Fix a nuisance compiler warning from clang (#12144)

Summary:
Example:

```
cache/clock_cache.cc:56:7: error: fallthrough annotation in unreachable code [-Werror,-Wimplicit-fallthrough]
      FALLTHROUGH_INTENDED;
      ^
./port/lang.h:10:30: note: expanded from macro 'FALLTHROUGH_INTENDED'
                             ^
```

In clang < 14, this is annoyingly generated from -Wimplicit-fallthrough, but was changed to -Wunreachable-code-fallthrough (implied by -Wunreachable-code) in clang 14. See https://reviews.llvm.org/D107933 for how this nuisance pattern generated false positives similar to ours in the Linux kernel.

Just to underscore the ridiculousness of this warning, here an error is reported on the annotation, not the call to do_something(), depending on the constexpr value (https://godbolt.org/z/EvxqdPTdr):

```
#include <atomic>
void do_something();
void test(int v) {
    switch (v) {
        case 1:
            if constexpr (std::atomic<long>::is_always_lock_free) {
                return;
            } else {
                do_something();
                [[fallthrough]];
            }
        case 2:
            return;
    }
}
```

Pull Request resolved: https://github.com/facebook/rocksdb/pull/12144

Test Plan: Added the warning to our Makefile for USE_CLANG, which reproduced the warning-as-error as shown above, but is now fixed.

Reviewed By: jaykorean

Differential Revision: D52139615

Pulled By: pdillinger

fbshipit-source-id: ba967ae700c0916d1a478bc465cf917633e337d9

.circleci/config.yml

@@ -54,6 +54,7 @@ commands:
           command: |
             echo "Installing CMake..."
             choco install cmake --installargs 'ADD_CMAKE_TO_PATH=System' -y
+            choco install liberica8jdk -y
             mkdir $Env:THIRDPARTY_HOME
             cd $Env:THIRDPARTY_HOME
             echo "Building Snappy dependency..."
@@ -67,9 +68,10 @@ commands:
       - run:
           name: "Build RocksDB"
           command: |
+            $env:Path = $env:JAVA_HOME + ";" + $env:Path
             mkdir build
             cd build
-            & $Env:CMAKE_BIN -G "$Env:CMAKE_GENERATOR" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DJNI=1 ..
+            & $Env:CMAKE_BIN -G "$Env:CMAKE_GENERATOR" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE="$Env:CMAKE_PORTABLE" -DSNAPPY=1 -DJNI=1 ..
             cd ..
             echo "Building with VS version: $Env:CMAKE_GENERATOR"
             msbuild.exe build/rocksdb.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
@@ -78,6 +80,11 @@ commands:
           shell: powershell.exe
           command: |
             build_tools\run_ci_db_test.ps1 -SuiteRun arena_test,db_basic_test,db_test,db_test2,db_merge_operand_test,bloom_test,c_test,coding_test,crc32c_test,dynamic_bloom_test,env_basic_test,env_test,hash_test,random_test -Concurrency 16
+      - run:
+          name: "Test RocksJava"
+          command: |
+            cd build\java
+            & $Env:CTEST_BIN -C Debug -j 16
   pre-steps-macos:
       steps:
         - pre-steps
@@ -105,6 +112,15 @@ commands:
           path: /tmp/core_dumps
           when: on_fail
 
+  post-pmd-steps:
+    steps:
+      - store_artifacts:
+          path: /home/circleci/project/java/target/pmd.xml
+          when: on_fail
+      - store_artifacts:
+          path: /home/circleci/project/java/target/site
+          when: on_fail
+
   upgrade-cmake:
     steps:
       - run:
@@ -127,6 +143,13 @@ commands:
           command: |
             HOMEBREW_NO_AUTO_UPDATE=1 brew install gflags
 
+  install-maven:
+    steps:
+      - run:
+          name: Install maven
+          command: |
+            sudo apt-get update -y && sudo apt-get install -y maven
+
   setup-folly:
     steps:
       - run:
@@ -206,6 +229,9 @@ executors:
       #  $ docker run --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -it zjay437/rocksdb:0.5 bash
       # option `--cap-add=SYS_PTRACE --security-opt seccomp=unconfined` is used to enable gdb to attach an existing process
       - image: zjay437/rocksdb:0.6
+  linux-java-docker:
+    docker:
+      - image: evolvedbinary/rocksjava:centos6_x64-be
 
 jobs:
   build-macos:
@@ -542,6 +568,24 @@ jobs:
           no_output_timeout: 100m
       - post-steps
 
+  build-windows-vs2022-avx2:
+    executor:
+      name: win/server-2022
+      size: 2xlarge
+    environment:
+      THIRDPARTY_HOME: C:/Users/circleci/thirdparty
+      CMAKE_HOME: C:/Program Files/CMake
+      CMAKE_BIN: C:/Program Files/CMake/bin/cmake.exe
+      CTEST_BIN: C:/Program Files/CMake/bin/ctest.exe
+      JAVA_HOME: C:/Program Files/BellSoft/LibericaJDK-8
+      SNAPPY_HOME: C:/Users/circleci/thirdparty/snappy-1.1.8
+      SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.8;C:/Users/circleci/thirdparty/snappy-1.1.8/build
+      SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
+      CMAKE_GENERATOR: Visual Studio 17 2022
+      CMAKE_PORTABLE: AVX2
+    steps:
+      - windows-build-steps
+
   build-windows-vs2022:
     executor:
       name: win/server-2022
@@ -550,10 +594,13 @@ jobs:
       THIRDPARTY_HOME: C:/Users/circleci/thirdparty
       CMAKE_HOME: C:/Program Files/CMake
       CMAKE_BIN: C:/Program Files/CMake/bin/cmake.exe
+      CTEST_BIN: C:/Program Files/CMake/bin/ctest.exe
+      JAVA_HOME: C:/Program Files/BellSoft/LibericaJDK-8
       SNAPPY_HOME: C:/Users/circleci/thirdparty/snappy-1.1.8
       SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.8;C:/Users/circleci/thirdparty/snappy-1.1.8/build
       SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
       CMAKE_GENERATOR: Visual Studio 17 2022
+      CMAKE_PORTABLE: 1
     steps:
       - windows-build-steps
 
@@ -565,10 +612,13 @@ jobs:
       THIRDPARTY_HOME: C:/Users/circleci/thirdparty
       CMAKE_HOME: C:/Program Files/CMake
       CMAKE_BIN: C:/Program Files/CMake/bin/cmake.exe
+      CTEST_BIN: C:/Program Files/CMake/bin/ctest.exe
+      JAVA_HOME: C:/Program Files/BellSoft/LibericaJDK-8
       SNAPPY_HOME: C:/Users/circleci/thirdparty/snappy-1.1.8
       SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.8;C:/Users/circleci/thirdparty/snappy-1.1.8/build
       SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
       CMAKE_GENERATOR: Visual Studio 16 2019
+      CMAKE_PORTABLE: 1
     steps:
       - windows-build-steps
 
@@ -589,8 +639,29 @@ jobs:
           command: make V=1 J=8 -j8 jtest
       - post-steps
 
+  build-linux-java-pmd:
+    machine:
+      image: ubuntu-2004:202111-02
+    resource_class: large
+    environment:
+      JAVA_HOME: /usr/lib/jvm/java-8-openjdk-amd64
+    steps:
+      - install-maven
+      - pre-steps
+      - run:
+          name: "Set Java Environment"
+          command: |
+            echo "JAVA_HOME=${JAVA_HOME}"
+            echo 'export PATH=$JAVA_HOME/bin:$PATH' >> $BASH_ENV
+            which java && java -version
+            which javac && javac -version
+      - run:
+          name: "PMD RocksDBJava"
+          command: make V=1 J=8 -j8 jpmd
+      - post-pmd-steps
+
   build-linux-java-static:
-    executor: linux-docker
+    executor: linux-java-docker
     resource_class: large
     steps:
       - pre-steps
@@ -603,7 +674,7 @@ jobs:
             which javac && javac -version
       - run:
           name: "Build RocksDBJava Static Library"
-          command: make V=1 J=8 -j8 rocksdbjavastatic
+          command: scl enable devtoolset-7 'make V=1 J=8 -j8 rocksdbjavastatic'
       - post-steps
 
   build-macos-java:
@@ -843,6 +914,7 @@ workflows:
       - build-linux-mini-crashtest
   jobs-windows:
     jobs:
+      - build-windows-vs2022-avx2
       - build-windows-vs2022
       - build-windows-vs2019
       - build-cmake-mingw
@@ -853,6 +925,7 @@ workflows:
       - build-macos-java
       - build-macos-java-static
       - build-macos-java-static-universal
+      - build-linux-java-pmd
   jobs-macos:
     jobs:
       - build-macos

CMakeLists.txt

@@ -772,6 +772,7 @@ set(SOURCES
         options/configurable.cc
         options/customizable.cc
         options/db_options.cc
+        options/offpeak_time_info.cc
         options/options.cc
         options/options_helper.cc
         options/options_parser.cc
@@ -1105,11 +1106,15 @@ set(BUILD_VERSION_CC ${CMAKE_BINARY_DIR}/build_version.cc)
 configure_file(util/build_version.cc.in ${BUILD_VERSION_CC} @ONLY)
 
 add_library(${ROCKSDB_STATIC_LIB} STATIC ${SOURCES} ${BUILD_VERSION_CC})
+target_include_directories(${ROCKSDB_STATIC_LIB} PUBLIC
+  $<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}/include>)
 target_link_libraries(${ROCKSDB_STATIC_LIB} PRIVATE
   ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
 
 if(ROCKSDB_BUILD_SHARED)
   add_library(${ROCKSDB_SHARED_LIB} SHARED ${SOURCES} ${BUILD_VERSION_CC})
+  target_include_directories(${ROCKSDB_SHARED_LIB} PUBLIC
+    $<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}/include>)
   target_link_libraries(${ROCKSDB_SHARED_LIB} PRIVATE
     ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
 
@@ -1596,6 +1601,3 @@ option(WITH_BENCHMARK "build benchmark tests" OFF)
 if(WITH_BENCHMARK)
   add_subdirectory(${PROJECT_SOURCE_DIR}/microbench/)
 endif()
-
-target_include_directories(${PROJECT_NAME} PUBLIC
-  $<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}/include>)

HISTORY.md

@@ -1,6 +1,108 @@
 # Rocksdb Change Log
 > NOTE: Entries for next release do not go here. Follow instructions in `unreleased_history/README.txt`
 
+## 8.9.2 (12/12/2023)
+### Bug Fixes
+* A lookup by MultiGet in a TieredCache that goes to the local flash cache and finishes with very low latency, i.e before the subsequent call to WaitAll, is ignored, resulting in a false negative and a memory leak.
+* Sanitize the cache_opts->secondary_cache field in TieredCacheOptions. It can be mistakenly set by the user, which would conflict with nvm_sec_cache in TieredCacheOptions if not sanitized to nullptr.
+
+## 8.9.1 (12/8/2023)
+### Bug Fixes
+* Avoid destroying the periodic task scheduler's default timer in order to prevent static destruction order issues.
+
+## 8.9.0 (11/17/2023)
+### New Features
+* Add GetEntity() and PutEntity() API implementation for Attribute Group support. Through the use of Column Families, AttributeGroup enables users to logically group wide-column entities.
+
+### Public API Changes
+* Added rocksdb_ratelimiter_create_auto_tuned API to create an auto-tuned GenericRateLimiter.
+* Added clipColumnFamily() to the Java API to clip the entries in the CF according to the range [begin_key, end_key).
+* Make the `EnableFileDeletion` API not default to force enabling. For users that rely on this default behavior and still
+want to continue to use force enabling, they need to explicitly pass a `true` to `EnableFileDeletion`.
+* Add new Cache APIs GetSecondaryCacheCapacity() and GetSecondaryCachePinnedUsage() to return the configured capacity, and cache reservation charged to the secondary cache.
+
+### Behavior Changes
+* During off-peak hours defined by `daily_offpeak_time_utc`, the compaction picker will select a larger number of files for periodic compaction. This selection will include files that are projected to expire by the next off-peak start time, ensuring that these files are not chosen for periodic compaction outside of off-peak hours.
+* If an error occurs when writing to a trace file after `DB::StartTrace()`, the subsequent trace writes are skipped to avoid writing to a file that has previously seen error. In this case, `DB::EndTrace()` will also return a non-ok status with info about the error occured previously in its status message.
+* Deleting stale files upon recovery are delegated to SstFileManger if available so they can be rate limited.
+* Make RocksDB only call `TablePropertiesCollector::Finish()` once.
+* When `WAL_ttl_seconds > 0`, we now process archived WALs for deletion at least every `WAL_ttl_seconds / 2` seconds. Previously it could be less frequent in case of small `WAL_ttl_seconds` values when size-based expiration (`WAL_size_limit_MB > 0 `) was simultaneously enabled.
+
+### Bug Fixes
+* Fixed a crash or assertion failure bug in experimental new HyperClockCache variant, especially when running with a SecondaryCache.
+* Fix a race between flush error recovery and db destruction that can lead to db crashing.
+* Fixed some bugs in the index builder/reader path for user-defined timestamps in Memtable only feature.
+
+## 8.8.0 (10/23/2023)
+### New Features
+* Introduce AttributeGroup by adding the first AttributeGroup support API, MultiGetEntity(). Through the use of Column Families, AttributeGroup enables users to logically group wide-column entities. More APIs to support AttributeGroup will come soon, including GetEntity, PutEntity, and others.
+* Added new tickers `rocksdb.fifo.{max.size|ttl}.compactions` to count FIFO compactions that drop files for different reasons
+* Add an experimental offpeak duration awareness by setting `DBOptions::daily_offpeak_time_utc` in "HH:mm-HH:mm" format. This information will be used for resource optimization in the future
+* Users can now change the max bytes granted in a single refill period (i.e, burst) during runtime by `SetSingleBurstBytes()` for RocksDB rate limiter
+
+### Public API Changes
+* The default value of `DBOptions::fail_if_options_file_error` changed from `false` to `true`. Operations that set in-memory options (e.g., `DB::Open*()`, `DB::SetOptions()`, `DB::CreateColumnFamily*()`, and `DB::DropColumnFamily()`) but fail to persist the change will now return a non-OK `Status` by default.
+
+### Behavior Changes
+* For non direct IO, eliminate the file system prefetching attempt for compaction read when `Options::compaction_readahead_size` is 0
+* During a write stop, writes now block on in-progress recovery attempts
+
+### Bug Fixes
+* Fix a bug in auto_readahead_size where first_internal_key of index blocks wasn't copied properly resulting in corruption error when first_internal_key was used for comparison.
+* Fixed a bug where compaction read under non direct IO still falls back to RocksDB internal prefetching after file system's prefetching returns non-OK status other than `Status::NotSupported()`
+* Add bounds check in WBWIIteratorImpl and make BaseDeltaIterator, WriteUnpreparedTxn and WritePreparedTxn respect the upper bound and lower bound in ReadOption. See 11680.
+* Fixed the handling of wide-column base values in the `max_successive_merges` logic.
+* Fixed a rare race bug involving a concurrent combination of Create/DropColumnFamily and/or Set(DB)Options that could lead to inconsistency between (a) the DB's reported options state, (b) the DB options in effect, and (c) the latest persisted OPTIONS file.
+* Fixed a possible underflow when computing the compressed secondary cache share of memory reservations while updating the compressed secondary to total block cache ratio.
+
+### Performance Improvements
+* Improved the I/O efficiency of DB::Open a new DB with `create_missing_column_families=true` and many column families.
+
+## 8.7.0 (09/22/2023)
+### New Features
+* Added an experimental new "automatic" variant of HyperClockCache that does not require a prior estimate of the average size of cache entries. This variant is activated when HyperClockCacheOptions::estimated\_entry\_charge = 0 and has essentially the same concurrency benefits as the existing HyperClockCache.
+* Add a new statistic `COMPACTION_CPU_TOTAL_TIME` that records cumulative compaction cpu time. This ticker is updated regularly while a compaction is running.
+* Add `GetEntity()` API for ReadOnly DB and Secondary DB.
+* Add a new iterator API `Iterator::Refresh(const Snapshot *)` that allows iterator to be refreshed while using the input snapshot to read.
+* Added a new read option `merge_operand_count_threshold`. When the number of merge operands applied during a successful point lookup exceeds this threshold, the query will return a special OK status with a new subcode `kMergeOperandThresholdExceeded`. Applications might use this signal to take action to reduce the number of merge operands for the affected key(s), for example by running a compaction.
+* For `NewRibbonFilterPolicy()`, made the `bloom_before_level` option mutable through the Configurable interface and the SetOptions API, allowing dynamic switching between all-Bloom and all-Ribbon configurations, and configurations in between. See comments on `NewRibbonFilterPolicy()`
+* RocksDB now allows the block cache to be stacked on top of a compressed secondary cache and a non-volatile secondary cache, thus creating a three-tier cache. To set it up, use the `NewTieredCache()` API in rocksdb/cache.h..
+* Added a new wide-column aware full merge API called `FullMergeV3` to `MergeOperator`. `FullMergeV3` supports wide columns both as base value and merge result, which enables the application to perform more general transformations during merges. For backward compatibility, the default implementation implements the earlier logic of applying the merge operation to the default column of any wide-column entities. Specifically, if there is no base value or the base value is a plain key-value, the default implementation falls back to `FullMergeV2`. If the base value is a wide-column entity, the default implementation invokes `FullMergeV2` to perform the merge on the default column, and leaves any other columns unchanged.
+* Add wide column support to ldb commands (scan, dump, idump, dump_wal) and sst_dump tool's scan command
+
+### Public API Changes
+* Expose more information about input files used in table creation (if any) in `CompactionFilter::Context`. See `CompactionFilter::Context::input_start_level`,`CompactionFilter::Context::input_table_properties` for more.
+* `Options::compaction_readahead_size` 's default value is changed from 0 to 2MB.
+* When using LZ4 compression, the `acceleration` parameter is configurable by setting the negated value in `CompressionOptions::level`. For example, `CompressionOptions::level=-10` will set `acceleration=10`
+* The `NewTieredCache` API has been changed to take the total cache capacity (inclusive of both the primary and the compressed secondary cache) and the ratio of total capacity to allocate to the compressed cache. These are specified in `TieredCacheOptions`. Any capacity specified in `LRUCacheOptions`, `HyperClockCacheOptions` and `CompressedSecondaryCacheOptions` is ignored. A new API, `UpdateTieredCache` is provided to dynamically update the total capacity, ratio of compressed cache, and admission policy.
+* The `NewTieredVolatileCache()` API in rocksdb/cache.h has been renamed to `NewTieredCache()`.
+
+### Behavior Changes
+* Compaction read performance will regress when `Options::compaction_readahead_size` is explicitly set to 0
+* Universal size amp compaction will conditionally exclude some of the newest L0 files when selecting input with a small negative impact to size amp. This is to prevent a large number of L0 files from being locked by a size amp compaction, potentially leading to write stop with a few more flushes.
+* Change ldb scan command delimiter from ':' to '==>'.
+
+### Bug Fixes
+* Fix a bug where if there is an error reading from offset 0 of a file from L1+ and that the file is not the first file in the sorted run, data can be lost in compaction and read/scan can return incorrect results.
+* Fix a bug where iterator may return incorrect result for DeleteRange() users if there was an error reading from a file.
+* Fix a bug with atomic_flush=true that can cause DB to stuck after a flush fails (#11872).
+* Fix a bug where RocksDB (with atomic_flush=false) can delete output SST files of pending flushes when a previous concurrent flush fails (#11865). This can result in DB entering read-only state with error message like `IO error: No such file or directory: While open a file for random read: /tmp/rocksdbtest-501/db_flush_test_87732_4230653031040984171/000013.sst`.
+* Fix an assertion fault during seek with async_io when readahead trimming is enabled.
+* When the compressed secondary cache capacity is reduced to 0, it should be completely disabled. Before this fix, inserts and lookups would still go to the backing `LRUCache` before returning, thus incurring locking overhead. With this fix, inserts and lookups are no-ops and do not add any overhead.
+* Updating the tiered cache (cache allocated using NewTieredCache()) by calling SetCapacity() on it was not working properly. The initial creation would set the primary cache capacity to the combined primary and compressed secondary cache capacity. But SetCapacity() would just set the primary cache capacity. With this fix, the user always specifies the total budget and compressed secondary cache ratio on creation. Subsequently, SetCapacity() will distribute the new capacity across the two caches by the same ratio.
+* Fixed a bug in `MultiGet` for cleaning up SuperVersion acquired with locking db mutex.
+* Fix a bug where row cache can falsely return kNotFound even though row cache entry is hit.
+* Fixed a race condition in `GenericRateLimiter` that could cause it to stop granting requests
+* Fix a bug (Issue #10257) where DB can hang after write stall since no compaction is scheduled (#11764).
+* Add a fix for async_io where during seek, when reading a block for seeking a target key in a file without any readahead, the iterator aligned the read on a page boundary and reading more than necessary. This increased the storage read bandwidth usage.
+* Fix an issue in sst dump tool to handle bounds specified for data with user-defined timestamps.
+* When auto_readahead_size is enabled, update readahead upper bound during readahead trimming when reseek changes iterate_upper_bound dynamically.
+* Fixed a bug where `rocksdb.file.read.verify.file.checksums.micros` is not populated
+
+### Performance Improvements
+* Added additional improvements in tuning readahead_size during Scans when auto_readahead_size is enabled. However it's not supported with Iterator::Prev operation and will return NotSupported error.
+* During async_io, the Seek happens in 2 phases. Phase 1 starts an asynchronous read on a block cache miss, and phase 2 waits for it to complete and finishes the seek. In both phases, it tries to lookup the block cache for the data block first before looking in the prefetch buffer. It's optimized by doing the block cache lookup only in the first phase that would save some CPU.
+
 ## 8.6.0 (08/18/2023)
 ### New Features
 * Added enhanced data integrity checking on SST files with new format_version=6. Performance impact is very small or negligible. Previously if SST data was misplaced or re-arranged by the storage layer, it could pass block checksum with higher than 1 in 4 billion probability. With format_version=6, block checksums depend on what file they are in and location within the file. This way, misplaced SST data is no more likely to pass checksum verification than randomly corrupted data. Also in format_version=6, SST footers are checksum-protected.

INSTALL.md

@@ -17,7 +17,7 @@ There are few options when compiling RocksDB:
 * `make check` will compile and run all the unit tests. `make check` will compile RocksDB in debug mode.
 
 * `make all` will compile our static library, and all our tools and unit tests. Our tools
-depend on gflags. You will need to have gflags installed to run `make all`. This will compile RocksDB in debug mode. Don't
+depend on gflags 2.2.0 or newer. You will need to have gflags installed to run `make all`. This will compile RocksDB in debug mode. Don't
 use binaries compiled by `make all` in production.
 
 * By default the binary we produce is optimized for the CPU you're compiling on
@@ -77,7 +77,7 @@ most processors made since roughly 2013.
 
               git clone https://github.com/gflags/gflags.git
               cd gflags
-              git checkout v2.0
+              git checkout v2.2.0
               ./configure && make && sudo make install
 
       **Notice**: Once installed, please add the include path for gflags to your `CPATH` environment variable and the

Makefile

@@ -539,7 +539,7 @@ endif
 
 ifdef USE_CLANG
 	# Used by some teams in Facebook
-	WARNING_FLAGS += -Wshift-sign-overflow -Wambiguous-reversed-operator
+	WARNING_FLAGS += -Wshift-sign-overflow -Wambiguous-reversed-operator -Wimplicit-fallthrough
 endif
 
 ifeq ($(PLATFORM), OS_OPENBSD)
@@ -994,7 +994,7 @@ endif
 
 .PHONY: check_0
 check_0:
-	printf '%s\n' ''						\
+	@printf '%s\n' ''						\
 	  'To monitor subtest <duration,pass/fail,name>,'		\
 	  '  run "make watch-log" in a separate window' '';		\
 	{ \
@@ -1016,7 +1016,7 @@ valgrind-exclude-regexp = InlineSkipTest.ConcurrentInsert|TransactionStressTest.
 .PHONY: valgrind_check_0
 valgrind_check_0: test_log_prefix := valgrind_
 valgrind_check_0:
-	printf '%s\n' ''						\
+	@printf '%s\n' ''						\
 	  'To monitor subtest <duration,pass/fail,name>,'		\
 	  '  run "make watch-log" in a separate window' '';		\
 	{								\
@@ -2094,8 +2094,8 @@ ROCKSDB_JAVADOCS_JAR = rocksdbjni-$(ROCKSDB_JAVA_VERSION)-javadoc.jar
 ROCKSDB_SOURCES_JAR = rocksdbjni-$(ROCKSDB_JAVA_VERSION)-sources.jar
 SHA256_CMD = sha256sum
 
-ZLIB_VER ?= 1.2.13
-ZLIB_SHA256 ?= b3a24de97a8fdbc835b9833169501030b8977031bcb54b3b3ac13740f846ab30
+ZLIB_VER ?= 1.3
+ZLIB_SHA256 ?= ff0ba4c292013dbc27530b3a81e1f9a813cd39de01ca5e0f8bf355702efa593e
 ZLIB_DOWNLOAD_BASE ?= http://zlib.net
 BZIP2_VER ?= 1.0.8
 BZIP2_SHA256 ?= ab5a03176ee106d3f0fa90e381da478ddae405918153cca248e682cd0c4a2269
@@ -2103,11 +2103,11 @@ BZIP2_DOWNLOAD_BASE ?= http://sourceware.org/pub/bzip2
 SNAPPY_VER ?= 1.1.8
 SNAPPY_SHA256 ?= 16b677f07832a612b0836178db7f374e414f94657c138e6993cbfc5dcc58651f
 SNAPPY_DOWNLOAD_BASE ?= https://github.com/google/snappy/archive
-LZ4_VER ?= 1.9.3
-LZ4_SHA256 ?= 030644df4611007ff7dc962d981f390361e6c97a34e5cbc393ddfbe019ffe2c1
+LZ4_VER ?= 1.9.4
+LZ4_SHA256 ?= 0b0e3aa07c8c063ddf40b082bdf7e37a1562bda40a0ff5272957f3e987e0e54b
 LZ4_DOWNLOAD_BASE ?= https://github.com/lz4/lz4/archive
-ZSTD_VER ?= 1.4.9
-ZSTD_SHA256 ?= acf714d98e3db7b876e5b540cbf6dee298f60eb3c0723104f6d3f065cd60d6a8
+ZSTD_VER ?= 1.5.5
+ZSTD_SHA256 ?= 98e9c3d949d1b924e28e01eccb7deed865eefebf25c2f21c702e5cd5b63b85e1
 ZSTD_DOWNLOAD_BASE ?= https://github.com/facebook/zstd/archive
 CURL_SSL_OPTS ?= --tlsv1
 
@@ -2419,6 +2419,9 @@ jtest_run:
 jtest: rocksdbjava
 	cd java;$(MAKE) sample test
 
+jpmd: rocksdbjava rocksdbjavageneratepom
+	cd java;$(MAKE) pmd
+
 jdb_bench:
 	cd java;$(MAKE) db_bench;
 

PLUGINS.md

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

TARGETS

@@ -3,8 +3,6 @@
 # --> DO NOT EDIT MANUALLY <--
 # This file is a Facebook-specific integration for buck builds, so can
 # only be validated by Facebook employees.
-#
-# @noautodeps @nocodemods
 load("//rocks/buckifier:defs.bzl", "cpp_library_wrapper","rocks_cpp_library_wrapper","cpp_binary_wrapper","cpp_unittest_wrapper","fancy_bench_wrapper","add_c_test_wrapper")
 
 
@@ -165,6 +163,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "options/configurable.cc",
         "options/customizable.cc",
         "options/db_options.cc",
+        "options/offpeak_time_info.cc",
         "options/options.cc",
         "options/options_helper.cc",
         "options/options_parser.cc",
@@ -395,6 +394,7 @@ rocks_cpp_library_wrapper(name="rocksdb_stress_lib", srcs=[
         "db_stress_tool/db_stress_stat.cc",
         "db_stress_tool/db_stress_test_base.cc",
         "db_stress_tool/db_stress_tool.cc",
+        "db_stress_tool/db_stress_wide_merge_operator.cc",
         "db_stress_tool/expected_state.cc",
         "db_stress_tool/expected_value.cc",
         "db_stress_tool/multi_ops_txns_stress.cc",
@@ -407,6 +407,8 @@ rocks_cpp_library_wrapper(name="rocksdb_stress_lib", srcs=[
 
 cpp_binary_wrapper(name="db_stress", srcs=["db_stress_tool/db_stress.cc"], deps=[":rocksdb_stress_lib"], extra_preprocessor_flags=[], extra_bench_libs=False)
 
+cpp_binary_wrapper(name="cache_bench", srcs=["cache/cache_bench.cc"], deps=[":rocksdb_cache_bench_tools_lib"], extra_preprocessor_flags=[], extra_bench_libs=False)
+
 cpp_binary_wrapper(name="ribbon_bench", srcs=["microbench/ribbon_bench.cc"], deps=[], extra_preprocessor_flags=[], extra_bench_libs=True)
 
 cpp_binary_wrapper(name="db_basic_bench", srcs=["microbench/db_basic_bench.cc"], deps=[], extra_preprocessor_flags=[], extra_bench_libs=True)
@@ -5619,3 +5621,5 @@ cpp_unittest_wrapper(name="write_unprepared_transaction_test",
             deps=[":rocksdb_test_lib"],
             extra_compiler_flags=[])
 
+
+export_file(name = "tools/db_crashtest.py")

USERS.md

@@ -152,6 +152,9 @@ LzLabs is using RocksDB as a storage engine in their multi-database distributed
 ## ArangoDB
 [ArangoDB](https://www.arangodb.com/) is a native multi-model database with flexible data models for documents, graphs, and key-values, for building high performance applications using a convenient SQL-like query language or JavaScript extensions. It uses RocksDB as its storage engine.
 
+## Qdrant
+[Qdrant](https://qdrant.tech/) is an open source vector database, it [uses](https://qdrant.tech/documentation/concepts/storage/) RocksDB as its persistent storage.
+
 ## Milvus
 [Milvus](https://milvus.io/) is an open source vector database for unstructured data. It uses RocksDB not only as one of the supported kv storage engines, but also as a message queue.
 

buckifier/buckify_rocksdb.py

@@ -197,6 +197,10 @@ def generate_targets(repo_path, deps_map):
     TARGETS.add_binary(
         "db_stress", ["db_stress_tool/db_stress.cc"], [":rocksdb_stress_lib"]
     )
+    # cache_bench binary
+    TARGETS.add_binary(
+        "cache_bench", ["cache/cache_bench.cc"], [":rocksdb_cache_bench_tools_lib"]
+    )
     # bench binaries
     for src in src_mk.get("MICROBENCH_SOURCES", []):
         name = src.rsplit("/", 1)[1].split(".")[0] if "/" in src else src.split(".")[0]
@@ -299,6 +303,7 @@ def generate_targets(repo_path, deps_map):
                     deps=json.dumps(deps["extra_deps"] + [":rocksdb_test_lib"]),
                     extra_compiler_flags=json.dumps(deps["extra_compiler_flags"]),
                 )
+    TARGETS.export_file("tools/db_crashtest.py")
 
     print(ColorString.info("Generated TARGETS Summary:"))
     print(ColorString.info("- %d libs" % TARGETS.total_lib))

buckifier/targets_builder.py

@@ -148,3 +148,9 @@ add_c_test_wrapper()
                 ).encode("utf-8")
             )
         self.total_test = self.total_test + 1
+
+    def export_file(self, name):
+        with open(self.path, "a") as targets_file:
+            targets_file.write(
+                targets_cfg.export_file_template.format(name=name)
+            )

buckifier/targets_cfg.py

@@ -6,8 +6,6 @@ rocksdb_target_header_template = """# This file \100generated by:
 # --> DO NOT EDIT MANUALLY <--
 # This file is a Facebook-specific integration for buck builds, so can
 # only be validated by Facebook employees.
-#
-# @noautodeps @nocodemods
 load("//rocks/buckifier:defs.bzl", "cpp_library_wrapper","rocks_cpp_library_wrapper","cpp_binary_wrapper","cpp_unittest_wrapper","fancy_bench_wrapper","add_c_test_wrapper")
 
 """
@@ -39,3 +37,7 @@ fancy_bench_template = """
 fancy_bench_wrapper(suite_name="{name}", binary_to_bench_to_metric_list_map={bench_config}, slow={slow}, expected_runtime={expected_runtime}, sl_iterations={sl_iterations}, regression_threshold={regression_threshold})
 
 """
+
+export_file_template = """
+export_file(name = "{name}")
+"""

build_tools/build_detect_platform

@@ -674,7 +674,7 @@ else
   fi
 
   if [[ "${PLATFORM}" == "OS_MACOSX" ]]; then
-    # For portability compile for macOS 10.14 or newer
+    # For portability compile for macOS 10.14 (2018) or newer
     COMMON_FLAGS="$COMMON_FLAGS -mmacosx-version-min=10.14"
     PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -mmacosx-version-min=10.14"
     # -mmacosx-version-min must come first here.

cache/cache_bench_tool.cc

@@ -63,11 +63,22 @@ DEFINE_int32(
 DEFINE_uint32(skew, 5, "Degree of skew in key selection. 0 = no skew");
 DEFINE_bool(populate_cache, true, "Populate cache before operations");
 
-DEFINE_uint32(lookup_insert_percent, 87,
+DEFINE_double(pinned_ratio, 0.25,
+              "Keep roughly this portion of entries pinned in cache.");
+DEFINE_double(
+    vary_capacity_ratio, 0.0,
+    "If greater than 0.0, will periodically vary the capacity between this "
+    "ratio less than full size and full size. If vary_capacity_ratio + "
+    "pinned_ratio is close to or exceeds 1.0, the cache might thrash.");
+
+DEFINE_uint32(lookup_insert_percent, 82,
               "Ratio of lookup (+ insert on not found) to total workload "
               "(expressed as a percentage)");
 DEFINE_uint32(insert_percent, 2,
               "Ratio of insert to total workload (expressed as a percentage)");
+DEFINE_uint32(blind_insert_percent, 5,
+              "Ratio of insert without keeping handle to total workload "
+              "(expressed as a percentage)");
 DEFINE_uint32(lookup_percent, 10,
               "Ratio of lookup to total workload (expressed as a percentage)");
 DEFINE_uint32(erase_percent, 1,
@@ -101,7 +112,6 @@ DEFINE_uint32(seed, 0, "Hashing/random seed to use. 0 = choose at random");
 
 DEFINE_string(secondary_cache_uri, "",
               "Full URI for creating a custom secondary cache object");
-static class std::shared_ptr<ROCKSDB_NAMESPACE::SecondaryCache> secondary_cache;
 
 DEFINE_string(cache_type, "lru_cache", "Type of block cache.");
 
@@ -200,16 +210,19 @@ class SharedState {
 
   bool Started() const { return start_; }
 
-  void AddLookupStats(uint64_t hits, uint64_t misses) {
+  void AddLookupStats(uint64_t hits, uint64_t misses, size_t pinned_count) {
     MutexLock l(&mu_);
     lookup_count_ += hits + misses;
     lookup_hits_ += hits;
+    pinned_count_ += pinned_count;
   }
 
   double GetLookupHitRatio() const {
     return 1.0 * lookup_hits_ / lookup_count_;
   }
 
+  size_t GetPinnedCount() const { return pinned_count_; }
+
  private:
   port::Mutex mu_;
   port::CondVar cv_;
@@ -221,6 +234,7 @@ class SharedState {
   uint64_t num_done_ = 0;
   uint64_t lookup_count_ = 0;
   uint64_t lookup_hits_ = 0;
+  size_t pinned_count_ = 0;
 };
 
 // Per-thread state for concurrent executions of the same benchmark.
@@ -313,6 +327,28 @@ Cache::CacheItemHelper helper2(CacheEntryRole::kIndexBlock, DeleteFn, SizeFn,
 Cache::CacheItemHelper helper3_wos(CacheEntryRole::kFilterBlock, DeleteFn);
 Cache::CacheItemHelper helper3(CacheEntryRole::kFilterBlock, DeleteFn, SizeFn,
                                SaveToFn, CreateFn, &helper3_wos);
+
+void ConfigureSecondaryCache(ShardedCacheOptions& opts) {
+  if (!FLAGS_secondary_cache_uri.empty()) {
+    std::shared_ptr<SecondaryCache> secondary_cache;
+    Status s = SecondaryCache::CreateFromString(
+        ConfigOptions(), FLAGS_secondary_cache_uri, &secondary_cache);
+    if (secondary_cache == nullptr) {
+      fprintf(stderr,
+              "No secondary cache registered matching string: %s status=%s\n",
+              FLAGS_secondary_cache_uri.c_str(), s.ToString().c_str());
+      exit(1);
+    }
+    opts.secondary_cache = secondary_cache;
+  }
+}
+
+ShardedCacheBase* AsShardedCache(Cache* c) {
+  if (!FLAGS_secondary_cache_uri.empty()) {
+    c = static_cast_with_check<CacheWrapper>(c)->GetTarget().get();
+  }
+  return static_cast_with_check<ShardedCacheBase>(c);
+}
 }  // namespace
 
 class CacheBench {
@@ -327,7 +363,9 @@ class CacheBench {
                                  FLAGS_lookup_insert_percent),
         insert_threshold_(lookup_insert_threshold_ +
                           kHundredthUint64 * FLAGS_insert_percent),
-        lookup_threshold_(insert_threshold_ +
+        blind_insert_threshold_(insert_threshold_ +
+                                kHundredthUint64 * FLAGS_blind_insert_percent),
+        lookup_threshold_(blind_insert_threshold_ +
                           kHundredthUint64 * FLAGS_lookup_percent),
         erase_threshold_(lookup_threshold_ +
                          kHundredthUint64 * FLAGS_erase_percent) {
@@ -366,6 +404,7 @@ class CacheBench {
         fprintf(stderr, "Cache type not supported.\n");
         exit(1);
       }
+      ConfigureSecondaryCache(opts);
       cache_ = opts.MakeSharedCache();
     } else if (FLAGS_cache_type == "lru_cache") {
       LRUCacheOptions opts(FLAGS_cache_size, FLAGS_num_shard_bits,
@@ -373,19 +412,7 @@ class CacheBench {
                            0.5 /* high_pri_pool_ratio */);
       opts.hash_seed = BitwiseAnd(FLAGS_seed, INT32_MAX);
       opts.memory_allocator = allocator;
-      if (!FLAGS_secondary_cache_uri.empty()) {
-        Status s = SecondaryCache::CreateFromString(
-            ConfigOptions(), FLAGS_secondary_cache_uri, &secondary_cache);
-        if (secondary_cache == nullptr) {
-          fprintf(
-              stderr,
-              "No secondary cache registered matching string: %s status=%s\n",
-              FLAGS_secondary_cache_uri.c_str(), s.ToString().c_str());
-          exit(1);
-        }
-        opts.secondary_cache = secondary_cache;
-      }
-
+      ConfigureSecondaryCache(opts);
       cache_ = NewLRUCache(opts);
     } else {
       fprintf(stderr, "Cache type not supported.\n");
@@ -505,6 +532,8 @@ class CacheBench {
     size_t slot = cache_->GetTableAddressCount();
     printf("Final load factor: %g (%zu / %zu)\n", 1.0 * occ / slot, occ, slot);
 
+    printf("Final pinned count: %zu\n", shared.GetPinnedCount());
+
     if (FLAGS_histograms) {
       printf("\nOperation latency (ns):\n");
       HistogramImpl combined;
@@ -536,6 +565,7 @@ class CacheBench {
   // Cumulative thresholds in the space of a random uint64_t
   const uint64_t lookup_insert_threshold_;
   const uint64_t insert_threshold_;
+  const uint64_t blind_insert_threshold_;
   const uint64_t lookup_threshold_;
   const uint64_t erase_threshold_;
 
@@ -651,29 +681,44 @@ class CacheBench {
     uint64_t lookup_misses = 0;
     uint64_t lookup_hits = 0;
     // To hold handles for a non-trivial amount of time
-    Cache::Handle* handle = nullptr;
+    std::deque<Cache::Handle*> pinned;
+    size_t total_pin_count = static_cast<size_t>(
+        (FLAGS_cache_size * FLAGS_pinned_ratio) / FLAGS_value_bytes + 0.999999);
+    // For this thread. Some round up, some round down, as appropriate
+    size_t pin_count = (total_pin_count + thread->tid) / FLAGS_threads;
+
     KeyGen gen;
     const auto clock = SystemClock::Default().get();
     uint64_t start_time = clock->NowMicros();
     StopWatchNano timer(clock);
     auto system_clock = SystemClock::Default();
+    size_t steps_to_next_capacity_change = 0;
 
     for (uint64_t i = 0; i < FLAGS_ops_per_thread; i++) {
       Slice key = gen.GetRand(thread->rnd, max_key_, FLAGS_skew);
       uint64_t random_op = thread->rnd.Next();
 
+      if (FLAGS_vary_capacity_ratio > 0.0 && thread->tid == 0) {
+        if (steps_to_next_capacity_change == 0) {
+          double cut_ratio = static_cast<double>(thread->rnd.Next()) /
+                             static_cast<double>(UINT64_MAX) *
+                             FLAGS_vary_capacity_ratio;
+          cache_->SetCapacity(FLAGS_cache_size * (1.0 - cut_ratio));
+          steps_to_next_capacity_change =
+              static_cast<size_t>(FLAGS_ops_per_thread / 100);
+        } else {
+          --steps_to_next_capacity_change;
+        }
+      }
+
       if (FLAGS_histograms) {
         timer.Start();
       }
 
       if (random_op < lookup_insert_threshold_) {
-        if (handle) {
-          cache_->Release(handle);
-          handle = nullptr;
-        }
         // do lookup
-        handle = cache_->Lookup(key, &helper2, /*context*/ nullptr,
-                                Cache::Priority::LOW);
+        auto handle = cache_->Lookup(key, &helper2, /*context*/ nullptr,
+                                     Cache::Priority::LOW);
         if (handle) {
           ++lookup_hits;
           if (!FLAGS_lean) {
@@ -681,32 +726,31 @@ class CacheBench {
             result += NPHash64(static_cast<char*>(cache_->Value(handle)),
                                FLAGS_value_bytes);
           }
+          pinned.push_back(handle);
         } else {
           ++lookup_misses;
           // do insert
           Status s = cache_->Insert(
               key, createValue(thread->rnd, cache_->memory_allocator()),
-              &helper2, FLAGS_value_bytes, &handle);
+              &helper2, FLAGS_value_bytes, &pinned.emplace_back());
           assert(s.ok());
         }
       } else if (random_op < insert_threshold_) {
-        if (handle) {
-          cache_->Release(handle);
-          handle = nullptr;
-        }
         // do insert
         Status s = cache_->Insert(
             key, createValue(thread->rnd, cache_->memory_allocator()), &helper3,
-            FLAGS_value_bytes, &handle);
+            FLAGS_value_bytes, &pinned.emplace_back());
+        assert(s.ok());
+      } else if (random_op < blind_insert_threshold_) {
+        // insert without keeping a handle
+        Status s = cache_->Insert(
+            key, createValue(thread->rnd, cache_->memory_allocator()), &helper3,
+            FLAGS_value_bytes);
         assert(s.ok());
       } else if (random_op < lookup_threshold_) {
-        if (handle) {
-          cache_->Release(handle);
-          handle = nullptr;
-        }
         // do lookup
-        handle = cache_->Lookup(key, &helper2, /*context*/ nullptr,
-                                Cache::Priority::LOW);
+        auto handle = cache_->Lookup(key, &helper2, /*context*/ nullptr,
+                                     Cache::Priority::LOW);
         if (handle) {
           ++lookup_hits;
           if (!FLAGS_lean) {
@@ -714,6 +758,7 @@ class CacheBench {
             result += NPHash64(static_cast<char*>(cache_->Value(handle)),
                                FLAGS_value_bytes);
           }
+          pinned.push_back(handle);
         } else {
           ++lookup_misses;
         }
@@ -727,7 +772,6 @@ class CacheBench {
       if (FLAGS_histograms) {
         thread->latency_ns_hist.Add(timer.ElapsedNanos());
       }
-      thread->shared->AddLookupStats(lookup_hits, lookup_misses);
       if (FLAGS_usleep > 0) {
         unsigned us =
             static_cast<unsigned>(thread->rnd.Uniform(FLAGS_usleep + 1));
@@ -735,12 +779,17 @@ class CacheBench {
           system_clock->SleepForMicroseconds(us);
         }
       }
+      while (pinned.size() > pin_count) {
+        cache_->Release(pinned.front());
+        pinned.pop_front();
+      }
     }
     if (FLAGS_early_exit) {
       MutexLock l(thread->shared->GetMutex());
       exit(0);
     }
-    if (handle) {
+    thread->shared->AddLookupStats(lookup_hits, lookup_misses, pinned.size());
+    for (auto handle : pinned) {
       cache_->Release(handle);
       handle = nullptr;
     }
@@ -769,8 +818,7 @@ class CacheBench {
     printf("Cache size          : %s\n",
            BytesToHumanString(FLAGS_cache_size).c_str());
     printf("Num shard bits      : %d\n",
-           static_cast_with_check<ShardedCacheBase>(cache_.get())
-               ->GetNumShardBits());
+           AsShardedCache(cache_.get())->GetNumShardBits());
     printf("Max key             : %" PRIu64 "\n", max_key_);
     printf("Resident ratio      : %g\n", FLAGS_resident_ratio);
     printf("Skew degree         : %u\n", FLAGS_skew);

cache/clock_cache.h

@@ -24,6 +24,7 @@
 #include "port/port.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/secondary_cache.h"
+#include "util/atomic.h"
 #include "util/autovector.h"
 #include "util/math.h"
 
@@ -368,7 +369,7 @@ struct ClockHandle : public ClockHandleBasicData {
   // TODO: make these coundown values tuning parameters for eviction?
 
