Update version and HISTORY for 6.16.5

Summary:
`CompareKeyContext::operator()` on the trunk has a bug: when comparing
column family IDs, `lhs` is used for both sides of the comparison. This
results in the `KeyContext`s getting sorted solely based on key, which
in turn means that keys with the same column family do not necessarily
form a single range in the sorted list. This violates an assumption of the
batched `MultiGet` logic, leading to the same column family
showing up multiple times in the list of `MultiGetColumnFamilyData`.
The end result is the code attempting to check out the thread-local
`SuperVersion` for the same CF multiple times, causing an
assertion violation in debug builds and memory corruption/crash in
release builds.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/8633

Test Plan: `make check`

Reviewed By: riversand963

Differential Revision: D30169182

Pulled By: ltamasi

fbshipit-source-id: a47710652df7e95b14b40fb710924c11a8478023

.circleci/config.yml

@@ -79,7 +79,7 @@ commands:
           name: Install gflags
           command: |
             sudo apt-get update -y && sudo apt-get install -y libgflags-dev
-  
+
   install-gflags-on-macos:
     steps:
       - run:
@@ -93,7 +93,7 @@ commands:
           name: Install gtest-parallel
           command: |
             git clone --single-branch --branch master --depth 1 https://github.com/google/gtest-parallel.git ~/gtest-parallel
-            echo "export PATH=$HOME/gtest-parallel:$PATH" >> $BASH_ENV
+            echo 'export PATH=$HOME/gtest-parallel:$PATH' >> $BASH_ENV
 
 executors:
   windows-2xlarge:
@@ -459,7 +459,7 @@ jobs:
   build-examples:
     machine:
       image: ubuntu-1604:202007-01
-    resource_class: medium
+    resource_class: large
     steps:
       - pre-steps
       - install-gflags
@@ -469,6 +469,29 @@ jobs:
             OPT=-DTRAVIS V=1 make -j4 static_lib && cd examples && make -j4 | ../.circleci/cat_ignore_eagain
       - post-steps
 
+  build-cmake-mingw:
+    machine:
+      image: ubuntu-1604:202007-01
+    resource_class: 2xlarge
+    steps:
+      - pre-steps
+      - install-gflags
+      - run: sudo apt-get update -y && sudo apt-get install -y mingw-w64
+      - run: sudo update-alternatives --set x86_64-w64-mingw32-g++ /usr/bin/x86_64-w64-mingw32-g++-posix
+      - run:
+          name: "Build cmake-mingw"
+          command: |
+            sudo apt-get install snapd && sudo snap install cmake --beta --classic
+            export PATH=/snap/bin:$PATH
+            sudo apt-get install -y openjdk-8-jdk
+            export JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64
+            export PATH=$JAVA_HOME/bin:$PATH
+            echo "JAVA_HOME=${JAVA_HOME}"
+            which java && java -version
+            which javac && javac -version
+            mkdir build && cd build && cmake -DJNI=1 -DWITH_GFLAGS=OFF .. -DCMAKE_C_COMPILER=x86_64-w64-mingw32-gcc -DCMAKE_CXX_COMPILER=x86_64-w64-mingw32-g++ -DCMAKE_SYSTEM_NAME=Windows && make -j4 rocksdb rocksdbjni
+      - post-steps
+
   build-linux-non-shm:
     machine:
       image: ubuntu-1604:202007-01
@@ -496,7 +519,7 @@ jobs:
             sed -i 's/[[:space:]]*$//; s/ / \.\//g; s/.*/.\/&/' /tmp/test_list
             cat /tmp/test_list
             export TEST_TMPDIR=/tmp/rocksdb_test_tmp
-            /usr/bin/python ../gtest-parallel/gtest-parallel $(</tmp/test_list) --output_dir=/tmp | cat  # pipe to cat to continuously output status on circleci UI. Otherwise, no status will be printed while the job is running.
+            gtest-parallel $(</tmp/test_list) --output_dir=/tmp | cat  # pipe to cat to continuously output status on circleci UI. Otherwise, no status will be printed while the job is running.
       - post-steps
 
 workflows:
@@ -591,3 +614,9 @@ workflows:
   build-macos:
     jobs:
       - build-macos
+  build-macos-cmake:
+    jobs:
+      - build-macos-cmake
+  build-cmake-mingw:
+    jobs:
+      - build-cmake-mingw

.gitignore

@@ -90,3 +90,6 @@ buckifier/__pycache__
 compile_commands.json
 clang-format-diff.py
 .py3/
+
+fuzz/proto/gen/
+fuzz/crash-*

.travis.yml

@@ -105,6 +105,13 @@ matrix:
     os: linux
     arch: ppc64le
     env: JOB_NAME=cmake
+  # Exclude all osx since it should be covered by CircleCI
+  - if: type = pull_request AND commit_message !~ /FULL_CI/
+    os: osx
+    env: TEST_GROUP=platform_dependent
+  - if: type = pull_request AND commit_message !~ /FULL_CI/
+    os: osx
+    env: JOB_NAME=cmake
   # NB: the cmake build is a partial java test
   - if: type = pull_request AND commit_message !~ /FULL_CI/
     os: osx

CMakeLists.txt

@@ -831,6 +831,20 @@ set(SOURCES
         utilities/write_batch_with_index/write_batch_with_index_internal.cc
         $<TARGET_OBJECTS:build_version>)
 
+list(APPEND SOURCE
+  utilities/transactions/lock/range/range_tree/lib/locktree/concurrent_tree.cc
+  utilities/transactions/lock/range/range_tree/lib/locktree/keyrange.cc
+  utilities/transactions/lock/range/range_tree/lib/locktree/lock_request.cc
+  utilities/transactions/lock/range/range_tree/lib/locktree/locktree.cc
+  utilities/transactions/lock/range/range_tree/lib/locktree/manager.cc
+  utilities/transactions/lock/range/range_tree/lib/locktree/range_buffer.cc
+  utilities/transactions/lock/range/range_tree/lib/locktree/treenode.cc
+  utilities/transactions/lock/range/range_tree/lib/locktree/txnid_set.cc
+  utilities/transactions/lock/range/range_tree/lib/locktree/wfg.cc
+  utilities/transactions/lock/range/range_tree/lib/standalone_port.cc
+  utilities/transactions/lock/range/range_tree/lib/util/dbt.cc
+  utilities/transactions/lock/range/range_tree/lib/util/memarena.cc)
+
 if(HAVE_SSE42 AND NOT MSVC)
   set_source_files_properties(
     util/crc32c.cc

HISTORY.md

@@ -1,27 +1,47 @@
 # Rocksdb Change Log
-## 6.15.5 (02/05/2021)
+## 6.16.5 (09/08/2021)
+### Bug Fixes
+* Fixed a bug affecting the batched `MultiGet` API when used with keys spanning multiple column families and `sorted_input == false`.
+
+## 6.16.4 (03/30/2021)
+### Bug Fixes
+* Fix build on ppc64 and musl build.
+
+## 6.16.3 (02/05/2021)
 ### Bug Fixes
 * Since 6.15.0, `TransactionDB` returns error `Status`es from calls to `DeleteRange()` and calls to `Write()` where the `WriteBatch` contains a range deletion. Previously such operations may have succeeded while not providing the expected transactional guarantees. There are certain cases where range deletion can still be used on such DBs; see the API doc on `TransactionDB::DeleteRange()` for details.
 * `OptimisticTransactionDB` now returns error `Status`es from calls to `DeleteRange()` and calls to `Write()` where the `WriteBatch` contains a range deletion. Previously such operations may have succeeded while not providing the expected transactional guarantees.
 
-## 6.15.4 (01/21/2021)
+## 6.16.2 (1/21/2021)
 ### Bug Fixes
 * Fix a race condition between DB startups and shutdowns in managing the periodic background worker threads. One effect of this race condition could be the process being terminated.
 
-## 6.15.3 (01/07/2021)
+## 6.16.1 (1/20/2021)
 ### Bug Fixes
-* For Java builds, fix errors due to missing compression library includes.
+* Version older than 6.15 cannot decode VersionEdits `WalAddition` and `WalDeletion`, fixed this by changing the encoded format of them to be ignorable by older versions.
 
-## 6.15.2 (12/22/2020)
-### Bug Fixes
-* Fix failing RocksJava test compilation and add CI jobs
-* Fix jemalloc compilation issue on macOS
-* Fix build issues - compatibility with older gcc, older jemalloc libraries, docker warning when building i686 binaries
+## 6.16.0 (12/18/2020)
+### Behavior Changes
+* Attempting to write a merge operand without explicitly configuring `merge_operator` now fails immediately, causing the DB to enter read-only mode. Previously, failure was deferred until the `merge_operator` was needed by a user read or a background operation.
+* Since RocksDB does not continue write the same file if a file write fails for any reason, the file scope write IO error is treated the same as retryable IO error. More information about error handling of file scope IO error is included in `ErrorHandler::SetBGError`.
 
-## 6.15.1 (12/01/2020)
 ### Bug Fixes
-* Truncated WALs ending in incomplete records can no longer produce gaps in the recovered data when `WALRecoveryMode::kPointInTimeRecovery` is used. Gaps are still possible when WALs are truncated exactly on record boundaries.
+* Truncated WALs ending in incomplete records can no longer produce gaps in the recovered data when `WALRecoveryMode::kPointInTimeRecovery` is used. Gaps are still possible when WALs are truncated exactly on record boundaries; for complete protection, users should enable `track_and_verify_wals_in_manifest`.
 * Fix a bug where compressed blocks read by MultiGet are not inserted into the compressed block cache when use_direct_reads = true.
+* Fixed the issue of full scanning on obsolete files when there are too many outstanding compactions with ConcurrentTaskLimiter enabled.
+* Fixed the logic of populating native data structure for `read_amp_bytes_per_bit` during OPTIONS file parsing on big-endian architecture. Without this fix, original code introduced in PR7659, when running on big-endian machine, can mistakenly store read_amp_bytes_per_bit (an uint32) in little endian format. Future access to `read_amp_bytes_per_bit` will give wrong values. Little endian architecture is not affected.
+* Fixed prefix extractor with timestamp issues.
+* Fixed a bug in atomic flush: in two-phase commit mode, the minimum WAL log number to keep is incorrect.
+* Fixed a bug related to checkpoint in PR7789: if there are multiple column families, and the checkpoint is not opened as read only, then in rare cases, data loss may happen in the checkpoint. Since backup engine relies on checkpoint, it may also be affected.
+
+### New Features
+* User defined timestamp feature supports `CompactRange` and `GetApproximateSizes`.
+* Support getting aggregated table properties (kAggregatedTableProperties and kAggregatedTablePropertiesAtLevel) with DB::GetMapProperty, for easier access to the data in a structured format.
+* Experimental option BlockBasedTableOptions::optimize_filters_for_memory now works with experimental Ribbon filter (as well as Bloom filter).
+
+### Public API Change
+* Deprecated public but rarely-used FilterBitsBuilder::CalculateNumEntry, which is replaced with ApproximateNumEntries taking a size_t parameter and returning size_t.
+* Added a new option `track_and_verify_wals_in_manifest`. If `true`, the log numbers and sizes of the synced WALs are tracked in MANIFEST, then during DB recovery, if a synced WAL is missing from disk, or the WAL's size does not match the recorded size in MANIFEST, an error will be reported and the recovery will be aborted. Note that this option does not work with secondary instance.
 
 ## 6.15.0 (11/13/2020)
 ### Bug Fixes
@@ -38,8 +58,7 @@
 * Fixed a potential bug caused by evaluating `TableBuilder::NeedCompact()` before `TableBuilder::Finish()` in compaction job. For example, the `NeedCompact()` method of `CompactOnDeletionCollector` returned by built-in `CompactOnDeletionCollectorFactory` requires `BlockBasedTable::Finish()` to return the correct result. The bug can cause a compaction-generated file not to be marked for future compaction based on deletion ratio.
 * Fixed a seek issue with prefix extractor and timestamp.
 * Fixed a bug of encoding and parsing BlockBasedTableOptions::read_amp_bytes_per_bit as a 64-bit integer.
-* Fixed the logic of populating native data structure for `read_amp_bytes_per_bit` during OPTIONS file parsing on big-endian architecture. Without this fix, original code introduced in PR7659, when running on big-endian machine, can mistakenly store read_amp_bytes_per_bit (an uint32) in little endian format. Future access to `read_amp_bytes_per_bit` will give wrong values. Little endian architecture is not affected.
-
+* Fixed a bug of a recovery corner case, details in PR7621.
 
 ### Public API Change
 * Deprecate `BlockBasedTableOptions::pin_l0_filter_and_index_blocks_in_cache` and `BlockBasedTableOptions::pin_top_level_index_and_filter`. These options still take effect until users migrate to the replacement APIs in `BlockBasedTableOptions::metadata_cache_options`. Migration guidance can be found in the API comments on the deprecated options.

Makefile

@@ -510,6 +510,12 @@ ifeq ($(USE_FOLLY_DISTRIBUTED_MUTEX),1)
   LIB_OBJECTS += $(patsubst %.cpp, $(OBJ_DIR)/%.o, $(FOLLY_SOURCES))
 endif
 
+# range_tree is not compatible with non GNU libc on ppc64
+# see https://jira.percona.com/browse/PS-7559
+ifneq ($(PPC_LIBC_IS_GNU),0)
+  LIB_OBJECTS += $(patsubst %.cc, $(OBJ_DIR)/%.o, $(RANGE_TREE_SOURCES))
+endif
+
 GTEST = $(OBJ_DIR)/$(GTEST_DIR)/gtest/gtest-all.o
 TESTUTIL = $(OBJ_DIR)/test_util/testutil.o
 TESTHARNESS = $(OBJ_DIR)/test_util/testharness.o $(TESTUTIL) $(GTEST)
@@ -589,12 +595,21 @@ ifdef ASSERT_STATUS_CHECKED
 		cassandra_row_merge_test \
 		cassandra_serialize_test \
 		cleanable_test \
+		checkpoint_test \
 		coding_test \
 		crc32c_test \
 		dbformat_test \
 		db_basic_test \
 		db_blob_basic_test \
+		db_blob_index_test \
+		db_block_cache_test \
 		db_flush_test \
+		db_iterator_test \
+		db_logical_block_size_cache_test \
+		db_memtable_test \
+		db_merge_operand_test \
+		db_merge_operator_test \
+		db_wal_test \
 		db_with_timestamp_basic_test \
 		db_with_timestamp_compaction_test \
 		db_options_test \
@@ -609,6 +624,7 @@ ifdef ASSERT_STATUS_CHECKED
 		env_logger_test \
 		event_logger_test \
 		error_handler_fs_test \
+		external_sst_file_basic_test \
 		auto_roll_logger_test \
 		file_indexer_test \
 		flush_job_test \
@@ -624,6 +640,7 @@ ifdef ASSERT_STATUS_CHECKED
 		merger_test \
 		mock_env_test \
 		object_registry_test \
+		optimistic_transaction_test \
 		prefix_test \
 		plain_table_db_test \
 		repair_test \
@@ -631,6 +648,7 @@ ifdef ASSERT_STATUS_CHECKED
 		customizable_test \
 		options_settable_test \
 		options_test \
+		point_lock_manager_test \
 		random_test \
 		range_del_aggregator_test \
 		sst_file_reader_test \
@@ -644,6 +662,7 @@ ifdef ASSERT_STATUS_CHECKED
 		stats_history_test \
 		thread_local_test \
 		trace_analyzer_test \
+		transaction_test \
 		env_timed_test \
 		filelock_test \
 		timer_queue_test \
@@ -659,6 +678,8 @@ ifdef ASSERT_STATUS_CHECKED
 		version_edit_test \
 		work_queue_test \
 		write_controller_test \
+		write_prepared_transaction_test \
+		write_unprepared_transaction_test \
 		compaction_iterator_test \
 		compaction_job_test \
 		compaction_job_stats_test \
@@ -677,6 +698,8 @@ ifdef ASSERT_STATUS_CHECKED
 		import_column_family_test \
 		memory_test \
 		table_test \
+		write_batch_test \
+		write_batch_with_index_test \
 
 ifeq ($(USE_FOLLY_DISTRIBUTED_MUTEX),1)
 TESTS_PASSING_ASC += folly_synchronization_distributed_mutex_test
@@ -2122,7 +2145,7 @@ zlib-$(ZLIB_VER).tar.gz:
 libz.a: zlib-$(ZLIB_VER).tar.gz
 	-rm -rf zlib-$(ZLIB_VER)
 	tar xvzf zlib-$(ZLIB_VER).tar.gz
-	cd zlib-$(ZLIB_VER) && CFLAGS='-fPIC ${EXTRA_CFLAGS}' LDFLAGS='${EXTRA_LDFLAGS}' ./configure --static && $(MAKE)
+	cd zlib-$(ZLIB_VER) && CFLAGS='-fPIC ${JAVA_STATIC_DEPS_CCFLAGS} ${EXTRA_CFLAGS}' LDFLAGS='${JAVA_STATIC_DEPS_LDFLAGS} ${EXTRA_LDFLAGS}' ./configure --static && $(MAKE)
 	cp zlib-$(ZLIB_VER)/libz.a .
 
 bzip2-$(BZIP2_VER).tar.gz:
@@ -2136,7 +2159,7 @@ bzip2-$(BZIP2_VER).tar.gz:
 libbz2.a: bzip2-$(BZIP2_VER).tar.gz
 	-rm -rf bzip2-$(BZIP2_VER)
 	tar xvzf bzip2-$(BZIP2_VER).tar.gz
-	cd bzip2-$(BZIP2_VER) && $(MAKE) CFLAGS='-fPIC -O2 -g -D_FILE_OFFSET_BITS=64 ${EXTRA_CFLAGS}' AR='ar ${EXTRA_ARFLAGS}'
+	cd bzip2-$(BZIP2_VER) && $(MAKE) CFLAGS='-fPIC -O2 -g -D_FILE_OFFSET_BITS=64 ${JAVA_STATIC_DEPS_CCFLAGS} ${EXTRA_CFLAGS}' LDFLAGS='${JAVA_STATIC_DEPS_LDFLAGS} ${EXTRA_LDFLAGS}' AR='ar ${EXTRA_ARFLAGS}'
 	cp bzip2-$(BZIP2_VER)/libbz2.a .
 
 snappy-$(SNAPPY_VER).tar.gz:
@@ -2151,7 +2174,7 @@ libsnappy.a: snappy-$(SNAPPY_VER).tar.gz
 	-rm -rf snappy-$(SNAPPY_VER)
 	tar xvzf snappy-$(SNAPPY_VER).tar.gz
 	mkdir snappy-$(SNAPPY_VER)/build
-	cd snappy-$(SNAPPY_VER)/build && CFLAGS='${EXTRA_CFLAGS}' CXXFLAGS='${EXTRA_CXXFLAGS}' LDFLAGS='${EXTRA_LDFLAGS}' cmake -DCMAKE_POSITION_INDEPENDENT_CODE=ON .. && $(MAKE) ${SNAPPY_MAKE_TARGET}
+	cd snappy-$(SNAPPY_VER)/build && CFLAGS='${JAVA_STATIC_DEPS_CCFLAGS} ${EXTRA_CFLAGS}' CXXFLAGS='${JAVA_STATIC_DEPS_CXXFLAGS} ${EXTRA_CXXFLAGS}' LDFLAGS='${JAVA_STATIC_DEPS_LDFLAGS} ${EXTRA_LDFLAGS}' cmake -DCMAKE_POSITION_INDEPENDENT_CODE=ON ${PLATFORM_CMAKE_FLAGS} .. && $(MAKE) ${SNAPPY_MAKE_TARGET}
 	cp snappy-$(SNAPPY_VER)/build/libsnappy.a .
 
 lz4-$(LZ4_VER).tar.gz:
@@ -2165,7 +2188,7 @@ lz4-$(LZ4_VER).tar.gz:
 liblz4.a: lz4-$(LZ4_VER).tar.gz
 	-rm -rf lz4-$(LZ4_VER)
 	tar xvzf lz4-$(LZ4_VER).tar.gz
-	cd lz4-$(LZ4_VER)/lib && $(MAKE) CFLAGS='-fPIC -O2 ${EXTRA_CFLAGS}' all
+	cd lz4-$(LZ4_VER)/lib && $(MAKE) CFLAGS='-fPIC -O2 ${JAVA_STATIC_DEPS_CCFLAGS} ${EXTRA_CFLAGS}' LDFLAGS='${JAVA_STATIC_DEPS_LDFLAGS} ${EXTRA_LDFLAGS}' all
 	cp lz4-$(LZ4_VER)/lib/liblz4.a .
 
 zstd-$(ZSTD_VER).tar.gz:
@@ -2179,7 +2202,7 @@ zstd-$(ZSTD_VER).tar.gz:
 libzstd.a: zstd-$(ZSTD_VER).tar.gz
 	-rm -rf zstd-$(ZSTD_VER)
 	tar xvzf zstd-$(ZSTD_VER).tar.gz
-	cd zstd-$(ZSTD_VER)/lib && DESTDIR=. PREFIX= $(MAKE) CFLAGS='-fPIC -O2 ${EXTRA_CFLAGS}' install
+	cd zstd-$(ZSTD_VER)/lib && DESTDIR=. PREFIX= $(MAKE) CFLAGS='-fPIC -O2 ${JAVA_STATIC_DEPS_CCFLAGS} ${EXTRA_CFLAGS}' LDFLAGS='${JAVA_STATIC_DEPS_LDFLAGS} ${EXTRA_LDFLAGS}' libzstd.a
 	cp zstd-$(ZSTD_VER)/lib/libzstd.a .
 
 # A version of each $(LIB_OBJECTS) compiled with -fPIC and a fixed set of static compression libraries

TARGETS

@@ -388,6 +388,18 @@ cpp_library(
         "utilities/transactions/lock/lock_manager.cc",
         "utilities/transactions/lock/point/point_lock_manager.cc",
         "utilities/transactions/lock/point/point_lock_tracker.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/concurrent_tree.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/keyrange.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/lock_request.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/locktree.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/manager.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/range_buffer.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/treenode.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/txnid_set.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/wfg.cc",
+        "utilities/transactions/lock/range/range_tree/lib/standalone_port.cc",
+        "utilities/transactions/lock/range/range_tree/lib/util/dbt.cc",
+        "utilities/transactions/lock/range/range_tree/lib/util/memarena.cc",
         "utilities/transactions/optimistic_transaction.cc",
         "utilities/transactions/optimistic_transaction_db_impl.cc",
         "utilities/transactions/pessimistic_transaction.cc",
@@ -678,6 +690,18 @@ cpp_library(
         "utilities/transactions/lock/lock_manager.cc",
         "utilities/transactions/lock/point/point_lock_manager.cc",
         "utilities/transactions/lock/point/point_lock_tracker.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/concurrent_tree.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/keyrange.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/lock_request.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/locktree.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/manager.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/range_buffer.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/treenode.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/txnid_set.cc",
+        "utilities/transactions/lock/range/range_tree/lib/locktree/wfg.cc",
+        "utilities/transactions/lock/range/range_tree/lib/standalone_port.cc",
+        "utilities/transactions/lock/range/range_tree/lib/util/dbt.cc",
+        "utilities/transactions/lock/range/range_tree/lib/util/memarena.cc",
         "utilities/transactions/optimistic_transaction.cc",
         "utilities/transactions/optimistic_transaction_db_impl.cc",
         "utilities/transactions/pessimistic_transaction.cc",
@@ -786,7 +810,7 @@ cpp_binary(
 ) if not is_opt_mode else None
 
 custom_unittest(
-    "c_test",
+    name = "c_test",
     command = [
         native.package_name() + "/buckifier/rocks_test_runner.sh",
         "$(location :{})".format("c_test_bin"),

USERS.md

@@ -44,7 +44,7 @@ Tango is using RocksDB as a graph storage to store all users' connection data an
 Turn is using RocksDB as a storage layer for their key/value store, serving at peak 2.4MM QPS out of different datacenters.
 Check out our RocksDB Protobuf merge operator at: https://github.com/vladb38/rocksdb_protobuf
 
-## Santanader UK/Cloudera Profession Services
+## Santander UK/Cloudera Profession Services
 Check out their blog post: http://blog.cloudera.com/blog/2015/08/inside-santanders-near-real-time-data-ingest-architecture/
 
 ## Airbnb
@@ -86,7 +86,7 @@ quasardb uses a heavily tuned RocksDB as its persistence layer.
 [Uber](http://eng.uber.com/cherami/) uses RocksDB as a durable and scalable task queue.
 
 ## 360 Pika
-[360](http://www.360.cn/) [Pika](https://github.com/Qihoo360/pika) is a nosql compatible with redis. With the huge amount of data stored, redis may suffer for a capacity bottleneck, and pika was born for solving it. It has widely been widely used in many company
+[360](http://www.360.cn/) [Pika](https://github.com/Qihoo360/pika) is a nosql compatible with redis. With the huge amount of data stored, redis may suffer for a capacity bottleneck, and pika was born for solving it. It has widely been used in many companies.
 
 ## LzLabs
 LzLabs is using RocksDB as a storage engine in their multi-database distributed framework to store application configuration and user data.
@@ -104,8 +104,11 @@ LzLabs is using RocksDB as a storage engine in their multi-database distributed
 [Crux](https://github.com/juxt/crux) is a document database that uses RocksDB for local [EAV](https://en.wikipedia.org/wiki/Entity%E2%80%93attribute%E2%80%93value_model) index storage to enable point-in-time bitemporal Datalog queries. The "unbundled" architecture uses Kafka to provide horizontal scalability.
 
 ## Nebula Graph
-
 [Nebula Graph](https://github.com/vesoft-inc/nebula) is a distributed, scalable, lightning-fast, open source graph database capable of hosting super large scale graphs with dozens of billions of vertices (nodes) and trillions of edges, with milliseconds of latency.
 
 ## YugabyteDB
 [YugabyteDB](https://www.yugabyte.com/) is an open source, high performance, distributed SQL database that uses RocksDB as its storage layer. For more information, please see https://github.com/yugabyte/yugabyte-db/.
+
+## 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 sotrage engine.
+

buckifier/buckify_rocksdb.py

@@ -135,11 +135,15 @@ def generate_targets(repo_path, deps_map):
     TARGETS.add_library(
         "rocksdb_lib",
         src_mk["LIB_SOURCES"] +
+        # always add range_tree, it's only excluded on ppc64, which we don't use internally
+        src_mk["RANGE_TREE_SOURCES"] +
         src_mk["TOOL_LIB_SOURCES"])
     # rocksdb_whole_archive_lib
     TARGETS.add_library(
         "rocksdb_whole_archive_lib",
         src_mk["LIB_SOURCES"] +
+        # always add range_tree, it's only excluded on ppc64, which we don't use internally
+        src_mk["RANGE_TREE_SOURCES"] +
         src_mk["TOOL_LIB_SOURCES"],
         deps=None,
         headers=None,

buckifier/targets_builder.py

@@ -90,7 +90,7 @@ cpp_binary(
 ) if not is_opt_mode else None
 
 custom_unittest(
-    "c_test",
+    name = "c_test",
     command = [
         native.package_name() + "/buckifier/rocks_test_runner.sh",
         "$(location :{})".format("c_test_bin"),

build_tools/build_detect_platform

@@ -655,6 +655,28 @@ else
     COMMON_FLAGS="$COMMON_FLAGS -mmacosx-version-min=10.12"
     PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -mmacosx-version-min=10.12"
     PLATFORM_SHARED_LDFLAGS="$PLATFORM_SHARED_LDFLAGS -mmacosx-version-min=10.12"
+    PLATFORM_CMAKE_FLAGS="-DCMAKE_OSX_DEPLOYMENT_TARGET=10.12"
+    JAVA_STATIC_DEPS_COMMON_FLAGS="-mmacosx-version-min=10.12"
+    JAVA_STATIC_DEPS_LDFLAGS="$JAVA_STATIC_DEPS_COMMON_FLAGS"
+    JAVA_STATIC_DEPS_CCFLAGS="$JAVA_STATIC_DEPS_COMMON_FLAGS"
+    JAVA_STATIC_DEPS_CXXFLAGS="$JAVA_STATIC_DEPS_COMMON_FLAGS"
+  fi
+fi
+
+if test -n "`echo $TARGET_ARCHITECTURE | grep ^ppc64`"; then
+  # check for GNU libc on ppc64
+  $CXX -x c++ - -o /dev/null 2>/dev/null <<EOF
+    #include <stdio.h>
+    #include <stdlib.h>
+    #include <gnu/libc-version.h>
+
+    int main(int argc, char *argv[]) {
+      printf("GNU libc version: %s\n", gnu_get_libc_version());
+      return 0;
+    }
+EOF
+  if [ "$?" != 0 ]; then
+      PPC_LIBC_IS_GNU=0
   fi
 fi
 
@@ -812,8 +834,12 @@ echo "CXX=$CXX" >> "$OUTPUT"
 echo "AR=$AR" >> "$OUTPUT"
 echo "PLATFORM=$PLATFORM" >> "$OUTPUT"
 echo "PLATFORM_LDFLAGS=$PLATFORM_LDFLAGS" >> "$OUTPUT"
+echo "PLATFORM_CMAKE_FLAGS=$PLATFORM_CMAKE_FLAGS" >> "$OUTPUT"
 echo "JAVA_LDFLAGS=$JAVA_LDFLAGS" >> "$OUTPUT"
 echo "JAVA_STATIC_LDFLAGS=$JAVA_STATIC_LDFLAGS" >> "$OUTPUT"
+echo "JAVA_STATIC_DEPS_CCFLAGS=$JAVA_STATIC_DEPS_CCFLAGS" >> "$OUTPUT"
+echo "JAVA_STATIC_DEPS_CXXFLAGS=$JAVA_STATIC_DEPS_CXXFLAGS" >> "$OUTPUT"
+echo "JAVA_STATIC_DEPS_LDFLAGS=$JAVA_STATIC_DEPS_LDFLAGS" >> "$OUTPUT"
 echo "JAVAC_ARGS=$JAVAC_ARGS" >> "$OUTPUT"
 echo "VALGRIND_VER=$VALGRIND_VER" >> "$OUTPUT"
 echo "PLATFORM_CCFLAGS=$PLATFORM_CCFLAGS" >> "$OUTPUT"
@@ -847,3 +873,6 @@ echo "LUA_PATH=$LUA_PATH" >> "$OUTPUT"
 if test -n "$USE_FOLLY_DISTRIBUTED_MUTEX"; then
   echo "USE_FOLLY_DISTRIBUTED_MUTEX=$USE_FOLLY_DISTRIBUTED_MUTEX" >> "$OUTPUT"
 fi
+if test -n "$PPC_LIBC_IS_GNU"; then
+  echo "PPC_LIBC_IS_GNU=$PPC_LIBC_IS_GNU" >> "$OUTPUT"
+fi

cache/lru_cache.h

@@ -67,7 +67,7 @@ struct LRUHandle {
     IS_HIGH_PRI = (1 << 1),
     // Whether this entry is in high-pri pool.
     IN_HIGH_PRI_POOL = (1 << 2),
-    // Wwhether this entry has had any lookups (hits).
+    // Whether this entry has had any lookups (hits).
     HAS_HIT = (1 << 3),
   };
 
@@ -130,7 +130,7 @@ struct LRUHandle {
     delete[] reinterpret_cast<char*>(this);
   }
 
-  // Caclculate the memory usage by metadata
+  // Calculate the memory usage by metadata
   inline size_t CalcTotalCharge(
       CacheMetadataChargePolicy metadata_charge_policy) {
     size_t meta_charge = 0;

db/arena_wrapped_db_iter.cc

@@ -30,20 +30,19 @@ Status ArenaWrappedDBIter::GetProperty(std::string prop_name,
   return db_iter_->GetProperty(prop_name, prop);
 }
 
-void ArenaWrappedDBIter::Init(Env* env, const ReadOptions& read_options,
-                              const ImmutableCFOptions& cf_options,
-                              const MutableCFOptions& mutable_cf_options,
-                              const SequenceNumber& sequence,
-                              uint64_t max_sequential_skip_in_iteration,
-                              uint64_t version_number,
-                              ReadCallback* read_callback, DBImpl* db_impl,
-                              ColumnFamilyData* cfd, bool allow_blob,
-                              bool allow_refresh) {
+void ArenaWrappedDBIter::Init(
+    Env* env, const ReadOptions& read_options,
+    const ImmutableCFOptions& cf_options,
+    const MutableCFOptions& mutable_cf_options, const Version* version,
+    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iteration,
+    uint64_t version_number, ReadCallback* read_callback, DBImpl* db_impl,
+    ColumnFamilyData* cfd, bool expose_blob_index, bool allow_refresh) {
   auto mem = arena_.AllocateAligned(sizeof(DBIter));
-  db_iter_ = new (mem) DBIter(env, read_options, cf_options, mutable_cf_options,
-                              cf_options.user_comparator, nullptr, sequence,
-                              true, max_sequential_skip_in_iteration,
-                              read_callback, db_impl, cfd, allow_blob);
+  db_iter_ =
+      new (mem) DBIter(env, read_options, cf_options, mutable_cf_options,
+                       cf_options.user_comparator, /* iter */ nullptr, version,
+                       sequence, true, max_sequential_skip_in_iteration,
+                       read_callback, db_impl, cfd, expose_blob_index);
   sv_number_ = version_number;
   read_options_ = read_options;
   allow_refresh_ = allow_refresh;
@@ -72,8 +71,9 @@ Status ArenaWrappedDBIter::Refresh() {
       read_callback_->Refresh(latest_seq);
     }
     Init(env, read_options_, *(cfd_->ioptions()), sv->mutable_cf_options,
-         latest_seq, sv->mutable_cf_options.max_sequential_skip_in_iterations,
-         cur_sv_number, read_callback_, db_impl_, cfd_, allow_blob_,
+         sv->current, latest_seq,
+         sv->mutable_cf_options.max_sequential_skip_in_iterations,
+         cur_sv_number, read_callback_, db_impl_, cfd_, expose_blob_index_,
          allow_refresh_);
 
     InternalIterator* internal_iter = db_impl_->NewInternalIterator(
@@ -90,16 +90,16 @@ Status ArenaWrappedDBIter::Refresh() {
 ArenaWrappedDBIter* NewArenaWrappedDbIterator(
     Env* env, const ReadOptions& read_options,
     const ImmutableCFOptions& cf_options,
-    const MutableCFOptions& mutable_cf_options, const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
-    ReadCallback* read_callback, DBImpl* db_impl, ColumnFamilyData* cfd,
-    bool allow_blob, bool allow_refresh) {
+    const MutableCFOptions& mutable_cf_options, const Version* version,
+    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
+    uint64_t version_number, ReadCallback* read_callback, DBImpl* db_impl,
+    ColumnFamilyData* cfd, bool expose_blob_index, bool allow_refresh) {
   ArenaWrappedDBIter* iter = new ArenaWrappedDBIter();
-  iter->Init(env, read_options, cf_options, mutable_cf_options, sequence,
-             max_sequential_skip_in_iterations, version_number, read_callback,
-             db_impl, cfd, allow_blob, allow_refresh);
+  iter->Init(env, read_options, cf_options, mutable_cf_options, version,
+             sequence, max_sequential_skip_in_iterations, version_number,
+             read_callback, db_impl, cfd, expose_blob_index, allow_refresh);
   if (db_impl != nullptr && cfd != nullptr && allow_refresh) {
-    iter->StoreRefreshInfo(db_impl, cfd, read_callback, allow_blob);
+    iter->StoreRefreshInfo(db_impl, cfd, read_callback, expose_blob_index);
   }
 
   return iter;

db/arena_wrapped_db_iter.h

@@ -23,6 +23,7 @@
 namespace ROCKSDB_NAMESPACE {
 
 class Arena;
+class Version;
 
 // A wrapper iterator which wraps DB Iterator and the arena, with which the DB
 // iterator is supposed to be allocated. This class is used as an entry point of
@@ -72,20 +73,20 @@ class ArenaWrappedDBIter : public Iterator {
 
   void Init(Env* env, const ReadOptions& read_options,
             const ImmutableCFOptions& cf_options,
-            const MutableCFOptions& mutable_cf_options,
+            const MutableCFOptions& mutable_cf_options, const Version* version,
             const SequenceNumber& sequence,
             uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
             ReadCallback* read_callback, DBImpl* db_impl, ColumnFamilyData* cfd,
-            bool allow_blob, bool allow_refresh);
+            bool expose_blob_index, bool allow_refresh);
 
   // Store some parameters so we can refresh the iterator at a later point
   // with these same params
   void StoreRefreshInfo(DBImpl* db_impl, ColumnFamilyData* cfd,
-                        ReadCallback* read_callback, bool allow_blob) {
+                        ReadCallback* read_callback, bool expose_blob_index) {
     db_impl_ = db_impl;
     cfd_ = cfd;
     read_callback_ = read_callback;
-    allow_blob_ = allow_blob;
+    expose_blob_index_ = expose_blob_index;
   }
 
  private:
@@ -96,7 +97,7 @@ class ArenaWrappedDBIter : public Iterator {
   DBImpl* db_impl_ = nullptr;
   ReadOptions read_options_;
   ReadCallback* read_callback_;
-  bool allow_blob_ = false;
+  bool expose_blob_index_ = false;
   bool allow_refresh_ = true;
 };
 
@@ -106,9 +107,9 @@ class ArenaWrappedDBIter : public Iterator {
 extern ArenaWrappedDBIter* NewArenaWrappedDbIterator(
     Env* env, const ReadOptions& read_options,
     const ImmutableCFOptions& cf_options,
-    const MutableCFOptions& mutable_cf_options, const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
-    ReadCallback* read_callback, DBImpl* db_impl = nullptr,
-    ColumnFamilyData* cfd = nullptr, bool allow_blob = false,
-    bool allow_refresh = true);
+    const MutableCFOptions& mutable_cf_options, const Version* version,
+    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
+    uint64_t version_number, ReadCallback* read_callback,
+    DBImpl* db_impl = nullptr, ColumnFamilyData* cfd = nullptr,
+    bool expose_blob_index = false, bool allow_refresh = true);
 }  // namespace ROCKSDB_NAMESPACE

db/blob/db_blob_basic_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 <array>
+
 #include "db/blob/blob_index.h"
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
@@ -44,6 +46,85 @@ TEST_F(DBBlobBasicTest, GetBlob) {
                   .IsIncomplete());
 }
 
+TEST_F(DBBlobBasicTest, MultiGetBlobs) {
+  constexpr size_t min_blob_size = 6;
+
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+  options.min_blob_size = min_blob_size;
+
+  Reopen(options);
+
+  // Put then retrieve three key-values. The first value is below the size limit
+  // and is thus stored inline; the other two are stored separately as blobs.
+  constexpr size_t num_keys = 3;
+
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "short";
+  static_assert(sizeof(first_value) - 1 < min_blob_size,
+                "first_value too long to be inlined");
+
+  ASSERT_OK(Put(first_key, first_value));
+
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "long_value";
+  static_assert(sizeof(second_value) - 1 >= min_blob_size,
+                "second_value too short to be stored as blob");
+
+  ASSERT_OK(Put(second_key, second_value));
+
+  constexpr char third_key[] = "third_key";
+  constexpr char third_value[] = "other_long_value";
+  static_assert(sizeof(third_value) - 1 >= min_blob_size,
+                "third_value too short to be stored as blob");
+
+  ASSERT_OK(Put(third_key, third_value));
+
+  ASSERT_OK(Flush());
+
+  ReadOptions read_options;
+
+  std::array<Slice, num_keys> keys{{first_key, second_key, third_key}};
+
+  {
+    std::array<PinnableSlice, num_keys> values;
+    std::array<Status, num_keys> statuses;
+
+    db_->MultiGet(read_options, db_->DefaultColumnFamily(), num_keys, &keys[0],
+                  &values[0], &statuses[0]);
+
+    ASSERT_OK(statuses[0]);
+    ASSERT_EQ(values[0], first_value);
+
+    ASSERT_OK(statuses[1]);
+    ASSERT_EQ(values[1], second_value);
+
+    ASSERT_OK(statuses[2]);
+    ASSERT_EQ(values[2], third_value);
+  }
+
+  // Try again with no I/O allowed. The table and the necessary blocks should
+  // already be in their respective caches. The first (inlined) value should be
+  // successfully read; however, the two blob values could only be read from the
+  // blob file, so for those the read should return Incomplete.
+  read_options.read_tier = kBlockCacheTier;
+
+  {
+    std::array<PinnableSlice, num_keys> values;
+    std::array<Status, num_keys> statuses;
+
+    db_->MultiGet(read_options, db_->DefaultColumnFamily(), num_keys, &keys[0],
+                  &values[0], &statuses[0]);
+
+    ASSERT_OK(statuses[0]);
+    ASSERT_EQ(values[0], first_value);
+
+    ASSERT_TRUE(statuses[1].IsIncomplete());
+
+    ASSERT_TRUE(statuses[2].IsIncomplete());
+  }
+}
+
 TEST_F(DBBlobBasicTest, GetBlob_CorruptIndex) {
   Options options = GetDefaultOptions();
   options.enable_blob_files = true;
@@ -175,6 +256,48 @@ TEST_P(DBBlobBasicIOErrorTest, GetBlob_IOError) {
   SyncPoint::GetInstance()->ClearAllCallBacks();
 }
 
+TEST_P(DBBlobBasicIOErrorTest, MultiGetBlobs_IOError) {
+  Options options = GetDefaultOptions();
+  options.env = fault_injection_env_.get();
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+
+  Reopen(options);
+
+  constexpr size_t num_keys = 2;
+
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "first_value";
+
+  ASSERT_OK(Put(first_key, first_value));
+
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "second_value";
+
+  ASSERT_OK(Put(second_key, second_value));
+
+  ASSERT_OK(Flush());
+
+  std::array<Slice, num_keys> keys{{first_key, second_key}};
+  std::array<PinnableSlice, num_keys> values;
+  std::array<Status, num_keys> statuses;
+
+  SyncPoint::GetInstance()->SetCallBack(sync_point_, [this](void* /* arg */) {
+    fault_injection_env_->SetFilesystemActive(false,
+                                              Status::IOError(sync_point_));
+  });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  db_->MultiGet(ReadOptions(), db_->DefaultColumnFamily(), num_keys, &keys[0],
+                &values[0], &statuses[0]);
+
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+
+  ASSERT_TRUE(statuses[0].IsIOError());
+  ASSERT_TRUE(statuses[1].IsIOError());
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/blob/db_blob_index_test.cc

@@ -98,7 +98,7 @@ class DBBlobIndexTest : public DBTestBase {
   ArenaWrappedDBIter* GetBlobIterator() {
     return dbfull()->NewIteratorImpl(
         ReadOptions(), cfd(), dbfull()->GetLatestSequenceNumber(),
-        nullptr /*read_callback*/, true /*allow_blob*/);
+        nullptr /*read_callback*/, true /*expose_blob_index*/);
   }
 
   Options GetTestOptions() {
@@ -238,8 +238,11 @@ TEST_F(DBBlobIndexTest, Updated) {
   }
 }
 
