FailMoreDbPaths) {
 }
 
 void CheckColumnFamilyMeta(
-    const ColumnFamilyMetaData& cf_meta,
+    const ColumnFamilyMetaData& cf_meta, const std::string& cf_name,
     const std::vector<std::vector<FileMetaData>>& files_by_level,
     uint64_t start_time, uint64_t end_time) {
-  ASSERT_EQ(cf_meta.name, kDefaultColumnFamilyName);
+  ASSERT_EQ(cf_meta.name, cf_name);
   ASSERT_EQ(cf_meta.levels.size(), files_by_level.size());
 
   uint64_t cf_size = 0;
@@ -1124,6 +1125,53 @@ void CheckLiveFilesMeta(
 }
 
 #ifndef ROCKSDB_LITE
+void AddBlobFile(const ColumnFamilyHandle* cfh, uint64_t blob_file_number,
+                 uint64_t total_blob_count, uint64_t total_blob_bytes,
+                 const std::string& checksum_method,
+                 const std::string& checksum_value,
+                 uint64_t garbage_blob_count = 0,
+                 uint64_t garbage_blob_bytes = 0) {
+  ColumnFamilyData* cfd =
+      (static_cast<const ColumnFamilyHandleImpl*>(cfh))->cfd();
+  assert(cfd);
+
+  Version* const version = cfd->current();
+  assert(version);
+
+  VersionStorageInfo* const storage_info = version->storage_info();
+  assert(storage_info);
+
+  // Add a live blob file.
+
+  auto shared_meta = SharedBlobFileMetaData::Create(
+      blob_file_number, total_blob_count, total_blob_bytes, checksum_method,
+      checksum_value);
+
+  auto meta = BlobFileMetaData::Create(std::move(shared_meta),
+                                       BlobFileMetaData::LinkedSsts(),
+                                       garbage_blob_count, garbage_blob_bytes);
+
+  storage_info->AddBlobFile(std::move(meta));
+}
+
+static void CheckBlobMetaData(
+    const BlobMetaData& bmd, uint64_t blob_file_number,
+    uint64_t total_blob_count, uint64_t total_blob_bytes,
+    const std::string& checksum_method, const std::string& checksum_value,
+    uint64_t garbage_blob_count = 0, uint64_t garbage_blob_bytes = 0) {
+  ASSERT_EQ(bmd.blob_file_number, blob_file_number);
+  ASSERT_EQ(bmd.blob_file_name, BlobFileName("", blob_file_number));
+  ASSERT_EQ(bmd.blob_file_size,
+            total_blob_bytes + BlobLogHeader::kSize + BlobLogFooter::kSize);
+
+  ASSERT_EQ(bmd.total_blob_count, total_blob_count);
+  ASSERT_EQ(bmd.total_blob_bytes, total_blob_bytes);
+  ASSERT_EQ(bmd.garbage_blob_count, garbage_blob_count);
+  ASSERT_EQ(bmd.garbage_blob_bytes, garbage_blob_bytes);
+  ASSERT_EQ(bmd.checksum_method, checksum_method);
+  ASSERT_EQ(bmd.checksum_value, checksum_value);
+}
+
 TEST_F(DBTest, MetaDataTest) {
   Options options = CurrentOptions();
   options.create_if_missing = true;
@@ -1164,13 +1212,71 @@ TEST_F(DBTest, MetaDataTest) {
 
   ColumnFamilyMetaData cf_meta;
   db_->GetColumnFamilyMetaData(&cf_meta);
-  CheckColumnFamilyMeta(cf_meta, files_by_level, start_time, end_time);
-
+  CheckColumnFamilyMeta(cf_meta, kDefaultColumnFamilyName, files_by_level,
+                        start_time, end_time);
   std::vector<LiveFileMetaData> live_file_meta;
   db_->GetLiveFilesMetaData(&live_file_meta);
   CheckLiveFilesMeta(live_file_meta, files_by_level);
 }
 