   // See above. Mutable for read reference counting.
-  mutable std::atomic<uint64_t> meta{};
+  mutable AcqRelAtomic<uint64_t> meta{};
 };  // struct ClockHandle
 
 class BaseClockTable {
@@ -395,19 +396,15 @@ class BaseClockTable {
 
   void Ref(ClockHandle& handle);
 
-  size_t GetOccupancy() const {
-    return occupancy_.load(std::memory_order_relaxed);
-  }
+  size_t GetOccupancy() const { return occupancy_.LoadRelaxed(); }
 
-  size_t GetUsage() const { return usage_.load(std::memory_order_relaxed); }
+  size_t GetUsage() const { return usage_.LoadRelaxed(); }
 
-  size_t GetStandaloneUsage() const {
-    return standalone_usage_.load(std::memory_order_relaxed);
-  }
+  size_t GetStandaloneUsage() const { return standalone_usage_.LoadRelaxed(); }
 
   uint32_t GetHashSeed() const { return hash_seed_; }
 
-  uint64_t GetYieldCount() const { return yield_count_.load(); }
+  uint64_t GetYieldCount() const { return yield_count_.LoadRelaxed(); }
 
   struct EvictionData {
     size_t freed_charge = 0;
@@ -460,21 +457,23 @@ class BaseClockTable {
   // operations in ClockCacheShard.
 
   // Clock algorithm sweep pointer.
-  std::atomic<uint64_t> clock_pointer_{};
+  // (Relaxed: only needs to be consistent with itself.)
+  RelaxedAtomic<uint64_t> clock_pointer_{};
 
   // Counter for number of times we yield to wait on another thread.
-  std::atomic<uint64_t> yield_count_{};
+  // (Relaxed: a simple stat counter.)
+  RelaxedAtomic<uint64_t> yield_count_{};
 
   // TODO: is this separation needed if we don't do background evictions?
   ALIGN_AS(CACHE_LINE_SIZE)
   // Number of elements in the table.
-  std::atomic<size_t> occupancy_{};
+  AcqRelAtomic<size_t> occupancy_{};
 
   // Memory usage by entries tracked by the cache (including standalone)
-  std::atomic<size_t> usage_{};
+  AcqRelAtomic<size_t> usage_{};
 
   // Part of usage by standalone entries (not in table)
-  std::atomic<size_t> standalone_usage_{};
+  AcqRelAtomic<size_t> standalone_usage_{};
 
   ALIGN_AS(CACHE_LINE_SIZE)
   const CacheMetadataChargePolicy metadata_charge_policy_;
@@ -500,7 +499,11 @@ class FixedHyperClockTable : public BaseClockTable {
   struct ALIGN_AS(64U) HandleImpl : public ClockHandle {
     // The number of elements that hash to this slot or a lower one, but wind
     // up in this slot or a higher one.
-    std::atomic<uint32_t> displacements{};
+    // (Relaxed: within a Cache op, does not need consistency with entries
+    // inserted/removed during that op. For example, a Lookup() that
+    // happens-after an Insert() will see an appropriate displacements value
+    // for the entry to be in a published state.)
+    RelaxedAtomic<uint32_t> displacements{};
 
     // Whether this is a "deteched" handle that is independently allocated
     // with `new` (so must be deleted with `delete`).
@@ -787,17 +790,16 @@ class AutoHyperClockTable : public BaseClockTable {
 
     // See above. The head pointer is logically independent of the rest of
     // the entry, including the chain next pointer.
-    std::atomic<uint64_t> head_next_with_shift{kUnusedMarker};
-    std::atomic<uint64_t> chain_next_with_shift{kUnusedMarker};
+    AcqRelAtomic<uint64_t> head_next_with_shift{kUnusedMarker};
+    AcqRelAtomic<uint64_t> chain_next_with_shift{kUnusedMarker};
 
     // For supporting CreateStandalone and some fallback cases.
     inline bool IsStandalone() const {
-      return head_next_with_shift.load(std::memory_order_acquire) ==
-             kStandaloneMarker;
+      return head_next_with_shift.Load() == kStandaloneMarker;
     }
 
     inline void SetStandalone() {
-      head_next_with_shift.store(kStandaloneMarker, std::memory_order_release);
+      head_next_with_shift.Store(kStandaloneMarker);
     }
   };  // struct HandleImpl
 
@@ -939,22 +941,25 @@ class AutoHyperClockTable : public BaseClockTable {
   // log time to find the correct chain, but normally this value enables
   // readers to find the correct chain on the first try.
   //
-  // NOTES: length_info_ is only updated at the end of a Grow operation,
-  // so that waiting in Grow operations isn't done while entries are pinned
-  // for internal operation purposes. Thus, Lookup and Insert have to
-  // detect and support cases where length_info hasn't caught up to updated
-  // chains. Winning grow thread is the one that transitions
-  // head_next_with_shift from zeros. Grow threads can spin/yield wait for
-  // preconditions and postconditions to be met.
-  std::atomic<uint64_t> length_info_;
+  // To maximize parallelization of Grow() operations, this field is only
+  // updated opportunistically after Grow() operations and in DoInsert() where
+  // it is found to be out-of-date. See CatchUpLengthInfoNoWait().
+  AcqRelAtomic<uint64_t> length_info_;
 
   // An already-computed version of the usable length times the max load
   // factor. Could be slightly out of date but GrowIfNeeded()/Grow() handle
   // that internally.
-  std::atomic<size_t> occupancy_limit_;
+  // (Relaxed: allowed to lag behind length_info_ by a little)
+  RelaxedAtomic<size_t> occupancy_limit_;
+
+  // The next index to use from array_ upon the next Grow(). Might be ahead of
+  // length_info_.
+  // (Relaxed: self-contained source of truth for next grow home)
+  RelaxedAtomic<size_t> grow_frontier_;
 
   // See explanation in AutoHyperClockTable::Evict
-  std::atomic<size_t> clock_pointer_mask_;
+  // (Relaxed: allowed to lag behind clock_pointer_ and length_info_ state)
+  RelaxedAtomic<size_t> clock_pointer_mask_;
 };  // class AutoHyperClockTable
 
 // A single shard of sharded cache.
@@ -1070,10 +1075,12 @@ class ALIGN_AS(CACHE_LINE_SIZE) ClockCacheShard final : public CacheShardBase {
   Table table_;
 
   // Maximum total charge of all elements stored in the table.
-  std::atomic<size_t> capacity_;
+  // (Relaxed: eventual consistency/update is OK)
+  RelaxedAtomic<size_t> capacity_;
 
   // Whether to reject insertion if cache reaches its full capacity.
-  std::atomic<bool> strict_capacity_limit_;
+  // (Relaxed: eventual consistency/update is OK)
+  RelaxedAtomic<bool> strict_capacity_limit_;
 };  // class ClockCacheShard
 
 template <class Table>

cache/compressed_secondary_cache.cc

@@ -33,9 +33,13 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
     Cache::CreateContext* create_context, bool /*wait*/, bool advise_erase,
     bool& kept_in_sec_cache) {
   assert(helper);
+  // This is a minor optimization. Its ok to skip it in TSAN in order to
+  // avoid a false positive.
+#ifndef __SANITIZE_THREAD__
   if (disable_cache_) {
     return nullptr;
   }
+#endif
 
   std::unique_ptr<SecondaryCacheResultHandle> handle;
   kept_in_sec_cache = false;

cache/compressed_secondary_cache_test.cc

@@ -1224,13 +1224,11 @@ TEST_P(CompressedSecCacheTestWithTiered, DynamicUpdate) {
   ASSERT_OK(sec_cache->GetCapacity(sec_capacity));
   ASSERT_EQ(sec_capacity, 0);
 
-  ASSERT_NOK(UpdateTieredCache(tiered_cache, -1, 0.3));
-  // Only check usage for LRU cache. HCC shows a 64KB usage for some reason
-  if (std::get<0>(GetParam()) == PrimaryCacheType::kCacheTypeLRU) {
-    ASSERT_EQ(GetCache()->GetUsage(), 0);
-  }
+  ASSERT_OK(UpdateTieredCache(tiered_cache, -1, 0.3));
+  EXPECT_PRED3(CacheUsageWithinBounds, GetCache()->GetUsage(), (30 << 20),
+               GetPercent(30 << 20, 1));
   ASSERT_OK(sec_cache->GetCapacity(sec_capacity));
-  ASSERT_EQ(sec_capacity, 0);
+  ASSERT_EQ(sec_capacity, (30 << 20));
 }
 
 TEST_P(CompressedSecCacheTestWithTiered, DynamicUpdateWithReservation) {
@@ -1276,6 +1274,13 @@ TEST_P(CompressedSecCacheTestWithTiered, DynamicUpdateWithReservation) {
   ASSERT_OK(sec_cache->GetCapacity(sec_capacity));
   ASSERT_EQ(sec_capacity, (30 << 20));
 
+  ASSERT_OK(tiered_cache->GetSecondaryCacheCapacity(sec_capacity));
+  ASSERT_EQ(sec_capacity, 30 << 20);
+  size_t sec_usage;
+  ASSERT_OK(tiered_cache->GetSecondaryCachePinnedUsage(sec_usage));
+  EXPECT_PRED3(CacheUsageWithinBounds, sec_usage, 3 << 20,
+               GetPercent(3 << 20, 1));
+
   ASSERT_OK(UpdateTieredCache(tiered_cache, -1, 0.39));
   EXPECT_PRED3(CacheUsageWithinBounds, GetCache()->GetUsage(), (45 << 20),
                GetPercent(45 << 20, 1));
@@ -1309,6 +1314,49 @@ TEST_P(CompressedSecCacheTestWithTiered, DynamicUpdateWithReservation) {
   ASSERT_OK(sec_cache->GetCapacity(sec_capacity));
   ASSERT_EQ(sec_capacity, 0);
 
+  ASSERT_OK(UpdateTieredCache(tiered_cache, -1, 0.3));
+  EXPECT_PRED3(CacheUsageWithinBounds, GetCache()->GetUsage(), (37 << 20),
+               GetPercent(37 << 20, 1));
+  EXPECT_PRED3(CacheUsageWithinBounds, sec_cache->TEST_GetUsage(), (3 << 20),
+               GetPercent(3 << 20, 1));
+  ASSERT_OK(sec_cache->GetCapacity(sec_capacity));
+  ASSERT_EQ(sec_capacity, 30 << 20);
+
+  ASSERT_OK(cache_res_mgr()->UpdateCacheReservation(0));
+}
+
+TEST_P(CompressedSecCacheTestWithTiered, ReservationOverCapacity) {
+  CompressedSecondaryCache* sec_cache =
+      reinterpret_cast<CompressedSecondaryCache*>(GetSecondaryCache());
+  std::shared_ptr<Cache> tiered_cache = GetTieredCache();
+
+  ASSERT_OK(cache_res_mgr()->UpdateCacheReservation(110 << 20));
+  // Use EXPECT_PRED3 instead of EXPECT_NEAR to void too many size_t to
+  // double explicit casts
+  EXPECT_PRED3(CacheUsageWithinBounds, GetCache()->GetUsage(), (110 << 20),
+               GetPercent(110 << 20, 1));
+  EXPECT_PRED3(CacheUsageWithinBounds, sec_cache->TEST_GetUsage(), (30 << 20),
+               GetPercent(30 << 20, 1));
+  size_t sec_capacity;
+  ASSERT_OK(sec_cache->GetCapacity(sec_capacity));
+  ASSERT_EQ(sec_capacity, (30 << 20));
+
+  ASSERT_OK(UpdateTieredCache(tiered_cache, -1, 0.39));
+  EXPECT_PRED3(CacheUsageWithinBounds, GetCache()->GetUsage(), (110 << 20),
+               GetPercent(110 << 20, 1));
+  EXPECT_PRED3(CacheUsageWithinBounds, sec_cache->TEST_GetUsage(), (39 << 20),
+               GetPercent(39 << 20, 1));
+  ASSERT_OK(sec_cache->GetCapacity(sec_capacity));
+  ASSERT_EQ(sec_capacity, (39 << 20));
+
+  ASSERT_OK(cache_res_mgr()->UpdateCacheReservation(90 << 20));
+  EXPECT_PRED3(CacheUsageWithinBounds, GetCache()->GetUsage(), (94 << 20),
+               GetPercent(94 << 20, 1));
+  EXPECT_PRED3(CacheUsageWithinBounds, sec_cache->TEST_GetUsage(), (35 << 20),
+               GetPercent(35 << 20, 1));
+  ASSERT_OK(sec_cache->GetCapacity(sec_capacity));
+  ASSERT_EQ(sec_capacity, (39 << 20));
+
   ASSERT_OK(cache_res_mgr()->UpdateCacheReservation(0));
 }
 

cache/secondary_cache_adapter.cc

@@ -5,8 +5,11 @@
 
 #include "cache/secondary_cache_adapter.h"
 
+#include <atomic>
+
 #include "cache/tiered_secondary_cache.h"
 #include "monitoring/perf_context_imp.h"
+#include "test_util/sync_point.h"
 #include "util/cast_util.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -80,7 +83,10 @@ CacheWithSecondaryAdapter::CacheWithSecondaryAdapter(
     : CacheWrapper(std::move(target)),
       secondary_cache_(std::move(secondary_cache)),
       adm_policy_(adm_policy),
-      distribute_cache_res_(distribute_cache_res) {
+      distribute_cache_res_(distribute_cache_res),
+      placeholder_usage_(0),
+      reserved_usage_(0),
+      sec_reserved_(0) {
   target_->SetEvictionCallback(
       [this](const Slice& key, Handle* handle, bool was_hit) {
         return EvictionHandler(key, handle, was_hit);
@@ -109,10 +115,12 @@ CacheWithSecondaryAdapter::~CacheWithSecondaryAdapter() {
   // use after free
   target_->SetEvictionCallback({});
 #ifndef NDEBUG
-  if (distribute_cache_res_ && !ratio_changed_) {
+  if (distribute_cache_res_) {
     size_t sec_capacity = 0;
     Status s = secondary_cache_->GetCapacity(sec_capacity);
     assert(s.ok());
+    assert(placeholder_usage_ == 0);
+    assert(reserved_usage_ == 0);
     assert(pri_cache_res_->GetTotalMemoryUsed() == sec_capacity);
   }
 #endif  // NDEBUG
@@ -232,18 +240,38 @@ Status CacheWithSecondaryAdapter::Insert(const Slice& key, ObjectPtr value,
                                          const Slice& compressed_value,
                                          CompressionType type) {
   Status s = target_->Insert(key, value, helper, charge, handle, priority);
-  if (s.ok() && value == nullptr && distribute_cache_res_) {
-    size_t sec_charge = static_cast<size_t>(charge * (sec_cache_res_ratio_));
-    s = secondary_cache_->Deflate(sec_charge);
-    assert(s.ok());
-    s = pri_cache_res_->UpdateCacheReservation(sec_charge, /*increase=*/false);
-    assert(s.ok());
+  if (s.ok() && value == nullptr && distribute_cache_res_ && handle) {
+    charge = target_->GetCharge(*handle);
+
+    MutexLock l(&cache_res_mutex_);
+    placeholder_usage_ += charge;
+    // Check if total placeholder reservation is more than the overall
+    // cache capacity. If it is, then we don't try to charge the
+    // secondary cache because we don't want to overcharge it (beyond
+    // its capacity).
+    // In order to make this a bit more lightweight, we also check if
+    // the difference between placeholder_usage_ and reserved_usage_ is
+    // atleast kReservationChunkSize and avoid any adjustments if not.
+    if ((placeholder_usage_ <= target_->GetCapacity()) &&
+        ((placeholder_usage_ - reserved_usage_) >= kReservationChunkSize)) {
+      reserved_usage_ = placeholder_usage_ & ~(kReservationChunkSize - 1);
+      size_t new_sec_reserved =
+          static_cast<size_t>(reserved_usage_ * sec_cache_res_ratio_);
+      size_t sec_charge = new_sec_reserved - sec_reserved_;
+      s = secondary_cache_->Deflate(sec_charge);
+      assert(s.ok());
+      s = pri_cache_res_->UpdateCacheReservation(sec_charge,
+                                                 /*increase=*/false);
+      assert(s.ok());
+      sec_reserved_ += sec_charge;
+    }
   }
   // Warm up the secondary cache with the compressed block. The secondary
   // cache may choose to ignore it based on the admission policy.
-  if (value != nullptr && !compressed_value.empty()) {
+  if (value != nullptr && !compressed_value.empty() &&
+      adm_policy_ == TieredAdmissionPolicy::kAdmPolicyThreeQueue) {
     Status status = secondary_cache_->InsertSaved(key, compressed_value, type);
-    assert(status.ok());
+    assert(status.ok() || status.IsNotSupported());
   }
 
   return s;
@@ -281,11 +309,27 @@ bool CacheWithSecondaryAdapter::Release(Handle* handle,
     ObjectPtr v = target_->Value(handle);
     if (v == nullptr && distribute_cache_res_) {
       size_t charge = target_->GetCharge(handle);
-      size_t sec_charge = static_cast<size_t>(charge * (sec_cache_res_ratio_));
-      Status s = secondary_cache_->Inflate(sec_charge);
-      assert(s.ok());
-      s = pri_cache_res_->UpdateCacheReservation(sec_charge, /*increase=*/true);
-      assert(s.ok());
+
+      MutexLock l(&cache_res_mutex_);
+      placeholder_usage_ -= charge;
+      // Check if total placeholder reservation is more than the overall
+      // cache capacity. If it is, then we do nothing as reserved_usage_ must
+      // be already maxed out
+      if ((placeholder_usage_ <= target_->GetCapacity()) &&
+          (placeholder_usage_ < reserved_usage_)) {
+        // Adjust reserved_usage_ in chunks of kReservationChunkSize, so
+        // we don't hit this slow path too often.
+        reserved_usage_ = placeholder_usage_ & ~(kReservationChunkSize - 1);
+        size_t new_sec_reserved =
+            static_cast<size_t>(reserved_usage_ * sec_cache_res_ratio_);
+        size_t sec_charge = sec_reserved_ - new_sec_reserved;
+        Status s = secondary_cache_->Inflate(sec_charge);
+        assert(s.ok());
+        s = pri_cache_res_->UpdateCacheReservation(sec_charge,
+                                                   /*increase=*/true);
+        assert(s.ok());
+        sec_reserved_ -= sec_charge;
+      }
     }
   }
   return target_->Release(handle, erase_if_last_ref);
@@ -436,7 +480,7 @@ void CacheWithSecondaryAdapter::SetCapacity(size_t capacity) {
   size_t old_sec_capacity = 0;
 
   if (distribute_cache_res_) {
-    MutexLock m(&mutex_);
+    MutexLock m(&cache_res_mutex_);
 
     Status s = secondary_cache_->GetCapacity(old_sec_capacity);
     if (!s.ok()) {
@@ -451,9 +495,17 @@ void CacheWithSecondaryAdapter::SetCapacity(size_t capacity) {
       // 3. Decrease the primary cache capacity to the total budget
       s = secondary_cache_->SetCapacity(sec_capacity);
       if (s.ok()) {
+        if (placeholder_usage_ > capacity) {
+          // Adjust reserved_usage_ down
+          reserved_usage_ = capacity & ~(kReservationChunkSize - 1);
+        }
+        size_t new_sec_reserved =
+            static_cast<size_t>(reserved_usage_ * sec_cache_res_ratio_);
         s = pri_cache_res_->UpdateCacheReservation(
-            old_sec_capacity - sec_capacity,
+            (old_sec_capacity - sec_capacity) -
+                (sec_reserved_ - new_sec_reserved),
             /*increase=*/false);
+        sec_reserved_ = new_sec_reserved;
         assert(s.ok());
         target_->SetCapacity(capacity);
       }
@@ -478,6 +530,29 @@ void CacheWithSecondaryAdapter::SetCapacity(size_t capacity) {
   }
 }
 
+Status CacheWithSecondaryAdapter::GetSecondaryCacheCapacity(
+    size_t& size) const {
+  return secondary_cache_->GetCapacity(size);
+}
+
+Status CacheWithSecondaryAdapter::GetSecondaryCachePinnedUsage(
+    size_t& size) const {
+  Status s;
+  if (distribute_cache_res_) {
+    MutexLock m(&cache_res_mutex_);
+    size_t capacity = 0;
+    s = secondary_cache_->GetCapacity(capacity);
+    if (s.ok()) {
+      size = capacity - pri_cache_res_->GetTotalMemoryUsed();
+    } else {
+      size = 0;
+    }
+  } else {
+    size = 0;
+  }
+  return s;
+}
+
 // Update the secondary/primary allocation ratio (remember, the primary
 // capacity is the total memory budget when distribute_cache_res_ is true).
 // When the ratio changes, we may accumulate some error in the calculations
@@ -492,11 +567,11 @@ void CacheWithSecondaryAdapter::SetCapacity(size_t capacity) {
 // in the future.
 Status CacheWithSecondaryAdapter::UpdateCacheReservationRatio(
     double compressed_secondary_ratio) {
-  if (!distribute_cache_res_ || sec_cache_res_ratio_ == 0.0) {
+  if (!distribute_cache_res_) {
     return Status::NotSupported();
   }
 
-  MutexLock m(&mutex_);
+  MutexLock m(&cache_res_mutex_);
   size_t pri_capacity = target_->GetCapacity();
   size_t sec_capacity =
       static_cast<size_t>(pri_capacity * compressed_secondary_ratio);
@@ -506,19 +581,17 @@ Status CacheWithSecondaryAdapter::UpdateCacheReservationRatio(
     return s;
   }
 
-  assert(old_sec_capacity >= pri_cache_res_->GetTotalMemoryUsed());
-  size_t old_sec_reserved =
-      old_sec_capacity - pri_cache_res_->GetTotalMemoryUsed();
   // Calculate the new secondary cache reservation
-  size_t sec_reserved = static_cast<size_t>(
-      old_sec_reserved *
-      (double)(compressed_secondary_ratio / sec_cache_res_ratio_));
+  // reserved_usage_ will never be > the cache capacity, so we don't
+  // have to worry about adjusting it here.
   sec_cache_res_ratio_ = compressed_secondary_ratio;
+  size_t new_sec_reserved =
+      static_cast<size_t>(reserved_usage_ * sec_cache_res_ratio_);
   if (sec_capacity > old_sec_capacity) {
     // We're increasing the ratio, thus ending up with a larger secondary
     // cache and a smaller usable primary cache capacity. Similar to
     // SetCapacity(), we try to avoid a temporary increase in total usage
-    // beyond teh configured capacity -
+    // beyond the configured capacity -
     // 1. A higher secondary cache ratio means it gets a higher share of
     //    cache reservations. So first account for that by deflating the
     //    secondary cache
@@ -526,13 +599,13 @@ Status CacheWithSecondaryAdapter::UpdateCacheReservationRatio(
     //    cache utilization (increase in capacity - increase in share of cache
     //    reservation)
     // 3. Increase secondary cache capacity
-    assert(sec_reserved > old_sec_reserved || sec_reserved == 0);
-    s = secondary_cache_->Deflate(sec_reserved - old_sec_reserved);
+    s = secondary_cache_->Deflate(new_sec_reserved - sec_reserved_);
     assert(s.ok());
     s = pri_cache_res_->UpdateCacheReservation(
-        (sec_capacity - old_sec_capacity) - (sec_reserved - old_sec_reserved),
+        (sec_capacity - old_sec_capacity) - (new_sec_reserved - sec_reserved_),
         /*increase=*/true);
     assert(s.ok());
+    sec_reserved_ = new_sec_reserved;
     s = secondary_cache_->SetCapacity(sec_capacity);
     assert(s.ok());
   } else {
@@ -543,24 +616,19 @@ Status CacheWithSecondaryAdapter::UpdateCacheReservationRatio(
     //    reservations)
     // 3. Inflate the secondary cache to give it back the reduction in its
     //    share of cache reservations
-    assert(old_sec_reserved > sec_reserved || sec_reserved == 0);
     s = secondary_cache_->SetCapacity(sec_capacity);
     if (s.ok()) {
       s = pri_cache_res_->UpdateCacheReservation(
-          (old_sec_capacity - sec_capacity) - (old_sec_reserved - sec_reserved),
+          (old_sec_capacity - sec_capacity) -
+              (sec_reserved_ - new_sec_reserved),
           /*increase=*/false);
       assert(s.ok());
-      s = secondary_cache_->Inflate(old_sec_reserved - sec_reserved);
+      s = secondary_cache_->Inflate(sec_reserved_ - new_sec_reserved);
       assert(s.ok());
+      sec_reserved_ = new_sec_reserved;
     }
   }
 
-#ifndef NDEBUG
-  // As mentioned in the function comments, we may accumulate some erros when
-  // the ratio is changed. We set a flag here which disables some assertions
-  // in the destructor
-  ratio_changed_ = true;
-#endif
   return s;
 }
 
@@ -575,23 +643,54 @@ std::shared_ptr<Cache> NewTieredCache(const TieredCacheOptions& _opts) {
     return nullptr;
   }
 
-  if (_opts.adm_policy >= TieredAdmissionPolicy::kAdmPolicyMax) {
-    return nullptr;
+  TieredCacheOptions opts = _opts;
+  {
+    bool valid_adm_policy = true;
+
+    switch (_opts.adm_policy) {
+      case TieredAdmissionPolicy::kAdmPolicyAuto:
+        // Select an appropriate default policy
+        if (opts.adm_policy == TieredAdmissionPolicy::kAdmPolicyAuto) {
+          if (opts.nvm_sec_cache) {
+            opts.adm_policy = TieredAdmissionPolicy::kAdmPolicyThreeQueue;
+          } else {
+            opts.adm_policy = TieredAdmissionPolicy::kAdmPolicyPlaceholder;
+          }
+        }
+        break;
+      case TieredAdmissionPolicy::kAdmPolicyPlaceholder:
+      case TieredAdmissionPolicy::kAdmPolicyAllowCacheHits:
+        if (opts.nvm_sec_cache) {
+          valid_adm_policy = false;
+        }
+        break;
+      case TieredAdmissionPolicy::kAdmPolicyThreeQueue:
+        if (!opts.nvm_sec_cache) {
+          valid_adm_policy = false;
+        }
+        break;
+      default:
+        valid_adm_policy = false;
+    }
+    if (!valid_adm_policy) {
+      return nullptr;
+    }
   }
 
-  TieredCacheOptions opts = _opts;
   std::shared_ptr<Cache> cache;
   if (opts.cache_type == PrimaryCacheType::kCacheTypeLRU) {
     LRUCacheOptions cache_opts =
         *(static_cast_with_check<LRUCacheOptions, ShardedCacheOptions>(
             opts.cache_opts));
     cache_opts.capacity = opts.total_capacity;
+    cache_opts.secondary_cache = nullptr;
     cache = cache_opts.MakeSharedCache();
   } else if (opts.cache_type == PrimaryCacheType::kCacheTypeHCC) {
     HyperClockCacheOptions cache_opts =
         *(static_cast_with_check<HyperClockCacheOptions, ShardedCacheOptions>(
             opts.cache_opts));
     cache_opts.capacity = opts.total_capacity;
+    cache_opts.secondary_cache = nullptr;
     cache = cache_opts.MakeSharedCache();
   } else {
     return nullptr;
@@ -602,8 +701,7 @@ std::shared_ptr<Cache> NewTieredCache(const TieredCacheOptions& _opts) {
   sec_cache = NewCompressedSecondaryCache(opts.comp_cache_opts);
 
   if (opts.nvm_sec_cache) {
-    if (opts.adm_policy == TieredAdmissionPolicy::kAdmPolicyThreeQueue ||
-        opts.adm_policy == TieredAdmissionPolicy::kAdmPolicyAuto) {
+    if (opts.adm_policy == TieredAdmissionPolicy::kAdmPolicyThreeQueue) {
       sec_cache = std::make_shared<TieredSecondaryCache>(
           sec_cache, opts.nvm_sec_cache,
           TieredAdmissionPolicy::kAdmPolicyThreeQueue);

cache/secondary_cache_adapter.h

@@ -47,6 +47,10 @@ class CacheWithSecondaryAdapter : public CacheWrapper {
 
   void SetCapacity(size_t capacity) override;
 
+  Status GetSecondaryCacheCapacity(size_t& size) const override;
+
+  Status GetSecondaryCachePinnedUsage(size_t& size) const override;
+
   Status UpdateCacheReservationRatio(double ratio);
 
   Status UpdateAdmissionPolicy(TieredAdmissionPolicy adm_policy);
@@ -56,6 +60,8 @@ class CacheWithSecondaryAdapter : public CacheWrapper {
   SecondaryCache* TEST_GetSecondaryCache() { return secondary_cache_.get(); }
 
  private:
+  static constexpr size_t kReservationChunkSize = 1 << 20;
+
   bool EvictionHandler(const Slice& key, Handle* handle, bool was_hit);
 
   void StartAsyncLookupOnMySecondary(AsyncLookupHandle& async_handle);
@@ -81,10 +87,17 @@ class CacheWithSecondaryAdapter : public CacheWrapper {
   // Fraction of a cache memory reservation to be assigned to the secondary
   // cache
   double sec_cache_res_ratio_;
-  port::Mutex mutex_;
-#ifndef NDEBUG
-  bool ratio_changed_ = false;
-#endif
+  // Mutex for use when managing cache memory reservations. Should not be used
+  // for other purposes, as it may risk causing deadlocks.
+  mutable port::Mutex cache_res_mutex_;
+  // Total memory reserved by placeholder entriesin the cache
+  size_t placeholder_usage_;
+  // Total placeholoder memory charged to both the primary and secondary
+  // caches. Will be <= placeholder_usage_.
+  size_t reserved_usage_;
+  // Amount of memory reserved in the secondary cache. This should be
+  // reserved_usage_ * sec_cache_res_ratio_ in steady state.
+  size_t sec_reserved_;
 };
 
 }  // namespace ROCKSDB_NAMESPACE

cache/sharded_cache.cc

@@ -83,6 +83,16 @@ size_t ShardedCacheBase::GetCapacity() const {
   return capacity_;
 }
 
+Status ShardedCacheBase::GetSecondaryCacheCapacity(size_t& size) const {
+  size = 0;
+  return Status::OK();
+}
+
+Status ShardedCacheBase::GetSecondaryCachePinnedUsage(size_t& size) const {
+  size = 0;
+  return Status::OK();
+}
+
 bool ShardedCacheBase::HasStrictCapacityLimit() const {
   MutexLock l(&config_mutex_);
   return strict_capacity_limit_;

cache/sharded_cache.h

@@ -99,6 +99,8 @@ class ShardedCacheBase : public Cache {
 
   bool HasStrictCapacityLimit() const override;
   size_t GetCapacity() const override;
+  Status GetSecondaryCacheCapacity(size_t& size) const override;
+  Status GetSecondaryCachePinnedUsage(size_t& size) const override;
 
   using Cache::GetUsage;
   size_t GetUsage(Handle* handle) const override;

cache/tiered_secondary_cache.cc

@@ -109,10 +109,8 @@ void TieredSecondaryCache::WaitAll(
   }
   nvm_sec_cache_->WaitAll(nvm_handles);
   for (auto handle : my_handles) {
-    assert(handle->IsReady());
-    auto nvm_handle = handle->inner_handle();
-    handle->SetSize(nvm_handle->Size());
-    handle->SetValue(nvm_handle->Value());
+    assert(handle->inner_handle()->IsReady());
+    handle->Complete();
   }
 }
 

cache/tiered_secondary_cache.h

@@ -79,7 +79,10 @@ class TieredSecondaryCache : public SecondaryCacheWrapper {
     ~ResultHandle() override {}
 
     bool IsReady() override {
-      return !inner_handle_ || inner_handle_->IsReady();
+      if (inner_handle_ && inner_handle_->IsReady()) {
+        Complete();
+      }
+      return ready_;
     }
 
     void Wait() override {
@@ -92,10 +95,10 @@ class TieredSecondaryCache : public SecondaryCacheWrapper {
     Cache::ObjectPtr Value() override { return value_; }
 
     void Complete() {
-      assert(IsReady());
       size_ = inner_handle_->Size();
       value_ = inner_handle_->Value();
       inner_handle_.reset();
+      ready_ = true;
     }
 
     void SetInnerHandle(std::unique_ptr<SecondaryCacheResultHandle>&& handle) {
@@ -115,6 +118,7 @@ class TieredSecondaryCache : public SecondaryCacheWrapper {
     CreateContext ctx_;
     size_t size_;
     Cache::ObjectPtr value_;
+    bool ready_ = false;
   };
 
   static void NoopDelete(Cache::ObjectPtr /*obj*/,

cache/tiered_secondary_cache_test.cc

@@ -3,6 +3,8 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 //
+#include "cache/compressed_secondary_cache.h"
+#include "cache/secondary_cache_adapter.h"
 #include "db/db_test_util.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/secondary_cache.h"
@@ -13,10 +15,11 @@ namespace ROCKSDB_NAMESPACE {
 
 class TestSecondaryCache : public SecondaryCache {
  public:
-  explicit TestSecondaryCache(size_t capacity)
+  explicit TestSecondaryCache(size_t capacity, bool ready_before_wait)
       : cache_(NewLRUCache(capacity, 0, false, 0.5 /* high_pri_pool_ratio */,
                            nullptr, kDefaultToAdaptiveMutex,
                            kDontChargeCacheMetadata)),
+        ready_before_wait_(ready_before_wait),
         num_insert_saved_(0),
         num_hits_(0),
         num_misses_(0) {}
@@ -86,7 +89,8 @@ class TestSecondaryCache : public SecondaryCache {
                             /*alloc*/ nullptr, &value, &charge);
       if (s.ok()) {
         secondary_handle.reset(new TestSecondaryCacheResultHandle(
-            cache_.get(), handle, value, charge, /*ready=*/wait));
+            cache_.get(), handle, value, charge,
+            /*ready=*/wait || ready_before_wait_));
         kept_in_sec_cache = true;
       } else {
         cache_.Release(handle);
@@ -166,6 +170,7 @@ class TestSecondaryCache : public SecondaryCache {
       BasicTypedSharedCacheInterface<char[], CacheEntryRole::kMisc>;
   using TypedHandle = SharedCache::TypedHandle;
   SharedCache cache_;
+  bool ready_before_wait_;
   uint32_t num_insert_saved_;
   uint32_t num_hits_;
   uint32_t num_misses_;
@@ -177,9 +182,10 @@ class DBTieredSecondaryCacheTest : public DBTestBase {
   DBTieredSecondaryCacheTest()
       : DBTestBase("db_tiered_secondary_cache_test", /*env_do_fsync=*/true) {}
 
-  std::shared_ptr<Cache> NewCache(size_t pri_capacity,
-                                  size_t compressed_capacity,
-                                  size_t nvm_capacity) {
+  std::shared_ptr<Cache> NewCache(
+      size_t pri_capacity, size_t compressed_capacity, size_t nvm_capacity,
+      TieredAdmissionPolicy adm_policy = TieredAdmissionPolicy::kAdmPolicyAuto,
+      bool ready_before_wait = false) {
     LRUCacheOptions lru_opts;
     TieredCacheOptions opts;
     lru_opts.capacity = 0;
@@ -187,33 +193,41 @@ class DBTieredSecondaryCacheTest : public DBTestBase {
     lru_opts.high_pri_pool_ratio = 0;
     opts.cache_opts = &lru_opts;
     opts.cache_type = PrimaryCacheType::kCacheTypeLRU;
-    opts.adm_policy = TieredAdmissionPolicy::kAdmPolicyThreeQueue;
     opts.comp_cache_opts.capacity = 0;
     opts.comp_cache_opts.num_shard_bits = 0;
     opts.total_capacity = pri_capacity + compressed_capacity;
-    opts.compressed_secondary_ratio =
+    opts.compressed_secondary_ratio = compressed_secondary_ratio_ =
         (double)compressed_capacity / opts.total_capacity;
-    nvm_sec_cache_.reset(new TestSecondaryCache(nvm_capacity));
-    opts.nvm_sec_cache = nvm_sec_cache_;
+    if (nvm_capacity > 0) {
+      nvm_sec_cache_.reset(
+          new TestSecondaryCache(nvm_capacity, ready_before_wait));
+      opts.nvm_sec_cache = nvm_sec_cache_;
+    }
+    opts.adm_policy = adm_policy;
     cache_ = NewTieredCache(opts);
     assert(cache_ != nullptr);
 
-#if 0
-    CacheWithSecondaryAdapter* adapter_cache_ =
-      static_cast<CacheWithSecondaryAdapter*>(cache_.get());
-    TieredSecondaryCache* tiered_cache_ =
-      static_cast<TieredSecondaryCache*>(
-          adapter_cache_->TEST_GetSecondaryCache());
-#endif
-
     return cache_;
   }
 
+  void ClearPrimaryCache() {
+    ASSERT_EQ(UpdateTieredCache(cache_, -1, 1.0), Status::OK());
+    ASSERT_EQ(UpdateTieredCache(cache_, -1, compressed_secondary_ratio_),
+              Status::OK());
+  }
+
   TestSecondaryCache* nvm_sec_cache() { return nvm_sec_cache_.get(); }
 
+  CompressedSecondaryCache* compressed_secondary_cache() {
+    return static_cast<CompressedSecondaryCache*>(
+        static_cast<CacheWithSecondaryAdapter*>(cache_.get())
+            ->TEST_GetSecondaryCache());
+  }
+
  private:
   std::shared_ptr<Cache> cache_;
   std::shared_ptr<TestSecondaryCache> nvm_sec_cache_;
+  double compressed_secondary_ratio_;
 };
 
 // In this test, the block size is set to 4096. Each value is 1007 bytes, so
@@ -578,6 +592,116 @@ TEST_F(DBTieredSecondaryCacheTest, WaitAllTest) {
   Destroy(options);
 }
 
+TEST_F(DBTieredSecondaryCacheTest, ReadyBeforeWaitAllTest) {
+  if (!LZ4_Supported()) {
+    ROCKSDB_GTEST_SKIP("This test requires LZ4 support.");
+    return;
+  }
+
+  BlockBasedTableOptions table_options;
+  table_options.block_cache = NewCache(250 * 1024, 20 * 1024, 256 * 1024,
+                                       TieredAdmissionPolicy::kAdmPolicyAuto,
+                                       /*ready_before_wait=*/true);
+  table_options.block_size = 4 * 1024;
+  table_options.cache_index_and_filter_blocks = false;
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  options.statistics = CreateDBStatistics();
+
+  options.paranoid_file_checks = false;
+  DestroyAndReopen(options);
+  Random rnd(301);
+  const int N = 256;
+  for (int i = 0; i < N; i++) {
+    std::string p_v;
+    test::CompressibleString(&rnd, 0.5, 1007, &p_v);
+    ASSERT_OK(Put(Key(i), p_v));
+  }
+
+  ASSERT_OK(Flush());
+
+  std::vector<std::string> keys;
+  std::vector<std::string> values;
+
+  keys.push_back(Key(0));
+  keys.push_back(Key(4));
+  keys.push_back(Key(8));
+  values = MultiGet(keys, /*snapshot=*/nullptr, /*async=*/true);
+  ASSERT_EQ(values.size(), keys.size());
+  for (auto value : values) {
+    ASSERT_EQ(1007, value.size());
+  }
+  ASSERT_EQ(nvm_sec_cache()->num_insert_saved(), 3u);
+  ASSERT_EQ(nvm_sec_cache()->num_misses(), 3u);
+  ASSERT_EQ(nvm_sec_cache()->num_hits(), 0u);
+  ASSERT_EQ(options.statistics->getTickerCount(BLOCK_CACHE_MISS), 3u);
+
+  keys.clear();
+  values.clear();
+  keys.push_back(Key(12));
+  keys.push_back(Key(16));
+  keys.push_back(Key(20));
+  values = MultiGet(keys, /*snapshot=*/nullptr, /*async=*/true);
+  ASSERT_EQ(values.size(), keys.size());
+  for (auto value : values) {
+    ASSERT_EQ(1007, value.size());
+  }
+  ASSERT_EQ(nvm_sec_cache()->num_insert_saved(), 6u);
+  ASSERT_EQ(nvm_sec_cache()->num_misses(), 6u);
+  ASSERT_EQ(nvm_sec_cache()->num_hits(), 0u);
+  ASSERT_EQ(options.statistics->getTickerCount(BLOCK_CACHE_MISS), 6u);
+
+  keys.clear();
+  values.clear();
+  keys.push_back(Key(0));
+  keys.push_back(Key(4));
+  keys.push_back(Key(8));
+  values = MultiGet(keys, /*snapshot=*/nullptr, /*async=*/true);
+  ASSERT_EQ(values.size(), keys.size());
+  for (auto value : values) {
+    ASSERT_EQ(1007, value.size());
+  }
+  ASSERT_EQ(nvm_sec_cache()->num_insert_saved(), 6u);
+  ASSERT_EQ(nvm_sec_cache()->num_misses(), 6u);
+  ASSERT_EQ(nvm_sec_cache()->num_hits(), 3u);
+  ASSERT_EQ(options.statistics->getTickerCount(BLOCK_CACHE_MISS), 6u);
+
+  ClearPrimaryCache();
+
+  keys.clear();
+  values.clear();
+  keys.push_back(Key(0));
+  keys.push_back(Key(32));
+  keys.push_back(Key(36));
+  values = MultiGet(keys, /*snapshot=*/nullptr, /*async=*/true);
+  ASSERT_EQ(values.size(), keys.size());
+  for (auto value : values) {
+    ASSERT_EQ(1007, value.size());
+  }
+  ASSERT_EQ(nvm_sec_cache()->num_insert_saved(), 8u);
+  ASSERT_EQ(nvm_sec_cache()->num_misses(), 8u);
+  ASSERT_EQ(nvm_sec_cache()->num_hits(), 4u);
+  ASSERT_EQ(options.statistics->getTickerCount(BLOCK_CACHE_MISS), 8u);
+
+  keys.clear();
+  values.clear();
+  keys.push_back(Key(0));
+  keys.push_back(Key(32));
+  keys.push_back(Key(36));
+  values = MultiGet(keys, /*snapshot=*/nullptr, /*async=*/true);
+  ASSERT_EQ(values.size(), keys.size());
+  for (auto value : values) {
+    ASSERT_EQ(1007, value.size());
+  }
+  ASSERT_EQ(nvm_sec_cache()->num_insert_saved(), 8u);
+  ASSERT_EQ(nvm_sec_cache()->num_misses(), 8u);
+  ASSERT_EQ(nvm_sec_cache()->num_hits(), 4u);
+  ASSERT_EQ(options.statistics->getTickerCount(BLOCK_CACHE_MISS), 8u);
+
+  Destroy(options);
+}
+
 // This test is for iteration. It iterates through a set of keys in two
 // passes. First pass loads the compressed blocks into the nvm tier, and
 // the second pass should hit all of those blocks.
@@ -636,6 +760,68 @@ TEST_F(DBTieredSecondaryCacheTest, IterateTest) {
   Destroy(options);
 }
 
+class DBTieredAdmPolicyTest
+    : public DBTieredSecondaryCacheTest,
+      public testing::WithParamInterface<TieredAdmissionPolicy> {};
+
+TEST_P(DBTieredAdmPolicyTest, CompressedOnlyTest) {
+  if (!LZ4_Supported()) {
+    ROCKSDB_GTEST_SKIP("This test requires LZ4 support.");
+    return;
+  }
+
+  BlockBasedTableOptions table_options;
+  // We want a block cache of size 10KB, and a compressed secondary cache of
+  // size 10KB. However, we specify a block cache size of 256KB here in order
+  // to take into account the cache reservation in the block cache on
+  // behalf of the compressed cache. The unit of cache reservation is 256KB.
+  // The effective block cache capacity will be calculated as 256 + 10 = 266KB,
+  // and 256KB will be reserved for the compressed cache, leaving 10KB for
+  // the primary block cache. We only have to worry about this here because
+  // the cache size is so small.
+  table_options.block_cache = NewCache(256 * 1024, 10 * 1024, 0, GetParam());
+  table_options.block_size = 4 * 1024;
+  table_options.cache_index_and_filter_blocks = false;
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+
+  size_t comp_cache_usage = compressed_secondary_cache()->TEST_GetUsage();
+  // Disable paranoid_file_checks so that flush will not read back the newly
+  // written file
+  options.paranoid_file_checks = false;
+  DestroyAndReopen(options);
+  Random rnd(301);
+  const int N = 256;
+  for (int i = 0; i < N; i++) {
+    std::string p_v;
+    test::CompressibleString(&rnd, 0.5, 1007, &p_v);
+    ASSERT_OK(Put(Key(i), p_v));
+  }
+
+  ASSERT_OK(Flush());
+
+  // The first 2 Gets, for keys 0 and 5, will load the corresponding data
+  // blocks as they will be cache misses. Since this is a 2-tier cache (
+  // primary and compressed), no warm-up should happen with the compressed
+  // blocks.
+  std::string v = Get(Key(0));
+  ASSERT_EQ(1007, v.size());
+
+  v = Get(Key(5));
+  ASSERT_EQ(1007, v.size());
+
+  ASSERT_EQ(compressed_secondary_cache()->TEST_GetUsage(), comp_cache_usage);
+
+  Destroy(options);
+}
+
+INSTANTIATE_TEST_CASE_P(
+    DBTieredAdmPolicyTest, DBTieredAdmPolicyTest,
+    ::testing::Values(TieredAdmissionPolicy::kAdmPolicyAuto,
+                      TieredAdmissionPolicy::kAdmPolicyPlaceholder,
+                      TieredAdmissionPolicy::kAdmPolicyAllowCacheHits));
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/blob/db_blob_basic_test.cc