-// Iterator should get blob value if allow_blob flag is set,
-// otherwise return Status::NotSupported status.
+// Note: the following test case pertains to the StackableDB-based BlobDB
+// implementation. When a blob iterator is used, it should set the
+// expose_blob_index flag for the underlying DBIter, and retrieve/return the
+// corresponding blob value. If a regular DBIter is created (i.e.
+// expose_blob_index is not set), it should return Status::Corruption.
 TEST_F(DBBlobIndexTest, Iterate) {
   const std::vector<std::vector<ValueType>> data = {
       /*00*/ {kTypeValue},
@@ -302,6 +305,7 @@ TEST_F(DBBlobIndexTest, Iterate) {
                     std::function<void(Iterator*)> extra_check = nullptr) {
     // Seek
     auto* iterator = create_iterator();
+    ASSERT_OK(iterator->status());
     ASSERT_OK(iterator->Refresh());
     iterator->Seek(get_key(index));
     check_iterator(iterator, expected_status, forward_value);
@@ -315,6 +319,7 @@ TEST_F(DBBlobIndexTest, Iterate) {
     ASSERT_OK(iterator->Refresh());
     iterator->Seek(get_key(index - 1));
     ASSERT_TRUE(iterator->Valid());
+    ASSERT_OK(iterator->status());
     iterator->Next();
     check_iterator(iterator, expected_status, forward_value);
     if (extra_check) {
@@ -324,6 +329,7 @@ TEST_F(DBBlobIndexTest, Iterate) {
 
     // SeekForPrev
     iterator = create_iterator();
+    ASSERT_OK(iterator->status());
     ASSERT_OK(iterator->Refresh());
     iterator->SeekForPrev(get_key(index));
     check_iterator(iterator, expected_status, backward_value);
@@ -336,6 +342,7 @@ TEST_F(DBBlobIndexTest, Iterate) {
     iterator = create_iterator();
     iterator->Seek(get_key(index + 1));
     ASSERT_TRUE(iterator->Valid());
+    ASSERT_OK(iterator->status());
     iterator->Prev();
     check_iterator(iterator, expected_status, backward_value);
     if (extra_check) {
@@ -373,7 +380,7 @@ TEST_F(DBBlobIndexTest, Iterate) {
             ASSERT_OK(Write(&batch));
             break;
           default:
-            assert(false);
+            FAIL();
         };
       }
       snapshots.push_back(dbfull()->GetSnapshot());
@@ -384,8 +391,8 @@ TEST_F(DBBlobIndexTest, Iterate) {
     MoveDataTo(tier);
 
     // Normal iterator
-    verify(1, Status::kNotSupported, "", "", create_normal_iterator);
-    verify(3, Status::kNotSupported, "", "", create_normal_iterator);
+    verify(1, Status::kCorruption, "", "", create_normal_iterator);
+    verify(3, Status::kCorruption, "", "", create_normal_iterator);
     verify(5, Status::kOk, get_value(5, 0), get_value(5, 0),
            create_normal_iterator);
     verify(7, Status::kOk, get_value(8, 0), get_value(6, 0),

db/column_family.cc

@@ -449,9 +449,7 @@ void SuperVersion::Cleanup() {
     to_delete.push_back(m);
   }
   current->Unref();
-  if (cfd->Unref()) {
-    delete cfd;
-  }
+  cfd->UnrefAndTryDelete(this);
 }
 
 void SuperVersion::Init(ColumnFamilyData* new_cfd, MemTable* new_mem,
@@ -660,7 +658,7 @@ ColumnFamilyData::~ColumnFamilyData() {
   }
 }
 
-bool ColumnFamilyData::UnrefAndTryDelete() {
+bool ColumnFamilyData::UnrefAndTryDelete(SuperVersion* sv_under_cleanup) {
   int old_refs = refs_.fetch_sub(1);
   assert(old_refs > 0);
 
@@ -670,7 +668,11 @@ bool ColumnFamilyData::UnrefAndTryDelete() {
     return true;
   }
 
-  if (old_refs == 2 && super_version_ != nullptr) {
+  // If called under SuperVersion::Cleanup, we should not re-enter Cleanup on
+  // the same SuperVersion. (But while installing a new SuperVersion, this
+  // cfd could be referenced only by two SuperVersions.)
+  if (old_refs == 2 && super_version_ != nullptr &&
+      super_version_ != sv_under_cleanup) {
     // Only the super_version_ holds me
     SuperVersion* sv = super_version_;
     super_version_ = nullptr;
@@ -708,9 +710,7 @@ uint64_t ColumnFamilyData::OldestLogToKeep() {
   auto current_log = GetLogNumber();
 
   if (allow_2pc_) {
-    autovector<MemTable*> empty_list;
-    auto imm_prep_log =
-        imm()->PrecomputeMinLogContainingPrepSection(empty_list);
+    auto imm_prep_log = imm()->PrecomputeMinLogContainingPrepSection();
     auto mem_prep_log = mem()->GetMinLogContainingPrepSection();
 
     if (imm_prep_log > 0 && imm_prep_log < current_log) {
@@ -1339,6 +1339,15 @@ Status ColumnFamilyData::ValidateOptions(
           "Block-Based Table format. ");
     }
   }
+
+  if (cf_options.enable_blob_garbage_collection &&
+      (cf_options.blob_garbage_collection_age_cutoff < 0.0 ||
+       cf_options.blob_garbage_collection_age_cutoff > 1.0)) {
+    return Status::InvalidArgument(
+        "The age cutoff for blob garbage collection should be in the range "
+        "[0.0, 1.0].");
+  }
+
   return s;
 }
 

db/column_family.h

@@ -279,21 +279,11 @@ class ColumnFamilyData {
   // holding a DB mutex, or as the leader in a write batch group).
   void Ref() { refs_.fetch_add(1); }
 
-  // Unref decreases the reference count, but does not handle deletion
-  // when the count goes to 0.  If this method returns true then the
-  // caller should delete the instance immediately, or later, by calling
-  // FreeDeadColumnFamilies().  Unref() can only be called while holding
-  // a DB mutex, or during single-threaded recovery.
-  bool Unref() {
-    int old_refs = refs_.fetch_sub(1);
-    assert(old_refs > 0);
-    return old_refs == 1;
-  }
-
   // UnrefAndTryDelete() decreases the reference count and do free if needed,
   // return true if this is freed else false, UnrefAndTryDelete() can only
   // be called while holding a DB mutex, or during single-threaded recovery.
-  bool UnrefAndTryDelete();
+  // sv_under_cleanup is only provided when called from SuperVersion::Cleanup.
+  bool UnrefAndTryDelete(SuperVersion* sv_under_cleanup = nullptr);
 
   // SetDropped() can only be called under following conditions:
   // 1) Holding a DB mutex,
@@ -360,6 +350,11 @@ class ColumnFamilyData {
 
   MemTableList* imm() { return &imm_; }
   MemTable* mem() { return mem_; }
+
+  bool IsEmpty() {
+    return mem()->GetFirstSequenceNumber() == 0 && imm()->NumNotFlushed() == 0;
+  }
+
   Version* current() { return current_; }
   Version* dummy_versions() { return dummy_versions_; }
   void SetCurrent(Version* _current);
@@ -509,6 +504,21 @@ class ColumnFamilyData {
 
   FSDirectory* GetDataDir(size_t path_id) const;
 
+  // full_history_ts_low_ can only increase.
+  void SetFullHistoryTsLow(std::string ts_low) {
+    assert(!ts_low.empty());
+    const Comparator* ucmp = user_comparator();
+    assert(ucmp);
+    if (full_history_ts_low_.empty() ||
+        ucmp->CompareTimestamp(ts_low, full_history_ts_low_) > 0) {
+      full_history_ts_low_ = std::move(ts_low);
+    }
+  }
+
+  const std::string& GetFullHistoryTsLow() const {
+    return full_history_ts_low_;
+  }
+
   ThreadLocalPtr* TEST_GetLocalSV() { return local_sv_.get(); }
 
  private:
@@ -605,6 +615,8 @@ class ColumnFamilyData {
   std::vector<std::shared_ptr<FSDirectory>> data_dirs_;
 
   bool db_paths_registered_;
+
+  std::string full_history_ts_low_;
 };
 
 // ColumnFamilySet has interesting thread-safety requirements

db/column_family_test.cc

@@ -3391,6 +3391,30 @@ TEST_P(ColumnFamilyTest, MultipleCFPathsTest) {
   }
 }
 
+TEST(ColumnFamilyTest, ValidateBlobGCCutoff) {
+  DBOptions db_options;
+
+  ColumnFamilyOptions cf_options;
+  cf_options.enable_blob_garbage_collection = true;
+
+  cf_options.blob_garbage_collection_age_cutoff = -0.5;
+  ASSERT_TRUE(ColumnFamilyData::ValidateOptions(db_options, cf_options)
+                  .IsInvalidArgument());
+
+  cf_options.blob_garbage_collection_age_cutoff = 0.0;
+  ASSERT_OK(ColumnFamilyData::ValidateOptions(db_options, cf_options));
+
+  cf_options.blob_garbage_collection_age_cutoff = 0.5;
+  ASSERT_OK(ColumnFamilyData::ValidateOptions(db_options, cf_options));
+
+  cf_options.blob_garbage_collection_age_cutoff = 1.0;
+  ASSERT_OK(ColumnFamilyData::ValidateOptions(db_options, cf_options));
+
+  cf_options.blob_garbage_collection_age_cutoff = 1.5;
+  ASSERT_TRUE(ColumnFamilyData::ValidateOptions(db_options, cf_options)
+                  .IsInvalidArgument());
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 #ifdef ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS

db/compaction/compaction_iterator.cc

@@ -5,9 +5,11 @@
 
 #include "db/compaction/compaction_iterator.h"
 
-#include <cinttypes>
+#include <iterator>
+#include <limits>
 
 #include "db/blob/blob_file_builder.h"
+#include "db/blob/blob_index.h"
 #include "db/snapshot_checker.h"
 #include "port/likely.h"
 #include "rocksdb/listener.h"
@@ -94,6 +96,8 @@ CompactionIterator::CompactionIterator(
       current_user_key_sequence_(0),
       current_user_key_snapshot_(0),
       merge_out_iter_(merge_helper_),
+      blob_garbage_collection_cutoff_file_number_(
+          ComputeBlobGarbageCollectionCutoffFileNumber(compaction_.get())),
       current_key_committed_(false),
       cmp_with_history_ts_low_(0) {
   assert(compaction_filter_ == nullptr || compaction_ != nullptr);
@@ -733,40 +737,136 @@ void CompactionIterator::NextFromInput() {
   }
 }
 
+bool CompactionIterator::ExtractLargeValueIfNeededImpl() {
+  if (!blob_file_builder_) {
+    return false;
+  }
+
+  blob_index_.clear();
+  const Status s = blob_file_builder_->Add(user_key(), value_, &blob_index_);
+
+  if (!s.ok()) {
+    status_ = s;
+    valid_ = false;
+
+    return false;
+  }
+
+  if (blob_index_.empty()) {
+    return false;
+  }
+
+  value_ = blob_index_;
+
+  return true;
+}
+
+void CompactionIterator::ExtractLargeValueIfNeeded() {
+  assert(ikey_.type == kTypeValue);
+
+  if (!ExtractLargeValueIfNeededImpl()) {
+    return;
+  }
+
+  ikey_.type = kTypeBlobIndex;
+  current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+}
+
+void CompactionIterator::GarbageCollectBlobIfNeeded() {
+  assert(ikey_.type == kTypeBlobIndex);
+
+  if (!compaction_) {
+    return;
+  }
+
+  // GC for integrated BlobDB
+  if (compaction_->enable_blob_garbage_collection()) {
+    BlobIndex blob_index;
+
+    {
+      const Status s = blob_index.DecodeFrom(value_);
+
+      if (!s.ok()) {
+        status_ = s;
+        valid_ = false;
+
+        return;
+      }
+    }
+
+    if (blob_index.IsInlined() || blob_index.HasTTL()) {
+      status_ = Status::Corruption("Unexpected TTL/inlined blob index");
+      valid_ = false;
+
+      return;
+    }
+
+    if (blob_index.file_number() >=
+        blob_garbage_collection_cutoff_file_number_) {
+      return;
+    }
+
+    const Version* const version = compaction_->input_version();
+    assert(version);
+
+    {
+      const Status s =
+          version->GetBlob(ReadOptions(), user_key(), blob_index, &blob_value_);
+
+      if (!s.ok()) {
+        status_ = s;
+        valid_ = false;
+
+        return;
+      }
+    }
+
+    value_ = blob_value_;
+
+    if (ExtractLargeValueIfNeededImpl()) {
+      return;
+    }
+
+    ikey_.type = kTypeValue;
+    current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+
+    return;
+  }
+
+  // GC for stacked BlobDB
+  if (compaction_filter_) {
+    const auto blob_decision = compaction_filter_->PrepareBlobOutput(
+        user_key(), value_, &compaction_filter_value_);
+
+    if (blob_decision == CompactionFilter::BlobDecision::kCorruption) {
+      status_ =
+          Status::Corruption("Corrupted blob reference encountered during GC");
+      valid_ = false;
+
+      return;
+    }
+
+    if (blob_decision == CompactionFilter::BlobDecision::kIOError) {
+      status_ = Status::IOError("Could not relocate blob during GC");
+      valid_ = false;
+
+      return;
+    }
+
+    if (blob_decision == CompactionFilter::BlobDecision::kChangeValue) {
+      value_ = compaction_filter_value_;
+
+      return;
+    }
+  }
+}
+
 void CompactionIterator::PrepareOutput() {
   if (valid_) {
     if (ikey_.type == kTypeValue) {
-      if (blob_file_builder_) {
-        blob_index_.clear();
-        const Status s =
-            blob_file_builder_->Add(user_key(), value_, &blob_index_);
-
-        if (!s.ok()) {
-          status_ = s;
-          valid_ = false;
-        } else if (!blob_index_.empty()) {
-          value_ = blob_index_;
-          ikey_.type = kTypeBlobIndex;
-          current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
-        }
-      }
+      ExtractLargeValueIfNeeded();
     } else if (ikey_.type == kTypeBlobIndex) {
-      if (compaction_filter_) {
-        const auto blob_decision = compaction_filter_->PrepareBlobOutput(
-            user_key(), value_, &compaction_filter_value_);
-
-        if (blob_decision == CompactionFilter::BlobDecision::kCorruption) {
-          status_ = Status::Corruption(
-              "Corrupted blob reference encountered during GC");
-          valid_ = false;
-        } else if (blob_decision == CompactionFilter::BlobDecision::kIOError) {
-          status_ = Status::IOError("Could not relocate blob during GC");
-          valid_ = false;
-        } else if (blob_decision ==
-                   CompactionFilter::BlobDecision::kChangeValue) {
-          value_ = compaction_filter_value_;
-        }
-      }
+      GarbageCollectBlobIfNeeded();
     }
 
     // Zeroing out the sequence number leads to better compression.
@@ -894,4 +994,30 @@ bool CompactionIterator::IsInEarliestSnapshot(SequenceNumber sequence) {
   return in_snapshot == SnapshotCheckerResult::kInSnapshot;
 }
 
+uint64_t CompactionIterator::ComputeBlobGarbageCollectionCutoffFileNumber(
+    const CompactionProxy* compaction) {
+  if (!compaction) {
+    return 0;
+  }
+
+  if (!compaction->enable_blob_garbage_collection()) {
+    return 0;
+  }
+
+  Version* const version = compaction->input_version();
+  assert(version);
+
+  const VersionStorageInfo* const storage_info = version->storage_info();
+  assert(storage_info);
+
+  const auto& blob_files = storage_info->GetBlobFiles();
+
+  auto it = blob_files.begin();
+  std::advance(
+      it, compaction->blob_garbage_collection_age_cutoff() * blob_files.size());
+
+  return it != blob_files.end() ? it->first
+                                : std::numeric_limits<uint64_t>::max();
+}
+
 }  // namespace ROCKSDB_NAMESPACE

db/compaction/compaction_iterator.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <algorithm>
+#include <cinttypes>
 #include <deque>
 #include <string>
 #include <unordered_set>
@@ -45,6 +46,12 @@ class CompactionIterator {
     virtual bool allow_ingest_behind() const = 0;
 
     virtual bool preserve_deletes() const = 0;
+
+    virtual bool enable_blob_garbage_collection() const = 0;
+
+    virtual double blob_garbage_collection_age_cutoff() const = 0;
+
+    virtual Version* input_version() const = 0;
   };
 
   class RealCompaction : public CompactionProxy {
@@ -53,6 +60,7 @@ class CompactionIterator {
         : compaction_(compaction) {
       assert(compaction_);
       assert(compaction_->immutable_cf_options());
+      assert(compaction_->mutable_cf_options());
     }
 
     int level() const override { return compaction_->level(); }
@@ -80,6 +88,19 @@ class CompactionIterator {
       return compaction_->immutable_cf_options()->preserve_deletes;
     }
 
+    bool enable_blob_garbage_collection() const override {
+      return compaction_->mutable_cf_options()->enable_blob_garbage_collection;
+    }
+
+    double blob_garbage_collection_age_cutoff() const override {
+      return compaction_->mutable_cf_options()
+          ->blob_garbage_collection_age_cutoff;
+    }
+
+    Version* input_version() const override {
+      return compaction_->input_version();
+    }
+
    private:
     const Compaction* compaction_;
   };
@@ -146,11 +167,30 @@ class CompactionIterator {
   // Processes the input stream to find the next output
   void NextFromInput();
 
-  // Do last preparations before presenting the output to the callee. At this
-  // point this only zeroes out the sequence number if possible for better
-  // compression.
+  // Do final preparations before presenting the output to the callee.
   void PrepareOutput();
 
+  // Passes the output value to the blob file builder (if any), and replaces it
+  // with the corresponding blob reference if it has been actually written to a
+  // blob file (i.e. if it passed the value size check). Returns true if the
+  // value got extracted to a blob file, false otherwise.
+  bool ExtractLargeValueIfNeededImpl();
+
+  // Extracts large values as described above, and updates the internal key's
+  // type to kTypeBlobIndex if the value got extracted. Should only be called
+  // for regular values (kTypeValue).
+  void ExtractLargeValueIfNeeded();
+
+  // Relocates valid blobs residing in the oldest blob files if garbage
+  // collection is enabled. Relocated blobs are written to new blob files or
+  // inlined in the LSM tree depending on the current settings (i.e.
+  // enable_blob_files and min_blob_size). Should only be called for blob
+  // references (kTypeBlobIndex).
+  //
+  // Note: the stacked BlobDB implementation's compaction filter based GC
+  // algorithm is also called from here.
+  void GarbageCollectBlobIfNeeded();
+
   // Invoke compaction filter if needed.
   // Return true on success, false on failures (e.g.: kIOError).
   bool InvokeFilterIfNeeded(bool* need_skip, Slice* skip_until);
@@ -191,6 +231,9 @@ class CompactionIterator {
     }
   }
 
+  static uint64_t ComputeBlobGarbageCollectionCutoffFileNumber(
+      const CompactionProxy* compaction);
+
   InternalIterator* input_;
   const Comparator* cmp_;
   MergeHelper* merge_helper_;
@@ -273,7 +316,11 @@ class CompactionIterator {
   // PinnedIteratorsManager used to pin input_ Iterator blocks while reading
   // merge operands and then releasing them after consuming them.
   PinnedIteratorsManager pinned_iters_mgr_;
+
+  uint64_t blob_garbage_collection_cutoff_file_number_;
+
   std::string blob_index_;
+  PinnableSlice blob_value_;
   std::string compaction_filter_value_;
   InternalKey compaction_filter_skip_until_;
   // "level_ptrs" holds indices that remember which file of an associated

db/compaction/compaction_iterator_test.cc

@@ -176,6 +176,12 @@ class FakeCompaction : public CompactionIterator::CompactionProxy {
 
   bool preserve_deletes() const override { return false; }
 
+  bool enable_blob_garbage_collection() const override { return false; }
+
+  double blob_garbage_collection_age_cutoff() const override { return 0.0; }
+
+  Version* input_version() const override { return nullptr; }
+
   bool key_not_exists_beyond_output_level = false;
 
   bool is_bottommost_level = false;

db/compaction/compaction_job.cc

@@ -1555,9 +1555,7 @@ Status CompactionJob::FinishCompactionOutputFile(
           "CompactionJob::FinishCompactionOutputFile:"
           "MaxAllowedSpaceReached");
       InstrumentedMutexLock l(db_mutex_);
-      // Should handle return error?
-      db_error_handler_->SetBGError(s, BackgroundErrorReason::kCompaction)
-          .PermitUncheckedError();
+      db_error_handler_->SetBGError(s, BackgroundErrorReason::kCompaction);
     }
   }
 #endif

db/corruption_test.cc

@@ -21,6 +21,7 @@
 #include "db/version_set.h"
 #include "env/composite_env_wrapper.h"
 #include "file/filename.h"
+#include "port/stack_trace.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/convenience.h"
 #include "rocksdb/db.h"
@@ -42,7 +43,8 @@ static constexpr int kValueSize = 1000;
 
 class CorruptionTest : public testing::Test {
  public:
-  test::ErrorEnv env_;
+  std::shared_ptr<Env> env_guard_;
+  test::ErrorEnv* env_;
   std::string dbname_;
   std::shared_ptr<Cache> tiny_cache_;
   Options options_;
@@ -53,9 +55,19 @@ class CorruptionTest : public testing::Test {
     // set it to 0), test SequenceNumberRecovery will fail, likely because of a
     // bug in recovery code. Keep it 4 for now to make the test passes.
     tiny_cache_ = NewLRUCache(100, 4);
+    Env* base_env = Env::Default();
+#ifndef ROCKSDB_LITE
+    const char* test_env_uri = getenv("TEST_ENV_URI");
+    if (test_env_uri) {
+      Status s = Env::LoadEnv(test_env_uri, &base_env, &env_guard_);
+      EXPECT_OK(s);
+      EXPECT_NE(Env::Default(), base_env);
+    }
+#endif  //! ROCKSDB_LITE
+    env_ = new test::ErrorEnv(base_env);
     options_.wal_recovery_mode = WALRecoveryMode::kTolerateCorruptedTailRecords;
-    options_.env = &env_;
-    dbname_ = test::PerThreadDBPath("corruption_test");
+    options_.env = env_;
+    dbname_ = test::PerThreadDBPath(env_, "corruption_test");
     Status s = DestroyDB(dbname_, options_);
     EXPECT_OK(s);
 
@@ -77,8 +89,11 @@ class CorruptionTest : public testing::Test {
     if (getenv("KEEP_DB")) {
       fprintf(stdout, "db is still at %s\n", dbname_.c_str());
     } else {
-      EXPECT_OK(DestroyDB(dbname_, Options()));
+      Options opts;
+      opts.env = env_->target();
+      EXPECT_OK(DestroyDB(dbname_, opts));
     }
+    delete env_;
   }
 
   void CloseDb() {
@@ -93,7 +108,7 @@ class CorruptionTest : public testing::Test {
     if (opt.env == Options().env) {
       // If env is not overridden, replace it with ErrorEnv.
       // Otherwise, the test already uses a non-default Env.
-      opt.env = &env_;
+      opt.env = env_;
     }
     opt.arena_block_size = 4096;
     BlockBasedTableOptions table_options;
@@ -176,7 +191,7 @@ class CorruptionTest : public testing::Test {
   void Corrupt(FileType filetype, int offset, int bytes_to_corrupt) {
     // Pick file to corrupt
     std::vector<std::string> filenames;
-    ASSERT_OK(env_.GetChildren(dbname_, &filenames));
+    ASSERT_OK(env_->GetChildren(dbname_, &filenames));
     uint64_t number;
     FileType type;
     std::string fname;
@@ -191,7 +206,7 @@ class CorruptionTest : public testing::Test {
     }
     ASSERT_TRUE(!fname.empty()) << filetype;
 
-    ASSERT_OK(test::CorruptFile(&env_, fname, offset, bytes_to_corrupt));
+    ASSERT_OK(test::CorruptFile(env_, fname, offset, bytes_to_corrupt));
   }
 
   // corrupts exactly one file at level `level`. if no file found at level,
@@ -201,7 +216,7 @@ class CorruptionTest : public testing::Test {
     db_->GetLiveFilesMetaData(&metadata);
     for (const auto& m : metadata) {
       if (m.level == level) {
-        ASSERT_OK(test::CorruptFile(&env_, dbname_ + "/" + m.name, offset,
+        ASSERT_OK(test::CorruptFile(env_, dbname_ + "/" + m.name, offset,
                                     bytes_to_corrupt));
         return;
       }
@@ -268,14 +283,14 @@ TEST_F(CorruptionTest, Recovery) {
 }
 
 TEST_F(CorruptionTest, RecoverWriteError) {
-  env_.writable_file_error_ = true;
+  env_->writable_file_error_ = true;
   Status s = TryReopen();
   ASSERT_TRUE(!s.ok());
 }
 
 TEST_F(CorruptionTest, NewFileErrorDuringWrite) {
   // Do enough writing to force minor compaction
-  env_.writable_file_error_ = true;
+  env_->writable_file_error_ = true;
   const int num =
       static_cast<int>(3 + (Options().write_buffer_size / kValueSize));
   std::string value_storage;
@@ -291,8 +306,8 @@ TEST_F(CorruptionTest, NewFileErrorDuringWrite) {
     ASSERT_TRUE(!failed || !s.ok());
   }
   ASSERT_TRUE(!s.ok());
-  ASSERT_GE(env_.num_writable_file_errors_, 1);
-  env_.writable_file_error_ = false;
+  ASSERT_GE(env_->num_writable_file_errors_, 1);
+  env_->writable_file_error_ = false;
   Reopen();
 }
 
@@ -310,7 +325,7 @@ TEST_F(CorruptionTest, TableFile) {
 
 TEST_F(CorruptionTest, VerifyChecksumReadahead) {
   Options options;
-  SpecialEnv senv(Env::Default());
+  SpecialEnv senv(env_->target());
   options.env = &senv;
   // Disable block cache as we are going to check checksum for
   // the same file twice and measure number of reads.
@@ -432,6 +447,7 @@ TEST_F(CorruptionTest, CorruptedDescriptor) {
 
 TEST_F(CorruptionTest, CompactionInputError) {
   Options options;
+  options.env = env_;
   Reopen(&options);
   Build(10);
   DBImpl* dbi = static_cast_with_check<DBImpl>(db_);
@@ -452,6 +468,7 @@ TEST_F(CorruptionTest, CompactionInputError) {
 
 TEST_F(CorruptionTest, CompactionInputErrorParanoid) {
   Options options;
+  options.env = env_;
   options.paranoid_checks = true;
   options.write_buffer_size = 131072;
   options.max_write_buffer_number = 2;
@@ -537,7 +554,7 @@ TEST_F(CorruptionTest, RangeDeletionCorrupted) {
       ImmutableCFOptions(options_), kRangeDelBlock, &range_del_handle));
 
   ASSERT_OK(TryReopen());
-  ASSERT_OK(test::CorruptFile(&env_, filename,
+  ASSERT_OK(test::CorruptFile(env_, filename,
                               static_cast<int>(range_del_handle.offset()), 1));
   ASSERT_TRUE(TryReopen().IsCorruption());
 }
@@ -545,6 +562,7 @@ TEST_F(CorruptionTest, RangeDeletionCorrupted) {
 TEST_F(CorruptionTest, FileSystemStateCorrupted) {
   for (int iter = 0; iter < 2; ++iter) {
     Options options;
+    options.env = env_;
     options.paranoid_checks = true;
     options.create_if_missing = true;
     Reopen(&options);
@@ -561,13 +579,13 @@ TEST_F(CorruptionTest, FileSystemStateCorrupted) {
 
     if (iter == 0) {  // corrupt file size
       std::unique_ptr<WritableFile> file;
-      env_.NewWritableFile(filename, &file, EnvOptions());
+      env_->NewWritableFile(filename, &file, EnvOptions());
       ASSERT_OK(file->Append(Slice("corrupted sst")));
       file.reset();
       Status x = TryReopen(&options);
       ASSERT_TRUE(x.IsCorruption());
     } else {  // delete the file
-      ASSERT_OK(env_.DeleteFile(filename));
+      ASSERT_OK(env_->DeleteFile(filename));
       Status x = TryReopen(&options);
       ASSERT_TRUE(x.IsCorruption());
     }
@@ -583,6 +601,7 @@ static const auto& corruption_modes = {
 
 TEST_F(CorruptionTest, ParanoidFileChecksOnFlush) {
   Options options;
+  options.env = env_;
   options.check_flush_compaction_key_order = false;
   options.paranoid_file_checks = true;
   options.create_if_missing = true;
@@ -610,6 +629,7 @@ TEST_F(CorruptionTest, ParanoidFileChecksOnFlush) {
 
 TEST_F(CorruptionTest, ParanoidFileChecksOnCompact) {
   Options options;
+  options.env = env_;
   options.paranoid_file_checks = true;
   options.create_if_missing = true;
   options.check_flush_compaction_key_order = false;
@@ -639,6 +659,7 @@ TEST_F(CorruptionTest, ParanoidFileChecksOnCompact) {
 
 TEST_F(CorruptionTest, ParanoidFileChecksWithDeleteRangeFirst) {
   Options options;
+  options.env = env_;
   options.check_flush_compaction_key_order = false;
   options.paranoid_file_checks = true;
   options.create_if_missing = true;
@@ -671,6 +692,7 @@ TEST_F(CorruptionTest, ParanoidFileChecksWithDeleteRangeFirst) {
 
 TEST_F(CorruptionTest, ParanoidFileChecksWithDeleteRange) {
   Options options;
+  options.env = env_;
   options.check_flush_compaction_key_order = false;
   options.paranoid_file_checks = true;
   options.create_if_missing = true;
@@ -706,6 +728,7 @@ TEST_F(CorruptionTest, ParanoidFileChecksWithDeleteRange) {
 
 TEST_F(CorruptionTest, ParanoidFileChecksWithDeleteRangeLast) {
   Options options;
+  options.env = env_;
   options.check_flush_compaction_key_order = false;
   options.paranoid_file_checks = true;
   options.create_if_missing = true;
@@ -738,6 +761,7 @@ TEST_F(CorruptionTest, ParanoidFileChecksWithDeleteRangeLast) {
 
 TEST_F(CorruptionTest, LogCorruptionErrorsInCompactionIterator) {
   Options options;
+  options.env = env_;
   options.create_if_missing = true;
   options.allow_data_in_errors = true;
   auto mode = mock::MockTableFactory::kCorruptKey;
@@ -763,6 +787,7 @@ TEST_F(CorruptionTest, LogCorruptionErrorsInCompactionIterator) {
 
 TEST_F(CorruptionTest, CompactionKeyOrderCheck) {
   Options options;
+  options.env = env_;
   options.paranoid_file_checks = false;
   options.create_if_missing = true;
   options.check_flush_compaction_key_order = false;
@@ -786,6 +811,7 @@ TEST_F(CorruptionTest, CompactionKeyOrderCheck) {
 
 TEST_F(CorruptionTest, FlushKeyOrderCheck) {
   Options options;
+  options.env = env_;
   options.paranoid_file_checks = false;
   options.create_if_missing = true;
   ASSERT_OK(db_->SetOptions({{"check_flush_compaction_key_order", "true"}}));
@@ -814,7 +840,6 @@ TEST_F(CorruptionTest, FlushKeyOrderCheck) {
 }
 
 TEST_F(CorruptionTest, DisableKeyOrderCheck) {
-  Options options;
   ASSERT_OK(db_->SetOptions({{"check_flush_compaction_key_order", "false"}}));
   DBImpl* dbi = static_cast_with_check<DBImpl>(db_);
 
@@ -836,7 +861,7 @@ TEST_F(CorruptionTest, DisableKeyOrderCheck) {
 TEST_F(CorruptionTest, VerifyWholeTableChecksum) {
   CloseDb();
   Options options;
-  options.env = &env_;
+  options.env = env_;
   ASSERT_OK(DestroyDB(dbname_, options));
   options.create_if_missing = true;
   options.file_checksum_gen_factory =
@@ -870,8 +895,18 @@ TEST_F(CorruptionTest, VerifyWholeTableChecksum) {
 
 }  // namespace ROCKSDB_NAMESPACE
 
+#ifdef ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS
+extern "C" {
+void RegisterCustomObjects(int argc, char** argv);
+}
+#else
+void RegisterCustomObjects(int /*argc*/, char** /*argv*/) {}
+#endif  // !ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS
+
 int main(int argc, char** argv) {
+  ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
+  RegisterCustomObjects(argc, argv);
   return RUN_ALL_TESTS();
 }
 

db/db_basic_test.cc

@@ -38,7 +38,10 @@ TEST_F(DBBasicTest, OpenWhenOpen) {
   options.env = env_;
   DB* db2 = nullptr;
   Status s = DB::Open(options, dbname_, &db2);
-  ASSERT_NOK(s);
+  ASSERT_NOK(s) << [db2]() {
+    delete db2;
+    return "db2 open: ok";
+  }();
   ASSERT_EQ(Status::Code::kIOError, s.code());
   ASSERT_EQ(Status::SubCode::kNone, s.subcode());
   ASSERT_TRUE(strstr(s.getState(), "lock ") != nullptr);
@@ -402,13 +405,16 @@ TEST_F(DBBasicTest, GetSnapshot) {
 
 TEST_F(DBBasicTest, CheckLock) {
   do {
-    DB* localdb;
+    DB* localdb = nullptr;
     Options options = CurrentOptions();
     ASSERT_OK(TryReopen(options));
 
     // second open should fail
     Status s = DB::Open(options, dbname_, &localdb);
-    ASSERT_NOK(s);
+    ASSERT_NOK(s) << [localdb]() {
+      delete localdb;
+      return "localdb open: ok";
+    }();
 #ifdef OS_LINUX
     ASSERT_TRUE(s.ToString().find("lock ") != std::string::npos);
 #endif  // OS_LINUX
@@ -1355,6 +1361,28 @@ TEST_P(DBMultiGetTestWithParam, MultiGetMultiCFSnapshot) {
   }
 }
 
+TEST_P(DBMultiGetTestWithParam, MultiGetMultiCFUnsorted) {
+  Options options = CurrentOptions();
+  CreateAndReopenWithCF({"one", "two"}, options);
+
+  ASSERT_OK(Put(1, "foo", "bar"));
+  ASSERT_OK(Put(2, "baz", "xyz"));
+  ASSERT_OK(Put(1, "abc", "def"));
+
+  // Note: keys for the same CF do not form a consecutive range
+  std::vector<int> cfs{1, 2, 1};
+  std::vector<std::string> keys{"foo", "baz", "abc"};
+  std::vector<std::string> values;
+
+  values =
+      MultiGet(cfs, keys, /* snapshot */ nullptr, /* batched */ GetParam());
+
+  ASSERT_EQ(values.size(), 3);
+  ASSERT_EQ(values[0], "bar");
+  ASSERT_EQ(values[1], "xyz");
+  ASSERT_EQ(values[2], "def");
+}
+
 INSTANTIATE_TEST_CASE_P(DBMultiGetTestWithParam, DBMultiGetTestWithParam,
                         testing::Bool());
 
@@ -2499,6 +2527,42 @@ TEST_F(DBBasicTest, SkipWALIfMissingTableFiles) {
   ASSERT_FALSE(iter->Valid());
   ASSERT_OK(iter->status());
 }
+
+TEST_F(DBBasicTest, DisableTrackWal) {
+  // If WAL tracking was enabled, and then disabled during reopen,
+  // the previously tracked WALs should be removed from MANIFEST.
+
+  Options options = CurrentOptions();
+  options.track_and_verify_wals_in_manifest = true;
+  // extremely small write buffer size,
+  // so that new WALs are created more frequently.
+  options.write_buffer_size = 100;
+  options.env = env_;
+  DestroyAndReopen(options);
+  for (int i = 0; i < 100; i++) {
+    ASSERT_OK(Put("foo" + std::to_string(i), "value" + std::to_string(i)));
+  }
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable());
+  ASSERT_OK(db_->SyncWAL());
+  // Some WALs are tracked.
+  ASSERT_FALSE(dbfull()->TEST_GetVersionSet()->GetWalSet().GetWals().empty());
+  Close();
+
+  // Disable WAL tracking.
+  options.track_and_verify_wals_in_manifest = false;
+  options.create_if_missing = false;
+  ASSERT_OK(TryReopen(options));
+  // Previously tracked WALs are cleared.
+  ASSERT_TRUE(dbfull()->TEST_GetVersionSet()->GetWalSet().GetWals().empty());
+  Close();
+
+  // Re-enable WAL tracking again.
+  options.track_and_verify_wals_in_manifest = true;
+  options.create_if_missing = false;
+  ASSERT_OK(TryReopen(options));
+  ASSERT_TRUE(dbfull()->TEST_GetVersionSet()->GetWalSet().GetWals().empty());
+  Close();
+}
 #endif  // !ROCKSDB_LITE
 
 TEST_F(DBBasicTest, ManifestChecksumMismatch) {
@@ -2528,6 +2592,64 @@ TEST_F(DBBasicTest, ManifestChecksumMismatch) {
   ASSERT_TRUE(s.IsCorruption());
 }
 
+#ifndef ROCKSDB_LITE
+class DBBasicTestTrackWal : public DBTestBase,
+                            public testing::WithParamInterface<bool> {
+ public:
+  DBBasicTestTrackWal()
+      : DBTestBase("/db_basic_test_track_wal", /*env_do_fsync=*/false) {}
+
+  int CountWalFiles() {
+    VectorLogPtr log_files;
+    dbfull()->GetSortedWalFiles(log_files);
+    return static_cast<int>(log_files.size());
+  };
+};
+
+TEST_P(DBBasicTestTrackWal, DoNotTrackObsoleteWal) {
+  // If a WAL becomes obsolete after flushing, but is not deleted from disk yet,
+  // then if SyncWAL is called afterwards, the obsolete WAL should not be
+  // tracked in MANIFEST.
+
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.track_and_verify_wals_in_manifest = true;
+  options.atomic_flush = GetParam();
+
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"cf"}, options);
+  ASSERT_EQ(handles_.size(), 2);  // default, cf
+  // Do not delete WALs.
+  ASSERT_OK(db_->DisableFileDeletions());
+  constexpr int n = 10;
+  std::vector<std::unique_ptr<LogFile>> wals(n);
+  for (size_t i = 0; i < n; i++) {
+    // Generate a new WAL for each key-value.
+    const int cf = i % 2;
+    ASSERT_OK(db_->GetCurrentWalFile(&wals[i]));
+    ASSERT_OK(Put(cf, "k" + std::to_string(i), "v" + std::to_string(i)));
+    ASSERT_OK(Flush({0, 1}));
+  }
+  ASSERT_EQ(CountWalFiles(), n);
+  // Since all WALs are obsolete, no WAL should be tracked in MANIFEST.
+  ASSERT_OK(db_->SyncWAL());
+
+  // Manually delete all WALs.
+  Close();
+  for (const auto& wal : wals) {
+    ASSERT_OK(env_->DeleteFile(LogFileName(dbname_, wal->LogNumber())));
+  }
+
+  // If SyncWAL tracks the obsolete WALs in MANIFEST,
+  // reopen will fail because the WALs are missing from disk.
+  ASSERT_OK(TryReopenWithColumnFamilies({"default", "cf"}, options));
+  Destroy(options);
+}
+
+INSTANTIATE_TEST_CASE_P(DBBasicTestTrackWal, DBBasicTestTrackWal,
+                        testing::Bool());
+#endif  // ROCKSDB_LITE
+
 class DBBasicTestMultiGet : public DBTestBase {
  public:
   DBBasicTestMultiGet(std::string test_dir, int num_cfs, bool compressed_cache,

db/db_block_cache_test.cc

@@ -677,7 +677,7 @@ TEST_F(DBBlockCacheTest, ParanoidFileChecks) {
   // Create a new SST file. This will further trigger a compaction
   // and generate another file.
   ASSERT_OK(Flush(1));
-  dbfull()->TEST_WaitForCompact();
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
   ASSERT_EQ(3, /* Totally 3 files created up to now */
             TestGetTickerCount(options, BLOCK_CACHE_ADD));
 
@@ -692,7 +692,7 @@ TEST_F(DBBlockCacheTest, ParanoidFileChecks) {
   ASSERT_OK(Put(1, "1_key4", "val4"));
   ASSERT_OK(Put(1, "9_key4", "val4"));
   ASSERT_OK(Flush(1));
-  dbfull()->TEST_WaitForCompact();
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
   ASSERT_EQ(3, /* Totally 3 files created up to now */
             TestGetTickerCount(options, BLOCK_CACHE_ADD));
 }
@@ -860,7 +860,7 @@ TEST_F(DBBlockCacheTest, CacheCompressionDict) {
       }
       ASSERT_OK(Flush());
     }
-    dbfull()->TEST_WaitForCompact();
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
     ASSERT_EQ(0, NumTableFilesAtLevel(0));
     ASSERT_EQ(kNumFiles, NumTableFilesAtLevel(1));
 

db/db_compaction_test.cc

@@ -7,6 +7,9 @@
 // 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 <tuple>
+
+#include "db/blob/blob_index.h"
 #include "db/db_test_util.h"
 #include "port/port.h"
 #include "port/stack_trace.h"
@@ -28,6 +31,31 @@ class DBCompactionTest : public DBTestBase {
  public:
   DBCompactionTest()
       : DBTestBase("/db_compaction_test", /*env_do_fsync=*/true) {}
+
+  std::vector<uint64_t> GetBlobFileNumbers() {
+    VersionSet* const versions = dbfull()->TEST_GetVersionSet();
+    assert(versions);
+
+    ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
+    assert(cfd);
+
+    Version* const current = cfd->current();
+    assert(current);
+
+    const VersionStorageInfo* const storage_info = current->storage_info();
+    assert(storage_info);
+
+    const auto& blob_files = storage_info->GetBlobFiles();
+
+    std::vector<uint64_t> result;
+    result.reserve(blob_files.size());
+
+    for (const auto& blob_file : blob_files) {
+      result.emplace_back(blob_file.first);
+    }
+
+    return result;
+  }
 };
 
 class DBCompactionTestWithParam
@@ -6020,6 +6048,233 @@ TEST_P(DBCompactionTestBlobError, CompactionError) {
   }
 }
 