+TEST_F(DBTest, AllMetaDataTest) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.disable_auto_compactions = true;
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  constexpr uint64_t blob_file_number = 234;
+  constexpr uint64_t total_blob_count = 555;
+  constexpr uint64_t total_blob_bytes = 66666;
+  constexpr char checksum_method[] = "CRC32";
+  constexpr char checksum_value[] = "\x3d\x87\xff\x57";
+
+  int64_t temp_time = 0;
+  options.env->GetCurrentTime(&temp_time).PermitUncheckedError();
+  uint64_t start_time = static_cast<uint64_t>(temp_time);
+
+  Random rnd(301);
+  dbfull()->TEST_LockMutex();
+  for (int cf = 0; cf < 2; cf++) {
+    AddBlobFile(handles_[cf], blob_file_number * (cf + 1),
+                total_blob_count * (cf + 1), total_blob_bytes * (cf + 1),
+                checksum_method, checksum_value);
+  }
+  dbfull()->TEST_UnlockMutex();
+
+  std::vector<ColumnFamilyMetaData> all_meta;
+  db_->GetAllColumnFamilyMetaData(&all_meta);
+
+  std::vector<std::vector<FileMetaData>> default_files_by_level;
+  std::vector<std::vector<FileMetaData>> pikachu_files_by_level;
+  dbfull()->TEST_GetFilesMetaData(handles_[0], &default_files_by_level);
+  dbfull()->TEST_GetFilesMetaData(handles_[1], &pikachu_files_by_level);
+
+  options.env->GetCurrentTime(&temp_time).PermitUncheckedError();
+  uint64_t end_time = static_cast<uint64_t>(temp_time);
+
+  ASSERT_EQ(all_meta.size(), 2);
+  for (int cf = 0; cf < 2; cf++) {
+    const auto& cfmd = all_meta[cf];
+    if (cf == 0) {
+      CheckColumnFamilyMeta(cfmd, "default", default_files_by_level, start_time,
+                            end_time);
+    } else {
+      CheckColumnFamilyMeta(cfmd, "pikachu", pikachu_files_by_level, start_time,
+                            end_time);
+    }
+    ASSERT_EQ(cfmd.blob_files.size(), 1U);
+    const auto& bmd = cfmd.blob_files[0];
+    ASSERT_EQ(cfmd.blob_file_count, 1U);
+    ASSERT_EQ(cfmd.blob_file_size, bmd.blob_file_size);
+    ASSERT_EQ(NormalizePath(bmd.blob_file_path), NormalizePath(dbname_));
+    CheckBlobMetaData(bmd, blob_file_number * (cf + 1),
+                      total_blob_count * (cf + 1), total_blob_bytes * (cf + 1),
+                      checksum_method, checksum_value);
+  }
+}
+
 namespace {
 void MinLevelHelper(DBTest* self, Options& options) {
   Random rnd(301);
@@ -2344,41 +2450,19 @@ TEST_F(DBTest, GetLiveBlobFiles) {
   Options options = CurrentOptions();
   options.stats_dump_period_sec = 0;
 
-  Reopen(options);
-
-  VersionSet* const versions = dbfull()->TEST_GetVersionSet();
-  assert(versions);
-  assert(versions->GetColumnFamilySet());
-
-  ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
-  assert(cfd);
-
-  Version* const version = cfd->current();
-  assert(version);
-
-  VersionStorageInfo* const storage_info = version->storage_info();
-  assert(storage_info);
-
-  // Add a live blob file.
   constexpr uint64_t blob_file_number = 234;
   constexpr uint64_t total_blob_count = 555;
   constexpr uint64_t total_blob_bytes = 66666;
   constexpr char checksum_method[] = "CRC32";
-  constexpr char checksum_value[] = "3d87ff57";
-
-  auto shared_meta = SharedBlobFileMetaData::Create(
-      blob_file_number, total_blob_count, total_blob_bytes, checksum_method,
-      checksum_value);
-
+  constexpr char checksum_value[] = "\x3d\x87\xff\x57";
   constexpr uint64_t garbage_blob_count = 0;
   constexpr uint64_t garbage_blob_bytes = 0;
 
-  auto meta = BlobFileMetaData::Create(std::move(shared_meta),
-                                       BlobFileMetaData::LinkedSsts(),
-                                       garbage_blob_count, garbage_blob_bytes);
-
-  storage_info->AddBlobFile(std::move(meta));
+  Reopen(options);
 
+  AddBlobFile(db_->DefaultColumnFamily(), blob_file_number, total_blob_count,
+              total_blob_bytes, checksum_method, checksum_value,
+              garbage_blob_count, garbage_blob_bytes);
   // Make sure it appears in the results returned by GetLiveFiles.
   uint64_t manifest_size = 0;
   std::vector<std::string> files;
@@ -2386,6 +2470,19 @@ TEST_F(DBTest, GetLiveBlobFiles) {
 
   ASSERT_FALSE(files.empty());
   ASSERT_EQ(files[0], BlobFileName("", blob_file_number));
+
+  ColumnFamilyMetaData cfmd;
+
+  db_->GetColumnFamilyMetaData(&cfmd);
+  ASSERT_EQ(cfmd.blob_files.size(), 1);
+  const BlobMetaData& bmd = cfmd.blob_files[0];
+
+  CheckBlobMetaData(bmd, blob_file_number, total_blob_count, total_blob_bytes,
+                    checksum_method, checksum_value, garbage_blob_count,
+                    garbage_blob_bytes);
+  ASSERT_EQ(NormalizePath(bmd.blob_file_path), NormalizePath(dbname_));
+  ASSERT_EQ(cfmd.blob_file_count, 1U);
+  ASSERT_EQ(cfmd.blob_file_size, bmd.blob_file_size);
 }
 #endif
 