@@ -168,6 +168,7 @@ TEST_F(DBBlobBasicTest, IterateBlobsFromCache) {
       ASSERT_EQ(iter->value().ToString(), blobs[i]);
       ++i;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(i, num_blobs);
     ASSERT_EQ(options.statistics->getAndResetTickerCount(BLOB_DB_CACHE_ADD), 0);
   }
@@ -203,6 +204,7 @@ TEST_F(DBBlobBasicTest, IterateBlobsFromCache) {
       ASSERT_EQ(iter->value().ToString(), blobs[i]);
       ++i;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(i, num_blobs);
     ASSERT_EQ(options.statistics->getAndResetTickerCount(BLOB_DB_CACHE_ADD),
               num_blobs);
@@ -224,6 +226,7 @@ TEST_F(DBBlobBasicTest, IterateBlobsFromCache) {
       ASSERT_EQ(iter->value().ToString(), blobs[i]);
       ++i;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(i, num_blobs);
     ASSERT_EQ(options.statistics->getAndResetTickerCount(BLOB_DB_CACHE_ADD), 0);
   }
@@ -2123,6 +2126,7 @@ TEST_F(DBBlobWithTimestampTest, IterateBlobs) {
                        /*key_is_internal*/ false);
       iter->Prev();
     }
+    ASSERT_OK(iter->status());
   }
 
   // Backward iteration, then reverse to forward.
@@ -2169,6 +2173,7 @@ TEST_F(DBBlobWithTimestampTest, IterateBlobs) {
         iter->Next();
       }
     }
+    ASSERT_OK(iter->status());
   }
 
   // Backward iterating multiple versions of the same key, get in this order:
@@ -2187,6 +2192,7 @@ TEST_F(DBBlobWithTimestampTest, IterateBlobs) {
         iter->Prev();
       }
     }
+    ASSERT_OK(iter->status());
   }
 
   int upper_bound_idx = num_blobs - 2;
@@ -2209,6 +2215,7 @@ TEST_F(DBBlobWithTimestampTest, IterateBlobs) {
         iter->Next();
       }
     }
+    ASSERT_OK(iter->status());
   }
 
   // Backward iteration with upper and lower bound.
@@ -2224,6 +2231,7 @@ TEST_F(DBBlobWithTimestampTest, IterateBlobs) {
         iter->Prev();
       }
     }
+    ASSERT_OK(iter->status());
   }
 }
 

db/builder.cc

@@ -294,12 +294,12 @@ Status BuildTable(
     if (!s.ok() || empty) {
       builder->Abandon();
     } else {
-      std::string seqno_time_mapping_str;
+      std::string seqno_to_time_mapping_str;
       seqno_to_time_mapping.Encode(
-          seqno_time_mapping_str, meta->fd.smallest_seqno,
+          seqno_to_time_mapping_str, meta->fd.smallest_seqno,
           meta->fd.largest_seqno, meta->file_creation_time);
       builder->SetSeqnoTimeTableProperties(
-          seqno_time_mapping_str,
+          seqno_to_time_mapping_str,
           ioptions.compaction_style == CompactionStyle::kCompactionStyleFIFO
               ? meta->file_creation_time
               : meta->oldest_ancester_time);

db/c.cc

@@ -45,6 +45,7 @@
 #include "rocksdb/utilities/transaction_db.h"
 #include "rocksdb/utilities/write_batch_with_index.h"
 #include "rocksdb/write_batch.h"
+#include "rocksdb/write_buffer_manager.h"
 #include "utilities/merge_operators.h"
 
 using ROCKSDB_NAMESPACE::BackupEngine;
@@ -125,6 +126,7 @@ using ROCKSDB_NAMESPACE::WALRecoveryMode;
 using ROCKSDB_NAMESPACE::WritableFile;
 using ROCKSDB_NAMESPACE::WriteBatch;
 using ROCKSDB_NAMESPACE::WriteBatchWithIndex;
+using ROCKSDB_NAMESPACE::WriteBufferManager;
 using ROCKSDB_NAMESPACE::WriteOptions;
 
 using std::unordered_set;
@@ -219,6 +221,9 @@ struct rocksdb_memory_allocator_t {
 struct rocksdb_cache_t {
   std::shared_ptr<Cache> rep;
 };
+struct rocksdb_write_buffer_manager_t {
+  std::shared_ptr<WriteBufferManager> rep;
+};
 struct rocksdb_livefiles_t {
   std::vector<LiveFileMetaData> rep;
 };
@@ -2904,6 +2909,16 @@ void rocksdb_options_set_db_paths(rocksdb_options_t* opt,
   opt->rep.db_paths = db_paths;
 }
 
+void rocksdb_options_set_cf_paths(rocksdb_options_t* opt,
+                                  const rocksdb_dbpath_t** dbpath_values,
+                                  size_t num_paths) {
+  std::vector<DbPath> cf_paths(num_paths);
+  for (size_t i = 0; i < num_paths; ++i) {
+    cf_paths[i] = dbpath_values[i]->rep;
+  }
+  opt->rep.cf_paths = cf_paths;
+}
+
 void rocksdb_options_set_env(rocksdb_options_t* opt, rocksdb_env_t* env) {
   opt->rep.env = (env ? env->rep : nullptr);
 }
@@ -2935,6 +2950,11 @@ void rocksdb_options_set_write_buffer_size(rocksdb_options_t* opt, size_t s) {
   opt->rep.write_buffer_size = s;
 }
 
+void rocksdb_options_set_write_buffer_manager(
+    rocksdb_options_t* opt, rocksdb_write_buffer_manager_t* wbm) {
+  opt->rep.write_buffer_manager = wbm->rep;
+}
+
 size_t rocksdb_options_get_write_buffer_size(rocksdb_options_t* opt) {
   return opt->rep.write_buffer_size;
 }
@@ -3029,6 +3049,16 @@ void rocksdb_options_set_max_bytes_for_level_multiplier_additional(
   }
 }
 
+void rocksdb_options_set_periodic_compaction_seconds(rocksdb_options_t* opt,
+                                                     uint64_t seconds) {
+  opt->rep.periodic_compaction_seconds = seconds;
+}
+
+uint64_t rocksdb_options_get_periodic_compaction_seconds(
+    rocksdb_options_t* opt) {
+  return opt->rep.periodic_compaction_seconds;
+}
+
 void rocksdb_options_enable_statistics(rocksdb_options_t* opt) {
   opt->rep.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
 }
@@ -3957,6 +3987,16 @@ rocksdb_ratelimiter_t* rocksdb_ratelimiter_create(int64_t rate_bytes_per_sec,
   return rate_limiter;
 }
 
+rocksdb_ratelimiter_t* rocksdb_ratelimiter_create_auto_tuned(
+    int64_t rate_bytes_per_sec, int64_t refill_period_us, int32_t fairness) {
+  rocksdb_ratelimiter_t* rate_limiter = new rocksdb_ratelimiter_t;
+  rate_limiter->rep.reset(NewGenericRateLimiter(rate_bytes_per_sec,
+                                                refill_period_us, fairness,
+                                                RateLimiter::Mode::kWritesOnly,
+                                                true));  // auto_tuned
+  return rate_limiter;
+}
+
 void rocksdb_ratelimiter_destroy(rocksdb_ratelimiter_t* limiter) {
   delete limiter;
 }
@@ -4866,6 +4906,60 @@ size_t rocksdb_cache_get_occupancy_count(const rocksdb_cache_t* cache) {
   return cache->rep->GetOccupancyCount();
 }
 
+rocksdb_write_buffer_manager_t* rocksdb_write_buffer_manager_create(
+    size_t buffer_size, bool allow_stall) {
+  rocksdb_write_buffer_manager_t* wbm = new rocksdb_write_buffer_manager_t;
+  wbm->rep.reset(new WriteBufferManager(buffer_size, {}, allow_stall));
+  return wbm;
+}
+
+rocksdb_write_buffer_manager_t* rocksdb_write_buffer_manager_create_with_cache(
+    size_t buffer_size, const rocksdb_cache_t* cache, bool allow_stall) {
+  rocksdb_write_buffer_manager_t* wbm = new rocksdb_write_buffer_manager_t;
+  wbm->rep.reset(new WriteBufferManager(buffer_size, cache->rep, allow_stall));
+  return wbm;
+}
+
+void rocksdb_write_buffer_manager_destroy(rocksdb_write_buffer_manager_t* wbm) {
+  delete wbm;
+}
+
+bool rocksdb_write_buffer_manager_enabled(rocksdb_write_buffer_manager_t* wbm) {
+  return wbm->rep->enabled();
+}
+
+bool rocksdb_write_buffer_manager_cost_to_cache(
+    rocksdb_write_buffer_manager_t* wbm) {
+  return wbm->rep->cost_to_cache();
+}
+
+size_t rocksdb_write_buffer_manager_memory_usage(
+    rocksdb_write_buffer_manager_t* wbm) {
+  return wbm->rep->memory_usage();
+}
+
+size_t rocksdb_write_buffer_manager_mutable_memtable_memory_usage(
+    rocksdb_write_buffer_manager_t* wbm) {
+  return wbm->rep->mutable_memtable_memory_usage();
+}
+
+size_t rocksdb_write_buffer_manager_dummy_entries_in_cache_usage(
+    rocksdb_write_buffer_manager_t* wbm) {
+  return wbm->rep->dummy_entries_in_cache_usage();
+}
+size_t rocksdb_write_buffer_manager_buffer_size(
+    rocksdb_write_buffer_manager_t* wbm) {
+  return wbm->rep->buffer_size();
+}
+void rocksdb_write_buffer_manager_set_buffer_size(
+    rocksdb_write_buffer_manager_t* wbm, size_t new_size) {
+  wbm->rep->SetBufferSize(new_size);
+}
+ROCKSDB_LIBRARY_API void rocksdb_write_buffer_manager_set_allow_stall(
+    rocksdb_write_buffer_manager_t* wbm, bool new_allow_stall) {
+  wbm->rep->SetAllowStall(new_allow_stall);
+}
+
 rocksdb_dbpath_t* rocksdb_dbpath_create(const char* path,
                                         uint64_t target_size) {
   rocksdb_dbpath_t* result = new rocksdb_dbpath_t;
@@ -5464,6 +5558,11 @@ char* rocksdb_sst_file_metadata_get_relative_filename(
   return strdup(file_meta->rep->relative_filename.c_str());
 }
 
+char* rocksdb_sst_file_metadata_get_directory(
+    rocksdb_sst_file_metadata_t* file_meta) {
+  return strdup(file_meta->rep->directory.c_str());
+}
+
 uint64_t rocksdb_sst_file_metadata_get_size(
     rocksdb_sst_file_metadata_t* file_meta) {
   return file_meta->rep->size;

db/c_test.c

@@ -713,6 +713,11 @@ int main(int argc, char** argv) {
   rocksdb_options_set_ratelimiter(options, rate_limiter);
   rocksdb_ratelimiter_destroy(rate_limiter);
 
+  rate_limiter =
+      rocksdb_ratelimiter_create_auto_tuned(1000 * 1024 * 1024, 100 * 1000, 10);
+  rocksdb_options_set_ratelimiter(options, rate_limiter);
+  rocksdb_ratelimiter_destroy(rate_limiter);
+
   roptions = rocksdb_readoptions_create();
   rocksdb_readoptions_set_verify_checksums(roptions, 1);
   rocksdb_readoptions_set_fill_cache(roptions, 1);
@@ -1474,10 +1479,20 @@ int main(int argc, char** argv) {
     CheckCondition(cflen == 2);
     rocksdb_list_column_families_destroy(column_fams, cflen);
 
-    rocksdb_options_t* cf_options = rocksdb_options_create();
+    rocksdb_options_t* cf_options_1 = rocksdb_options_create();
+    rocksdb_options_t* cf_options_2 = rocksdb_options_create();
+
+    // use dbpathname2 as the cf_path for "cf1"
+    rocksdb_dbpath_t* dbpath2;
+    char dbpathname2[200];
+    snprintf(dbpathname2, sizeof(dbpathname2), "%s/rocksdb_c_test-%d-dbpath2",
+             GetTempDir(), ((int)geteuid()));
+    dbpath2 = rocksdb_dbpath_create(dbpathname2, 1024 * 1024);
+    const rocksdb_dbpath_t* cf_paths[1] = {dbpath2};
+    rocksdb_options_set_cf_paths(cf_options_2, cf_paths, 1);
 
     const char* cf_names[2] = {"default", "cf1"};
-    const rocksdb_options_t* cf_opts[2] = {cf_options, cf_options};
+    const rocksdb_options_t* cf_opts[2] = {cf_options_1, cf_options_2};
     rocksdb_column_family_handle_t* handles[2];
 
     LoadAndCheckLatestOptions(dbname, env, false, cache, NULL, 2, cf_names,
@@ -1505,6 +1520,37 @@ int main(int argc, char** argv) {
     rocksdb_flushoptions_t* flush_options = rocksdb_flushoptions_create();
     rocksdb_flushoptions_set_wait(flush_options, 1);
     rocksdb_flush_cf(db, flush_options, handles[1], &err);
+
+    // make sure all files in "cf1" are under the specified cf path
+    {
+      rocksdb_column_family_metadata_t* cf_meta =
+          rocksdb_get_column_family_metadata_cf(db, handles[1]);
+      size_t cf_file_count = rocksdb_column_family_metadata_get_size(cf_meta);
+      assert(cf_file_count > 0);
+      size_t level_count =
+          rocksdb_column_family_metadata_get_level_count(cf_meta);
+      assert(level_count > 0);
+      for (size_t l = 0; l < level_count; ++l) {
+        rocksdb_level_metadata_t* level_meta =
+            rocksdb_column_family_metadata_get_level_metadata(cf_meta, l);
+        assert(level_meta);
+
+        size_t file_count = rocksdb_level_metadata_get_file_count(level_meta);
+        for (size_t f = 0; f < file_count; ++f) {
+          rocksdb_sst_file_metadata_t* file_meta =
+              rocksdb_level_metadata_get_sst_file_metadata(level_meta, f);
+          assert(file_meta);
+          char* file_path = rocksdb_sst_file_metadata_get_directory(file_meta);
+          assert(strcmp(file_path, dbpathname2) == 0);
+          Free(&file_path);
+          rocksdb_sst_file_metadata_destroy(file_meta);
+        }
+        rocksdb_level_metadata_destroy(level_meta);
+      }
+
+      rocksdb_column_family_metadata_destroy(cf_meta);
+    }
+
     CheckNoError(err) rocksdb_flushoptions_destroy(flush_options);
 
     CheckGetCF(db, roptions, handles[1], "foo", "hello");
@@ -1668,7 +1714,9 @@ int main(int argc, char** argv) {
     }
     rocksdb_destroy_db(options, dbname, &err);
     rocksdb_options_destroy(db_options);
-    rocksdb_options_destroy(cf_options);
+    rocksdb_options_destroy(cf_options_1);
+    rocksdb_options_destroy(cf_options_2);
+    rocksdb_dbpath_destroy(dbpath2);
   }
 
   StartPhase("prefix");
@@ -1873,6 +1921,10 @@ int main(int argc, char** argv) {
     CheckCondition(2.0 ==
                    rocksdb_options_get_max_bytes_for_level_multiplier(o));
 
+    rocksdb_options_set_periodic_compaction_seconds(o, 100000);
+    CheckCondition(100000 ==
+                   rocksdb_options_get_periodic_compaction_seconds(o));
+
     rocksdb_options_set_skip_stats_update_on_db_open(o, 1);
     CheckCondition(1 == rocksdb_options_get_skip_stats_update_on_db_open(o));
 
@@ -2303,6 +2355,12 @@ int main(int argc, char** argv) {
     CheckCondition(2.0 ==
                    rocksdb_options_get_max_bytes_for_level_multiplier(o));
 
+    rocksdb_options_set_periodic_compaction_seconds(copy, 8000);
+    CheckCondition(8000 ==
+                   rocksdb_options_get_periodic_compaction_seconds(copy));
+    CheckCondition(100000 ==
+                   rocksdb_options_get_periodic_compaction_seconds(o));
+
     rocksdb_options_set_skip_stats_update_on_db_open(copy, 0);
     CheckCondition(0 == rocksdb_options_get_skip_stats_update_on_db_open(copy));
     CheckCondition(1 == rocksdb_options_get_skip_stats_update_on_db_open(o));
@@ -3723,6 +3781,30 @@ int main(int argc, char** argv) {
     rocksdb_wait_for_compact_options_destroy(wco);
   }
 
+  StartPhase("write_buffer_manager");
+  {
+    rocksdb_cache_t* lru;
+    lru = rocksdb_cache_create_lru(100);
+
+    rocksdb_write_buffer_manager_t* write_buffer_manager;
+    write_buffer_manager =
+        rocksdb_write_buffer_manager_create_with_cache(200, lru, false);
+
+    CheckCondition(true ==
+                   rocksdb_write_buffer_manager_enabled(write_buffer_manager));
+    CheckCondition(true == rocksdb_write_buffer_manager_cost_to_cache(
+                               write_buffer_manager));
+    CheckCondition(
+        200 == rocksdb_write_buffer_manager_buffer_size(write_buffer_manager));
+
+    rocksdb_write_buffer_manager_set_buffer_size(write_buffer_manager, 300);
+    CheckCondition(
+        300 == rocksdb_write_buffer_manager_buffer_size(write_buffer_manager));
+
+    rocksdb_write_buffer_manager_destroy(write_buffer_manager);
+    rocksdb_cache_destroy(lru);
+  }
+
   StartPhase("cancel_all_background_work");
   rocksdb_cancel_all_background_work(db, 1);
 

db/column_family.cc

@@ -1748,4 +1748,20 @@ const Comparator* GetColumnFamilyUserComparator(
   return nullptr;
 }
 
+const ImmutableOptions& GetImmutableOptions(ColumnFamilyHandle* column_family) {
+  assert(column_family);
+
+  ColumnFamilyHandleImpl* const handle =
+      static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
+  assert(handle);
+
+  const ColumnFamilyData* const cfd = handle->cfd();
+  assert(cfd);
+
+  const ImmutableOptions* ioptions = cfd->ioptions();
+  assert(ioptions);
+
+  return *ioptions;
+}
+
 }  // namespace ROCKSDB_NAMESPACE

db/column_family.h

@@ -877,4 +877,7 @@ extern uint32_t GetColumnFamilyID(ColumnFamilyHandle* column_family);
 extern const Comparator* GetColumnFamilyUserComparator(
     ColumnFamilyHandle* column_family);
 
+extern const ImmutableOptions& GetImmutableOptions(
+    ColumnFamilyHandle* column_family);
+
 }  // namespace ROCKSDB_NAMESPACE

db/column_family_test.cc

@@ -8,6 +8,7 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include <algorithm>
+#include <atomic>
 #include <string>
 #include <thread>
 #include <vector>
@@ -27,6 +28,7 @@
 #include "test_util/testharness.h"
 #include "test_util/testutil.h"
 #include "util/coding.h"
+#include "util/defer.h"
 #include "util/string_util.h"
 #include "utilities/fault_injection_env.h"
 #include "utilities/merge_operators.h"
@@ -2169,13 +2171,57 @@ TEST_P(ColumnFamilyTest, FlushStaleColumnFamilies) {
   Close();
 }
 
+namespace {
+struct CountOptionsFilesFs : public FileSystemWrapper {
+  explicit CountOptionsFilesFs(const std::shared_ptr<FileSystem>& t)
+      : FileSystemWrapper(t) {}
+  const char* Name() const override { return "CountOptionsFilesFs"; }
+
+  IOStatus NewWritableFile(const std::string& f, const FileOptions& file_opts,
+                           std::unique_ptr<FSWritableFile>* r,
+                           IODebugContext* dbg) override {
+    if (f.find("OPTIONS-") != std::string::npos) {
+      options_files_created.fetch_add(1, std::memory_order_relaxed);
+    }
+    return FileSystemWrapper::NewWritableFile(f, file_opts, r, dbg);
+  }
+
+  std::atomic<int> options_files_created{};
+};
+}  // namespace
+
 TEST_P(ColumnFamilyTest, CreateMissingColumnFamilies) {
-  Status s = TryOpen({"one", "two"});
-  ASSERT_TRUE(!s.ok());
-  db_options_.create_missing_column_families = true;
-  s = TryOpen({"default", "one", "two"});
-  ASSERT_TRUE(s.ok());
+  // Can't accidentally add CFs to an existing DB
+  Open();
   Close();
+  ASSERT_FALSE(db_options_.create_missing_column_families);
+  ASSERT_NOK(TryOpen({"one", "two"}));
+
+  // Nor accidentally create in a new DB
+  Destroy();
+  db_options_.create_if_missing = true;
+  ASSERT_NOK(TryOpen({"one", "two"}));
+
+  // Only with the option (new DB case)
+  db_options_.create_missing_column_families = true;
+  // Also setup to count number of options files created (see check below)
+  auto my_fs =
+      std::make_shared<CountOptionsFilesFs>(db_options_.env->GetFileSystem());
+  auto my_env = std::make_unique<CompositeEnvWrapper>(db_options_.env, my_fs);
+  SaveAndRestore<Env*> save_restore_env(&db_options_.env, my_env.get());
+
+  ASSERT_OK(TryOpen({"default", "one", "two"}));
+  Close();
+
+  // An older version would write an updated options file for each column
+  // family created under create_missing_column_families, which would be
+  // quadratic I/O in the number of column families.
+  ASSERT_EQ(my_fs->options_files_created.load(), 1);
+
+  // Add to existing DB case
+  ASSERT_OK(TryOpen({"default", "one", "two", "three", "four"}));
+  Close();
+  ASSERT_EQ(my_fs->options_files_created.load(), 2);
 }
 
 TEST_P(ColumnFamilyTest, SanitizeOptions) {
@@ -2427,7 +2473,10 @@ void DropSingleColumnFamily(ColumnFamilyTest* cf_test, int cf_id,
 }
 }  // anonymous namespace
 
-TEST_P(ColumnFamilyTest, CreateAndDropRace) {
+// This test attempts to set up a race condition in a way that is no longer
+// possible, causing the test to hang. If DBImpl::options_mutex_ is removed
+// in the future, this test might become relevant again.
+TEST_P(ColumnFamilyTest, DISABLED_CreateAndDropRace) {
   const int kCfCount = 5;
   std::vector<ColumnFamilyOptions> cf_opts;
   std::vector<Comparator*> comparators;
@@ -2489,6 +2538,53 @@ TEST_P(ColumnFamilyTest, CreateAndDropRace) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }
 
+TEST_P(ColumnFamilyTest, CreateAndDropPeriodicRace) {
+  // This is a mini-stress test looking for inconsistency between the set of
+  // CFs in the DB, particularly whether any use preserve_internal_time_seconds,
+  // and whether that is accurately reflected in the periodic task setup.
+  constexpr size_t kNumThreads = 12;
+  std::vector<std::thread> threads;
+  bool last_cf_on = Random::GetTLSInstance()->OneIn(2);
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::RegisterRecordSeqnoTimeWorker:BeforePeriodicTaskType",
+      [&](void* /*arg*/) { std::this_thread::yield(); });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_EQ(column_family_options_.preserve_internal_time_seconds, 0U);
+  ColumnFamilyOptions other_opts = column_family_options_;
+  ColumnFamilyOptions last_opts = column_family_options_;
+  (last_cf_on ? last_opts : other_opts).preserve_internal_time_seconds =
+      1000000;
+  Open();
+
+  for (size_t i = 0; i < kNumThreads; i++) {
+    threads.emplace_back([this, &other_opts, i]() {
+      ColumnFamilyHandle* cfh;
+      ASSERT_OK(db_->CreateColumnFamily(other_opts, std::to_string(i), &cfh));
+      ASSERT_OK(db_->DropColumnFamily(cfh));
+      ASSERT_OK(db_->DestroyColumnFamilyHandle(cfh));
+    });
+  }
+
+  ColumnFamilyHandle* last_cfh;
+  ASSERT_OK(db_->CreateColumnFamily(last_opts, "last", &last_cfh));
+
+  for (auto& t : threads) {
+    t.join();
+  }
+
+  bool task_enabled = dbfull()->TEST_GetPeriodicTaskScheduler().TEST_HasTask(
+      PeriodicTaskType::kRecordSeqnoTime);
+  ASSERT_EQ(last_cf_on, task_enabled);
+
+  ASSERT_OK(db_->DropColumnFamily(last_cfh));
+  ASSERT_OK(db_->DestroyColumnFamilyHandle(last_cfh));
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+}
+
 TEST_P(ColumnFamilyTest, WriteStallSingleColumnFamily) {
   const uint64_t kBaseRate = 800000u;
   db_options_.delayed_write_rate = kBaseRate;

db/compaction/compaction.cc

@@ -115,6 +115,42 @@ void Compaction::GetBoundaryKeys(
   }
 }
 
+void Compaction::GetBoundaryInternalKeys(
+    VersionStorageInfo* vstorage,
+    const std::vector<CompactionInputFiles>& inputs, InternalKey* smallest_key,
+    InternalKey* largest_key, int exclude_level) {
+  bool initialized = false;
+  const InternalKeyComparator* icmp = vstorage->InternalComparator();
+  for (size_t i = 0; i < inputs.size(); ++i) {
+    if (inputs[i].files.empty() || inputs[i].level == exclude_level) {
+      continue;
+    }
+    if (inputs[i].level == 0) {
+      // we need to consider all files on level 0
+      for (const auto* f : inputs[i].files) {
+        if (!initialized || icmp->Compare(f->smallest, *smallest_key) < 0) {
+          *smallest_key = f->smallest;
+        }
+        if (!initialized || icmp->Compare(f->largest, *largest_key) > 0) {
+          *largest_key = f->largest;
+        }
+        initialized = true;
+      }
+    } else {
+      // we only need to consider the first and last file
+      if (!initialized ||
+          icmp->Compare(inputs[i].files[0]->smallest, *smallest_key) < 0) {
+        *smallest_key = inputs[i].files[0]->smallest;
+      }
+      if (!initialized ||
+          icmp->Compare(inputs[i].files.back()->largest, *largest_key) > 0) {
+        *largest_key = inputs[i].files.back()->largest;
+      }
+      initialized = true;
+    }
+  }
+}
+
 std::vector<CompactionInputFiles> Compaction::PopulateWithAtomicBoundaries(
     VersionStorageInfo* vstorage, std::vector<CompactionInputFiles> inputs) {
   const Comparator* ucmp = vstorage->InternalComparator()->user_comparator();
@@ -399,9 +435,14 @@ void Compaction::PopulatePenultimateLevelOutputRange() {
     }
   }
 
-  GetBoundaryKeys(input_vstorage_, inputs_,
-                  &penultimate_level_smallest_user_key_,
-                  &penultimate_level_largest_user_key_, exclude_level);
+  // FIXME: should make use of `penultimate_output_range_type_`.
+  // FIXME: when last level's input range does not overlap with
+  //  penultimate level, and penultimate level input is empty,
+  //  this call will not set penultimate_level_smallest_ or
+  //  penultimate_level_largest_. No keys will be compacted up.
+  GetBoundaryInternalKeys(input_vstorage_, inputs_,
+                          &penultimate_level_smallest_,
+                          &penultimate_level_largest_, exclude_level);
 }
 
 Compaction::~Compaction() {
@@ -426,33 +467,40 @@ bool Compaction::OverlapPenultimateLevelOutputRange(
   if (!SupportsPerKeyPlacement()) {
     return false;
   }
+
+  // See FIXME in Compaction::PopulatePenultimateLevelOutputRange().
+  // We do not compact any key up in this case.
+  if (penultimate_level_smallest_.size() == 0 ||
+      penultimate_level_largest_.size() == 0) {
+    return false;
+  }
+
   const Comparator* ucmp =
       input_vstorage_->InternalComparator()->user_comparator();
 
   return ucmp->CompareWithoutTimestamp(
-             smallest_key, penultimate_level_largest_user_key_) <= 0 &&
+             smallest_key, penultimate_level_largest_.user_key()) <= 0 &&
          ucmp->CompareWithoutTimestamp(
-             largest_key, penultimate_level_smallest_user_key_) >= 0;
+             largest_key, penultimate_level_smallest_.user_key()) >= 0;
 }
 
 // key includes timestamp if user-defined timestamp is enabled.
-bool Compaction::WithinPenultimateLevelOutputRange(const Slice& key) const {
+bool Compaction::WithinPenultimateLevelOutputRange(
+    const ParsedInternalKey& ikey) const {
   if (!SupportsPerKeyPlacement()) {
     return false;
   }
 
-  if (penultimate_level_smallest_user_key_.empty() ||
-      penultimate_level_largest_user_key_.empty()) {
+  if (penultimate_level_smallest_.size() == 0 ||
+      penultimate_level_largest_.size() == 0) {
     return false;
   }
 
-  const Comparator* ucmp =
-      input_vstorage_->InternalComparator()->user_comparator();
+  const InternalKeyComparator* icmp = input_vstorage_->InternalComparator();
 
-  return ucmp->CompareWithoutTimestamp(
-             key, penultimate_level_smallest_user_key_) >= 0 &&
-         ucmp->CompareWithoutTimestamp(
-             key, penultimate_level_largest_user_key_) <= 0;
+  // op_type of a key can change during compaction, e.g. Merge -> Put.
+  return icmp->CompareKeySeq(ikey, penultimate_level_smallest_.Encode()) >= 0 &&
+         icmp->CompareKeySeq(ikey, penultimate_level_largest_.Encode()) <= 0;
 }
 
 bool Compaction::InputCompressionMatchesOutput() const {

db/compaction/compaction.h

@@ -353,14 +353,6 @@ class Compaction {
 
   Slice GetLargestUserKey() const { return largest_user_key_; }
 
-  Slice GetPenultimateLevelSmallestUserKey() const {
-    return penultimate_level_smallest_user_key_;
-  }
-
-  Slice GetPenultimateLevelLargestUserKey() const {
-    return penultimate_level_largest_user_key_;
-  }
-
   PenultimateOutputRangeType GetPenultimateOutputRangeType() const {
     return penultimate_output_range_type_;
   }
@@ -383,10 +375,8 @@ class Compaction {
   // per_key_placement feature, which is safe to place the key to the
   // penultimate level. different compaction strategy has different rules.
   // If per_key_placement is not supported, always return false.
-  // TODO: currently it doesn't support moving data from the last level to the
-  //  penultimate level
   //  key includes timestamp if user-defined timestamp is enabled.
-  bool WithinPenultimateLevelOutputRange(const Slice& key) const;
+  bool WithinPenultimateLevelOutputRange(const ParsedInternalKey& ikey) const;
 
   CompactionReason compaction_reason() const { return compaction_reason_; }
 
@@ -456,6 +446,13 @@ class Compaction {
                               Slice* smallest_key, Slice* largest_key,
                               int exclude_level = -1);
 
+  // get the smallest and largest internal key present in files to be compacted
+  static void GetBoundaryInternalKeys(
+      VersionStorageInfo* vstorage,
+      const std::vector<CompactionInputFiles>& inputs,
+      InternalKey* smallest_key, InternalKey* largest_key,
+      int exclude_level = -1);
+
   // populate penultimate level output range, which will be used to determine if
   // a key is safe to output to the penultimate level (details see
   // `Compaction::WithinPenultimateLevelOutputRange()`.
@@ -568,8 +565,8 @@ class Compaction {
   // Key range for penultimate level output
   // includes timestamp if user-defined timestamp is enabled.
   // penultimate_output_range_type_ shows the range type
-  Slice penultimate_level_smallest_user_key_;
-  Slice penultimate_level_largest_user_key_;
+  InternalKey penultimate_level_smallest_;
+  InternalKey penultimate_level_largest_;
   PenultimateOutputRangeType penultimate_output_range_type_ =
       PenultimateOutputRangeType::kNotSupported;
 };

db/compaction/compaction_iterator.cc

@@ -1227,7 +1227,7 @@ void CompactionIterator::DecideOutputLevel() {
     // not from this compaction.
     // TODO: add statistic for declined output_to_penultimate_level
     bool safe_to_penultimate_level =
-        compaction_->WithinPenultimateLevelOutputRange(ikey_.user_key);
+        compaction_->WithinPenultimateLevelOutputRange(ikey_);
     if (!safe_to_penultimate_level) {
       output_to_penultimate_level_ = false;
       // It could happen when disable/enable `last_level_temperature` while
@@ -1256,10 +1256,13 @@ void CompactionIterator::PrepareOutput() {
       } else if (ikey_.type == kTypeBlobIndex) {
         GarbageCollectBlobIfNeeded();
       }
-    }
 
-    if (compaction_ != nullptr && compaction_->SupportsPerKeyPlacement()) {
-      DecideOutputLevel();
+      // For range del sentinel, we don't use it to cut files for bottommost
+      // compaction. So it should not make a difference which output level we
+      // decide.
+      if (compaction_ != nullptr && compaction_->SupportsPerKeyPlacement()) {
+        DecideOutputLevel();
+      }
     }
 
     // Zeroing out the sequence number leads to better compression.

db/compaction/compaction_iterator.h

@@ -119,7 +119,8 @@ class CompactionIterator {
     virtual bool SupportsPerKeyPlacement() const = 0;
 
     // `key` includes timestamp if user-defined timestamp is enabled.
-    virtual bool WithinPenultimateLevelOutputRange(const Slice& key) const = 0;
+    virtual bool WithinPenultimateLevelOutputRange(
+        const ParsedInternalKey&) const = 0;
   };
 
   class RealCompaction : public CompactionProxy {
@@ -186,8 +187,9 @@ class CompactionIterator {
     // Check if key is within penultimate level output range, to see if it's
     // safe to output to the penultimate level for per_key_placement feature.
     // `key` includes timestamp if user-defined timestamp is enabled.
-    bool WithinPenultimateLevelOutputRange(const Slice& key) const override {
-      return compaction_->WithinPenultimateLevelOutputRange(key);
+    bool WithinPenultimateLevelOutputRange(
+        const ParsedInternalKey& ikey) const override {
+      return compaction_->WithinPenultimateLevelOutputRange(ikey);
     }
 
    private:

db/compaction/compaction_iterator_test.cc

@@ -184,8 +184,9 @@ class FakeCompaction : public CompactionIterator::CompactionProxy {
     return supports_per_key_placement;
   }
 
-  bool WithinPenultimateLevelOutputRange(const Slice& key) const override {
-    return (!key.starts_with("unsafe_pb"));
+  bool WithinPenultimateLevelOutputRange(
+      const ParsedInternalKey& key) const override {
+    return (!key.user_key.starts_with("unsafe_pb"));
   }
 
   bool key_not_exists_beyond_output_level = false;

db/compaction/compaction_job.cc

@@ -288,23 +288,23 @@ void CompactionJob::Prepare() {
 
   if (preserve_time_duration > 0) {
     const ReadOptions read_options(Env::IOActivity::kCompaction);
-    // setup seqno_time_mapping_
-    seqno_time_mapping_.SetMaxTimeDuration(preserve_time_duration);
+    // setup seqno_to_time_mapping_
+    seqno_to_time_mapping_.SetMaxTimeDuration(preserve_time_duration);
     for (const auto& each_level : *c->inputs()) {
       for (const auto& fmd : each_level.files) {
         std::shared_ptr<const TableProperties> tp;
         Status s =
             cfd->current()->GetTableProperties(read_options, &tp, fmd, nullptr);
         if (s.ok()) {
-          seqno_time_mapping_.Add(tp->seqno_to_time_mapping)
+          seqno_to_time_mapping_.Add(tp->seqno_to_time_mapping)
               .PermitUncheckedError();
-          seqno_time_mapping_.Add(fmd->fd.smallest_seqno,
-                                  fmd->oldest_ancester_time);
+          seqno_to_time_mapping_.Add(fmd->fd.smallest_seqno,
+                                     fmd->oldest_ancester_time);
         }
       }
     }
 
-    auto status = seqno_time_mapping_.Sort();
+    auto status = seqno_to_time_mapping_.Sort();
     if (!status.ok()) {
       ROCKS_LOG_WARN(db_options_.info_log,
                      "Invalid sequence number to time mapping: Status: %s",
@@ -320,13 +320,17 @@ void CompactionJob::Prepare() {
       preserve_time_min_seqno_ = 0;
       preclude_last_level_min_seqno_ = 0;
     } else {
-      seqno_time_mapping_.TruncateOldEntries(_current_time);
+      seqno_to_time_mapping_.TruncateOldEntries(_current_time);
       uint64_t preserve_time =
           static_cast<uint64_t>(_current_time) > preserve_time_duration
               ? _current_time - preserve_time_duration
               : 0;
+      // GetProximalSeqnoBeforeTime tells us the last seqno known to have been
+      // written at or before the given time. + 1 to get the minimum we should
+      // preserve without excluding anything that might have been written on or
+      // after the given time.
       preserve_time_min_seqno_ =
-          seqno_time_mapping_.GetOldestSequenceNum(preserve_time);
+          seqno_to_time_mapping_.GetProximalSeqnoBeforeTime(preserve_time) + 1;
       if (c->immutable_options()->preclude_last_level_data_seconds > 0) {
         uint64_t preclude_last_level_time =
             static_cast<uint64_t>(_current_time) >
@@ -335,7 +339,9 @@ void CompactionJob::Prepare() {
                       c->immutable_options()->preclude_last_level_data_seconds
                 : 0;
         preclude_last_level_min_seqno_ =
-            seqno_time_mapping_.GetOldestSequenceNum(preclude_last_level_time);
+            seqno_to_time_mapping_.GetProximalSeqnoBeforeTime(
+                preclude_last_level_time) +
+            1;
       }
     }
   }
@@ -1570,7 +1576,7 @@ Status CompactionJob::FinishCompactionOutputFile(
 
   const uint64_t current_entries = outputs.NumEntries();
 
-  s = outputs.Finish(s, seqno_time_mapping_);
+  s = outputs.Finish(s, seqno_to_time_mapping_);
 
   if (s.ok()) {
     // With accurate smallest and largest key, we can get a slightly more
@@ -1894,6 +1900,8 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
           sub_compact->start.has_value() ? &tmp_start : nullptr,
           sub_compact->end.has_value() ? &tmp_end : nullptr);
   if (oldest_ancester_time == std::numeric_limits<uint64_t>::max()) {
+    // TODO: fix DBSSTTest.GetTotalSstFilesSize and use
+    //  kUnknownOldestAncesterTime
     oldest_ancester_time = current_time;
   }
 
@@ -1938,6 +1946,7 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
       db_options_.stats, listeners, db_options_.file_checksum_gen_factory.get(),
       tmp_set.Contains(FileType::kTableFile), false));
 
+  // TODO(hx235): pass in the correct `oldest_key_time` instead of `0`
   TableBuilderOptions tboptions(
       *cfd->ioptions(), *(sub_compact->compaction->mutable_cf_options()),
       cfd->internal_comparator(), cfd->int_tbl_prop_collector_factories(),

db/compaction/compaction_job.h

@@ -350,7 +350,7 @@ class CompactionJob {
 
   // Stores the sequence number to time mapping gathered from all input files
   // it also collects the smallest_seqno -> oldest_ancester_time from the SST.
-  SeqnoToTimeMapping seqno_time_mapping_;
+  SeqnoToTimeMapping seqno_to_time_mapping_;
 
   // Minimal sequence number for preserving the time information. The time info
   // older than this sequence number won't be preserved after the compaction and

db/compaction/compaction_job_test.cc

@@ -215,7 +215,9 @@ class CompactionJobTestBase : public testing::Test {
             dbname_, &db_options_, env_options_, table_cache_.get(),
             &write_buffer_manager_, &write_controller_,
             /*block_cache_tracer=*/nullptr,
-            /*io_tracer=*/nullptr, /*db_id*/ "", /*db_session_id*/ "")),
+            /*io_tracer=*/nullptr, /*db_id=*/"", /*db_session_id=*/"",
+            /*daily_offpeak_time_utc=*/"",
+            /*error_handler=*/nullptr)),
         shutting_down_(false),
         mock_table_factory_(new mock::MockTableFactory()),
         error_handler_(nullptr, db_options_, &mutex_),
@@ -540,11 +542,12 @@ class CompactionJobTestBase : public testing::Test {
     ASSERT_OK(s);
     db_options_.info_log = info_log;
 
-    versions_.reset(
-        new VersionSet(dbname_, &db_options_, env_options_, table_cache_.get(),
-                       &write_buffer_manager_, &write_controller_,
-                       /*block_cache_tracer=*/nullptr, /*io_tracer=*/nullptr,
-                       /*db_id*/ "", /*db_session_id*/ ""));
+    versions_.reset(new VersionSet(
+        dbname_, &db_options_, env_options_, table_cache_.get(),
+        &write_buffer_manager_, &write_controller_,
+        /*block_cache_tracer=*/nullptr, /*io_tracer=*/nullptr,
+        /*db_id=*/"", /*db_session_id=*/"", /*daily_offpeak_time_utc=*/"",
+        /*error_handler=*/nullptr));
     compaction_job_stats_.Reset();
     ASSERT_OK(SetIdentityFile(env_, dbname_));
 
@@ -1524,13 +1527,15 @@ TEST_F(CompactionJobTest, VerifyPenultimateLevelOutput) {
       {files0, files1, files2, files3}, input_levels,
       /*verify_func=*/[&](Compaction& comp) {
         for (char c = 'a'; c <= 'z'; c++) {
-          std::string c_str;
-          c_str = c;
-          const Slice key(c_str);
           if (c == 'a') {
-            ASSERT_FALSE(comp.WithinPenultimateLevelOutputRange(key));
+            ParsedInternalKey pik("a", 0U, kTypeValue);
+            ASSERT_FALSE(comp.WithinPenultimateLevelOutputRange(pik));
           } else {
-            ASSERT_TRUE(comp.WithinPenultimateLevelOutputRange(key));
+            std::string c_str{c};
+            // WithinPenultimateLevelOutputRange checks internal key range.
+            // 'z' is the last key, so set seqno properly.
+            ParsedInternalKey pik(c_str, c == 'z' ? 12U : 0U, kTypeValue);
+            ASSERT_TRUE(comp.WithinPenultimateLevelOutputRange(pik));
           }
         }
       });

db/compaction/compaction_outputs.cc

@@ -18,16 +18,18 @@ void CompactionOutputs::NewBuilder(const TableBuilderOptions& tboptions) {
   builder_.reset(NewTableBuilder(tboptions, file_writer_.get()));
 }
 
-Status CompactionOutputs::Finish(const Status& intput_status,
-                                 const SeqnoToTimeMapping& seqno_time_mapping) {
+Status CompactionOutputs::Finish(
+    const Status& intput_status,
+    const SeqnoToTimeMapping& seqno_to_time_mapping) {
   FileMetaData* meta = GetMetaData();
   assert(meta != nullptr);
   Status s = intput_status;
   if (s.ok()) {
-    std::string seqno_time_mapping_str;
-    seqno_time_mapping.Encode(seqno_time_mapping_str, meta->fd.smallest_seqno,
-                              meta->fd.largest_seqno, meta->file_creation_time);
-    builder_->SetSeqnoTimeTableProperties(seqno_time_mapping_str,
+    std::string seqno_to_time_mapping_str;
+    seqno_to_time_mapping.Encode(
+        seqno_to_time_mapping_str, meta->fd.smallest_seqno,
+        meta->fd.largest_seqno, meta->file_creation_time);
+    builder_->SetSeqnoTimeTableProperties(seqno_to_time_mapping_str,
                                           meta->oldest_ancester_time);
     s = builder_->Finish();
 

db/compaction/compaction_outputs.h

@@ -107,7 +107,7 @@ class CompactionOutputs {
 
   // Finish the current output file
   Status Finish(const Status& intput_status,
-                const SeqnoToTimeMapping& seqno_time_mapping);
+                const SeqnoToTimeMapping& seqno_to_time_mapping);
 
   // Update output table properties from table builder
   void UpdateTableProperties() {

db/compaction/compaction_picker_fifo.cc

@@ -17,6 +17,9 @@
 #include "logging/log_buffer.h"
 #include "logging/logging.h"
 #include "options/options_helper.h"
+#include "rocksdb/listener.h"
+#include "rocksdb/statistics.h"
+#include "rocksdb/status.h"
 #include "util/string_util.h"
 
 namespace ROCKSDB_NAMESPACE {

db/compaction/compaction_picker_test.cc

@@ -85,7 +85,8 @@ class CompactionPickerTestBase : public testing::Test {
     vstorage_.reset(new VersionStorageInfo(
         &icmp_, ucmp_, options_.num_levels, style, nullptr, false,
         EpochNumberRequirement::kMustPresent, ioptions_.clock,
-        options_.bottommost_file_compaction_delay));
+        options_.bottommost_file_compaction_delay,
+        OffpeakTimeOption(mutable_db_options_.daily_offpeak_time_utc)));
     vstorage_->PrepareForVersionAppend(ioptions_, mutable_cf_options_);
   }
 