+class DBCompactionTestBlobGC
+    : public DBCompactionTest,
+      public testing::WithParamInterface<std::tuple<double, bool>> {
+ public:
+  DBCompactionTestBlobGC()
+      : blob_gc_age_cutoff_(std::get<0>(GetParam())),
+        updated_enable_blob_files_(std::get<1>(GetParam())) {}
+
+  double blob_gc_age_cutoff_;
+  bool updated_enable_blob_files_;
+};
+
+INSTANTIATE_TEST_CASE_P(DBCompactionTestBlobGC, DBCompactionTestBlobGC,
+                        ::testing::Combine(::testing::Values(0.0, 0.5, 1.0),
+                                           ::testing::Bool()));
+
+TEST_P(DBCompactionTestBlobGC, CompactionWithBlobGC) {
+  Options options;
+  options.env = env_;
+  options.disable_auto_compactions = true;
+  options.enable_blob_files = true;
+  options.blob_file_size = 32;  // one blob per file
+  options.enable_blob_garbage_collection = true;
+  options.blob_garbage_collection_age_cutoff = blob_gc_age_cutoff_;
+
+  Reopen(options);
+
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "first_value";
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "second_value";
+
+  ASSERT_OK(Put(first_key, first_value));
+  ASSERT_OK(Put(second_key, second_value));
+  ASSERT_OK(Flush());
+
+  constexpr char third_key[] = "third_key";
+  constexpr char third_value[] = "third_value";
+  constexpr char fourth_key[] = "fourth_key";
+  constexpr char fourth_value[] = "fourth_value";
+
+  ASSERT_OK(Put(third_key, third_value));
+  ASSERT_OK(Put(fourth_key, fourth_value));
+  ASSERT_OK(Flush());
+
+  const std::vector<uint64_t> original_blob_files = GetBlobFileNumbers();
+
+  ASSERT_EQ(original_blob_files.size(), 4);
+
+  const size_t cutoff_index = static_cast<size_t>(
+      options.blob_garbage_collection_age_cutoff * original_blob_files.size());
+
+  // Note: turning off enable_blob_files before the compaction results in
+  // garbage collected values getting inlined.
+  size_t expected_number_of_files = original_blob_files.size();
+
+  if (!updated_enable_blob_files_) {
+    ASSERT_OK(db_->SetOptions({{"enable_blob_files", "false"}}));
+
+    expected_number_of_files -= cutoff_index;
+  }
+
+  constexpr Slice* begin = nullptr;
+  constexpr Slice* end = nullptr;
+
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), begin, end));
+
+  ASSERT_EQ(Get(first_key), first_value);
+  ASSERT_EQ(Get(second_key), second_value);
+  ASSERT_EQ(Get(third_key), third_value);
+  ASSERT_EQ(Get(fourth_key), fourth_value);
+
+  const std::vector<uint64_t> new_blob_files = GetBlobFileNumbers();
+
+  ASSERT_EQ(new_blob_files.size(), expected_number_of_files);
+
+  // Original blob files below the cutoff should be gone, original blob files at
+  // or above the cutoff should be still there
+  for (size_t i = cutoff_index; i < original_blob_files.size(); ++i) {
+    ASSERT_EQ(new_blob_files[i - cutoff_index], original_blob_files[i]);
+  }
+}
+
+TEST_F(DBCompactionTest, CompactionWithBlobGCError_CorruptIndex) {
+  Options options;
+  options.env = env_;
+  options.disable_auto_compactions = true;
+  options.enable_blob_files = true;
+  options.enable_blob_garbage_collection = true;
+  options.blob_garbage_collection_age_cutoff = 1.0;
+
+  Reopen(options);
+
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "first_value";
+  ASSERT_OK(Put(first_key, first_value));
+
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "second_value";
+  ASSERT_OK(Put(second_key, second_value));
+
+  ASSERT_OK(Flush());
+
+  constexpr char third_key[] = "third_key";
+  constexpr char third_value[] = "third_value";
+  ASSERT_OK(Put(third_key, third_value));
+
+  constexpr char fourth_key[] = "fourth_key";
+  constexpr char corrupt_blob_index[] = "foobar";
+
+  WriteBatch batch;
+  ASSERT_OK(WriteBatchInternal::PutBlobIndex(&batch, 0, fourth_key,
+                                             corrupt_blob_index));
+  ASSERT_OK(db_->Write(WriteOptions(), &batch));
+
+  ASSERT_OK(Flush());
+
+  constexpr Slice* begin = nullptr;
+  constexpr Slice* end = nullptr;
+
+  ASSERT_TRUE(
+      db_->CompactRange(CompactRangeOptions(), begin, end).IsCorruption());
+}
+
+TEST_F(DBCompactionTest, CompactionWithBlobGCError_InlinedTTLIndex) {
+  constexpr uint64_t min_blob_size = 10;
+
+  Options options;
+  options.env = env_;
+  options.disable_auto_compactions = true;
+  options.enable_blob_files = true;
+  options.min_blob_size = min_blob_size;
+  options.enable_blob_garbage_collection = true;
+  options.blob_garbage_collection_age_cutoff = 1.0;
+
+  Reopen(options);
+
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "first_value";
+  ASSERT_OK(Put(first_key, first_value));
+
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "second_value";
+  ASSERT_OK(Put(second_key, second_value));
+
+  ASSERT_OK(Flush());
+
+  constexpr char third_key[] = "third_key";
+  constexpr char third_value[] = "third_value";
+  ASSERT_OK(Put(third_key, third_value));
+
+  constexpr char fourth_key[] = "fourth_key";
+  constexpr char blob[] = "short";
+  static_assert(sizeof(short) - 1 < min_blob_size,
+                "Blob too long to be inlined");
+
+  // Fake an inlined TTL blob index.
+  std::string blob_index;
+
+  constexpr uint64_t expiration = 1234567890;
+
+  BlobIndex::EncodeInlinedTTL(&blob_index, expiration, blob);
+
+  WriteBatch batch;
+  ASSERT_OK(
+      WriteBatchInternal::PutBlobIndex(&batch, 0, fourth_key, blob_index));
+  ASSERT_OK(db_->Write(WriteOptions(), &batch));
+
+  ASSERT_OK(Flush());
+
+  constexpr Slice* begin = nullptr;
+  constexpr Slice* end = nullptr;
+
+  ASSERT_TRUE(
+      db_->CompactRange(CompactRangeOptions(), begin, end).IsCorruption());
+}
+
+TEST_F(DBCompactionTest, CompactionWithBlobGCError_IndexWithInvalidFileNumber) {
+  Options options;
+  options.env = env_;
+  options.disable_auto_compactions = true;
+  options.enable_blob_files = true;
+  options.enable_blob_garbage_collection = true;
+  options.blob_garbage_collection_age_cutoff = 1.0;
+
+  Reopen(options);
+
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "first_value";
+  ASSERT_OK(Put(first_key, first_value));
+
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "second_value";
+  ASSERT_OK(Put(second_key, second_value));
+
+  ASSERT_OK(Flush());
+
+  constexpr char third_key[] = "third_key";
+  constexpr char third_value[] = "third_value";
+  ASSERT_OK(Put(third_key, third_value));
+
+  constexpr char fourth_key[] = "fourth_key";
+
+  // Fake a blob index referencing a non-existent blob file.
+  std::string blob_index;
+
+  constexpr uint64_t blob_file_number = 1000;
+  constexpr uint64_t offset = 1234;
+  constexpr uint64_t size = 5678;
+
+  BlobIndex::EncodeBlob(&blob_index, blob_file_number, offset, size,
+                        kNoCompression);
+
+  WriteBatch batch;
+  ASSERT_OK(
+      WriteBatchInternal::PutBlobIndex(&batch, 0, fourth_key, blob_index));
+  ASSERT_OK(db_->Write(WriteOptions(), &batch));
+
+  ASSERT_OK(Flush());
+
+  constexpr Slice* begin = nullptr;
+  constexpr Slice* end = nullptr;
+
+  ASSERT_TRUE(
+      db_->CompactRange(CompactRangeOptions(), begin, end).IsCorruption());
+}
+
 #endif  // !defined(ROCKSDB_LITE)
 
 }  // namespace ROCKSDB_NAMESPACE

db/db_flush_test.cc

@@ -14,6 +14,7 @@
 #include "file/filename.h"
 #include "port/port.h"
 #include "port/stack_trace.h"
+#include "rocksdb/utilities/transaction_db.h"
 #include "test_util/sync_point.h"
 #include "util/cast_util.h"
 #include "util/mutexlock.h"
@@ -620,6 +621,96 @@ TEST_P(DBFlushTestBlobError, FlushError) {
 #endif  // ROCKSDB_LITE
 }
 
+#ifndef ROCKSDB_LITE
+TEST_P(DBAtomicFlushTest, ManualFlushUnder2PC) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.allow_2pc = true;
+  options.atomic_flush = GetParam();
+  // 64MB so that memtable flush won't be trigger by the small writes.
+  options.write_buffer_size = (static_cast<size_t>(64) << 20);
+
+  // Destroy the DB to recreate as a TransactionDB.
+  Close();
+  Destroy(options, true);
+
+  // Create a TransactionDB.
+  TransactionDB* txn_db = nullptr;
+  TransactionDBOptions txn_db_opts;
+  txn_db_opts.write_policy = TxnDBWritePolicy::WRITE_COMMITTED;
+  ASSERT_OK(TransactionDB::Open(options, txn_db_opts, dbname_, &txn_db));
+  ASSERT_NE(txn_db, nullptr);
+  db_ = txn_db;
+
+  // Create two more columns other than default CF.
+  std::vector<std::string> cfs = {"puppy", "kitty"};
+  CreateColumnFamilies(cfs, options);
+  ASSERT_EQ(handles_.size(), 2);
+  ASSERT_EQ(handles_[0]->GetName(), cfs[0]);
+  ASSERT_EQ(handles_[1]->GetName(), cfs[1]);
+  const size_t kNumCfToFlush = options.atomic_flush ? 2 : 1;
+
+  WriteOptions wopts;
+  TransactionOptions txn_opts;
+  // txn1 only prepare, but does not commit.
+  // The WAL containing the prepared but uncommitted data must be kept.
+  Transaction* txn1 = txn_db->BeginTransaction(wopts, txn_opts, nullptr);
+  // txn2 not only prepare, but also commit.
+  Transaction* txn2 = txn_db->BeginTransaction(wopts, txn_opts, nullptr);
+  ASSERT_NE(txn1, nullptr);
+  ASSERT_NE(txn2, nullptr);
+  for (size_t i = 0; i < kNumCfToFlush; i++) {
+    ASSERT_OK(txn1->Put(handles_[i], "k1", "v1"));
+    ASSERT_OK(txn2->Put(handles_[i], "k2", "v2"));
+  }
+  // A txn must be named before prepare.
+  ASSERT_OK(txn1->SetName("txn1"));
+  ASSERT_OK(txn2->SetName("txn2"));
+  // Prepare writes to WAL, but not to memtable. (WriteCommitted)
+  ASSERT_OK(txn1->Prepare());
+  ASSERT_OK(txn2->Prepare());
+  // Commit writes to memtable.
+  ASSERT_OK(txn2->Commit());
+  delete txn1;
+  delete txn2;
+
+  // There are still data in memtable not flushed.
+  // But since data is small enough to reside in the active memtable,
+  // there are no immutable memtable.
+  for (size_t i = 0; i < kNumCfToFlush; i++) {
+    auto cfh = static_cast<ColumnFamilyHandleImpl*>(handles_[i]);
+    ASSERT_EQ(0, cfh->cfd()->imm()->NumNotFlushed());
+    ASSERT_FALSE(cfh->cfd()->mem()->IsEmpty());
+  }
+
+  // Atomic flush memtables,
+  // the min log with prepared data should be written to MANIFEST.
+  std::vector<ColumnFamilyHandle*> cfs_to_flush(kNumCfToFlush);
+  for (size_t i = 0; i < kNumCfToFlush; i++) {
+    cfs_to_flush[i] = handles_[i];
+  }
+  ASSERT_OK(txn_db->Flush(FlushOptions(), cfs_to_flush));
+
+  // There are no remaining data in memtable after flush.
+  for (size_t i = 0; i < kNumCfToFlush; i++) {
+    auto cfh = static_cast<ColumnFamilyHandleImpl*>(handles_[i]);
+    ASSERT_EQ(0, cfh->cfd()->imm()->NumNotFlushed());
+    ASSERT_TRUE(cfh->cfd()->mem()->IsEmpty());
+    ASSERT_EQ(cfh->cfd()->GetFlushReason(), FlushReason::kManualFlush);
+  }
+
+  // The recovered min log number with prepared data should be non-zero.
+  // In 2pc mode, MinLogNumberToKeep returns the
+  // VersionSet::min_log_number_to_keep_2pc recovered from MANIFEST, if it's 0,
+  // it means atomic flush didn't write the min_log_number_to_keep to MANIFEST.
+  cfs.push_back(kDefaultColumnFamilyName);
+  ASSERT_OK(TryReopenWithColumnFamilies(cfs, options));
+  DBImpl* db_impl = reinterpret_cast<DBImpl*>(db_);
+  ASSERT_TRUE(db_impl->allow_2pc());
+  ASSERT_NE(db_impl->MinLogNumberToKeep(), 0);
+}
+#endif  // ROCKSDB_LITE
+
 TEST_P(DBAtomicFlushTest, ManualAtomicFlush) {
   Options options = CurrentOptions();
   options.create_if_missing = true;
@@ -634,18 +725,84 @@ TEST_P(DBAtomicFlushTest, ManualAtomicFlush) {
   for (size_t i = 0; i != num_cfs; ++i) {
     ASSERT_OK(Put(static_cast<int>(i) /*cf*/, "key", "value", wopts));
   }
+
+  for (size_t i = 0; i != num_cfs; ++i) {
+    auto cfh = static_cast<ColumnFamilyHandleImpl*>(handles_[i]);
+    ASSERT_EQ(0, cfh->cfd()->imm()->NumNotFlushed());
+    ASSERT_FALSE(cfh->cfd()->mem()->IsEmpty());
+  }
+
   std::vector<int> cf_ids;
   for (size_t i = 0; i != num_cfs; ++i) {
     cf_ids.emplace_back(static_cast<int>(i));
   }
   ASSERT_OK(Flush(cf_ids));
+
   for (size_t i = 0; i != num_cfs; ++i) {
     auto cfh = static_cast<ColumnFamilyHandleImpl*>(handles_[i]);
+    ASSERT_EQ(cfh->cfd()->GetFlushReason(), FlushReason::kManualFlush);
     ASSERT_EQ(0, cfh->cfd()->imm()->NumNotFlushed());
     ASSERT_TRUE(cfh->cfd()->mem()->IsEmpty());
   }
 }
 
+TEST_P(DBAtomicFlushTest, PrecomputeMinLogNumberToKeepNon2PC) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.atomic_flush = GetParam();
+  options.write_buffer_size = (static_cast<size_t>(64) << 20);
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  const size_t num_cfs = handles_.size();
+  ASSERT_EQ(num_cfs, 2);
+  WriteOptions wopts;
+  for (size_t i = 0; i != num_cfs; ++i) {
+    ASSERT_OK(Put(static_cast<int>(i) /*cf*/, "key", "value", wopts));
+  }
+
+  {
+    // Flush the default CF only.
+    std::vector<int> cf_ids{0};
+    ASSERT_OK(Flush(cf_ids));
+
+    autovector<ColumnFamilyData*> flushed_cfds;
+    autovector<autovector<VersionEdit*>> flush_edits;
+    auto flushed_cfh = static_cast<ColumnFamilyHandleImpl*>(handles_[0]);
+    flushed_cfds.push_back(flushed_cfh->cfd());
+    flush_edits.push_back({});
+    auto unflushed_cfh = static_cast<ColumnFamilyHandleImpl*>(handles_[1]);
+
+    ASSERT_EQ(PrecomputeMinLogNumberToKeepNon2PC(dbfull()->TEST_GetVersionSet(),
+                                                 flushed_cfds, flush_edits),
+              unflushed_cfh->cfd()->GetLogNumber());
+  }
+
+  {
+    // Flush all CFs.
+    std::vector<int> cf_ids;
+    for (size_t i = 0; i != num_cfs; ++i) {
+      cf_ids.emplace_back(static_cast<int>(i));
+    }
+    ASSERT_OK(Flush(cf_ids));
+    uint64_t log_num_after_flush = dbfull()->TEST_GetCurrentLogNumber();
+
+    uint64_t min_log_number_to_keep = port::kMaxUint64;
+    autovector<ColumnFamilyData*> flushed_cfds;
+    autovector<autovector<VersionEdit*>> flush_edits;
+    for (size_t i = 0; i != num_cfs; ++i) {
+      auto cfh = static_cast<ColumnFamilyHandleImpl*>(handles_[i]);
+      flushed_cfds.push_back(cfh->cfd());
+      flush_edits.push_back({});
+      min_log_number_to_keep =
+          std::min(min_log_number_to_keep, cfh->cfd()->GetLogNumber());
+    }
+    ASSERT_EQ(min_log_number_to_keep, log_num_after_flush);
+    ASSERT_EQ(PrecomputeMinLogNumberToKeepNon2PC(dbfull()->TEST_GetVersionSet(),
+                                                 flushed_cfds, flush_edits),
+              min_log_number_to_keep);
+  }
+}
+
 TEST_P(DBAtomicFlushTest, AtomicFlushTriggeredByMemTableFull) {
   Options options = CurrentOptions();
   options.create_if_missing = true;

db/db_impl/db_impl.cc

@@ -1323,7 +1323,8 @@ Status DBImpl::MarkLogsSynced(uint64_t up_to, bool synced_dir) {
     auto& wal = *it;
     assert(wal.getting_synced);
     if (logs_.size() > 1) {
-      if (immutable_db_options_.track_and_verify_wals_in_manifest) {
+      if (immutable_db_options_.track_and_verify_wals_in_manifest &&
+          wal.writer->file()->GetFileSize() > 0) {
         synced_wals.AddWal(wal.number,
                            WalMetadata(wal.writer->file()->GetFileSize()));
       }
@@ -2163,20 +2164,18 @@ void DBImpl::MultiGet(const ReadOptions& read_options, const size_t num_keys,
       multiget_cf_data;
   size_t cf_start = 0;
   ColumnFamilyHandle* cf = sorted_keys[0]->column_family;
+
   for (size_t i = 0; i < num_keys; ++i) {
     KeyContext* key_ctx = sorted_keys[i];
     if (key_ctx->column_family != cf) {
-      multiget_cf_data.emplace_back(
-          MultiGetColumnFamilyData(cf, cf_start, i - cf_start, nullptr));
+      multiget_cf_data.emplace_back(cf, cf_start, i - cf_start, nullptr);
       cf_start = i;
       cf = key_ctx->column_family;
     }
   }
-  {
-    // multiget_cf_data.emplace_back(
-    // MultiGetColumnFamilyData(cf, cf_start, num_keys - cf_start, nullptr));
-    multiget_cf_data.emplace_back(cf, cf_start, num_keys - cf_start, nullptr);
-  }
+
+  multiget_cf_data.emplace_back(cf, cf_start, num_keys - cf_start, nullptr);
+
   std::function<MultiGetColumnFamilyData*(
       autovector<MultiGetColumnFamilyData,
                  MultiGetContext::MAX_BATCH_SIZE>::iterator&)>
@@ -2204,7 +2203,7 @@ void DBImpl::MultiGet(const ReadOptions& read_options, const size_t num_keys,
   for (; cf_iter != multiget_cf_data.end(); ++cf_iter) {
     s = MultiGetImpl(read_options, cf_iter->start, cf_iter->num_keys,
                      &sorted_keys, cf_iter->super_version, consistent_seqnum,
-                     read_callback, nullptr);
+                     read_callback);
     if (!s.ok()) {
       break;
     }
@@ -2236,7 +2235,7 @@ struct CompareKeyContext {
         static_cast<ColumnFamilyHandleImpl*>(lhs->column_family);
     uint32_t cfd_id1 = cfh->cfd()->GetID();
     const Comparator* comparator = cfh->cfd()->user_comparator();
-    cfh = static_cast<ColumnFamilyHandleImpl*>(lhs->column_family);
+    cfh = static_cast<ColumnFamilyHandleImpl*>(rhs->column_family);
     uint32_t cfd_id2 = cfh->cfd()->GetID();
 
     if (cfd_id1 < cfd_id2) {
@@ -2260,39 +2259,24 @@ struct CompareKeyContext {
 void DBImpl::PrepareMultiGetKeys(
     size_t num_keys, bool sorted_input,
     autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* sorted_keys) {
-#ifndef NDEBUG
   if (sorted_input) {
-    for (size_t index = 0; index < sorted_keys->size(); ++index) {
-      if (index > 0) {
-        KeyContext* lhs = (*sorted_keys)[index - 1];
-        KeyContext* rhs = (*sorted_keys)[index];
-        ColumnFamilyHandleImpl* cfh =
-            static_cast_with_check<ColumnFamilyHandleImpl>(lhs->column_family);
-        uint32_t cfd_id1 = cfh->cfd()->GetID();
-        const Comparator* comparator = cfh->cfd()->user_comparator();
-        cfh =
-            static_cast_with_check<ColumnFamilyHandleImpl>(lhs->column_family);
-        uint32_t cfd_id2 = cfh->cfd()->GetID();
+#ifndef NDEBUG
+    CompareKeyContext key_context_less;
 
-        assert(cfd_id1 <= cfd_id2);
-        if (cfd_id1 < cfd_id2) {
-          continue;
-        }
+    for (size_t index = 1; index < sorted_keys->size(); ++index) {
+      const KeyContext* const lhs = (*sorted_keys)[index - 1];
+      const KeyContext* const rhs = (*sorted_keys)[index];
 
-        // Both keys are from the same column family
-        int cmp = comparator->CompareWithoutTimestamp(
-            *(lhs->key), /*a_has_ts=*/false, *(rhs->key), /*b_has_ts=*/false);
-        assert(cmp <= 0);
-      }
-      index++;
+      // lhs should be <= rhs, or in other words, rhs should NOT be < lhs
+      assert(!key_context_less(rhs, lhs));
     }
-  }
 #endif
-  if (!sorted_input) {
-    CompareKeyContext sort_comparator;
-    std::sort(sorted_keys->begin(), sorted_keys->begin() + num_keys,
-              sort_comparator);
+
+    return;
   }
+
+  std::sort(sorted_keys->begin(), sorted_keys->begin() + num_keys,
+            CompareKeyContext());
 }
 
 void DBImpl::MultiGet(const ReadOptions& read_options,
@@ -2376,7 +2360,7 @@ void DBImpl::MultiGetWithCallback(
 
   Status s = MultiGetImpl(read_options, 0, num_keys, sorted_keys,
                           multiget_cf_data[0].super_version, consistent_seqnum,
-                          read_callback, nullptr);
+                          read_callback);
   assert(s.ok() || s.IsTimedOut() || s.IsAborted());
   ReturnAndCleanupSuperVersion(multiget_cf_data[0].cfd,
                                multiget_cf_data[0].super_version);
@@ -2395,7 +2379,7 @@ Status DBImpl::MultiGetImpl(
     const ReadOptions& read_options, size_t start_key, size_t num_keys,
     autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* sorted_keys,
     SuperVersion* super_version, SequenceNumber snapshot,
-    ReadCallback* callback, bool* is_blob_index) {
+    ReadCallback* callback) {
   PERF_CPU_TIMER_GUARD(get_cpu_nanos, env_);
   StopWatch sw(env_, stats_, DB_MULTIGET);
 
@@ -2434,11 +2418,9 @@ Status DBImpl::MultiGetImpl(
         (read_options.read_tier == kPersistedTier &&
          has_unpersisted_data_.load(std::memory_order_relaxed));
     if (!skip_memtable) {
-      super_version->mem->MultiGet(read_options, &range, callback,
-                                   is_blob_index);
+      super_version->mem->MultiGet(read_options, &range, callback);
       if (!range.empty()) {
-        super_version->imm->MultiGet(read_options, &range, callback,
-                                     is_blob_index);
+        super_version->imm->MultiGet(read_options, &range, callback);
       }
       if (!range.empty()) {
         lookup_current = true;
@@ -2448,8 +2430,7 @@ Status DBImpl::MultiGetImpl(
     }
     if (lookup_current) {
       PERF_TIMER_GUARD(get_from_output_files_time);
-      super_version->current->MultiGet(read_options, &range, callback,
-                                       is_blob_index);
+      super_version->current->MultiGet(read_options, &range, callback);
     }
     curr_value_size = range.GetValueSize();
     if (curr_value_size > read_options.value_size_soft_limit) {
@@ -2803,7 +2784,7 @@ Iterator* DBImpl::NewIterator(const ReadOptions& read_options,
                                     /* allow_unprepared_value */ true);
     result = NewDBIterator(
         env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
-        cfd->user_comparator(), iter, kMaxSequenceNumber,
+        cfd->user_comparator(), iter, sv->current, kMaxSequenceNumber,
         sv->mutable_cf_options.max_sequential_skip_in_iterations, read_callback,
         this, cfd);
 #endif
@@ -2824,7 +2805,7 @@ ArenaWrappedDBIter* DBImpl::NewIteratorImpl(const ReadOptions& read_options,
                                             ColumnFamilyData* cfd,
                                             SequenceNumber snapshot,
                                             ReadCallback* read_callback,
-                                            bool allow_blob,
+                                            bool expose_blob_index,
                                             bool allow_refresh) {
   SuperVersion* sv = cfd->GetReferencedSuperVersion(this);
 
@@ -2889,9 +2870,9 @@ ArenaWrappedDBIter* DBImpl::NewIteratorImpl(const ReadOptions& read_options,
   // likely that any iterator pointer is close to the iterator it points to so
   // that they are likely to be in the same cache line and/or page.
   ArenaWrappedDBIter* db_iter = NewArenaWrappedDbIterator(
-      env_, read_options, *cfd->ioptions(), sv->mutable_cf_options, snapshot,
-      sv->mutable_cf_options.max_sequential_skip_in_iterations,
-      sv->version_number, read_callback, this, cfd, allow_blob,
+      env_, read_options, *cfd->ioptions(), sv->mutable_cf_options, sv->current,
+      snapshot, sv->mutable_cf_options.max_sequential_skip_in_iterations,
+      sv->version_number, read_callback, this, cfd, expose_blob_index,
       read_options.snapshot != nullptr ? false : allow_refresh);
 
   InternalIterator* internal_iter = NewInternalIterator(
@@ -2929,7 +2910,7 @@ Status DBImpl::NewIterators(
                                       /* allow_unprepared_value */ true);
       iterators->push_back(NewDBIterator(
           env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
-          cfd->user_comparator(), iter, kMaxSequenceNumber,
+          cfd->user_comparator(), iter, sv->current, kMaxSequenceNumber,
           sv->mutable_cf_options.max_sequential_skip_in_iterations,
           read_callback, this, cfd));
     }
@@ -3400,6 +3381,10 @@ Status DBImpl::GetApproximateSizes(const SizeApproximationOptions& options,
     return Status::InvalidArgument("Invalid options");
   }
 
+  const Comparator* const ucmp = column_family->GetComparator();
+  assert(ucmp);
+  size_t ts_sz = ucmp->timestamp_size();
+
   Version* v;
   auto cfh = static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
   auto cfd = cfh->cfd();
@@ -3407,9 +3392,23 @@ Status DBImpl::GetApproximateSizes(const SizeApproximationOptions& options,
   v = sv->current;
 
   for (int i = 0; i < n; i++) {
+    Slice start = range[i].start;
+    Slice limit = range[i].limit;
+
+    // Add timestamp if needed
+    std::string start_with_ts, limit_with_ts;
+    if (ts_sz > 0) {
+      // Maximum timestamp means including all key with any timestamp
+      AppendKeyWithMaxTimestamp(&start_with_ts, start, ts_sz);
+      // Append a maximum timestamp as the range limit is exclusive:
+      // [start, limit)
+      AppendKeyWithMaxTimestamp(&limit_with_ts, limit, ts_sz);
+      start = start_with_ts;
+      limit = limit_with_ts;
+    }
     // Convert user_key into a corresponding internal key.
-    InternalKey k1(range[i].start, kMaxSequenceNumber, kValueTypeForSeek);
-    InternalKey k2(range[i].limit, kMaxSequenceNumber, kValueTypeForSeek);
+    InternalKey k1(start, kMaxSequenceNumber, kValueTypeForSeek);
+    InternalKey k2(limit, kMaxSequenceNumber, kValueTypeForSeek);
     sizes[i] = 0;
     if (options.include_files) {
       sizes[i] += versions_->ApproximateSize(
@@ -4588,8 +4587,7 @@ Status DBImpl::IngestExternalFiles(
       // TODO: distinguish between MANIFEST write and CURRENT renaming
       const IOStatus& io_s = versions_->io_status();
       // Should handle return error?
-      error_handler_.SetBGError(io_s, BackgroundErrorReason::kManifestWrite)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(io_s, BackgroundErrorReason::kManifestWrite);
     }
 
     // Resume writes to the DB

db/db_impl/db_impl.h

@@ -522,7 +522,7 @@ class DBImpl : public DB {
                                       ColumnFamilyData* cfd,
                                       SequenceNumber snapshot,
                                       ReadCallback* read_callback,
-                                      bool allow_blob = false,
+                                      bool expose_blob_index = false,
                                       bool allow_refresh = true);
 
   virtual SequenceNumber GetLastPublishedSequence() const {
@@ -922,7 +922,7 @@ class DBImpl : public DB {
                            ColumnFamilyHandle* column_family = nullptr,
                            bool disallow_trivial_move = false);
 
-  void TEST_SwitchWAL();
+  Status TEST_SwitchWAL();
 
   bool TEST_UnableToReleaseOldestLog() { return unable_to_release_oldest_log_; }
 
@@ -1017,6 +1017,12 @@ class DBImpl : public DB {
 
   VersionSet* TEST_GetVersionSet() const { return versions_.get(); }
 
+  uint64_t TEST_GetCurrentLogNumber() const {
+    InstrumentedMutexLock l(mutex());
+    assert(!logs_.empty());
+    return logs_.back().number;
+  }
+
   const std::unordered_set<uint64_t>& TEST_GetFilesGrabbedForPurge() const {
     return files_grabbed_for_purge_;
   }
@@ -1104,6 +1110,15 @@ class DBImpl : public DB {
   // If need_enter_write_thread = false, the method will enter write thread.
   Status WriteOptionsFile(bool need_mutex_lock, bool need_enter_write_thread);
 
+  Status CompactRangeInternal(const CompactRangeOptions& options,
+                              ColumnFamilyHandle* column_family,
+                              const Slice* begin, const Slice* end);
+
+  Status GetApproximateSizesInternal(const SizeApproximationOptions& options,
+                                     ColumnFamilyHandle* column_family,
+                                     const Range* range, int n,
+                                     uint64_t* sizes);
+
   // The following two functions can only be called when:
   // 1. WriteThread::Writer::EnterUnbatched() is used.
   // 2. db_mutex is NOT held
@@ -1851,8 +1866,7 @@ class DBImpl : public DB {
   Status MultiGetImpl(
       const ReadOptions& read_options, size_t start_key, size_t num_keys,
       autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* sorted_keys,
-      SuperVersion* sv, SequenceNumber snap_seqnum, ReadCallback* callback,
-      bool* is_blob_index);
+      SuperVersion* sv, SequenceNumber snap_seqnum, ReadCallback* callback);
 
   Status DisableFileDeletionsWithLock();
 
@@ -2219,12 +2233,22 @@ extern uint64_t PrecomputeMinLogNumberToKeep2PC(
     const autovector<VersionEdit*>& edit_list,
     const autovector<MemTable*>& memtables_to_flush,
     LogsWithPrepTracker* prep_tracker);
+// For atomic flush.
+extern uint64_t PrecomputeMinLogNumberToKeep2PC(
+    VersionSet* vset, const autovector<ColumnFamilyData*>& cfds_to_flush,
+    const autovector<autovector<VersionEdit*>>& edit_lists,
+    const autovector<const autovector<MemTable*>*>& memtables_to_flush,
+    LogsWithPrepTracker* prep_tracker);
 
 // In non-2PC mode, WALs with log number < the returned number can be
 // deleted after the cfd_to_flush column family is flushed successfully.
 extern uint64_t PrecomputeMinLogNumberToKeepNon2PC(
     VersionSet* vset, const ColumnFamilyData& cfd_to_flush,
     const autovector<VersionEdit*>& edit_list);
+// For atomic flush.
+extern uint64_t PrecomputeMinLogNumberToKeepNon2PC(
+    VersionSet* vset, const autovector<ColumnFamilyData*>& cfds_to_flush,
+    const autovector<autovector<VersionEdit*>>& edit_lists);
 
 // `cfd_to_flush` is the column family whose memtable will be flushed and thus
 // will not depend on any WAL file. nullptr means no memtable is being flushed.
@@ -2232,6 +2256,10 @@ extern uint64_t PrecomputeMinLogNumberToKeepNon2PC(
 extern uint64_t FindMinPrepLogReferencedByMemTable(
     VersionSet* vset, const ColumnFamilyData* cfd_to_flush,
     const autovector<MemTable*>& memtables_to_flush);
+// For atomic flush.
+extern uint64_t FindMinPrepLogReferencedByMemTable(
+    VersionSet* vset, const autovector<ColumnFamilyData*>& cfds_to_flush,
+    const autovector<const autovector<MemTable*>*>& memtables_to_flush);
 
 // Fix user-supplied options to be reasonable
 template <class T, class V>

db/db_impl/db_impl_compaction_flush.cc

@@ -130,12 +130,10 @@ IOStatus DBImpl::SyncClosedLogs(JobContext* job_context) {
     }
     if (!io_s.ok()) {
       if (total_log_size_ > 0) {
-        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlush)
-            .PermitUncheckedError();
+        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlush);
       } else {
         // If the WAL is empty, we use different error reason
-        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlushNoWAL)
-            .PermitUncheckedError();
+        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlushNoWAL);
       }
       TEST_SYNC_POINT("DBImpl::SyncClosedLogs:Failed");
       return io_s;
@@ -159,14 +157,14 @@ Status DBImpl::FlushMemTableToOutputFile(
 
   FlushJob flush_job(
       dbname_, cfd, immutable_db_options_, mutable_cf_options,
-      nullptr /* memtable_id */, file_options_for_compaction_, versions_.get(),
-      &mutex_, &shutting_down_, snapshot_seqs, earliest_write_conflict_snapshot,
-      snapshot_checker, job_context, log_buffer, directories_.GetDbDir(),
-      GetDataDir(cfd, 0U),
+      port::kMaxUint64 /* memtable_id */, file_options_for_compaction_,
+      versions_.get(), &mutex_, &shutting_down_, snapshot_seqs,
+      earliest_write_conflict_snapshot, snapshot_checker, job_context,
+      log_buffer, directories_.GetDbDir(), GetDataDir(cfd, 0U),
       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_, db_id_, db_session_id_);
+      io_tracer_, db_id_, db_session_id_, cfd->GetFullHistoryTsLow());
   FileMetaData file_meta;
 
   TEST_SYNC_POINT("DBImpl::FlushMemTableToOutputFile:BeforePickMemtables");
@@ -192,8 +190,7 @@ Status DBImpl::FlushMemTableToOutputFile(
     s = io_s;
     if (!io_s.ok() && !io_s.IsShutdownInProgress() &&
         !io_s.IsColumnFamilyDropped()) {
-      error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlush)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlush);
     }
   } else {
     TEST_SYNC_POINT("DBImpl::SyncClosedLogs:Skip");
@@ -253,24 +250,23 @@ Status DBImpl::FlushMemTableToOutputFile(
       // be pessimistic and try write to a new MANIFEST.
       // TODO: distinguish between MANIFEST write and CURRENT renaming
       if (!versions_->io_status().ok()) {
-        // Should handle return error?
-        error_handler_.SetBGError(io_s, BackgroundErrorReason::kManifestWrite)
-            .PermitUncheckedError();
+        if (total_log_size_ > 0) {
+          // If the WAL is empty, we use different error reason
+          error_handler_.SetBGError(io_s,
+                                    BackgroundErrorReason::kManifestWrite);
+        } else {
+          error_handler_.SetBGError(io_s,
+                                    BackgroundErrorReason::kManifestWriteNoWAL);
+        }
       } else if (total_log_size_ > 0) {
-        // Should handle return error?
-        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlush)
-            .PermitUncheckedError();
+        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlush);
       } else {
         // If the WAL is empty, we use different error reason
-        // Should handle return error?
-        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlushNoWAL)
-            .PermitUncheckedError();
+        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlushNoWAL);
       }
     } else {
       Status new_bg_error = s;
-      // Should handle return error?
-      error_handler_.SetBGError(new_bg_error, BackgroundErrorReason::kFlush)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(new_bg_error, BackgroundErrorReason::kFlush);
     }
   } else {
     // If we got here, then we decided not to care about the i_os status (either
@@ -295,9 +291,7 @@ Status DBImpl::FlushMemTableToOutputFile(
         TEST_SYNC_POINT_CALLBACK(
             "DBImpl::FlushMemTableToOutputFile:MaxAllowedSpaceReached",
             &new_bg_error);
-        // Should handle this error?
-        error_handler_.SetBGError(new_bg_error, BackgroundErrorReason::kFlush)
-            .PermitUncheckedError();
+        error_handler_.SetBGError(new_bg_error, BackgroundErrorReason::kFlush);
       }
     }
 #endif  // ROCKSDB_LITE
@@ -313,30 +307,22 @@ Status DBImpl::FlushMemTablesToOutputFiles(
     return AtomicFlushMemTablesToOutputFiles(
         bg_flush_args, made_progress, job_context, log_buffer, thread_pri);
   }
+  assert(bg_flush_args.size() == 1);
   std::vector<SequenceNumber> snapshot_seqs;
   SequenceNumber earliest_write_conflict_snapshot;
   SnapshotChecker* snapshot_checker;
   GetSnapshotContext(job_context, &snapshot_seqs,
                      &earliest_write_conflict_snapshot, &snapshot_checker);
-  Status status;
-  for (auto& arg : bg_flush_args) {
-    ColumnFamilyData* cfd = arg.cfd_;
-    MutableCFOptions mutable_cf_options = *cfd->GetLatestMutableCFOptions();
-    SuperVersionContext* superversion_context = arg.superversion_context_;
-    Status s = FlushMemTableToOutputFile(
-        cfd, mutable_cf_options, made_progress, job_context,
-        superversion_context, snapshot_seqs, earliest_write_conflict_snapshot,
-        snapshot_checker, log_buffer, thread_pri);
-    if (!s.ok()) {
-      status = s;
-      if (!s.IsShutdownInProgress() && !s.IsColumnFamilyDropped()) {
-        // At this point, DB is not shutting down, nor is cfd dropped.
-        // Something is wrong, thus we break out of the loop.
-        break;
-      }
-    }
-  }
-  return status;
+  const auto& bg_flush_arg = bg_flush_args[0];
+  ColumnFamilyData* cfd = bg_flush_arg.cfd_;
+  MutableCFOptions mutable_cf_options = *cfd->GetLatestMutableCFOptions();
+  SuperVersionContext* superversion_context =
+      bg_flush_arg.superversion_context_;
+  Status s = FlushMemTableToOutputFile(
+      cfd, mutable_cf_options, made_progress, job_context, superversion_context,
+      snapshot_seqs, earliest_write_conflict_snapshot, snapshot_checker,
+      log_buffer, thread_pri);
+  return s;
 }
 
 /*
@@ -399,7 +385,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
 
     all_mutable_cf_options.emplace_back(*cfd->GetLatestMutableCFOptions());
     const MutableCFOptions& mutable_cf_options = all_mutable_cf_options.back();
-    const uint64_t* max_memtable_id = &(bg_flush_args[i].max_memtable_id_);
+    uint64_t max_memtable_id = bg_flush_args[i].max_memtable_id_;
     jobs.emplace_back(new FlushJob(
         dbname_, cfd, immutable_db_options_, mutable_cf_options,
         max_memtable_id, file_options_for_compaction_, versions_.get(), &mutex_,
@@ -408,7 +394,8 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
         data_dir, 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_, db_id_, db_session_id_));
+        thread_pri, io_tracer_, db_id_, db_session_id_,
+        cfd->GetFullHistoryTsLow()));
     jobs.back()->PickMemTable();
   }
 
@@ -585,7 +572,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
 
     s = InstallMemtableAtomicFlushResults(
         nullptr /* imm_lists */, tmp_cfds, mutable_cf_options_list, mems_list,
-        versions_.get(), &mutex_, tmp_file_meta,
+        versions_.get(), &logs_with_prep_tracker_, &mutex_, tmp_file_meta,
         &job_context->memtables_to_free, directories_.GetDbDir(), log_buffer);
   }
 
@@ -645,9 +632,8 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
             error_handler_.GetBGError().ok()) {
           Status new_bg_error =
               Status::SpaceLimit("Max allowed space was reached");
-          // Should Handle this error?
-          error_handler_.SetBGError(new_bg_error, BackgroundErrorReason::kFlush)
-              .PermitUncheckedError();
+          error_handler_.SetBGError(new_bg_error,
+                                    BackgroundErrorReason::kFlush);
         }
       }
     }
@@ -664,24 +650,23 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
       // be pessimistic and try write to a new MANIFEST.
       // TODO: distinguish between MANIFEST write and CURRENT renaming
       if (!versions_->io_status().ok()) {
-        // Should Handle this error?
-        error_handler_.SetBGError(io_s, BackgroundErrorReason::kManifestWrite)
-            .PermitUncheckedError();
+        if (total_log_size_ > 0) {
+          // If the WAL is empty, we use different error reason
+          error_handler_.SetBGError(io_s,
+                                    BackgroundErrorReason::kManifestWrite);
+        } else {
+          error_handler_.SetBGError(io_s,
+                                    BackgroundErrorReason::kManifestWriteNoWAL);
+        }
       } else if (total_log_size_ > 0) {
-        // Should Handle this error?
-        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlush)
-            .PermitUncheckedError();
+        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlush);
       } else {
         // If the WAL is empty, we use different error reason
-        // Should Handle this error?
-        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlushNoWAL)
-            .PermitUncheckedError();
+        error_handler_.SetBGError(io_s, BackgroundErrorReason::kFlushNoWAL);
       }
     } else {
       Status new_bg_error = s;
-      // Should Handle this error?
-      error_handler_.SetBGError(new_bg_error, BackgroundErrorReason::kFlush)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(new_bg_error, BackgroundErrorReason::kFlush);
     }
   }
 
@@ -781,7 +766,41 @@ void DBImpl::NotifyOnFlushCompleted(
 
 Status DBImpl::CompactRange(const CompactRangeOptions& options,
                             ColumnFamilyHandle* column_family,
-                            const Slice* begin, const Slice* end) {
+                            const Slice* begin_without_ts,
+                            const Slice* end_without_ts) {
+  const Comparator* const ucmp = column_family->GetComparator();
+  assert(ucmp);
+  size_t ts_sz = ucmp->timestamp_size();
+  if (ts_sz == 0) {
+    return CompactRangeInternal(options, column_family, begin_without_ts,
+                                end_without_ts);
+  }
+
+  std::string begin_str;
+  std::string end_str;
+
+  // CompactRange compact all keys: [begin, end] inclusively. Add maximum
+  // timestamp to include all `begin` keys, and add minimal timestamp to include
+  // all `end` keys.
+  if (begin_without_ts != nullptr) {
+    AppendKeyWithMaxTimestamp(&begin_str, *begin_without_ts, ts_sz);
+  }
+  if (end_without_ts != nullptr) {
+    AppendKeyWithMinTimestamp(&end_str, *end_without_ts, ts_sz);
+  }
+  Slice begin(begin_str);
+  Slice end(end_str);
+
+  Slice* begin_with_ts = begin_without_ts ? &begin : nullptr;
+  Slice* end_with_ts = end_without_ts ? &end : nullptr;
+
+  return CompactRangeInternal(options, column_family, begin_with_ts,
+                              end_with_ts);
+}
+
+Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
+                                    ColumnFamilyHandle* column_family,
+                                    const Slice* begin, const Slice* end) {
   auto cfh = static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
   auto cfd = cfh->cfd();
 
@@ -1160,7 +1179,8 @@ Status DBImpl::CompactFilesImpl(
       c->mutable_cf_options()->paranoid_file_checks,
       c->mutable_cf_options()->report_bg_io_stats, dbname_,
       &compaction_job_stats, Env::Priority::USER, io_tracer_,
-      &manual_compaction_paused_, db_id_, db_session_id_);
+      &manual_compaction_paused_, db_id_, db_session_id_,
+      c->column_family_data()->GetFullHistoryTsLow());
 
   // Creating a compaction influences the compaction score because the score
   // takes running compactions into account (by skipping files that are already
@@ -1224,11 +1244,9 @@ Status DBImpl::CompactFilesImpl(
                    job_context->job_id, status.ToString().c_str());
     IOStatus io_s = compaction_job.io_status();
     if (!io_s.ok()) {
-      error_handler_.SetBGError(io_s, BackgroundErrorReason::kCompaction)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(io_s, BackgroundErrorReason::kCompaction);
     } else {
-      error_handler_.SetBGError(status, BackgroundErrorReason::kCompaction)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(status, BackgroundErrorReason::kCompaction);
     }
   }
 