@@ -3114,7 +3211,7 @@ class ModelDB : public DB {
   std::string name_ = "";
 };
 
-#ifndef ROCKSDB_VALGRIND_RUN
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 static std::string RandomKey(Random* rnd, int minimum = 0) {
   int len;
   do {
@@ -3269,7 +3366,7 @@ TEST_P(DBTestRandomized, Randomized) {
   if (model_snap != nullptr) model.ReleaseSnapshot(model_snap);
   if (db_snap != nullptr) db_->ReleaseSnapshot(db_snap);
 }
-#endif  // ROCKSDB_VALGRIND_RUN
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 TEST_F(DBTest, BlockBasedTablePrefixIndexTest) {
   // create a DB with block prefix index

db/db_test_util.cc

@@ -68,9 +68,13 @@ DBTestBase::DBTestBase(const std::string path, bool env_do_fsync)
 #ifndef ROCKSDB_LITE
   if (getenv("ENCRYPTED_ENV")) {
     std::shared_ptr<EncryptionProvider> provider;
-    Status s = EncryptionProvider::CreateFromString(
-        config_options, std::string("test://") + getenv("ENCRYPTED_ENV"),
-        &provider);
+    std::string provider_id = getenv("ENCRYPTED_ENV");
+    if (provider_id.find("=") == std::string::npos &&
+        !EndsWith(provider_id, "://test")) {
+      provider_id = provider_id + "://test";
+    }
+    EXPECT_OK(EncryptionProvider::CreateFromString(ConfigOptions(), provider_id,
+                                                   &provider));
     encrypted_env_ = NewEncryptedEnv(mem_env_ ? mem_env_ : base_env, provider);
   }
 #endif  // !ROCKSDB_LITE

db/db_test_util.h

@@ -403,7 +403,7 @@ class SpecialEnv : public EnvWrapper {
       Status Sync() override {
         ++env_->sync_counter_;
         if (env_->corrupt_in_sync_) {
-          Append(std::string(33000, ' '));
+          EXPECT_OK(Append(std::string(33000, ' ')));
           return Status::IOError("Ingested Sync Failure");
         }
         if (env_->skip_fsync_) {
@@ -675,6 +675,14 @@ class SpecialEnv : public EnvWrapper {
     }
   }
 
+  Status RenameFile(const std::string& src, const std::string& dest) override {
+    rename_count_.fetch_add(1);
+    if (rename_error_.load(std::memory_order_acquire)) {
+      return Status::NotSupported("Simulated `RenameFile()` error.");
+    }
+    return target()->RenameFile(src, dest);
+  }
+
   // Something to return when mocking current time
   const int64_t maybe_starting_time_;
 
@@ -702,6 +710,9 @@ class SpecialEnv : public EnvWrapper {
   // Force write to log files to fail while this pointer is non-nullptr
   std::atomic<bool> log_write_error_;
 
+  // Force `RenameFile()` to fail while this pointer is non-nullptr
+  std::atomic<bool> rename_error_{false};
+
   // Slow down every log write, in micro-seconds.
   std::atomic<int> log_write_slowdown_;
 
@@ -745,6 +756,8 @@ class SpecialEnv : public EnvWrapper {
 
   std::atomic<int> delete_count_;
 
+  std::atomic<int> rename_count_{0};
+
   std::atomic<bool> is_wal_sync_thread_safe_{true};
 
   std::atomic<size_t> compaction_readahead_size_{};
@@ -764,6 +777,8 @@ class SpecialEnv : public EnvWrapper {
 class OnFileDeletionListener : public EventListener {
  public:
   OnFileDeletionListener() : matched_count_(0), expected_file_name_("") {}
+  const char* Name() const override { return kClassName(); }
+  static const char* kClassName() { return "OnFileDeletionListener"; }
 
   void SetExpectedFileName(const std::string file_name) {
     expected_file_name_ = file_name;
@@ -788,6 +803,8 @@ class OnFileDeletionListener : public EventListener {
 
 class FlushCounterListener : public EventListener {
  public:
+  const char* Name() const override { return kClassName(); }
+  static const char* kClassName() { return "FlushCounterListener"; }
   std::atomic<int> count{0};
   std::atomic<FlushReason> expected_flush_reason{FlushReason::kOthers};
 