@@ -95,7 +96,8 @@ class CompactionPickerTestBase : public testing::Test {
     temp_vstorage_.reset(new VersionStorageInfo(
         &icmp_, ucmp_, options_.num_levels, ioptions_.compaction_style,
         vstorage_.get(), false, EpochNumberRequirement::kMustPresent,
-        ioptions_.clock, options_.bottommost_file_compaction_delay));
+        ioptions_.clock, options_.bottommost_file_compaction_delay,
+        OffpeakTimeOption(mutable_db_options_.daily_offpeak_time_utc)));
   }
 
   void DeleteVersionStorage() {

db/compaction/tiered_compaction_test.cc

@@ -1202,20 +1202,126 @@ TEST_P(TieredCompactionTest, RangeBasedTieredStorageLevel) {
   ASSERT_EQ("0,0,0,0,0,0,1", FilesPerLevel());
   ASSERT_EQ(GetSstSizeHelper(Temperature::kUnknown), 0);
   ASSERT_GT(GetSstSizeHelper(Temperature::kCold), 0);
-
   ASSERT_EQ(
       options.statistics->getTickerCount(COMPACTION_RANGE_DEL_DROP_OBSOLETE),
       1);
+
+  // Tests that we only compact keys up to penultimate level
+  // that are within penultimate level input's internal key range.
+  {
+    MutexLock l(&mutex);
+    hot_start = Key(0);
+    hot_end = Key(100);
+  }
+  const Snapshot* temp_snap = db_->GetSnapshot();
+  // Key(0) and Key(1) here are inserted with higher sequence number
+  // than Key(0) and Key(1) inserted above.
+  // Only Key(0) in last level will be compacted up, not Key(1).
+  ASSERT_OK(Put(Key(0), "value" + std::to_string(0)));
+  ASSERT_OK(Put(Key(1), "value" + std::to_string(100)));
+  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
+  ASSERT_EQ("0,0,0,0,0,1,1", FilesPerLevel());
+  {
+    std::vector<LiveFileMetaData> metas;
+    db_->GetLiveFilesMetaData(&metas);
+    for (const auto& f : metas) {
+      if (f.temperature == Temperature::kUnknown) {
+        // Expect Key(0), Key(0), Key(1)
+        ASSERT_EQ(f.num_entries, 3);
+        ASSERT_EQ(f.smallestkey, Key(0));
+        ASSERT_EQ(f.largestkey, Key(1));
+      } else {
+        ASSERT_EQ(f.temperature, Temperature::kCold);
+        // Key(2)-Key(49) and Key(100).
+        ASSERT_EQ(f.num_entries, 50);
+      }
+    }
+  }
+  db_->ReleaseSnapshot(temp_snap);
 }
 
 INSTANTIATE_TEST_CASE_P(TieredCompactionTest, TieredCompactionTest,
                         testing::Bool());
 
+TEST_P(TieredCompactionTest, CheckInternalKeyRange) {
+  // When compacting keys from the last level to penultimate level,
+  // output to penultimate level should be within internal key range
+  // of input files from penultimate level.
+  // Set up:
+  // L5:
+  //  File 1: DeleteRange[1, 3)@4, File 2: [3@5, 100@6]
+  // L6:
+  // File 3: [2@1, 3@2], File 4: [50@3]
+  //
+  // When File 1 and File 3 are being compacted,
+  // Key(3) cannot be compacted up, otherwise it causes
+  // inconsistency where File 3's Key(3) has a lower sequence number
+  // than File 2's Key(3).
+  const int kNumLevels = 7;
+  auto options = CurrentOptions();
+  SetColdTemperature(options);
+  options.level_compaction_dynamic_level_bytes = true;
+  options.num_levels = kNumLevels;
+  options.statistics = CreateDBStatistics();
+  options.max_subcompactions = 10;
+  options.preclude_last_level_data_seconds = 10000;
+  DestroyAndReopen(options);
+  auto cmp = options.comparator;
+
+  std::string hot_start = Key(0);
+  std::string hot_end = Key(0);
+  SyncPoint::GetInstance()->SetCallBack(
+      "CompactionIterator::PrepareOutput.context", [&](void* arg) {
+        auto context = static_cast<PerKeyPlacementContext*>(arg);
+        context->output_to_penultimate_level =
+            cmp->Compare(context->key, hot_start) >= 0 &&
+            cmp->Compare(context->key, hot_end) < 0;
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+  // File 1
+  ASSERT_OK(Put(Key(2), "val2"));
+  ASSERT_OK(Put(Key(3), "val3"));
+  ASSERT_OK(Flush());
+  MoveFilesToLevel(6);
+  // File 2
+  ASSERT_OK(Put(Key(50), "val50"));
+  ASSERT_OK(Flush());
+  MoveFilesToLevel(6);
+
+  const Snapshot* snapshot = db_->GetSnapshot();
+  hot_end = Key(100);
+  std::string start = Key(1);
+  std::string end = Key(3);
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), start, end));
+  ASSERT_OK(Flush());
+  MoveFilesToLevel(5);
+  // File 3
+  ASSERT_OK(Put(Key(3), "vall"));
+  ASSERT_OK(Put(Key(100), "val100"));
+  ASSERT_OK(Flush());
+  MoveFilesToLevel(5);
+  // Try to compact keys up
+  CompactRangeOptions cro;
+  cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+  start = Key(1);
+  end = Key(2);
+  Slice begin_slice(start);
+  Slice end_slice(end);
+  ASSERT_OK(db_->CompactRange(cro, &begin_slice, &end_slice));
+  // Without internal key range checking, we get the following error:
+  // Corruption: force_consistency_checks(DEBUG): VersionBuilder: L5 has
+  // overlapping ranges: file #18 largest key: '6B6579303030303033' seq:102,
+  // type:1 vs. file #15 smallest key: '6B6579303030303033' seq:104, type:1
+  db_->ReleaseSnapshot(snapshot);
+}
+
 class PrecludeLastLevelTest : public DBTestBase {
  public:
   PrecludeLastLevelTest()
       : DBTestBase("preclude_last_level_test", /*env_do_fsync=*/false) {
     mock_clock_ = std::make_shared<MockSystemClock>(env_->GetSystemClock());
+    mock_clock_->SetCurrentTime(kMockStartTime);
     mock_env_ = std::make_unique<CompositeEnvWrapper>(env_, mock_clock_);
   }
 
@@ -1223,6 +1329,10 @@ class PrecludeLastLevelTest : public DBTestBase {
   std::unique_ptr<Env> mock_env_;
   std::shared_ptr<MockSystemClock> mock_clock_;
 
+  // Sufficient starting time that preserve time doesn't under-flow into
+  // pre-history
+  static constexpr uint32_t kMockStartTime = 10000000;
+
   void SetUp() override {
     mock_clock_->InstallTimedWaitFixCallback();
     SyncPoint::GetInstance()->SetCallBack(
@@ -1231,7 +1341,7 @@ class PrecludeLastLevelTest : public DBTestBase {
               reinterpret_cast<PeriodicTaskScheduler*>(arg);
           periodic_task_scheduler_ptr->TEST_OverrideTimer(mock_clock_.get());
         });
-    mock_clock_->SetCurrentTime(0);
+    mock_clock_->SetCurrentTime(kMockStartTime);
   }
 };
 
@@ -1249,11 +1359,6 @@ TEST_F(PrecludeLastLevelTest, MigrationFromPreserveTimeManualCompaction) {
   options.num_levels = kNumLevels;
   DestroyAndReopen(options);
 
-  // pass some time first, otherwise the first a few keys write time are going
-  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
-  dbfull()->TEST_WaitForPeriodicTaskRun(
-      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec)); });
-
   int sst_num = 0;
   // Write files that are overlap and enough to trigger compaction
   for (; sst_num < kNumTrigger; sst_num++) {
@@ -1311,11 +1416,6 @@ TEST_F(PrecludeLastLevelTest, MigrationFromPreserveTimeAutoCompaction) {
   options.num_levels = kNumLevels;
   DestroyAndReopen(options);
 
-  // pass some time first, otherwise the first a few keys write time are going
-  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
-  dbfull()->TEST_WaitForPeriodicTaskRun(
-      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec)); });
-
   int sst_num = 0;
   // Write files that are overlap and enough to trigger compaction
   for (; sst_num < kNumTrigger; sst_num++) {
@@ -1387,11 +1487,6 @@ TEST_F(PrecludeLastLevelTest, MigrationFromPreserveTimePartial) {
   options.num_levels = kNumLevels;
   DestroyAndReopen(options);
 
-  // pass some time first, otherwise the first a few keys write time are going
-  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
-  dbfull()->TEST_WaitForPeriodicTaskRun(
-      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec)); });
-
   int sst_num = 0;
   // Write files that are overlap and enough to trigger compaction
   for (; sst_num < kNumTrigger; sst_num++) {
@@ -1514,11 +1609,6 @@ TEST_F(PrecludeLastLevelTest, LastLevelOnlyCompactionPartial) {
   options.num_levels = kNumLevels;
   DestroyAndReopen(options);
 
-  // pass some time first, otherwise the first a few keys write time are going
-  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
-  dbfull()->TEST_WaitForPeriodicTaskRun(
-      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec)); });
-
   int sst_num = 0;
   // Write files that are overlap and enough to trigger compaction
   for (; sst_num < kNumTrigger; sst_num++) {
@@ -1592,11 +1682,6 @@ TEST_P(PrecludeLastLevelTestWithParms, LastLevelOnlyCompactionNoPreclude) {
   options.num_levels = kNumLevels;
   DestroyAndReopen(options);
 
-  // pass some time first, otherwise the first a few keys write time are going
-  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
-  dbfull()->TEST_WaitForPeriodicTaskRun(
-      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(kKeyPerSec)); });
-
   Random rnd(301);
   int sst_num = 0;
   // Write files that are overlap and enough to trigger compaction
@@ -1906,11 +1991,6 @@ TEST_F(PrecludeLastLevelTest, PartialPenultimateLevelCompaction) {
   options.num_levels = kNumLevels;
   DestroyAndReopen(options);
 
-  // pass some time first, otherwise the first a few keys write time are going
-  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
-  dbfull()->TEST_WaitForPeriodicTaskRun(
-      [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(10)); });
-
   Random rnd(301);
 
   for (int i = 0; i < 300; i++) {
@@ -2017,12 +2097,6 @@ TEST_F(PrecludeLastLevelTest, RangeDelsCauseFileEndpointsToOverlap) {
   options.target_file_size_base = kFileBytes;
   DestroyAndReopen(options);
 
-  // pass some time first, otherwise the first a few keys write time are going
-  // to be zero, and internally zero has special meaning: kUnknownSeqnoTime
-  dbfull()->TEST_WaitForPeriodicTaskRun([&] {
-    mock_clock_->MockSleepForSeconds(static_cast<int>(kSecondsPerKey));
-  });
-
   // Flush an L0 file with the following contents (new to old):
   //
   // Range deletions [4, 6) [7, 8) [9, 11)
@@ -2139,7 +2213,6 @@ TEST_F(PrecludeLastLevelTest, RangeDelsCauseFileEndpointsToOverlap) {
   Close();
 }
 
-
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/comparator_db_test.cc

@@ -165,6 +165,7 @@ void DoRandomIteraratorTest(DB* db, std::vector<std::string> source_strings,
     AssertItersEqual(iter.get(), result_iter.get());
     is_valid = iter->Valid();
   }
+  ASSERT_OK(iter->status());
 }
 
 class DoubleComparator : public Comparator {

db/db_basic_test.cc

@@ -138,6 +138,7 @@ TEST_F(DBBasicTest, ReadOnlyDB) {
       ASSERT_OK(iter->status());
       ++count;
     }
+    ASSERT_OK(iter->status());
     // Always expect two keys: "foo" and "bar"
     ASSERT_EQ(count, 2);
   };
@@ -4414,6 +4415,8 @@ TEST_F(DBBasicTest, ManifestWriteFailure) {
   options.create_if_missing = true;
   options.disable_auto_compactions = true;
   options.env = env_;
+  options.enable_blob_files = true;
+  options.blob_file_size = 0;
   DestroyAndReopen(options);
   ASSERT_OK(Put("foo", "bar"));
   ASSERT_OK(Flush());
@@ -4434,6 +4437,11 @@ TEST_F(DBBasicTest, ManifestWriteFailure) {
   SyncPoint::GetInstance()->ClearAllCallBacks();
   SyncPoint::GetInstance()->EnableProcessing();
   Reopen(options);
+  // The IO error was a mocked one from the `AfterSyncManifest` callback. The
+  // Flush's VersionEdit actually made it into the Manifest. So these keys can
+  // be read back. Read them to check all live sst files and blob files.
+  ASSERT_EQ("bar", Get("foo"));
+  ASSERT_EQ("value", Get("key"));
 }
 
 TEST_F(DBBasicTest, DestroyDefaultCfHandle) {

db/db_block_cache_test.cc

@@ -389,6 +389,7 @@ TEST_F(DBBlockCacheTest, FillCacheAndIterateDB) {
   while (iter->Valid()) {
     iter->Next();
   }
+  ASSERT_OK(iter->status());
   delete iter;
   iter = nullptr;
 }

db/db_bloom_filter_test.cc

@@ -1905,6 +1905,7 @@ TEST_F(DBBloomFilterTest, PrefixExtractorWithFilter2) {
   for (iter->Seek("zzzzz_AAAA"); iter->Valid(); iter->Next()) {
     iter_res.emplace_back(iter->value().ToString());
   }
+  ASSERT_OK(iter->status());
 
   std::vector<std::string> expected_res = {"val1", "val2", "val3", "val4"};
   ASSERT_EQ(iter_res, expected_res);

db/db_compaction_filter_test.cc

@@ -805,6 +805,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilterIgnoreSnapshot) {
       count++;
       iter->Next();
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(count, 6);
     read_options.snapshot = nullptr;
     std::unique_ptr<Iterator> iter1(db_->NewIterator(read_options));
@@ -815,6 +816,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilterIgnoreSnapshot) {
       count++;
       iter1->Next();
     }
+    ASSERT_OK(iter1->status());
     // We have deleted 10 keys from 40 using the compaction filter
     //  Keys 6-9 before the snapshot and 100-105 after the snapshot
     ASSERT_EQ(count, 30);

db/db_compaction_test.cc

@@ -19,7 +19,7 @@
 #include "rocksdb/concurrent_task_limiter.h"
 #include "rocksdb/experimental.h"
 #include "rocksdb/sst_file_writer.h"
-#include "rocksdb/utilities/convenience.h"
+#include "test_util/mock_time_env.h"
 #include "test_util/sync_point.h"
 #include "test_util/testutil.h"
 #include "util/concurrent_task_limiter_impl.h"
@@ -2673,6 +2673,7 @@ TEST_P(DBCompactionTestWithParam, ConvertCompactionStyle) {
     keys_in_db.append(iter->key().ToString());
     keys_in_db.push_back(',');
   }
+  ASSERT_OK(iter->status());
   delete iter;
 
   std::string expected_keys;
@@ -4793,9 +4794,9 @@ TEST_F(DBCompactionTest, LevelTtlCascadingCompactions) {
       ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
           "VersionEdit::EncodeTo:VarintOldestAncesterTime", [&](void* arg) {
             if (if_restart && if_open_all_files) {
-              std::string* encoded_fieled = static_cast<std::string*>(arg);
-              *encoded_fieled = "";
-              PutVarint64(encoded_fieled, 0);
+              std::string* encoded_field = static_cast<std::string*>(arg);
+              *encoded_field = "";
+              PutVarint64(encoded_field, 0);
             }
           });
       ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
@@ -4941,9 +4942,9 @@ TEST_F(DBCompactionTest, LevelPeriodicCompaction) {
       ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
           "VersionEdit::EncodeTo:VarintFileCreationTime", [&](void* arg) {
             if (if_restart && if_open_all_files) {
-              std::string* encoded_fieled = static_cast<std::string*>(arg);
-              *encoded_fieled = "";
-              PutVarint64(encoded_fieled, 0);
+              std::string* encoded_field = static_cast<std::string*>(arg);
+              *encoded_field = "";
+              PutVarint64(encoded_field, 0);
             }
           });
       ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
@@ -5019,6 +5020,106 @@ TEST_F(DBCompactionTest, LevelPeriodicCompaction) {
   }
 }
 
+TEST_F(DBCompactionTest, LevelPeriodicCompactionOffpeak) {
+  // This test simply checks if offpeak adjustment works in Leveled
+  // Compactions. For testing offpeak periodic compactions in various
+  // scenarios, please refer to
+  // DBTestUniversalCompaction2::PeriodicCompactionOffpeak
+  constexpr int kNumKeysPerFile = 32;
+  constexpr int kNumLevelFiles = 2;
+  constexpr int kValueSize = 100;
+  constexpr int kSecondsPerDay = 86400;
+  constexpr int kSecondsPerHour = 3600;
+  constexpr int kSecondsPerMinute = 60;
+
+  for (bool if_restart : {false, true}) {
+    SCOPED_TRACE("if_restart=" + std::to_string(if_restart));
+    Options options = CurrentOptions();
+    options.ttl = 0;
+    options.periodic_compaction_seconds = 5 * kSecondsPerDay;  // 5 days
+    // In the case where all files are opened and doing DB restart
+    // forcing the file creation time in manifest file to be 0 to
+    // simulate the case of reading from an old version.
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+        "VersionEdit::EncodeTo:VarintFileCreationTime", [&](void* arg) {
+          if (if_restart) {
+            std::string* encoded_field = static_cast<std::string*>(arg);
+            *encoded_field = "";
+            PutVarint64(encoded_field, 0);
+          }
+        });
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+    // Just to add some extra random days to current time
+    Random rnd(test::RandomSeed());
+    int days = rnd.Uniform(100);
+
+    int periodic_compactions = 0;
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+        "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
+          Compaction* compaction = static_cast<Compaction*>(arg);
+          auto compaction_reason = compaction->compaction_reason();
+          if (compaction_reason == CompactionReason::kPeriodicCompaction) {
+            periodic_compactions++;
+          }
+        });
+
+    // Starting at 12:15AM
+    int now_hour = 0;
+    int now_minute = 15;
+    auto mock_clock = std::make_shared<MockSystemClock>(env_->GetSystemClock());
+    auto mock_env = std::make_unique<CompositeEnvWrapper>(env_, mock_clock);
+    options.env = mock_env.get();
+    mock_clock->SetCurrentTime(days * kSecondsPerDay +
+                               now_hour * kSecondsPerHour +
+                               now_minute * kSecondsPerMinute);
+    // Offpeak is set from 12:30AM to 4:30AM
+    options.daily_offpeak_time_utc = "00:30-04:30";
+    Reopen(options);
+
+    for (int i = 0; i < kNumLevelFiles; ++i) {
+      for (int j = 0; j < kNumKeysPerFile; ++j) {
+        ASSERT_OK(
+            Put(Key(i * kNumKeysPerFile + j), rnd.RandomString(kValueSize)));
+      }
+      ASSERT_OK(Flush());
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ("2", FilesPerLevel());
+    ASSERT_EQ(0, periodic_compactions);
+
+    // Move clock forward by 1 hour. Now at 1:15AM Day 0. No compaction.
+    mock_clock->MockSleepForSeconds(1 * kSecondsPerHour);
+    ASSERT_OK(Put("a", "1"));
+    ASSERT_OK(Flush());
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    // Assert that the files stay in the same level
+    ASSERT_EQ("3", FilesPerLevel());
+    ASSERT_EQ(0, periodic_compactions);
+    MoveFilesToLevel(1);
+    ASSERT_EQ("0,3", FilesPerLevel());
+
+    // Move clock forward by 4 days and check if it triggers periodic
+    // comapaction at 1:15AM Day 4. Files created on Day 0 at 12:15AM is
+    // expected to expire before the offpeak starts next day at 12:30AM
+    mock_clock->MockSleepForSeconds(4 * kSecondsPerDay);
+    ASSERT_OK(Put("b", "2"));
+    if (if_restart) {
+      Reopen(options);
+    } else {
+      ASSERT_OK(Flush());
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ("1,3", FilesPerLevel());
+    // The two old files go through the periodic compaction process
+    ASSERT_EQ(2, periodic_compactions);
+
+    Destroy(options);
+
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
+
 TEST_F(DBCompactionTest, LevelPeriodicCompactionWithOldDB) {
   // This test makes sure that periodic compactions are working with a DB
   // where file_creation_time of some files is 0.

db/db_filesnapshot.cc

@@ -121,7 +121,7 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
 
   // DisableFileDeletions / EnableFileDeletions not supported in read-only DB
   if (deletions_disabled.ok()) {
-    Status s2 = EnableFileDeletions(/*force*/ false);
+    Status s2 = EnableFileDeletions(/*force=*/false);
     assert(s2.ok());
     s2.PermitUncheckedError();
   } else {

db/db_flush_test.cc

@@ -1377,6 +1377,7 @@ TEST_F(DBFlushTest, MemPurgeDeleteAndDeleteRange) {
         ASSERT_EQ(value, NOT_FOUND);
       count++;
     }
+    ASSERT_OK(iter->status());
 
     // Expected count here is 3: KEY3, KEY4, KEY5.
     ASSERT_EQ(count, EXPECTED_COUNT_FORLOOP);

db/db_impl/db_impl.cc

@@ -13,7 +13,6 @@
 #include <alloca.h>
 #endif
 
-#include <algorithm>
 #include <cinttypes>
 #include <cstdio>
 #include <map>
@@ -273,13 +272,15 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
   periodic_task_functions_.emplace(PeriodicTaskType::kFlushInfoLog,
                                    [this]() { this->FlushInfoLog(); });
   periodic_task_functions_.emplace(
-      PeriodicTaskType::kRecordSeqnoTime,
-      [this]() { this->RecordSeqnoToTimeMapping(); });
+      PeriodicTaskType::kRecordSeqnoTime, [this]() {
+        this->RecordSeqnoToTimeMapping(/*populate_historical_seconds=*/0);
+      });
 
-  versions_.reset(new VersionSet(dbname_, &immutable_db_options_, file_options_,
-                                 table_cache_.get(), write_buffer_manager_,
-                                 &write_controller_, &block_cache_tracer_,
-                                 io_tracer_, db_id_, db_session_id_));
+  versions_.reset(new VersionSet(
+      dbname_, &immutable_db_options_, file_options_, table_cache_.get(),
+      write_buffer_manager_, &write_controller_, &block_cache_tracer_,
+      io_tracer_, db_id_, db_session_id_, options.daily_offpeak_time_utc,
+      &error_handler_));
   column_family_memtables_.reset(
       new ColumnFamilyMemTablesImpl(versions_->GetColumnFamilySet()));
 
@@ -354,16 +355,13 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
   }
 
   // Make sure the IO Status stored in version set is set to OK.
-  bool file_deletion_disabled = !IsFileDeletionsEnabled();
   if (s.ok()) {
     IOStatus io_s = versions_->io_status();
     if (io_s.IsIOError()) {
       // If resuming from IOError resulted from MANIFEST write, then assert
       // that we must have already set the MANIFEST writer to nullptr during
-      // clean-up phase MANIFEST writing. We must have also disabled file
-      // deletions.
+      // clean-up phase MANIFEST writing.
       assert(!versions_->descriptor_log_);
-      assert(file_deletion_disabled);
       // Since we are trying to recover from MANIFEST write error, we need to
       // switch to a new MANIFEST anyway. The old MANIFEST can be corrupted.
       // Therefore, force writing a dummy version edit because we do not know
@@ -387,13 +385,18 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
     }
   }
 
-  // We cannot guarantee consistency of the WAL. So force flush Memtables of
-  // all the column families
   if (s.ok()) {
-    FlushOptions flush_opts;
-    // We allow flush to stall write since we are trying to resume from error.
-    flush_opts.allow_write_stall = true;
-    s = FlushAllColumnFamilies(flush_opts, context.flush_reason);
+    if (context.flush_reason == FlushReason::kErrorRecoveryRetryFlush) {
+      s = RetryFlushesForErrorRecovery(FlushReason::kErrorRecoveryRetryFlush,
+                                       true /* wait */);
+    } else {
+      // We cannot guarantee consistency of the WAL. So force flush Memtables of
+      // all the column families
+      FlushOptions flush_opts;
+      // We allow flush to stall write since we are trying to resume from error.
+      flush_opts.allow_write_stall = true;
+      s = FlushAllColumnFamilies(flush_opts, context.flush_reason);
+    }
     if (!s.ok()) {
       ROCKS_LOG_INFO(immutable_db_options_.info_log,
                      "DB resume requested but failed due to Flush failure [%s]",
@@ -401,34 +404,6 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
     }
   }
 
-  JobContext job_context(0);
-  FindObsoleteFiles(&job_context, true);
-  mutex_.Unlock();
-
-  job_context.manifest_file_number = 1;
-  if (job_context.HaveSomethingToDelete()) {
-    PurgeObsoleteFiles(job_context);
-  }
-  job_context.Clean();
-
-  if (s.ok()) {
-    assert(versions_->io_status().ok());
-    // If we reach here, we should re-enable file deletions if it was disabled
-    // during previous error handling.
-    if (file_deletion_disabled) {
-      // Always return ok
-      s = EnableFileDeletions(/*force=*/true);
-      if (!s.ok()) {
-        ROCKS_LOG_INFO(
-            immutable_db_options_.info_log,
-            "DB resume requested but could not enable file deletions [%s]",
-            s.ToString().c_str());
-        assert(false);
-      }
-    }
-  }
-
-  mutex_.Lock();
   if (s.ok()) {
     // This will notify and unblock threads waiting for error recovery to
     // finish. Those previouly waiting threads can now proceed, which may
@@ -441,6 +416,15 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
     error_handler_.GetRecoveryError().PermitUncheckedError();
   }
 
+  JobContext job_context(0);
+  FindObsoleteFiles(&job_context, true);
+  mutex_.Unlock();
+  job_context.manifest_file_number = 1;
+  if (job_context.HaveSomethingToDelete()) {
+    PurgeObsoleteFiles(job_context);
+  }
+  job_context.Clean();
+
   if (s.ok()) {
     ROCKS_LOG_INFO(immutable_db_options_.info_log, "Successfully resumed DB");
   } else {
@@ -448,20 +432,18 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
                    s.ToString().c_str());
   }
 
+  mutex_.Lock();
   // Check for shutdown again before scheduling further compactions,
   // since we released and re-acquired the lock above
   if (shutdown_initiated_) {
     s = Status::ShutdownInProgress();
   }
-  if (s.ok()) {
+  if (s.ok() && context.flush_after_recovery) {
     // Since we drop all non-recovery flush requests during recovery,
     // and new memtable may fill up during recovery,
     // schedule one more round of flush.
-    FlushOptions flush_opts;
-    flush_opts.allow_write_stall = false;
-    flush_opts.wait = false;
-    Status status = FlushAllColumnFamilies(
-        flush_opts, FlushReason::kCatchUpAfterErrorRecovery);
+    Status status = RetryFlushesForErrorRecovery(
+        FlushReason::kCatchUpAfterErrorRecovery, false /* wait */);
     if (!status.ok()) {
       // FlushAllColumnFamilies internally should take care of setting
       // background error if needed.
@@ -472,14 +454,16 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
     // FlushAllColumnFamilies releases and re-acquires mutex.
     if (shutdown_initiated_) {
       s = Status::ShutdownInProgress();
-    } else {
-      for (auto cfd : *versions_->GetColumnFamilySet()) {
-        SchedulePendingCompaction(cfd);
-      }
-      MaybeScheduleFlushOrCompaction();
     }
   }
 
+  if (s.ok()) {
+    for (auto cfd : *versions_->GetColumnFamilySet()) {
+      SchedulePendingCompaction(cfd);
+    }
+    MaybeScheduleFlushOrCompaction();
+  }
+
   // Wake up any waiters - in this case, it could be the shutdown thread
   bg_cv_.SignalAll();
 
@@ -811,42 +795,107 @@ Status DBImpl::StartPeriodicTaskScheduler() {
   return s;
 }
 
-Status DBImpl::RegisterRecordSeqnoTimeWorker() {
-  uint64_t min_time_duration = std::numeric_limits<uint64_t>::max();
-  uint64_t max_time_duration = std::numeric_limits<uint64_t>::min();
+Status DBImpl::RegisterRecordSeqnoTimeWorker(bool is_new_db) {
+  options_mutex_.AssertHeld();
+
+  uint64_t min_preserve_seconds = std::numeric_limits<uint64_t>::max();
+  uint64_t max_preserve_seconds = std::numeric_limits<uint64_t>::min();
+  bool mapping_was_empty = false;
   {
     InstrumentedMutexLock l(&mutex_);
 
     for (auto cfd : *versions_->GetColumnFamilySet()) {
       // preserve time is the max of 2 options.
-      uint64_t preserve_time_duration =
+      uint64_t preserve_seconds =
           std::max(cfd->ioptions()->preserve_internal_time_seconds,
                    cfd->ioptions()->preclude_last_level_data_seconds);
-      if (!cfd->IsDropped() && preserve_time_duration > 0) {
-        min_time_duration = std::min(preserve_time_duration, min_time_duration);
-        max_time_duration = std::max(preserve_time_duration, max_time_duration);
+      if (!cfd->IsDropped() && preserve_seconds > 0) {
+        min_preserve_seconds = std::min(preserve_seconds, min_preserve_seconds);
+        max_preserve_seconds = std::max(preserve_seconds, max_preserve_seconds);
       }
     }
-    if (min_time_duration == std::numeric_limits<uint64_t>::max()) {
-      seqno_time_mapping_.Resize(0, 0);
+    if (min_preserve_seconds == std::numeric_limits<uint64_t>::max()) {
+      seqno_to_time_mapping_.Resize(0, 0);
     } else {
-      seqno_time_mapping_.Resize(min_time_duration, max_time_duration);
+      seqno_to_time_mapping_.Resize(min_preserve_seconds, max_preserve_seconds);
     }
+    mapping_was_empty = seqno_to_time_mapping_.Empty();
   }
 
   uint64_t seqno_time_cadence = 0;
-  if (min_time_duration != std::numeric_limits<uint64_t>::max()) {
+  if (min_preserve_seconds != std::numeric_limits<uint64_t>::max()) {
     // round up to 1 when the time_duration is smaller than
     // kMaxSeqnoTimePairsPerCF
-    seqno_time_cadence =
-        (min_time_duration + SeqnoToTimeMapping::kMaxSeqnoTimePairsPerCF - 1) /
-        SeqnoToTimeMapping::kMaxSeqnoTimePairsPerCF;
+    seqno_time_cadence = (min_preserve_seconds +
+                          SeqnoToTimeMapping::kMaxSeqnoTimePairsPerCF - 1) /
+                         SeqnoToTimeMapping::kMaxSeqnoTimePairsPerCF;
   }
 
+  TEST_SYNC_POINT_CALLBACK(
+      "DBImpl::RegisterRecordSeqnoTimeWorker:BeforePeriodicTaskType", nullptr);
+
   Status s;
   if (seqno_time_cadence == 0) {
     s = periodic_task_scheduler_.Unregister(PeriodicTaskType::kRecordSeqnoTime);
   } else {
+    // Before registering the periodic task, we need to be sure to fulfill two
+    // promises:
+    // 1) Any DB created with preserve/preclude options set from the beginning
+    // will get pre-allocated seqnos with pre-populated time mappings back to
+    // the times we are interested in. (This will enable future import of data
+    // while preserving rough write time. We can only do this reliably from
+    // DB::Open, as otherwise there could be a race between CreateColumnFamily
+    // and the first Write to the DB, and seqno-to-time mappings need to be
+    // monotonic.
+    // 2) In any DB, any data written after setting preserve/preclude options
+    // must have a reasonable time estimate (so that we can accurately place
+    // the data), which means at least one entry in seqno_to_time_mapping_.
+    //
+    // FIXME: We don't currently guarantee that if the first column family with
+    // that setting is added or configured after initial DB::Open but before
+    // the first user Write. Fixing this causes complications with the crash
+    // test because if DB starts without preserve/preclude option, does some
+    // user writes but all those writes are lost in crash, then re-opens with
+    // preserve/preclude option, it sees seqno==1 which looks like one of the
+    // user writes was recovered, when actually it was not.
+    bool last_seqno_zero = GetLatestSequenceNumber() == 0;
+    assert(!is_new_db || last_seqno_zero);
+    if (is_new_db && last_seqno_zero) {
+      // Pre-allocate seqnos and pre-populate historical mapping
+      assert(mapping_was_empty);
+
+      // We can simply modify these, before writes are allowed
+      constexpr uint64_t kMax = SeqnoToTimeMapping::kMaxSeqnoTimePairsPerSST;
+      versions_->SetLastAllocatedSequence(kMax);
+      versions_->SetLastPublishedSequence(kMax);
+      versions_->SetLastSequence(kMax);
+
+      // And record in manifest, to avoid going backwards in seqno on re-open
+      // (potentially with different options). Concurrency is simple because we
+      // are in DB::Open
+      {
+        InstrumentedMutexLock l(&mutex_);
+        VersionEdit edit;
+        edit.SetLastSequence(kMax);
+        s = versions_->LogAndApplyToDefaultColumnFamily(
+            {}, &edit, &mutex_, directories_.GetDbDir());
+        if (!s.ok() && versions_->io_status().IsIOError()) {
+          s = error_handler_.SetBGError(versions_->io_status(),
+                                        BackgroundErrorReason::kManifestWrite);
+        }
+      }
+
+      // Pre-populate mappings for reserved sequence numbers.
+      RecordSeqnoToTimeMapping(max_preserve_seconds);
+    } else if (mapping_was_empty) {
+      if (!last_seqno_zero) {
+        // Ensure at least one mapping (or log a warning)
+        RecordSeqnoToTimeMapping(/*populate_historical_seconds=*/0);
+      } else {
+        // FIXME (see limitation described above)
+      }
+    }
+
     s = periodic_task_scheduler_.Register(
         PeriodicTaskType::kRecordSeqnoTime,
         periodic_task_functions_.at(PeriodicTaskType::kRecordSeqnoTime),
@@ -873,6 +922,7 @@ Status DBImpl::CancelPeriodicTaskScheduler() {
 
 // esitmate the total size of stats_history_
 size_t DBImpl::EstimateInMemoryStatsHistorySize() const {
+  stats_history_mutex_.AssertHeld();
   size_t size_total =
       sizeof(std::map<uint64_t, std::map<std::string, uint64_t>>);
   if (stats_history_.size() == 0) return size_total;
@@ -1163,6 +1213,7 @@ Status DBImpl::SetOptions(
     return Status::InvalidArgument("empty input");
   }
 
+  InstrumentedMutexLock ol(&options_mutex_);
   MutableCFOptions new_options;
   Status s;
   Status persist_options_status;
@@ -1182,8 +1233,7 @@ Status DBImpl::SetOptions(
       // thread.
       InstallSuperVersionAndScheduleWork(cfd, &sv_context, new_options);
 
-      persist_options_status = WriteOptionsFile(
-          false /*need_mutex_lock*/, true /*need_enter_write_thread*/);
+      persist_options_status = WriteOptionsFile(true /*db_mutex_already_held*/);
       bg_cv_.SignalAll();
     }
   }
@@ -1221,10 +1271,12 @@ Status DBImpl::SetDBOptions(
     return Status::InvalidArgument("empty input");
   }
 
+  InstrumentedMutexLock ol(&options_mutex_);
   MutableDBOptions new_options;
   Status s;
   Status persist_options_status = Status::OK();
-  bool wal_changed = false;
+  bool wal_size_option_changed = false;
+  bool wal_other_option_changed = false;
   WriteContext write_context;
   {
     InstrumentedMutexLock l(&mutex_);
@@ -1275,17 +1327,24 @@ Status DBImpl::SetDBOptions(
       const bool max_compactions_increased =
           new_bg_job_limits.max_compactions >
           current_bg_job_limits.max_compactions;
+      const bool offpeak_time_changed =
+          versions_->offpeak_time_option().daily_offpeak_time_utc !=
+          new_db_options.daily_offpeak_time_utc;
 
-      if (max_flushes_increased || max_compactions_increased) {
+      if (max_flushes_increased || max_compactions_increased ||
+          offpeak_time_changed) {
         if (max_flushes_increased) {
           env_->IncBackgroundThreadsIfNeeded(new_bg_job_limits.max_flushes,
                                              Env::Priority::HIGH);
         }
-
         if (max_compactions_increased) {
           env_->IncBackgroundThreadsIfNeeded(new_bg_job_limits.max_compactions,
                                              Env::Priority::LOW);
         }
+        if (offpeak_time_changed) {
+          versions_->ChangeOffpeakTimeOption(
+              new_db_options.daily_offpeak_time_utc);
+        }
 
         MaybeScheduleFlushOrCompaction();
       }
@@ -1325,8 +1384,10 @@ Status DBImpl::SetDBOptions(
       table_cache_.get()->SetCapacity(new_options.max_open_files == -1
                                           ? TableCache::kInfiniteCapacity
                                           : new_options.max_open_files - 10);
-      wal_changed = mutable_db_options_.wal_bytes_per_sync !=
-                    new_options.wal_bytes_per_sync;
+      wal_other_option_changed = mutable_db_options_.wal_bytes_per_sync !=
+                                 new_options.wal_bytes_per_sync;
+      wal_size_option_changed = mutable_db_options_.max_total_wal_size !=
+                                new_options.max_total_wal_size;
       mutable_db_options_ = new_options;
       file_options_for_compaction_ = FileOptions(new_db_options);
       file_options_for_compaction_ = fs_->OptimizeForCompactionTableWrite(
@@ -1337,19 +1398,21 @@ Status DBImpl::SetDBOptions(
           file_options_for_compaction_, immutable_db_options_);
       file_options_for_compaction_.compaction_readahead_size =
           mutable_db_options_.compaction_readahead_size;
-      WriteThread::Writer w;
-      write_thread_.EnterUnbatched(&w, &mutex_);
-      if (total_log_size_ > GetMaxTotalWalSize() || wal_changed) {
-        Status purge_wal_status = SwitchWAL(&write_context);
-        if (!purge_wal_status.ok()) {
-          ROCKS_LOG_WARN(immutable_db_options_.info_log,
-                         "Unable to purge WAL files in SetDBOptions() -- %s",
-                         purge_wal_status.ToString().c_str());
+      if (wal_other_option_changed || wal_size_option_changed) {
+        WriteThread::Writer w;
+        write_thread_.EnterUnbatched(&w, &mutex_);
+        if (wal_other_option_changed ||
+            total_log_size_ > GetMaxTotalWalSize()) {
+          Status purge_wal_status = SwitchWAL(&write_context);
+          if (!purge_wal_status.ok()) {
+            ROCKS_LOG_WARN(immutable_db_options_.info_log,
+                           "Unable to purge WAL files in SetDBOptions() -- %s",
+                           purge_wal_status.ToString().c_str());
+          }
         }
+        write_thread_.ExitUnbatched(&w);
       }
-      persist_options_status = WriteOptionsFile(
-          false /*need_mutex_lock*/, false /*need_enter_write_thread*/);
-      write_thread_.ExitUnbatched(&w);
+      persist_options_status = WriteOptionsFile(true /*db_mutex_already_held*/);
     } else {
       // To get here, we must have had invalid options and will not attempt to
       // persist the options, which means the status is "OK/Uninitialized.
@@ -1993,6 +2056,73 @@ Status DBImpl::GetEntity(const ReadOptions& _read_options,
   return GetImpl(read_options, key, get_impl_options);
 }
 
+Status DBImpl::GetEntity(const ReadOptions& _read_options, const Slice& key,
+                         PinnableAttributeGroups* result) {
+  if (!result) {
+    return Status::InvalidArgument(
+        "Cannot call GetEntity without PinnableAttributeGroups object");
+  }
+  Status s;
+  const size_t num_column_families = result->size();
+  if (_read_options.io_activity != Env::IOActivity::kUnknown &&
+      _read_options.io_activity != Env::IOActivity::kGetEntity) {
+    s = Status::InvalidArgument(
+        "Cannot call GetEntity with `ReadOptions::io_activity` != "
+        "`Env::IOActivity::kUnknown` or `Env::IOActivity::kGetEntity`");
+    for (size_t i = 0; i < num_column_families; ++i) {
+      (*result)[i].SetStatus(s);
+    }
+    return s;
+  }
+  // return early if no CF was passed in
+  if (num_column_families == 0) {
+    return s;
+  }
+  ReadOptions read_options(_read_options);
+  if (read_options.io_activity == Env::IOActivity::kUnknown) {
+    read_options.io_activity = Env::IOActivity::kGetEntity;
+  }
+  std::vector<Slice> keys;
+  std::vector<ColumnFamilyHandle*> column_families;
+  for (size_t i = 0; i < num_column_families; ++i) {
+    // If any of the CFH is null, break early since the entire query will fail
+    if (!(*result)[i].column_family()) {
+      s = Status::InvalidArgument(
+          "DB failed to query because one or more group(s) have null column "
+          "family handle");
+      (*result)[i].SetStatus(
+          Status::InvalidArgument("Column family handle cannot be null"));
+      break;
+    }
+    // Adding the same key slice for different CFs
+    keys.emplace_back(key);
+    column_families.emplace_back((*result)[i].column_family());
+  }
+  if (!s.ok()) {
+    for (size_t i = 0; i < num_column_families; ++i) {
+      if ((*result)[i].status().ok()) {
+        (*result)[i].SetStatus(
+            Status::Incomplete("DB not queried due to invalid argument(s) in "
+                               "one or more of the attribute groups"));
+      }
+    }
+    return s;
+  }
+  std::vector<PinnableWideColumns> columns(num_column_families);
+  std::vector<Status> statuses(num_column_families);
+  MultiGetCommon(
+      read_options, num_column_families, column_families.data(), keys.data(),
+      /* values */ nullptr, columns.data(),
+      /* timestamps */ nullptr, statuses.data(), /* sorted_input */ false);
+  // Set results
+  for (size_t i = 0; i < num_column_families; ++i) {
+    (*result)[i].Reset();
+    (*result)[i].SetStatus(statuses[i]);
+    (*result)[i].SetColumns(std::move(columns[i]));
+  }
+  return s;
+}
+
 bool DBImpl::ShouldReferenceSuperVersion(const MergeContext& merge_context) {
   // If both thresholds are reached, a function returning merge operands as
   // `PinnableSlice`s should reference the `SuperVersion` to avoid large and/or
@@ -2772,7 +2902,6 @@ void DBImpl::MultiGetCommon(const ReadOptions& read_options,
   bool should_fail = false;
   for (size_t i = 0; i < num_keys; ++i) {
     ColumnFamilyHandle* cfh = column_families[i];
-    assert(cfh);
     if (read_options.timestamp) {
       statuses[i] = FailIfTsMismatchCf(cfh, *(read_options.timestamp));
       if (!statuses[i].ok()) {
@@ -3292,14 +3421,86 @@ void DBImpl::MultiGetEntity(const ReadOptions& _read_options,
                  statuses, sorted_input);
 }
 
+void DBImpl::MultiGetEntity(const ReadOptions& _read_options, size_t num_keys,
+                            const Slice* keys,
+                            PinnableAttributeGroups* results) {
+  if (_read_options.io_activity != Env::IOActivity::kUnknown &&
+      _read_options.io_activity != Env::IOActivity::kMultiGetEntity) {
+    Status s = Status::InvalidArgument(
+        "Can only call MultiGetEntity with ReadOptions::io_activity` is "
+        "`Env::IOActivity::kUnknown` or `Env::IOActivity::kMultiGetEntity`");
+    for (size_t i = 0; i < num_keys; ++i) {
+      for (size_t j = 0; j < results[i].size(); ++j) {
+        results[i][j].SetStatus(s);
+      }
+    }
+    return;
+  }
+  ReadOptions read_options(_read_options);
+  if (read_options.io_activity == Env::IOActivity::kUnknown) {
+    read_options.io_activity = Env::IOActivity::kMultiGetEntity;
+  }
+
+  std::vector<ColumnFamilyHandle*> column_families;
+  std::vector<Slice> all_keys;
+  size_t total_count = 0;
+
+  for (size_t i = 0; i < num_keys; ++i) {
+    for (size_t j = 0; j < results[i].size(); ++j) {
+      // Adding the same key slice for different CFs
+      all_keys.emplace_back(keys[i]);
+      column_families.emplace_back(results[i][j].column_family());
+      ++total_count;
+    }
+  }
+  std::vector<Status> statuses(total_count);
+  std::vector<PinnableWideColumns> columns(total_count);
+  MultiGetCommon(read_options, total_count, column_families.data(),
+                 all_keys.data(),
+                 /* values */ nullptr, columns.data(),
+                 /* timestamps */ nullptr, statuses.data(),
+                 /* sorted_input */ false);
+
+  // Set results
+  size_t index = 0;
+  for (size_t i = 0; i < num_keys; ++i) {
+    for (size_t j = 0; j < results[i].size(); ++j) {
+      results[i][j].Reset();
+      results[i][j].SetStatus(std::move(statuses[index]));
+      results[i][j].SetColumns(std::move(columns[index]));
+      ++index;
+    }
+  }
+}
+
+Status DBImpl::WrapUpCreateColumnFamilies(
+    const std::vector<const ColumnFamilyOptions*>& cf_options) {
+  // NOTE: this function is skipped for create_missing_column_families and
+  // DB::Open, so new functionality here might need to go into Open also.
+  bool register_worker = false;
+  for (auto* opts_ptr : cf_options) {
+    if (opts_ptr->preserve_internal_time_seconds > 0 ||
+        opts_ptr->preclude_last_level_data_seconds > 0) {
+      register_worker = true;
+      break;
+    }
+  }
+  // Attempt both follow-up actions even if one fails
+  Status s = WriteOptionsFile(false /*db_mutex_already_held*/);
+  if (register_worker) {
+    s.UpdateIfOk(RegisterRecordSeqnoTimeWorker(/*from_db_open=*/false));
+  }
+  return s;
+}
+
 Status DBImpl::CreateColumnFamily(const ColumnFamilyOptions& cf_options,
                                   const std::string& column_family,
                                   ColumnFamilyHandle** handle) {
   assert(handle != nullptr);
+  InstrumentedMutexLock ol(&options_mutex_);
   Status s = CreateColumnFamilyImpl(cf_options, column_family, handle);
   if (s.ok()) {
-    s = WriteOptionsFile(true /*need_mutex_lock*/,
-                         true /*need_enter_write_thread*/);
+    s.UpdateIfOk(WrapUpCreateColumnFamilies({&cf_options}));
   }
   return s;
 }
@@ -3309,6 +3510,7 @@ Status DBImpl::CreateColumnFamilies(
     const std::vector<std::string>& column_family_names,
     std::vector<ColumnFamilyHandle*>* handles) {
   assert(handles != nullptr);
+  InstrumentedMutexLock ol(&options_mutex_);
   handles->clear();
   size_t num_cf = column_family_names.size();
   Status s;
@@ -3323,11 +3525,7 @@ Status DBImpl::CreateColumnFamilies(
     success_once = true;
   }
   if (success_once) {
-    Status persist_options_status = WriteOptionsFile(
-        true /*need_mutex_lock*/, true /*need_enter_write_thread*/);
-    if (s.ok() && !persist_options_status.ok()) {
-      s = persist_options_status;
-    }
+    s.UpdateIfOk(WrapUpCreateColumnFamilies({&cf_options}));
   }
   return s;
 }
@@ -3336,10 +3534,13 @@ Status DBImpl::CreateColumnFamilies(
     const std::vector<ColumnFamilyDescriptor>& column_families,
     std::vector<ColumnFamilyHandle*>* handles) {
   assert(handles != nullptr);
+  InstrumentedMutexLock ol(&options_mutex_);
   handles->clear();
   size_t num_cf = column_families.size();
   Status s;
   bool success_once = false;
+  std::vector<const ColumnFamilyOptions*> cf_opts;
+  cf_opts.reserve(num_cf);
   for (size_t i = 0; i < num_cf; i++) {
     ColumnFamilyHandle* handle;
     s = CreateColumnFamilyImpl(column_families[i].options,
@@ -3349,13 +3550,10 @@ Status DBImpl::CreateColumnFamilies(
     }
     handles->push_back(handle);
     success_once = true;
+    cf_opts.push_back(&column_families[i].options);
   }
   if (success_once) {
-    Status persist_options_status = WriteOptionsFile(
-        true /*need_mutex_lock*/, true /*need_enter_write_thread*/);
-    if (s.ok() && !persist_options_status.ok()) {
-      s = persist_options_status;
-    }
+    s.UpdateIfOk(WrapUpCreateColumnFamilies(cf_opts));
   }
   return s;
 }
@@ -3363,6 +3561,7 @@ Status DBImpl::CreateColumnFamilies(
 Status DBImpl::CreateColumnFamilyImpl(const ColumnFamilyOptions& cf_options,
                                       const std::string& column_family_name,
                                       ColumnFamilyHandle** handle) {
+  options_mutex_.AssertHeld();
   // TODO: plumb Env::IOActivity
   const ReadOptions read_options;
   Status s;
@@ -3443,10 +3642,6 @@ Status DBImpl::CreateColumnFamilyImpl(const ColumnFamilyOptions& cf_options,
     }
   }  // InstrumentedMutexLock l(&mutex_)
 