@@ -1697,8 +1715,9 @@ void DBImpl::GenerateFlushRequest(const autovector<ColumnFamilyData*>& cfds,
 Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
                              const FlushOptions& flush_options,
                              FlushReason flush_reason, bool writes_stopped) {
+  // This method should not be called if atomic_flush is true.
+  assert(!immutable_db_options_.atomic_flush);
   Status s;
-  uint64_t flush_memtable_id = 0;
   if (!flush_options.allow_write_stall) {
     bool flush_needed = true;
     s = WaitUntilFlushWouldNotStallWrites(cfd, &flush_needed);
@@ -1708,7 +1727,8 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
     }
   }
 
-  FlushRequest flush_req;
+  autovector<FlushRequest> flush_reqs;
+  autovector<uint64_t> memtable_ids_to_wait;
   {
     WriteContext context;
     InstrumentedMutexLock guard_lock(&mutex_);
@@ -1723,18 +1743,24 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
     }
     WaitForPendingWrites();
 
-    if (!cfd->mem()->IsEmpty() || !cached_recoverable_state_empty_.load()) {
-      if (flush_reason != FlushReason::kErrorRecoveryRetryFlush) {
-        s = SwitchMemtable(cfd, &context);
-      } else {
-        assert(cfd->imm()->NumNotFlushed() > 0);
-      }
+    if (flush_reason != FlushReason::kErrorRecoveryRetryFlush &&
+        (!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.
+      // 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.
+      s = SwitchMemtable(cfd, &context);
     }
+    const uint64_t flush_memtable_id = port::kMaxUint64;
     if (s.ok()) {
       if (cfd->imm()->NumNotFlushed() != 0 || !cfd->mem()->IsEmpty() ||
           !cached_recoverable_state_empty_.load()) {
-        flush_memtable_id = cfd->imm()->GetLatestMemTableID();
-        flush_req.emplace_back(cfd, flush_memtable_id);
+        FlushRequest req{{cfd, flush_memtable_id}};
+        flush_reqs.emplace_back(std::move(req));
+        memtable_ids_to_wait.emplace_back(cfd->imm()->GetLatestMemTableID());
       }
       if (immutable_db_options_.persist_stats_to_disk &&
           flush_reason != FlushReason::kErrorRecoveryRetryFlush) {
@@ -1760,15 +1786,19 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
                            "to avoid holding old logs",
                            cfd->GetName().c_str());
             s = SwitchMemtable(cfd_stats, &context);
-            flush_memtable_id = cfd_stats->imm()->GetLatestMemTableID();
-            flush_req.emplace_back(cfd_stats, flush_memtable_id);
+            FlushRequest req{{cfd_stats, flush_memtable_id}};
+            flush_reqs.emplace_back(std::move(req));
+            memtable_ids_to_wait.emplace_back(
+                cfd->imm()->GetLatestMemTableID());
           }
         }
       }
     }
-    if (s.ok() && !flush_req.empty()) {
-      for (auto& elem : flush_req) {
-        ColumnFamilyData* loop_cfd = elem.first;
+
+    if (s.ok() && !flush_reqs.empty()) {
+      for (const auto& req : flush_reqs) {
+        assert(req.size() == 1);
+        ColumnFamilyData* loop_cfd = req[0].first;
         loop_cfd->imm()->FlushRequested();
       }
       // If the caller wants to wait for this flush to complete, it indicates
@@ -1776,12 +1806,15 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
       // other threads which may drop the column family concurrently.
       // Therefore, we increase the cfd's ref count.
       if (flush_options.wait) {
-        for (auto& elem : flush_req) {
-          ColumnFamilyData* loop_cfd = elem.first;
+        for (const auto& req : flush_reqs) {
+          assert(req.size() == 1);
+          ColumnFamilyData* loop_cfd = req[0].first;
           loop_cfd->Ref();
         }
       }
-      SchedulePendingFlush(flush_req, flush_reason);
+      for (const auto& req : flush_reqs) {
+        SchedulePendingFlush(req, flush_reason);
+      }
       MaybeScheduleFlushOrCompaction();
     }
 
@@ -1797,9 +1830,11 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
   if (s.ok() && flush_options.wait) {
     autovector<ColumnFamilyData*> cfds;
     autovector<const uint64_t*> flush_memtable_ids;
-    for (auto& iter : flush_req) {
-      cfds.push_back(iter.first);
-      flush_memtable_ids.push_back(&(iter.second));
+    assert(flush_reqs.size() == memtable_ids_to_wait.size());
+    for (size_t i = 0; i < flush_reqs.size(); ++i) {
+      assert(flush_reqs[i].size() == 1);
+      cfds.push_back(flush_reqs[i][0].first);
+      flush_memtable_ids.push_back(&(memtable_ids_to_wait[i]));
     }
     s = WaitForFlushMemTables(
         cfds, flush_memtable_ids,
@@ -2224,6 +2259,17 @@ DBImpl::FlushRequest DBImpl::PopFirstFromFlushQueue() {
   assert(!flush_queue_.empty());
   FlushRequest flush_req = flush_queue_.front();
   flush_queue_.pop_front();
+  if (!immutable_db_options_.atomic_flush) {
+    assert(flush_req.size() == 1);
+  }
+  for (const auto& elem : flush_req) {
+    if (!immutable_db_options_.atomic_flush) {
+      ColumnFamilyData* cfd = elem.first;
+      assert(cfd);
+      assert(cfd->queued_for_flush());
+      cfd->set_queued_for_flush(false);
+    }
+  }
   // TODO: need to unset flush reason?
   return flush_req;
 }
@@ -2256,19 +2302,36 @@ ColumnFamilyData* DBImpl::PickCompactionFromQueue(
 
 void DBImpl::SchedulePendingFlush(const FlushRequest& flush_req,
                                   FlushReason flush_reason) {
+  mutex_.AssertHeld();
   if (flush_req.empty()) {
     return;
   }
-  for (auto& iter : flush_req) {
-    ColumnFamilyData* cfd = iter.first;
-    cfd->Ref();
-    cfd->SetFlushReason(flush_reason);
+  if (!immutable_db_options_.atomic_flush) {
+    // For the non-atomic flush case, we never schedule multiple column
+    // families in the same flush request.
+    assert(flush_req.size() == 1);
+    ColumnFamilyData* cfd = flush_req[0].first;
+    assert(cfd);
+    if (!cfd->queued_for_flush() && cfd->imm()->IsFlushPending()) {
+      cfd->Ref();
+      cfd->set_queued_for_flush(true);
+      cfd->SetFlushReason(flush_reason);
+      ++unscheduled_flushes_;
+      flush_queue_.push_back(flush_req);
+    }
+  } else {
+    for (auto& iter : flush_req) {
+      ColumnFamilyData* cfd = iter.first;
+      cfd->Ref();
+      cfd->SetFlushReason(flush_reason);
+    }
+    ++unscheduled_flushes_;
+    flush_queue_.push_back(flush_req);
   }
-  ++unscheduled_flushes_;
-  flush_queue_.push_back(flush_req);
 }
 
 void DBImpl::SchedulePendingCompaction(ColumnFamilyData* cfd) {
+  mutex_.AssertHeld();
   if (!cfd->queued_for_compaction() && cfd->NeedsCompaction()) {
     AddToCompactionQueue(cfd);
     ++unscheduled_compactions_;
@@ -2563,7 +2626,8 @@ void DBImpl::BackgroundCallCompaction(PrepickedCompaction* prepicked_compaction,
     // failure). Thus, we force full scan in FindObsoleteFiles()
     FindObsoleteFiles(&job_context, !s.ok() && !s.IsShutdownInProgress() &&
                                         !s.IsManualCompactionPaused() &&
-                                        !s.IsColumnFamilyDropped());
+                                        !s.IsColumnFamilyDropped() &&
+                                        !s.IsBusy());
     TEST_SYNC_POINT("DBImpl::BackgroundCallCompaction:FoundObsoleteFiles");
 
     // delete unnecessary files if any, this is done outside the mutex
@@ -2969,7 +3033,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
         c->mutable_cf_options()->report_bg_io_stats, dbname_,
         &compaction_job_stats, thread_pri, io_tracer_,
         is_manual ? &manual_compaction_paused_ : nullptr, db_id_,
-        db_session_id_);
+        db_session_id_, c->column_family_data()->GetFullHistoryTsLow());
     compaction_job.Prepare();
 
     NotifyOnCompactionBegin(c->column_family_data(), c.get(), status,
@@ -3034,10 +3098,9 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       auto err_reason = versions_->io_status().ok()
                             ? BackgroundErrorReason::kCompaction
                             : BackgroundErrorReason::kManifestWrite;
-      error_handler_.SetBGError(io_s, err_reason).PermitUncheckedError();
+      error_handler_.SetBGError(io_s, err_reason);
     } else {
-      error_handler_.SetBGError(status, BackgroundErrorReason::kCompaction)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(status, BackgroundErrorReason::kCompaction);
     }
     if (c != nullptr && !is_manual && !error_handler_.IsBGWorkStopped()) {
       // Put this cfd back in the compaction queue so we can retry after some

db/db_impl/db_impl_debug.cc

@@ -22,12 +22,13 @@ uint64_t DBImpl::TEST_GetLevel0TotalSize() {
   return default_cf_handle_->cfd()->current()->storage_info()->NumLevelBytes(0);
 }
 
-void DBImpl::TEST_SwitchWAL() {
+Status DBImpl::TEST_SwitchWAL() {
   WriteContext write_context;
   InstrumentedMutexLock l(&mutex_);
   void* writer = TEST_BeginWrite();
-  SwitchWAL(&write_context);
+  auto s = SwitchWAL(&write_context);
   TEST_EndWrite(writer);
+  return s;
 }
 
 bool DBImpl::TEST_WALBufferIsEmpty(bool lock) {
@@ -170,7 +171,7 @@ Status DBImpl::TEST_WaitForCompact(bool wait_unscheduled) {
   while ((bg_bottom_compaction_scheduled_ || bg_compaction_scheduled_ ||
           bg_flush_scheduled_ ||
           (wait_unscheduled && unscheduled_compactions_)) &&
-         (error_handler_.GetBGError() == Status::OK())) {
+         (error_handler_.GetBGError().ok())) {
     bg_cv_.Wait();
   }
   return error_handler_.GetBGError();

db/db_impl/db_impl_files.cc

@@ -6,16 +6,17 @@
 // Copyright (c) 2011 The LevelDB Authors. All rights reserved.
 // 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 "db/db_impl/db_impl.h"
-
 #include <cinttypes>
 #include <set>
 #include <unordered_set>
+
+#include "db/db_impl/db_impl.h"
 #include "db/event_helpers.h"
 #include "db/memtable_list.h"
 #include "file/file_util.h"
 #include "file/filename.h"
 #include "file/sst_file_manager_impl.h"
+#include "port/port.h"
 #include "util/autovector.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -658,13 +659,15 @@ uint64_t FindMinPrepLogReferencedByMemTable(
 
   // we must look through the memtables for two phase transactions
   // that have been committed but not yet flushed
+  std::unordered_set<MemTable*> memtables_to_flush_set(
+      memtables_to_flush.begin(), memtables_to_flush.end());
   for (auto loop_cfd : *vset->GetColumnFamilySet()) {
     if (loop_cfd->IsDropped() || loop_cfd == cfd_to_flush) {
       continue;
     }
 
     auto log = loop_cfd->imm()->PrecomputeMinLogContainingPrepSection(
-        memtables_to_flush);
+        &memtables_to_flush_set);
 
     if (log > 0 && (min_log == 0 || log < min_log)) {
       min_log = log;
@@ -680,6 +683,37 @@ uint64_t FindMinPrepLogReferencedByMemTable(
   return min_log;
 }
 
+uint64_t FindMinPrepLogReferencedByMemTable(
+    VersionSet* vset, const autovector<ColumnFamilyData*>& cfds_to_flush,
+    const autovector<const autovector<MemTable*>*>& memtables_to_flush) {
+  uint64_t min_log = 0;
+
+  std::unordered_set<ColumnFamilyData*> cfds_to_flush_set(cfds_to_flush.begin(),
+                                                          cfds_to_flush.end());
+  std::unordered_set<MemTable*> memtables_to_flush_set;
+  for (const autovector<MemTable*>* memtables : memtables_to_flush) {
+    memtables_to_flush_set.insert(memtables->begin(), memtables->end());
+  }
+  for (auto loop_cfd : *vset->GetColumnFamilySet()) {
+    if (loop_cfd->IsDropped() || cfds_to_flush_set.count(loop_cfd)) {
+      continue;
+    }
+
+    auto log = loop_cfd->imm()->PrecomputeMinLogContainingPrepSection(
+        &memtables_to_flush_set);
+    if (log > 0 && (min_log == 0 || log < min_log)) {
+      min_log = log;
+    }
+
+    log = loop_cfd->mem()->GetMinLogContainingPrepSection();
+    if (log > 0 && (min_log == 0 || log < min_log)) {
+      min_log = log;
+    }
+  }
+
+  return min_log;
+}
+
 uint64_t PrecomputeMinLogNumberToKeepNon2PC(
     VersionSet* vset, const ColumnFamilyData& cfd_to_flush,
     const autovector<VersionEdit*>& edit_list) {
@@ -710,6 +744,42 @@ uint64_t PrecomputeMinLogNumberToKeepNon2PC(
   return min_log_number_to_keep;
 }
 
+uint64_t PrecomputeMinLogNumberToKeepNon2PC(
+    VersionSet* vset, const autovector<ColumnFamilyData*>& cfds_to_flush,
+    const autovector<autovector<VersionEdit*>>& edit_lists) {
+  assert(vset != nullptr);
+  assert(!cfds_to_flush.empty());
+  assert(cfds_to_flush.size() == edit_lists.size());
+
+  uint64_t min_log_number_to_keep = port::kMaxUint64;
+  for (const auto& edit_list : edit_lists) {
+    uint64_t log = 0;
+    for (const auto& e : edit_list) {
+      if (e->HasLogNumber()) {
+        log = std::max(log, e->GetLogNumber());
+      }
+    }
+    if (log != 0) {
+      min_log_number_to_keep = std::min(min_log_number_to_keep, log);
+    }
+  }
+  if (min_log_number_to_keep == port::kMaxUint64) {
+    min_log_number_to_keep = cfds_to_flush[0]->GetLogNumber();
+    for (size_t i = 1; i < cfds_to_flush.size(); i++) {
+      min_log_number_to_keep =
+          std::min(min_log_number_to_keep, cfds_to_flush[i]->GetLogNumber());
+    }
+  }
+
+  std::unordered_set<const ColumnFamilyData*> flushed_cfds(
+      cfds_to_flush.begin(), cfds_to_flush.end());
+  min_log_number_to_keep =
+      std::min(min_log_number_to_keep,
+               vset->PreComputeMinLogNumberWithUnflushedData(flushed_cfds));
+
+  return min_log_number_to_keep;
+}
+
 uint64_t PrecomputeMinLogNumberToKeep2PC(
     VersionSet* vset, const ColumnFamilyData& cfd_to_flush,
     const autovector<VersionEdit*>& edit_list,
@@ -751,6 +821,38 @@ uint64_t PrecomputeMinLogNumberToKeep2PC(
   return min_log_number_to_keep;
 }
 
+uint64_t PrecomputeMinLogNumberToKeep2PC(
+    VersionSet* vset, const autovector<ColumnFamilyData*>& cfds_to_flush,
+    const autovector<autovector<VersionEdit*>>& edit_lists,
+    const autovector<const autovector<MemTable*>*>& memtables_to_flush,
+    LogsWithPrepTracker* prep_tracker) {
+  assert(vset != nullptr);
+  assert(prep_tracker != nullptr);
+  assert(cfds_to_flush.size() == edit_lists.size());
+  assert(cfds_to_flush.size() == memtables_to_flush.size());
+
+  uint64_t min_log_number_to_keep =
+      PrecomputeMinLogNumberToKeepNon2PC(vset, cfds_to_flush, edit_lists);
+
+  uint64_t min_log_in_prep_heap =
+      prep_tracker->FindMinLogContainingOutstandingPrep();
+
+  if (min_log_in_prep_heap != 0 &&
+      min_log_in_prep_heap < min_log_number_to_keep) {
+    min_log_number_to_keep = min_log_in_prep_heap;
+  }
+
+  uint64_t min_log_refed_by_mem = FindMinPrepLogReferencedByMemTable(
+      vset, cfds_to_flush, memtables_to_flush);
+
+  if (min_log_refed_by_mem != 0 &&
+      min_log_refed_by_mem < min_log_number_to_keep) {
+    min_log_number_to_keep = min_log_refed_by_mem;
+  }
+
+  return min_log_number_to_keep;
+}
+
 Status DBImpl::SetDBId() {
   Status s;
   // Happens when immutable_db_options_.write_dbid_to_manifest is set to true

db/db_impl/db_impl_open.cc

@@ -1368,6 +1368,9 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
                       cfd->GetName().c_str(), meta.fd.GetNumber(),
                       meta.fd.GetFileSize(), s.ToString().c_str());
       mutex_.Lock();
+
+      io_s.PermitUncheckedError();  // TODO(AR) is this correct, or should we
+                                    // return io_s if not ok()?
     }
   }
   ReleaseFileNumberFromPendingOutputs(pending_outputs_inserted_elem);
@@ -1563,6 +1566,7 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
       InstrumentedMutexLock wl(&impl->log_write_mutex_);
       impl->logfile_number_ = new_log_number;
       assert(new_log != nullptr);
+      assert(impl->logs_.empty());
       impl->logs_.emplace_back(new_log_number, new_log);
     }
 

db/db_impl/db_impl_readonly.cc

@@ -83,7 +83,7 @@ Iterator* DBImplReadOnly::NewIterator(const ReadOptions& read_options,
   ReadCallback* read_callback = nullptr;  // No read callback provided.
   auto db_iter = NewArenaWrappedDbIterator(
       env_, read_options, *cfd->ioptions(), super_version->mutable_cf_options,
-      read_seq,
+      super_version->current, read_seq,
       super_version->mutable_cf_options.max_sequential_skip_in_iterations,
       super_version->version_number, read_callback);
   auto internal_iter = NewInternalIterator(
@@ -115,7 +115,8 @@ Status DBImplReadOnly::NewIterators(
     auto* cfd = static_cast_with_check<ColumnFamilyHandleImpl>(cfh)->cfd();
     auto* sv = cfd->GetSuperVersion()->Ref();
     auto* db_iter = NewArenaWrappedDbIterator(
-        env_, read_options, *cfd->ioptions(), sv->mutable_cf_options, read_seq,
+        env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
+        sv->current, read_seq,
         sv->mutable_cf_options.max_sequential_skip_in_iterations,
         sv->version_number, read_callback);
     auto* internal_iter = NewInternalIterator(

db/db_impl/db_impl_secondary.cc

@@ -421,7 +421,7 @@ ArenaWrappedDBIter* DBImplSecondary::NewIteratorImpl(
   SuperVersion* super_version = cfd->GetReferencedSuperVersion(this);
   auto db_iter = NewArenaWrappedDbIterator(
       env_, read_options, *cfd->ioptions(), super_version->mutable_cf_options,
-      snapshot,
+      super_version->current, snapshot,
       super_version->mutable_cf_options.max_sequential_skip_in_iterations,
       super_version->version_number, read_callback);
   auto internal_iter = NewInternalIterator(

db/db_impl/db_impl_write.cc

@@ -666,7 +666,6 @@ Status DBImpl::WriteImplWALOnly(
     const uint64_t log_ref, uint64_t* seq_used, const size_t sub_batch_cnt,
     PreReleaseCallback* pre_release_callback, const AssignOrder assign_order,
     const PublishLastSeq publish_last_seq, const bool disable_memtable) {
-  Status status;
   PERF_TIMER_GUARD(write_pre_and_post_process_time);
   WriteThread::Writer w(write_options, my_batch, callback, log_ref,
                         disable_memtable, sub_batch_cnt, pre_release_callback);
@@ -688,6 +687,8 @@ Status DBImpl::WriteImplWALOnly(
   assert(w.state == WriteThread::STATE_GROUP_LEADER);
 
   if (publish_last_seq == kDoPublishLastSeq) {
+    Status status;
+
     // Currently we only use kDoPublishLastSeq in unordered_write
     assert(immutable_db_options_.unordered_write);
     WriteContext write_context;
@@ -764,6 +765,7 @@ Status DBImpl::WriteImplWALOnly(
     }
     seq_inc = total_batch_cnt;
   }
+  Status status;
   IOStatus io_s;
   if (!write_options.disableWAL) {
     io_s = ConcurrentWriteToWAL(write_group, log_used, &last_sequence, seq_inc);
@@ -850,8 +852,7 @@ void DBImpl::WriteStatusCheckOnLocked(const Status& status) {
   if (immutable_db_options_.paranoid_checks && !status.ok() &&
       !status.IsBusy() && !status.IsIncomplete()) {
     // Maybe change the return status to void?
-    error_handler_.SetBGError(status, BackgroundErrorReason::kWriteCallback)
-        .PermitUncheckedError();
+    error_handler_.SetBGError(status, BackgroundErrorReason::kWriteCallback);
   }
 }
 
@@ -863,8 +864,7 @@ void DBImpl::WriteStatusCheck(const Status& status) {
       !status.IsBusy() && !status.IsIncomplete()) {
     mutex_.Lock();
     // Maybe change the return status to void?
-    error_handler_.SetBGError(status, BackgroundErrorReason::kWriteCallback)
-        .PermitUncheckedError();
+    error_handler_.SetBGError(status, BackgroundErrorReason::kWriteCallback);
     mutex_.Unlock();
   }
 }
@@ -877,8 +877,7 @@ void DBImpl::IOStatusCheck(const IOStatus& io_status) {
       io_status.IsIOFenced()) {
     mutex_.Lock();
     // Maybe change the return status to void?
-    error_handler_.SetBGError(io_status, BackgroundErrorReason::kWriteCallback)
-        .PermitUncheckedError();
+    error_handler_.SetBGError(io_status, BackgroundErrorReason::kWriteCallback);
     mutex_.Unlock();
   }
 }
@@ -1335,10 +1334,17 @@ Status DBImpl::SwitchWAL(WriteContext* write_context) {
     }
     for (auto cfd : cfds) {
       cfd->imm()->FlushRequested();
+      if (!immutable_db_options_.atomic_flush) {
+        FlushRequest flush_req;
+        GenerateFlushRequest({cfd}, &flush_req);
+        SchedulePendingFlush(flush_req, FlushReason::kWriteBufferManager);
+      }
+    }
+    if (immutable_db_options_.atomic_flush) {
+      FlushRequest flush_req;
+      GenerateFlushRequest(cfds, &flush_req);
+      SchedulePendingFlush(flush_req, FlushReason::kWriteBufferManager);
     }
-    FlushRequest flush_req;
-    GenerateFlushRequest(cfds, &flush_req);
-    SchedulePendingFlush(flush_req, FlushReason::kWriteBufferManager);
     MaybeScheduleFlushOrCompaction();
   }
   return status;
@@ -1414,10 +1420,17 @@ Status DBImpl::HandleWriteBufferFull(WriteContext* write_context) {
     }
     for (const auto cfd : cfds) {
       cfd->imm()->FlushRequested();
+      if (!immutable_db_options_.atomic_flush) {
+        FlushRequest flush_req;
+        GenerateFlushRequest({cfd}, &flush_req);
+        SchedulePendingFlush(flush_req, FlushReason::kWriteBufferFull);
+      }
+    }
+    if (immutable_db_options_.atomic_flush) {
+      FlushRequest flush_req;
+      GenerateFlushRequest(cfds, &flush_req);
+      SchedulePendingFlush(flush_req, FlushReason::kWriteBufferFull);
     }
-    FlushRequest flush_req;
-    GenerateFlushRequest(cfds, &flush_req);
-    SchedulePendingFlush(flush_req, FlushReason::kWriteBufferFull);
     MaybeScheduleFlushOrCompaction();
   }
   return status;
@@ -1641,10 +1654,16 @@ Status DBImpl::ScheduleFlushes(WriteContext* context) {
   if (status.ok()) {
     if (immutable_db_options_.atomic_flush) {
       AssignAtomicFlushSeq(cfds);
+      FlushRequest flush_req;
+      GenerateFlushRequest(cfds, &flush_req);
+      SchedulePendingFlush(flush_req, FlushReason::kWriteBufferFull);
+    } else {
+      for (auto* cfd : cfds) {
+        FlushRequest flush_req;
+        GenerateFlushRequest({cfd}, &flush_req);
+        SchedulePendingFlush(flush_req, FlushReason::kWriteBufferFull);
+      }
     }
-    FlushRequest flush_req;
-    GenerateFlushRequest(cfds, &flush_req);
-    SchedulePendingFlush(flush_req, FlushReason::kWriteBufferFull);
     MaybeScheduleFlushOrCompaction();
   }
   return status;
@@ -1790,32 +1809,74 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
     }
     // We may have lost data from the WritableFileBuffer in-memory buffer for
     // the current log, so treat it as a fatal error and set bg_error
-    // Should handle return error?
     if (!io_s.ok()) {
-      // Should handle return error?
-      error_handler_.SetBGError(io_s, BackgroundErrorReason::kMemTable)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(io_s, BackgroundErrorReason::kMemTable);
     } else {
-      // Should handle return error?
-      error_handler_.SetBGError(s, BackgroundErrorReason::kMemTable)
-          .PermitUncheckedError();
+      error_handler_.SetBGError(s, BackgroundErrorReason::kMemTable);
     }
     // Read back bg_error in order to get the right severity
     s = error_handler_.GetBGError();
     return s;
   }
 
-  for (auto loop_cfd : *versions_->GetColumnFamilySet()) {
-    // all this is just optimization to delete logs that
-    // are no longer needed -- if CF is empty, that means it
-    // doesn't need that particular log to stay alive, so we just
-    // advance the log number. no need to persist this in the manifest
-    if (loop_cfd->mem()->GetFirstSequenceNumber() == 0 &&
-        loop_cfd->imm()->NumNotFlushed() == 0) {
-      if (creating_new_log) {
-        loop_cfd->SetLogNumber(logfile_number_);
+  bool empty_cf_updated = false;
+  if (immutable_db_options_.track_and_verify_wals_in_manifest &&
+      !immutable_db_options_.allow_2pc && creating_new_log) {
+    // In non-2pc mode, WALs become obsolete if they do not contain unflushed
+    // data. Updating the empty CF's log number might cause some WALs to become
+    // obsolete. So we should track the WAL obsoletion event before actually
+    // updating the empty CF's log number.
+    uint64_t min_wal_number_to_keep =
+        versions_->PreComputeMinLogNumberWithUnflushedData(logfile_number_);
+    if (min_wal_number_to_keep >
+        versions_->GetWalSet().GetMinWalNumberToKeep()) {
+      // Get a snapshot of the empty column families.
+      // LogAndApply may release and reacquire db
+      // mutex, during that period, column family may become empty (e.g. its
+      // flush succeeds), then it affects the computed min_log_number_to_keep,
+      // so we take a snapshot for consistency of column family data
+      // status. If a column family becomes non-empty afterwards, its active log
+      // should still be the created new log, so the min_log_number_to_keep is
+      // not affected.
+      autovector<ColumnFamilyData*> empty_cfs;
+      for (auto cf : *versions_->GetColumnFamilySet()) {
+        if (cf->IsEmpty()) {
+          empty_cfs.push_back(cf);
+        }
+      }
+
+      VersionEdit wal_deletion;
+      wal_deletion.DeleteWalsBefore(min_wal_number_to_keep);
+      s = versions_->LogAndApplyToDefaultColumnFamily(&wal_deletion, &mutex_);
+      if (!s.ok() && versions_->io_status().IsIOError()) {
+        s = error_handler_.SetBGError(versions_->io_status(),
+                                      BackgroundErrorReason::kManifestWrite);
+      }
+      if (!s.ok()) {
+        return s;
+      }
+
+      for (auto cf : empty_cfs) {
+        if (cf->IsEmpty()) {
+          cf->SetLogNumber(logfile_number_);
+          cf->mem()->SetCreationSeq(versions_->LastSequence());
+        }  // cf may become non-empty.
+      }
+      empty_cf_updated = true;
+    }
+  }
+  if (!empty_cf_updated) {
+    for (auto cf : *versions_->GetColumnFamilySet()) {
+      // all this is just optimization to delete logs that
+      // are no longer needed -- if CF is empty, that means it
+      // doesn't need that particular log to stay alive, so we just
+      // advance the log number. no need to persist this in the manifest
+      if (cf->IsEmpty()) {
+        if (creating_new_log) {
+          cf->SetLogNumber(logfile_number_);
+        }
+        cf->mem()->SetCreationSeq(versions_->LastSequence());
       }
-      loop_cfd->mem()->SetCreationSeq(versions_->LastSequence());
     }
   }
 

db/db_impl/db_secondary_test.cc

@@ -748,10 +748,12 @@ TEST_F(DBSecondaryTest, SwitchWALMultiColumnFamilies) {
     }
   };
   for (int k = 0; k != 8; ++k) {
-    ASSERT_OK(
-        Put(0 /*cf*/, "key" + std::to_string(k), "value" + std::to_string(k)));
-    ASSERT_OK(
-        Put(1 /*cf*/, "key" + std::to_string(k), "value" + std::to_string(k)));
+    for (int j = 0; j < 2; ++j) {
+      ASSERT_OK(Put(0 /*cf*/, "key" + std::to_string(k),
+                    "value" + std::to_string(k)));
+      ASSERT_OK(Put(1 /*cf*/, "key" + std::to_string(k),
+                    "value" + std::to_string(k)));
+    }
     TEST_SYNC_POINT(
         "DBSecondaryTest::SwitchWALMultipleColumnFamilies:BeforeCatchUp");
     ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());

db/db_iter.cc

@@ -36,16 +36,18 @@ namespace ROCKSDB_NAMESPACE {
 DBIter::DBIter(Env* _env, const ReadOptions& read_options,
                const ImmutableCFOptions& cf_options,
                const MutableCFOptions& mutable_cf_options,
-               const Comparator* cmp, InternalIterator* iter, SequenceNumber s,
-               bool arena_mode, uint64_t max_sequential_skip_in_iterations,
+               const Comparator* cmp, InternalIterator* iter,
+               const Version* version, SequenceNumber s, bool arena_mode,
+               uint64_t max_sequential_skip_in_iterations,
                ReadCallback* read_callback, DBImpl* db_impl,
-               ColumnFamilyData* cfd, bool allow_blob)
+               ColumnFamilyData* cfd, bool expose_blob_index)
     : prefix_extractor_(mutable_cf_options.prefix_extractor.get()),
       env_(_env),
       logger_(cf_options.info_log),
       user_comparator_(cmp),
       merge_operator_(cf_options.merge_operator),
       iter_(iter),
+      version_(version),
       read_callback_(read_callback),
       sequence_(s),
       statistics_(cf_options.statistics),
@@ -65,7 +67,9 @@ DBIter::DBIter(Env* _env, const ReadOptions& read_options,
       expect_total_order_inner_iter_(prefix_extractor_ == nullptr ||
                                      read_options.total_order_seek ||
                                      read_options.auto_prefix_mode),
-      allow_blob_(allow_blob),
+      read_tier_(read_options.read_tier),
+      verify_checksums_(read_options.verify_checksums),
+      expose_blob_index_(expose_blob_index),
       is_blob_(false),
       arena_mode_(arena_mode),
       range_del_agg_(&cf_options.internal_comparator, s),
@@ -165,6 +169,40 @@ void DBIter::Next() {
   }
 }
 
+bool DBIter::SetBlobValueIfNeeded(const Slice& user_key,
+                                  const Slice& blob_index) {
+  assert(!is_blob_);
+
+  if (expose_blob_index_) {  // Stacked BlobDB implementation
+    is_blob_ = true;
+    return true;
+  }
+
+  if (!version_) {
+    status_ = Status::Corruption("Encountered unexpected blob index.");
+    valid_ = false;
+    return false;
+  }
+
+  // TODO: consider moving ReadOptions from ArenaWrappedDBIter to DBIter to
+  // avoid having to copy options back and forth.
+  ReadOptions read_options;
+  read_options.read_tier = read_tier_;
+  read_options.verify_checksums = verify_checksums_;
+
+  const Status s =
+      version_->GetBlob(read_options, user_key, blob_index, &blob_value_);
+
+  if (!s.ok()) {
+    status_ = s;
+    valid_ = false;
+    return false;
+  }
+
+  is_blob_ = true;
+  return true;
+}
+
 // PRE: saved_key_ has the current user key if skipping_saved_key
 // POST: saved_key_ should have the next user key if valid_,
 //       if the current entry is a result of merge
@@ -221,25 +259,28 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
       is_key_seqnum_zero_ = false;
       return false;
     }
+    Slice user_key_without_ts =
+        StripTimestampFromUserKey(ikey_.user_key, timestamp_size_);
 
     is_key_seqnum_zero_ = (ikey_.sequence == 0);
 
     assert(iterate_upper_bound_ == nullptr ||
            iter_.UpperBoundCheckResult() != IterBoundCheck::kInbound ||
            user_comparator_.CompareWithoutTimestamp(
-               ikey_.user_key, /*a_has_ts=*/true, *iterate_upper_bound_,
+               user_key_without_ts, /*a_has_ts=*/false, *iterate_upper_bound_,
                /*b_has_ts=*/false) < 0);
     if (iterate_upper_bound_ != nullptr &&
         iter_.UpperBoundCheckResult() != IterBoundCheck::kInbound &&
         user_comparator_.CompareWithoutTimestamp(
-            ikey_.user_key, /*a_has_ts=*/true, *iterate_upper_bound_,
+            user_key_without_ts, /*a_has_ts=*/false, *iterate_upper_bound_,
             /*b_has_ts=*/false) >= 0) {
       break;
     }
 
     assert(prefix == nullptr || prefix_extractor_ != nullptr);
     if (prefix != nullptr &&
-        prefix_extractor_->Transform(ikey_.user_key).compare(*prefix) != 0) {
+        prefix_extractor_->Transform(user_key_without_ts).compare(*prefix) !=
+            0) {
       assert(prefix_same_as_start_);
       break;
     }
@@ -316,8 +357,14 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
           case kTypeBlobIndex:
             if (start_seqnum_ > 0) {
               if (ikey_.sequence >= start_seqnum_) {
-                assert(ikey_.type != kTypeBlobIndex);
                 saved_key_.SetInternalKey(ikey_);
+
+                if (ikey_.type == kTypeBlobIndex) {
+                  if (!SetBlobValueIfNeeded(ikey_.user_key, iter_.value())) {
+                    return false;
+                  }
+                }
+
                 valid_ = true;
                 return true;
               } else {
@@ -331,6 +378,13 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
               }
             } else if (timestamp_lb_) {
               saved_key_.SetInternalKey(ikey_);
+
+              if (ikey_.type == kTypeBlobIndex) {
+                if (!SetBlobValueIfNeeded(ikey_.user_key, iter_.value())) {
+                  return false;
+                }
+              }
+
               valid_ = true;
               return true;
             } else {
@@ -345,20 +399,13 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
                 num_skipped = 0;
                 reseek_done = false;
                 PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
-              } else if (ikey_.type == kTypeBlobIndex) {
-                if (!allow_blob_) {
-                  ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
-                  status_ = Status::NotSupported(
-                      "Encounter unexpected blob index. Please open DB with "
-                      "ROCKSDB_NAMESPACE::blob_db::BlobDB instead.");
-                  valid_ = false;
-                  return false;
+              } else {
+                if (ikey_.type == kTypeBlobIndex) {
+                  if (!SetBlobValueIfNeeded(ikey_.user_key, iter_.value())) {
+                    return false;
+                  }
                 }
 
-                is_blob_ = true;
-                valid_ = true;
-                return true;
-              } else {
                 valid_ = true;
                 return true;
               }
@@ -498,7 +545,6 @@ bool DBIter::MergeValuesNewToOld() {
   TEST_SYNC_POINT("DBIter::MergeValuesNewToOld:PushedFirstOperand");
 
   ParsedInternalKey ikey;
-  Status s;
   for (iter_.Next(); iter_.Valid(); iter_.Next()) {
     TEST_SYNC_POINT("DBIter::MergeValuesNewToOld:SteppedToNextOperand");
     if (!ParseKey(&ikey)) {
@@ -526,7 +572,7 @@ bool DBIter::MergeValuesNewToOld() {
       // hit a put, merge the put value with operands and store the
       // final result in saved_value_. We are done!
       const Slice val = iter_.value();
-      s = MergeHelper::TimedFullMerge(
+      Status s = MergeHelper::TimedFullMerge(
           merge_operator_, ikey.user_key, &val, merge_context_.GetOperands(),
           &saved_value_, logger_, statistics_, env_, &pinned_value_, true);
       if (!s.ok()) {
@@ -548,15 +594,7 @@ bool DBIter::MergeValuesNewToOld() {
           iter_.value(), iter_.iter()->IsValuePinned() /* operand_pinned */);
       PERF_COUNTER_ADD(internal_merge_count, 1);
     } else if (kTypeBlobIndex == ikey.type) {
-      if (!allow_blob_) {
-        ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
-        status_ = Status::NotSupported(
-            "Encounter unexpected blob index. Please open DB with "
-            "ROCKSDB_NAMESPACE::blob_db::BlobDB instead.");
-      } else {
-        status_ =
-            Status::NotSupported("Blob DB does not support merge operator.");
-      }
+      status_ = Status::NotSupported("BlobDB does not support merge operator.");
       valid_ = false;
       return false;
     } else {
@@ -577,10 +615,10 @@ bool DBIter::MergeValuesNewToOld() {
   // a deletion marker.
   // feed null as the existing value to the merge operator, such that
   // client can differentiate this scenario and do things accordingly.
-  s = MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetUserKey(),
-                                  nullptr, merge_context_.GetOperands(),
-                                  &saved_value_, logger_, statistics_, env_,
-                                  &pinned_value_, true);
+  Status s = MergeHelper::TimedFullMerge(
+      merge_operator_, saved_key_.GetUserKey(), nullptr,
+      merge_context_.GetOperands(), &saved_value_, logger_, statistics_, env_,
+      &pinned_value_, true);
   if (!s.ok()) {
     valid_ = false;
     status_ = s;
@@ -704,7 +742,9 @@ void DBIter::PrevInternal(const Slice* prefix) {
 
     assert(prefix == nullptr || prefix_extractor_ != nullptr);
     if (prefix != nullptr &&
-        prefix_extractor_->Transform(saved_key_.GetUserKey())
+        prefix_extractor_
+                ->Transform(StripTimestampFromUserKey(saved_key_.GetUserKey(),
+                                                      timestamp_size_))
                 .compare(*prefix) != 0) {
       assert(prefix_same_as_start_);
       // Current key does not have the same prefix as start
@@ -883,15 +923,8 @@ bool DBIter::FindValueForCurrentKey() {
             merge_context_.GetOperands(), &saved_value_, logger_, statistics_,
             env_, &pinned_value_, true);
       } else if (last_not_merge_type == kTypeBlobIndex) {
-        if (!allow_blob_) {
-          ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
-          status_ = Status::NotSupported(
-              "Encounter unexpected blob index. Please open DB with "
-              "ROCKSDB_NAMESPACE::blob_db::BlobDB instead.");
-        } else {
-          status_ =
-              Status::NotSupported("Blob DB does not support merge operator.");
-        }
+        status_ =
+            Status::NotSupported("BlobDB does not support merge operator.");
         valid_ = false;
         return false;
       } else {
@@ -906,15 +939,10 @@ bool DBIter::FindValueForCurrentKey() {
       // do nothing - we've already has value in pinned_value_
       break;
     case kTypeBlobIndex:
-      if (!allow_blob_) {
-        ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
-        status_ = Status::NotSupported(
-            "Encounter unexpected blob index. Please open DB with "
-            "ROCKSDB_NAMESPACE::blob_db::BlobDB instead.");
-        valid_ = false;
+      if (!SetBlobValueIfNeeded(saved_key_.GetUserKey(), pinned_value_)) {
         return false;
       }
-      is_blob_ = true;
+
       break;
     default:
       valid_ = false;
@@ -987,14 +1015,6 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
     valid_ = false;
     return true;
   }
-  if (ikey.type == kTypeBlobIndex && !allow_blob_) {
-    ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
-    status_ = Status::NotSupported(
-        "Encounter unexpected blob index. Please open DB with "
-        "ROCKSDB_NAMESPACE::blob_db::BlobDB instead.");
-    valid_ = false;
-    return false;
-  }
   if (!iter_.PrepareValue()) {
     valid_ = false;
     return false;
@@ -1002,7 +1022,12 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
   if (ikey.type == kTypeValue || ikey.type == kTypeBlobIndex) {
     assert(iter_.iter()->IsValuePinned());
     pinned_value_ = iter_.value();
-    is_blob_ = (ikey.type == kTypeBlobIndex);
+    if (ikey.type == kTypeBlobIndex) {
+      if (!SetBlobValueIfNeeded(ikey.user_key, pinned_value_)) {
+        return false;
+      }
+    }
+
     valid_ = true;
     return true;
   }
@@ -1058,15 +1083,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
           iter_.value(), iter_.iter()->IsValuePinned() /* operand_pinned */);
       PERF_COUNTER_ADD(internal_merge_count, 1);
     } else if (ikey.type == kTypeBlobIndex) {
-      if (!allow_blob_) {
-        ROCKS_LOG_ERROR(logger_, "Encounter unexpected blob index.");
-        status_ = Status::NotSupported(
-            "Encounter unexpected blob index. Please open DB with "
-            "ROCKSDB_NAMESPACE::blob_db::BlobDB instead.");
-      } else {
-        status_ =
-            Status::NotSupported("Blob DB does not support merge operator.");
-      }
+      status_ = Status::NotSupported("BlobDB does not support merge operator.");
       valid_ = false;
       return false;
     } else {
@@ -1399,7 +1416,8 @@ void DBIter::SeekToFirst() {
   }
   if (valid_ && prefix_same_as_start_) {
     assert(prefix_extractor_ != nullptr);
-    prefix_.SetUserKey(prefix_extractor_->Transform(saved_key_.GetUserKey()));
+    prefix_.SetUserKey(prefix_extractor_->Transform(
+        StripTimestampFromUserKey(saved_key_.GetUserKey(), timestamp_size_)));
   }
 }
 
@@ -1450,7 +1468,8 @@ void DBIter::SeekToLast() {
   }
   if (valid_ && prefix_same_as_start_) {
     assert(prefix_extractor_ != nullptr);
-    prefix_.SetUserKey(prefix_extractor_->Transform(saved_key_.GetUserKey()));
+    prefix_.SetUserKey(prefix_extractor_->Transform(
+        StripTimestampFromUserKey(saved_key_.GetUserKey(), timestamp_size_)));
   }
 }
 