db/db_universal_compaction_test.cc

@@ -45,7 +45,7 @@ class DBTestUniversalCompaction : public DBTestUniversalCompactionBase {
 class DBTestUniversalCompaction2 : public DBTestBase {
  public:
   DBTestUniversalCompaction2()
-      : DBTestBase("/db_universal_compaction_test2", /*env_do_fsync=*/false) {}
+      : DBTestBase("db_universal_compaction_test2", /*env_do_fsync=*/false) {}
 };
 
 namespace {
@@ -635,7 +635,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTargetLevel) {
   ASSERT_EQ("0,0,0,0,1", FilesPerLevel(0));
 }
 
-#ifndef ROCKSDB_VALGRIND_RUN
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 class DBTestUniversalCompactionMultiLevels
     : public DBTestUniversalCompactionBase {
  public:
@@ -912,7 +912,7 @@ TEST_P(DBTestUniversalCompactionParallel, PickByFileNumberBug) {
 INSTANTIATE_TEST_CASE_P(Parallel, DBTestUniversalCompactionParallel,
                         ::testing::Combine(::testing::Values(1, 10),
                                            ::testing::Values(false)));
-#endif  // ROCKSDB_VALGRIND_RUN
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 TEST_P(DBTestUniversalCompaction, UniversalCompactionOptions) {
   Options options = CurrentOptions();
@@ -1128,7 +1128,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionCompressRatio2) {
   ASSERT_LT(TotalSize(), 120000U * 12 * 0.82 + 120000 * 2);
 }
 
-#ifndef ROCKSDB_VALGRIND_RUN
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 // Test that checks trivial move in universal compaction
 TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest1) {
   int32_t trivial_move = 0;
@@ -1221,7 +1221,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {
 
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
 }
-#endif  // ROCKSDB_VALGRIND_RUN
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 TEST_P(DBTestUniversalCompaction, UniversalCompactionFourPaths) {
   Options options = CurrentOptions();

db/db_wal_test.cc

@@ -119,7 +119,7 @@ class EnrichedSpecialEnv : public SpecialEnv {
 class DBWALTestWithEnrichedEnv : public DBTestBase {
  public:
   DBWALTestWithEnrichedEnv()
-      : DBTestBase("/db_wal_test", /*env_do_fsync=*/true) {
+      : DBTestBase("db_wal_test", /*env_do_fsync=*/true) {
     enriched_env_ = new EnrichedSpecialEnv(env_->target());
     auto options = CurrentOptions();
     options.env = enriched_env_;

db/db_with_timestamp_basic_test.cc

@@ -1494,6 +1494,7 @@ TEST_F(DBBasicTestWithTimestamp, CompactDeletionWithTimestampMarkerToBottom) {
   Close();
 }
 
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 class DBBasicTestWithTimestampFilterPrefixSettings
     : public DBBasicTestWithTimestampBase,
       public testing::WithParamInterface<
@@ -1625,6 +1626,7 @@ INSTANTIATE_TEST_CASE_P(
             BlockBasedTableOptions::IndexType::kHashSearch,
             BlockBasedTableOptions::IndexType::kTwoLevelIndexSearch,
             BlockBasedTableOptions::IndexType::kBinarySearchWithFirstKey)));
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 class DataVisibilityTest : public DBBasicTestWithTimestampBase {
  public:
@@ -2215,6 +2217,7 @@ TEST_F(DataVisibilityTest, MultiGetCrossCF) {
   Close();
 }
 
+#if !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 class DBBasicTestWithTimestampCompressionSettings
     : public DBBasicTestWithTimestampBase,
       public testing::WithParamInterface<
@@ -2954,6 +2957,7 @@ INSTANTIATE_TEST_CASE_P(
             BlockBasedTableOptions::IndexType::kHashSearch,
             BlockBasedTableOptions::IndexType::kTwoLevelIndexSearch,
             BlockBasedTableOptions::IndexType::kBinarySearchWithFirstKey)));
+#endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
 
 }  // namespace ROCKSDB_NAMESPACE
 

db/db_with_timestamp_compaction_test.cc

@@ -31,7 +31,7 @@ std::string Timestamp(uint64_t ts) {
 class TimestampCompatibleCompactionTest : public DBTestBase {
  public:
   TimestampCompatibleCompactionTest()
-      : DBTestBase("/ts_compatible_compaction_test", /*env_do_fsync=*/true) {}
+      : DBTestBase("ts_compatible_compaction_test", /*env_do_fsync=*/true) {}
 
   std::string Get(const std::string& key, uint64_t ts) {
     ReadOptions read_opts;