-  if (cf_options.preserve_internal_time_seconds > 0 ||
-      cf_options.preclude_last_level_data_seconds > 0) {
-    s = RegisterRecordSeqnoTimeWorker();
-  }
   sv_context.Clean();
   // this is outside the mutex
   if (s.ok()) {
@@ -3458,16 +3653,17 @@ Status DBImpl::CreateColumnFamilyImpl(const ColumnFamilyOptions& cf_options,
 
 Status DBImpl::DropColumnFamily(ColumnFamilyHandle* column_family) {
   assert(column_family != nullptr);
+  InstrumentedMutexLock ol(&options_mutex_);
   Status s = DropColumnFamilyImpl(column_family);
   if (s.ok()) {
-    s = WriteOptionsFile(true /*need_mutex_lock*/,
-                         true /*need_enter_write_thread*/);
+    s = WriteOptionsFile(false /*db_mutex_already_held*/);
   }
   return s;
 }
 
 Status DBImpl::DropColumnFamilies(
     const std::vector<ColumnFamilyHandle*>& column_families) {
+  InstrumentedMutexLock ol(&options_mutex_);
   Status s;
   bool success_once = false;
   for (auto* handle : column_families) {
@@ -3478,8 +3674,8 @@ Status DBImpl::DropColumnFamilies(
     success_once = true;
   }
   if (success_once) {
-    Status persist_options_status = WriteOptionsFile(
-        true /*need_mutex_lock*/, true /*need_enter_write_thread*/);
+    Status persist_options_status =
+        WriteOptionsFile(false /*db_mutex_already_held*/);
     if (s.ok() && !persist_options_status.ok()) {
       s = persist_options_status;
     }
@@ -3540,7 +3736,7 @@ Status DBImpl::DropColumnFamilyImpl(ColumnFamilyHandle* column_family) {
 
   if (cfd->ioptions()->preserve_internal_time_seconds > 0 ||
       cfd->ioptions()->preclude_last_level_data_seconds > 0) {
-    s = RegisterRecordSeqnoTimeWorker();
+    s = RegisterRecordSeqnoTimeWorker(/*from_db_open=*/false);
   }
 
   if (s.ok()) {
@@ -4047,7 +4243,8 @@ void DBImpl::ReleaseSnapshot(const Snapshot* s) {
       CfdList cf_scheduled;
       for (auto* cfd : *versions_->GetColumnFamilySet()) {
         if (!cfd->ioptions()->allow_ingest_behind) {
-          cfd->current()->storage_info()->UpdateOldestSnapshot(oldest_snapshot);
+          cfd->current()->storage_info()->UpdateOldestSnapshot(
+              oldest_snapshot, /*allow_ingest_behind=*/false);
           if (!cfd->current()
                    ->storage_info()
                    ->BottommostFilesMarkedForCompaction()
@@ -5106,16 +5303,13 @@ Status DestroyDB(const std::string& dbname, const Options& options,
   return result;
 }
 
-Status DBImpl::WriteOptionsFile(bool need_mutex_lock,
-                                bool need_enter_write_thread) {
-  WriteThread::Writer w;
-  if (need_mutex_lock) {
-    mutex_.Lock();
-  } else {
+Status DBImpl::WriteOptionsFile(bool db_mutex_already_held) {
+  options_mutex_.AssertHeld();
+
+  if (db_mutex_already_held) {
     mutex_.AssertHeld();
-  }
-  if (need_enter_write_thread) {
-    write_thread_.EnterUnbatched(&w, &mutex_);
+  } else {
+    mutex_.Lock();
   }
 
   std::vector<std::string> cf_names;
@@ -5130,10 +5324,10 @@ Status DBImpl::WriteOptionsFile(bool need_mutex_lock,
     cf_opts.push_back(cfd->GetLatestCFOptions());
   }
 
-  // Unlock during expensive operations.  New writes cannot get here
-  // because the single write thread ensures all new writes get queued.
   DBOptions db_options =
       BuildDBOptions(immutable_db_options_, mutable_db_options_);
+
+  // Unlock during expensive operations.
   mutex_.Unlock();
 
   TEST_SYNC_POINT("DBImpl::WriteOptionsFile:1");
@@ -5158,22 +5352,22 @@ Status DBImpl::WriteOptionsFile(bool need_mutex_lock,
     }
   }
 
-  // restore lock
-  if (!need_mutex_lock) {
-    mutex_.Lock();
-  }
-  if (need_enter_write_thread) {
-    write_thread_.ExitUnbatched(&w);
-  }
   if (!s.ok()) {
     ROCKS_LOG_WARN(immutable_db_options_.info_log,
                    "Unnable to persist options -- %s", s.ToString().c_str());
     if (immutable_db_options_.fail_if_options_file_error) {
-      return Status::IOError("Unable to persist options.",
-                             s.ToString().c_str());
+      s = Status::IOError("Unable to persist options.", s.ToString().c_str());
+    } else {
+      // Ignore error
+      s = Status::OK();
     }
   }
-  return Status::OK();
+
+  // Restore lock if appropriate
+  if (db_mutex_already_held) {
+    mutex_.Lock();
+  }
+  return s;
 }
 
 namespace {
@@ -6368,19 +6562,51 @@ Status DBImpl::GetCreationTimeOfOldestFile(uint64_t* creation_time) {
   }
 }
 
-void DBImpl::RecordSeqnoToTimeMapping() {
-  // Get time first then sequence number, so the actual time of seqno is <=
-  // unix_time recorded
-  int64_t unix_time = 0;
-  immutable_db_options_.clock->GetCurrentTime(&unix_time)
-      .PermitUncheckedError();  // Ignore error
+void DBImpl::RecordSeqnoToTimeMapping(uint64_t populate_historical_seconds) {
+  // TECHNICALITY: Sample last sequence number *before* time, as prescribed
+  // for SeqnoToTimeMapping. We don't know how long it has been since the last
+  // sequence number was written, so we at least have a one-sided bound by
+  // sampling in this order.
   SequenceNumber seqno = GetLatestSequenceNumber();
+  int64_t unix_time_signed = 0;
+  immutable_db_options_.clock->GetCurrentTime(&unix_time_signed)
+      .PermitUncheckedError();  // Ignore error
+  uint64_t unix_time = static_cast<uint64_t>(unix_time_signed);
   bool appended = false;
   {
     InstrumentedMutexLock l(&mutex_);
-    appended = seqno_time_mapping_.Append(seqno, unix_time);
+    if (populate_historical_seconds > 0) {
+      if (seqno > 1 && unix_time > populate_historical_seconds) {
+        // seqno=0 is reserved
+        SequenceNumber from_seqno = 1;
+        appended = seqno_to_time_mapping_.PrePopulate(
+            from_seqno, seqno, unix_time - populate_historical_seconds,
+            unix_time);
+      } else {
+        // One of these will fail
+        assert(seqno > 1);
+        assert(unix_time > populate_historical_seconds);
+      }
+    } else {
+      // FIXME: assert(seqno > 0);
+      appended = seqno_to_time_mapping_.Append(seqno, unix_time);
+    }
   }
-  if (!appended) {
+  if (populate_historical_seconds > 0) {
+    if (appended) {
+      ROCKS_LOG_INFO(
+          immutable_db_options_.info_log,
+          "Pre-populated sequence number to time entries: [1,%" PRIu64
+          "] -> [%" PRIu64 ",%" PRIu64 "]",
+          seqno, unix_time - populate_historical_seconds, unix_time);
+    } else {
+      ROCKS_LOG_WARN(
+          immutable_db_options_.info_log,
+          "Failed to pre-populate sequence number to time entries: [1,%" PRIu64
+          "] -> [%" PRIu64 ",%" PRIu64 "]",
+          seqno, unix_time - populate_historical_seconds, unix_time);
+    }
+  } else if (!appended) {
     ROCKS_LOG_WARN(immutable_db_options_.info_log,
                    "Failed to insert sequence number to time entry: %" PRIu64
                    " -> %" PRIu64,

db/db_impl/db_impl.h

@@ -197,6 +197,8 @@ class DBImpl : public DB {
   Status PutEntity(const WriteOptions& options,
                    ColumnFamilyHandle* column_family, const Slice& key,
                    const WideColumns& columns) override;
+  Status PutEntity(const WriteOptions& options, const Slice& key,
+                   const AttributeGroups& attribute_groups) override;
 
   using DB::Merge;
   Status Merge(const WriteOptions& options, ColumnFamilyHandle* column_family,
@@ -242,6 +244,8 @@ class DBImpl : public DB {
   Status GetEntity(const ReadOptions& options,
                    ColumnFamilyHandle* column_family, const Slice& key,
                    PinnableWideColumns* columns) override;
+  Status GetEntity(const ReadOptions& options, const Slice& key,
+                   PinnableAttributeGroups* result) override;
 
   using DB::GetMergeOperands;
   Status GetMergeOperands(const ReadOptions& options,
@@ -311,6 +315,9 @@ class DBImpl : public DB {
                       ColumnFamilyHandle** column_families, const Slice* keys,
                       PinnableWideColumns* results, Status* statuses,
                       bool sorted_input) override;
+  void MultiGetEntity(const ReadOptions& options, size_t num_keys,
+                      const Slice* keys,
+                      PinnableAttributeGroups* results) override;
 
   virtual Status CreateColumnFamily(const ColumnFamilyOptions& cf_options,
                                     const std::string& column_family,
@@ -1183,6 +1190,7 @@ class DBImpl : public DB {
   size_t TEST_GetWalPreallocateBlockSize(uint64_t write_buffer_size) const;
   void TEST_WaitForPeriodicTaskRun(std::function<void()> callback) const;
   SeqnoToTimeMapping TEST_GetSeqnoToTimeMapping() const;
+  const autovector<uint64_t>& TEST_GetFilesToQuarantine() const;
   size_t TEST_EstimateInMemoryStatsHistorySize() const;
 
   uint64_t TEST_GetCurrentLogNumber() const {
@@ -1197,6 +1205,10 @@ class DBImpl : public DB {
 
   const PeriodicTaskScheduler& TEST_GetPeriodicTaskScheduler() const;
 
+  static Status TEST_ValidateOptions(const DBOptions& db_options) {
+    return ValidateOptions(db_options);
+  }
+
 #endif  // NDEBUG
 
   // persist stats to column family "_persistent_stats"
@@ -1208,8 +1220,11 @@ class DBImpl : public DB {
   // flush LOG out of application buffer
   void FlushInfoLog();
 
-  // record current sequence number to time mapping
-  void RecordSeqnoToTimeMapping();
+  // record current sequence number to time mapping. If
+  // populate_historical_seconds > 0 then pre-populate all the
+  // sequence numbers from [1, last] to map to [now minus
+  // populate_historical_seconds, now].
+  void RecordSeqnoToTimeMapping(uint64_t populate_historical_seconds);
 
   // Interface to block and signal the DB in case of stalling writes by
   // WriteBufferManager. Each DBImpl object contains ptr to WBMStallInterface.
@@ -1383,15 +1398,15 @@ class DBImpl : public DB {
     autovector<ColumnFamilyData*> cfds_;
     autovector<const MutableCFOptions*> mutable_cf_opts_;
     autovector<autovector<VersionEdit*>> edit_lists_;
-    // files_to_delete_ contains sst files
-    std::unordered_set<std::string> files_to_delete_;
+    // Stale SST files to delete found upon recovery. This stores a mapping from
+    // such a file's absolute path to its parent directory.
+    std::unordered_map<std::string, std::string> files_to_delete_;
+    bool is_new_db_ = false;
   };
 
-  // Except in DB::Open(), WriteOptionsFile can only be called when:
-  // Persist options to options file.
-  // If need_mutex_lock = false, the method will lock DB mutex.
-  // If need_enter_write_thread = false, the method will enter write thread.
-  Status WriteOptionsFile(bool need_mutex_lock, bool need_enter_write_thread);
+  // Persist options to options file. Must be holding options_mutex_.
+  // Will lock DB mutex if !db_mutex_already_held.
+  Status WriteOptionsFile(bool db_mutex_already_held);
 
   Status CompactRangeInternal(const CompactRangeOptions& options,
                               ColumnFamilyHandle* column_family,
@@ -1819,10 +1834,15 @@ class DBImpl : public DB {
 
   const Status CreateArchivalDirectory();
 
+  // Create a column family, without some of the follow-up work yet
   Status CreateColumnFamilyImpl(const ColumnFamilyOptions& cf_options,
                                 const std::string& cf_name,
                                 ColumnFamilyHandle** handle);
 
+  // Follow-up work to user creating a column family or (families)
+  Status WrapUpCreateColumnFamilies(
+      const std::vector<const ColumnFamilyOptions*>& cf_options);
+
   Status DropColumnFamilyImpl(ColumnFamilyHandle* column_family);
 
   // Delete any unneeded files and stale in-memory entries.
@@ -1852,7 +1872,8 @@ class DBImpl : public DB {
   void ReleaseFileNumberFromPendingOutputs(
       std::unique_ptr<std::list<uint64_t>::iterator>& v);
 
-  IOStatus SyncClosedLogs(JobContext* job_context, VersionEdit* synced_wals);
+  IOStatus SyncClosedLogs(JobContext* job_context, VersionEdit* synced_wals,
+                          bool error_recovery_in_prog);
 
   // Flush the in-memory write buffer to storage.  Switches to a new
   // log-file/memtable and writes a new descriptor iff successful. Then
@@ -1947,6 +1968,8 @@ class DBImpl : public DB {
       const autovector<ColumnFamilyData*>& provided_candidate_cfds = {},
       bool entered_write_thread = false);
 
+  Status RetryFlushesForErrorRecovery(FlushReason flush_reason, bool wait);
+
   // Wait until flushing this column family won't stall writes
   Status WaitUntilFlushWouldNotStallWrites(ColumnFamilyData* cfd,
                                            bool* flush_needed);
@@ -2095,6 +2118,12 @@ class DBImpl : public DB {
 #endif /* !NDEBUG */
   };
 
+  // In case of atomic flush, generates a `FlushRequest` for the latest atomic
+  // cuts for these `cfds`. Atomic cuts are recorded in
+  // `AssignAtomicFlushSeq()`. For each entry in `cfds`, all CFDs sharing the
+  // same latest atomic cut must also be present.
+  //
+  // REQUIRES: mutex held
   void GenerateFlushRequest(const autovector<ColumnFamilyData*>& cfds,
                             FlushReason flush_reason, FlushRequest* req);
 
@@ -2146,7 +2175,7 @@ class DBImpl : public DB {
   // Cancel scheduled periodic tasks
   Status CancelPeriodicTaskScheduler();
 
-  Status RegisterRecordSeqnoTimeWorker();
+  Status RegisterRecordSeqnoTimeWorker(bool is_new_db);
 
   void PrintStatistics();
 
@@ -2226,6 +2255,7 @@ class DBImpl : public DB {
   bool ShouldntRunManualCompaction(ManualCompactionState* m);
   bool HaveManualCompaction(ColumnFamilyData* cfd);
   bool MCOverlap(ManualCompactionState* m, ManualCompactionState* m1);
+  void UpdateDeletionCompactionStats(const std::unique_ptr<Compaction>& c);
   void BuildCompactionJobInfo(const ColumnFamilyData* cfd, Compaction* c,
                               const Status& st,
                               const CompactionJobStats& compaction_job_stats,
@@ -2359,9 +2389,19 @@ class DBImpl : public DB {
   // Lock over the persistent DB state.  Non-nullptr iff successfully acquired.
   FileLock* db_lock_;
 
-  // In addition to mutex_, log_write_mutex_ protected writes to stats_history_
+  // Guards changes to DB and CF options to ensure consistency between
+  // * In-memory options objects
+  // * Settings in effect
+  // * Options file contents
+  // while allowing the DB mutex to be released during slow operations like
+  // persisting options file or modifying global periodic task timer.
+  // Always acquired *before* DB mutex when this one is applicable.
+  InstrumentedMutex options_mutex_;
+
+  // Guards reads and writes to in-memory stats_history_.
   InstrumentedMutex stats_history_mutex_;
-  // In addition to mutex_, log_write_mutex_ protected writes to logs_ and
+
+  // In addition to mutex_, log_write_mutex_ protects writes to logs_ and
   // logfile_number_. With two_write_queues it also protects alive_log_files_,
   // and log_empty_. Refer to the definition of each variable below for more
   // details.
@@ -2748,9 +2788,9 @@ class DBImpl : public DB {
   // Pointer to WriteBufferManager stalling interface.
   std::unique_ptr<StallInterface> wbm_stall_;
 
-  // seqno_time_mapping_ stores the sequence number to time mapping, it's not
+  // seqno_to_time_mapping_ stores the sequence number to time mapping, it's not
   // thread safe, both read and write need db mutex hold.
-  SeqnoToTimeMapping seqno_time_mapping_;
+  SeqnoToTimeMapping seqno_to_time_mapping_;
 
   // Stop write token that is acquired when first LockWAL() is called.
   // Destroyed when last UnlockWAL() is called. Controlled by DB mutex.
@@ -2830,7 +2870,9 @@ static void ClipToRange(T* ptr, V minvalue, V maxvalue) {
 
 inline Status DBImpl::FailIfCfHasTs(
     const ColumnFamilyHandle* column_family) const {
-  column_family = column_family ? column_family : DefaultColumnFamily();
+  if (!column_family) {
+    return Status::InvalidArgument("column family handle cannot be null");
+  }
   assert(column_family);
   const Comparator* const ucmp = column_family->GetComparator();
   assert(ucmp);

db/db_impl/db_impl_compaction_flush.cc

@@ -113,7 +113,8 @@ bool DBImpl::ShouldRescheduleFlushRequestToRetainUDT(
 }
 
 IOStatus DBImpl::SyncClosedLogs(JobContext* job_context,
-                                VersionEdit* synced_wals) {
+                                VersionEdit* synced_wals,
+                                bool error_recovery_in_prog) {
   TEST_SYNC_POINT("DBImpl::SyncClosedLogs:Start");
   InstrumentedMutexLock l(&log_write_mutex_);
   autovector<log::Writer*, 1> logs_to_sync;
@@ -139,7 +140,7 @@ IOStatus DBImpl::SyncClosedLogs(JobContext* job_context,
       ROCKS_LOG_INFO(immutable_db_options_.info_log,
                      "[JOB %d] Syncing log #%" PRIu64, job_context->job_id,
                      log->get_log_number());
-      if (error_handler_.IsRecoveryInProgress()) {
+      if (error_recovery_in_prog) {
         log->file()->reset_seen_error();
       }
       io_s = log->file()->Sync(immutable_db_options_.use_fsync);
@@ -148,7 +149,7 @@ IOStatus DBImpl::SyncClosedLogs(JobContext* job_context,
       }
 
       if (immutable_db_options_.recycle_log_file_num > 0) {
-        if (error_handler_.IsRecoveryInProgress()) {
+        if (error_recovery_in_prog) {
           log->file()->reset_seen_error();
         }
         io_s = log->Close();
@@ -222,9 +223,10 @@ Status DBImpl::FlushMemTableToOutputFile(
   // `snapshot_seqs` has already been computed before this function starts.
   // Recording the max memtable ID ensures that the flush job does not flush
   // a memtable without knowing such snapshot(s).
-  uint64_t max_memtable_id = needs_to_sync_closed_wals
-                                 ? cfd->imm()->GetLatestMemTableID()
-                                 : std::numeric_limits<uint64_t>::max();
+  uint64_t max_memtable_id =
+      needs_to_sync_closed_wals
+          ? cfd->imm()->GetLatestMemTableID(false /* for_atomic_flush */)
+          : std::numeric_limits<uint64_t>::max();
 
   // If needs_to_sync_closed_wals is false, then the flush job will pick ALL
   // existing memtables of the column family when PickMemTable() is called
@@ -233,7 +235,7 @@ Status DBImpl::FlushMemTableToOutputFile(
   // releases and re-acquires the db mutex. In the meantime, the application
   // can still insert into the memtables and increase the db's sequence number.
   // The application can take a snapshot, hoping that the latest visible state
-  // to this snapshto is preserved. This is hard to guarantee since db mutex
+  // to this snapshot is preserved. This is hard to guarantee since db mutex
   // not held. This newly-created snapshot is not included in `snapshot_seqs`
   // and the flush job is unaware of its presence. Consequently, the flush job
   // may drop certain keys when generating the L0, causing incorrect data to be
@@ -250,7 +252,7 @@ Status DBImpl::FlushMemTableToOutputFile(
       GetCompressionFlush(*cfd->ioptions(), mutable_cf_options), stats_,
       &event_logger_, mutable_cf_options.report_bg_io_stats,
       true /* sync_output_directory */, true /* write_manifest */, thread_pri,
-      io_tracer_, seqno_time_mapping_, db_id_, db_session_id_,
+      io_tracer_, seqno_to_time_mapping_, db_id_, db_session_id_,
       cfd->GetFullHistoryTsLow(), &blob_callback_);
   FileMetaData file_meta;
 
@@ -261,8 +263,10 @@ Status DBImpl::FlushMemTableToOutputFile(
     // SyncClosedLogs() may unlock and re-lock the log_write_mutex multiple
     // times.
     VersionEdit synced_wals;
+    bool error_recovery_in_prog = error_handler_.IsRecoveryInProgress();
     mutex_.Unlock();
-    log_io_s = SyncClosedLogs(job_context, &synced_wals);
+    log_io_s =
+        SyncClosedLogs(job_context, &synced_wals, error_recovery_in_prog);
     mutex_.Lock();
     if (log_io_s.ok() && synced_wals.IsWalAddition()) {
       const ReadOptions read_options(Env::IOActivity::kFlush);
@@ -522,7 +526,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
         GetCompressionFlush(*cfd->ioptions(), mutable_cf_options), stats_,
         &event_logger_, mutable_cf_options.report_bg_io_stats,
         false /* sync_output_directory */, false /* write_manifest */,
-        thread_pri, io_tracer_, seqno_time_mapping_, db_id_, db_session_id_,
+        thread_pri, io_tracer_, seqno_to_time_mapping_, db_id_, db_session_id_,
         cfd->GetFullHistoryTsLow(), &blob_callback_));
   }
 
@@ -546,8 +550,10 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     // TODO (yanqin) investigate whether we should sync the closed logs for
     // single column family case.
     VersionEdit synced_wals;
+    bool error_recovery_in_prog = error_handler_.IsRecoveryInProgress();
     mutex_.Unlock();
-    log_io_s = SyncClosedLogs(job_context, &synced_wals);
+    log_io_s =
+        SyncClosedLogs(job_context, &synced_wals, error_recovery_in_prog);
     mutex_.Lock();
     if (log_io_s.ok() && synced_wals.IsWalAddition()) {
       const ReadOptions read_options(Env::IOActivity::kFlush);
@@ -2230,7 +2236,8 @@ void DBImpl::GenerateFlushRequest(const autovector<ColumnFamilyData*>& cfds,
       // cfd may be null, see DBImpl::ScheduleFlushes
       continue;
     }
-    uint64_t max_memtable_id = cfd->imm()->GetLatestMemTableID();
+    uint64_t max_memtable_id = cfd->imm()->GetLatestMemTableID(
+        immutable_db_options_.atomic_flush /* for_atomic_flush */);
     req->cfd_to_max_mem_id_to_persist.emplace(cfd, max_memtable_id);
   }
 }
@@ -2274,19 +2281,7 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
     }
     WaitForPendingWrites();
 
-    if (flush_reason != FlushReason::kErrorRecoveryRetryFlush &&
-        flush_reason != FlushReason::kCatchUpAfterErrorRecovery &&
-        (!cfd->mem()->IsEmpty() || !cached_recoverable_state_empty_.load())) {
-      // Note that, when flush reason is kErrorRecoveryRetryFlush, during the
-      // auto retry resume, we want to avoid creating new small memtables.
-      // If flush reason is kCatchUpAfterErrorRecovery, we try to flush any new
-      // memtable that filled up during recovery, and we also want to avoid
-      // switching memtable to create small memtables.
-      // Therefore, SwitchMemtable will not be called. Also, since ResumeImpl
-      // will iterate through all the CFs and call FlushMemtable during auto
-      // retry resume, it is possible that in some CFs,
-      // cfd->imm()->NumNotFlushed() = 0. In this case, so no flush request will
-      // be created and scheduled, status::OK() will be returned.
+    if (!cfd->mem()->IsEmpty() || !cached_recoverable_state_empty_.load()) {
       s = SwitchMemtable(cfd, &context);
     }
     const uint64_t flush_memtable_id = std::numeric_limits<uint64_t>::max();
@@ -2295,10 +2290,10 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
           !cached_recoverable_state_empty_.load()) {
         FlushRequest req{flush_reason, {{cfd, flush_memtable_id}}};
         flush_reqs.emplace_back(std::move(req));
-        memtable_ids_to_wait.emplace_back(cfd->imm()->GetLatestMemTableID());
+        memtable_ids_to_wait.emplace_back(
+            cfd->imm()->GetLatestMemTableID(false /* for_atomic_flush */));
       }
-      if (immutable_db_options_.persist_stats_to_disk &&
-          flush_reason != FlushReason::kErrorRecoveryRetryFlush) {
+      if (immutable_db_options_.persist_stats_to_disk) {
         ColumnFamilyData* cfd_stats =
             versions_->GetColumnFamilySet()->GetColumnFamily(
                 kPersistentStatsColumnFamilyName);
@@ -2324,7 +2319,8 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
             FlushRequest req{flush_reason, {{cfd_stats, flush_memtable_id}}};
             flush_reqs.emplace_back(std::move(req));
             memtable_ids_to_wait.emplace_back(
-                cfd_stats->imm()->GetLatestMemTableID());
+                cfd_stats->imm()->GetLatestMemTableID(
+                    false /* for_atomic_flush */));
           }
         }
       }
@@ -2375,8 +2371,7 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
     }
     s = WaitForFlushMemTables(
         cfds, flush_memtable_ids,
-        (flush_reason == FlushReason::kErrorRecovery ||
-         flush_reason == FlushReason::kErrorRecoveryRetryFlush));
+        flush_reason == FlushReason::kErrorRecovery /* resuming_from_bg_err */);
     InstrumentedMutexLock lock_guard(&mutex_);
     for (auto* tmp_cfd : cfds) {
       tmp_cfd->UnrefAndTryDelete();
@@ -2471,9 +2466,7 @@ Status DBImpl::AtomicFlushMemTables(
     }
 
     for (auto cfd : cfds) {
-      if ((cfd->mem()->IsEmpty() && cached_recoverable_state_empty_.load()) ||
-          flush_reason == FlushReason::kErrorRecoveryRetryFlush ||
-          flush_reason == FlushReason::kCatchUpAfterErrorRecovery) {
+      if (cfd->mem()->IsEmpty() && cached_recoverable_state_empty_.load()) {
         continue;
       }
       cfd->Ref();
@@ -2518,8 +2511,7 @@ Status DBImpl::AtomicFlushMemTables(
     }
     s = WaitForFlushMemTables(
         cfds, flush_memtable_ids,
-        (flush_reason == FlushReason::kErrorRecovery ||
-         flush_reason == FlushReason::kErrorRecoveryRetryFlush));
+        flush_reason == FlushReason::kErrorRecovery /* resuming_from_bg_err */);
     InstrumentedMutexLock lock_guard(&mutex_);
     for (auto* cfd : cfds) {
       cfd->UnrefAndTryDelete();
@@ -2528,6 +2520,68 @@ Status DBImpl::AtomicFlushMemTables(
   return s;
 }
 
+Status DBImpl::RetryFlushesForErrorRecovery(FlushReason flush_reason,
+                                            bool wait) {
+  mutex_.AssertHeld();
+  assert(flush_reason == FlushReason::kErrorRecoveryRetryFlush ||
+         flush_reason == FlushReason::kCatchUpAfterErrorRecovery);
+
+  // Collect referenced CFDs.
+  autovector<ColumnFamilyData*> cfds;
+  for (ColumnFamilyData* cfd : *versions_->GetColumnFamilySet()) {
+    if (!cfd->IsDropped() && cfd->initialized() &&
+        cfd->imm()->NumNotFlushed() != 0) {
+      cfd->Ref();
+      cfd->imm()->FlushRequested();
+      cfds.push_back(cfd);
+    }
+  }
+
+  // Submit flush requests for all immutable memtables needing flush.
+  // `flush_memtable_ids` will be populated such that all immutable
+  // memtables eligible for flush are waited on before this function
+  // returns.
+  autovector<uint64_t> flush_memtable_ids;
+  if (immutable_db_options_.atomic_flush) {
+    FlushRequest flush_req;
+    GenerateFlushRequest(cfds, flush_reason, &flush_req);
+    SchedulePendingFlush(flush_req);
+    for (auto& iter : flush_req.cfd_to_max_mem_id_to_persist) {
+      flush_memtable_ids.push_back(iter.second);
+    }
+  } else {
+    for (auto cfd : cfds) {
+      flush_memtable_ids.push_back(
+          cfd->imm()->GetLatestMemTableID(false /* for_atomic_flush */));
+      // Impose no bound on the highest memtable ID flushed. There is no
+      // reason to do so outside of atomic flush.
+      FlushRequest flush_req{
+          flush_reason,
+          {{cfd,
+            std::numeric_limits<uint64_t>::max() /* max_mem_id_to_persist */}}};
+      SchedulePendingFlush(flush_req);
+    }
+  }
+  MaybeScheduleFlushOrCompaction();
+
+  Status s;
+  if (wait) {
+    mutex_.Unlock();
+    autovector<const uint64_t*> flush_memtable_id_ptrs;
+    for (auto& flush_memtable_id : flush_memtable_ids) {
+      flush_memtable_id_ptrs.push_back(&flush_memtable_id);
+    }
+    s = WaitForFlushMemTables(cfds, flush_memtable_id_ptrs,
+                              true /* resuming_from_bg_err */);
+    mutex_.Lock();
+  }
+
+  for (auto* cfd : cfds) {
+    cfd->UnrefAndTryDelete();
+  }
+  return s;
+}
+
 // Calling FlushMemTable(), whether from DB::Flush() or from Backup Engine, can
 // cause write stall, for example if one memtable is being flushed already.
 // This method tries to avoid write stall (similar to CompactRange() behavior)
@@ -2729,6 +2783,7 @@ void DBImpl::EnableManualCompaction() {
 
 void DBImpl::MaybeScheduleFlushOrCompaction() {
   mutex_.AssertHeld();
+  TEST_SYNC_POINT("DBImpl::MaybeScheduleFlushOrCompaction:Start");
   if (!opened_successfully_) {
     // Compaction may introduce data race to DB open
     return;
@@ -3650,6 +3705,9 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
     ROCKS_LOG_BUFFER(log_buffer, "[%s] Deleted %d files\n",
                      c->column_family_data()->GetName().c_str(),
                      c->num_input_files(0));
+    if (status.ok() && io_s.ok()) {
+      UpdateDeletionCompactionStats(c);
+    }
     *made_progress = true;
     TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:AfterCompaction",
                              c->column_family_data());
@@ -4028,6 +4086,27 @@ bool DBImpl::MCOverlap(ManualCompactionState* m, ManualCompactionState* m1) {
   return false;
 }
 
+void DBImpl::UpdateDeletionCompactionStats(
+    const std::unique_ptr<Compaction>& c) {
+  if (c == nullptr) {
+    return;
+  }
+
+  CompactionReason reason = c->compaction_reason();
+
+  switch (reason) {
+    case CompactionReason::kFIFOMaxSize:
+      RecordTick(stats_, FIFO_MAX_SIZE_COMPACTIONS);
+      break;
+    case CompactionReason::kFIFOTtl:
+      RecordTick(stats_, FIFO_TTL_COMPACTIONS);
+      break;
+    default:
+      assert(false);
+      break;
+  }
+}
+
 void DBImpl::BuildCompactionJobInfo(
     const ColumnFamilyData* cfd, Compaction* c, const Status& st,
     const CompactionJobStats& compaction_job_stats, const int job_id,

db/db_impl/db_impl_debug.cc

@@ -306,11 +306,16 @@ const PeriodicTaskScheduler& DBImpl::TEST_GetPeriodicTaskScheduler() const {
 
 SeqnoToTimeMapping DBImpl::TEST_GetSeqnoToTimeMapping() const {
   InstrumentedMutexLock l(&mutex_);
-  return seqno_time_mapping_;
+  return seqno_to_time_mapping_;
 }
 
+const autovector<uint64_t>& DBImpl::TEST_GetFilesToQuarantine() const {
+  InstrumentedMutexLock l(&mutex_);
+  return error_handler_.GetFilesToQuarantine();
+}
 
 size_t DBImpl::TEST_EstimateInMemoryStatsHistorySize() const {
+  InstrumentedMutexLock l(&const_cast<DBImpl*>(this)->stats_history_mutex_);
   return EstimateInMemoryStatsHistorySize();
 }
 }  // namespace ROCKSDB_NAMESPACE

db/db_impl/db_impl_files.cc

@@ -146,6 +146,7 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,
   // mutex_ cannot be released. Otherwise, we might see no min_pending_output
   // here but later find newer generated unfinalized files while scanning.
   job_context->min_pending_output = MinObsoleteSstNumberToKeep();
+  job_context->files_to_quarantine = error_handler_.GetFilesToQuarantine();
 
   // Get obsolete files.  This function will also update the list of
   // pending files in VersionSet().
@@ -419,6 +420,8 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
                                              state.blob_live.end());
   std::unordered_set<uint64_t> log_recycle_files_set(
       state.log_recycle_files.begin(), state.log_recycle_files.end());
+  std::unordered_set<uint64_t> quarantine_files_set(
+      state.files_to_quarantine.begin(), state.files_to_quarantine.end());
 
   auto candidate_files = state.full_scan_candidate_files;
   candidate_files.reserve(
@@ -522,6 +525,10 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
       continue;
     }
 
+    if (quarantine_files_set.find(number) != quarantine_files_set.end()) {
+      continue;
+    }
+
     bool keep = true;
     switch (type) {
       case kWalFile:
@@ -995,7 +1002,7 @@ Status DBImpl::DeleteUnreferencedSstFiles(RecoveryContext* recovery_ctx) {
       if (type == kTableFile && number >= next_file_number &&
           recovery_ctx->files_to_delete_.find(normalized_fpath) ==
               recovery_ctx->files_to_delete_.end()) {
-        recovery_ctx->files_to_delete_.emplace(normalized_fpath);
+        recovery_ctx->files_to_delete_.emplace(normalized_fpath, path);
       }
     }
   }

db/db_impl/db_impl_open.cc

@@ -25,6 +25,7 @@
 #include "rocksdb/wal_filter.h"
 #include "test_util/sync_point.h"
 #include "util/rate_limiter_impl.h"
+#include "util/string_util.h"
 #include "util/udt_util.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -291,6 +292,18 @@ Status DBImpl::ValidateOptions(const DBOptions& db_options) {
         "writes in direct IO require writable_file_max_buffer_size > 0");
   }
 
+  if (db_options.daily_offpeak_time_utc != "") {
+    int start_time, end_time;
+    if (!TryParseTimeRangeString(db_options.daily_offpeak_time_utc, start_time,
+                                 end_time)) {
+      return Status::InvalidArgument(
+          "daily_offpeak_time_utc should be set in the format HH:mm-HH:mm "
+          "(e.g. 04:30-07:30)");
+    } else if (start_time == end_time) {
+      return Status::InvalidArgument(
+          "start_time and end_time cannot be the same");
+    }
+  }
   return Status::OK();
 }
 
@@ -405,7 +418,8 @@ Status DBImpl::Recover(
     uint64_t* recovered_seq, RecoveryContext* recovery_ctx) {
   mutex_.AssertHeld();
 
-  bool is_new_db = false;
+  bool tmp_is_new_db = false;
+  bool& is_new_db = recovery_ctx ? recovery_ctx->is_new_db_ : tmp_is_new_db;
   assert(db_lock_ == nullptr);
   std::vector<std::string> files_in_dbname;
   if (!read_only) {
@@ -858,7 +872,8 @@ Status DBImpl::PersistentStatsProcessFormatVersion() {
       if (s.ok()) {
         ColumnFamilyOptions cfo;
         OptimizeForPersistentStats(&cfo);
-        s = CreateColumnFamily(cfo, kPersistentStatsColumnFamilyName, &handle);
+        s = CreateColumnFamilyImpl(cfo, kPersistentStatsColumnFamilyName,
+                                   &handle);
       }
       if (s.ok()) {
         persist_stats_cf_handle_ = static_cast<ColumnFamilyHandleImpl*>(handle);
@@ -911,7 +926,7 @@ Status DBImpl::InitPersistStatsColumnFamily() {
     ColumnFamilyHandle* handle = nullptr;
     ColumnFamilyOptions cfo;
     OptimizeForPersistentStats(&cfo);
-    s = CreateColumnFamily(cfo, kPersistentStatsColumnFamilyName, &handle);
+    s = CreateColumnFamilyImpl(cfo, kPersistentStatsColumnFamilyName, &handle);
     persist_stats_cf_handle_ = static_cast<ColumnFamilyHandleImpl*>(handle);
     mutex_.Lock();
   }
@@ -927,8 +942,11 @@ Status DBImpl::LogAndApplyForRecovery(const RecoveryContext& recovery_ctx) {
       recovery_ctx.edit_lists_, &mutex_, directories_.GetDbDir());
   if (s.ok() && !(recovery_ctx.files_to_delete_.empty())) {
     mutex_.Unlock();
-    for (const auto& fname : recovery_ctx.files_to_delete_) {
-      s = env_->DeleteFile(fname);
+    for (const auto& stale_sst_file : recovery_ctx.files_to_delete_) {
+      s = DeleteDBFile(&immutable_db_options_, stale_sst_file.first,
+                       stale_sst_file.second,
+                       /*force_bg=*/false,
+                       /*force_fg=*/false);
       if (!s.ok()) {
         break;
       }
@@ -1233,6 +1251,13 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& wal_numbers,
       }
 
       SequenceNumber sequence = WriteBatchInternal::Sequence(batch_to_use);
+      if (sequence > kMaxSequenceNumber) {
+        reporter.Corruption(
+            record.size(),
+            Status::Corruption("sequence " + std::to_string(sequence) +
+                               " is too large"));
+        continue;
+      }
 
       if (immutable_db_options_.wal_recovery_mode ==
           WALRecoveryMode::kPointInTimeRecovery) {
@@ -1649,7 +1674,7 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
           TableFileCreationReason::kRecovery, 0 /* oldest_key_time */,
           0 /* file_creation_time */, db_id_, db_session_id_,
           0 /* target_file_size */, meta.fd.GetNumber());
-      SeqnoToTimeMapping empty_seqno_time_mapping;
+      SeqnoToTimeMapping empty_seqno_to_time_mapping;
       Version* version = cfd->current();
       version->Ref();
       const ReadOptions read_option(Env::IOActivity::kDBOpen);
@@ -1661,7 +1686,7 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
           snapshot_seqs, earliest_write_conflict_snapshot, kMaxSequenceNumber,
           snapshot_checker, paranoid_file_checks, cfd->internal_stats(), &io_s,
           io_tracer_, BlobFileCreationReason::kRecovery,
-          empty_seqno_time_mapping, &event_logger_, job_id, Env::IO_HIGH,
+          empty_seqno_to_time_mapping, &event_logger_, job_id, Env::IO_HIGH,
           nullptr /* table_properties */, write_hint,
           nullptr /*full_history_ts_low*/, &blob_callback_, version,
           &num_input_entries);
@@ -1975,6 +2000,7 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
 
   impl->wal_in_db_path_ = impl->immutable_db_options_.IsWalDirSameAsDBPath();
   RecoveryContext recovery_ctx;
+  impl->options_mutex_.Lock();
   impl->mutex_.Lock();
 
   // Handles create_if_missing, error_if_exists
@@ -2056,7 +2082,9 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
           // missing column family, create it
           ColumnFamilyHandle* handle = nullptr;
           impl->mutex_.Unlock();
-          s = impl->CreateColumnFamily(cf.options, cf.name, &handle);
+          // NOTE: the work normally done in WrapUpCreateColumnFamilies will
+          // be done separately below.
+          s = impl->CreateColumnFamilyImpl(cf.options, cf.name, &handle);
           impl->mutex_.Lock();
           if (s.ok()) {
             handles->push_back(handle);
@@ -2107,9 +2135,8 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
   if (s.ok()) {
     // Persist RocksDB Options before scheduling the compaction.
     // The WriteOptionsFile() will release and lock the mutex internally.
-    persist_options_status = impl->WriteOptionsFile(
-        false /*need_mutex_lock*/, false /*need_enter_write_thread*/);
-
+    persist_options_status =
+        impl->WriteOptionsFile(true /*db_mutex_already_held*/);
     *dbptr = impl;
     impl->opened_successfully_ = true;
     impl->DeleteObsoleteFiles();
@@ -2230,10 +2257,10 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
   if (s.ok()) {
     s = impl->StartPeriodicTaskScheduler();
   }
-
   if (s.ok()) {
-    s = impl->RegisterRecordSeqnoTimeWorker();
+    s = impl->RegisterRecordSeqnoTimeWorker(recovery_ctx.is_new_db_);
   }
+  impl->options_mutex_.Unlock();
   if (!s.ok()) {
     for (auto* h : *handles) {
       delete h;

db/db_impl/db_impl_readonly.h

@@ -53,6 +53,10 @@ class DBImplReadOnly : public DBImpl {
                    const WideColumns& /* columns */) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
+  Status PutEntity(const WriteOptions& /* options */, const Slice& /* key */,
+                   const AttributeGroups& /* attribute_groups */) override {
+    return Status::NotSupported("Not supported operation in read only mode.");
+  }
 
   using DBImpl::Merge;
   virtual Status Merge(const WriteOptions& /*options*/,

db/db_impl/db_impl_secondary.h

@@ -136,6 +136,10 @@ class DBImplSecondary : public DBImpl {
                    const WideColumns& /* columns */) override {
     return Status::NotSupported("Not supported operation in secondary mode.");
   }
+  Status PutEntity(const WriteOptions& /* options */, const Slice& /* key */,
+                   const AttributeGroups& /* attribute_groups */) override {
+    return Status::NotSupported("Not supported operation in secondary mode.");
+  }
 
   using DBImpl::Merge;
   Status Merge(const WriteOptions& /*options*/,

db/db_impl/db_impl_write.cc

@@ -48,6 +48,17 @@ Status DBImpl::PutEntity(const WriteOptions& options,
   return DB::PutEntity(options, column_family, key, columns);
 }
 