@@ -1458,15 +1477,16 @@ Iterator* NewDBIterator(Env* env, const ReadOptions& read_options,
                         const ImmutableCFOptions& cf_options,
                         const MutableCFOptions& mutable_cf_options,
                         const Comparator* user_key_comparator,
-                        InternalIterator* internal_iter,
+                        InternalIterator* internal_iter, const Version* version,
                         const SequenceNumber& sequence,
                         uint64_t max_sequential_skip_in_iterations,
                         ReadCallback* read_callback, DBImpl* db_impl,
-                        ColumnFamilyData* cfd, bool allow_blob) {
-  DBIter* db_iter = new DBIter(
-      env, read_options, cf_options, mutable_cf_options, user_key_comparator,
-      internal_iter, sequence, false, max_sequential_skip_in_iterations,
-      read_callback, db_impl, cfd, allow_blob);
+                        ColumnFamilyData* cfd, bool expose_blob_index) {
+  DBIter* db_iter =
+      new DBIter(env, read_options, cf_options, mutable_cf_options,
+                 user_key_comparator, internal_iter, version, sequence, false,
+                 max_sequential_skip_in_iterations, read_callback, db_impl, cfd,
+                 expose_blob_index);
   return db_iter;
 }
 

db/db_iter.h

@@ -22,6 +22,8 @@
 
 namespace ROCKSDB_NAMESPACE {
 
+class Version;
+
 // This file declares the factory functions of DBIter, in its original form
 // or a wrapped form with class ArenaWrappedDBIter, which is defined here.
 // Class DBIter, which is declared and implemented inside db_iter.cc, is
@@ -114,10 +116,10 @@ class DBIter final : public Iterator {
   DBIter(Env* _env, const ReadOptions& read_options,
          const ImmutableCFOptions& cf_options,
          const MutableCFOptions& mutable_cf_options, const Comparator* cmp,
-         InternalIterator* iter, SequenceNumber s, bool arena_mode,
-         uint64_t max_sequential_skip_in_iterations,
+         InternalIterator* iter, const Version* version, SequenceNumber s,
+         bool arena_mode, uint64_t max_sequential_skip_in_iterations,
          ReadCallback* read_callback, DBImpl* db_impl, ColumnFamilyData* cfd,
-         bool allow_blob);
+         bool expose_blob_index);
 
   // No copying allowed
   DBIter(const DBIter&) = delete;
@@ -159,7 +161,10 @@ class DBIter final : public Iterator {
   }
   Slice value() const override {
     assert(valid_);
-    if (current_entry_is_merged_) {
+
+    if (!expose_blob_index_ && is_blob_) {
+      return blob_value_;
+    } else if (current_entry_is_merged_) {
       // If pinned_value_ is set then the result of merge operator is one of
       // the merge operands and we should return it.
       return pinned_value_.data() ? pinned_value_ : saved_value_;
@@ -185,7 +190,7 @@ class DBIter final : public Iterator {
     return ExtractTimestampFromUserKey(ukey_and_ts, timestamp_size_);
   }
   bool IsBlob() const {
-    assert(valid_ && (allow_blob_ || !is_blob_));
+    assert(valid_);
     return is_blob_;
   }
 
@@ -287,12 +292,17 @@ class DBIter final : public Iterator {
                : user_comparator_.CompareWithoutTimestamp(a, b);
   }
 
+  // Retrieves the blob value for the specified user key using the given blob
+  // index when using the integrated BlobDB implementation.
+  bool SetBlobValueIfNeeded(const Slice& user_key, const Slice& blob_index);
+
   const SliceTransform* prefix_extractor_;
   Env* const env_;
   Logger* logger_;
   UserComparatorWrapper user_comparator_;
   const MergeOperator* const merge_operator_;
   IteratorWrapper iter_;
+  const Version* version_;
   ReadCallback* read_callback_;
   // Max visible sequence number. It is normally the snapshot seq unless we have
   // uncommitted data in db as in WriteUnCommitted.
@@ -306,6 +316,7 @@ class DBIter final : public Iterator {
   std::string saved_value_;
   Slice pinned_value_;
   // for prefix seek mode to support prev()
+  PinnableSlice blob_value_;
   Statistics* statistics_;
   uint64_t max_skip_;
   uint64_t max_skippable_internal_keys_;
@@ -335,7 +346,11 @@ class DBIter final : public Iterator {
   // Expect the inner iterator to maintain a total order.
   // prefix_extractor_ must be non-NULL if the value is false.
   const bool expect_total_order_inner_iter_;
-  bool allow_blob_;
+  ReadTier read_tier_;
+  bool verify_checksums_;
+  // Whether the iterator is allowed to expose blob references. Set to true when
+  // the stacked BlobDB implementation is used, false otherwise.
+  bool expose_blob_index_;
   bool is_blob_;
   bool arena_mode_;
   // List of operands for merge operator.
@@ -367,8 +382,9 @@ extern Iterator* NewDBIterator(
     const ImmutableCFOptions& cf_options,
     const MutableCFOptions& mutable_cf_options,
     const Comparator* user_key_comparator, InternalIterator* internal_iter,
-    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
-    ReadCallback* read_callback, DBImpl* db_impl = nullptr,
-    ColumnFamilyData* cfd = nullptr, bool allow_blob = false);
+    const Version* version, const SequenceNumber& sequence,
+    uint64_t max_sequential_skip_in_iterations, ReadCallback* read_callback,
+    DBImpl* db_impl = nullptr, ColumnFamilyData* cfd = nullptr,
+    bool expose_blob_index = false);
 
 }  // namespace ROCKSDB_NAMESPACE

db/db_iter_stress_test.cc

@@ -513,7 +513,7 @@ TEST_F(DBIteratorStressTest, StressTest) {
                   db_iter.reset(NewDBIterator(
                       env_, ropt, ImmutableCFOptions(options),
                       MutableCFOptions(options), BytewiseComparator(),
-                      internal_iter, sequence,
+                      internal_iter, nullptr /* version */, sequence,
                       options.max_sequential_skip_in_iterations,
                       nullptr /*read_callback*/));
                 }

db/db_iter_test.cc

@@ -253,8 +253,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
     ReadOptions ro;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
@@ -286,8 +287,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
     ReadOptions ro;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
@@ -313,8 +315,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
@@ -346,8 +349,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
@@ -382,8 +386,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(!db_iter->Valid());
@@ -412,8 +417,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 7, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 7 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     SetPerfLevel(kEnableCount);
     ASSERT_TRUE(GetPerfLevel() == kEnableCount);
@@ -450,8 +456,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 4, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 4 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
@@ -476,8 +483,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(!db_iter->Valid());
@@ -499,8 +507,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
@@ -535,8 +544,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 7, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 7 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     SetPerfLevel(kEnableCount);
     ASSERT_TRUE(GetPerfLevel() == kEnableCount);
@@ -565,8 +575,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
     ReadOptions ro;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToFirst();
     ASSERT_TRUE(db_iter->Valid());
@@ -608,8 +619,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
     ReadOptions ro;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 2, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 2 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -640,8 +652,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
     ReadOptions ro;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -671,8 +684,9 @@ TEST_F(DBIteratorTest, DBIteratorEmpty) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 0, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 0 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(!db_iter->Valid());
   }
@@ -683,8 +697,9 @@ TEST_F(DBIteratorTest, DBIteratorEmpty) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 0, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 0 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToFirst();
     ASSERT_TRUE(!db_iter->Valid());
   }
@@ -706,8 +721,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkipCountSkips) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 2,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      2 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
   db_iter->SeekToLast();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -751,8 +767,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
       options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
       std::unique_ptr<Iterator> db_iter(NewDBIterator(
           env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-          internal_iter, i + 2, options.max_sequential_skip_in_iterations,
-          nullptr /*read_callback*/));
+          internal_iter, nullptr /* version */, i + 2 /* sequence */,
+          options.max_sequential_skip_in_iterations,
+          nullptr /* read_callback */));
       db_iter->SeekToLast();
       ASSERT_TRUE(db_iter->Valid());
 
@@ -787,8 +804,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
 
       std::unique_ptr<Iterator> db_iter(NewDBIterator(
           env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-          internal_iter, i + 2, options.max_sequential_skip_in_iterations,
-          nullptr /*read_callback*/));
+          internal_iter, nullptr /* version */, i + 2 /* sequence */,
+          options.max_sequential_skip_in_iterations,
+          nullptr /* read_callback */));
       db_iter->SeekToLast();
       ASSERT_TRUE(db_iter->Valid());
 
@@ -816,8 +834,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
 
       std::unique_ptr<Iterator> db_iter(NewDBIterator(
           env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-          internal_iter, 202, options.max_sequential_skip_in_iterations,
-          nullptr /*read_callback*/));
+          internal_iter, nullptr /* version */, 202 /* sequence */,
+          options.max_sequential_skip_in_iterations,
+          nullptr /* read_callback */));
       db_iter->SeekToLast();
       ASSERT_TRUE(db_iter->Valid());
 
@@ -849,8 +868,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
       internal_iter->Finish();
       std::unique_ptr<Iterator> db_iter(NewDBIterator(
           env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-          internal_iter, i, options.max_sequential_skip_in_iterations,
-          nullptr /*read_callback*/));
+          internal_iter, nullptr /* version */, i /* sequence */,
+          options.max_sequential_skip_in_iterations,
+          nullptr /* read_callback */));
       db_iter->SeekToLast();
       ASSERT_TRUE(!db_iter->Valid());
 
@@ -866,8 +886,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
     internal_iter->Finish();
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 200, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 200 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -901,8 +922,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
 
       std::unique_ptr<Iterator> db_iter(NewDBIterator(
           env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-          internal_iter, i + 2, options.max_sequential_skip_in_iterations,
-          nullptr /*read_callback*/));
+          internal_iter, nullptr /* version */, i + 2 /* sequence */,
+          options.max_sequential_skip_in_iterations,
+          nullptr /* read_callback */));
       db_iter->SeekToLast();
       ASSERT_TRUE(db_iter->Valid());
 
@@ -936,8 +958,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
 
       std::unique_ptr<Iterator> db_iter(NewDBIterator(
           env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-          internal_iter, i + 2, options.max_sequential_skip_in_iterations,
-          nullptr /*read_callback*/));
+          internal_iter, nullptr /* version */, i + 2 /* sequence */,
+          options.max_sequential_skip_in_iterations,
+          nullptr /* read_callback */));
       db_iter->SeekToLast();
       ASSERT_TRUE(db_iter->Valid());
 
@@ -986,8 +1009,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
     ro.max_skippable_internal_keys = 0;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToFirst();
     ASSERT_TRUE(db_iter->Valid());
@@ -1033,8 +1057,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
     ro.max_skippable_internal_keys = 2;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToFirst();
     ASSERT_TRUE(db_iter->Valid());
@@ -1078,8 +1103,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
     ro.max_skippable_internal_keys = 2;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToFirst();
     ASSERT_TRUE(db_iter->Valid());
@@ -1117,8 +1143,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
     ro.max_skippable_internal_keys = 2;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToFirst();
     ASSERT_TRUE(db_iter->Valid());
@@ -1153,8 +1180,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
     ro.max_skippable_internal_keys = 2;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
@@ -1184,8 +1212,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
     ro.max_skippable_internal_keys = 2;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToFirst();
     ASSERT_TRUE(db_iter->Valid());
@@ -1222,8 +1251,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
     ro.max_skippable_internal_keys = 2;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToFirst();
     ASSERT_TRUE(db_iter->Valid());
@@ -1260,8 +1290,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
       ro.max_skippable_internal_keys = i;
       std::unique_ptr<Iterator> db_iter(NewDBIterator(
           env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-          internal_iter, 2 * i + 1, options.max_sequential_skip_in_iterations,
-          nullptr /*read_callback*/));
+          internal_iter, nullptr /* version */, 2 * i + 1 /* sequence */,
+          options.max_sequential_skip_in_iterations,
+          nullptr /* read_callback */));
 
       db_iter->SeekToFirst();
       ASSERT_TRUE(db_iter->Valid());
@@ -1314,8 +1345,9 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
       ro.max_skippable_internal_keys = i;
       std::unique_ptr<Iterator> db_iter(NewDBIterator(
           env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-          internal_iter, 2 * i + 1, options.max_sequential_skip_in_iterations,
-          nullptr /*read_callback*/));
+          internal_iter, nullptr /* version */, 2 * i + 1 /* sequence */,
+          options.max_sequential_skip_in_iterations,
+          nullptr /* read_callback */));
 
       db_iter->SeekToFirst();
       ASSERT_TRUE(db_iter->Valid());
@@ -1353,8 +1385,9 @@ TEST_F(DBIteratorTest, DBIterator1) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 1,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      1 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
   db_iter->SeekToFirst();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1381,8 +1414,9 @@ TEST_F(DBIteratorTest, DBIterator2) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 0,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      0 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
   db_iter->SeekToFirst();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1406,8 +1440,9 @@ TEST_F(DBIteratorTest, DBIterator3) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 2,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      2 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
   db_iter->SeekToFirst();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1431,8 +1466,9 @@ TEST_F(DBIteratorTest, DBIterator4) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 4,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      4 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
   db_iter->SeekToFirst();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1465,8 +1501,9 @@ TEST_F(DBIteratorTest, DBIterator5) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 0, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 0 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1488,8 +1525,9 @@ TEST_F(DBIteratorTest, DBIterator5) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 1, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 1 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1511,8 +1549,9 @@ TEST_F(DBIteratorTest, DBIterator5) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 2, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 2 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1534,8 +1573,9 @@ TEST_F(DBIteratorTest, DBIterator5) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 3, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 3 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1557,8 +1597,9 @@ TEST_F(DBIteratorTest, DBIterator5) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 4, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 4 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1580,8 +1621,9 @@ TEST_F(DBIteratorTest, DBIterator5) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 5, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 5 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1603,8 +1645,9 @@ TEST_F(DBIteratorTest, DBIterator5) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 6, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 6 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1624,8 +1667,9 @@ TEST_F(DBIteratorTest, DBIterator5) {
     internal_iter->Finish();
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 10, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 10 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->Seek("b");
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -1655,8 +1699,9 @@ TEST_F(DBIteratorTest, DBIterator6) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 0, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 0 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1678,8 +1723,9 @@ TEST_F(DBIteratorTest, DBIterator6) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 1, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 1 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1701,8 +1747,9 @@ TEST_F(DBIteratorTest, DBIterator6) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 2, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 2 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1724,8 +1771,9 @@ TEST_F(DBIteratorTest, DBIterator6) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 3, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 3 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(!db_iter->Valid());
   }
@@ -1743,8 +1791,9 @@ TEST_F(DBIteratorTest, DBIterator6) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 4, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 4 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1766,8 +1815,9 @@ TEST_F(DBIteratorTest, DBIterator6) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 5, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 5 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1789,8 +1839,9 @@ TEST_F(DBIteratorTest, DBIterator6) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 6, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 6 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1832,8 +1883,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 0, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 0 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
     ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1867,8 +1919,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 2, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 2 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
 
@@ -1908,8 +1961,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 4, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 4 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
 
@@ -1949,8 +2003,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 5, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 5 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
 
@@ -1995,8 +2050,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 6, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 6 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
 
@@ -2042,8 +2098,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 7, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 7 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
 
@@ -2083,8 +2140,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 9, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 9 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
 
@@ -2130,8 +2188,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 13, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 13 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
 
@@ -2178,8 +2237,9 @@ TEST_F(DBIteratorTest, DBIterator7) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, cf_options, mutable_cf_options, BytewiseComparator(),
-        internal_iter, 14, options.max_sequential_skip_in_iterations,
-        nullptr /*read_callback*/));
+        internal_iter, nullptr /* version */, 14 /* sequence */,
+        options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
 
@@ -2209,8 +2269,9 @@ TEST_F(DBIteratorTest, DBIterator8) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 10,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      10 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
   db_iter->SeekToLast();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -2240,8 +2301,9 @@ TEST_F(DBIteratorTest, DBIterator9) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-        BytewiseComparator(), internal_iter, 10,
-        options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+        BytewiseComparator(), internal_iter, nullptr /* version */,
+        10 /* sequence */, options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
 
     db_iter->SeekToLast();
     ASSERT_TRUE(db_iter->Valid());
@@ -2307,8 +2369,9 @@ TEST_F(DBIteratorTest, DBIterator10) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 10,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      10 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
 
   db_iter->Seek("c");
   ASSERT_TRUE(db_iter->Valid());
@@ -2347,8 +2410,8 @@ TEST_F(DBIteratorTest, SeekToLastOccurrenceSeq0) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 10, 0 /* force seek */,
-      nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      10 /* sequence */, 0 /* force seek */, nullptr /* read_callback */));
   db_iter->SeekToFirst();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -2376,8 +2439,9 @@ TEST_F(DBIteratorTest, DBIterator11) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 1,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      1 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
   db_iter->SeekToFirst();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -2403,7 +2467,8 @@ TEST_F(DBIteratorTest, DBIterator12) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 10, 0, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      10 /* sequence */, 0 /* force seek */, nullptr /* read_callback */));
   db_iter->SeekToLast();
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -2440,7 +2505,9 @@ TEST_F(DBIteratorTest, DBIterator13) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 2, 3, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      2 /* sequence */, 3 /* max_sequential_skip_in_iterations */,
+      nullptr /* read_callback */));
   db_iter->Seek("b");
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), key);
@@ -2468,7 +2535,9 @@ TEST_F(DBIteratorTest, DBIterator14) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 4, 1, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      4 /* sequence */, 1 /* max_sequential_skip_in_iterations */,
+      nullptr /* read_callback */));
   db_iter->Seek("b");
   ASSERT_TRUE(db_iter->Valid());
   ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -2495,8 +2564,9 @@ TEST_F(DBIteratorTest, DBIteratorTestDifferentialSnapshots) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-        BytewiseComparator(), internal_iter, 13,
-        options.max_sequential_skip_in_iterations, nullptr));
+        BytewiseComparator(), internal_iter, nullptr /* version */,
+        13 /* sequence */, options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     // Expecting InternalKeys in [5,8] range with correct type
     int seqnums[4] = {5,8,11,13};
     std::string user_keys[4] = {"1","2","3","4"};
@@ -2530,8 +2600,9 @@ TEST_F(DBIteratorTest, DBIteratorTestDifferentialSnapshots) {
 
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-        BytewiseComparator(), internal_iter, 13,
-        options.max_sequential_skip_in_iterations, nullptr));
+        BytewiseComparator(), internal_iter, nullptr /* version */,
+        13 /* sequence */, options.max_sequential_skip_in_iterations,
+        nullptr /* read_callback */));
     // Expecting InternalKeys in [5,8] range with correct type
     int seqnums[4] = {5,8,11,13};
     EntryType key_types[4] = {EntryType::kEntryDelete,EntryType::kEntryDelete,
@@ -2580,9 +2651,9 @@ class DBIterWithMergeIterTest : public testing::Test {
 
     db_iter_.reset(NewDBIterator(
         env_, ro_, ImmutableCFOptions(options_), MutableCFOptions(options_),
-        BytewiseComparator(), merge_iter,
+        BytewiseComparator(), merge_iter, nullptr /* version */,
         8 /* read data earlier than seqId 8 */,
-        3 /* max iterators before reseek */, nullptr /*read_callback*/));
+        3 /* max iterators before reseek */, nullptr /* read_callback */));
   }
 
   Env* env_;
@@ -3020,8 +3091,9 @@ TEST_F(DBIteratorTest, SeekPrefixTombstones) {
   ro.prefix_same_as_start = true;
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 10,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      10 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
 
   int skipped_keys = 0;
 
@@ -3056,8 +3128,8 @@ TEST_F(DBIteratorTest, SeekToFirstLowerBound) {
     Options options;
     std::unique_ptr<Iterator> db_iter(NewDBIterator(
         env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-        BytewiseComparator(), internal_iter, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
+        BytewiseComparator(), internal_iter, nullptr /* version */,
+        10 /* sequence */, options.max_sequential_skip_in_iterations,
         nullptr /* read_callback */));
 
     db_iter->SeekToFirst();
@@ -3095,8 +3167,9 @@ TEST_F(DBIteratorTest, PrevLowerBound) {
   Options options;
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 10 /* sequence */,
-      options.max_sequential_skip_in_iterations, nullptr /* read_callback */));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      10 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
 
   db_iter->SeekToLast();
   for (int i = kNumKeys; i >= kLowerBound; --i) {
@@ -3123,8 +3196,9 @@ TEST_F(DBIteratorTest, SeekLessLowerBound) {
   Options options;
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ro, ImmutableCFOptions(options), MutableCFOptions(options),
-      BytewiseComparator(), internal_iter, 10 /* sequence */,
-      options.max_sequential_skip_in_iterations, nullptr /* read_callback */));
+      BytewiseComparator(), internal_iter, nullptr /* version */,
+      10 /* sequence */, options.max_sequential_skip_in_iterations,
+      nullptr /* read_callback */));
 
   auto before_lower_bound_str = std::to_string(kLowerBound - 1);
   Slice before_lower_bound(lower_bound_str);
@@ -3148,8 +3222,9 @@ TEST_F(DBIteratorTest, ReverseToForwardWithDisappearingKeys) {
 
   std::unique_ptr<Iterator> db_iter(NewDBIterator(
       env_, ReadOptions(), ImmutableCFOptions(options),
-      MutableCFOptions(options), BytewiseComparator(), internal_iter, 10,
-      options.max_sequential_skip_in_iterations, nullptr /*read_callback*/));
+      MutableCFOptions(options), BytewiseComparator(), internal_iter,
+      nullptr /* version */, 10 /* sequence */,
+      options.max_sequential_skip_in_iterations, nullptr /* read_callback */));
 
   db_iter->SeekForPrev("a");
   ASSERT_TRUE(db_iter->Valid());

db/db_iterator_test.cc

@@ -67,8 +67,8 @@ TEST_P(DBIteratorTest, IteratorProperty) {
   // The test needs to be changed if kPersistedTier is supported in iterator.
   Options options = CurrentOptions();
   CreateAndReopenWithCF({"pikachu"}, options);
-  Put(1, "1", "2");
-  Delete(1, "2");
+  ASSERT_OK(Put(1, "1", "2"));
+  ASSERT_OK(Delete(1, "2"));
   ReadOptions ropt;
   ropt.pin_data = false;
   {
@@ -172,10 +172,10 @@ TEST_P(DBIteratorTest, NonBlockingIteration) {
 TEST_P(DBIteratorTest, IterSeekBeforePrev) {
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("0", "f"));
   ASSERT_OK(Put("1", "h"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("2", "j"));
   auto iter = NewIterator(ReadOptions());
   iter->Seek(Slice("c"));
@@ -199,7 +199,7 @@ TEST_P(DBIteratorTest, IterReseekNewUpperBound) {
   ASSERT_OK(Put("aabb", rnd.RandomString(400)));
   ASSERT_OK(Put("aaef", rnd.RandomString(400)));
   ASSERT_OK(Put("b", rnd.RandomString(400)));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ReadOptions opts;
   Slice ub = Slice("aa");
   opts.iterate_upper_bound = &ub;
@@ -215,10 +215,10 @@ TEST_P(DBIteratorTest, IterReseekNewUpperBound) {
 TEST_P(DBIteratorTest, IterSeekForPrevBeforeNext) {
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("0", "f"));
   ASSERT_OK(Put("1", "h"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("2", "j"));
   auto iter = NewIterator(ReadOptions());
   iter->SeekForPrev(Slice("0"));
@@ -238,7 +238,7 @@ TEST_P(DBIteratorTest, IterLongKeys) {
   ASSERT_OK(Put(MakeLongKey(20, 0), "0"));
   ASSERT_OK(Put(MakeLongKey(32, 2), "2"));
   ASSERT_OK(Put("a", "b"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put(MakeLongKey(50, 1), "1"));
   ASSERT_OK(Put(MakeLongKey(127, 3), "3"));
   ASSERT_OK(Put(MakeLongKey(64, 4), "4"));
@@ -276,7 +276,7 @@ TEST_P(DBIteratorTest, IterLongKeys) {
 
 TEST_P(DBIteratorTest, IterNextWithNewerSeq) {
   ASSERT_OK(Put("0", "0"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
   ASSERT_OK(Put("d", "e"));
@@ -302,7 +302,7 @@ TEST_P(DBIteratorTest, IterNextWithNewerSeq) {
 
 TEST_P(DBIteratorTest, IterPrevWithNewerSeq) {
   ASSERT_OK(Put("0", "0"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
   ASSERT_OK(Put("d", "e"));
@@ -333,7 +333,7 @@ TEST_P(DBIteratorTest, IterPrevWithNewerSeq) {
 
 TEST_P(DBIteratorTest, IterPrevWithNewerSeq2) {
   ASSERT_OK(Put("0", "0"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
   ASSERT_OK(Put("e", "f"));
@@ -377,6 +377,8 @@ TEST_P(DBIteratorTest, IterEmpty) {
     iter->SeekForPrev("foo");
     ASSERT_EQ(IterStatus(iter), "(invalid)");
 
+    ASSERT_OK(iter->status());
+
     delete iter;
   } while (ChangeCompactOptions());
 }
@@ -783,18 +785,18 @@ TEST_P(DBIteratorTest, IterWithSnapshot) {
 TEST_P(DBIteratorTest, IteratorPinsRef) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
-    Put(1, "foo", "hello");
+    ASSERT_OK(Put(1, "foo", "hello"));
 
     // Get iterator that will yield the current contents of the DB.
     Iterator* iter = NewIterator(ReadOptions(), handles_[1]);
 
     // Write to force compactions
-    Put(1, "foo", "newvalue1");
+    ASSERT_OK(Put(1, "foo", "newvalue1"));
     for (int i = 0; i < 100; i++) {
       // 100K values
       ASSERT_OK(Put(1, Key(i), Key(i) + std::string(100000, 'v')));
     }
-    Put(1, "foo", "newvalue2");
+    ASSERT_OK(Put(1, "foo", "newvalue2"));
 
     iter->SeekToFirst();
     ASSERT_TRUE(iter->Valid());
@@ -809,8 +811,8 @@ TEST_P(DBIteratorTest, IteratorPinsRef) {
 TEST_P(DBIteratorTest, IteratorDeleteAfterCfDelete) {
   CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
 
-  Put(1, "foo", "delete-cf-then-delete-iter");
-  Put(1, "hello", "value2");
+  ASSERT_OK(Put(1, "foo", "delete-cf-then-delete-iter"));
+  ASSERT_OK(Put(1, "hello", "value2"));
 
   ColumnFamilyHandle* cf = handles_[1];
   ReadOptions ro;
@@ -820,7 +822,7 @@ TEST_P(DBIteratorTest, IteratorDeleteAfterCfDelete) {
   ASSERT_EQ(IterStatus(iter), "foo->delete-cf-then-delete-iter");
 
   // delete CF handle
-  db_->DestroyColumnFamilyHandle(cf);
+  EXPECT_OK(db_->DestroyColumnFamilyHandle(cf));
   handles_.erase(std::begin(handles_) + 1);
 
   // delete Iterator after CF handle is deleted
@@ -832,7 +834,7 @@ TEST_P(DBIteratorTest, IteratorDeleteAfterCfDelete) {
 TEST_P(DBIteratorTest, IteratorDeleteAfterCfDrop) {
   CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
 
-  Put(1, "foo", "drop-cf-then-delete-iter");
+  ASSERT_OK(Put(1, "foo", "drop-cf-then-delete-iter"));
 
   ReadOptions ro;
   ColumnFamilyHandle* cf = handles_[1];
@@ -842,8 +844,8 @@ TEST_P(DBIteratorTest, IteratorDeleteAfterCfDrop) {
   ASSERT_EQ(IterStatus(iter), "foo->drop-cf-then-delete-iter");
 
   // drop and delete CF
-  db_->DropColumnFamily(cf);
-  db_->DestroyColumnFamilyHandle(cf);
+  EXPECT_OK(db_->DropColumnFamily(cf));
+  EXPECT_OK(db_->DestroyColumnFamilyHandle(cf));
   handles_.erase(std::begin(handles_) + 1);
 
   // delete Iterator after CF handle is dropped
@@ -1307,9 +1309,9 @@ TEST_P(DBIteratorTest, PrevAfterAndNextAfterMerge) {
 
   // write three entries with different keys using Merge()
   WriteOptions wopts;
-  db_->Merge(wopts, "1", "data1");
-  db_->Merge(wopts, "2", "data2");
-  db_->Merge(wopts, "3", "data3");
+  ASSERT_OK(db_->Merge(wopts, "1", "data1"));
+  ASSERT_OK(db_->Merge(wopts, "2", "data2"));
+  ASSERT_OK(db_->Merge(wopts, "3", "data3"));
 
   std::unique_ptr<Iterator> it(NewIterator(ReadOptions()));
 
@@ -1393,7 +1395,7 @@ class DBIteratorTestForPinnedData : public DBIteratorTest {
 
       if (run_config == TestConfig::FLUSH_EVERY_1000) {
         if (i && i % 1000 == 0) {
-          Flush();
+          ASSERT_OK(Flush());
         }
       }
     }
@@ -1402,7 +1404,7 @@ class DBIteratorTestForPinnedData : public DBIteratorTest {
       Close();
       Reopen(options);
     } else if (run_config == TestConfig::COMPACT_BEFORE_READ) {
-      db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+      ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
     }
 
     ReadOptions ro;
@@ -1768,6 +1770,7 @@ TEST_P(DBIteratorTest, IterSeekForPrevCrossingFiles) {
     Iterator* iter = NewIterator(ro);
     iter->SeekForPrev("c2");
     ASSERT_TRUE(!iter->Valid());
+    ASSERT_OK(iter->status());
     delete iter;
   }
 }
@@ -1823,6 +1826,7 @@ TEST_P(DBIteratorTest, IterSeekForPrevCrossingFilesCustomPrefixExtractor) {
     Iterator* iter = NewIterator(ro);
     iter->SeekForPrev("c2");
     ASSERT_TRUE(!iter->Valid());
+    ASSERT_OK(iter->status());
     delete iter;
   }
 }
@@ -2151,19 +2155,19 @@ TEST_P(DBIteratorTest, ReadAhead) {
 
   std::string value(1024, 'a');
   for (int i = 0; i < 100; i++) {
-    Put(Key(i), value);
+    ASSERT_OK(Put(Key(i), value));
   }
   ASSERT_OK(Flush());
   MoveFilesToLevel(2);
 
   for (int i = 0; i < 100; i++) {
-    Put(Key(i), value);
+    ASSERT_OK(Put(Key(i), value));
   }
   ASSERT_OK(Flush());
   MoveFilesToLevel(1);
 
   for (int i = 0; i < 100; i++) {
-    Put(Key(i), value);
+    ASSERT_OK(Put(Key(i), value));
   }
   ASSERT_OK(Flush());
 #ifndef ROCKSDB_LITE
@@ -2270,6 +2274,7 @@ TEST_P(DBIteratorTest, Refresh) {
   ASSERT_OK(Put("x", "y"));
 
   std::unique_ptr<Iterator> iter(NewIterator(ReadOptions()));
+  ASSERT_OK(iter->status());
   iter->Seek(Slice("a"));
   ASSERT_TRUE(iter->Valid());
   ASSERT_EQ(iter->key().compare(Slice("x")), 0);
@@ -2284,7 +2289,8 @@ TEST_P(DBIteratorTest, Refresh) {
   iter->Next();
   ASSERT_FALSE(iter->Valid());
 
-  iter->Refresh();
+  ASSERT_OK(iter->status());
+  ASSERT_OK(iter->Refresh());
 
   iter->Seek(Slice("a"));
   ASSERT_TRUE(iter->Valid());
@@ -2295,7 +2301,7 @@ TEST_P(DBIteratorTest, Refresh) {
   iter->Next();
   ASSERT_FALSE(iter->Valid());
 
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
 
   ASSERT_OK(Put("m", "n"));
 
@@ -2308,7 +2314,8 @@ TEST_P(DBIteratorTest, Refresh) {
   iter->Next();
   ASSERT_FALSE(iter->Valid());
 
-  iter->Refresh();
+  ASSERT_OK(iter->status());
+  ASSERT_OK(iter->Refresh());
 
   iter->Seek(Slice("a"));
   ASSERT_TRUE(iter->Valid());
@@ -2331,6 +2338,7 @@ TEST_P(DBIteratorTest, RefreshWithSnapshot) {
   ReadOptions options;
   options.snapshot = snapshot;
   Iterator* iter = NewIterator(options);
+  ASSERT_OK(iter->status());
 
   iter->Seek(Slice("a"));
   ASSERT_TRUE(iter->Valid());
@@ -2346,8 +2354,8 @@ TEST_P(DBIteratorTest, RefreshWithSnapshot) {
   iter->Next();
   ASSERT_FALSE(iter->Valid());
 
-  Status s;
-  s = iter->Refresh();
+  ASSERT_OK(iter->status());
+  Status s = iter->Refresh();
   ASSERT_TRUE(s.IsNotSupported());
   db_->ReleaseSnapshot(snapshot);
   delete iter;
@@ -2405,14 +2413,14 @@ TEST_P(DBIteratorTest, UpperBoundWithChangeDirection) {
 
 TEST_P(DBIteratorTest, TableFilter) {
   ASSERT_OK(Put("a", "1"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("b", "2"));
   ASSERT_OK(Put("c", "3"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
   ASSERT_OK(Put("d", "4"));
   ASSERT_OK(Put("e", "5"));
   ASSERT_OK(Put("f", "6"));
-  dbfull()->Flush(FlushOptions());
+  EXPECT_OK(dbfull()->Flush(FlushOptions()));
 
   // Ensure the table_filter callback is called once for each table.
   {
@@ -2597,13 +2605,13 @@ TEST_P(DBIteratorTest, SeekAfterHittingManyInternalKeys) {
   ReadOptions ropts;
   ropts.max_skippable_internal_keys = 2;
 
-  Put("1", "val_1");
+  ASSERT_OK(Put("1", "val_1"));
   // Add more tombstones than max_skippable_internal_keys so that Next() fails.
-  Delete("2");
-  Delete("3");
-  Delete("4");
-  Delete("5");
-  Put("6", "val_6");
+  ASSERT_OK(Delete("2"));
+  ASSERT_OK(Delete("3"));
+  ASSERT_OK(Delete("4"));
+  ASSERT_OK(Delete("5"));
+  ASSERT_OK(Put("6", "val_6"));
 
   std::unique_ptr<Iterator> iter(NewIterator(ropts));
   iter->SeekToFirst();
@@ -2645,9 +2653,9 @@ TEST_P(DBIteratorTest, NonBlockingIterationBugRepro) {
   DestroyAndReopen(options);
 
   // Two records in sst file, each in its own block.
-  Put("b", "");
-  Put("d", "");
-  Flush();
+  ASSERT_OK(Put("b", ""));
+  ASSERT_OK(Put("d", ""));
+  ASSERT_OK(Flush());
 
   // Create a nonblocking iterator before writing to memtable.
   ReadOptions ropt;
@@ -2657,7 +2665,7 @@ TEST_P(DBIteratorTest, NonBlockingIterationBugRepro) {
   // Overwrite a key in memtable many times to hit
   // max_sequential_skip_in_iterations (which is 8 by default).
   for (int i = 0; i < 20; ++i) {
-    Put("c", "");
+    ASSERT_OK(Put("c", ""));
   }
 
   // Load the second block in sst file into the block cache.
@@ -2674,9 +2682,9 @@ TEST_P(DBIteratorTest, NonBlockingIterationBugRepro) {
 }
 
 TEST_P(DBIteratorTest, SeekBackwardAfterOutOfUpperBound) {
-  Put("a", "");
-  Put("b", "");
-  Flush();
+  ASSERT_OK(Put("a", ""));
+  ASSERT_OK(Put("b", ""));
+  ASSERT_OK(Flush());
 
   ReadOptions ropt;
   Slice ub = "b";
@@ -2883,6 +2891,127 @@ TEST_P(DBIteratorTest, IterateWithLowerBoundAcrossFileBoundary) {
   ASSERT_OK(iter->status());
 }
 
+TEST_P(DBIteratorTest, Blob) {
+  Options options = CurrentOptions();
+  options.enable_blob_files = true;
+  options.max_sequential_skip_in_iterations = 2;
+  options.statistics = CreateDBStatistics();
+
+  Reopen(options);
+
+  // Note: we have 4 KVs (3 of which are hidden) for key "b" and
+  // max_sequential_skip_in_iterations is set to 2. Thus, we need to do a reseek
+  // anytime we move from "b" to "c" or vice versa.
+  ASSERT_OK(Put("a", "va"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("b", "vb0"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("b", "vb1"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("b", "vb2"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("b", "vb3"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("c", "vc"));
+  ASSERT_OK(Flush());
+
+  std::unique_ptr<Iterator> iter_guard(NewIterator(ReadOptions()));
+  Iterator* const iter = iter_guard.get();
+
+  iter->SeekToFirst();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 0);
+  ASSERT_EQ(IterStatus(iter), "a->va");
+  iter->Next();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 0);
+  ASSERT_EQ(IterStatus(iter), "b->vb3");
+  iter->Next();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 1);
+  ASSERT_EQ(IterStatus(iter), "c->vc");
+  iter->Next();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 1);
+  ASSERT_EQ(IterStatus(iter), "(invalid)");
+  iter->SeekToFirst();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 1);
+  ASSERT_EQ(IterStatus(iter), "a->va");
+  iter->Prev();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 1);
+  ASSERT_EQ(IterStatus(iter), "(invalid)");
+
+  iter->SeekToLast();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 1);
+  ASSERT_EQ(IterStatus(iter), "c->vc");
+  iter->Prev();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "b->vb3");
+  iter->Prev();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "a->va");
+  iter->Prev();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "(invalid)");
+  iter->SeekToLast();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "c->vc");
+  iter->Next();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "(invalid)");
+
+  iter->Seek("");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "a->va");
+  iter->Seek("a");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "a->va");
+  iter->Seek("ax");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "b->vb3");
+
+  iter->SeekForPrev("d");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "c->vc");
+  iter->SeekForPrev("c");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 2);
+  ASSERT_EQ(IterStatus(iter), "c->vc");
+  iter->SeekForPrev("bx");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 3);
+  ASSERT_EQ(IterStatus(iter), "b->vb3");
+
+  iter->Seek("b");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 3);
+  ASSERT_EQ(IterStatus(iter), "b->vb3");
+  iter->Seek("z");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 3);
+  ASSERT_EQ(IterStatus(iter), "(invalid)");
+  iter->SeekForPrev("b");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 4);
+  ASSERT_EQ(IterStatus(iter), "b->vb3");
+  iter->SeekForPrev("");
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 4);
+  ASSERT_EQ(IterStatus(iter), "(invalid)");
+
+  // Switch from reverse to forward
+  iter->SeekToLast();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 4);
+  iter->Prev();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 5);
+  iter->Prev();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 5);
+  iter->Next();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 5);
+  ASSERT_EQ(IterStatus(iter), "b->vb3");
+
+  // Switch from forward to reverse
+  iter->SeekToFirst();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 5);
+  iter->Next();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 5);
+  iter->Next();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 6);
+  iter->Prev();
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 7);
+  ASSERT_EQ(IterStatus(iter), "b->vb3");
+}
+
 INSTANTIATE_TEST_CASE_P(DBIteratorTestInstance, DBIteratorTest,
                         testing::Values(true, false));
 

db/db_logical_block_size_cache_test.cc

@@ -401,7 +401,7 @@ TEST_F(DBLogicalBlockSizeCacheTest, MultiDBWithDifferentPaths) {
   ColumnFamilyOptions cf_options0;
   cf_options0.cf_paths = {{cf_path_0_, 1024}};
   ColumnFamilyHandle* cf0;
-  db0->CreateColumnFamily(cf_options0, "cf", &cf0);
+  ASSERT_OK(db0->CreateColumnFamily(cf_options0, "cf", &cf0));
   ASSERT_EQ(2, cache_->Size());
   ASSERT_TRUE(cache_->Contains(data_path_0_));
   ASSERT_EQ(1, cache_->GetRefCount(data_path_0_));
@@ -421,7 +421,7 @@ TEST_F(DBLogicalBlockSizeCacheTest, MultiDBWithDifferentPaths) {
   ColumnFamilyOptions cf_options1;
   cf_options1.cf_paths = {{cf_path_1_, 1024}};
   ColumnFamilyHandle* cf1;
-  db1->CreateColumnFamily(cf_options1, "cf", &cf1);
+  ASSERT_OK(db1->CreateColumnFamily(cf_options1, "cf", &cf1));
   ASSERT_EQ(4, cache_->Size());
   ASSERT_TRUE(cache_->Contains(data_path_0_));
   ASSERT_EQ(1, cache_->GetRefCount(data_path_0_));
@@ -432,7 +432,7 @@ TEST_F(DBLogicalBlockSizeCacheTest, MultiDBWithDifferentPaths) {
   ASSERT_TRUE(cache_->Contains(cf_path_1_));
   ASSERT_EQ(1, cache_->GetRefCount(cf_path_1_));
 
-  db0->DestroyColumnFamilyHandle(cf0);
+  ASSERT_OK(db0->DestroyColumnFamilyHandle(cf0));
   delete db0;
   ASSERT_EQ(2, cache_->Size());
   ASSERT_TRUE(cache_->Contains(data_path_1_));
@@ -441,7 +441,7 @@ TEST_F(DBLogicalBlockSizeCacheTest, MultiDBWithDifferentPaths) {
   ASSERT_EQ(1, cache_->GetRefCount(cf_path_1_));
   ASSERT_OK(DestroyDB(data_path_0_, options, {{"cf", cf_options0}}));
 
-  db1->DestroyColumnFamilyHandle(cf1);
+  ASSERT_OK(db1->DestroyColumnFamilyHandle(cf1));
   delete db1;
   ASSERT_EQ(0, cache_->Size());
   ASSERT_OK(DestroyDB(data_path_1_, options, {{"cf", cf_options1}}));
@@ -466,7 +466,7 @@ TEST_F(DBLogicalBlockSizeCacheTest, MultiDBWithSamePaths) {
   ASSERT_EQ(1, cache_->GetRefCount(data_path_0_));
 
   ColumnFamilyHandle* cf0;
-  db0->CreateColumnFamily(cf_options, "cf", &cf0);
+  ASSERT_OK(db0->CreateColumnFamily(cf_options, "cf", &cf0));
   ASSERT_EQ(2, cache_->Size());
   ASSERT_TRUE(cache_->Contains(data_path_0_));
   ASSERT_EQ(1, cache_->GetRefCount(data_path_0_));
@@ -482,14 +482,14 @@ TEST_F(DBLogicalBlockSizeCacheTest, MultiDBWithSamePaths) {
   ASSERT_EQ(1, cache_->GetRefCount(cf_path_0_));
 
   ColumnFamilyHandle* cf1;
-  db1->CreateColumnFamily(cf_options, "cf", &cf1);
+  ASSERT_OK(db1->CreateColumnFamily(cf_options, "cf", &cf1));
   ASSERT_EQ(2, cache_->Size());
   ASSERT_TRUE(cache_->Contains(data_path_0_));
   ASSERT_EQ(2, cache_->GetRefCount(data_path_0_));
   ASSERT_TRUE(cache_->Contains(cf_path_0_));
   ASSERT_EQ(2, cache_->GetRefCount(cf_path_0_));
 
-  db0->DestroyColumnFamilyHandle(cf0);
+  ASSERT_OK(db0->DestroyColumnFamilyHandle(cf0));
   delete db0;
   ASSERT_EQ(2, cache_->Size());
   ASSERT_TRUE(cache_->Contains(data_path_0_));
@@ -498,7 +498,7 @@ TEST_F(DBLogicalBlockSizeCacheTest, MultiDBWithSamePaths) {
   ASSERT_EQ(1, cache_->GetRefCount(cf_path_0_));
   ASSERT_OK(DestroyDB(dbname_ + "/db0", options, {{"cf", cf_options}}));
 
-  db1->DestroyColumnFamilyHandle(cf1);
+  ASSERT_OK(db1->DestroyColumnFamilyHandle(cf1));
   delete db1;
   ASSERT_EQ(0, cache_->Size());
   ASSERT_OK(DestroyDB(dbname_ + "/db1", options, {{"cf", cf_options}}));

db/db_memtable_test.cc

@@ -129,7 +129,6 @@ class TestPrefixExtractor : public SliceTransform {
 TEST_F(DBMemTableTest, DuplicateSeq) {
   SequenceNumber seq = 123;
   std::string value;
-  Status s;
   MergeContext merge_context;
   Options options;
   InternalKeyComparator ikey_cmp(options.comparator);
@@ -146,22 +145,15 @@ TEST_F(DBMemTableTest, DuplicateSeq) {
                                kMaxSequenceNumber, 0 /* column_family_id */);
 
   // Write some keys and make sure it returns false on duplicates
-  bool res;
-  res = mem->Add(seq, kTypeValue, "key", "value2");
-  ASSERT_TRUE(res);
-  res = mem->Add(seq, kTypeValue, "key", "value2");
-  ASSERT_FALSE(res);
+  ASSERT_OK(mem->Add(seq, kTypeValue, "key", "value2"));
+  ASSERT_TRUE(mem->Add(seq, kTypeValue, "key", "value2").IsTryAgain());
   // Changing the type should still cause the duplicatae key
-  res = mem->Add(seq, kTypeMerge, "key", "value2");
-  ASSERT_FALSE(res);
+  ASSERT_TRUE(mem->Add(seq, kTypeMerge, "key", "value2").IsTryAgain());
   // Changing the seq number will make the key fresh
-  res = mem->Add(seq + 1, kTypeMerge, "key", "value2");
-  ASSERT_TRUE(res);
+  ASSERT_OK(mem->Add(seq + 1, kTypeMerge, "key", "value2"));
   // Test with different types for duplicate keys
-  res = mem->Add(seq, kTypeDeletion, "key", "");
-  ASSERT_FALSE(res);
-  res = mem->Add(seq, kTypeSingleDeletion, "key", "");
-  ASSERT_FALSE(res);
+  ASSERT_TRUE(mem->Add(seq, kTypeDeletion, "key", "").IsTryAgain());
+  ASSERT_TRUE(mem->Add(seq, kTypeSingleDeletion, "key", "").IsTryAgain());
 
   // Test the duplicate keys under stress
   for (int i = 0; i < 10000; i++) {
@@ -169,11 +161,11 @@ TEST_F(DBMemTableTest, DuplicateSeq) {
     if (!insert_dup) {
       seq++;
     }
-    res = mem->Add(seq, kTypeValue, "foo", "value" + ToString(seq));
+    Status s = mem->Add(seq, kTypeValue, "foo", "value" + ToString(seq));
     if (insert_dup) {
-      ASSERT_FALSE(res);
+      ASSERT_TRUE(s.IsTryAgain());
     } else {
-      ASSERT_TRUE(res);
+      ASSERT_OK(s);
     }
   }
   delete mem;
@@ -185,10 +177,8 @@ TEST_F(DBMemTableTest, DuplicateSeq) {
   mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
                      kMaxSequenceNumber, 0 /* column_family_id */);
   // Insert a duplicate key with _ in it
-  res = mem->Add(seq, kTypeValue, "key_1", "value");
-  ASSERT_TRUE(res);
-  res = mem->Add(seq, kTypeValue, "key_1", "value");
-  ASSERT_FALSE(res);
+  ASSERT_OK(mem->Add(seq, kTypeValue, "key_1", "value"));
+  ASSERT_TRUE(mem->Add(seq, kTypeValue, "key_1", "value").IsTryAgain());
   delete mem;
 
   // Test when InsertConcurrently will be invoked
@@ -197,10 +187,11 @@ TEST_F(DBMemTableTest, DuplicateSeq) {
   mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
                      kMaxSequenceNumber, 0 /* column_family_id */);
   MemTablePostProcessInfo post_process_info;
-  res = mem->Add(seq, kTypeValue, "key", "value", true, &post_process_info);
-  ASSERT_TRUE(res);
-  res = mem->Add(seq, kTypeValue, "key", "value", true, &post_process_info);
-  ASSERT_FALSE(res);
+  ASSERT_OK(
+      mem->Add(seq, kTypeValue, "key", "value", true, &post_process_info));
+  ASSERT_TRUE(
+      mem->Add(seq, kTypeValue, "key", "value", true, &post_process_info)
+          .IsTryAgain());
   delete mem;
 }
 
@@ -208,7 +199,6 @@ TEST_F(DBMemTableTest, DuplicateSeq) {
 TEST_F(DBMemTableTest, ConcurrentMergeWrite) {
   int num_ops = 1000;
   std::string value;
-  Status s;
   MergeContext merge_context;
   Options options;
   // A merge operator that is not sensitive to concurrent writes since in this
@@ -227,8 +217,7 @@ TEST_F(DBMemTableTest, ConcurrentMergeWrite) {
 
   // Put 0 as the base
   PutFixed64(&value, static_cast<uint64_t>(0));
-  bool res = mem->Add(0, kTypeValue, "key", value);
-  ASSERT_TRUE(res);
+  ASSERT_OK(mem->Add(0, kTypeValue, "key", value));
   value.clear();
 
   // Write Merge concurrently
@@ -237,9 +226,8 @@ TEST_F(DBMemTableTest, ConcurrentMergeWrite) {
     std::string v1;
     for (int seq = 1; seq < num_ops / 2; seq++) {
       PutFixed64(&v1, seq);
-      bool res1 =
-          mem->Add(seq, kTypeMerge, "key", v1, true, &post_process_info1);
-      ASSERT_TRUE(res1);
+      ASSERT_OK(
+          mem->Add(seq, kTypeMerge, "key", v1, true, &post_process_info1));
       v1.clear();
     }
   });
@@ -248,9 +236,8 @@ TEST_F(DBMemTableTest, ConcurrentMergeWrite) {
     std::string v2;
     for (int seq = num_ops / 2; seq < num_ops; seq++) {
       PutFixed64(&v2, seq);
-      bool res2 =
-          mem->Add(seq, kTypeMerge, "key", v2, true, &post_process_info2);
-      ASSERT_TRUE(res2);
+      ASSERT_OK(
+          mem->Add(seq, kTypeMerge, "key", v2, true, &post_process_info2));
       v2.clear();
     }
   });
@@ -261,8 +248,9 @@ TEST_F(DBMemTableTest, ConcurrentMergeWrite) {
   ReadOptions roptions;
   SequenceNumber max_covering_tombstone_seq = 0;
   LookupKey lkey("key", kMaxSequenceNumber);
-  res = mem->Get(lkey, &value, /*timestamp=*/nullptr, &status, &merge_context,
-                 &max_covering_tombstone_seq, roptions);
+  bool res = mem->Get(lkey, &value, /*timestamp=*/nullptr, &status,
+                      &merge_context, &max_covering_tombstone_seq, roptions);
+  ASSERT_OK(status);
   ASSERT_TRUE(res);
   uint64_t ivalue = DecodeFixed64(Slice(value).data());
   uint64_t sum = 0;

db/db_merge_operand_test.cc

@@ -59,29 +59,29 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   merge_operands_info.expected_max_number_of_operands = num_records;
 
   // k0 value in memtable
-  Put("k0", "PutARock");
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k0",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(Put("k0", "PutARock"));
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k0", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "PutARock");
 
   // k0.1 value in SST
-  Put("k0.1", "RockInSST");
+  ASSERT_OK(Put("k0.1", "RockInSST"));
   ASSERT_OK(Flush());
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k0.1",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k0.1", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "RockInSST");
 
   // All k1 values are in memtable.
   ASSERT_OK(Merge("k1", "a"));
-  Put("k1", "x");
+  ASSERT_OK(Put("k1", "x"));
   ASSERT_OK(Merge("k1", "b"));
   ASSERT_OK(Merge("k1", "c"));
   ASSERT_OK(Merge("k1", "d"));
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k1", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "x");
   ASSERT_EQ(values[1], "b");
   ASSERT_EQ(values[2], "c");
@@ -98,13 +98,13 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
 
   // All k1.1 values are in memtable.
   ASSERT_OK(Merge("k1.1", "r"));
-  Delete("k1.1");
+  ASSERT_OK(Delete("k1.1"));
   ASSERT_OK(Merge("k1.1", "c"));
   ASSERT_OK(Merge("k1.1", "k"));
   ASSERT_OK(Merge("k1.1", "s"));
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1.1",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k1.1", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "c");
   ASSERT_EQ(values[1], "k");
   ASSERT_EQ(values[2], "s");
@@ -115,9 +115,9 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   ASSERT_OK(Merge("k2", "e"));
   ASSERT_OK(Merge("k2", "r"));
   ASSERT_OK(Flush());
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k2",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k2", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "q");
   ASSERT_EQ(values[1], "w");
   ASSERT_EQ(values[2], "e");
@@ -125,30 +125,30 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
 
   // All k2.1 values are flushed to L0 into a single file.
   ASSERT_OK(Merge("k2.1", "m"));
-  Put("k2.1", "l");
+  ASSERT_OK(Put("k2.1", "l"));
   ASSERT_OK(Merge("k2.1", "n"));
   ASSERT_OK(Merge("k2.1", "o"));
   ASSERT_OK(Flush());
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k2.1",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k2.1", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "l,n,o");
 
   // All k2.2 values are flushed to L0 into a single file.
   ASSERT_OK(Merge("k2.2", "g"));
-  Delete("k2.2");
+  ASSERT_OK(Delete("k2.2"));
   ASSERT_OK(Merge("k2.2", "o"));
   ASSERT_OK(Merge("k2.2", "t"));
   ASSERT_OK(Flush());
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k2.2",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k2.2", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "o,t");
 
   // Do some compaction that will make the following tests more predictable
   //  Slice start("PutARock");
   //  Slice end("t");
-  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
 
   // All k3 values are flushed and are in different files.
   ASSERT_OK(Merge("k3", "ab"));
@@ -158,9 +158,9 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   ASSERT_OK(Merge("k3", "cd"));
   ASSERT_OK(Flush());
   ASSERT_OK(Merge("k3", "de"));
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k3",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k3", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "ab");
   ASSERT_EQ(values[1], "bc");
   ASSERT_EQ(values[2], "cd");
@@ -169,14 +169,14 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   // All k3.1 values are flushed and are in different files.
   ASSERT_OK(Merge("k3.1", "ab"));
   ASSERT_OK(Flush());
-  Put("k3.1", "bc");
+  ASSERT_OK(Put("k3.1", "bc"));
   ASSERT_OK(Flush());
   ASSERT_OK(Merge("k3.1", "cd"));
   ASSERT_OK(Flush());
   ASSERT_OK(Merge("k3.1", "de"));
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k3.1",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k3.1", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "bc");
   ASSERT_EQ(values[1], "cd");
   ASSERT_EQ(values[2], "de");
@@ -184,14 +184,14 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   // All k3.2 values are flushed and are in different files.
   ASSERT_OK(Merge("k3.2", "ab"));
   ASSERT_OK(Flush());
-  Delete("k3.2");
+  ASSERT_OK(Delete("k3.2"));
   ASSERT_OK(Flush());
   ASSERT_OK(Merge("k3.2", "cd"));
   ASSERT_OK(Flush());
   ASSERT_OK(Merge("k3.2", "de"));
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k3.2",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k3.2", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "cd");
   ASSERT_EQ(values[1], "de");
 
@@ -206,9 +206,9 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   ASSERT_OK(Flush());
   MoveFilesToLevel(1);
   ASSERT_OK(Merge("k4", "ed"));
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k4",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k4", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "ba");
   ASSERT_EQ(values[1], "cb");
   ASSERT_EQ(values[2], "dc");
@@ -219,14 +219,14 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   ASSERT_OK(Merge("k5", "am"));
   ASSERT_OK(Merge("k5", "i"));
   ASSERT_OK(Flush());
-  Put("k5", "remember");
+  ASSERT_OK(Put("k5", "remember"));
   ASSERT_OK(Merge("k5", "i"));
   ASSERT_OK(Merge("k5", "am"));
   ASSERT_OK(Merge("k5", "rocks"));
-  dbfull()->TEST_SwitchMemtable();
-  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k5",
-                        values.data(), &merge_operands_info,
-                        &number_of_operands);
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable());
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k5", values.data(), &merge_operands_info,
+                                  &number_of_operands));
   ASSERT_EQ(values[0], "remember");
   ASSERT_EQ(values[1], "i");
   ASSERT_EQ(values[2], "am");

db/db_merge_operator_test.cc

@@ -94,7 +94,7 @@ TEST_F(DBMergeOperatorTest, LimitMergeOperands) {
   ASSERT_OK(Merge("k1", "c"));
   ASSERT_OK(Merge("k1", "d"));
   std::string value;
-  ASSERT_TRUE(db_->Get(ReadOptions(), "k1", &value).ok());
+  ASSERT_OK(db_->Get(ReadOptions(), "k1", &value));
   // Make sure that only the latest two merge operands are used. If this was
   // not the case the value would be "a,b,c,d".
   ASSERT_EQ(value, "c,d");
@@ -105,7 +105,7 @@ TEST_F(DBMergeOperatorTest, LimitMergeOperands) {
   ASSERT_OK(Merge("k2", "c"));
   ASSERT_OK(Merge("k2", "d"));
   ASSERT_OK(Flush());
-  ASSERT_TRUE(db_->Get(ReadOptions(), "k2", &value).ok());
+  ASSERT_OK(db_->Get(ReadOptions(), "k2", &value));
   ASSERT_EQ(value, "c,d");
 
   // All K3 values are flushed and are in different files.
@@ -116,7 +116,7 @@ TEST_F(DBMergeOperatorTest, LimitMergeOperands) {
   ASSERT_OK(Merge("k3", "cd"));
   ASSERT_OK(Flush());
   ASSERT_OK(Merge("k3", "de"));
-  ASSERT_TRUE(db_->Get(ReadOptions(), "k3", &value).ok());
+  ASSERT_OK(db_->Get(ReadOptions(), "k3", &value));
   ASSERT_EQ(value, "cd,de");
 
   // All K4 values are in different levels
@@ -130,7 +130,7 @@ TEST_F(DBMergeOperatorTest, LimitMergeOperands) {
   ASSERT_OK(Flush());
   MoveFilesToLevel(1);
   ASSERT_OK(Merge("k4", "de"));
-  ASSERT_TRUE(db_->Get(ReadOptions(), "k4", &value).ok());
+  ASSERT_OK(db_->Get(ReadOptions(), "k4", &value));
   ASSERT_EQ(value, "cd,de");
 }
 
@@ -344,8 +344,9 @@ TEST_P(MergeOperatorPinningTest, EvictCacheBeforeMerge) {
   // Code executed before merge operation
   merge_hook->before_merge_ = [&]() {
     // Evict all tables from cache before every merge operation
+    auto* table_cache = dbfull()->TEST_table_cache();
     for (uint64_t num : file_numbers) {
-      TableCache::Evict(dbfull()->TEST_table_cache(), num);
+      TableCache::Evict(table_cache, num);
     }
     // Decrease cache capacity to force all unrefed blocks to be evicted
     if (bbto.block_cache) {
@@ -366,7 +367,7 @@ TEST_P(MergeOperatorPinningTest, EvictCacheBeforeMerge) {
   VerifyDBFromMap(true_data, &total_reads);
   ASSERT_EQ(merge_cnt, total_reads);
 
-  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
 
   VerifyDBFromMap(true_data, &total_reads);
 }
@@ -385,7 +386,7 @@ TEST_P(MergeOperatorPinningTest, TailingIterator) {
   std::function<void()> writer_func = [&]() {
     int k = 0;
     for (int i = 0; i < kNumWrites; i++) {
-      db_->Merge(WriteOptions(), Key(k), Key(k));
+      ASSERT_OK(db_->Merge(WriteOptions(), Key(k), Key(k)));
 
       if (i && i % kNumOperands == 0) {
         k++;
@@ -403,7 +404,7 @@ TEST_P(MergeOperatorPinningTest, TailingIterator) {
     ReadOptions ro;
     ro.tailing = true;
     Iterator* iter = db_->NewIterator(ro);
-
+    ASSERT_OK(iter->status());
     iter->SeekToFirst();
     for (int i = 0; i < (kNumWrites / kNumOperands); i++) {
       while (!iter->Valid()) {
@@ -416,6 +417,7 @@ TEST_P(MergeOperatorPinningTest, TailingIterator) {
 
       iter->Next();
     }
+    ASSERT_OK(iter->status());
 
     delete iter;
   };
@@ -449,12 +451,13 @@ TEST_F(DBMergeOperatorTest, TailingIteratorMemtableUnrefedBySomeoneElse) {
   //    ForwardIterator to not pin it in some circumstances. This test
   //    reproduces it.
 
-  db_->Merge(WriteOptions(), "key", "sst");
-  db_->Flush(FlushOptions()); // Switch to SuperVersion A
-  db_->Merge(WriteOptions(), "key", "memtable");
+  ASSERT_OK(db_->Merge(WriteOptions(), "key", "sst"));
+  ASSERT_OK(db_->Flush(FlushOptions()));  // Switch to SuperVersion A
+  ASSERT_OK(db_->Merge(WriteOptions(), "key", "memtable"));
 
   // Pin SuperVersion A
   std::unique_ptr<Iterator> someone_else(db_->NewIterator(ReadOptions()));
+  ASSERT_OK(someone_else->status());
 
   bool pushed_first_operand = false;
   bool stepped_to_next_operand = false;
@@ -462,7 +465,7 @@ TEST_F(DBMergeOperatorTest, TailingIteratorMemtableUnrefedBySomeoneElse) {
       "DBIter::MergeValuesNewToOld:PushedFirstOperand", [&](void*) {
         EXPECT_FALSE(pushed_first_operand);
         pushed_first_operand = true;
-        db_->Flush(FlushOptions()); // Switch to SuperVersion B
+        EXPECT_OK(db_->Flush(FlushOptions()));  // Switch to SuperVersion B
       });
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
       "DBIter::MergeValuesNewToOld:SteppedToNextOperand", [&](void*) {
@@ -477,7 +480,7 @@ TEST_F(DBMergeOperatorTest, TailingIteratorMemtableUnrefedBySomeoneElse) {
   std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
   iter->Seek("key");
 
-  ASSERT_TRUE(iter->status().ok());
+  ASSERT_OK(iter->status());
   ASSERT_TRUE(iter->Valid());
   EXPECT_EQ(std::string("sst,memtable"), iter->value().ToString());
   EXPECT_TRUE(pushed_first_operand);

db/db_test2.cc

@@ -174,7 +174,9 @@ TEST_F(DBTest2, PartitionedIndexUserToInternalKey) {
     ASSERT_OK(Put("keykey_" + std::to_string(j), value));
     snapshots.push_back(db_->GetSnapshot());
   }
-  Flush();
+  ASSERT_OK(Flush());
+  dbfull()->TEST_WaitForFlushMemTable();
+
   for (auto s : snapshots) {
     db_->ReleaseSnapshot(s);
   }

db/db_test_util.cc

@@ -200,28 +200,28 @@ bool DBTestBase::ChangeCompactOptions() {
     Destroy(last_options_);
     auto options = CurrentOptions();
     options.create_if_missing = true;
-    TryReopen(options);
+    Reopen(options);
     return true;
   } else if (option_config_ == kUniversalCompaction) {
     option_config_ = kUniversalCompactionMultiLevel;
     Destroy(last_options_);
     auto options = CurrentOptions();
     options.create_if_missing = true;
-    TryReopen(options);
+    Reopen(options);
     return true;
   } else if (option_config_ == kUniversalCompactionMultiLevel) {
     option_config_ = kLevelSubcompactions;
     Destroy(last_options_);
     auto options = CurrentOptions();
     assert(options.max_subcompactions > 1);
-    TryReopen(options);
+    Reopen(options);
     return true;
   } else if (option_config_ == kLevelSubcompactions) {
     option_config_ = kUniversalSubcompactions;
     Destroy(last_options_);
     auto options = CurrentOptions();
     assert(options.max_subcompactions > 1);
-    TryReopen(options);
+    Reopen(options);
     return true;
   } else {
     return false;
@@ -680,7 +680,7 @@ void DBTestBase::Close() {
 void DBTestBase::DestroyAndReopen(const Options& options) {
   // Destroy using last options
   Destroy(last_options_);
-  ASSERT_OK(TryReopen(options));
+  Reopen(options);
 }
 
 void DBTestBase::Destroy(const Options& options, bool delete_cf_paths) {

db/db_wal_test.cc

@@ -358,16 +358,16 @@ TEST_F(DBWALTest, RecoverWithBlob) {
   // There should be no files just yet since we haven't flushed.
   {
     VersionSet* const versions = dbfull()->TEST_GetVersionSet();
-    assert(versions);
+    ASSERT_NE(versions, nullptr);
 
     ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
-    assert(cfd);
+    ASSERT_NE(cfd, nullptr);
 
     Version* const current = cfd->current();
-    assert(current);
+    ASSERT_NE(current, nullptr);
 
     const VersionStorageInfo* const storage_info = current->storage_info();
-    assert(storage_info);
+    ASSERT_NE(storage_info, nullptr);
 
     ASSERT_EQ(storage_info->num_non_empty_levels(), 0);
     ASSERT_TRUE(storage_info->GetBlobFiles().empty());
@@ -388,28 +388,28 @@ TEST_F(DBWALTest, RecoverWithBlob) {
   ASSERT_EQ(Get("key2"), long_value);
 
   VersionSet* const versions = dbfull()->TEST_GetVersionSet();
-  assert(versions);
+  ASSERT_NE(versions, nullptr);
 
   ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
-  assert(cfd);
+  ASSERT_NE(cfd, nullptr);
 
   Version* const current = cfd->current();
-  assert(current);
+  ASSERT_NE(current, nullptr);
 
   const VersionStorageInfo* const storage_info = current->storage_info();
-  assert(storage_info);
+  ASSERT_NE(storage_info, nullptr);
 
   const auto& l0_files = storage_info->LevelFiles(0);
   ASSERT_EQ(l0_files.size(), 1);
 
   const FileMetaData* const table_file = l0_files[0];
-  assert(table_file);
+  ASSERT_NE(table_file, nullptr);
 
   const auto& blob_files = storage_info->GetBlobFiles();
   ASSERT_EQ(blob_files.size(), 1);
 
   const auto& blob_file = blob_files.begin()->second;
-  assert(blob_file);
+  ASSERT_NE(blob_file, nullptr);
 
   ASSERT_EQ(table_file->smallest.user_key(), "key1");
   ASSERT_EQ(table_file->largest.user_key(), "key2");
@@ -422,7 +422,7 @@ TEST_F(DBWALTest, RecoverWithBlob) {
 
 #ifndef ROCKSDB_LITE
   const InternalStats* const internal_stats = cfd->internal_stats();
-  assert(internal_stats);
+  ASSERT_NE(internal_stats, nullptr);
 
   const uint64_t expected_bytes =
       table_file->fd.GetFileSize() + blob_file->GetTotalBlobBytes();
@@ -502,12 +502,12 @@ TEST_F(DBWALTest, IgnoreRecoveredLog) {
 
   do {
     // delete old files in backup_logs directory
-    env_->CreateDirIfMissing(backup_logs);
+    ASSERT_OK(env_->CreateDirIfMissing(backup_logs));
     std::vector<std::string> old_files;
-    env_->GetChildren(backup_logs, &old_files);
+    ASSERT_OK(env_->GetChildren(backup_logs, &old_files));
     for (auto& file : old_files) {
       if (file != "." && file != "..") {
-        env_->DeleteFile(backup_logs + "/" + file);
+        ASSERT_OK(env_->DeleteFile(backup_logs + "/" + file));
       }
     }
     Options options = CurrentOptions();
@@ -526,7 +526,7 @@ TEST_F(DBWALTest, IgnoreRecoveredLog) {
 
     // copy the logs to backup
     std::vector<std::string> logs;
-    env_->GetChildren(options.wal_dir, &logs);
+    ASSERT_OK(env_->GetChildren(options.wal_dir, &logs));
     for (auto& log : logs) {
       if (log != ".." && log != ".") {
         CopyFile(options.wal_dir + "/" + log, backup_logs + "/" + log);
@@ -557,7 +557,7 @@ TEST_F(DBWALTest, IgnoreRecoveredLog) {
     Close();
 
     // copy the logs from backup back to wal dir
-    env_->CreateDirIfMissing(options.wal_dir);
+    ASSERT_OK(env_->CreateDirIfMissing(options.wal_dir));
     for (auto& log : logs) {
       if (log != ".." && log != ".") {
         CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
@@ -572,16 +572,16 @@ TEST_F(DBWALTest, IgnoreRecoveredLog) {
     // Recovery will fail if DB directory doesn't exist.
     Destroy(options);
     // copy the logs from backup back to wal dir
-    env_->CreateDirIfMissing(options.wal_dir);
+    ASSERT_OK(env_->CreateDirIfMissing(options.wal_dir));
     for (auto& log : logs) {
       if (log != ".." && log != ".") {
         CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
         // we won't be needing this file no more
-        env_->DeleteFile(backup_logs + "/" + log);
+        ASSERT_OK(env_->DeleteFile(backup_logs + "/" + log));
       }
     }
     Status s = TryReopen(options);
-    ASSERT_TRUE(!s.ok());
+    ASSERT_NOK(s);
     Destroy(options);
   } while (ChangeWalOptions());
 }
@@ -619,9 +619,9 @@ TEST_F(DBWALTest, PreallocateBlock) {
         called.fetch_add(1);
       });
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-  Put("", "");
-  Flush();
-  Put("", "");
+  ASSERT_OK(Put("", ""));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("", ""));
   Close();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ASSERT_EQ(2, called.load());
@@ -638,9 +638,9 @@ TEST_F(DBWALTest, PreallocateBlock) {
         called.fetch_add(1);
       });
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-  Put("", "");
-  Flush();
-  Put("", "");
+  ASSERT_OK(Put("", ""));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("", ""));
   Close();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ASSERT_EQ(2, called.load());
@@ -658,9 +658,9 @@ TEST_F(DBWALTest, PreallocateBlock) {
         called.fetch_add(1);
       });
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-  Put("", "");
-  Flush();
-  Put("", "");
+  ASSERT_OK(Put("", ""));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("", ""));
   Close();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ASSERT_EQ(2, called.load());
@@ -679,9 +679,9 @@ TEST_F(DBWALTest, PreallocateBlock) {
         called.fetch_add(1);
       });
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-  Put("", "");
-  Flush();
-  Put("", "");
+  ASSERT_OK(Put("", ""));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("", ""));
   Close();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ASSERT_EQ(2, called.load());
@@ -907,7 +907,7 @@ TEST_F(DBWALTest, RecoverCheckFileAmountWithSmallWriteBuffer) {
   // Make 'dobrynia' to be flushed and new WAL file to be created
   ASSERT_OK(Put(2, Key(10), DummyString(7500000)));
   ASSERT_OK(Put(2, Key(1), DummyString(1)));
-  dbfull()->TEST_WaitForFlushMemTable(handles_[2]);
+  ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[2]));
   {
     auto tables = ListTableFiles(env_, dbname_);
     ASSERT_EQ(tables.size(), static_cast<size_t>(1));
@@ -961,7 +961,7 @@ TEST_F(DBWALTest, RecoverCheckFileAmount) {
   // Make 'nikitich' memtable to be flushed
   ASSERT_OK(Put(3, Key(10), DummyString(1002400)));
   ASSERT_OK(Put(3, Key(1), DummyString(1)));
-  dbfull()->TEST_WaitForFlushMemTable(handles_[3]);
+  ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[3]));
   // 4 memtable are not flushed, 1 sst file
   {
     auto tables = ListTableFiles(env_, dbname_);
@@ -981,7 +981,7 @@ TEST_F(DBWALTest, RecoverCheckFileAmount) {
   ASSERT_OK(Put(3, Key(10), DummyString(1002400)));
   // make it flush
   ASSERT_OK(Put(3, Key(1), DummyString(1)));
-  dbfull()->TEST_WaitForFlushMemTable(handles_[3]);
+  ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[3]));
   // There are still 4 memtable not flushed, and 2 sst tables
   ASSERT_OK(Put(0, Key(1), DummyString(1)));
   ASSERT_OK(Put(1, Key(1), DummyString(1)));
@@ -1029,10 +1029,10 @@ TEST_F(DBWALTest, SyncMultipleLogs) {
   for (uint64_t b = 0; b < kNumBatches; b++) {
     batch.Clear();
     for (int i = 0; i < kBatchSize; i++) {
-      batch.Put(Key(i), DummyString(128));
+      ASSERT_OK(batch.Put(Key(i), DummyString(128)));
     }
 
-    dbfull()->Write(wo, &batch);
+    ASSERT_OK(dbfull()->Write(wo, &batch));
   }
 
   ASSERT_OK(dbfull()->SyncWAL());
@@ -1060,7 +1060,7 @@ TEST_F(DBWALTest, PartOfWritesWithWALDisabled) {
   ASSERT_OK(Flush(0));
   ASSERT_OK(Put(0, "key", "v5", wal_on));  // seq id 5
   ASSERT_EQ("v5", Get(0, "key"));
-  dbfull()->FlushWAL(false);
+  ASSERT_OK(dbfull()->FlushWAL(false));
   // Simulate a crash.
   fault_env->SetFilesystemActive(false);
   Close();
@@ -1128,12 +1128,13 @@ class RecoveryTestHelper {
       for (int i = 0; i < kKeysPerWALFile; i++) {
         std::string key = "key" + ToString((*count)++);
         std::string value = test->DummyString(kValueSize);
-        assert(current_log_writer.get() != nullptr);
+        ASSERT_NE(current_log_writer.get(), nullptr);
         uint64_t seq = versions->LastSequence() + 1;
         batch.Clear();
-        batch.Put(key, value);
+        ASSERT_OK(batch.Put(key, value));
         WriteBatchInternal::SetSequence(&batch, seq);
-        current_log_writer->AddRecord(WriteBatchInternal::Contents(&batch));
+        ASSERT_OK(current_log_writer->AddRecord(
+            WriteBatchInternal::Contents(&batch)));
         versions->SetLastAllocatedSequence(seq);
         versions->SetLastPublishedSequence(seq);
         versions->SetLastSequence(seq);
@@ -1309,10 +1310,11 @@ TEST_F(DBWALTest, kPointInTimeRecoveryCFConsistency) {
 
   ASSERT_OK(Put(1, "key3", "val3"));
   // Corrupt WAL at location of key3
-  test::CorruptFile(env, fname, static_cast<int>(offset_to_corrupt), 4, false);
+  ASSERT_OK(test::CorruptFile(env, fname, static_cast<int>(offset_to_corrupt),
+                              4, false));
   ASSERT_OK(Put(2, "key4", "val4"));
   ASSERT_OK(Put(1, "key5", "val5"));
-  Flush(2);
+  ASSERT_OK(Flush(2));
 
   // PIT recovery & verify
   options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
@@ -1466,7 +1468,7 @@ TEST_F(DBWALTest, WalCleanupAfterAvoidFlushDuringRecovery) {
   for (int i = 0; i < 2; ++i) {
     if (i > 0) {
       // Flush() triggers deletion of obsolete tracked files
-      Flush();
+      ASSERT_OK(Flush());
     }
     VectorLogPtr log_files;
     ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
@@ -1508,7 +1510,7 @@ TEST_F(DBWALTest, RecoverWithoutFlush) {
   ASSERT_EQ(Get("foo"), "foo_v2");
   ASSERT_EQ(Get("bar"), "bar_v2");
   // manual flush and insert again
-  Flush();
+  ASSERT_OK(Flush());
   ASSERT_EQ(Get("foo"), "foo_v2");
   ASSERT_EQ(Get("bar"), "bar_v2");
   ASSERT_OK(Put("foo", "foo_v3"));
@@ -1529,7 +1531,9 @@ TEST_F(DBWALTest, RecoverWithoutFlushMultipleCF) {
 
   auto countWalFiles = [this]() {
     VectorLogPtr log_files;
-    dbfull()->GetSortedWalFiles(log_files);
+    if (!dbfull()->GetSortedWalFiles(log_files).ok()) {
+      return size_t{0};
+    }
     return log_files.size();
   };
 
@@ -1537,11 +1541,11 @@ TEST_F(DBWALTest, RecoverWithoutFlushMultipleCF) {
   CreateAndReopenWithCF({"one", "two"}, options);
   ASSERT_OK(Put(0, "key1", kSmallValue));
   ASSERT_OK(Put(1, "key2", kLargeValue));
-  Flush(1);
+  ASSERT_OK(Flush(1));
   ASSERT_EQ(1, countWalFiles());
   ASSERT_OK(Put(0, "key3", kSmallValue));
   ASSERT_OK(Put(2, "key4", kLargeValue));
-  Flush(2);
+  ASSERT_OK(Flush(2));
   ASSERT_EQ(2, countWalFiles());
 
   // Reopen, insert and flush.
@@ -1555,9 +1559,9 @@ TEST_F(DBWALTest, RecoverWithoutFlushMultipleCF) {
   ASSERT_OK(Put(0, "key5", kLargeValue));
   ASSERT_OK(Put(1, "key6", kLargeValue));
   ASSERT_EQ(3, countWalFiles());
-  Flush(1);
+  ASSERT_OK(Flush(1));
   ASSERT_OK(Put(2, "key7", kLargeValue));
-  dbfull()->FlushWAL(false);
+  ASSERT_OK(dbfull()->FlushWAL(false));
   ASSERT_EQ(4, countWalFiles());
 
   // Reopen twice and validate.
@@ -1646,7 +1650,7 @@ TEST_F(DBWALTest, RestoreTotalLogSizeAfterRecoverWithoutFlush) {
 
     void OnFlushBegin(DB* /*db*/, const FlushJobInfo& flush_job_info) override {
       count++;
-      assert(FlushReason::kWriteBufferManager == flush_job_info.flush_reason);
+      ASSERT_EQ(FlushReason::kWriteBufferManager, flush_job_info.flush_reason);
     }
   };
   std::shared_ptr<TestFlushListener> test_listener =
@@ -1690,7 +1694,9 @@ TEST_F(DBWALTest, RestoreTotalLogSizeAfterRecoverWithoutFlush) {
             1 * kMB);
   // Write one more key to trigger flush.
   ASSERT_OK(Put(0, "foo", "v2"));
-  dbfull()->TEST_WaitForFlushMemTable();
+  for (auto* h : handles_) {
+    ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(h));
+  }
   // Flushed two column families.
   ASSERT_EQ(2, test_listener->count.load());
 }
@@ -1764,9 +1770,9 @@ TEST_F(DBWALTest, WalTermTest) {
   wo.disableWAL = false;
 
   WriteBatch batch;
-  batch.Put("foo", "bar");
+  ASSERT_OK(batch.Put("foo", "bar"));
   batch.MarkWalTerminationPoint();
-  batch.Put("foo2", "bar2");
+  ASSERT_OK(batch.Put("foo2", "bar2"));
 
   ASSERT_OK(dbfull()->Write(wo, &batch));
 

db/db_with_timestamp_basic_test.cc

@@ -32,6 +32,22 @@ class DBBasicTestWithTimestampBase : public DBTestBase {
     return ret;
   }
 
+  static std::string KeyWithPrefix(std::string prefix, uint64_t k) {
+    std::string ret;
+    PutFixed64(&ret, k);
+    std::reverse(ret.begin(), ret.end());
+    return prefix + ret;
+  }
+
+  static std::vector<Slice> ConvertStrToSlice(
+      std::vector<std::string>& strings) {
+    std::vector<Slice> ret;
+    for (const auto& s : strings) {
+      ret.emplace_back(s);
+    }
+    return ret;
+  }
+
   class TestComparator : public Comparator {
    private:
     const Comparator* cmp_without_ts_;
@@ -179,6 +195,113 @@ class DBBasicTestWithTimestamp : public DBBasicTestWithTimestampBase {
       : DBBasicTestWithTimestampBase("db_basic_test_with_timestamp") {}
 };
 
+TEST_F(DBBasicTestWithTimestamp, CompactRangeWithSpecifiedRange) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+  DestroyAndReopen(options);
+
+  WriteOptions write_opts;
+  std::string ts_str = Timestamp(1, 0);
+  Slice ts = ts_str;
+  write_opts.timestamp = &ts;
+
+  ASSERT_OK(db_->Put(write_opts, "foo1", "bar"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(db_->Put(write_opts, "foo2", "bar"));
+  ASSERT_OK(Flush());
+
+  std::string start_str = "foo";
+  std::string end_str = "foo2";
+  Slice start(start_str), end(end_str);
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), &start, &end));
+
+  Close();
+}
+
+TEST_F(DBBasicTestWithTimestamp, GetApproximateSizes) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 100000000;  // Large write buffer
+  options.compression = kNoCompression;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+  DestroyAndReopen(options);
+  auto default_cf = db_->DefaultColumnFamily();
+
+  WriteOptions write_opts;
+  std::string ts_str = Timestamp(1, 0);
+  Slice ts = ts_str;
+  write_opts.timestamp = &ts;
+
+  const int N = 128;
+  Random rnd(301);
+  for (int i = 0; i < N; i++) {
+    ASSERT_OK(db_->Put(write_opts, Key(i), rnd.RandomString(1024)));
+  }
+
+  uint64_t size;
+  std::string start = Key(50);
+  std::string end = Key(60);
+  Range r(start, end);
+  SizeApproximationOptions size_approx_options;
+  size_approx_options.include_memtabtles = true;
+  size_approx_options.include_files = true;
+  ASSERT_OK(
+      db_->GetApproximateSizes(size_approx_options, default_cf, &r, 1, &size));
+  ASSERT_GT(size, 6000);
+  ASSERT_LT(size, 204800);
+
+  // test multiple ranges
+  std::vector<Range> ranges;
+  std::string start_tmp = Key(10);
+  std::string end_tmp = Key(20);
+  ranges.emplace_back(Range(start_tmp, end_tmp));
+  ranges.emplace_back(Range(start, end));
+  uint64_t range_sizes[2];
+  ASSERT_OK(db_->GetApproximateSizes(size_approx_options, default_cf,
+                                     ranges.data(), 2, range_sizes));
+
+  ASSERT_EQ(range_sizes[1], size);
+
+  // Zero if not including mem table
+  db_->GetApproximateSizes(&r, 1, &size);
+  ASSERT_EQ(size, 0);
+
+  start = Key(500);
+  end = Key(600);
+  r = Range(start, end);
+  ASSERT_OK(
+      db_->GetApproximateSizes(size_approx_options, default_cf, &r, 1, &size));
+  ASSERT_EQ(size, 0);
+
+  // Test range boundaries
+  ASSERT_OK(db_->Put(write_opts, Key(1000), rnd.RandomString(1024)));
+  // Should include start key
+  start = Key(1000);
+  end = Key(1100);
+  r = Range(start, end);
+  ASSERT_OK(
+      db_->GetApproximateSizes(size_approx_options, default_cf, &r, 1, &size));
+  ASSERT_GT(size, 0);
+
+  // Should exclude end key
+  start = Key(900);
+  end = Key(1000);
+  r = Range(start, end);
+  ASSERT_OK(
+      db_->GetApproximateSizes(size_approx_options, default_cf, &r, 1, &size));
+  ASSERT_EQ(size, 0);
+  std::cout << size << std::endl;
+
+  Close();
+}
+
 TEST_F(DBBasicTestWithTimestamp, SimpleForwardIterate) {
   const int kNumKeysPerFile = 128;
   const uint64_t kMaxKey = 1024;
@@ -249,6 +372,8 @@ TEST_F(DBBasicTestWithTimestamp, SeekWithPrefixLessThanKey) {
   options.env = env_;
   options.create_if_missing = true;
   options.prefix_extractor.reset(NewFixedPrefixTransform(3));
+  options.memtable_whole_key_filtering = true;
+  options.memtable_prefix_bloom_size_ratio = 0.1;
   BlockBasedTableOptions bbto;
   bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
   bbto.cache_index_and_filter_blocks = true;
@@ -265,16 +390,16 @@ TEST_F(DBBasicTestWithTimestamp, SeekWithPrefixLessThanKey) {
   write_opts.timestamp = &ts;
 
   ASSERT_OK(db_->Put(write_opts, "foo1", "bar"));
-  Flush();
+  ASSERT_OK(Flush());
 
   ASSERT_OK(db_->Put(write_opts, "foo2", "bar"));
-  Flush();
+  ASSERT_OK(Flush());
 
   // Move sst file to next level
   ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
 
   ASSERT_OK(db_->Put(write_opts, "foo3", "bar"));
-  Flush();
+  ASSERT_OK(Flush());
 
   ReadOptions read_opts;
   std::string read_ts = Timestamp(1, 0);
@@ -285,6 +410,10 @@ TEST_F(DBBasicTestWithTimestamp, SeekWithPrefixLessThanKey) {
     iter->Seek("foo");
     ASSERT_TRUE(iter->Valid());
     ASSERT_OK(iter->status());
+    iter->Next();
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_OK(iter->status());
+
     iter->Seek("bbb");
     ASSERT_FALSE(iter->Valid());
     ASSERT_OK(iter->status());
@@ -298,6 +427,8 @@ TEST_F(DBBasicTestWithTimestamp, SeekWithPrefixLargerThanKey) {
   options.env = env_;
   options.create_if_missing = true;
   options.prefix_extractor.reset(NewFixedPrefixTransform(20));
+  options.memtable_whole_key_filtering = true;
+  options.memtable_prefix_bloom_size_ratio = 0.1;
   BlockBasedTableOptions bbto;
   bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
   bbto.cache_index_and_filter_blocks = true;
@@ -314,21 +445,76 @@ TEST_F(DBBasicTestWithTimestamp, SeekWithPrefixLargerThanKey) {
   write_opts.timestamp = &ts;
 
   ASSERT_OK(db_->Put(write_opts, "foo1", "bar"));
-  Flush();
+  ASSERT_OK(Flush());
 
   ASSERT_OK(db_->Put(write_opts, "foo2", "bar"));
-  Flush();
+  ASSERT_OK(Flush());
 
   // Move sst file to next level
   ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
 
   ASSERT_OK(db_->Put(write_opts, "foo3", "bar"));
-  Flush();
+  ASSERT_OK(Flush());
 
   ReadOptions read_opts;
   std::string read_ts = Timestamp(2, 0);
   ts = read_ts;
   read_opts.timestamp = &ts;
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_opts));
+    // Make sure the prefix extractor doesn't include timestamp, otherwise it
+    // may return invalid result.
+    iter->Seek("foo");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_OK(iter->status());
+    iter->Next();
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_OK(iter->status());
+  }
+
+  Close();
+}
+
+TEST_F(DBBasicTestWithTimestamp, SeekWithBound) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.prefix_extractor.reset(NewFixedPrefixTransform(2));
+  BlockBasedTableOptions bbto;
+  bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  bbto.cache_index_and_filter_blocks = true;
+  bbto.whole_key_filtering = true;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+  DestroyAndReopen(options);
+
+  WriteOptions write_opts;
+  std::string ts_str = Timestamp(1, 0);
+  Slice ts = ts_str;
+  write_opts.timestamp = &ts;
+
+  ASSERT_OK(db_->Put(write_opts, "foo1", "bar"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(db_->Put(write_opts, "foo2", "bar"));
+  ASSERT_OK(Flush());
+
+  // Move sst file to next level
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+
+  ASSERT_OK(db_->Put(write_opts, "foo3", "bar"));
+  ASSERT_OK(Flush());
+
+  ReadOptions read_opts;
+  std::string read_ts = Timestamp(2, 0);
+  ts = read_ts;
+  read_opts.timestamp = &ts;
+  std::string up_bound = "foo5";
+  Slice up_bound_slice = up_bound;
+  read_opts.iterate_upper_bound = &up_bound_slice;
+  read_opts.auto_prefix_mode = true;
   {
     std::unique_ptr<Iterator> iter(db_->NewIterator(read_opts));
     // Make sure the prefix extractor doesn't include timestamp, otherwise it
@@ -611,7 +797,7 @@ TEST_F(DBBasicTestWithTimestamp, MultiGetWithFastLocalBloom) {
 
   ASSERT_OK(db_->Put(write_opts, "foo", "bar"));
 