+Status DBImpl::PutEntity(const WriteOptions& options, const Slice& key,
+                         const AttributeGroups& attribute_groups) {
+  for (const AttributeGroup& ag : attribute_groups) {
+    const Status s = FailIfCfHasTs(ag.column_family());
+    if (!s.ok()) {
+      return s;
+    }
+  }
+  return DB::PutEntity(options, key, attribute_groups);
+}
+
 Status DBImpl::Merge(const WriteOptions& o, ColumnFamilyHandle* column_family,
                      const Slice& key, const Slice& val) {
   const Status s = FailIfCfHasTs(column_family);
@@ -403,17 +414,6 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   IOStatus io_s;
   Status pre_release_cb_status;
   if (status.ok()) {
-    // TODO: this use of operator bool on `tracer_` can avoid unnecessary lock
-    // grabs but does not seem thread-safe.
-    if (tracer_) {
-      InstrumentedMutexLock lock(&trace_mutex_);
-      if (tracer_ && tracer_->IsWriteOrderPreserved()) {
-        for (auto* writer : write_group) {
-          // TODO: maybe handle the tracing status?
-          tracer_->Write(writer->batch).PermitUncheckedError();
-        }
-      }
-    }
     // Rules for when we can update the memtable concurrently
     // 1. supported by memtable
     // 2. Puts are not okay if inplace_update_support
@@ -446,6 +446,20 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
         }
       }
     }
+    // TODO: this use of operator bool on `tracer_` can avoid unnecessary lock
+    // grabs but does not seem thread-safe.
+    if (tracer_) {
+      InstrumentedMutexLock lock(&trace_mutex_);
+      if (tracer_ && tracer_->IsWriteOrderPreserved()) {
+        for (auto* writer : write_group) {
+          if (writer->CallbackFailed()) {
+            continue;
+          }
+          // TODO: maybe handle the tracing status?
+          tracer_->Write(writer->batch).PermitUncheckedError();
+        }
+      }
+    }
     // Note about seq_per_batch_: either disableWAL is set for the entire write
     // group or not. In either case we inc seq for each write batch with no
     // failed callback. This means that there could be a batch with
@@ -1811,11 +1825,9 @@ uint64_t DBImpl::GetMaxTotalWalSize() const {
 Status DBImpl::DelayWrite(uint64_t num_bytes, WriteThread& write_thread,
                           const WriteOptions& write_options) {
   mutex_.AssertHeld();
-  uint64_t time_delayed = 0;
+  uint64_t start_time = 0;
   bool delayed = false;
   {
-    StopWatch sw(immutable_db_options_.clock, stats_, WRITE_STALL,
-                 Histograms::HISTOGRAM_ENUM_MAX, &time_delayed);
     // To avoid parallel timed delays (bad throttling), only support them
     // on the primary write queue.
     uint64_t delay;
@@ -1831,6 +1843,7 @@ Status DBImpl::DelayWrite(uint64_t num_bytes, WriteThread& write_thread,
       if (write_options.no_slowdown) {
         return Status::Incomplete("Write stall");
       }
+      start_time = immutable_db_options_.clock->NowMicros();
       TEST_SYNC_POINT("DBImpl::DelayWrite:Sleep");
 
       // Notify write_thread about the stall so it can setup a barrier and
@@ -1843,7 +1856,7 @@ Status DBImpl::DelayWrite(uint64_t num_bytes, WriteThread& write_thread,
       // (slightly longer because WriteController minimum delay is 1ms, in
       // case of sleep imprecision, rounding, etc.)
       const uint64_t kDelayInterval = 1001;
-      uint64_t stall_end = sw.start_time() + delay;
+      uint64_t stall_end = start_time + delay;
       while (write_controller_.NeedsDelay()) {
         if (immutable_db_options_.clock->NowMicros() >= stall_end) {
           // We already delayed this write `delay` microseconds
@@ -1858,11 +1871,11 @@ Status DBImpl::DelayWrite(uint64_t num_bytes, WriteThread& write_thread,
       write_thread.EndWriteStall();
     }
 
-    // Don't wait if there's a background error, even if its a soft error. We
-    // might wait here indefinitely as the background compaction may never
-    // finish successfully, resulting in the stall condition lasting
-    // indefinitely
-    while (error_handler_.GetBGError().ok() && write_controller_.IsStopped() &&
+    // Don't wait if there's a background error that is not pending recovery
+    // since recovery might never be attempted.
+    while ((error_handler_.GetBGError().ok() ||
+            error_handler_.IsRecoveryInProgress()) &&
+           write_controller_.IsStopped() &&
            !shutting_down_.load(std::memory_order_relaxed)) {
       if (write_options.no_slowdown) {
         return Status::Incomplete("Write stall");
@@ -1884,9 +1897,11 @@ Status DBImpl::DelayWrite(uint64_t num_bytes, WriteThread& write_thread,
   }
   assert(!delayed || !write_options.no_slowdown);
   if (delayed) {
+    auto time_delayed = immutable_db_options_.clock->NowMicros() - start_time;
     default_cf_internal_stats_->AddDBStats(
         InternalStats::kIntStatsWriteStallMicros, time_delayed);
     RecordTick(stats_, STALL_MICROS, time_delayed);
+    RecordInHistogram(stats_, WRITE_STALL, time_delayed);
   }
 
   // If DB is not in read-only mode and write_controller is not stopping
@@ -1953,9 +1968,13 @@ Status DBImpl::ThrottleLowPriWritesIfNeeded(const WriteOptions& write_options,
       // a chance to run. Now we guarantee we are still slowly making
       // progress.
       PERF_TIMER_GUARD(write_delay_time);
-      write_controller_.low_pri_rate_limiter()->Request(
-          my_batch->GetDataSize(), Env::IO_HIGH, nullptr /* stats */,
-          RateLimiter::OpType::kWrite);
+      auto data_size = my_batch->GetDataSize();
+      while (data_size > 0) {
+        size_t allowed = write_controller_.low_pri_rate_limiter()->RequestToken(
+            data_size, 0 /* alignment */, Env::IO_HIGH, nullptr /* stats */,
+            RateLimiter::OpType::kWrite);
+        data_size -= allowed;
+      }
     }
   }
   return Status::OK();
@@ -2378,6 +2397,22 @@ Status DB::PutEntity(const WriteOptions& options,
   return Write(options, &batch);
 }
 
+Status DB::PutEntity(const WriteOptions& options, const Slice& key,
+                     const AttributeGroups& attribute_groups) {
+  ColumnFamilyHandle* default_cf = DefaultColumnFamily();
+  assert(default_cf);
+  const Comparator* const default_cf_ucmp = default_cf->GetComparator();
+  assert(default_cf_ucmp);
+  WriteBatch batch(0 /* reserved_bytes */, 0 /* max_bytes */,
+                   options.protection_bytes_per_key,
+                   default_cf_ucmp->timestamp_size());
+  const Status s = batch.PutEntity(key, attribute_groups);
+  if (!s.ok()) {
+    return s;
+  }
+  return Write(options, &batch);
+}
+
 Status DB::Delete(const WriteOptions& opt, ColumnFamilyHandle* column_family,
                   const Slice& key) {
   WriteBatch batch(0 /* reserved_bytes */, 0 /* max_bytes */,

db/db_iter.cc

@@ -83,7 +83,8 @@ DBIter::DBIter(Env* _env, const ReadOptions& read_options,
       cfd_(cfd),
       timestamp_ub_(read_options.timestamp),
       timestamp_lb_(read_options.iter_start_ts),
-      timestamp_size_(timestamp_ub_ ? timestamp_ub_->size() : 0) {
+      timestamp_size_(timestamp_ub_ ? timestamp_ub_->size() : 0),
+      auto_readahead_size_(read_options.auto_readahead_size) {
   RecordTick(statistics_, NO_ITERATOR_CREATED);
   if (pin_thru_lifetime_) {
     pinned_iters_mgr_.StartPinning();
@@ -113,6 +114,9 @@ Status DBIter::GetProperty(std::string prop_name, std::string* prop) {
   } else if (prop_name == "rocksdb.iterator.internal-key") {
     *prop = saved_key_.GetUserKey().ToString();
     return Status::OK();
+  } else if (prop_name == "rocksdb.iterator.write-time") {
+    // TODO(yuzhangyu): implement return the actual write time.
+    return Status::NotSupported("write time property is under construction");
   }
   return Status::InvalidArgument("Unidentified property.");
 }
@@ -743,15 +747,22 @@ bool DBIter::ReverseToBackward() {
   // When current_entry_is_merged_ is true, iter_ may be positioned on the next
   // key, which may not exist or may have prefix different from current.
   // If that's the case, seek to saved_key_.
-  if (current_entry_is_merged_ &&
-      (!expect_total_order_inner_iter() || !iter_.Valid())) {
+  //
+  // In case of auto_readahead_size enabled, index_iter moves forward during
+  // forward scan for block cache lookup and points to different block. If Prev
+  // op is called, it needs to call SeekForPrev to point to right index_iter_ in
+  // BlockBasedTableIterator. This only happens when direction is changed from
+  // forward to backward.
+  if ((current_entry_is_merged_ &&
+       (!expect_total_order_inner_iter() || !iter_.Valid())) ||
+      auto_readahead_size_) {
     IterKey last_key;
     // Using kMaxSequenceNumber and kValueTypeForSeek
     // (not kValueTypeForSeekForPrev) to seek to a key strictly smaller
     // than saved_key_.
     last_key.SetInternalKey(ParsedInternalKey(
         saved_key_.GetUserKey(), kMaxSequenceNumber, kValueTypeForSeek));
-    if (!expect_total_order_inner_iter()) {
+    if (!expect_total_order_inner_iter() || auto_readahead_size_) {
       iter_.SeekForPrev(last_key.GetInternalKey());
     } else {
       // Some iterators may not support SeekForPrev(), so we avoid using it

db/db_iter.h

@@ -402,6 +402,7 @@ class DBIter final : public Iterator {
   const Slice* const timestamp_lb_;
   const size_t timestamp_size_;
   std::string saved_timestamp_;
+  bool auto_readahead_size_;
 };
 
 // Return a new iterator that converts internal keys (yielded by

db/db_iter_test.cc

@@ -275,6 +275,7 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     db_iter->Next();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
   // Test to check the SeekToLast() with iterate_upper_bound not set
   {
@@ -1415,6 +1416,7 @@ TEST_F(DBIteratorTest, DBIterator1) {
   ASSERT_EQ(db_iter->key().ToString(), "b");
   db_iter->Next();
   ASSERT_FALSE(db_iter->Valid());
+  ASSERT_OK(db_iter->status());
 }
 
 TEST_F(DBIteratorTest, DBIterator2) {
@@ -1528,6 +1530,7 @@ TEST_F(DBIteratorTest, DBIterator5) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1552,6 +1555,7 @@ TEST_F(DBIteratorTest, DBIterator5) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1,merge_2");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1576,6 +1580,7 @@ TEST_F(DBIteratorTest, DBIterator5) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1,merge_2,merge_3");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1600,6 +1605,7 @@ TEST_F(DBIteratorTest, DBIterator5) {
     ASSERT_EQ(db_iter->value().ToString(), "put_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1624,6 +1630,7 @@ TEST_F(DBIteratorTest, DBIterator5) {
     ASSERT_EQ(db_iter->value().ToString(), "put_1,merge_4");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1648,6 +1655,7 @@ TEST_F(DBIteratorTest, DBIterator5) {
     ASSERT_EQ(db_iter->value().ToString(), "put_1,merge_4,merge_5");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1672,6 +1680,7 @@ TEST_F(DBIteratorTest, DBIterator5) {
     ASSERT_EQ(db_iter->value().ToString(), "put_1,merge_4,merge_5,merge_6");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1726,6 +1735,7 @@ TEST_F(DBIteratorTest, DBIterator6) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1750,6 +1760,7 @@ TEST_F(DBIteratorTest, DBIterator6) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1,merge_2");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1774,6 +1785,7 @@ TEST_F(DBIteratorTest, DBIterator6) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1,merge_2,merge_3");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1794,6 +1806,7 @@ TEST_F(DBIteratorTest, DBIterator6) {
         nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1818,6 +1831,7 @@ TEST_F(DBIteratorTest, DBIterator6) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_4");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1842,6 +1856,7 @@ TEST_F(DBIteratorTest, DBIterator6) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_4,merge_5");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1866,6 +1881,7 @@ TEST_F(DBIteratorTest, DBIterator6) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_4,merge_5,merge_6");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 }
 
@@ -1910,6 +1926,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1952,6 +1969,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -1994,6 +2012,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -2041,6 +2060,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -2089,6 +2109,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -2131,6 +2152,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -2179,6 +2201,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -2228,6 +2251,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 
   {
@@ -2271,6 +2295,7 @@ TEST_F(DBIteratorTest, DBIterator7) {
     ASSERT_EQ(db_iter->value().ToString(), "merge_1");
     db_iter->Prev();
     ASSERT_TRUE(!db_iter->Valid());
+    ASSERT_OK(db_iter->status());
   }
 }
 
@@ -2440,6 +2465,7 @@ TEST_F(DBIteratorTest, SeekToLastOccurrenceSeq0) {
   ASSERT_EQ(db_iter->value().ToString(), "2");
   db_iter->Next();
   ASSERT_FALSE(db_iter->Valid());
+  ASSERT_OK(db_iter->status());
 }
 
 TEST_F(DBIteratorTest, DBIterator11) {
@@ -2469,6 +2495,7 @@ TEST_F(DBIteratorTest, DBIterator11) {
   ASSERT_EQ(db_iter->key().ToString(), "b");
   db_iter->Next();
   ASSERT_FALSE(db_iter->Valid());
+  ASSERT_OK(db_iter->status());
 }
 
 TEST_F(DBIteratorTest, DBIterator12) {
@@ -2497,6 +2524,7 @@ TEST_F(DBIteratorTest, DBIterator12) {
   ASSERT_EQ(db_iter->value().ToString(), "1");
   db_iter->Prev();
   ASSERT_FALSE(db_iter->Valid());
+  ASSERT_OK(db_iter->status());
 }
 
 TEST_F(DBIteratorTest, DBIterator13) {
@@ -2635,6 +2663,7 @@ TEST_F(DBIterWithMergeIterTest, InnerMergeIterator1) {
   ASSERT_EQ(db_iter_->value().ToString(), "3");
   db_iter_->Next();
   ASSERT_FALSE(db_iter_->Valid());
+  ASSERT_OK(db_iter_->status());
 }
 
 TEST_F(DBIterWithMergeIterTest, InnerMergeIterator2) {

db/db_iterator_test.cc

@@ -183,6 +183,7 @@ TEST_P(DBIteratorTest, NonBlockingIteration) {
       ASSERT_OK(iter->status());
       count++;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(count, 1);
     delete iter;
 
@@ -217,6 +218,7 @@ TEST_P(DBIteratorTest, NonBlockingIteration) {
       ASSERT_OK(iter->status());
       count++;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(count, 1);
     ASSERT_EQ(numopen, TestGetTickerCount(options, NO_FILE_OPENS));
     ASSERT_EQ(cache_added, TestGetTickerCount(options, BLOCK_CACHE_ADD));
@@ -870,6 +872,7 @@ TEST_P(DBIteratorTest, IterWithSnapshot) {
       }
     }
     db_->ReleaseSnapshot(snapshot);
+    ASSERT_OK(iter->status());
     delete iter;
   } while (ChangeOptions());
 }
@@ -1214,6 +1217,7 @@ TEST_P(DBIteratorTest, DBIteratorBoundOptimizationTest) {
 
     iter->Next();
     ASSERT_FALSE(iter->Valid());
+    ASSERT_OK(iter->status());
     ASSERT_EQ(upper_bound_hits, 1);
   }
 }
@@ -1338,6 +1342,7 @@ TEST_P(DBIteratorTest, IndexWithFirstKey) {
 
     iter->Next();
     ASSERT_FALSE(iter->Valid());
+    ASSERT_OK(iter->status());
     EXPECT_EQ(3, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
     EXPECT_EQ(7, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
   }
@@ -1579,6 +1584,7 @@ class DBIteratorTestForPinnedData : public DBIteratorTest {
         ASSERT_EQ("1", prop_value);
         all_keys.push_back(iter->key());
       }
+      ASSERT_OK(iter->status());
       ASSERT_EQ(all_keys.size(), true_data.size());
 
       // Verify that all keys slices are valid (backward)
@@ -1682,7 +1688,7 @@ TEST_P(DBIteratorTest, PinnedDataIteratorMultipleFiles) {
     ASSERT_EQ(kv.first, data_iter->first);
     ASSERT_EQ(kv.second, data_iter->second);
   }
-
+  ASSERT_OK(iter->status());
   delete iter;
 }
 
@@ -1728,6 +1734,7 @@ TEST_P(DBIteratorTest, PinnedDataIteratorMergeOperator) {
     ASSERT_EQ("1", prop_value);
     results.emplace_back(iter->key(), iter->value().ToString());
   }
+  ASSERT_OK(iter->status());
 
   ASSERT_EQ(results.size(), 1000);
   for (size_t i = 0; i < results.size(); i++) {
@@ -1785,6 +1792,7 @@ TEST_P(DBIteratorTest, PinnedDataIteratorReadAfterUpdate) {
     ASSERT_EQ("1", prop_value);
     results.emplace_back(iter->key(), iter->value().ToString());
   }
+  ASSERT_OK(iter->status());
 
   auto data_iter = true_data.begin();
   for (size_t i = 0; i < results.size(); i++, data_iter++) {
@@ -2079,6 +2087,7 @@ TEST_P(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
       ASSERT_EQ(iter->value().ToString(), data_iter->second);
       data_iter++;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(data_iter, true_data.rend());
 
     delete iter;
@@ -2136,6 +2145,7 @@ TEST_P(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
       entries_right++;
       data_iter++;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(data_iter, true_data.rend());
 
     delete iter;
@@ -2175,6 +2185,7 @@ TEST_P(DBIteratorTest, IteratorWithLocalStatistics) {
       total_next++;
 
       if (!iter->Valid()) {
+        EXPECT_OK(iter->status());
         break;
       }
       total_next_found++;
@@ -2202,6 +2213,7 @@ TEST_P(DBIteratorTest, IteratorWithLocalStatistics) {
       total_prev++;
 
       if (!iter->Valid()) {
+        EXPECT_OK(iter->status());
         break;
       }
       total_prev_found++;
@@ -2416,6 +2428,7 @@ TEST_P(DBIteratorTest, Refresh) {
   ASSERT_EQ(iter->key().compare(Slice("x")), 0);
   iter->Next();
   ASSERT_FALSE(iter->Valid());
+  ASSERT_OK(iter->status());
 
   iter.reset();
 }
@@ -2469,32 +2482,32 @@ TEST_P(DBIteratorTest, RefreshWithSnapshot) {
     verify_iter(1, 3);
     // Refresh to same snapshot
     ASSERT_OK(iter->Refresh(snapshot));
-    ASSERT_TRUE(iter->status().ok() && !iter->Valid());
+    ASSERT_TRUE(!iter->Valid() && iter->status().ok());
     iter->Seek(Key(3));
     verify_iter(3, 6);
-    ASSERT_TRUE(iter->status().ok() && !iter->Valid());
+    ASSERT_TRUE(!iter->Valid() && iter->status().ok());
 
     // Refresh to a newer snapshot
     ASSERT_OK(iter->Refresh(snapshot2));
-    ASSERT_TRUE(iter->status().ok() && !iter->Valid());
+    ASSERT_TRUE(!iter->Valid() && iter->status().ok());
     iter->SeekToFirst();
     verify_iter(0, 4, /*new_key2=*/true);
-    ASSERT_TRUE(iter->status().ok() && !iter->Valid());
+    ASSERT_TRUE(!iter->Valid() && iter->status().ok());
 
     // Refresh to an older snapshot
     ASSERT_OK(iter->Refresh(snapshot));
-    ASSERT_TRUE(iter->status().ok() && !iter->Valid());
+    ASSERT_TRUE(!iter->Valid() && iter->status().ok());
     iter->Seek(Key(3));
     verify_iter(3, 6);
-    ASSERT_TRUE(iter->status().ok() && !iter->Valid());
+    ASSERT_TRUE(!iter->Valid() && iter->status().ok());
 
     // Refresh to no snapshot
     ASSERT_OK(iter->Refresh());
-    ASSERT_TRUE(iter->status().ok() && !iter->Valid());
+    ASSERT_TRUE(!iter->Valid() && iter->status().ok());
     iter->Seek(Key(2));
     verify_iter(2, 4, /*new_key2=*/true);
     verify_iter(6, 7);
-    ASSERT_TRUE(iter->status().ok() && !iter->Valid());
+    ASSERT_TRUE(!iter->Valid() && iter->status().ok());
 
     // Change LSM shape, new SuperVersion is created.
     ASSERT_OK(Flush());
@@ -2599,6 +2612,7 @@ TEST_P(DBIteratorTest, TableFilter) {
     ASSERT_EQ(IterStatus(iter), "f->6");
     iter->Next();
     ASSERT_FALSE(iter->Valid());
+    ASSERT_OK(iter->status());
     ASSERT_TRUE(unseen.empty());
     delete iter;
   }
@@ -2621,6 +2635,7 @@ TEST_P(DBIteratorTest, TableFilter) {
     ASSERT_EQ(IterStatus(iter), "f->6");
     iter->Next();
     ASSERT_FALSE(iter->Valid());
+    ASSERT_OK(iter->status());
     delete iter;
   }
 }
@@ -2705,6 +2720,7 @@ TEST_P(DBIteratorTest, SkipStatistics) {
     ASSERT_OK(iter->status());
     count++;
   }
+  ASSERT_OK(iter->status());
   ASSERT_EQ(count, 3);
   delete iter;
   skip_count += 8;  // Same as above, but in reverse order
@@ -2740,6 +2756,7 @@ TEST_P(DBIteratorTest, SkipStatistics) {
     ASSERT_OK(iter->status());
     count++;
   }
+  ASSERT_OK(iter->status());
   ASSERT_EQ(count, 2);
   delete iter;
   // 3 deletes + 3 original keys + lower sequence of "a"
@@ -3322,6 +3339,7 @@ TEST_F(DBIteratorTest, BackwardIterationOnInplaceUpdateMemtable) {
     for (iter->SeekToLast(); iter->Valid(); iter->Prev()) {
       ++count;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(kNumKeys, count);
   }
 

db/db_log_iter_test.cc

@@ -236,7 +236,7 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorCorruptedLog) {
     ASSERT_OK(test::TruncateFile(env_, logfile_path,
                                  wal_files.front()->SizeFileBytes() / 2));
 
-    ASSERT_OK(db_->EnableFileDeletions());
+    ASSERT_OK(db_->EnableFileDeletions(/*force=*/false));
 
     // Insert a new entry to a new log file
     ASSERT_OK(Put("key1025", DummyString(10)));

db/db_merge_operator_test.cc

@@ -6,10 +6,12 @@
 #include <vector>
 
 #include "db/db_test_util.h"
+#include "db/dbformat.h"
 #include "db/forward_iterator.h"
 #include "port/stack_trace.h"
 #include "rocksdb/merge_operator.h"
 #include "rocksdb/snapshot.h"
+#include "rocksdb/utilities/debug.h"
 #include "util/random.h"
 #include "utilities/merge_operators.h"
 #include "utilities/merge_operators/string_append/stringappend2.h"
@@ -949,6 +951,98 @@ TEST_P(PerConfigMergeOperatorPinningTest, Randomized) {
   VerifyDBFromMap(true_data);
 }
 
+TEST_F(DBMergeOperatorTest, MaxSuccessiveMergesBaseValues) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.merge_operator = MergeOperators::CreatePutOperator();
+  options.max_successive_merges = 1;
+  options.env = env_;
+  Reopen(options);
+
+  constexpr char foo[] = "foo";
+  constexpr char bar[] = "bar";
+  constexpr char baz[] = "baz";
+  constexpr char qux[] = "qux";
+  constexpr char corge[] = "corge";
+
+  // No base value
+  {
+    constexpr char key[] = "key1";
+
+    ASSERT_OK(db_->Merge(WriteOptions(), db_->DefaultColumnFamily(), key, foo));
+    ASSERT_OK(db_->Merge(WriteOptions(), db_->DefaultColumnFamily(), key, bar));
+
+    PinnableSlice result;
+    ASSERT_OK(
+        db_->Get(ReadOptions(), db_->DefaultColumnFamily(), key, &result));
+    ASSERT_EQ(result, bar);
+
+    // We expect the second Merge to be converted to a Put because of
+    // max_successive_merges.
+    constexpr size_t max_key_versions = 8;
+    std::vector<KeyVersion> key_versions;
+    ASSERT_OK(GetAllKeyVersions(db_, db_->DefaultColumnFamily(), key, key,
+                                max_key_versions, &key_versions));
+    ASSERT_EQ(key_versions.size(), 2);
+    ASSERT_EQ(key_versions[0].type, kTypeValue);
+    ASSERT_EQ(key_versions[1].type, kTypeMerge);
+  }
+
+  // Plain base value
+  {
+    constexpr char key[] = "key2";
+
+    ASSERT_OK(db_->Put(WriteOptions(), db_->DefaultColumnFamily(), key, foo));
+    ASSERT_OK(db_->Merge(WriteOptions(), db_->DefaultColumnFamily(), key, bar));
+    ASSERT_OK(db_->Merge(WriteOptions(), db_->DefaultColumnFamily(), key, baz));
+
+    PinnableSlice result;
+    ASSERT_OK(
+        db_->Get(ReadOptions(), db_->DefaultColumnFamily(), key, &result));
+    ASSERT_EQ(result, baz);
+
+    // We expect the second Merge to be converted to a Put because of
+    // max_successive_merges.
+    constexpr size_t max_key_versions = 8;
+    std::vector<KeyVersion> key_versions;
+    ASSERT_OK(GetAllKeyVersions(db_, db_->DefaultColumnFamily(), key, key,
+                                max_key_versions, &key_versions));
+    ASSERT_EQ(key_versions.size(), 3);
+    ASSERT_EQ(key_versions[0].type, kTypeValue);
+    ASSERT_EQ(key_versions[1].type, kTypeMerge);
+    ASSERT_EQ(key_versions[2].type, kTypeValue);
+  }
+
+  // Wide-column base value
+  {
+    constexpr char key[] = "key3";
+    const WideColumns columns{{kDefaultWideColumnName, foo}, {bar, baz}};
+
+    ASSERT_OK(db_->PutEntity(WriteOptions(), db_->DefaultColumnFamily(), key,
+                             columns));
+    ASSERT_OK(db_->Merge(WriteOptions(), db_->DefaultColumnFamily(), key, qux));
+    ASSERT_OK(
+        db_->Merge(WriteOptions(), db_->DefaultColumnFamily(), key, corge));
+
+    PinnableWideColumns result;
+    ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(), key,
+                             &result));
+    const WideColumns expected{{kDefaultWideColumnName, corge}, {bar, baz}};
+    ASSERT_EQ(result.columns(), expected);
+
+    // We expect the second Merge to be converted to a PutEntity because of
+    // max_successive_merges.
+    constexpr size_t max_key_versions = 8;
+    std::vector<KeyVersion> key_versions;
+    ASSERT_OK(GetAllKeyVersions(db_, db_->DefaultColumnFamily(), key, key,
+                                max_key_versions, &key_versions));
+    ASSERT_EQ(key_versions.size(), 3);
+    ASSERT_EQ(key_versions[0].type, kTypeWideColumnEntity);
+    ASSERT_EQ(key_versions[1].type, kTypeMerge);
+    ASSERT_EQ(key_versions[2].type, kTypeWideColumnEntity);
+  }
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/db_options_test.cc

@@ -19,6 +19,8 @@
 #include "rocksdb/convenience.h"
 #include "rocksdb/rate_limiter.h"
 #include "rocksdb/stats_history.h"
+#include "rocksdb/utilities/options_util.h"
+#include "test_util/mock_time_env.h"
 #include "test_util/sync_point.h"
 #include "test_util/testutil.h"
 #include "util/random.h"
@@ -740,6 +742,55 @@ TEST_F(DBOptionsTest, SetStatsDumpPeriodSec) {
   Close();
 }
 
+TEST_F(DBOptionsTest, SetStatsDumpPeriodSecRace) {
+  // This is a mini-stress test looking for inconsistency between the reported
+  // state of the option and the behavior in effect for the DB, after the last
+  // modification to that option (indefinite inconsistency).
+  std::vector<std::thread> threads;
+  for (int i = 0; i < 12; i++) {
+    threads.emplace_back([this, i]() {
+      ASSERT_OK(dbfull()->SetDBOptions(
+          {{"stats_dump_period_sec", i % 2 ? "100" : "0"}}));
+    });
+  }
+
+  for (auto& t : threads) {
+    t.join();
+  }
+
+  bool stats_dump_set = dbfull()->GetDBOptions().stats_dump_period_sec > 0;
+  bool task_enabled = dbfull()->TEST_GetPeriodicTaskScheduler().TEST_HasTask(
+      PeriodicTaskType::kDumpStats);
+
+  ASSERT_EQ(stats_dump_set, task_enabled);
+}
+
+TEST_F(DBOptionsTest, SetOptionsAndFileRace) {
+  // This is a mini-stress test looking for inconsistency between the reported
+  // state of the option and what is persisted in the options file, after the
+  // last modification to that option (indefinite inconsistency).
+  std::vector<std::thread> threads;
+  for (int i = 0; i < 12; i++) {
+    threads.emplace_back([this, i]() {
+      ASSERT_OK(dbfull()->SetOptions({{"ttl", std::to_string(i * 100)}}));
+    });
+  }
+
+  for (auto& t : threads) {
+    t.join();
+  }
+
+  auto setting_in_mem = dbfull()->GetOptions().ttl;
+
+  std::vector<ColumnFamilyDescriptor> cf_descs;
+  DBOptions db_options;
+  ConfigOptions cfg;
+  cfg.env = env_;
+  ASSERT_OK(LoadLatestOptions(cfg, dbname_, &db_options, &cf_descs, nullptr));
+  ASSERT_EQ(cf_descs.size(), 1);
+  ASSERT_EQ(setting_in_mem, cf_descs[0].options.ttl);
+}
+
 TEST_F(DBOptionsTest, SetOptionsStatsPersistPeriodSec) {
   Options options;
   options.create_if_missing = true;
@@ -908,6 +959,7 @@ TEST_F(DBOptionsTest, SetFIFOCompactionOptions) {
   Options options;
   options.env = CurrentOptions().env;
   options.compaction_style = kCompactionStyleFIFO;
+  options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
   options.write_buffer_size = 10 << 10;  // 10KB
   options.arena_block_size = 4096;
   options.compression = kNoCompression;
@@ -941,6 +993,9 @@ TEST_F(DBOptionsTest, SetFIFOCompactionOptions) {
   ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
   ASSERT_EQ(NumTableFilesAtLevel(0), 10);
 
+  ASSERT_EQ(options.statistics->getTickerCount(FIFO_TTL_COMPACTIONS), 0);
+  ASSERT_EQ(options.statistics->getTickerCount(FIFO_MAX_SIZE_COMPACTIONS), 0);
+
   // Set ttl to 1 minute. So all files should get deleted.
   ASSERT_OK(dbfull()->SetOptions({{"ttl", "60"}}));
   ASSERT_EQ(dbfull()->GetOptions().ttl, 60);
@@ -948,6 +1003,10 @@ TEST_F(DBOptionsTest, SetFIFOCompactionOptions) {
   ASSERT_OK(dbfull()->TEST_WaitForCompact());
   ASSERT_EQ(NumTableFilesAtLevel(0), 0);
 
+  ASSERT_GT(options.statistics->getTickerCount(FIFO_TTL_COMPACTIONS), 0);
+  ASSERT_EQ(options.statistics->getTickerCount(FIFO_MAX_SIZE_COMPACTIONS), 0);
+  ASSERT_OK(options.statistics->Reset());
+
   // NOTE: Presumed unnecessary and removed: resetting mock time in env
 
   // Test dynamically changing compaction_options_fifo.max_table_files_size
@@ -971,6 +1030,9 @@ TEST_F(DBOptionsTest, SetFIFOCompactionOptions) {
   ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
   ASSERT_EQ(NumTableFilesAtLevel(0), 10);
 
+  ASSERT_EQ(options.statistics->getTickerCount(FIFO_MAX_SIZE_COMPACTIONS), 0);
+  ASSERT_EQ(options.statistics->getTickerCount(FIFO_TTL_COMPACTIONS), 0);
+
   // Set max_table_files_size to 12 KB. So only 1 file should remain now.
   ASSERT_OK(dbfull()->SetOptions(
       {{"compaction_options_fifo", "{max_table_files_size=12288;}"}}));
@@ -980,6 +1042,10 @@ TEST_F(DBOptionsTest, SetFIFOCompactionOptions) {
   ASSERT_OK(dbfull()->TEST_WaitForCompact());
   ASSERT_EQ(NumTableFilesAtLevel(0), 1);
 
+  ASSERT_GT(options.statistics->getTickerCount(FIFO_MAX_SIZE_COMPACTIONS), 0);
+  ASSERT_EQ(options.statistics->getTickerCount(FIFO_TTL_COMPACTIONS), 0);
+  ASSERT_OK(options.statistics->Reset());
+
   // Test dynamically changing compaction_options_fifo.allow_compaction
   options.compaction_options_fifo.max_table_files_size = 500 << 10;  // 500KB
   options.ttl = 0;
@@ -1033,6 +1099,184 @@ TEST_F(DBOptionsTest, SetFIFOCompactionOptions) {
   ASSERT_EQ(fifo_temp_opt[1].age, 30000);
 }
 
+TEST_F(DBOptionsTest, OffpeakTimes) {
+  Options options;
+  options.create_if_missing = true;
+  Random rnd(test::RandomSeed());
+
+  auto verify_invalid = [&]() {
+    Status s = DBImpl::TEST_ValidateOptions(options);
+    ASSERT_NOK(s);
+    ASSERT_TRUE(s.IsInvalidArgument());
+  };
+
+  auto verify_valid = [&]() {
+    Status s = DBImpl::TEST_ValidateOptions(options);
+    ASSERT_OK(s);
+    ASSERT_FALSE(s.IsInvalidArgument());
+  };
+  std::vector<std::string> invalid_cases = {
+      "06:30-",
+      "-23:30",  // Both need to be set
+      "00:00-00:00",
+      "06:30-06:30"  //  Start time cannot be the same as end time
+      "12:30 PM-23:30",
+      "12:01AM-11:00PM",  // Invalid format
+      "01:99-22:00",      // Invalid value for minutes
+      "00:00-24:00",      // 24:00 is an invalid value
+      "6-7",
+      "6:-7",
+      "06:31.42-7:00",
+      "6.31:42-7:00",
+      "6:0-7:",
+      "15:0.2-3:.7",
+      ":00-00:02",
+      "02:00-:00",
+      "random-value",
+      "No:No-Hi:Hi",
+  };
+
+  std::vector<std::string> valid_cases = {
+      "",  // Not enabled. Valid case
+      "06:30-11:30",
+      "06:30-23:30",
+      "13:30-14:30",
+      "00:00-23:59",  // Entire Day
+      "23:30-01:15",  // From 11:30PM to 1:15AM next day. Valid case.
+      "1:0000000000000-2:000000000042",  // Weird, but we can parse the int.
+  };
+
+  for (std::string invalid_case : invalid_cases) {
+    options.daily_offpeak_time_utc = invalid_case;
+    verify_invalid();
+  }
+  for (std::string valid_case : valid_cases) {
+    options.daily_offpeak_time_utc = valid_case;
+    verify_valid();
+  }
+
+  auto verify_offpeak_info = [&](bool expected_is_now_off_peak,
+                                 int expected_seconds_till_next_offpeak_start,
+                                 int now_utc_hour, int now_utc_minute,
+                                 int now_utc_second = 0) {
+    auto mock_clock = std::make_shared<MockSystemClock>(env_->GetSystemClock());
+    // Add some extra random days to current time
+    int days = rnd.Uniform(100);
+    mock_clock->SetCurrentTime(
+        days * OffpeakTimeOption::kSecondsPerDay +
+        now_utc_hour * OffpeakTimeOption::kSecondsPerHour +
+        now_utc_minute * OffpeakTimeOption::kSecondsPerMinute + now_utc_second);
+    Status s = DBImpl::TEST_ValidateOptions(options);
+    ASSERT_OK(s);
+    auto offpeak_option = OffpeakTimeOption(options.daily_offpeak_time_utc);
+    int64_t now;
+    ASSERT_OK(mock_clock.get()->GetCurrentTime(&now));
+    auto offpeak_info = offpeak_option.GetOffpeakTimeInfo(now);
+    ASSERT_EQ(expected_is_now_off_peak, offpeak_info.is_now_offpeak);
+    ASSERT_EQ(expected_seconds_till_next_offpeak_start,
+              offpeak_info.seconds_till_next_offpeak_start);
+  };
+
+  options.daily_offpeak_time_utc = "";
+  verify_offpeak_info(false, 0, 12, 30);
+
+  options.daily_offpeak_time_utc = "06:30-11:30";
+  verify_offpeak_info(false, 1 * OffpeakTimeOption::kSecondsPerHour, 5, 30);
+  verify_offpeak_info(true, 24 * OffpeakTimeOption::kSecondsPerHour, 6, 30);
+  verify_offpeak_info(true, 20 * OffpeakTimeOption::kSecondsPerHour, 10, 30);
+  verify_offpeak_info(true, 19 * OffpeakTimeOption::kSecondsPerHour, 11, 30);
+  verify_offpeak_info(false, 17 * OffpeakTimeOption::kSecondsPerHour, 13, 30);
+
+  options.daily_offpeak_time_utc = "23:30-04:30";
+  verify_offpeak_info(false, 17 * OffpeakTimeOption::kSecondsPerHour, 6, 30);
+  verify_offpeak_info(true, 24 * OffpeakTimeOption::kSecondsPerHour, 23, 30);
+  verify_offpeak_info(true,
+                      23 * OffpeakTimeOption::kSecondsPerHour +
+                          30 * OffpeakTimeOption::kSecondsPerMinute,
+                      0, 0);
+  verify_offpeak_info(true,
+                      22 * OffpeakTimeOption::kSecondsPerHour +
+                          30 * OffpeakTimeOption::kSecondsPerMinute,
+                      1, 0);
+  verify_offpeak_info(true, 19 * OffpeakTimeOption::kSecondsPerHour, 4, 30);
+  verify_offpeak_info(false,
+                      18 * OffpeakTimeOption::kSecondsPerHour +
+                          59 * OffpeakTimeOption::kSecondsPerMinute,
+                      4, 31);
+
+  // Entire day offpeak
+  options.daily_offpeak_time_utc = "00:00-23:59";
+  verify_offpeak_info(true, 24 * OffpeakTimeOption::kSecondsPerHour, 0, 0);
+  verify_offpeak_info(true, 12 * OffpeakTimeOption::kSecondsPerHour, 12, 00);
+  verify_offpeak_info(true, 1 * OffpeakTimeOption::kSecondsPerMinute, 23, 59);
+  verify_offpeak_info(true, 59, 23, 59, 1);
+  verify_offpeak_info(true, 1, 23, 59, 59);
+
+  // Start with a valid option
+  options.daily_offpeak_time_utc = "01:30-04:15";
+  DestroyAndReopen(options);
+  ASSERT_EQ("01:30-04:15", dbfull()->GetDBOptions().daily_offpeak_time_utc);
+
+  int may_schedule_compaction_called = 0;
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::MaybeScheduleFlushOrCompaction:Start",
+      [&](void*) { may_schedule_compaction_called++; });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  // Make sure calling SetDBOptions with invalid option does not change the
+  // value nor call MaybeScheduleFlushOrCompaction()
+  for (std::string invalid_case : invalid_cases) {
+    ASSERT_NOK(
+        dbfull()->SetDBOptions({{"daily_offpeak_time_utc", invalid_case}}));
+    ASSERT_EQ("01:30-04:15", dbfull()
+                                 ->GetVersionSet()
+                                 ->offpeak_time_option()
+                                 .daily_offpeak_time_utc);
+    ASSERT_EQ(1 * kSecondInHour + 30 * kSecondInMinute,
+              dbfull()
+                  ->GetVersionSet()
+                  ->offpeak_time_option()
+                  .daily_offpeak_start_time_utc);
+    ASSERT_EQ(4 * kSecondInHour + 15 * kSecondInMinute,
+              dbfull()
+                  ->GetVersionSet()
+                  ->offpeak_time_option()
+                  .daily_offpeak_end_time_utc);
+  }
+  ASSERT_EQ(0, may_schedule_compaction_called);
+
+  // Changing to new valid values should call MaybeScheduleFlushOrCompaction()
+  // and sets the offpeak_time_option in VersionSet
+  int expected_count = 0;
+  for (std::string valid_case : valid_cases) {
+    if (dbfull()
+            ->GetVersionSet()
+            ->offpeak_time_option()
+            .daily_offpeak_time_utc != valid_case) {
+      expected_count++;
+    }
+    ASSERT_OK(dbfull()->SetDBOptions({{"daily_offpeak_time_utc", valid_case}}));
+    ASSERT_EQ(valid_case, dbfull()->GetDBOptions().daily_offpeak_time_utc);
+    ASSERT_EQ(valid_case, dbfull()
+                              ->GetVersionSet()
+                              ->offpeak_time_option()
+                              .daily_offpeak_time_utc);
+  }
+  ASSERT_EQ(expected_count, may_schedule_compaction_called);
+
+  // Changing to the same value should not call MaybeScheduleFlushOrCompaction()
+  ASSERT_OK(
+      dbfull()->SetDBOptions({{"daily_offpeak_time_utc", "06:30-11:30"}}));
+  may_schedule_compaction_called = 0;
+  ASSERT_OK(
+      dbfull()->SetDBOptions({{"daily_offpeak_time_utc", "06:30-11:30"}}));
+  ASSERT_EQ(0, may_schedule_compaction_called);
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+  Close();
+}
+
 TEST_F(DBOptionsTest, CompactionReadaheadSizeChange) {
   for (bool use_direct_reads : {true, false}) {
     SpecialEnv env(env_);

db/db_properties_test.cc

@@ -107,12 +107,12 @@ TEST_F(DBPropertiesTest, Empty) {
         dbfull()->GetProperty("rocksdb.is-file-deletions-enabled", &num));
     ASSERT_EQ("0", num);
 
-    ASSERT_OK(db_->EnableFileDeletions(false));
+    ASSERT_OK(db_->EnableFileDeletions(/*force=*/false));
     ASSERT_TRUE(
         dbfull()->GetProperty("rocksdb.is-file-deletions-enabled", &num));
     ASSERT_EQ("0", num);
 
-    ASSERT_OK(db_->EnableFileDeletions());
+    ASSERT_OK(db_->EnableFileDeletions(/*force=*/true));
     ASSERT_TRUE(
         dbfull()->GetProperty("rocksdb.is-file-deletions-enabled", &num));
     ASSERT_EQ("1", num);
@@ -1084,12 +1084,14 @@ class CountingUserTblPropCollector : public TablePropertiesCollector {
   const char* Name() const override { return "CountingUserTblPropCollector"; }
 
   Status Finish(UserCollectedProperties* properties) override {
+    assert(!finish_called_);
     std::string encoded;
     PutVarint32(&encoded, count_);
     *properties = UserCollectedProperties{
         {"CountingUserTblPropCollector", message_},
         {"Count", encoded},
     };
+    finish_called_ = true;
     return Status::OK();
   }
 
@@ -1101,12 +1103,14 @@ class CountingUserTblPropCollector : public TablePropertiesCollector {
   }
 
   UserCollectedProperties GetReadableProperties() const override {
+    assert(finish_called_);
     return UserCollectedProperties{};
   }
 
  private:
   std::string message_ = "Rocksdb";
   uint32_t count_ = 0;
+  bool finish_called_ = false;
 };
 
 class CountingUserTblPropCollectorFactory
@@ -1740,7 +1744,7 @@ TEST_F(DBPropertiesTest, SstFilesSize) {
   ASSERT_EQ(obsolete_sst_size, sst_size);
 
   // Let the obsolete files be deleted.
-  ASSERT_OK(db_->EnableFileDeletions());
+  ASSERT_OK(db_->EnableFileDeletions(/*force=*/false));
   ASSERT_TRUE(db_->GetIntProperty(DB::Properties::kObsoleteSstFilesSize,
                                   &obsolete_sst_size));
   ASSERT_EQ(obsolete_sst_size, 0);

db/db_range_del_test.cc

@@ -682,6 +682,7 @@ TEST_F(DBRangeDelTest, TableEvictedDuringScan) {
     // soon as its refcount drops to zero.
     bbto.block_cache->EraseUnRefEntries();
   }
+  ASSERT_OK(iter->status());
   ASSERT_EQ(kNum, expected);
   delete iter;
   db_->ReleaseSnapshot(snapshot);
@@ -840,6 +841,7 @@ TEST_F(DBRangeDelTest, IteratorRemovesCoveredKeys) {
       ++expected;
     }
   }
+  ASSERT_OK(iter->status());
   ASSERT_EQ(kNum, expected);
   delete iter;
 }
@@ -908,6 +910,7 @@ TEST_F(DBRangeDelTest, IteratorIgnoresRangeDeletions) {
     std::string key;
     ASSERT_EQ(expected[i], iter->key());
   }
+  ASSERT_OK(iter->status());
   ASSERT_EQ(3, i);
   delete iter;
   db_->ReleaseSnapshot(snapshot);
@@ -1382,6 +1385,7 @@ TEST_F(DBRangeDelTest, UntruncatedTombstoneDoesNotDeleteNewerKey) {
     for (; iter->Valid(); iter->Next()) {
       ++keys_found;
     }
+    EXPECT_OK(iter->status());
     delete iter;
     return keys_found;
   };
@@ -1485,6 +1489,7 @@ TEST_F(DBRangeDelTest, DeletedMergeOperandReappearsIterPrev) {
   for (; iter->Valid(); iter->Prev()) {
     ++keys_found;
   }
+  ASSERT_OK(iter->status());
   delete iter;
   ASSERT_EQ(kNumKeys, keys_found);
 
@@ -1519,6 +1524,7 @@ TEST_F(DBRangeDelTest, SnapshotPreventsDroppedKeys) {
 
   iter->Next();
   ASSERT_FALSE(iter->Valid());
+  ASSERT_OK(iter->status());
 
   delete iter;
   db_->ReleaseSnapshot(snapshot);
@@ -1564,6 +1570,7 @@ TEST_F(DBRangeDelTest, SnapshotPreventsDroppedKeysInImmMemTables) {
 
   iter->Next();
   ASSERT_FALSE(iter->Valid());
+  ASSERT_OK(iter->status());
 }
 
 TEST_F(DBRangeDelTest, RangeTombstoneWrittenToMinimalSsts) {
@@ -1978,6 +1985,7 @@ TEST_F(DBRangeDelTest, IteratorRefresh) {
     ASSERT_EQ("key1", iter->key());
     iter->Next();
     ASSERT_FALSE(iter->Valid());
+    ASSERT_OK(iter->status());
 
     delete iter;
   }

db/db_rate_limiter_test.cc

@@ -220,6 +220,7 @@ TEST_P(DBRateLimiterOnReadTest, Iterator) {
       ++expected;
     }
   }
+  ASSERT_OK(iter->status());
   // Reverse scan does not read evenly (one block per iteration) due to
   // descending seqno ordering, so wait until after the loop to check total.
   ASSERT_EQ(expected, options_.rate_limiter->GetTotalRequests(Env::IO_USER));

db/db_readonly_with_timestamp_test.cc

@@ -240,6 +240,7 @@ TEST_F(DBReadOnlyTestWithTimestamp, IteratorAndGet) {
       get_value_and_check(db_, read_opts, it->key(), it->value(),
                           write_timestamps[i]);
     }
+    ASSERT_OK(it->status());
     size_t expected_count = kMaxKey - start_keys[i] + 1;
     ASSERT_EQ(expected_count, count);
 