-  Flush();
+  ASSERT_OK(Flush());
 
   // Read with MultiGet
   ReadOptions read_opts;
@@ -652,7 +838,7 @@ TEST_F(DBBasicTestWithTimestamp, MultiGetWithPrefix) {
 
   ASSERT_OK(db_->Put(write_opts, "foo", "bar"));
 
-  Flush();
+  ASSERT_OK(Flush());
 
   // Read with MultiGet
   ReadOptions read_opts;
@@ -740,7 +926,7 @@ TEST_F(DBBasicTestWithTimestamp, MultiGetRangeFiltering) {
     Slice key_slice = key;
     Slice value_slice = value;
     ASSERT_OK(db_->Put(write_opts, key_slice, value_slice));
-    Flush();
+    ASSERT_OK(Flush());
   }
 
   // Make num_levels to 2 to do key range filtering of sst files
@@ -748,7 +934,7 @@ TEST_F(DBBasicTestWithTimestamp, MultiGetRangeFiltering) {
 
   ASSERT_OK(db_->Put(write_opts, "foo", "bar"));
 
-  Flush();
+  ASSERT_OK(Flush());
 
   // Read with MultiGet
   ts_str = Timestamp(2, 0);
@@ -791,7 +977,7 @@ TEST_F(DBBasicTestWithTimestamp, MultiGetPrefixFilter) {
 
   ASSERT_OK(db_->Put(write_opts, "foo", "bar"));
 
-  Flush();
+  ASSERT_OK(Flush());
   // Read with MultiGet
   ts_str = Timestamp(2, 0);
   ts = ts_str;
@@ -909,6 +1095,131 @@ TEST_F(DBBasicTestWithTimestamp, CompactDeletionWithTimestampMarkerToBottom) {
   Close();
 }
 
+class DBBasicTestWithTimestampFilterPrefixSettings
+    : public DBBasicTestWithTimestampBase,
+      public testing::WithParamInterface<
+          std::tuple<std::shared_ptr<const FilterPolicy>, bool, bool,
+                     std::shared_ptr<const SliceTransform>, bool, double>> {
+ public:
+  DBBasicTestWithTimestampFilterPrefixSettings()
+      : DBBasicTestWithTimestampBase(
+            "db_basic_test_with_timestamp_filter_prefix") {}
+};
+
+TEST_P(DBBasicTestWithTimestampFilterPrefixSettings, GetAndMultiGet) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  BlockBasedTableOptions bbto;
+  bbto.filter_policy = std::get<0>(GetParam());
+  bbto.whole_key_filtering = std::get<1>(GetParam());
+  bbto.cache_index_and_filter_blocks = std::get<2>(GetParam());
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  options.prefix_extractor = std::get<3>(GetParam());
+  options.memtable_whole_key_filtering = std::get<4>(GetParam());
+  options.memtable_prefix_bloom_size_ratio = std::get<5>(GetParam());
+
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+  DestroyAndReopen(options);
+  const int kMaxKey = 1000;
+
+  // Write any value
+  WriteOptions write_opts;
+  std::string ts_str = Timestamp(1, 0);
+  Slice ts = ts_str;
+  write_opts.timestamp = &ts;
+
+  int idx = 0;
+  for (; idx < kMaxKey / 4; idx++) {
+    ASSERT_OK(db_->Put(write_opts, Key1(idx), "bar"));
+    ASSERT_OK(db_->Put(write_opts, KeyWithPrefix("foo", idx), "bar"));
+  }
+
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+
+  for (; idx < kMaxKey / 2; idx++) {
+    ASSERT_OK(db_->Put(write_opts, Key1(idx), "bar"));
+    ASSERT_OK(db_->Put(write_opts, KeyWithPrefix("foo", idx), "bar"));
+  }
+
+  ASSERT_OK(Flush());
+
+  for (; idx < kMaxKey; idx++) {
+    ASSERT_OK(db_->Put(write_opts, Key1(idx), "bar"));
+    ASSERT_OK(db_->Put(write_opts, KeyWithPrefix("foo", idx), "bar"));
+  }
+
+  // Read with MultiGet
+  ReadOptions read_opts;
+  read_opts.timestamp = &ts;
+
+  ReadOptions read_opts_total_order;
+  read_opts_total_order.timestamp = &ts;
+  read_opts_total_order.total_order_seek = true;
+
+  for (idx = 0; idx < kMaxKey; idx++) {
+    size_t batch_size = 4;
+    std::vector<std::string> keys_str(batch_size);
+    std::vector<PinnableSlice> values(batch_size);
+    std::vector<Status> statuses(batch_size);
+    ColumnFamilyHandle* cfh = db_->DefaultColumnFamily();
+
+    keys_str[0] = Key1(idx);
+    keys_str[1] = KeyWithPrefix("foo", idx);
+    keys_str[2] = Key1(kMaxKey + idx);
+    keys_str[3] = KeyWithPrefix("foo", kMaxKey + idx);
+
+    auto keys = ConvertStrToSlice(keys_str);
+
+    db_->MultiGet(read_opts, cfh, batch_size, keys.data(), values.data(),
+                  statuses.data());
+
+    for (int i = 0; i < 2; i++) {
+      ASSERT_OK(statuses[i]);
+    }
+    for (int i = 2; i < 4; i++) {
+      ASSERT_TRUE(statuses[i].IsNotFound());
+    }
+
+    for (int i = 0; i < 2; i++) {
+      std::string value;
+      ASSERT_OK(db_->Get(read_opts, keys[i], &value));
+      std::unique_ptr<Iterator> it1(db_->NewIterator(read_opts));
+      ASSERT_NE(nullptr, it1);
+      ASSERT_OK(it1->status());
+      // TODO(zjay) Fix seek with prefix
+      // it1->Seek(keys[i]);
+      // ASSERT_TRUE(it1->Valid());
+    }
+
+    for (int i = 2; i < 4; i++) {
+      std::string value;
+      Status s = db_->Get(read_opts, keys[i], &value);
+      ASSERT_TRUE(s.IsNotFound());
+    }
+  }
+  Close();
+}
+
+INSTANTIATE_TEST_CASE_P(
+    Timestamp, DBBasicTestWithTimestampFilterPrefixSettings,
+    ::testing::Combine(
+        ::testing::Values(
+            std::shared_ptr<const FilterPolicy>(nullptr),
+            std::shared_ptr<const FilterPolicy>(NewBloomFilterPolicy(10, true)),
+            std::shared_ptr<const FilterPolicy>(NewBloomFilterPolicy(10,
+                                                                     false))),
+        ::testing::Bool(), ::testing::Bool(),
+        ::testing::Values(
+            std::shared_ptr<const SliceTransform>(NewFixedPrefixTransform(1)),
+            std::shared_ptr<const SliceTransform>(NewFixedPrefixTransform(4)),
+            std::shared_ptr<const SliceTransform>(NewFixedPrefixTransform(7)),
+            std::shared_ptr<const SliceTransform>(NewFixedPrefixTransform(8))),
+        ::testing::Bool(), ::testing::Values(0, 0.1)));
+
 class DataVisibilityTest : public DBBasicTestWithTimestampBase {
  public:
   DataVisibilityTest() : DBBasicTestWithTimestampBase("data_visibility_test") {
@@ -1372,7 +1683,7 @@ TEST_F(DataVisibilityTest, MultiGetWithTimestamp) {
   VerifyDefaultCF(snap0);
   VerifyDefaultCF(snap1);
 
-  Flush();
+  ASSERT_OK(Flush());
 
   const Snapshot* snap2 = db_->GetSnapshot();
   PutTestData(2);
@@ -1458,7 +1769,7 @@ TEST_F(DataVisibilityTest, MultiGetCrossCF) {
   VerifyDefaultCF(snap0);
   VerifyDefaultCF(snap1);
 
-  Flush();
+  ASSERT_OK(Flush());
 
   const Snapshot* snap2 = db_->GetSnapshot();
   PutTestData(2);
@@ -1839,6 +2150,8 @@ TEST_F(DBBasicTestWithTimestamp, BatchWriteAndMultiGet) {
   options.create_if_missing = true;
   options.env = env_;
   options.memtable_factory.reset(new SpecialSkipListFactory(kNumKeysPerFile));
+  options.memtable_prefix_bloom_size_ratio = 0.1;
+  options.memtable_whole_key_filtering = true;
 
   size_t ts_sz = Timestamp(0, 0).size();
   TestComparator test_cmp(ts_sz);
@@ -2093,6 +2406,7 @@ INSTANTIATE_TEST_CASE_P(
     Timestamp, DBBasicTestWithTimestampPrefixSeek,
     ::testing::Combine(
         ::testing::Values(
+            std::shared_ptr<const SliceTransform>(NewFixedPrefixTransform(1)),
             std::shared_ptr<const SliceTransform>(NewFixedPrefixTransform(4)),
             std::shared_ptr<const SliceTransform>(NewFixedPrefixTransform(7)),
             std::shared_ptr<const SliceTransform>(NewFixedPrefixTransform(8))),

db/db_write_test.cc

@@ -320,7 +320,7 @@ TEST_P(DBWriteTest, ManualWalFlushInEffect) {
   ASSERT_TRUE(dbfull()->FlushWAL(false).ok());
   ASSERT_TRUE(dbfull()->TEST_WALBufferIsEmpty());
   // try the 2nd wal created during SwitchWAL
-  dbfull()->TEST_SwitchWAL();
+  ASSERT_OK(dbfull()->TEST_SwitchWAL());
   ASSERT_TRUE(Put("key" + ToString(0), "value").ok());
   ASSERT_TRUE(options.manual_wal_flush != dbfull()->TEST_WALBufferIsEmpty());
   ASSERT_TRUE(dbfull()->FlushWAL(false).ok());
@@ -395,7 +395,7 @@ TEST_P(DBWriteTest, LockWalInEffect) {
   ASSERT_TRUE(dbfull()->TEST_WALBufferIsEmpty(false));
   ASSERT_OK(dbfull()->UnlockWAL());
   // try the 2nd wal created during SwitchWAL
-  dbfull()->TEST_SwitchWAL();
+  ASSERT_OK(dbfull()->TEST_SwitchWAL());
   ASSERT_OK(Put("key" + ToString(0), "value"));
   ASSERT_TRUE(options.manual_wal_flush != dbfull()->TEST_WALBufferIsEmpty());
   ASSERT_OK(dbfull()->LockWAL());

db/dbformat.cc

@@ -78,6 +78,22 @@ void AppendInternalKeyFooter(std::string* result, SequenceNumber s,
   PutFixed64(result, PackSequenceAndType(s, t));
 }
 
+void AppendKeyWithMinTimestamp(std::string* result, const Slice& key,
+                               size_t ts_sz) {
+  assert(ts_sz > 0);
+  const std::string kTsMin(ts_sz, static_cast<unsigned char>(0));
+  result->append(key.data(), key.size());
+  result->append(kTsMin.data(), ts_sz);
+}
+
+void AppendKeyWithMaxTimestamp(std::string* result, const Slice& key,
+                               size_t ts_sz) {
+  assert(ts_sz > 0);
+  const std::string kTsMax(ts_sz, static_cast<unsigned char>(0xff));
+  result->append(key.data(), key.size());
+  result->append(kTsMax.data(), ts_sz);
+}
+
 std::string ParsedInternalKey::DebugString(bool log_err_key, bool hex) const {
   std::string result = "'";
   if (log_err_key) {

db/dbformat.h

@@ -167,6 +167,14 @@ extern void AppendInternalKeyWithDifferentTimestamp(
 extern void AppendInternalKeyFooter(std::string* result, SequenceNumber s,
                                     ValueType t);
 
+// Append the key and a minimal timestamp to *result
+extern void AppendKeyWithMinTimestamp(std::string* result, const Slice& key,
+                                      size_t ts_sz);
+
+// Append the key and a maximal timestamp to *result
+extern void AppendKeyWithMaxTimestamp(std::string* result, const Slice& key,
+                                      size_t ts_sz);
+
 // Attempt to parse an internal key from "internal_key".  On success,
 // stores the parsed data in "*result", and returns true.
 //

db/error_handler.cc

@@ -111,6 +111,23 @@ std::map<std::tuple<BackgroundErrorReason, Status::Code, Status::SubCode, bool>,
                          Status::Code::kIOError, Status::SubCode::kIOFenced,
                          false),
          Status::Severity::kFatalError},
+        // Errors during MANIFEST write when WAL is disabled
+        {std::make_tuple(BackgroundErrorReason::kManifestWriteNoWAL,
+                         Status::Code::kIOError, Status::SubCode::kNoSpace,
+                         true),
+         Status::Severity::kHardError},
+        {std::make_tuple(BackgroundErrorReason::kManifestWriteNoWAL,
+                         Status::Code::kIOError, Status::SubCode::kNoSpace,
+                         false),
+         Status::Severity::kHardError},
+        {std::make_tuple(BackgroundErrorReason::kManifestWriteNoWAL,
+                         Status::Code::kIOError, Status::SubCode::kIOFenced,
+                         true),
+         Status::Severity::kFatalError},
+        {std::make_tuple(BackgroundErrorReason::kManifestWriteNoWAL,
+                         Status::Code::kIOError, Status::SubCode::kIOFenced,
+                         false),
+         Status::Severity::kFatalError},
 
 };
 
@@ -175,6 +192,12 @@ std::map<std::tuple<BackgroundErrorReason, Status::Code, bool>,
         {std::make_tuple(BackgroundErrorReason::kFlushNoWAL,
                          Status::Code::kIOError, false),
          Status::Severity::kNoError},
+        {std::make_tuple(BackgroundErrorReason::kManifestWriteNoWAL,
+                         Status::Code::kIOError, true),
+         Status::Severity::kFatalError},
+        {std::make_tuple(BackgroundErrorReason::kManifestWriteNoWAL,
+                         Status::Code::kIOError, false),
+         Status::Severity::kFatalError},
 };
 
 std::map<std::tuple<BackgroundErrorReason, bool>, Status::Severity>
@@ -244,10 +267,11 @@ void ErrorHandler::CancelErrorRecovery() {
 // This can also get called as part of a recovery operation. In that case, we
 // also track the error separately in recovery_error_ so we can tell in the
 // end whether recovery succeeded or not
-Status ErrorHandler::SetBGError(const Status& bg_err, BackgroundErrorReason reason) {
+const Status& ErrorHandler::SetBGError(const Status& bg_err,
+                                       BackgroundErrorReason reason) {
   db_mutex_->AssertHeld();
   if (bg_err.ok()) {
-    return Status::OK();
+    return bg_err;
   }
 
   bool paranoid = db_options_.paranoid_checks;
@@ -324,11 +348,32 @@ Status ErrorHandler::SetBGError(const Status& bg_err, BackgroundErrorReason reas
   return bg_error_;
 }
 
-Status ErrorHandler::SetBGError(const IOStatus& bg_io_err,
-                                BackgroundErrorReason reason) {
+// This is the main function for looking at IO related error during the
+// background operations. The main logic is:
+// 1) File scope IO error is treated as retryable IO error in the write
+//    path. In RocksDB, If a file has write IO error and it is at file scope,
+//    RocksDB never write to the same file again. RocksDB will create a new
+//    file and rewrite the whole content. Thus, it is retryable.
+// 1) if the error is caused by data loss, the error is mapped to
+//    unrecoverable error. Application/user must take action to handle
+//    this situation (File scope case is excluded).
+// 2) if the error is a Retryable IO error (i.e., it is a file scope IO error,
+//     or its retryable flag is set and not a data loss error), auto resume
+//     will be called and the auto resume can be controlled by resume count
+//     and resume interval options. There are three sub-cases:
+//    a) if the error happens during compaction, it is mapped to a soft error.
+//       the compaction thread will reschedule a new compaction.
+//    b) if the error happens during flush and also WAL is empty, it is mapped
+//       to a soft error. Note that, it includes the case that IO error happens
+//       in SST or manifest write during flush.
+//    c) all other errors are mapped to hard error.
+// 3) for other cases, SetBGError(const Status& bg_err, BackgroundErrorReason
+//    reason) will be called to handle other error cases.
+const Status& ErrorHandler::SetBGError(const IOStatus& bg_io_err,
+                                       BackgroundErrorReason reason) {
   db_mutex_->AssertHeld();
   if (bg_io_err.ok()) {
-    return Status::OK();
+    return bg_io_err;
   }
   ROCKS_LOG_WARN(db_options_.info_log, "Background IO error %s",
                  bg_io_err.ToString().c_str());
@@ -336,37 +381,41 @@ Status ErrorHandler::SetBGError(const IOStatus& bg_io_err,
   if (recovery_in_prog_ && recovery_io_error_.ok()) {
     recovery_io_error_ = bg_io_err;
   }
-  if (BackgroundErrorReason::kManifestWrite == reason) {
+  if (BackgroundErrorReason::kManifestWrite == reason ||
+      BackgroundErrorReason::kManifestWriteNoWAL == reason) {
     // Always returns ok
     db_->DisableFileDeletionsWithLock().PermitUncheckedError();
   }
 
   Status new_bg_io_err = bg_io_err;
-  Status s;
   DBRecoverContext context;
-  if (bg_io_err.GetDataLoss()) {
-    // First, data loss is treated as unrecoverable error. So it can directly
-    // overwrite any existing bg_error_.
+  if (bg_io_err.GetScope() != IOStatus::IOErrorScope::kIOErrorScopeFile &&
+      bg_io_err.GetDataLoss()) {
+    // First, data loss (non file scope) is treated as unrecoverable error. So
+    // it can directly overwrite any existing bg_error_.
     bool auto_recovery = false;
     Status bg_err(new_bg_io_err, Status::Severity::kUnrecoverableError);
     bg_error_ = bg_err;
     if (recovery_in_prog_ && recovery_error_.ok()) {
       recovery_error_ = bg_err;
     }
-    EventHelpers::NotifyOnBackgroundError(db_options_.listeners, reason, &s,
-                                          db_mutex_, &auto_recovery);
+    EventHelpers::NotifyOnBackgroundError(db_options_.listeners, reason,
+                                          &bg_err, db_mutex_, &auto_recovery);
     recover_context_ = context;
     return bg_error_;
-  } else if (bg_io_err.GetRetryable()) {
-    // Second, check if the error is a retryable IO error or not. if it is
-    // retryable error and its severity is higher than bg_error_, overwrite
-    // the bg_error_ with new error.
-    // In current stage, for retryable IO error of compaction, treat it as
-    // soft error. In other cases, treat the retryable IO error as hard
-    // error.
+  } else if (bg_io_err.GetScope() ==
+                 IOStatus::IOErrorScope::kIOErrorScopeFile ||
+             bg_io_err.GetRetryable()) {
+    // Second, check if the error is a retryable IO error (file scope IO error
+    // is also treated as retryable IO error in RocksDB write path). if it is
+    // retryable error and its severity is higher than bg_error_, overwrite the
+    // bg_error_ with new error. In current stage, for retryable IO error of
+    // compaction, treat it as soft error. In other cases, treat the retryable
+    // IO error as hard error.
     bool auto_recovery = false;
-    EventHelpers::NotifyOnBackgroundError(db_options_.listeners, reason, &s,
-                                          db_mutex_, &auto_recovery);
+    EventHelpers::NotifyOnBackgroundError(db_options_.listeners, reason,
+                                          &new_bg_io_err, db_mutex_,
+                                          &auto_recovery);
     if (BackgroundErrorReason::kCompaction == reason) {
       Status bg_err(new_bg_io_err, Status::Severity::kSoftError);
       if (bg_err.severity() > bg_error_.severity()) {
@@ -374,7 +423,8 @@ Status ErrorHandler::SetBGError(const IOStatus& bg_io_err,
       }
       recover_context_ = context;
       return bg_error_;
-    } else if (BackgroundErrorReason::kFlushNoWAL == reason) {
+    } else if (BackgroundErrorReason::kFlushNoWAL == reason ||
+               BackgroundErrorReason::kManifestWriteNoWAL == reason) {
       // When the BG Retryable IO error reason is flush without WAL,
       // We map it to a soft error. At the same time, all the background work
       // should be stopped except the BG work from recovery. Therefore, we
@@ -405,12 +455,11 @@ Status ErrorHandler::SetBGError(const IOStatus& bg_io_err,
       return StartRecoverFromRetryableBGIOError(bg_io_err);
     }
   } else {
-    s = SetBGError(new_bg_io_err, reason);
+    return SetBGError(new_bg_io_err, reason);
   }
-  return s;
 }
 
-Status ErrorHandler::OverrideNoSpaceError(Status bg_error,
+Status ErrorHandler::OverrideNoSpaceError(const Status& bg_error,
                                           bool* auto_recovery) {
 #ifndef ROCKSDB_LITE
   if (bg_error.severity() >= Status::Severity::kFatalError) {
@@ -466,7 +515,11 @@ Status ErrorHandler::ClearBGError() {
   // Signal that recovery succeeded
   if (recovery_error_.ok()) {
     Status old_bg_error = bg_error_;
+    // Clear and check the recovery IO and BG error
     bg_error_ = Status::OK();
+    recovery_io_error_ = IOStatus::OK();
+    bg_error_.PermitUncheckedError();
+    recovery_io_error_.PermitUncheckedError();
     recovery_in_prog_ = false;
     soft_error_no_bg_work_ = false;
     EventHelpers::NotifyOnErrorRecoveryCompleted(db_options_.listeners,
@@ -516,6 +569,7 @@ Status ErrorHandler::RecoverFromBGError(bool is_manual) {
   // 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_.PermitUncheckedError();
   Status s = db_->ResumeImpl(recover_context_);
   if (s.ok()) {
     soft_error_no_bg_work_ = false;
@@ -537,24 +591,34 @@ Status ErrorHandler::RecoverFromBGError(bool is_manual) {
 #endif
 }
 
-Status ErrorHandler::StartRecoverFromRetryableBGIOError(IOStatus io_error) {
+const Status& ErrorHandler::StartRecoverFromRetryableBGIOError(
+    const IOStatus& io_error) {
 #ifndef ROCKSDB_LITE
   db_mutex_->AssertHeld();
-  if (bg_error_.ok() || io_error.ok()) {
-    return Status::OK();
-  }
-  if (db_options_.max_bgerror_resume_count <= 0 || recovery_in_prog_ ||
-      recovery_thread_) {
+  if (bg_error_.ok()) {
+    return bg_error_;
+  } else if (io_error.ok()) {
+    return io_error;
+  } 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_;
   }
 
+  if (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();
+    db_mutex_->Lock();
+  }
+
   recovery_in_prog_ = true;
   recovery_thread_.reset(
       new port::Thread(&ErrorHandler::RecoverFromRetryableBGIOError, this));
 
   if (recovery_io_error_.ok() && recovery_error_.ok()) {
-    return Status::OK();
+    return recovery_error_;
   } else {
     TEST_SYNC_POINT("StartRecoverRetryableBGIOError:RecoverFail");
     return bg_error_;
@@ -619,6 +683,7 @@ void ErrorHandler::RecoverFromRetryableBGIOError() {
         TEST_SYNC_POINT("RecoverFromRetryableBGIOError:RecoverSuccess");
         Status old_bg_error = bg_error_;
         bg_error_ = Status::OK();
+        bg_error_.PermitUncheckedError();
         EventHelpers::NotifyOnErrorRecoveryCompleted(db_options_.listeners,
                                                      old_bg_error, db_mutex_);
         recovery_in_prog_ = false;

db/error_handler.h

@@ -31,17 +31,14 @@ class ErrorHandler {
                 InstrumentedMutex* db_mutex)
        : db_(db),
          db_options_(db_options),
-         bg_error_(Status::OK()),
-         recovery_error_(Status::OK()),
-         recovery_io_error_(IOStatus::OK()),
          cv_(db_mutex),
          end_recovery_(false),
          recovery_thread_(nullptr),
          db_mutex_(db_mutex),
          auto_recovery_(false),
          recovery_in_prog_(false),
-         soft_error_no_bg_work_(false) {}
-   ~ErrorHandler() {
+         soft_error_no_bg_work_(false) {
+     // Clear the checked flag for uninitialized errors
      bg_error_.PermitUncheckedError();
      recovery_error_.PermitUncheckedError();
      recovery_io_error_.PermitUncheckedError();
@@ -53,13 +50,14 @@ class ErrorHandler {
                                      Status::Code code,
                                      Status::SubCode subcode);
 
-   Status SetBGError(const Status& bg_err, BackgroundErrorReason reason);
+   const Status& SetBGError(const Status& bg_err, BackgroundErrorReason reason);
 
-   Status SetBGError(const IOStatus& bg_io_err, BackgroundErrorReason reason);
+   const Status& SetBGError(const IOStatus& bg_io_err,
+                            BackgroundErrorReason reason);
 
-   Status GetBGError() { return bg_error_; }
+   Status GetBGError() const { return bg_error_; }
 
-   Status GetRecoveryError() { return recovery_error_; }
+   Status GetRecoveryError() const { return recovery_error_; }
 
    Status ClearBGError();
 
@@ -110,9 +108,9 @@ class ErrorHandler {
     // Used to store the context for recover, such as flush reason.
     DBRecoverContext recover_context_;
 
-    Status OverrideNoSpaceError(Status bg_error, bool* auto_recovery);
+    Status OverrideNoSpaceError(const Status& bg_error, bool* auto_recovery);
     void RecoverFromNoSpace();
-    Status StartRecoverFromRetryableBGIOError(IOStatus io_error);
+    const Status& StartRecoverFromRetryableBGIOError(const IOStatus& io_error);
     void RecoverFromRetryableBGIOError();
 };
 

db/error_handler_fs_test.cc

@@ -241,6 +241,90 @@ TEST_F(DBErrorHandlingFSTest, FLushWritRetryableError) {
   Destroy(options);
 }
 
+TEST_F(DBErrorHandlingFSTest, FLushWritFileScopeError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener(
+      new ErrorHandlerFSListener());
+  Options options = GetDefaultOptions();
+  options.env = fault_env_.get();
+  options.create_if_missing = true;
+  options.listeners.emplace_back(listener);
+  options.max_bgerror_resume_count = 0;
+  Status s;
+
+  listener->EnableAutoRecovery(false);
+  DestroyAndReopen(options);
+
+  IOStatus error_msg = IOStatus::IOError("File Scope Data Loss Error");
+  error_msg.SetDataLoss(true);
+  error_msg.SetScope(
+      ROCKSDB_NAMESPACE::IOStatus::IOErrorScope::kIOErrorScopeFile);
+  error_msg.SetRetryable(false);
+
+  ASSERT_OK(Put(Key(1), "val1"));
+  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::kHardError);
+  SyncPoint::GetInstance()->DisableProcessing();
+  fault_fs_->SetFilesystemActive(true);
+  s = dbfull()->Resume();
+  ASSERT_OK(s);
+  Reopen(options);
+  ASSERT_EQ("val1", Get(Key(1)));
+
+  ASSERT_OK(Put(Key(2), "val2"));
+  SyncPoint::GetInstance()->SetCallBack(
+      "BuildTable:BeforeSyncTable",
+      [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
+  SyncPoint::GetInstance()->EnableProcessing();
+  s = Flush();
+  ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
+  SyncPoint::GetInstance()->DisableProcessing();
+  fault_fs_->SetFilesystemActive(true);
+  s = dbfull()->Resume();
+  ASSERT_OK(s);
+  Reopen(options);
+  ASSERT_EQ("val2", Get(Key(2)));
+
+  ASSERT_OK(Put(Key(3), "val3"));
+  SyncPoint::GetInstance()->SetCallBack(
+      "BuildTable:BeforeCloseTableFile",
+      [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
+  SyncPoint::GetInstance()->EnableProcessing();
+  s = Flush();
+  ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
+  SyncPoint::GetInstance()->DisableProcessing();
+  fault_fs_->SetFilesystemActive(true);
+  s = dbfull()->Resume();
+  ASSERT_OK(s);
+  Reopen(options);
+  ASSERT_EQ("val3", Get(Key(3)));
+
+  // not file scope, but retyrable set
+  error_msg.SetDataLoss(false);
+  error_msg.SetScope(
+      ROCKSDB_NAMESPACE::IOStatus::IOErrorScope::kIOErrorScopeFileSystem);
+  error_msg.SetRetryable(true);
+
+  ASSERT_OK(Put(Key(3), "val3"));
+  SyncPoint::GetInstance()->SetCallBack(
+      "BuildTable:BeforeCloseTableFile",
+      [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
+  SyncPoint::GetInstance()->EnableProcessing();
+  s = Flush();
+  ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
+  SyncPoint::GetInstance()->DisableProcessing();
+  fault_fs_->SetFilesystemActive(true);
+  s = dbfull()->Resume();
+  ASSERT_OK(s);
+  Reopen(options);
+  ASSERT_EQ("val3", Get(Key(3)));
+
+  Destroy(options);
+}
+
 TEST_F(DBErrorHandlingFSTest, FLushWritNoWALRetryableError1) {
   std::shared_ptr<ErrorHandlerFSListener> listener(
       new ErrorHandlerFSListener());
@@ -453,6 +537,97 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableError) {
   Close();
 }
 
+TEST_F(DBErrorHandlingFSTest, ManifestWriteFileScopeError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener(
+      new ErrorHandlerFSListener());
+  Options options = GetDefaultOptions();
+  options.env = fault_env_.get();
+  options.create_if_missing = true;
+  options.listeners.emplace_back(listener);
+  options.max_bgerror_resume_count = 0;
+  Status s;
+  std::string old_manifest;
+  std::string new_manifest;
+
+  listener->EnableAutoRecovery(false);
+  DestroyAndReopen(options);
+  old_manifest = GetManifestNameFromLiveFiles();
+
+  IOStatus error_msg = IOStatus::IOError("File Scope Data Loss Error");
+  error_msg.SetDataLoss(true);
+  error_msg.SetScope(
+      ROCKSDB_NAMESPACE::IOStatus::IOErrorScope::kIOErrorScopeFile);
+  error_msg.SetRetryable(false);
+
+  ASSERT_OK(Put(Key(0), "val"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put(Key(1), "val"));
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::LogAndApply:WriteManifest",
+      [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
+  SyncPoint::GetInstance()->EnableProcessing();
+  s = Flush();
+  ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  SyncPoint::GetInstance()->DisableProcessing();
+  fault_fs_->SetFilesystemActive(true);
+  s = dbfull()->Resume();
+  ASSERT_OK(s);
+
+  new_manifest = GetManifestNameFromLiveFiles();
+  ASSERT_NE(new_manifest, old_manifest);
+
+  Reopen(options);
+  ASSERT_EQ("val", Get(Key(0)));
+  ASSERT_EQ("val", Get(Key(1)));
+  Close();
+}
+
+TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener(
+      new ErrorHandlerFSListener());
+  Options options = GetDefaultOptions();
+  options.env = fault_env_.get();
+  options.create_if_missing = true;
+  options.listeners.emplace_back(listener);
+  options.max_bgerror_resume_count = 0;
+  Status s;
+  std::string old_manifest;
+  std::string new_manifest;
+
+  listener->EnableAutoRecovery(false);
+  DestroyAndReopen(options);
+  old_manifest = GetManifestNameFromLiveFiles();
+
+  IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
+  error_msg.SetRetryable(true);
+
+  WriteOptions wo = WriteOptions();
+  wo.disableWAL = true;
+  ASSERT_OK(Put(Key(0), "val", wo));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put(Key(1), "val", wo));
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::LogAndApply:WriteManifest",
+      [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
+  SyncPoint::GetInstance()->EnableProcessing();
+  s = Flush();
+  ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  SyncPoint::GetInstance()->DisableProcessing();
+  fault_fs_->SetFilesystemActive(true);
+  s = dbfull()->Resume();
+  ASSERT_EQ(s, Status::OK());
+
+  new_manifest = GetManifestNameFromLiveFiles();
+  ASSERT_NE(new_manifest, old_manifest);
+
+  Reopen(options);
+  ASSERT_EQ("val", Get(Key(0)));
+  ASSERT_EQ("val", Get(Key(1)));
+  Close();
+}
+
 TEST_F(DBErrorHandlingFSTest, DoubleManifestWriteError) {
   std::shared_ptr<ErrorHandlerFSListener> listener(
       new ErrorHandlerFSListener());
@@ -734,6 +909,54 @@ TEST_F(DBErrorHandlingFSTest, CompactionWriteRetryableError) {
   Destroy(options);
 }
 
+TEST_F(DBErrorHandlingFSTest, CompactionWriteFileScopeError) {
+  std::shared_ptr<ErrorHandlerFSListener> listener(
+      new ErrorHandlerFSListener());
+  Options options = GetDefaultOptions();
+  options.env = fault_env_.get();
+  options.create_if_missing = true;
+  options.level0_file_num_compaction_trigger = 2;
+  options.listeners.emplace_back(listener);
+  options.max_bgerror_resume_count = 0;
+  Status s;
+  DestroyAndReopen(options);
+
+  IOStatus error_msg = IOStatus::IOError("File Scope Data Loss Error");
+  error_msg.SetDataLoss(true);
+  error_msg.SetScope(
+      ROCKSDB_NAMESPACE::IOStatus::IOErrorScope::kIOErrorScopeFile);
+  error_msg.SetRetryable(false);
+
+  ASSERT_OK(Put(Key(0), "va;"));
+  ASSERT_OK(Put(Key(2), "va;"));
+  s = Flush();
+  ASSERT_OK(s);
+
+  listener->OverrideBGError(Status(error_msg, Status::Severity::kHardError));
+  listener->EnableAutoRecovery(false);
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::FlushMemTable:FlushMemTableFinished",
+        "BackgroundCallCompaction:0"}});
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "CompactionJob::OpenCompactionOutputFile",
+      [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put(Key(1), "val"));
+  s = Flush();
+  ASSERT_OK(s);
+
+  s = dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
+
+  fault_fs_->SetFilesystemActive(true);
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  SyncPoint::GetInstance()->DisableProcessing();
+  s = dbfull()->Resume();
+  ASSERT_OK(s);
+  Destroy(options);
+}
+
 TEST_F(DBErrorHandlingFSTest, CorruptionError) {
   Options options = GetDefaultOptions();
   options.env = fault_env_.get();
@@ -1789,6 +2012,61 @@ TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableErrorAutoRecover) {
   Close();
 }
 
+TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableErrorAutoRecover) {
+  // Fail the first resume and let the second resume be successful
+  std::shared_ptr<ErrorHandlerFSListener> listener(
+      new ErrorHandlerFSListener());
+  Options options = GetDefaultOptions();
+  options.env = fault_env_.get();
+  options.create_if_missing = true;
+  options.listeners.emplace_back(listener);
+  options.max_bgerror_resume_count = 2;
+  options.bgerror_resume_retry_interval = 100000;  // 0.1 second
+  Status s;
+  std::string old_manifest;
+  std::string new_manifest;
+
+  listener->EnableAutoRecovery(false);
+  DestroyAndReopen(options);
+  old_manifest = GetManifestNameFromLiveFiles();
+
+  IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
+  error_msg.SetRetryable(true);
+
+  WriteOptions wo = WriteOptions();
+  wo.disableWAL = true;
+  ASSERT_OK(Put(Key(0), "val", wo));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put(Key(1), "val", wo));
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+      {{"RecoverFromRetryableBGIOError:BeforeStart",
+        "ManifestWriteNoWALRetryableErrorAutoRecover:0"},
+       {"ManifestWriteNoWALRetryableErrorAutoRecover:1",
+        "RecoverFromRetryableBGIOError:BeforeWait1"},
+       {"RecoverFromRetryableBGIOError:RecoverSuccess",
+        "ManifestWriteNoWALRetryableErrorAutoRecover:2"}});
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::LogAndApply:WriteManifest",
+      [&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
+  SyncPoint::GetInstance()->EnableProcessing();
+  s = Flush();
+  ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
+  TEST_SYNC_POINT("ManifestWriteNoWALRetryableErrorAutoRecover:0");
+  fault_fs_->SetFilesystemActive(true);
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+  TEST_SYNC_POINT("ManifestWriteNoWALRetryableErrorAutoRecover:1");
+  TEST_SYNC_POINT("ManifestWriteNoWALRetryableErrorAutoRecover:2");
+  SyncPoint::GetInstance()->DisableProcessing();
+
+  new_manifest = GetManifestNameFromLiveFiles();
+  ASSERT_NE(new_manifest, old_manifest);
+
+  Reopen(options);
+  ASSERT_EQ("val", Get(Key(0)));
+  ASSERT_EQ("val", Get(Key(1)));
+  Close();
+}
+
 TEST_F(DBErrorHandlingFSTest,
        CompactionManifestWriteRetryableErrorAutoRecover) {
   std::shared_ptr<ErrorHandlerFSListener> listener(

db/external_sst_file_basic_test.cc

@@ -29,8 +29,8 @@ class ExternalSSTFileBasicTest
   }
 
   void DestroyAndRecreateExternalSSTFilesDir() {
-    DestroyDir(env_, sst_files_dir_);
-    env_->CreateDir(sst_files_dir_);
+    ASSERT_OK(DestroyDir(env_, sst_files_dir_));
+    ASSERT_OK(env_->CreateDir(sst_files_dir_));
   }
 
   Status DeprecatedAddFile(const std::vector<std::string>& files,
@@ -162,7 +162,9 @@ class ExternalSSTFileBasicTest
         write_global_seqno, verify_checksums_before_ingest, true_data);
   }
 
-  ~ExternalSSTFileBasicTest() override { DestroyDir(env_, sst_files_dir_); }
+  ~ExternalSSTFileBasicTest() override {
+    DestroyDir(env_, sst_files_dir_).PermitUncheckedError();
+  }
 
  protected:
   std::string sst_files_dir_;
@@ -186,7 +188,7 @@ TEST_F(ExternalSSTFileBasicTest, Basic) {
   }
   ExternalSstFileInfo file1_info;
   Status s = sst_file_writer.Finish(&file1_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
 
   // Current file size should be non-zero after success write.
   ASSERT_GT(sst_file_writer.FileSize(), 0);
@@ -202,14 +204,14 @@ TEST_F(ExternalSSTFileBasicTest, Basic) {
   ASSERT_EQ(file1_info.file_checksum_func_name, kUnknownFileChecksumFuncName);
   // sst_file_writer already finished, cannot add this value
   s = sst_file_writer.Put(Key(100), "bad_val");
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
   s = sst_file_writer.DeleteRange(Key(100), Key(200));
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
 
   DestroyAndReopen(options);
   // Add file using file path
   s = DeprecatedAddFile({file1});
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(db_->GetLatestSequenceNumber(), 0U);
   for (int k = 0; k < 100; k++) {
     ASSERT_EQ(Get(Key(k)), Key(k) + "_val");
@@ -286,7 +288,7 @@ TEST_F(ExternalSSTFileBasicTest, BasicWithFileChecksumCrc32c) {
   }
   ExternalSstFileInfo file1_info;
   Status s = sst_file_writer.Finish(&file1_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   std::string file_checksum, file_checksum_func_name;
   ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
       file1, &file_checksum, &file_checksum_func_name));
@@ -305,14 +307,14 @@ TEST_F(ExternalSSTFileBasicTest, BasicWithFileChecksumCrc32c) {
   ASSERT_EQ(file1_info.file_checksum_func_name, file_checksum_func_name);
   // sst_file_writer already finished, cannot add this value
   s = sst_file_writer.Put(Key(100), "bad_val");
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
   s = sst_file_writer.DeleteRange(Key(100), Key(200));
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
 
   DestroyAndReopen(options);
   // Add file using file path
   s = DeprecatedAddFile({file1});
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(db_->GetLatestSequenceNumber(), 0U);
   for (int k = 0; k < 100; k++) {
     ASSERT_EQ(Get(Key(k)), Key(k) + "_val");
@@ -338,7 +340,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   }
   ExternalSstFileInfo file1_info;
   Status s = sst_file_writer.Finish(&file1_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file1_info.file_path, file1);
   ASSERT_EQ(file1_info.num_entries, 100);
   ASSERT_EQ(file1_info.smallest_key, Key(1000));
@@ -357,7 +359,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   }
   ExternalSstFileInfo file2_info;
   s = sst_file_writer.Finish(&file2_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file2_info.file_path, file2);
   ASSERT_EQ(file2_info.num_entries, 200);
   ASSERT_EQ(file2_info.smallest_key, Key(1100));
@@ -376,7 +378,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   }
   ExternalSstFileInfo file3_info;
   s = sst_file_writer.Finish(&file3_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file3_info.file_path, file3);
   ASSERT_EQ(file3_info.num_entries, 200);
   ASSERT_EQ(file3_info.smallest_key, Key(1300));
@@ -395,7 +397,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   }
   ExternalSstFileInfo file4_info;
   s = sst_file_writer.Finish(&file4_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file4_info.file_path, file4);
   ASSERT_EQ(file4_info.num_entries, 300);
   ASSERT_EQ(file4_info.smallest_key, Key(1500));
@@ -414,7 +416,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   }
   ExternalSstFileInfo file5_info;
   s = sst_file_writer.Finish(&file5_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file5_info.file_path, file5);
   ASSERT_EQ(file5_info.num_entries, 200);
   ASSERT_EQ(file5_info.smallest_key, Key(1800));
@@ -433,7 +435,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   }
   ExternalSstFileInfo file6_info;
   s = sst_file_writer.Finish(&file6_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file6_info.file_path, file6);
   ASSERT_EQ(file6_info.num_entries, 200);
   ASSERT_EQ(file6_info.smallest_key, Key(2000));
@@ -447,7 +449,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   s = AddFileWithFileChecksum({file1}, {file_checksum1, "xyz"},
                               {file_checksum1}, true, false, false, false);
   // does not care the checksum input since db does not enable file checksum
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_OK(env_->FileExists(file1));
   std::vector<LiveFileMetaData> live_files;
   dbfull()->GetLiveFilesMetaData(&live_files);
@@ -465,26 +467,26 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   s = AddFileWithFileChecksum({file2}, {file_checksum2, "xyz"},
                               {file_checksum_func_name2}, true, false, false,
                               false);
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
 
   // Enable verify_file_checksum option
   // The checksum name does not match, fail the ingestion
   s = AddFileWithFileChecksum({file2}, {file_checksum2}, {"xyz"}, true, false,
                               false, false);
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
 
   // Enable verify_file_checksum option
   // The checksum itself does not match, fail the ingestion
   s = AddFileWithFileChecksum({file2}, {"xyz"}, {file_checksum_func_name2},
                               true, false, false, false);
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
 
   // Enable verify_file_checksum option
   // All matches, ingestion is successful
   s = AddFileWithFileChecksum({file2}, {file_checksum2},
                               {file_checksum_func_name2}, true, false, false,
                               false);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   std::vector<LiveFileMetaData> live_files1;
   dbfull()->GetLiveFilesMetaData(&live_files1);
   for (auto f : live_files1) {
@@ -501,7 +503,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   std::vector<std::string> checksum, checksum_func;
   s = AddFileWithFileChecksum({file3}, checksum, checksum_func, true, false,
                               false, false);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   std::vector<LiveFileMetaData> live_files2;
   dbfull()->GetLiveFilesMetaData(&live_files2);
   for (auto f : live_files2) {
@@ -511,20 +513,20 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
       set1.insert(f.name);
     }
   }
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_OK(env_->FileExists(file3));
 
   // Does not enable verify_file_checksum options
   // The checksum name does not match, fail the ingestion
   s = AddFileWithFileChecksum({file4}, {file_checksum4}, {"xyz"}, false, false,
                               false, false);
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
 
   // Does not enable verify_file_checksum options
   // Checksum function name matches, store the checksum being ingested.
   s = AddFileWithFileChecksum({file4}, {"asd"}, {file_checksum_func_name4},
                               false, false, false, false);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   std::vector<LiveFileMetaData> live_files3;
   dbfull()->GetLiveFilesMetaData(&live_files3);
   for (auto f : live_files3) {
@@ -535,7 +537,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
       set1.insert(f.name);
     }
   }
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_OK(env_->FileExists(file4));
 