@@ -252,6 +253,7 @@ TEST_F(DBReadOnlyTestWithTimestamp, IteratorAndGet) {
       get_value_and_check(db_, read_opts, it->key(), it->value(),
                           write_timestamps[i]);
     }
+    ASSERT_OK(it->status());
     ASSERT_EQ(static_cast<size_t>(kMaxKey) - start_keys[i] + 1, count);
 
     // SeekToFirst()/SeekToLast() with lower/upper bounds.
@@ -273,6 +275,7 @@ TEST_F(DBReadOnlyTestWithTimestamp, IteratorAndGet) {
         get_value_and_check(db_, read_opts, it->key(), it->value(),
                             write_timestamps[i]);
       }
+      ASSERT_OK(it->status());
       ASSERT_EQ(r - std::max(l, start_keys[i]), count);
 
       for (it->SeekToLast(), key = std::min(r, kMaxKey + 1), count = 0;
@@ -282,6 +285,7 @@ TEST_F(DBReadOnlyTestWithTimestamp, IteratorAndGet) {
         get_value_and_check(db_, read_opts, it->key(), it->value(),
                             write_timestamps[i]);
       }
+      ASSERT_OK(it->status());
       l += (kMaxKey / 100);
       r -= (kMaxKey / 100);
     }
@@ -328,6 +332,7 @@ TEST_F(DBReadOnlyTestWithTimestamp, Iterators) {
     CheckIterUserEntry(iters[0], Key1(key), kTypeValue,
                        "value" + std::to_string(key), write_timestamp);
   }
+  ASSERT_OK(iters[0]->status());
 
   size_t expected_count = kMaxKey - 0 + 1;
   ASSERT_EQ(expected_count, count);

db/db_secondary_test.cc

@@ -200,6 +200,7 @@ TEST_F(DBSecondaryTest, ReopenAsSecondary) {
     }
     ++count;
   }
+  ASSERT_OK(iter->status());
   delete iter;
   ASSERT_EQ(3, count);
 }
@@ -534,6 +535,8 @@ TEST_F(DBSecondaryTest, SecondaryCloseFiles) {
     }
     ASSERT_FALSE(iter1->Valid());
     ASSERT_FALSE(iter2->Valid());
+    ASSERT_OK(iter1->status());
+    ASSERT_OK(iter2->status());
   };
 
   ASSERT_OK(Put("a", "value"));
@@ -806,6 +809,7 @@ TEST_F(DBSecondaryTest, MissingTableFileDuringOpen) {
   for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
     ++count;
   }
+  ASSERT_OK(iter->status());
   ASSERT_EQ(2, count);
   delete iter;
 }
@@ -863,6 +867,7 @@ TEST_F(DBSecondaryTest, MissingTableFile) {
   for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
     ++count;
   }
+  ASSERT_OK(iter->status());
   ASSERT_EQ(2, count);
   delete iter;
 }
@@ -935,6 +940,7 @@ TEST_F(DBSecondaryTest, SwitchManifest) {
       ASSERT_EQ("value_" + std::to_string(kNumFiles - 1),
                 iter->value().ToString());
     }
+    EXPECT_OK(iter->status());
   };
 
   range_scan_db();
@@ -1485,6 +1491,7 @@ TEST_F(DBSecondaryTestWithTimestamp, IteratorAndGet) {
       get_value_and_check(db_, read_opts, it->key(), it->value(),
                           write_timestamps[i]);
     }
+    ASSERT_OK(it->status());
     size_t expected_count = kMaxKey - start_keys[i] + 1;
     ASSERT_EQ(expected_count, count);
 
@@ -1497,6 +1504,7 @@ TEST_F(DBSecondaryTestWithTimestamp, IteratorAndGet) {
       get_value_and_check(db_, read_opts, it->key(), it->value(),
                           write_timestamps[i]);
     }
+    ASSERT_OK(it->status());
     ASSERT_EQ(static_cast<size_t>(kMaxKey) - start_keys[i] + 1, count);
 
     // SeekToFirst()/SeekToLast() with lower/upper bounds.
@@ -1518,6 +1526,7 @@ TEST_F(DBSecondaryTestWithTimestamp, IteratorAndGet) {
         get_value_and_check(db_, read_opts, it->key(), it->value(),
                             write_timestamps[i]);
       }
+      ASSERT_OK(it->status());
       ASSERT_EQ(r - std::max(l, start_keys[i]), count);
 
       for (it->SeekToLast(), key = std::min(r, kMaxKey + 1), count = 0;
@@ -1527,6 +1536,7 @@ TEST_F(DBSecondaryTestWithTimestamp, IteratorAndGet) {
         get_value_and_check(db_, read_opts, it->key(), it->value(),
                             write_timestamps[i]);
       }
+      ASSERT_OK(it->status());
       l += (kMaxKey / 100);
       r -= (kMaxKey / 100);
     }
@@ -1736,6 +1746,7 @@ TEST_F(DBSecondaryTestWithTimestamp, Iterators) {
     CheckIterUserEntry(iters[0], Key1(key), kTypeValue,
                        "value" + std::to_string(key), write_timestamp);
   }
+  ASSERT_OK(iters[0]->status());
 
   size_t expected_count = kMaxKey - 0 + 1;
   ASSERT_EQ(expected_count, count);

db/db_sst_test.cc

@@ -937,12 +937,21 @@ INSTANTIATE_TEST_CASE_P(DBWALTestWithParam, DBWALTestWithParam,
                         ::testing::Values(std::make_tuple("", true),
                                           std::make_tuple("_wal_dir", false)));
 
-TEST_F(DBSSTTest, OpenDBWithExistingTrash) {
+TEST_F(DBSSTTest, OpenDBWithExistingTrashAndObsoleteSstFile) {
   Options options = CurrentOptions();
-
   options.sst_file_manager.reset(
       NewSstFileManager(env_, nullptr, "", 1024 * 1024 /* 1 MB/sec */));
   auto sfm = static_cast<SstFileManagerImpl*>(options.sst_file_manager.get());
+  // Set an extra high trash ratio to prevent immediate/non-rate limited
+  // deletions
+  sfm->SetDeleteRateBytesPerSecond(1024 * 1024);
+  sfm->delete_scheduler()->SetMaxTrashDBRatio(1000.0);
+
+  int bg_delete_file = 0;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DeleteScheduler::DeleteTrashFile:DeleteFile",
+      [&](void* /*arg*/) { bg_delete_file++; });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
 
   Destroy(last_options_);
 
@@ -951,13 +960,22 @@ TEST_F(DBSSTTest, OpenDBWithExistingTrash) {
   ASSERT_OK(WriteStringToFile(env_, "abc", dbname_ + "/" + "001.sst.trash"));
   ASSERT_OK(WriteStringToFile(env_, "abc", dbname_ + "/" + "002.sst.trash"));
   ASSERT_OK(WriteStringToFile(env_, "abc", dbname_ + "/" + "003.sst.trash"));
+  // Manually add an obsolete sst file. Obsolete SST files are discovered and
+  // deleted upon recovery.
+  constexpr uint64_t kSstFileNumber = 100;
+  const std::string kObsoleteSstFile =
+      MakeTableFileName(dbname_, kSstFileNumber);
+  ASSERT_OK(WriteStringToFile(env_, "abc", kObsoleteSstFile));
 
-  // Reopen the DB and verify that it deletes existing trash files
+  // Reopen the DB and verify that it deletes existing trash files and obsolete
+  // SST files with rate limiting.
   Reopen(options);
   sfm->WaitForEmptyTrash();
   ASSERT_NOK(env_->FileExists(dbname_ + "/" + "001.sst.trash"));
   ASSERT_NOK(env_->FileExists(dbname_ + "/" + "002.sst.trash"));
   ASSERT_NOK(env_->FileExists(dbname_ + "/" + "003.sst.trash"));
+  ASSERT_NOK(env_->FileExists(kObsoleteSstFile));
+  ASSERT_EQ(bg_delete_file, 4);
 }
 
 // Create a DB with 2 db_paths, and generate multiple files in the 2
@@ -1520,6 +1538,11 @@ TEST_F(DBSSTTest, OpenDBWithInfiniteMaxOpenFilesSubjectToMemoryLimit) {
 }
 
 TEST_F(DBSSTTest, GetTotalSstFilesSize) {
+  // FIXME: L0 file and L1+ file also differ in size of `oldest_key_time`.
+  //  L0 file has non-zero `oldest_key_time` while L1+ files have 0.
+  //  The test passes since L1+ file uses current time instead of 0
+  //  as oldest_ancestor_time.
+  //
   // We don't propagate oldest-key-time table property on compaction and
   // just write 0 as default value. This affect the exact table size, since
   // we encode table properties as varint64. Force time to be 0 to work around

db/db_tailing_iter_test.cc

@@ -52,6 +52,7 @@ TEST_P(DBTestTailingIterator, TailingIteratorSingle) {
 
   iter->Next();
   ASSERT_TRUE(!iter->Valid());
+  ASSERT_OK(iter->status());
 }
 
 TEST_P(DBTestTailingIterator, TailingIteratorKeepAdding) {
@@ -361,7 +362,7 @@ TEST_P(DBTestTailingIterator, TailingIteratorDeletes) {
     int count = 0;
     for (; iter->Valid(); iter->Next(), ++count)
       ;
-
+    ASSERT_OK(iter->status());
     ASSERT_EQ(count, num_records);
   }
   Close();
@@ -408,6 +409,7 @@ TEST_P(DBTestTailingIterator, TailingIteratorPrefixSeek) {
 
     iter->Next();
     ASSERT_TRUE(!iter->Valid());
+    ASSERT_OK(iter->status());
   }
   Close();
 }

db/db_test.cc

@@ -127,6 +127,7 @@ TEST_F(DBTest, MockEnvTest) {
     iterator->Next();
   }
   ASSERT_TRUE(!iterator->Valid());
+  ASSERT_OK(iterator->status());
   delete iterator;
 
   DBImpl* dbi = static_cast_with_check<DBImpl>(db);
@@ -171,6 +172,7 @@ TEST_F(DBTest, MemEnvTest) {
     iterator->Next();
   }
   ASSERT_TRUE(!iterator->Valid());
+  ASSERT_OK(iterator->status());
   delete iterator;
 
   DBImpl* dbi = static_cast_with_check<DBImpl>(db);
@@ -2983,6 +2985,7 @@ TEST_F(DBTest, GroupCommitTest) {
       itr->Next();
     }
     ASSERT_TRUE(!itr->Valid());
+    ASSERT_OK(itr->status());
     delete itr;
 
     HistogramData hist_data;
@@ -3511,6 +3514,8 @@ static bool CompareIterators(int step, DB* model, DB* db,
       ok = false;
     }
   }
+  EXPECT_OK(miter->status());
+  EXPECT_OK(dbiter->status());
   (void)count;
   delete miter;
   delete dbiter;
@@ -6008,6 +6013,7 @@ TEST_F(DBTest, MergeTestTime) {
     ASSERT_OK(iter->status());
     ++count;
   }
+  ASSERT_OK(iter->status());
 
   ASSERT_EQ(1, count);
   ASSERT_EQ(4000000, TestGetTickerCount(options, MERGE_OPERATION_TOTAL_TIME));

db/db_test2.cc

@@ -1805,6 +1805,7 @@ TEST_P(CompressionFailuresTest, CompressionFailures) {
       ASSERT_EQ(key_value_written[key], value);
       key_value_written.erase(key);
     }
+    ASSERT_OK(db_iter->status());
     ASSERT_EQ(0, key_value_written.size());
   } else if (compression_failure_type_ == kTestDecompressionFail) {
     ASSERT_EQ(std::string(s.getState()),
@@ -3801,6 +3802,7 @@ TEST_F(DBTest2, MemtableOnlyIterator) {
     count++;
   }
   ASSERT_TRUE(!it->Valid());
+  ASSERT_OK(it->status());
   ASSERT_EQ(2, count);
   delete it;
 
@@ -3880,16 +3882,26 @@ TEST_F(DBTest2, LowPriWrite) {
   ASSERT_OK(Put("", "", wo));
   ASSERT_EQ(1, rate_limit_count.load());
 
+  wo.low_pri = true;
+  std::string big_value = std::string(1 * 1024 * 1024, 'x');
+  ASSERT_OK(Put("", big_value, wo));
+  ASSERT_LT(1, rate_limit_count.load());
+  // Reset
+  rate_limit_count = 0;
+  wo.low_pri = false;
+  ASSERT_OK(Put("", big_value, wo));
+  ASSERT_EQ(0, rate_limit_count.load());
+
   TEST_SYNC_POINT("DBTest.LowPriWrite:0");
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
 
   ASSERT_OK(dbfull()->TEST_WaitForCompact());
   wo.low_pri = true;
   ASSERT_OK(Put("", "", wo));
-  ASSERT_EQ(1, rate_limit_count.load());
+  ASSERT_EQ(0, rate_limit_count.load());
   wo.low_pri = false;
   ASSERT_OK(Put("", "", wo));
-  ASSERT_EQ(1, rate_limit_count.load());
+  ASSERT_EQ(0, rate_limit_count.load());
 }
 
 TEST_F(DBTest2, RateLimitedCompactionReads) {
@@ -4123,7 +4135,7 @@ TEST_F(DBTest2, LiveFilesOmitObsoleteFiles) {
     ASSERT_OK(env_->FileExists(LogFileName(dbname_, log_file->LogNumber())));
   }
 
-  ASSERT_OK(db_->EnableFileDeletions());
+  ASSERT_OK(db_->EnableFileDeletions(/*force=*/false));
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
 }
 
@@ -7588,6 +7600,7 @@ TEST_F(DBTest2, BestEffortsRecoveryWithSstUniqueIdVerification) {
       ASSERT_EQ(std::to_string(cnt), it->key());
       ASSERT_EQ(expected_v, it->value());
     }
+    EXPECT_OK(it->status());
     ASSERT_EQ(expected_count, cnt);
   };
 

db/db_test_util.cc

@@ -699,6 +699,7 @@ void DBTestBase::Destroy(const Options& options, bool delete_cf_paths) {
 }
 
 Status DBTestBase::ReadOnlyReopen(const Options& options) {
+  Close();
   MaybeInstallTimeElapseOnlySleep(options);
   return DB::OpenForReadOnly(options, dbname_, &db_);
 }
@@ -942,6 +943,7 @@ std::string DBTestBase::Contents(int cf) {
     EXPECT_EQ(IterStatus(iter), forward[forward.size() - matched - 1]);
     matched++;
   }
+  EXPECT_OK(iter->status());
   EXPECT_EQ(matched, forward.size());
 
   delete iter;
@@ -1364,6 +1366,7 @@ std::string DBTestBase::IterStatus(Iterator* iter) {
   if (iter->Valid()) {
     result = iter->key().ToString() + "->" + iter->value().ToString();
   } else {
+    EXPECT_OK(iter->status());
     result = "(invalid)";
   }
   return result;
@@ -1582,6 +1585,7 @@ void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
       iter_cnt++;
       total_reads++;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(data_iter, true_data.end())
         << iter_cnt << " / " << true_data.size();
     delete iter;
@@ -1605,6 +1609,7 @@ void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
       iter_cnt++;
       total_reads++;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(data_rev, true_data.rend())
         << iter_cnt << " / " << true_data.size();
 

db/db_test_util.h

@@ -233,6 +233,7 @@ class SpecialEnv : public EnvWrapper {
       size_t GetUniqueId(char* id, size_t max_size) const override {
         return base_->GetUniqueId(id, max_size);
       }
+      uint64_t GetFileSize() final { return base_->GetFileSize(); }
     };
     class ManifestFile : public WritableFile {
      public:
@@ -345,6 +346,7 @@ class SpecialEnv : public EnvWrapper {
       Status Allocate(uint64_t offset, uint64_t len) override {
         return base_->Allocate(offset, len);
       }
+      uint64_t GetFileSize() final { return base_->GetFileSize(); }
 
      private:
       SpecialEnv* env_;

db/db_universal_compaction_test.cc

@@ -7,9 +7,12 @@
 // 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 <string>
+
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/utilities/table_properties_collectors.h"
+#include "test_util/mock_time_env.h"
 #include "test_util/sync_point.h"
 #include "test_util/testutil.h"
 #include "util/random.h"
@@ -1469,6 +1472,7 @@ TEST_P(DBTestUniversalCompaction, IncreaseUniversalCompactionNumLevels) {
       keys_in_db.append(iter->key().ToString());
       keys_in_db.push_back(',');
     }
+    EXPECT_OK(iter->status());
     delete iter;
 
     std::string expected_keys;
@@ -2229,6 +2233,141 @@ TEST_F(DBTestUniversalCompaction2, PeriodicCompaction) {
   ASSERT_EQ(4, output_level);
 }
 
+TEST_F(DBTestUniversalCompaction2, PeriodicCompactionOffpeak) {
+  constexpr int kSecondsPerDay = 86400;
+  constexpr int kSecondsPerHour = 3600;
+  constexpr int kSecondsPerMinute = 60;
+
+  Options opts = CurrentOptions();
+  opts.compaction_style = kCompactionStyleUniversal;
+  opts.level0_file_num_compaction_trigger = 10;
+  opts.max_open_files = -1;
+  opts.compaction_options_universal.size_ratio = 10;
+  opts.compaction_options_universal.min_merge_width = 2;
+  opts.compaction_options_universal.max_size_amplification_percent = 200;
+  opts.periodic_compaction_seconds = 5 * kSecondsPerDay;  // 5 days
+  opts.num_levels = 5;
+
+  // Just to add some extra random days to current time
+  Random rnd(test::RandomSeed());
+  int days = rnd.Uniform(100);
+
+  int periodic_compactions = 0;
+  int start_level = -1;
+  int output_level = -1;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "UniversalCompactionPicker::PickPeriodicCompaction:Return",
+      [&](void* arg) {
+        Compaction* compaction = reinterpret_cast<Compaction*>(arg);
+        ASSERT_TRUE(arg != nullptr);
+        ASSERT_TRUE(compaction->compaction_reason() ==
+                    CompactionReason::kPeriodicCompaction);
+        start_level = compaction->start_level();
+        output_level = compaction->output_level();
+        periodic_compactions++;
+      });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  for (std::string preset_offpeak_time : {"", "00:30-04:30", "10:30-02:30"}) {
+    SCOPED_TRACE("preset_offpeak_time=" + preset_offpeak_time);
+    for (std::string new_offpeak_time : {"", "23:30-02:30"}) {
+      SCOPED_TRACE("new_offpeak_time=" + new_offpeak_time);
+      std::vector<std::pair<int, int>> times_to_test = {
+          {0, 0}, {2, 30}, {3, 15}, {5, 10}, {13, 30}, {23, 30}};
+      for (std::pair<int, int> now : times_to_test) {
+        int now_hour = now.first;
+        int now_minute = now.second;
+        SCOPED_TRACE("now=" + std::to_string(now_hour) + ":" +
+                     std::to_string(now_minute));
+
+        auto mock_clock =
+            std::make_shared<MockSystemClock>(env_->GetSystemClock());
+        auto mock_env = std::make_unique<CompositeEnvWrapper>(env_, mock_clock);
+        opts.env = mock_env.get();
+        mock_clock->SetCurrentTime(days * kSecondsPerDay +
+                                   now_hour * kSecondsPerHour +
+                                   now_minute * kSecondsPerMinute);
+        opts.daily_offpeak_time_utc = preset_offpeak_time;
+        Reopen(opts);
+
+        ASSERT_OK(Put("foo", "bar1"));
+        ASSERT_OK(Flush());
+        ASSERT_EQ(0, periodic_compactions);
+
+        // Move clock forward by 8 hours. There should be no periodic
+        // compaction, yet.
+        mock_clock->MockSleepForSeconds(8 * kSecondsPerHour);
+        ASSERT_OK(Put("foo", "bar2"));
+        ASSERT_OK(Flush());
+        ASSERT_OK(dbfull()->TEST_WaitForCompact());
+        ASSERT_EQ(0, periodic_compactions);
+
+        // Move clock forward by 4 days
+        mock_clock->MockSleepForSeconds(4 * kSecondsPerDay);
+        ASSERT_OK(Put("foo", "bar3"));
+        ASSERT_OK(Flush());
+        ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+        int64_t mock_now;
+        ASSERT_OK(mock_clock->GetCurrentTime(&mock_now));
+
+        auto offpeak_time_info =
+            dbfull()->GetVersionSet()->offpeak_time_option().GetOffpeakTimeInfo(
+                mock_now);
+        // At this point, the first file is 4 days and 8 hours old.
+        // If it's offpeak now and the file is expected to expire before the
+        // next offpeak starts
+        if (offpeak_time_info.is_now_offpeak &&
+            offpeak_time_info.seconds_till_next_offpeak_start /
+                    kSecondsPerHour >
+                16) {
+          ASSERT_EQ(1, periodic_compactions);
+        } else {
+          ASSERT_EQ(0, periodic_compactions);
+          // Change offpeak option by SetDBOption()
+          if (preset_offpeak_time != new_offpeak_time) {
+            ASSERT_OK(dbfull()->SetDBOptions(
+                {{"daily_offpeak_time_utc", new_offpeak_time}}));
+            ASSERT_OK(Put("foo", "bar4"));
+            ASSERT_OK(Flush());
+            ASSERT_OK(dbfull()->TEST_WaitForCompact());
+            offpeak_time_info = dbfull()
+                                    ->GetVersionSet()
+                                    ->offpeak_time_option()
+                                    .GetOffpeakTimeInfo(mock_now);
+            // if the first file is now eligible to be picked up
+            if (offpeak_time_info.is_now_offpeak &&
+                offpeak_time_info.seconds_till_next_offpeak_start /
+                        kSecondsPerHour >
+                    16) {
+              ASSERT_OK(Put("foo", "bar5"));
+              ASSERT_OK(Flush());
+              ASSERT_OK(dbfull()->TEST_WaitForCompact());
+              ASSERT_EQ(1, periodic_compactions);
+            }
+          }
+
+          // If the file has not been picked up yet (no offpeak set, or offpeak
+          // set but then unset before the file becomes eligible)
+          if (periodic_compactions == 0) {
+            // move clock forward by one more day
+            mock_clock->MockSleepForSeconds(1 * kSecondsPerDay);
+            ASSERT_OK(Put("foo", "bar6"));
+            ASSERT_OK(Flush());
+            ASSERT_OK(dbfull()->TEST_WaitForCompact());
+          }
+        }
+        ASSERT_EQ(1, periodic_compactions);
+        ASSERT_EQ(0, start_level);
+        ASSERT_EQ(4, output_level);
+        Destroy(opts);
+
+        periodic_compactions = 0;
+      }
+    }
+  }
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 

db/db_wal_test.cc

@@ -1195,7 +1195,7 @@ TEST_F(DBWALTest, DISABLED_FullPurgePreservesLogPendingReuse) {
     ROCKSDB_NAMESPACE::port::Thread thread([&]() {
       TEST_SYNC_POINT(
           "DBWALTest::FullPurgePreservesLogPendingReuse:PreFullPurge");
-      ASSERT_OK(db_->EnableFileDeletions(true));
+      ASSERT_OK(db_->EnableFileDeletions(/*force=*/true));
       TEST_SYNC_POINT(
           "DBWALTest::FullPurgePreservesLogPendingReuse:PostFullPurge");
     });
@@ -1541,7 +1541,9 @@ class RecoveryTestHelper {
         test->dbname_, &db_options, file_options, table_cache.get(),
         &write_buffer_manager, &write_controller,
         /*block_cache_tracer=*/nullptr,
-        /*io_tracer=*/nullptr, /*db_id*/ "", /*db_session_id*/ ""));
+        /*io_tracer=*/nullptr, /*db_id=*/"", /*db_session_id=*/"",
+        options.daily_offpeak_time_utc,
+        /*error_handler=*/nullptr));
 
     wal_manager.reset(
         new WalManager(db_options, file_options, /*io_tracer=*/nullptr));
@@ -2228,6 +2230,7 @@ TEST_P(DBWALTestWithParamsVaryingRecoveryMode,
       data.push_back(
           std::make_pair(iter->key().ToString(), iter->value().ToString()));
     }
+    EXPECT_OK(iter->status());
     delete iter;
     return data;
   };

db/db_with_timestamp_basic_test.cc

@@ -525,6 +525,7 @@ TEST_F(DBBasicTestWithTimestamp, SimpleIterate) {
       CheckIterUserEntry(it.get(), Key1(key), kTypeValue,
                          "value" + std::to_string(i), write_timestamps[i]);
     }
+    ASSERT_OK(it->status());
     ASSERT_EQ(static_cast<size_t>(kMaxKey) - start_keys[i] + 1, count);
 
     // SeekToFirst()/SeekToLast() with lower/upper bounds.
@@ -544,6 +545,7 @@ TEST_F(DBBasicTestWithTimestamp, SimpleIterate) {
         CheckIterUserEntry(it.get(), Key1(key), kTypeValue,
                            "value" + std::to_string(i), write_timestamps[i]);
       }
+      ASSERT_OK(it->status());
       ASSERT_EQ(r - std::max(l, start_keys[i]), count);
 
       for (it->SeekToLast(), key = std::min(r, kMaxKey + 1), count = 0;
@@ -551,6 +553,7 @@ TEST_F(DBBasicTestWithTimestamp, SimpleIterate) {
         CheckIterUserEntry(it.get(), Key1(key - 1), kTypeValue,
                            "value" + std::to_string(i), write_timestamps[i]);
       }
+      ASSERT_OK(it->status());
       l += (kMaxKey / 100);
       r -= (kMaxKey / 100);
     }
@@ -733,6 +736,7 @@ TEST_P(DBBasicTestWithTimestampTableOptions, GetAndMultiGet) {
       ASSERT_EQ(it->value(), value_from_get);
       ASSERT_EQ(Timestamp(1, 0), timestamp);
     }
+    ASSERT_OK(it->status());
 
     // verify MultiGet()
     constexpr uint64_t step = 2;
@@ -1065,6 +1069,7 @@ TEST_F(DBBasicTestWithTimestamp, SimpleForwardIterateLowerTsBound) {
                        write_timestamps[i - 1]);
       }
     }
+    ASSERT_OK(it->status());
     size_t expected_count = kMaxKey + 1;
     ASSERT_EQ(expected_count, count);
   }
@@ -1143,6 +1148,7 @@ TEST_F(DBBasicTestWithTimestamp, BackwardIterateLowerTsBound) {
                        write_timestamps[1]);
       }
     }
+    ASSERT_OK(it->status());
     size_t expected_count = kMaxKey + 1;
     ASSERT_EQ(expected_count, count);
   }
@@ -1173,6 +1179,7 @@ TEST_F(DBBasicTestWithTimestamp, BackwardIterateLowerTsBound) {
       CheckIterEntry(it.get(), Key1(key), kTypeDeletionWithTimestamp, Slice(),
                      write_timestamp);
     }
+    ASSERT_OK(it->status());
     ASSERT_EQ(kMaxKey + 1, count);
   }
   Close();
@@ -1278,6 +1285,7 @@ TEST_F(DBBasicTestWithTimestamp, BackwardIterateLowerTsBound_Reseek) {
       CheckIterEntry(it.get(), "a", kTypeValue, "v" + std::to_string(4 + i),
                      Timestamp(4 + i, 0));
     }
+    ASSERT_OK(it->status());
   }
 
   Close();
@@ -3145,6 +3153,7 @@ TEST_P(DBBasicTestWithTimestampPrefixSeek, IterateWithPrefix) {
                          "value" + std::to_string(i), write_ts_list[i]);
       iter->Next();
       ASSERT_FALSE(iter->Valid());
+      ASSERT_OK(iter->status());
 
       // Seek to kMinKey
       iter->Seek(Key1(kMinKey));
@@ -3152,6 +3161,7 @@ TEST_P(DBBasicTestWithTimestampPrefixSeek, IterateWithPrefix) {
                          "value" + std::to_string(i), write_ts_list[i]);
       iter->Prev();
       ASSERT_FALSE(iter->Valid());
+      ASSERT_OK(iter->status());
     }
     const std::vector<uint64_t> targets = {kMinKey, kMinKey + 0x10,
                                            kMinKey + 0x100, kMaxKey};
@@ -3190,6 +3200,7 @@ TEST_P(DBBasicTestWithTimestampPrefixSeek, IterateWithPrefix) {
           ++expected_key;
           it->Next();
         }
+        ASSERT_OK(it->status());
         ASSERT_EQ(expected_ub - targets[j] + 1, count);
 
         count = 0;
@@ -3208,6 +3219,7 @@ TEST_P(DBBasicTestWithTimestampPrefixSeek, IterateWithPrefix) {
           --expected_key;
           it->Prev();
         }
+        ASSERT_OK(it->status());
         ASSERT_EQ(targets[j] - std::max(expected_lb, kMinKey) + 1, count);
       }
     }
@@ -3313,6 +3325,7 @@ TEST_P(DBBasicTestWithTsIterTombstones, IterWithDelete) {
       ASSERT_EQ(Key1(key), iter->key());
       ASSERT_EQ("value1" + std::to_string(key), iter->value());
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ((kMaxKey - kMinKey + 1) / 2, count);
   }
   Close();
@@ -3932,6 +3945,7 @@ TEST_P(DBBasicTestWithTimestampTableOptions, DeleteRangeBaiscReadAndIterate) {
         ++expected;
       }
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(kNum, expected);
 
     expected = kNum / 2;
@@ -3939,6 +3953,7 @@ TEST_P(DBBasicTestWithTimestampTableOptions, DeleteRangeBaiscReadAndIterate) {
       ASSERT_EQ(Key1(expected), iter->key());
       ++expected;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(kNum, expected);
 
     expected = kRangeBegin - 1;
@@ -3946,6 +3961,7 @@ TEST_P(DBBasicTestWithTimestampTableOptions, DeleteRangeBaiscReadAndIterate) {
       ASSERT_EQ(Key1(expected), iter->key());
       --expected;
     }
+    ASSERT_OK(iter->status());
     ASSERT_EQ(-1, expected);
 
     read_ts = Timestamp(0, 0);
@@ -4227,6 +4243,7 @@ TEST_F(DBBasicTestWithTimestamp, MergeBasic) {
         ASSERT_EQ(value, it->value());
         ASSERT_EQ(write_ts_strs[i], it->timestamp());
       }
+      EXPECT_OK(it->status());
       ASSERT_EQ(kNumOfUniqKeys, key_int_val);
 
       key_int_val = kNumOfUniqKeys - 1;
@@ -4238,6 +4255,7 @@ TEST_F(DBBasicTestWithTimestamp, MergeBasic) {
         ASSERT_EQ(value, it->value());
         ASSERT_EQ(write_ts_strs[i], it->timestamp());
       }
+      ASSERT_OK(it->status());
       ASSERT_EQ(std::numeric_limits<size_t>::max(), key_int_val);
 
       value_suffix = value_suffix + "." + std::to_string(i + 1);

db/dbformat.h

@@ -355,6 +355,7 @@ class InternalKeyComparator
 
   // Same as Compare except that it excludes the value type from comparison
   int CompareKeySeq(const Slice& a, const Slice& b) const;
+  int CompareKeySeq(const ParsedInternalKey& a, const Slice& b) const;
 
   const Comparator* user_comparator() const {
     return user_comparator_.user_comparator();
@@ -976,6 +977,26 @@ inline int InternalKeyComparator::CompareKeySeq(const Slice& akey,
   return r;
 }
 
+inline int InternalKeyComparator::CompareKeySeq(const ParsedInternalKey& a,
+                                                const Slice& b) const {
+  // Order by:
+  //    increasing user key (according to user-supplied comparator)
+  //    decreasing sequence number
+  int r = user_comparator_.Compare(a.user_key, ExtractUserKey(b));
+  if (r == 0) {
+    // Shift the number to exclude the last byte which contains the value type
+    const uint64_t anum = a.sequence;
+    const uint64_t bnum =
+        DecodeFixed64(b.data() + b.size() - kNumInternalBytes) >> 8;
+    if (anum > bnum) {
+      r = -1;
+    } else if (anum < bnum) {
+      r = +1;
+    }
+  }
+  return r;
+}
+
 inline int InternalKeyComparator::Compare(const Slice& a,
                                           SequenceNumber a_global_seqno,
                                           const Slice& b,

db/deletefile_test.cc

@@ -575,6 +575,7 @@ TEST_F(DeleteFileTest, DeleteNonDefaultColumnFamily) {
       ASSERT_OK(itr->status());
       ++count;
     }
+    ASSERT_OK(itr->status());
     ASSERT_EQ(count, 1000);
   }
 
@@ -588,6 +589,7 @@ TEST_F(DeleteFileTest, DeleteNonDefaultColumnFamily) {
       ASSERT_OK(itr->status());
       ++count;
     }
+    ASSERT_OK(itr->status());
     ASSERT_EQ(count, 1000);
   }
 }

db/error_handler.cc

@@ -320,7 +320,7 @@ const Status& ErrorHandler::HandleKnownErrors(const Status& bg_err,
   // Check if recovery is currently in progress. If it is, we will save this
   // error so we can check it at the end to see if recovery succeeded or not
   if (recovery_in_prog_ && recovery_error_.ok()) {
-    recovery_error_ = new_bg_err;
+    recovery_error_ = status_to_io_status(Status(new_bg_err));
   }
 
   bool auto_recovery = auto_recovery_;
@@ -396,16 +396,6 @@ const Status& ErrorHandler::SetBGError(const Status& bg_status,
   ROCKS_LOG_WARN(db_options_.info_log, "Background IO error %s",
                  bg_io_err.ToString().c_str());
 
-  if (recovery_in_prog_ && recovery_io_error_.ok()) {
-    recovery_io_error_ = bg_io_err;
-  }
-  if (BackgroundErrorReason::kManifestWrite == reason ||
-      BackgroundErrorReason::kManifestWriteNoWAL == reason) {
-    // Always returns ok
-    ROCKS_LOG_INFO(db_options_.info_log, "Disabling File Deletions");
-    db_->DisableFileDeletionsWithLock().PermitUncheckedError();
-  }
-
   Status new_bg_io_err = bg_io_err;
   DBRecoverContext context;
   if (bg_io_err.GetScope() != IOStatus::IOErrorScope::kIOErrorScopeFile &&
@@ -502,14 +492,35 @@ const Status& ErrorHandler::SetBGError(const Status& bg_status,
       RecordTick(bg_error_stats_.get(),
                  ERROR_HANDLER_BG_IO_ERROR_COUNT_MISSPELLED);
     }
-    // HandleKnownErrors() will use recovery_error_, so ignore
-    // recovery_io_error_.
-    // TODO: Do some refactoring and use only one recovery_error_
-    recovery_io_error_.PermitUncheckedError();
     return HandleKnownErrors(new_bg_io_err, reason);
   }
 }
 
+void ErrorHandler::AddFilesToQuarantine(
+    autovector<const autovector<uint64_t>*> files_to_quarantine) {
+  db_mutex_->AssertHeld();
+  std::ostringstream quarantine_files_oss;
+  bool is_first_one = true;
+  for (const auto* files : files_to_quarantine) {
+    assert(files);
+    for (uint64_t file_number : *files) {
+      files_to_quarantine_.push_back(file_number);
+      quarantine_files_oss << (is_first_one ? "" : ", ") << file_number;
+      is_first_one = false;
+    }
+  }
+  ROCKS_LOG_INFO(db_options_.info_log,
+                 "ErrorHandler: added file numbers %s to quarantine.\n",
+                 quarantine_files_oss.str().c_str());
+}
+
+void ErrorHandler::ClearFilesToQuarantine() {
+  db_mutex_->AssertHeld();
+  files_to_quarantine_.clear();
+  ROCKS_LOG_INFO(db_options_.info_log,
+                 "ErrorHandler: cleared files in quarantine.\n");
+}
+
 Status ErrorHandler::OverrideNoSpaceError(const Status& bg_error,
                                           bool* auto_recovery) {
   if (bg_error.severity() >= Status::Severity::kFatalError) {
@@ -557,14 +568,16 @@ Status ErrorHandler::ClearBGError() {
 
   // Signal that recovery succeeded
   if (recovery_error_.ok()) {
+    assert(files_to_quarantine_.empty());
     Status old_bg_error = bg_error_;
     // old_bg_error is only for notifying listeners, so may not be checked
     old_bg_error.PermitUncheckedError();
     // Clear and check the recovery IO and BG error
+    is_db_stopped_.store(false, std::memory_order_release);
     bg_error_ = Status::OK();
-    recovery_io_error_ = IOStatus::OK();
+    recovery_error_ = IOStatus::OK();
     bg_error_.PermitUncheckedError();
-    recovery_io_error_.PermitUncheckedError();
+    recovery_error_.PermitUncheckedError();
     recovery_in_prog_ = false;
     soft_error_no_bg_work_ = false;
     EventHelpers::NotifyOnErrorRecoveryEnd(db_options_.listeners, old_bg_error,
@@ -602,14 +615,14 @@ Status ErrorHandler::RecoverFromBGError(bool is_manual) {
   if (bg_error_.severity() == Status::Severity::kSoftError &&
       recover_context_.flush_reason == FlushReason::kErrorRecovery) {
     // Simply clear the background error and return
-    recovery_error_ = Status::OK();
+    recovery_error_ = IOStatus::OK();
     return ClearBGError();
   }
 
   // Reset recovery_error_. We will use this to record any errors that happen
   // during the recovery process. While recovering, the only operations that
   // can generate background errors should be the flush operations
-  recovery_error_ = Status::OK();
+  recovery_error_ = IOStatus::OK();
   recovery_error_.PermitUncheckedError();
   Status s = db_->ResumeImpl(recover_context_);
   if (s.ok()) {
@@ -638,6 +651,13 @@ const Status& ErrorHandler::StartRecoverFromRetryableBGIOError(
   } else if (db_options_.max_bgerror_resume_count <= 0 || recovery_in_prog_) {
     // Auto resume BG error is not enabled, directly return bg_error_.
     return bg_error_;
+  } else if (end_recovery_) {
+    // Can temporarily release db mutex
+    EventHelpers::NotifyOnErrorRecoveryEnd(db_options_.listeners, bg_error_,
+                                           Status::ShutdownInProgress(),
+                                           db_mutex_);
+    db_mutex_->AssertHeld();
+    return bg_error_;
   }
   if (bg_error_stats_ != nullptr) {
     RecordTick(bg_error_stats_.get(), ERROR_HANDLER_AUTORESUME_COUNT);
@@ -645,21 +665,30 @@ const Status& ErrorHandler::StartRecoverFromRetryableBGIOError(
   ROCKS_LOG_INFO(
       db_options_.info_log,
       "ErrorHandler: Call StartRecoverFromRetryableBGIOError to resume\n");
+  // Needs to be set in the same lock hold as setting BG error, otherwise
+  // intervening writes could see a BG error without a recovery and bail out.
+  recovery_in_prog_ = true;
+
   if (recovery_thread_) {
+    // Ensure only one thread can execute the join().
+    std::unique_ptr<port::Thread> old_recovery_thread(
+        std::move(recovery_thread_));
     // In this case, if recovery_in_prog_ is false, current thread should
     // wait the previous recover thread to finish and create a new thread
     // to recover from the bg error.
     db_mutex_->Unlock();
-    recovery_thread_->join();
+    TEST_SYNC_POINT(
+        "StartRecoverFromRetryableBGIOError:BeforeWaitingForOtherThread");
+    old_recovery_thread->join();
+    TEST_SYNC_POINT(
+        "StartRecoverFromRetryableBGIOError:AfterWaitingForOtherThread");
     db_mutex_->Lock();
   }
 
-  recovery_in_prog_ = true;
-  TEST_SYNC_POINT("StartRecoverFromRetryableBGIOError::in_progress");
   recovery_thread_.reset(
       new port::Thread(&ErrorHandler::RecoverFromRetryableBGIOError, this));
 
-  if (recovery_io_error_.ok() && recovery_error_.ok()) {
+  if (recovery_error_.ok()) {
     return recovery_error_;
   } else {
     return bg_error_;
@@ -669,6 +698,7 @@ const Status& ErrorHandler::StartRecoverFromRetryableBGIOError(
 // Automatic recover from Retryable BG IO error. Must be called after db
 // mutex is released.
 void ErrorHandler::RecoverFromRetryableBGIOError() {
+  assert(recovery_in_prog_);
   TEST_SYNC_POINT("RecoverFromRetryableBGIOError:BeforeStart");
   TEST_SYNC_POINT("RecoverFromRetryableBGIOError:BeforeStart2");
   InstrumentedMutexLock l(db_mutex_);
@@ -676,9 +706,12 @@ void ErrorHandler::RecoverFromRetryableBGIOError() {
     EventHelpers::NotifyOnErrorRecoveryEnd(db_options_.listeners, bg_error_,
                                            Status::ShutdownInProgress(),
                                            db_mutex_);
+
+    recovery_in_prog_ = false;
     return;
   }
   DBRecoverContext context = recover_context_;
+  context.flush_after_recovery = true;
   int resume_count = db_options_.max_bgerror_resume_count;
   uint64_t wait_interval = db_options_.bgerror_resume_retry_interval;
   uint64_t retry_count = 0;
@@ -688,12 +721,12 @@ void ErrorHandler::RecoverFromRetryableBGIOError() {
       EventHelpers::NotifyOnErrorRecoveryEnd(db_options_.listeners, bg_error_,
                                              Status::ShutdownInProgress(),
                                              db_mutex_);
+      recovery_in_prog_ = false;
       return;
     }
     TEST_SYNC_POINT("RecoverFromRetryableBGIOError:BeforeResume0");
     TEST_SYNC_POINT("RecoverFromRetryableBGIOError:BeforeResume1");
-    recovery_io_error_ = IOStatus::OK();
-    recovery_error_ = Status::OK();
+    recovery_error_ = IOStatus::OK();
     retry_count++;
     Status s = db_->ResumeImpl(context);
     if (bg_error_stats_ != nullptr) {
@@ -713,9 +746,9 @@ void ErrorHandler::RecoverFromRetryableBGIOError() {
                                              bg_error_, db_mutex_);
       return;
     }
-    if (!recovery_io_error_.ok() &&
+    if (!recovery_error_.ok() &&
         recovery_error_.severity() <= Status::Severity::kHardError &&
-        recovery_io_error_.GetRetryable()) {
+        recovery_error_.GetRetryable()) {
       // If new BG IO error happens during auto recovery and it is retryable
       // and its severity is Hard Error or lower, the auto resmue sleep for
       // a period of time and redo auto resume if it is allowed.
@@ -724,33 +757,23 @@ void ErrorHandler::RecoverFromRetryableBGIOError() {
       int64_t wait_until = db_options_.clock->NowMicros() + wait_interval;
       cv_.TimedWait(wait_until);
     } else {
-      // There are three possibility: 1) recover_io_error is set during resume
+      // There are three possibility: 1) recovery_error_ is set during resume
       // and the error is not retryable, 2) recover is successful, 3) other
       // error happens during resume and cannot be resumed here.
-      if (recovery_io_error_.ok() && recovery_error_.ok() && s.ok()) {
+      if (recovery_error_.ok() && s.ok()) {
         // recover from the retryable IO error and no other BG errors. Clean
         // the bg_error and notify user.
         TEST_SYNC_POINT("RecoverFromRetryableBGIOError:RecoverSuccess");
-        Status old_bg_error = bg_error_;
-        is_db_stopped_.store(false, std::memory_order_release);
-        bg_error_ = Status::OK();
-        bg_error_.PermitUncheckedError();
-        EventHelpers::NotifyOnErrorRecoveryEnd(
-            db_options_.listeners, old_bg_error, bg_error_, db_mutex_);
         if (bg_error_stats_ != nullptr) {
           RecordTick(bg_error_stats_.get(),
                      ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT);
           RecordInHistogram(bg_error_stats_.get(),
                             ERROR_HANDLER_AUTORESUME_RETRY_COUNT, retry_count);
         }
-        recovery_in_prog_ = false;
-        if (soft_error_no_bg_work_) {
-          soft_error_no_bg_work_ = false;
-        }
         return;
       } else {
-        // In this case: 1) recovery_io_error is more serious or not retryable
-        // 2) other Non IO recovery_error happens. The auto recovery stops.
+        // In this case: 1) recovery_error_ is more serious or not retryable
+        // 2) other error happens. The auto recovery stops.
         recovery_in_prog_ = false;
         if (bg_error_stats_ != nullptr) {
           RecordInHistogram(bg_error_stats_.get(),
@@ -758,10 +781,7 @@ void ErrorHandler::RecoverFromRetryableBGIOError() {
         }
         EventHelpers::NotifyOnErrorRecoveryEnd(
             db_options_.listeners, bg_error_,
-            !recovery_io_error_.ok()
-                ? recovery_io_error_
-                : (!recovery_error_.ok() ? recovery_error_ : s),
-            db_mutex_);
+            !recovery_error_.ok() ? recovery_error_ : s, db_mutex_);
         return;
       }
     }
@@ -781,7 +801,7 @@ void ErrorHandler::RecoverFromRetryableBGIOError() {
 
 void ErrorHandler::CheckAndSetRecoveryAndBGError(const Status& bg_err) {
   if (recovery_in_prog_ && recovery_error_.ok()) {
-    recovery_error_ = bg_err;
+    recovery_error_ = status_to_io_status(Status(bg_err));
   }
   if (bg_err.severity() > bg_error_.severity()) {
     bg_error_ = bg_err;
@@ -797,12 +817,16 @@ void ErrorHandler::EndAutoRecovery() {
   if (!end_recovery_) {
     end_recovery_ = true;
   }
-  cv_.SignalAll();
-  db_mutex_->Unlock();
   if (recovery_thread_) {
-    recovery_thread_->join();
+    // Ensure only one thread can execute the join().
+    std::unique_ptr<port::Thread> old_recovery_thread(
+        std::move(recovery_thread_));
+    db_mutex_->Unlock();
+    cv_.SignalAll();
+    old_recovery_thread->join();
+    db_mutex_->Lock();
   }
-  db_mutex_->Lock();
+  TEST_SYNC_POINT("PostEndAutoRecovery");
   return;
 }
 

db/error_handler.h

@@ -4,11 +4,14 @@
 //  (found in the LICENSE.Apache file in the root directory).
 #pragma once
 
+#include <sstream>
+
 #include "monitoring/instrumented_mutex.h"
 #include "options/db_options.h"
 #include "rocksdb/io_status.h"
 #include "rocksdb/listener.h"
 #include "rocksdb/status.h"
+#include "util/autovector.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -19,10 +22,13 @@ class DBImpl;
 // FlushReason, which tells the flush job why this flush is called.
 struct DBRecoverContext {
   FlushReason flush_reason;
+  bool flush_after_recovery;
 
-  DBRecoverContext() : flush_reason(FlushReason::kErrorRecovery) {}
-
-  DBRecoverContext(FlushReason reason) : flush_reason(reason) {}
+  DBRecoverContext()
+      : flush_reason(FlushReason::kErrorRecovery),
+        flush_after_recovery(false) {}
+  DBRecoverContext(FlushReason reason)
+      : flush_reason(reason), flush_after_recovery(false) {}
 };
 
 class ErrorHandler {
@@ -43,7 +49,6 @@ class ErrorHandler {
     // Clear the checked flag for uninitialized errors
     bg_error_.PermitUncheckedError();
     recovery_error_.PermitUncheckedError();
-    recovery_io_error_.PermitUncheckedError();
   }
 
   void EnableAutoRecovery() { auto_recovery_ = true; }
@@ -78,16 +83,23 @@ class ErrorHandler {
 
   void EndAutoRecovery();
 
+  void AddFilesToQuarantine(
+      autovector<const autovector<uint64_t>*> files_to_quarantine);
+
+  const autovector<uint64_t>& GetFilesToQuarantine() const {
+    db_mutex_->AssertHeld();
+    return files_to_quarantine_;
+  }
+
+  void ClearFilesToQuarantine();
+
  private:
   DBImpl* db_;
   const ImmutableDBOptions& db_options_;
   Status bg_error_;
   // A separate Status variable used to record any errors during the
   // recovery process from hard errors
-  Status recovery_error_;
-  // A separate IO Status variable used to record any IO errors during
-  // the recovery process. At the same time, recovery_error_ is also set.
-  IOStatus recovery_io_error_;
+  IOStatus recovery_error_;
   // The condition variable used with db_mutex during auto resume for time
   // wait.
   InstrumentedCondVar cv_;
@@ -109,6 +121,13 @@ class ErrorHandler {
   // The pointer of DB statistics.
   std::shared_ptr<Statistics> bg_error_stats_;
 
+  // During recovery from manifest IO errors, files whose VersionEdits entries
+  // could be in an ambiguous state are quarantined and file deletion refrain
+  // from deleting them. Successful recovery will clear this vector. Files are
+  // added to this vector while DB mutex was locked, this data structure is
+  // unsorted.
+  autovector<uint64_t> files_to_quarantine_;
+
   const Status& HandleKnownErrors(const Status& bg_err,
                                   BackgroundErrorReason reason);
   Status OverrideNoSpaceError(const Status& bg_error, bool* auto_recovery);

db/error_handler_fs_test.cc

@@ -7,12 +7,15 @@
 // 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 <memory>
+
 #include "db/db_test_util.h"
 #include "file/sst_file_manager_impl.h"
 #include "port/stack_trace.h"
 #include "rocksdb/io_status.h"
 #include "rocksdb/sst_file_manager.h"
 #include "test_util/sync_point.h"
+#include "test_util/testharness.h"
 #include "util/random.h"
 #include "utilities/fault_injection_env.h"
 #include "utilities/fault_injection_fs.h"
@@ -152,9 +155,9 @@ class ErrorHandlerFSListener : public EventListener {
   FaultInjectionTestFS* fault_fs_;
 };
 
-TEST_F(DBErrorHandlingFSTest, FLushWriteError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWriteError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -197,9 +200,9 @@ TEST_F(DBErrorHandlingFSTest, FLushWriteError) {
 // All the NoSpace IOError will be handled as the regular BG Error no matter the
 // retryable flag is set of not. So the auto resume for retryable IO Error will
 // not be triggered. Also, it is mapped as hard error.
-TEST_F(DBErrorHandlingFSTest, FLushWriteNoSpaceError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWriteNoSpaceError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -241,9 +244,9 @@ TEST_F(DBErrorHandlingFSTest, FLushWriteNoSpaceError) {
   Destroy(options);
 }
 
-TEST_F(DBErrorHandlingFSTest, FLushWriteRetryableError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWriteRetryableError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -315,9 +318,9 @@ TEST_F(DBErrorHandlingFSTest, FLushWriteRetryableError) {
   Destroy(options);
 }
 
-TEST_F(DBErrorHandlingFSTest, FLushWriteFileScopeError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWriteFileScopeError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -399,9 +402,9 @@ TEST_F(DBErrorHandlingFSTest, FLushWriteFileScopeError) {
   Destroy(options);
 }
 
-TEST_F(DBErrorHandlingFSTest, FLushWALWriteRetryableError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWALWriteRetryableError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -443,9 +446,9 @@ TEST_F(DBErrorHandlingFSTest, FLushWALWriteRetryableError) {
   Destroy(options);
 }
 
-TEST_F(DBErrorHandlingFSTest, FLushWALAtomicWriteRetryableError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWALAtomicWriteRetryableError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -489,9 +492,9 @@ TEST_F(DBErrorHandlingFSTest, FLushWALAtomicWriteRetryableError) {
 }
 
 // The flush error is injected before we finish the table build
-TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableError1) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWritNoWALRetryableError1) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -545,9 +548,9 @@ TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableError1) {
 }
 
 // The retryable IO error is injected before we sync table
-TEST_F(DBErrorHandlingFSTest, FLushWriteNoWALRetryableError2) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWriteNoWALRetryableError2) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -589,9 +592,9 @@ TEST_F(DBErrorHandlingFSTest, FLushWriteNoWALRetryableError2) {
 }
 
 // The retryable IO error is injected before we close the table file
-TEST_F(DBErrorHandlingFSTest, FLushWriteNoWALRetryableError3) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_F(DBErrorHandlingFSTest, FlushWriteNoWALRetryableError3) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -633,8 +636,8 @@ TEST_F(DBErrorHandlingFSTest, FLushWriteNoWALRetryableError3) {
 }
 
 TEST_F(DBErrorHandlingFSTest, ManifestWriteError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -658,6 +661,7 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteError) {
   SyncPoint::GetInstance()->EnableProcessing();
   s = Flush();
   ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
+  ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
   SyncPoint::GetInstance()->ClearAllCallBacks();
   SyncPoint::GetInstance()->DisableProcessing();
   fault_fs_->SetFilesystemActive(true);
@@ -666,6 +670,7 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteError) {
 
   new_manifest = GetManifestNameFromLiveFiles();
   ASSERT_NE(new_manifest, old_manifest);
+  ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
 
   Reopen(options);
   ASSERT_EQ("val", Get(Key(0)));
@@ -674,8 +679,8 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -701,6 +706,7 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableError) {
   SyncPoint::GetInstance()->EnableProcessing();
   s = Flush();
   ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
+  ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
   SyncPoint::GetInstance()->ClearAllCallBacks();
   SyncPoint::GetInstance()->DisableProcessing();
   fault_fs_->SetFilesystemActive(true);
@@ -709,6 +715,7 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableError) {
 
   new_manifest = GetManifestNameFromLiveFiles();
   ASSERT_NE(new_manifest, old_manifest);
+  ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
 
   Reopen(options);
   ASSERT_EQ("val", Get(Key(0)));
@@ -717,8 +724,8 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, ManifestWriteFileScopeError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -746,6 +753,7 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteFileScopeError) {
       [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
   SyncPoint::GetInstance()->EnableProcessing();
   s = Flush();
+  ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
   ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
   SyncPoint::GetInstance()->ClearAllCallBacks();
   SyncPoint::GetInstance()->DisableProcessing();
@@ -755,6 +763,7 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteFileScopeError) {
 
   new_manifest = GetManifestNameFromLiveFiles();
   ASSERT_NE(new_manifest, old_manifest);
+  ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
 
   Reopen(options);
   ASSERT_EQ("val", Get(Key(0)));
@@ -763,8 +772,8 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteFileScopeError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -792,6 +801,7 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableError) {
   SyncPoint::GetInstance()->EnableProcessing();
   s = Flush();
   ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
+  ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
   SyncPoint::GetInstance()->ClearAllCallBacks();
   SyncPoint::GetInstance()->DisableProcessing();
   fault_fs_->SetFilesystemActive(true);
@@ -800,6 +810,7 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableError) {
 
   new_manifest = GetManifestNameFromLiveFiles();
   ASSERT_NE(new_manifest, old_manifest);
+  ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
 
   Reopen(options);
   ASSERT_EQ("val", Get(Key(0)));
@@ -808,8 +819,8 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, DoubleManifestWriteError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -833,11 +844,13 @@ TEST_F(DBErrorHandlingFSTest, DoubleManifestWriteError) {
   SyncPoint::GetInstance()->EnableProcessing();
   s = Flush();
   ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
+  ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
   fault_fs_->SetFilesystemActive(true);
 
   // This Resume() will attempt to create a new manifest file and fail again
   s = dbfull()->Resume();
   ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
+  ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
   fault_fs_->SetFilesystemActive(true);
   SyncPoint::GetInstance()->ClearAllCallBacks();
   SyncPoint::GetInstance()->DisableProcessing();
@@ -848,6 +861,7 @@ TEST_F(DBErrorHandlingFSTest, DoubleManifestWriteError) {
 
   new_manifest = GetManifestNameFromLiveFiles();
   ASSERT_NE(new_manifest, old_manifest);
+  ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
 
   Reopen(options);
   ASSERT_EQ("val", Get(Key(0)));
@@ -860,8 +874,8 @@ TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteError) {
     ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
     return;
   }
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -914,6 +928,7 @@ TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteError) {
   fault_fs_->SetFilesystemActive(true);
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
   TEST_SYNC_POINT("CompactionManifestWriteError:1");
+  ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
   TEST_SYNC_POINT("CompactionManifestWriteError:2");
 
   s = dbfull()->TEST_WaitForCompact();
@@ -922,6 +937,7 @@ TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteError) {
 
   new_manifest = GetManifestNameFromLiveFiles();
   ASSERT_NE(new_manifest, old_manifest);
+  ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
   Reopen(options);
   ASSERT_EQ("val", Get(Key(0)));
   ASSERT_EQ("val", Get(Key(1)));
@@ -930,8 +946,8 @@ TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteRetryableError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -980,6 +996,7 @@ TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteRetryableError) {
   ASSERT_OK(s);
 
   TEST_SYNC_POINT("CompactionManifestWriteError:0");
+  ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
   TEST_SYNC_POINT("CompactionManifestWriteError:1");
 
   s = dbfull()->TEST_WaitForCompact();
@@ -993,6 +1010,7 @@ TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteRetryableError) {
 
   new_manifest = GetManifestNameFromLiveFiles();
   ASSERT_NE(new_manifest, old_manifest);
+  ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
 
   Reopen(options);
   ASSERT_EQ("val", Get(Key(0)));
@@ -1002,8 +1020,8 @@ TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteRetryableError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, CompactionWriteError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1044,8 +1062,8 @@ TEST_F(DBErrorHandlingFSTest, CompactionWriteError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, DISABLED_CompactionWriteRetryableError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1091,8 +1109,8 @@ TEST_F(DBErrorHandlingFSTest, DISABLED_CompactionWriteRetryableError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, DISABLED_CompactionWriteFileScopeError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1183,8 +1201,8 @@ TEST_F(DBErrorHandlingFSTest, AutoRecoverFlushError) {
     ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
     return;
   }
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1228,8 +1246,8 @@ TEST_F(DBErrorHandlingFSTest, AutoRecoverFlushError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, FailRecoverFlushError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1257,8 +1275,8 @@ TEST_F(DBErrorHandlingFSTest, WALWriteError) {
     ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
     return;
   }
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1330,8 +1348,8 @@ TEST_F(DBErrorHandlingFSTest, WALWriteError) {
 }
 
 TEST_F(DBErrorHandlingFSTest, WALWriteRetryableError) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1425,8 +1443,8 @@ TEST_F(DBErrorHandlingFSTest, MultiCFWALWriteError) {
     ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
     return;
   }
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1757,10 +1775,10 @@ TEST_F(DBErrorHandlingFSTest, MultiDBVariousErrors) {
 // to soft error and trigger auto resume. During auto resume, SwitchMemtable
 // is disabled to avoid small SST tables. Write can still be applied before
 // the bg error is cleaned unless the memtable is full.
-TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableErrorAutoRecover1) {
+TEST_F(DBErrorHandlingFSTest, FlushWritNoWALRetryableErrorAutoRecover1) {
   // Activate the FS before the first resume
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1781,7 +1799,7 @@ TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableErrorAutoRecover1) {
   ASSERT_OK(Put(Key(1), "val1", wo));
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
       {{"RecoverFromRetryableBGIOError:LoopOut",
-        "FLushWritNoWALRetryableeErrorAutoRecover1:1"}});
+        "FlushWritNoWALRetryableeErrorAutoRecover1:1"}});
   SyncPoint::GetInstance()->SetCallBack(
       "BuildTable:BeforeFinishBuildTable",
       [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
@@ -1790,7 +1808,7 @@ TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableErrorAutoRecover1) {
   s = Flush();
   ASSERT_EQ("val1", Get(Key(1)));
   ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
-  TEST_SYNC_POINT("FLushWritNoWALRetryableeErrorAutoRecover1:1");
+  TEST_SYNC_POINT("FlushWritNoWALRetryableeErrorAutoRecover1:1");
   ASSERT_EQ("val1", Get(Key(1)));
   ASSERT_EQ("val1", Get(Key(1)));
   SyncPoint::GetInstance()->DisableProcessing();
@@ -1827,10 +1845,94 @@ TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableErrorAutoRecover1) {
   Destroy(options);
 }
 
-TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableErrorAutoRecover2) {
+TEST_F(DBErrorHandlingFSTest, MultipleRecoveryThreads) {
+  // This test creates a scenario where second write's recovery can get started
+  // while mutex is released for a short period during
+  // NotifyOnErrorRecoveryEnd() from the first write's recovery. This is to make
+  // sure RecoverFromRetryableBGIOError() from the second write's recovery
+  // thread does not start with recovery_in_prog_ = false;
+
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
+  Options options = GetDefaultOptions();
+  options.env = fault_env_.get();
+  options.create_if_missing = true;
+  options.listeners.emplace_back(listener);
+  options.max_bgerror_resume_count = 100;
+  options.bgerror_resume_retry_interval = 1000000;  // 1 second
+  options.statistics = CreateDBStatistics();
+
+  listener->EnableAutoRecovery(false);
+  DestroyAndReopen(options);
+
+  IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
+  error_msg.SetRetryable(true);
+
+  WriteOptions wo = WriteOptions();
+  wo.disableWAL = true;
+  fault_fs_->SetFilesystemActive(false, error_msg);
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+      {{"NotifyOnErrorRecoveryEnd:MutexUnlocked:1",
+        "MultipleRecoveryThreads:1"},
+       {"MultipleRecoveryThreads:2",
+        "NotifyOnErrorRecoveryEnd:MutexUnlocked:2"},
+       {"StartRecoverFromRetryableBGIOError:BeforeWaitingForOtherThread",
+        "MultipleRecoveryThreads:3"},
+       {"RecoverFromRetryableBGIOError:RecoverSuccess",
+        "MultipleRecoveryThreads:4"},
+       {"MultipleRecoveryThreads:4",
+        "StartRecoverFromRetryableBGIOError:AfterWaitingForOtherThread"}});
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  // First write with read fault injected and recovery will start
+  {
+    ASSERT_OK(Put(Key(1), "val1", wo));
+    Status s = Flush();
+    ASSERT_NOK(s);
+  }
+  // Remove read fault injection so that first recovery can go through
+  fault_fs_->SetFilesystemActive(true);
+
+  // At this point, first recovery is now at NotifyOnErrorRecoveryEnd. Mutex is
+  // released.
+  TEST_SYNC_POINT("MultipleRecoveryThreads:1");
+
+  ROCKSDB_NAMESPACE::port::Thread second_write([&] {
+    // Second write with read fault injected
+    fault_fs_->SetFilesystemActive(false, error_msg);
+    ASSERT_OK(Put(Key(2), "val2", wo));
+    Status s = Flush();
+    ASSERT_NOK(s);
+  });
+  // Second bg thread before waiting for the first thread's recovery thread
+  TEST_SYNC_POINT("MultipleRecoveryThreads:3");
+  // First thread's recovery thread continues
+  TEST_SYNC_POINT("MultipleRecoveryThreads:2");
+  // Wait for the first thread's recovery to finish
+  // (this sets recovery_in_prog_ = false)
+  // And second thread continues and starts recovery thread
+  TEST_SYNC_POINT("MultipleRecoveryThreads:4");
+  second_write.join();
+  // Remove error injection so that second thread recovery can go through
+  fault_fs_->SetFilesystemActive(true);
+
+  // Set up sync point so that we can wait for the recovery thread to finish
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+      {{"RecoverFromRetryableBGIOError:RecoverSuccess",
+        "MultipleRecoveryThreads:6"}});
+
+  // Wait for the second thread's recovery to be done
+  TEST_SYNC_POINT("MultipleRecoveryThreads:6");
+
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  Destroy(options);
+}
+
+TEST_F(DBErrorHandlingFSTest, FlushWritNoWALRetryableErrorAutoRecover2) {
   // Activate the FS before the first resume
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1888,10 +1990,10 @@ TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableErrorAutoRecover2) {
 
 // Auto resume fromt the flush retryable IO error. Activate the FS before the
 // first resume. Resume is successful
-TEST_F(DBErrorHandlingFSTest, FLushWritRetryableErrorAutoRecover1) {
+TEST_F(DBErrorHandlingFSTest, FlushWritRetryableErrorAutoRecover1) {
   // Activate the FS before the first resume
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1930,10 +2032,10 @@ TEST_F(DBErrorHandlingFSTest, FLushWritRetryableErrorAutoRecover1) {
 
 // Auto resume fromt the flush retryable IO error and set the retry limit count.
 // Never activate the FS and auto resume should fail at the end
-TEST_F(DBErrorHandlingFSTest, FLushWritRetryableErrorAutoRecover2) {
+TEST_F(DBErrorHandlingFSTest, FlushWritRetryableErrorAutoRecover2) {
   // Fail all the resume and let user to resume
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -1950,18 +2052,18 @@ TEST_F(DBErrorHandlingFSTest, FLushWritRetryableErrorAutoRecover2) {
 
   ASSERT_OK(Put(Key(1), "val1"));
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
-      {{"FLushWritRetryableeErrorAutoRecover2:0",
+      {{"FlushWritRetryableeErrorAutoRecover2:0",
         "RecoverFromRetryableBGIOError:BeforeStart"},
        {"RecoverFromRetryableBGIOError:LoopOut",
-        "FLushWritRetryableeErrorAutoRecover2:1"}});
+        "FlushWritRetryableeErrorAutoRecover2:1"}});
   SyncPoint::GetInstance()->SetCallBack(
       "BuildTable:BeforeFinishBuildTable",
       [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
   SyncPoint::GetInstance()->EnableProcessing();
   s = Flush();
   ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
-  TEST_SYNC_POINT("FLushWritRetryableeErrorAutoRecover2:0");
-  TEST_SYNC_POINT("FLushWritRetryableeErrorAutoRecover2:1");
+  TEST_SYNC_POINT("FlushWritRetryableeErrorAutoRecover2:0");
+  TEST_SYNC_POINT("FlushWritRetryableeErrorAutoRecover2:1");
   fault_fs_->SetFilesystemActive(true);
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
   SyncPoint::GetInstance()->DisableProcessing();
@@ -1983,8 +2085,8 @@ TEST_F(DBErrorHandlingFSTest, FLushWritRetryableErrorAutoRecover2) {
 // Fail the first resume and let the second resume be successful.
 TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableErrorAutoRecover) {
   // Fail the first resume and let the second resume be successful
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2036,8 +2138,8 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableErrorAutoRecover) {
 
 TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableErrorAutoRecover) {
   // Fail the first resume and let the second resume be successful
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2091,8 +2193,8 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableErrorAutoRecover) {
 
 TEST_F(DBErrorHandlingFSTest,
        CompactionManifestWriteRetryableErrorAutoRecover) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2181,8 +2283,8 @@ TEST_F(DBErrorHandlingFSTest, CompactionWriteRetryableErrorAutoRecover) {
   // compaction, the FS is set to active and compaction is successful, so
   // the test will hit the CompactionJob::FinishCompactionOutputFile1 sync
   // point.
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2235,8 +2337,8 @@ TEST_F(DBErrorHandlingFSTest, CompactionWriteRetryableErrorAutoRecover) {
 }
 
 TEST_F(DBErrorHandlingFSTest, WALWriteRetryableErrorAutoRecover1) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2338,8 +2440,8 @@ TEST_F(DBErrorHandlingFSTest, WALWriteRetryableErrorAutoRecover1) {
 
 TEST_F(DBErrorHandlingFSTest, WALWriteRetryableErrorAutoRecover2) {
   // Fail the first recover and try second time.
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2439,10 +2541,10 @@ TEST_F(DBErrorHandlingFSTest, WALWriteRetryableErrorAutoRecover2) {
 
 // Fail auto resume from a flush retryable error and verify that
 // OnErrorRecoveryEnd listener callback is called
-TEST_F(DBErrorHandlingFSTest, FLushWritRetryableErrorAbortRecovery) {
+TEST_F(DBErrorHandlingFSTest, FlushWritRetryableErrorAbortRecovery) {
   // Activate the FS before the first resume
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2473,6 +2575,59 @@ TEST_F(DBErrorHandlingFSTest, FLushWritRetryableErrorAbortRecovery) {
   Destroy(options);
 }
 
+TEST_F(DBErrorHandlingFSTest, FlushErrorRecoveryRaceWithDBDestruction) {
+  Options options = GetDefaultOptions();
+  options.env = fault_env_.get();
+  options.create_if_missing = true;
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
+  options.listeners.emplace_back(listener);
+  DestroyAndReopen(options);
+  ASSERT_OK(Put("k1", "val"));
+
+  // Inject retryable flush error
+  bool error_set = false;
+  SyncPoint::GetInstance()->SetCallBack(
+      "BuildTable:BeforeOutputValidation", [&](void*) {
+        if (error_set) {
+          return;
+        }
+        IOStatus st = IOStatus::IOError("Injected");
+        st.SetRetryable(true);
+        fault_fs_->SetFilesystemActive(false, st);
+        error_set = true;
+      });
+
+  port::Thread db_close_thread;
+  SyncPoint::GetInstance()->SetCallBack(
+      "BuildTable:BeforeDeleteFile", [&](void*) {
+        // Clear retryable flush error injection
+        fault_fs_->SetFilesystemActive(true);
+
+        // Coerce race between ending auto recovery in db destruction and flush
+        // error recovery
+        ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+            {{"PostEndAutoRecovery", "FlushJob::WriteLevel0Table"}});
+        db_close_thread = port::Thread([&] { Close(); });
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  Status s = Flush();
+  ASSERT_NOK(s);
+
+  int placeholder = 1;
+  listener->WaitForRecovery(placeholder);
+  ASSERT_TRUE(listener->new_bg_error().IsShutdownInProgress());
+
+  // Prior to the fix, the db close will crash due to the recovery thread for
+  // flush error is not joined by the time of destruction.
+  db_close_thread.join();
+
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  Destroy(options);
+}
+
 TEST_F(DBErrorHandlingFSTest, FlushReadError) {
   std::shared_ptr<ErrorHandlerFSListener> listener =
       std::make_shared<ErrorHandlerFSListener>();
@@ -2680,9 +2835,9 @@ TEST_F(DBErrorHandlingFSTest, CompactionReadRetryableErrorAutoRecover) {
 class DBErrorHandlingFencingTest : public DBErrorHandlingFSTest,
                                    public testing::WithParamInterface<bool> {};
 
-TEST_P(DBErrorHandlingFencingTest, FLushWriteFenced) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+TEST_P(DBErrorHandlingFencingTest, FlushWriteFenced) {
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2709,8 +2864,8 @@ TEST_P(DBErrorHandlingFencingTest, FLushWriteFenced) {
 }
 
 TEST_P(DBErrorHandlingFencingTest, ManifestWriteFenced) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2744,8 +2899,8 @@ TEST_P(DBErrorHandlingFencingTest, ManifestWriteFenced) {
 }
 
 TEST_P(DBErrorHandlingFencingTest, CompactionWriteFenced) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;
@@ -2785,8 +2940,8 @@ TEST_P(DBErrorHandlingFencingTest, CompactionWriteFenced) {
 }
 
 TEST_P(DBErrorHandlingFencingTest, WALWriteFenced) {
-  std::shared_ptr<ErrorHandlerFSListener> listener(
-      new ErrorHandlerFSListener());
+  std::shared_ptr<ErrorHandlerFSListener> listener =
+      std::make_shared<ErrorHandlerFSListener>();
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
   options.create_if_missing = true;

db/event_helpers.cc

@@ -230,6 +230,8 @@ void EventHelpers::NotifyOnErrorRecoveryEnd(
     db_mutex->AssertHeld();
     // release lock while notifying events
     db_mutex->Unlock();
+    TEST_SYNC_POINT("NotifyOnErrorRecoveryEnd:MutexUnlocked:1");
+    TEST_SYNC_POINT("NotifyOnErrorRecoveryEnd:MutexUnlocked:2");
     for (auto& listener : listeners) {
       BackgroundErrorRecoveryInfo info;
       info.old_bg_error = old_bg_error;

db/external_sst_file_basic_test.cc

@@ -1548,6 +1548,11 @@ TEST_F(ExternalSSTFileBasicTest, RangeDeletionEndComesBeforeStart) {
 }
 
 TEST_P(ExternalSSTFileBasicTest, IngestFileWithBadBlockChecksum) {
+  bool verify_checksums_before_ingest = std::get<1>(GetParam());
+  if (!verify_checksums_before_ingest) {
+    ROCKSDB_GTEST_BYPASS("Bypassing test when !verify_checksums_before_ingest");
+    return;
+  }
   bool change_checksum_called = false;
   const auto& change_checksum = [&](void* arg) {
     if (!change_checksum_called) {
@@ -1565,24 +1570,20 @@ TEST_P(ExternalSSTFileBasicTest, IngestFileWithBadBlockChecksum) {
   SyncPoint::GetInstance()->EnableProcessing();
   int file_id = 0;
   bool write_global_seqno = std::get<0>(GetParam());
-  bool verify_checksums_before_ingest = std::get<1>(GetParam());
   do {
     Options options = CurrentOptions();
     DestroyAndReopen(options);
     std::map<std::string, std::string> true_data;
     Status s = GenerateAndAddExternalFile(
         options, {1, 2, 3, 4, 5, 6}, ValueType::kTypeValue, file_id++,
-        write_global_seqno, verify_checksums_before_ingest, &true_data);
-    if (verify_checksums_before_ingest) {
-      ASSERT_NOK(s);
-    } else {
-      ASSERT_OK(s);
-    }
+        write_global_seqno, /*verify_checksums_before_ingest=*/true,
+        &true_data);
+    ASSERT_NOK(s);
     change_checksum_called = false;
   } while (ChangeOptionsForFileIngestionTest());
 }
 
-TEST_P(ExternalSSTFileBasicTest, IngestFileWithFirstByteTampered) {
+TEST_P(ExternalSSTFileBasicTest, IngestFileWithCorruptedDataBlock) {
   if (!random_rwfile_supported_) {
     ROCKSDB_GTEST_SKIP("Test requires NewRandomRWFile support");
     return;
@@ -1590,15 +1591,21 @@ TEST_P(ExternalSSTFileBasicTest, IngestFileWithFirstByteTampered) {
   SyncPoint::GetInstance()->DisableProcessing();
   int file_id = 0;
   EnvOptions env_options;
+  Random rnd(301);
   do {
     Options options = CurrentOptions();
+    options.compression = kNoCompression;
+    BlockBasedTableOptions table_options;
+    table_options.block_size = 4 * 1024;
+    options.table_factory.reset(NewBlockBasedTableFactory(table_options));
     std::string file_path = sst_files_dir_ + std::to_string(file_id++);
     SstFileWriter sst_file_writer(env_options, options);
     Status s = sst_file_writer.Open(file_path);
     ASSERT_OK(s);
+    // This should write more than 2 data blocks.
     for (int i = 0; i != 100; ++i) {
       std::string key = Key(i);
-      std::string value = Key(i) + std::to_string(0);
+      std::string value = rnd.RandomString(200);
       ASSERT_OK(sst_file_writer.Put(key, value));
     }
     ASSERT_OK(sst_file_writer.Finish());
@@ -1609,11 +1616,11 @@ TEST_P(ExternalSSTFileBasicTest, IngestFileWithFirstByteTampered) {
       ASSERT_GT(file_size, 8);
       std::unique_ptr<RandomRWFile> rwfile;
       ASSERT_OK(env_->NewRandomRWFile(file_path, &rwfile, EnvOptions()));
-      // Manually corrupt the file
-      // We deterministically corrupt the first byte because we currently
-      // cannot choose a random offset. The reason for this limitation is that
-      // we do not checksum property block at present.
-      const uint64_t offset = 0;
+      // Corrupt the second data block.
+      // We need to corrupt a non-first and non-last data block
+      // since we access them to get smallest and largest internal
+      // key in the file in GetIngestedFileInfo().
+      const uint64_t offset = 5000;
       char scratch[8] = {0};
       Slice buf;
       ASSERT_OK(rwfile->Read(offset, sizeof(scratch), &buf, scratch));

db/external_sst_file_ingestion_job.cc

@@ -769,8 +769,6 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
   std::unique_ptr<InternalIterator> iter(table_reader->NewIterator(
       ro, sv->mutable_cf_options.prefix_extractor.get(), /*arena=*/nullptr,
       /*skip_filters=*/false, TableReaderCaller::kExternalSSTIngestion));
-  std::unique_ptr<InternalIterator> range_del_iter(
-      table_reader->NewRangeTombstoneIterator(ro));
 
   // Get first (smallest) and last (largest) key from file.
   file_to_ingest->smallest_internal_key =
@@ -792,8 +790,33 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
     }
     file_to_ingest->smallest_internal_key.SetFrom(key);
 
-    iter->SeekToLast();
-    pik_status = ParseInternalKey(iter->key(), &key, allow_data_in_errors);
+    Slice largest;
+    if (strcmp(cfd_->ioptions()->table_factory->Name(), "PlainTable") == 0) {
+      // PlainTable iterator does not support SeekToLast().
+      largest = iter->key();
+      for (; iter->Valid(); iter->Next()) {
+        if (cfd_->internal_comparator().Compare(iter->key(), largest) > 0) {
+          largest = iter->key();
+        }
+      }
+      if (!iter->status().ok()) {
+        return iter->status();
+      }
+    } else {
+      iter->SeekToLast();
+      if (!iter->Valid()) {
+        if (iter->status().ok()) {
+          // The file contains at least 1 key since iter is valid after
+          // SeekToFirst().
+          return Status::Corruption("Can not find largest key in sst file");
+        } else {
+          return iter->status();
+        }
+      }
+      largest = iter->key();
+    }
+
+    pik_status = ParseInternalKey(largest, &key, allow_data_in_errors);
     if (!pik_status.ok()) {
       return Status::Corruption("Corrupted key in external file. ",
                                 pik_status.getState());
@@ -804,8 +827,12 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
     file_to_ingest->largest_internal_key.SetFrom(key);
 
     bounds_set = true;
+  } else if (!iter->status().ok()) {
+    return iter->status();
   }
 
+  std::unique_ptr<InternalIterator> range_del_iter(
+      table_reader->NewRangeTombstoneIterator(ro));
   // We may need to adjust these key bounds, depending on whether any range
   // deletion tombstones extend past them.
   const Comparator* ucmp = cfd_->internal_comparator().user_comparator();

db/external_sst_file_test.cc

@@ -2623,6 +2623,7 @@ TEST_P(ExternalSSTFileTest,
       "AfterRead");
   ingest_thread.join();
   for (auto* iter : iters) {
+    ASSERT_OK(iter->status());
     delete iter;
   }
   iters.clear();

db/flush_job.cc

@@ -100,7 +100,7 @@ FlushJob::FlushJob(
     Statistics* stats, EventLogger* event_logger, bool measure_io_stats,
     const bool sync_output_directory, const bool write_manifest,
     Env::Priority thread_pri, const std::shared_ptr<IOTracer>& io_tracer,
-    const SeqnoToTimeMapping& seqno_time_mapping, const std::string& db_id,
+    const SeqnoToTimeMapping& seqno_to_time_mapping, const std::string& db_id,
     const std::string& db_session_id, std::string full_history_ts_low,
     BlobFileCompletionCallback* blob_callback)
     : dbname_(dbname),
@@ -136,7 +136,7 @@ FlushJob::FlushJob(
       clock_(db_options_.clock),
       full_history_ts_low_(std::move(full_history_ts_low)),
       blob_callback_(blob_callback),
-      db_impl_seqno_time_mapping_(seqno_time_mapping) {
+      db_impl_seqno_to_time_mapping_(seqno_to_time_mapping) {
   // Update the thread status to indicate flush.
   ReportStartedFlush();
   TEST_SYNC_POINT("FlushJob::FlushJob()");
@@ -851,10 +851,11 @@ Status FlushJob::WriteLevel0Table() {
   Status s;
 
   SequenceNumber smallest_seqno = mems_.front()->GetEarliestSequenceNumber();
-  if (!db_impl_seqno_time_mapping_.Empty()) {
-    // make a local copy, as the seqno_time_mapping from db_impl is not thread
-    // safe, which will be used while not holding the db_mutex.
-    seqno_to_time_mapping_ = db_impl_seqno_time_mapping_.Copy(smallest_seqno);
+  if (!db_impl_seqno_to_time_mapping_.Empty()) {
+    // make a local copy, as the seqno_to_time_mapping from db_impl is not
+    // thread safe, which will be used while not holding the db_mutex.
+    seqno_to_time_mapping_ =
+        db_impl_seqno_to_time_mapping_.Copy(smallest_seqno);
   }
 
   std::vector<BlobFileAddition> blob_file_additions;

db/flush_job.h

@@ -210,9 +210,9 @@ class FlushJob {
   const std::string full_history_ts_low_;
   BlobFileCompletionCallback* blob_callback_;
 
-  // reference to the seqno_time_mapping_ in db_impl.h, not safe to read without
-  // db mutex
-  const SeqnoToTimeMapping& db_impl_seqno_time_mapping_;
+  // reference to the seqno_to_time_mapping_ in db_impl.h, not safe to read
+  // without db mutex
+  const SeqnoToTimeMapping& db_impl_seqno_to_time_mapping_;
   SeqnoToTimeMapping seqno_to_time_mapping_;
 
   // Keeps track of the newest user-defined timestamp for this flush job if

db/flush_job_test.cc

@@ -127,11 +127,12 @@ class FlushJobTestBase : public testing::Test {
       column_families.emplace_back(cf_name, cf_options_);
     }
 
-    versions_.reset(
-        new VersionSet(dbname_, &db_options_, env_options_, table_cache_.get(),
-                       &write_buffer_manager_, &write_controller_,
-                       /*block_cache_tracer=*/nullptr, /*io_tracer=*/nullptr,
-                       /*db_id*/ "", /*db_session_id*/ ""));
+    versions_.reset(new VersionSet(
+        dbname_, &db_options_, env_options_, table_cache_.get(),
+        &write_buffer_manager_, &write_controller_,
+        /*block_cache_tracer=*/nullptr, /*io_tracer=*/nullptr,
+        /*db_id=*/"", /*db_session_id=*/"", /*daily_offpeak_time_utc=*/"",
+        /*error_handler=*/nullptr));
     EXPECT_OK(versions_->Recover(column_families, false));
   }
 
@@ -457,7 +458,8 @@ TEST_F(FlushJobTest, FlushMemtablesMultipleColumnFamilies) {
     // Verify that imm is empty
     ASSERT_EQ(std::numeric_limits<uint64_t>::max(),
               all_cfds[k]->imm()->GetEarliestMemTableID());
-    ASSERT_EQ(0, all_cfds[k]->imm()->GetLatestMemTableID());
+    ASSERT_EQ(0, all_cfds[k]->imm()->GetLatestMemTableID(
+                     false /* for_atomic_flush */));
     ++k;
   }
 

db/import_column_family_job.cc

@@ -186,7 +186,8 @@ Status ImportColumnFamilyJob::Run() {
       cfd_->NumberLevels(), cfd_->ioptions()->compaction_style,
       nullptr /* src_vstorage */, cfd_->ioptions()->force_consistency_checks,
       EpochNumberRequirement::kMightMissing, cfd_->ioptions()->clock,
-      cfd_->GetLatestMutableCFOptions()->bottommost_file_compaction_delay);
+      cfd_->GetLatestMutableCFOptions()->bottommost_file_compaction_delay,
+      cfd_->current()->version_set()->offpeak_time_option());
   Status s;
 
   for (size_t i = 0; s.ok() && i < files_to_import_.size(); ++i) {
@@ -365,9 +366,35 @@ Status ImportColumnFamilyJob::GetIngestedFileInfo(
     bool bound_set = false;
     if (iter->Valid()) {
       file_to_import->smallest_internal_key.DecodeFrom(iter->key());
-      iter->SeekToLast();
-      file_to_import->largest_internal_key.DecodeFrom(iter->key());
+      Slice largest;
+      if (strcmp(cfd_->ioptions()->table_factory->Name(), "PlainTable") == 0) {
+        // PlainTable iterator does not support SeekToLast().
+        largest = iter->key();
+        for (; iter->Valid(); iter->Next()) {
+          if (cfd_->internal_comparator().Compare(iter->key(), largest) > 0) {
+            largest = iter->key();
+          }
+        }
+        if (!iter->status().ok()) {
+          return iter->status();
+        }
+      } else {
+        iter->SeekToLast();
+        if (!iter->Valid()) {
+          if (iter->status().ok()) {
+            // The file contains at least 1 key since iter is valid after
+            // SeekToFirst().
+            return Status::Corruption("Can not find largest key in sst file");
+          } else {
+            return iter->status();
+          }
+        }
+        largest = iter->key();
+      }
+      file_to_import->largest_internal_key.DecodeFrom(largest);
       bound_set = true;
+    } else if (!iter->status().ok()) {
+      return iter->status();
     }
 
     std::unique_ptr<InternalIterator> range_del_iter{
@@ -429,4 +456,4 @@ Status ImportColumnFamilyJob::GetIngestedFileInfo(
 
   return status;
 }
-}  // namespace ROCKSDB_NAMESPACE
+}  // namespace ROCKSDB_NAMESPACE

db/job_context.h

@@ -15,6 +15,7 @@
 #include "db/column_family.h"
 #include "db/log_writer.h"
 #include "db/version_set.h"
+#include "util/autovector.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -170,6 +171,16 @@ struct JobContext {
   // will be reused later
   std::vector<uint64_t> log_recycle_files;
 
+  // Files quarantined from deletion. This list contains file numbers for files
+  // that are in an ambiguous states. This includes newly generated SST files
+  // and blob files from flush and compaction job whose VersionEdits' persist
+  // state in Manifest are unclear. An old manifest file whose immediately
+  // following new manifest file's CURRENT file creation is in an unclear state.
+  // WAL logs don't have this premature deletion risk since
+  // min_log_number_to_keep is only updated after successful manifest commits.
+  // So this data structure doesn't track log files.
+  autovector<uint64_t> files_to_quarantine;
+
   // a list of manifest files that we need to delete
   std::vector<std::string> manifest_delete_files;
 

db/log_reader.cc

@@ -469,12 +469,14 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result, size_t* drop_size,
     const unsigned int type = header[6];
     const uint32_t length = a | (b << 8);
     int header_size = kHeaderSize;
-    if ((type >= kRecyclableFullType && type <= kRecyclableLastType) ||
-        type == kRecyclableUserDefinedTimestampSizeType) {
+    const bool is_recyclable_type =
+        ((type >= kRecyclableFullType && type <= kRecyclableLastType) ||
+         type == kRecyclableUserDefinedTimestampSizeType);
+    if (is_recyclable_type) {
+      header_size = kRecyclableHeaderSize;
       if (end_of_buffer_offset_ - buffer_.size() == 0) {
         recycled_ = true;
       }
-      header_size = kRecyclableHeaderSize;
       // We need enough for the larger header
       if (buffer_.size() < static_cast<size_t>(kRecyclableHeaderSize)) {
         int r = kEof;
@@ -483,11 +485,8 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result, size_t* drop_size,
         }
         continue;
       }
-      const uint32_t log_num = DecodeFixed32(header + 7);
-      if (log_num != log_number_) {
-        return kOldRecord;
-      }
     }
+
     if (header_size + length > buffer_.size()) {
       assert(buffer_.size() >= static_cast<size_t>(header_size));
       *drop_size = buffer_.size();
@@ -499,6 +498,14 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result, size_t* drop_size,
       return kBadRecordLen;
     }
 
+    if (is_recyclable_type) {
+      const uint32_t log_num = DecodeFixed32(header + 7);
+      if (log_num != log_number_) {
+        buffer_.remove_prefix(header_size + length);
+        return kOldRecord;
+      }
+    }
+
     if (type == kZeroType && length == 0) {
       // Skip zero length record without reporting any drops since
       // such records are produced by the mmap based writing code in

db/manual_compaction_test.cc

@@ -124,6 +124,7 @@ TEST_F(ManualCompactionTest, CompactTouchesAllKeys) {
     ASSERT_EQ("key3", itr->key().ToString());
     itr->Next();
     ASSERT_TRUE(!itr->Valid());
+    ASSERT_OK(itr->status());
     delete itr;
 
     delete options.compaction_filter;
@@ -179,6 +180,7 @@ TEST_F(ManualCompactionTest, Test) {
   for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
     num_keys++;
   }
+  ASSERT_OK(iter->status());
   delete iter;
   ASSERT_EQ(kNumKeys, num_keys) << "Bad number of keys";
 

db/memtable_list.h

@@ -386,11 +386,21 @@ class MemTableList {
     return memlist.back()->GetID();
   }
 
-  uint64_t GetLatestMemTableID() const {
+  uint64_t GetLatestMemTableID(bool for_atomic_flush) const {
     auto& memlist = current_->memlist_;
     if (memlist.empty()) {
       return 0;
     }
+    if (for_atomic_flush) {
+      // Scan the memtable list from new to old
+      for (auto it = memlist.begin(); it != memlist.end(); ++it) {
+        MemTable* m = *it;
+        if (m->atomic_flush_seqno_ != kMaxSequenceNumber) {
+          return m->GetID();
+        }
+      }
+      return 0;
+    }
     return memlist.front()->GetID();
   }
 

db/memtable_list_test.cc

@@ -106,8 +106,9 @@ class MemTableListTest : public testing::Test {
     VersionSet versions(dbname, &immutable_db_options, env_options,
                         table_cache.get(), &write_buffer_manager,
                         &write_controller, /*block_cache_tracer=*/nullptr,
-                        /*io_tracer=*/nullptr, /*db_id*/ "",
-                        /*db_session_id*/ "");
+                        /*io_tracer=*/nullptr, /*db_id=*/"",
+                        /*db_session_id=*/"", /*daily_offpeak_time_utc=*/"",
+                        /*error_handler=*/nullptr);
     std::vector<ColumnFamilyDescriptor> cf_descs;
     cf_descs.emplace_back(kDefaultColumnFamilyName, ColumnFamilyOptions());
     cf_descs.emplace_back("one", ColumnFamilyOptions());
@@ -157,8 +158,9 @@ class MemTableListTest : public testing::Test {
     VersionSet versions(dbname, &immutable_db_options, env_options,
                         table_cache.get(), &write_buffer_manager,
                         &write_controller, /*block_cache_tracer=*/nullptr,
-                        /*io_tracer=*/nullptr, /*db_id*/ "",
-                        /*db_session_id*/ "");
+                        /*io_tracer=*/nullptr, /*db_id=*/"",
+                        /*db_session_id=*/"", /*daily_offpeak_time_utc=*/"",
+                        /*error_handler=*/nullptr);
     std::vector<ColumnFamilyDescriptor> cf_descs;
     cf_descs.emplace_back(kDefaultColumnFamilyName, ColumnFamilyOptions());
     cf_descs.emplace_back("one", ColumnFamilyOptions());
@@ -833,7 +835,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   // Add another table
   list.Add(tables[5], &to_delete);
   ASSERT_EQ(1, list.NumNotFlushed());
-  ASSERT_EQ(5, list.GetLatestMemTableID());
+  ASSERT_EQ(5, list.GetLatestMemTableID(false /* for_atomic_flush */));
   memtable_id = 4;
   // Pick tables to flush. The tables to pick must have ID smaller than or
   // equal to 4. Therefore, no table will be selected in this case.