   // enable verify_file_checksum options, DB enable checksum, and enable
@@ -544,8 +546,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   s = AddFileWithFileChecksum({file5}, {file_checksum5},
                               {file_checksum_func_name5}, true, false, false,
                               true);
-  ASSERT_OK(s);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   std::vector<LiveFileMetaData> live_files4;
   dbfull()->GetLiveFilesMetaData(&live_files4);
   for (auto f : live_files4) {
@@ -558,7 +559,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
       set1.insert(f.name);
     }
   }
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_OK(env_->FileExists(file5));
 
   // Does not enable verify_file_checksum options and also the ingested file
@@ -567,7 +568,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
   std::vector<std::string> files_c6, files_name6;
   s = AddFileWithFileChecksum({file6}, files_c6, files_name6, false, false,
                               false, false);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   std::vector<LiveFileMetaData> live_files6;
   dbfull()->GetLiveFilesMetaData(&live_files6);
   for (auto f : live_files6) {
@@ -577,7 +578,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
       set1.insert(f.name);
     }
   }
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_OK(env_->FileExists(file6));
 }
 
@@ -595,7 +596,7 @@ TEST_F(ExternalSSTFileBasicTest, NoCopy) {
   }
   ExternalSstFileInfo file1_info;
   Status s = sst_file_writer.Finish(&file1_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file1_info.file_path, file1);
   ASSERT_EQ(file1_info.num_entries, 100);
   ASSERT_EQ(file1_info.smallest_key, Key(0));
@@ -609,7 +610,7 @@ TEST_F(ExternalSSTFileBasicTest, NoCopy) {
   }
   ExternalSstFileInfo file2_info;
   s = sst_file_writer.Finish(&file2_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file2_info.file_path, file2);
   ASSERT_EQ(file2_info.num_entries, 200);
   ASSERT_EQ(file2_info.smallest_key, Key(100));
@@ -623,23 +624,23 @@ TEST_F(ExternalSSTFileBasicTest, NoCopy) {
   }
   ExternalSstFileInfo file3_info;
   s = sst_file_writer.Finish(&file3_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file3_info.file_path, file3);
   ASSERT_EQ(file3_info.num_entries, 15);
   ASSERT_EQ(file3_info.smallest_key, Key(110));
   ASSERT_EQ(file3_info.largest_key, Key(124));
 
   s = DeprecatedAddFile({file1}, true /* move file */);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(Status::NotFound(), env_->FileExists(file1));
 
   s = DeprecatedAddFile({file2}, false /* copy file */);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_OK(env_->FileExists(file2));
 
   // This file has overlapping values with the existing data
   s = DeprecatedAddFile({file3}, true /* move file */);
-  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_NOK(s) << s.ToString();
   ASSERT_OK(env_->FileExists(file3));
 
   for (int k = 0; k < 300; k++) {
@@ -1126,7 +1127,7 @@ TEST_F(ExternalSSTFileBasicTest, SyncFailure) {
     if (i == 2) {
       ingest_opt.write_global_seqno = true;
     }
-    ASSERT_FALSE(db_->IngestExternalFile({file_name}, ingest_opt).ok());
+    ASSERT_NOK(db_->IngestExternalFile({file_name}, ingest_opt));
     db_->ReleaseSnapshot(snapshot);
 
     SyncPoint::GetInstance()->DisableProcessing();
@@ -1326,7 +1327,7 @@ TEST_F(ExternalSSTFileBasicTest, AdjacentRangeDeletionTombstones) {
   ASSERT_OK(sst_file_writer.DeleteRange(Key(300), Key(400)));
   ExternalSstFileInfo file8_info;
   Status s = sst_file_writer.Finish(&file8_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file8_info.file_path, file8);
   ASSERT_EQ(file8_info.num_entries, 0);
   ASSERT_EQ(file8_info.smallest_key, "");
@@ -1341,7 +1342,7 @@ TEST_F(ExternalSSTFileBasicTest, AdjacentRangeDeletionTombstones) {
   ASSERT_OK(sst_file_writer.DeleteRange(Key(400), Key(500)));
   ExternalSstFileInfo file9_info;
   s = sst_file_writer.Finish(&file9_info);
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(file9_info.file_path, file9);
   ASSERT_EQ(file9_info.num_entries, 0);
   ASSERT_EQ(file9_info.smallest_key, "");
@@ -1353,7 +1354,7 @@ TEST_F(ExternalSSTFileBasicTest, AdjacentRangeDeletionTombstones) {
   // Range deletion tombstones are exclusive on their end key, so these SSTs
   // should not be considered as overlapping.
   s = DeprecatedAddFile({file8, file9});
-  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(s) << s.ToString();
   ASSERT_EQ(db_->GetLatestSequenceNumber(), 0U);
   DestroyAndRecreateExternalSSTFilesDir();
 }

db/external_sst_file_ingestion_job.cc

@@ -335,6 +335,12 @@ Status ExternalSstFileIngestionJob::Run() {
   // with the files we are ingesting
   bool need_flush = false;
   status = NeedsFlush(&need_flush, super_version);
+  if (!status.ok()) {
+    return status;
+  }
+  if (need_flush) {
+    return Status::TryAgain();
+  }
   assert(status.ok() && need_flush == false);
 #endif
 

db/external_sst_file_test.cc

@@ -1345,15 +1345,18 @@ TEST_F(ExternalSSTFileTest, PickedLevelBug) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
 
   // While writing the MANIFEST start a thread that will ask for compaction
+  Status bg_compact_status;
   ROCKSDB_NAMESPACE::port::Thread bg_compact([&]() {
-    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+    bg_compact_status =
+        db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
   });
   TEST_SYNC_POINT("ExternalSSTFileTest::PickedLevelBug:2");
 
   // Start a thread that will ingest a new file
+  Status bg_addfile_status;
   ROCKSDB_NAMESPACE::port::Thread bg_addfile([&]() {
     file_keys = {1, 2, 3};
-    ASSERT_OK(GenerateAndAddExternalFile(options, file_keys, 1));
+    bg_addfile_status = GenerateAndAddExternalFile(options, file_keys, 1);
   });
 
   // Wait for AddFile to start picking levels and writing MANIFEST
@@ -1374,6 +1377,9 @@ TEST_F(ExternalSSTFileTest, PickedLevelBug) {
   bg_addfile.join();
   bg_compact.join();
 
+  ASSERT_OK(bg_addfile_status);
+  ASSERT_OK(bg_compact_status);
+
   dbfull()->TEST_WaitForCompact();
 
   int total_keys = 0;

db/flush_job.cc

@@ -80,22 +80,24 @@ const char* GetFlushReasonString (FlushReason flush_reason) {
   }
 }
 
-FlushJob::FlushJob(
-    const std::string& dbname, ColumnFamilyData* cfd,
-    const ImmutableDBOptions& db_options,
-    const MutableCFOptions& mutable_cf_options, const uint64_t* max_memtable_id,
-    const FileOptions& file_options, VersionSet* versions,
-    InstrumentedMutex* db_mutex, std::atomic<bool>* shutting_down,
-    std::vector<SequenceNumber> existing_snapshots,
-    SequenceNumber earliest_write_conflict_snapshot,
-    SnapshotChecker* snapshot_checker, JobContext* job_context,
-    LogBuffer* log_buffer, FSDirectory* db_directory,
-    FSDirectory* output_file_directory, CompressionType output_compression,
-    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 std::string& db_id, const std::string& db_session_id,
-    std::string full_history_ts_low)
+FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
+                   const ImmutableDBOptions& db_options,
+                   const MutableCFOptions& mutable_cf_options,
+                   uint64_t max_memtable_id, const FileOptions& file_options,
+                   VersionSet* versions, InstrumentedMutex* db_mutex,
+                   std::atomic<bool>* shutting_down,
+                   std::vector<SequenceNumber> existing_snapshots,
+                   SequenceNumber earliest_write_conflict_snapshot,
+                   SnapshotChecker* snapshot_checker, JobContext* job_context,
+                   LogBuffer* log_buffer, FSDirectory* db_directory,
+                   FSDirectory* output_file_directory,
+                   CompressionType output_compression, 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 std::string& db_id, const std::string& db_session_id,
+                   std::string full_history_ts_low)
     : dbname_(dbname),
       db_id_(db_id),
       db_session_id_(db_session_id),

db/flush_job.h

@@ -60,10 +60,9 @@ class FlushJob {
   // IMPORTANT: mutable_cf_options needs to be alive while FlushJob is alive
   FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
            const ImmutableDBOptions& db_options,
-           const MutableCFOptions& mutable_cf_options,
-           const uint64_t* max_memtable_id, const FileOptions& file_options,
-           VersionSet* versions, InstrumentedMutex* db_mutex,
-           std::atomic<bool>* shutting_down,
+           const MutableCFOptions& mutable_cf_options, uint64_t max_memtable_id,
+           const FileOptions& file_options, VersionSet* versions,
+           InstrumentedMutex* db_mutex, std::atomic<bool>* shutting_down,
            std::vector<SequenceNumber> existing_snapshots,
            SequenceNumber earliest_write_conflict_snapshot,
            SnapshotChecker* snapshot_checker, JobContext* job_context,
@@ -110,12 +109,11 @@ class FlushJob {
   ColumnFamilyData* cfd_;
   const ImmutableDBOptions& db_options_;
   const MutableCFOptions& mutable_cf_options_;
-  // Pointer to a variable storing the largest memtable id to flush in this
+  // A variable storing the largest memtable id to flush in this
   // flush job. RocksDB uses this variable to select the memtables to flush in
   // this job. All memtables in this column family with an ID smaller than or
-  // equal to *max_memtable_id_ will be selected for flush. If null, then all
-  // memtables in the column family will be selected.
-  const uint64_t* max_memtable_id_;
+  // equal to max_memtable_id_ will be selected for flush.
+  uint64_t max_memtable_id_;
   const FileOptions file_options_;
   VersionSet* versions_;
   InstrumentedMutex* db_mutex_;

db/flush_job_test.cc

@@ -158,7 +158,7 @@ TEST_F(FlushJobTest, Empty) {
   SnapshotChecker* snapshot_checker = nullptr;  // not relavant
   FlushJob flush_job(
       dbname_, versions_->GetColumnFamilySet()->GetDefault(), db_options_,
-      *cfd->GetLatestMutableCFOptions(), nullptr /* memtable_id */,
+      *cfd->GetLatestMutableCFOptions(), port::kMaxUint64 /* memtable_id */,
       env_options_, versions_.get(), &mutex_, &shutting_down_, {},
       kMaxSequenceNumber, snapshot_checker, &job_context, nullptr, nullptr,
       nullptr, kNoCompression, nullptr, &event_logger, false,
@@ -187,7 +187,7 @@ TEST_F(FlushJobTest, NonEmpty) {
   for (int i = 1; i < 10000; ++i) {
     std::string key(ToString((i + 1000) % 10000));
     std::string value("value" + key);
-    new_mem->Add(SequenceNumber(i), kTypeValue, key, value);
+    ASSERT_OK(new_mem->Add(SequenceNumber(i), kTypeValue, key, value));
     if ((i + 1000) % 10000 < 9995) {
       InternalKey internal_key(key, SequenceNumber(i), kTypeValue);
       inserted_keys.push_back({internal_key.Encode().ToString(), value});
@@ -195,7 +195,8 @@ TEST_F(FlushJobTest, NonEmpty) {
   }
 
   {
-    new_mem->Add(SequenceNumber(10000), kTypeRangeDeletion, "9995", "9999a");
+    ASSERT_OK(new_mem->Add(SequenceNumber(10000), kTypeRangeDeletion, "9995",
+                           "9999a"));
     InternalKey internal_key("9995", SequenceNumber(10000), kTypeRangeDeletion);
     inserted_keys.push_back({internal_key.Encode().ToString(), "9999a"});
   }
@@ -222,7 +223,7 @@ TEST_F(FlushJobTest, NonEmpty) {
     }
 
     const SequenceNumber seq(i + 10001);
-    new_mem->Add(seq, kTypeBlobIndex, key, blob_index);
+    ASSERT_OK(new_mem->Add(seq, kTypeBlobIndex, key, blob_index));
 
     InternalKey internal_key(key, seq, kTypeBlobIndex);
     inserted_keys.push_back({internal_key.Encode().ToString(), blob_index});
@@ -239,7 +240,7 @@ TEST_F(FlushJobTest, NonEmpty) {
   SnapshotChecker* snapshot_checker = nullptr;  // not relavant
   FlushJob flush_job(
       dbname_, versions_->GetColumnFamilySet()->GetDefault(), db_options_,
-      *cfd->GetLatestMutableCFOptions(), nullptr /* memtable_id */,
+      *cfd->GetLatestMutableCFOptions(), port::kMaxUint64 /* memtable_id */,
       env_options_, versions_.get(), &mutex_, &shutting_down_, {},
       kMaxSequenceNumber, snapshot_checker, &job_context, nullptr, nullptr,
       nullptr, kNoCompression, db_options_.statistics.get(), &event_logger,
@@ -283,8 +284,8 @@ TEST_F(FlushJobTest, FlushMemTablesSingleColumnFamily) {
     for (size_t j = 0; j < num_keys_per_table; ++j) {
       std::string key(ToString(j + i * num_keys_per_table));
       std::string value("value" + key);
-      mem->Add(SequenceNumber(j + i * num_keys_per_table), kTypeValue, key,
-               value);
+      ASSERT_OK(mem->Add(SequenceNumber(j + i * num_keys_per_table), kTypeValue,
+                         key, value));
     }
   }
 
@@ -301,7 +302,7 @@ TEST_F(FlushJobTest, FlushMemTablesSingleColumnFamily) {
   uint64_t flush_memtable_id = smallest_memtable_id + num_mems_to_flush - 1;
   FlushJob flush_job(
       dbname_, versions_->GetColumnFamilySet()->GetDefault(), db_options_,
-      *cfd->GetLatestMutableCFOptions(), &flush_memtable_id, env_options_,
+      *cfd->GetLatestMutableCFOptions(), flush_memtable_id, env_options_,
       versions_.get(), &mutex_, &shutting_down_, {}, kMaxSequenceNumber,
       snapshot_checker, &job_context, nullptr, nullptr, nullptr, kNoCompression,
       db_options_.statistics.get(), &event_logger, true,
@@ -356,7 +357,7 @@ TEST_F(FlushJobTest, FlushMemtablesMultipleColumnFamilies) {
       for (size_t j = 0; j != num_keys_per_memtable; ++j) {
         std::string key(ToString(j + i * num_keys_per_memtable));
         std::string value("value" + key);
-        mem->Add(curr_seqno++, kTypeValue, key, value);
+        ASSERT_OK(mem->Add(curr_seqno++, kTypeValue, key, value));
       }
 
       cfd->imm()->Add(mem, &to_delete);
@@ -373,7 +374,7 @@ TEST_F(FlushJobTest, FlushMemtablesMultipleColumnFamilies) {
     std::vector<SequenceNumber> snapshot_seqs;
     flush_jobs.emplace_back(new FlushJob(
         dbname_, cfd, db_options_, *cfd->GetLatestMutableCFOptions(),
-        &memtable_ids[k], env_options_, versions_.get(), &mutex_,
+        memtable_ids[k], env_options_, versions_.get(), &mutex_,
         &shutting_down_, snapshot_seqs, kMaxSequenceNumber, snapshot_checker,
         &job_context, nullptr, nullptr, nullptr, kNoCompression,
         db_options_.statistics.get(), &event_logger, true,
@@ -411,8 +412,9 @@ TEST_F(FlushJobTest, FlushMemtablesMultipleColumnFamilies) {
 
   Status s = InstallMemtableAtomicFlushResults(
       nullptr /* imm_lists */, all_cfds, mutable_cf_options_list, mems_list,
-      versions_.get(), &mutex_, file_meta_ptrs, &job_context.memtables_to_free,
-      nullptr /* db_directory */, nullptr /* log_buffer */);
+      versions_.get(), nullptr /* prep_tracker */, &mutex_, file_meta_ptrs,
+      &job_context.memtables_to_free, nullptr /* db_directory */,
+      nullptr /* log_buffer */);
   ASSERT_OK(s);
 
   mutex_.Unlock();
@@ -466,7 +468,7 @@ TEST_F(FlushJobTest, Snapshots) {
     for (int j = 0; j < insertions; ++j) {
       std::string value(rnd.HumanReadableString(10));
       auto seqno = ++current_seqno;
-      new_mem->Add(SequenceNumber(seqno), kTypeValue, key, value);
+      ASSERT_OK(new_mem->Add(SequenceNumber(seqno), kTypeValue, key, value));
       // a key is visible only if:
       // 1. it's the last one written (j == insertions - 1)
       // 2. there's a snapshot pointing at it
@@ -490,7 +492,7 @@ TEST_F(FlushJobTest, Snapshots) {
   SnapshotChecker* snapshot_checker = nullptr;  // not relavant
   FlushJob flush_job(
       dbname_, versions_->GetColumnFamilySet()->GetDefault(), db_options_,
-      *cfd->GetLatestMutableCFOptions(), nullptr /* memtable_id */,
+      *cfd->GetLatestMutableCFOptions(), port::kMaxUint64 /* memtable_id */,
       env_options_, versions_.get(), &mutex_, &shutting_down_, snapshots,
       kMaxSequenceNumber, snapshot_checker, &job_context, nullptr, nullptr,
       nullptr, kNoCompression, db_options_.statistics.get(), &event_logger,
@@ -518,7 +520,7 @@ class FlushJobTimestampTest : public FlushJobTestBase {
                              Slice value) {
     std::string key_str(std::move(key));
     PutFixed64(&key_str, ts);
-    memtable->Add(seq, value_type, key_str, value);
+    ASSERT_OK(memtable->Add(seq, value_type, key_str, value));
   }
 
  protected:
@@ -557,8 +559,8 @@ TEST_F(FlushJobTimestampTest, AllKeysExpired) {
   PutFixed64(&full_history_ts_low, std::numeric_limits<uint64_t>::max());
   FlushJob flush_job(
       dbname_, cfd, db_options_, *cfd->GetLatestMutableCFOptions(),
-      nullptr /* memtable_id */, env_options_, versions_.get(), &mutex_,
-      &shutting_down_, snapshots, kMaxSequenceNumber, snapshot_checker,
+      port::kMaxUint64 /* memtable_id */, env_options_, versions_.get(),
+      &mutex_, &shutting_down_, snapshots, kMaxSequenceNumber, snapshot_checker,
       &job_context, nullptr, nullptr, nullptr, kNoCompression,
       db_options_.statistics.get(), &event_logger, true,
       true /* sync_output_directory */, true /* write_manifest */,
@@ -608,8 +610,8 @@ TEST_F(FlushJobTimestampTest, NoKeyExpired) {
   PutFixed64(&full_history_ts_low, 0);
   FlushJob flush_job(
       dbname_, cfd, db_options_, *cfd->GetLatestMutableCFOptions(),
-      nullptr /* memtable_id */, env_options_, versions_.get(), &mutex_,
-      &shutting_down_, snapshots, kMaxSequenceNumber, snapshot_checker,
+      port::kMaxUint64 /* memtable_id */, env_options_, versions_.get(),
+      &mutex_, &shutting_down_, snapshots, kMaxSequenceNumber, snapshot_checker,
       &job_context, nullptr, nullptr, nullptr, kNoCompression,
       db_options_.statistics.get(), &event_logger, true,
       true /* sync_output_directory */, true /* write_manifest */,

db/forward_iterator.cc

@@ -46,7 +46,13 @@ class ForwardLevelIterator : public InternalIterator {
         file_iter_(nullptr),
         pinned_iters_mgr_(nullptr),
         prefix_extractor_(prefix_extractor),
-        allow_unprepared_value_(allow_unprepared_value) {}
+        allow_unprepared_value_(allow_unprepared_value) {
+    /*
+    NOTE needed for ASSERT_STATUS_CHECKED
+    in MergeOperatorPinningTest/MergeOperatorPinningTest.TailingIterator
+    */
+    status_.PermitUncheckedError();
+  }
 
   ~ForwardLevelIterator() override {
     // Reset current pointer

db/internal_stats.cc

@@ -61,6 +61,7 @@ void PrintLevelStatsHeader(char* buf, size_t len, const std::string& cf_name,
                            const std::string& group_by) {
   int written_size =
       snprintf(buf, len, "\n** Compaction Stats [%s] **\n", cf_name.c_str());
+  written_size = std::min(written_size, static_cast<int>(len));
   auto hdr = [](LevelStatType t) {
     return InternalStats::compaction_level_stats.at(t).header_name.c_str();
   };
@@ -80,6 +81,7 @@ void PrintLevelStatsHeader(char* buf, size_t len, const std::string& cf_name,
       hdr(LevelStatType::KEY_DROP));
 
   written_size += line_size;
+  written_size = std::min(written_size, static_cast<int>(len));
   snprintf(buf + written_size, len - written_size, "%s\n",
            std::string(line_size, '-').c_str());
 }
@@ -376,10 +378,11 @@ const std::unordered_map<std::string, DBPropertyInfo>
          {false, &InternalStats::HandleSsTables, nullptr, nullptr, nullptr}},
         {DB::Properties::kAggregatedTableProperties,
          {false, &InternalStats::HandleAggregatedTableProperties, nullptr,
-          nullptr, nullptr}},
+          &InternalStats::HandleAggregatedTablePropertiesMap, nullptr}},
         {DB::Properties::kAggregatedTablePropertiesAtLevel,
          {false, &InternalStats::HandleAggregatedTablePropertiesAtLevel,
-          nullptr, nullptr, nullptr}},
+          nullptr, &InternalStats::HandleAggregatedTablePropertiesAtLevelMap,
+          nullptr}},
         {DB::Properties::kNumImmutableMemTable,
          {false, nullptr, &InternalStats::HandleNumImmutableMemTable, nullptr,
           nullptr}},
@@ -508,11 +511,12 @@ bool InternalStats::GetStringProperty(const DBPropertyInfo& property_info,
 }
 
 bool InternalStats::GetMapProperty(const DBPropertyInfo& property_info,
-                                   const Slice& /*property*/,
+                                   const Slice& property,
                                    std::map<std::string, std::string>* value) {
   assert(value != nullptr);
   assert(property_info.handle_map != nullptr);
-  return (this->*(property_info.handle_map))(value);
+  Slice arg = GetPropertyNameAndArg(property).second;
+  return (this->*(property_info.handle_map))(value, arg);
 }
 
 bool InternalStats::GetIntProperty(const DBPropertyInfo& property_info,
@@ -588,7 +592,7 @@ bool InternalStats::HandleStats(std::string* value, Slice suffix) {
 }
 
 bool InternalStats::HandleCFMapStats(
-    std::map<std::string, std::string>* cf_stats) {
+    std::map<std::string, std::string>* cf_stats, Slice /*suffix*/) {
   DumpCFMapStats(cf_stats);
   return true;
 }
@@ -632,7 +636,27 @@ bool InternalStats::HandleAggregatedTableProperties(std::string* value,
   return true;
 }
 
-bool InternalStats::HandleAggregatedTablePropertiesAtLevel(std::string* value,
+static std::map<std::string, std::string> MapUint64ValuesToString(
+    const std::map<std::string, uint64_t>& from) {
+  std::map<std::string, std::string> to;
+  for (const auto& e : from) {
+    to[e.first] = ToString(e.second);
+  }
+  return to;
+}
+
+bool InternalStats::HandleAggregatedTablePropertiesMap(
+    std::map<std::string, std::string>* values, Slice /*suffix*/) {
+  std::shared_ptr<const TableProperties> tp;
+  auto s = cfd_->current()->GetAggregatedTableProperties(&tp);
+  if (!s.ok()) {
+    return false;
+  }
+  *values = MapUint64ValuesToString(tp->GetAggregatablePropertiesAsMap());
+  return true;
+}
+
+bool InternalStats::HandleAggregatedTablePropertiesAtLevel(std::string* values,
                                                            Slice suffix) {
   uint64_t level;
   bool ok = ConsumeDecimalNumber(&suffix, &level) && suffix.empty();
@@ -645,7 +669,24 @@ bool InternalStats::HandleAggregatedTablePropertiesAtLevel(std::string* value,
   if (!s.ok()) {
     return false;
   }
-  *value = tp->ToString();
+  *values = tp->ToString();
+  return true;
+}
+
+bool InternalStats::HandleAggregatedTablePropertiesAtLevelMap(
+    std::map<std::string, std::string>* values, Slice suffix) {
+  uint64_t level;
+  bool ok = ConsumeDecimalNumber(&suffix, &level) && suffix.empty();
+  if (!ok || static_cast<int>(level) >= number_levels_) {
+    return false;
+  }
+  std::shared_ptr<const TableProperties> tp;
+  auto s = cfd_->current()->GetAggregatedTableProperties(
+      &tp, static_cast<int>(level));
+  if (!s.ok()) {
+    return false;
+  }
+  *values = MapUint64ValuesToString(tp->GetAggregatablePropertiesAsMap());
   return true;
 }
 

db/internal_stats.h

@@ -44,7 +44,9 @@ struct DBPropertyInfo {
                                     Version* version);
 
   // @param props Map of general properties to populate
-  bool (InternalStats::*handle_map)(std::map<std::string, std::string>* props);
+  // @param suffix Argument portion of the property. (see handle_string)
+  bool (InternalStats::*handle_map)(std::map<std::string, std::string>* props,
+                                    Slice suffix);
 
   // handle the string type properties rely on DBImpl methods
   // @param value Value-result argument for storing the property's string value
@@ -525,7 +527,8 @@ class InternalStats {
   bool HandleCompressionRatioAtLevelPrefix(std::string* value, Slice suffix);
   bool HandleLevelStats(std::string* value, Slice suffix);
   bool HandleStats(std::string* value, Slice suffix);
-  bool HandleCFMapStats(std::map<std::string, std::string>* compaction_stats);
+  bool HandleCFMapStats(std::map<std::string, std::string>* compaction_stats,
+                        Slice suffix);
   bool HandleCFStats(std::string* value, Slice suffix);
   bool HandleCFStatsNoFileHistogram(std::string* value, Slice suffix);
   bool HandleCFFileHistogram(std::string* value, Slice suffix);
@@ -533,6 +536,10 @@ class InternalStats {
   bool HandleSsTables(std::string* value, Slice suffix);
   bool HandleAggregatedTableProperties(std::string* value, Slice suffix);
   bool HandleAggregatedTablePropertiesAtLevel(std::string* value, Slice suffix);
+  bool HandleAggregatedTablePropertiesMap(
+      std::map<std::string, std::string>* values, Slice suffix);
+  bool HandleAggregatedTablePropertiesAtLevelMap(
+      std::map<std::string, std::string>* values, Slice suffix);
   bool HandleNumImmutableMemTable(uint64_t* value, DBImpl* db,
                                   Version* version);
   bool HandleNumImmutableMemTableFlushed(uint64_t* value, DBImpl* db,

db/memtable.cc

@@ -328,9 +328,11 @@ class MemTableIterator : public InternalIterator {
     PERF_COUNTER_ADD(seek_on_memtable_count, 1);
     if (bloom_) {
       // iterator should only use prefix bloom filter
-      Slice user_k(ExtractUserKey(k));
-      if (prefix_extractor_->InDomain(user_k) &&
-          !bloom_->MayContain(prefix_extractor_->Transform(user_k))) {
+      auto ts_sz = comparator_.comparator.user_comparator()->timestamp_size();
+      Slice user_k_without_ts(ExtractUserKeyAndStripTimestamp(k, ts_sz));
+      if (prefix_extractor_->InDomain(user_k_without_ts) &&
+          !bloom_->MayContain(
+              prefix_extractor_->Transform(user_k_without_ts))) {
         PERF_COUNTER_ADD(bloom_memtable_miss_count, 1);
         valid_ = false;
         return;
@@ -345,9 +347,11 @@ class MemTableIterator : public InternalIterator {
     PERF_TIMER_GUARD(seek_on_memtable_time);
     PERF_COUNTER_ADD(seek_on_memtable_count, 1);
     if (bloom_) {
-      Slice user_k(ExtractUserKey(k));
-      if (prefix_extractor_->InDomain(user_k) &&
-          !bloom_->MayContain(prefix_extractor_->Transform(user_k))) {
+      auto ts_sz = comparator_.comparator.user_comparator()->timestamp_size();
+      Slice user_k_without_ts(ExtractUserKeyAndStripTimestamp(k, ts_sz));
+      if (prefix_extractor_->InDomain(user_k_without_ts) &&
+          !bloom_->MayContain(
+              prefix_extractor_->Transform(user_k_without_ts))) {
         PERF_COUNTER_ADD(bloom_memtable_miss_count, 1);
         valid_ = false;
         return;
@@ -480,10 +484,10 @@ MemTable::MemTableStats MemTable::ApproximateStats(const Slice& start_ikey,
   return {entry_count * (data_size / n), entry_count};
 }
 
-bool MemTable::Add(SequenceNumber s, ValueType type,
-                   const Slice& key, /* user key */
-                   const Slice& value, bool allow_concurrent,
-                   MemTablePostProcessInfo* post_process_info, void** hint) {
+Status MemTable::Add(SequenceNumber s, ValueType type,
+                     const Slice& key, /* user key */
+                     const Slice& value, bool allow_concurrent,
+                     MemTablePostProcessInfo* post_process_info, void** hint) {
   // Format of an entry is concatenation of:
   //  key_size     : varint32 of internal_key.size()
   //  key bytes    : char[internal_key.size()]
@@ -511,6 +515,7 @@ bool MemTable::Add(SequenceNumber s, ValueType type,
   memcpy(p, value.data(), val_size);
   assert((unsigned)(p + val_size - buf) == (unsigned)encoded_len);
   size_t ts_sz = GetInternalKeyComparator().user_comparator()->timestamp_size();
+  Slice key_without_ts = StripTimestampFromUserKey(key, ts_sz);
 
   if (!allow_concurrent) {
     // Extract prefix for insert with hint.
@@ -519,12 +524,12 @@ bool MemTable::Add(SequenceNumber s, ValueType type,
       Slice prefix = insert_with_hint_prefix_extractor_->Transform(key_slice);
       bool res = table->InsertKeyWithHint(handle, &insert_hints_[prefix]);
       if (UNLIKELY(!res)) {
-        return res;
+        return Status::TryAgain("key+seq exists");
       }
     } else {
       bool res = table->InsertKey(handle);
       if (UNLIKELY(!res)) {
-        return res;
+        return Status::TryAgain("key+seq exists");
       }
     }
 
@@ -540,12 +545,11 @@ bool MemTable::Add(SequenceNumber s, ValueType type,
     }
 
     if (bloom_filter_ && prefix_extractor_ &&
-        prefix_extractor_->InDomain(key)) {
-      bloom_filter_->Add(
-          prefix_extractor_->Transform(StripTimestampFromUserKey(key, ts_sz)));
+        prefix_extractor_->InDomain(key_without_ts)) {
+      bloom_filter_->Add(prefix_extractor_->Transform(key_without_ts));
     }
     if (bloom_filter_ && moptions_.memtable_whole_key_filtering) {
-      bloom_filter_->Add(StripTimestampFromUserKey(key, ts_sz));
+      bloom_filter_->Add(key_without_ts);
     }
 
     // The first sequence number inserted into the memtable
@@ -566,7 +570,7 @@ bool MemTable::Add(SequenceNumber s, ValueType type,
                    ? table->InsertKeyConcurrently(handle)
                    : table->InsertKeyWithHintConcurrently(handle, hint);
     if (UNLIKELY(!res)) {
-      return res;
+      return Status::TryAgain("key+seq exists");
     }
 
     assert(post_process_info != nullptr);
@@ -577,11 +581,12 @@ bool MemTable::Add(SequenceNumber s, ValueType type,
     }
 
     if (bloom_filter_ && prefix_extractor_ &&
-        prefix_extractor_->InDomain(key)) {
-      bloom_filter_->AddConcurrently(prefix_extractor_->Transform(key));
+        prefix_extractor_->InDomain(key_without_ts)) {
+      bloom_filter_->AddConcurrently(
+          prefix_extractor_->Transform(key_without_ts));
     }
     if (bloom_filter_ && moptions_.memtable_whole_key_filtering) {
-      bloom_filter_->AddConcurrently(StripTimestampFromUserKey(key, ts_sz));
+      bloom_filter_->AddConcurrently(key_without_ts);
     }
 
     // atomically update first_seqno_ and earliest_seqno_.
@@ -600,7 +605,7 @@ bool MemTable::Add(SequenceNumber s, ValueType type,
     is_range_del_table_empty_.store(false, std::memory_order_relaxed);
   }
   UpdateOldestKeyTime();
-  return true;
+  return Status::OK();
 }
 
 // Callback from MemTable::Get()
@@ -821,22 +826,21 @@ bool MemTable::Get(const LookupKey& key, std::string* value,
                  range_del_iter->MaxCoveringTombstoneSeqnum(key.user_key()));
   }
 
-  Slice user_key = key.user_key();
   bool found_final_value = false;
   bool merge_in_progress = s->IsMergeInProgress();
   bool may_contain = true;
   size_t ts_sz = GetInternalKeyComparator().user_comparator()->timestamp_size();
+  Slice user_key_without_ts = StripTimestampFromUserKey(key.user_key(), ts_sz);
   if (bloom_filter_) {
     // when both memtable_whole_key_filtering and prefix_extractor_ are set,
     // only do whole key filtering for Get() to save CPU
     if (moptions_.memtable_whole_key_filtering) {
-      may_contain =
-          bloom_filter_->MayContain(StripTimestampFromUserKey(user_key, ts_sz));
+      may_contain = bloom_filter_->MayContain(user_key_without_ts);
     } else {
       assert(prefix_extractor_);
-      may_contain =
-          !prefix_extractor_->InDomain(user_key) ||
-          bloom_filter_->MayContain(prefix_extractor_->Transform(user_key));
+      may_contain = !prefix_extractor_->InDomain(user_key_without_ts) ||
+                    bloom_filter_->MayContain(
+                        prefix_extractor_->Transform(user_key_without_ts));
     }
   }
 
@@ -893,7 +897,7 @@ void MemTable::GetFromTable(const LookupKey& key,
 }
 
 void MemTable::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
-                        ReadCallback* callback, bool* is_blob) {
+                        ReadCallback* callback) {
   // The sequence number is updated synchronously in version_set.h
   if (IsEmpty()) {
     // Avoiding recording stats for speed.
@@ -946,9 +950,9 @@ void MemTable::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
           range_del_iter->MaxCoveringTombstoneSeqnum(iter->lkey->user_key()));
     }
     GetFromTable(*(iter->lkey), iter->max_covering_tombstone_seq, true,
-                 callback, is_blob, iter->value->GetSelf(), iter->timestamp,
-                 iter->s, &(iter->merge_context), &seq, &found_final_value,
-                 &merge_in_progress);
+                 callback, &iter->is_blob_index, iter->value->GetSelf(),
+                 iter->timestamp, iter->s, &(iter->merge_context), &seq,
+                 &found_final_value, &merge_in_progress);
 
     if (!found_final_value && merge_in_progress) {
       *(iter->s) = Status::MergeInProgress();
@@ -973,9 +977,8 @@ void MemTable::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
   PERF_COUNTER_ADD(get_from_memtable_count, 1);
 }
 
-void MemTable::Update(SequenceNumber seq,
-                      const Slice& key,
-                      const Slice& value) {
+Status MemTable::Update(SequenceNumber seq, const Slice& key,
+                        const Slice& value) {
   LookupKey lkey(key, seq);
   Slice mem_key = lkey.memtable_key();
 
@@ -1019,22 +1022,18 @@ void MemTable::Update(SequenceNumber seq,
                  (unsigned)(VarintLength(key_length) + key_length +
                             VarintLength(value.size()) + value.size()));
           RecordTick(moptions_.statistics, NUMBER_KEYS_UPDATED);
-          return;
+          return Status::OK();
         }
       }
     }
   }
 
-  // key doesn't exist
-  bool add_res __attribute__((__unused__));
-  add_res = Add(seq, kTypeValue, key, value);
-  // We already checked unused != seq above. In that case, Add should not fail.
-  assert(add_res);
+  // The latest value is not `kTypeValue` or key doesn't exist
+  return Add(seq, kTypeValue, key, value);
 }
 
-bool MemTable::UpdateCallback(SequenceNumber seq,
-                              const Slice& key,
-                              const Slice& delta) {
+Status MemTable::UpdateCallback(SequenceNumber seq, const Slice& key,
+                                const Slice& delta) {
   LookupKey lkey(key, seq);
   Slice memkey = lkey.memtable_key();
 
@@ -1088,16 +1087,17 @@ bool MemTable::UpdateCallback(SequenceNumber seq,
             }
             RecordTick(moptions_.statistics, NUMBER_KEYS_UPDATED);
             UpdateFlushState();
-            return true;
+            return Status::OK();
           } else if (status == UpdateStatus::UPDATED) {
-            Add(seq, kTypeValue, key, Slice(str_value));
+            Status s = Add(seq, kTypeValue, key, Slice(str_value));
             RecordTick(moptions_.statistics, NUMBER_KEYS_WRITTEN);
             UpdateFlushState();
-            return true;
+            return s;
           } else if (status == UpdateStatus::UPDATE_FAILED) {
-            // No action required. Return.
+            // `UPDATE_FAILED` is named incorrectly. It indicates no update
+            // happened. It does not indicate a failure happened.
             UpdateFlushState();
-            return true;
+            return Status::OK();
           }
         }
         default:
@@ -1105,9 +1105,8 @@ bool MemTable::UpdateCallback(SequenceNumber seq,
       }
     }
   }
-  // If the latest value is not kTypeValue
-  // or key doesn't exist
-  return false;
+  // The latest value is not `kTypeValue` or key doesn't exist
+  return Status::NotFound();
 }
 
 size_t MemTable::CountSuccessiveMergeEntries(const LookupKey& key) {

db/memtable.h

@@ -182,12 +182,13 @@ class MemTable {
   // REQUIRES: if allow_concurrent = false, external synchronization to prevent
   // simultaneous operations on the same MemTable.
   //
-  // Returns false if MemTableRepFactory::CanHandleDuplicatedKey() is true and
-  // the <key, seq> already exists.
-  bool Add(SequenceNumber seq, ValueType type, const Slice& key,
-           const Slice& value, bool allow_concurrent = false,
-           MemTablePostProcessInfo* post_process_info = nullptr,
-           void** hint = nullptr);
+  // Returns `Status::TryAgain` if the `seq`, `key` combination already exists
+  // in the memtable and `MemTableRepFactory::CanHandleDuplicatedKey()` is true.
+  // The next attempt should try a larger value for `seq`.
+  Status Add(SequenceNumber seq, ValueType type, const Slice& key,
+             const Slice& value, bool allow_concurrent = false,
+             MemTablePostProcessInfo* post_process_info = nullptr,
+             void** hint = nullptr);
 
   // Used to Get value associated with key or Get Merge Operands associated
   // with key.
@@ -237,37 +238,36 @@ class MemTable {
   }
 
   void MultiGet(const ReadOptions& read_options, MultiGetRange* range,
-                ReadCallback* callback, bool* is_blob);
+                ReadCallback* callback);
 
-  // Attempts to update the new_value inplace, else does normal Add
-  // Pseudocode
-  //   if key exists in current memtable && prev_value is of type kTypeValue
-  //     if new sizeof(new_value) <= sizeof(prev_value)
-  //       update inplace
-  //     else add(key, new_value)
-  //   else add(key, new_value)
+  // If `key` exists in current memtable with type `kTypeValue` and the existing
+  // value is at least as large as the new value, updates it in-place. Otherwise
+  // adds the new value to the memtable out-of-place.
+  //
+  // Returns `Status::TryAgain` if the `seq`, `key` combination already exists
+  // in the memtable and `MemTableRepFactory::CanHandleDuplicatedKey()` is true.
+  // The next attempt should try a larger value for `seq`.
   //
   // REQUIRES: external synchronization to prevent simultaneous
   // operations on the same MemTable.
-  void Update(SequenceNumber seq,
-              const Slice& key,
-              const Slice& value);
+  Status Update(SequenceNumber seq, const Slice& key, const Slice& value);
 
-  // If prev_value for key exists, attempts to update it inplace.
-  // else returns false
-  // Pseudocode
-  //   if key exists in current memtable && prev_value is of type kTypeValue
-  //     new_value = delta(prev_value)
-  //     if sizeof(new_value) <= sizeof(prev_value)
-  //       update inplace
-  //     else add(key, new_value)
-  //   else return false
+  // If `key` exists in current memtable with type `kTypeValue` and the existing
+  // value is at least as large as the new value, updates it in-place. Otherwise
+  // if `key` exists in current memtable with type `kTypeValue`, adds the new
+  // value to the memtable out-of-place.
+  //
+  // Returns `Status::NotFound` if `key` does not exist in current memtable or
+  // the latest version of `key` does not have `kTypeValue`.
+  //
+  // Returns `Status::TryAgain` if the `seq`, `key` combination already exists
+  // in the memtable and `MemTableRepFactory::CanHandleDuplicatedKey()` is true.
+  // The next attempt should try a larger value for `seq`.
   //
   // REQUIRES: external synchronization to prevent simultaneous
   // operations on the same MemTable.
-  bool UpdateCallback(SequenceNumber seq,
-                      const Slice& key,
-                      const Slice& delta);
+  Status UpdateCallback(SequenceNumber seq, const Slice& key,
+                        const Slice& delta);
 
   // Returns the number of successive merge entries starting from the newest
   // entry for the key up to the last non-merge entry or last entry for the

db/memtable_list.cc

@@ -113,10 +113,10 @@ bool MemTableListVersion::Get(const LookupKey& key, std::string* value,
 }
 
 void MemTableListVersion::MultiGet(const ReadOptions& read_options,
-                                   MultiGetRange* range, ReadCallback* callback,
-                                   bool* is_blob) {
+                                   MultiGetRange* range,
+                                   ReadCallback* callback) {
   for (auto memtable : memlist_) {
-    memtable->MultiGet(read_options, range, callback, is_blob);
+    memtable->MultiGet(read_options, range, callback);
     if (range->empty()) {
       return;
     }
@@ -334,7 +334,7 @@ bool MemTableList::IsFlushPending() const {
 }
 
 // Returns the memtables that need to be flushed.
-void MemTableList::PickMemtablesToFlush(const uint64_t* max_memtable_id,
+void MemTableList::PickMemtablesToFlush(uint64_t max_memtable_id,
                                         autovector<MemTable*>* ret) {
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_PICK_MEMTABLES_TO_FLUSH);
@@ -345,7 +345,7 @@ void MemTableList::PickMemtablesToFlush(const uint64_t* max_memtable_id,
     if (!atomic_flush && m->atomic_flush_seqno_ != kMaxSequenceNumber) {
       atomic_flush = true;
     }
-    if (max_memtable_id != nullptr && m->GetID() > *max_memtable_id) {
+    if (m->GetID() > max_memtable_id) {
       break;
     }
     if (!m->flush_in_progress_) {
@@ -481,15 +481,16 @@ Status MemTableList::TryInstallMemtableFlushResults(
         // We piggyback the information of  earliest log file to keep in the
         // manifest entry for the last file flushed.
         edit_list.back()->SetMinLogNumberToKeep(min_wal_number_to_keep);
-      } else {
-        min_wal_number_to_keep =
-            PrecomputeMinLogNumberToKeepNon2PC(vset, *cfd, edit_list);
       }
 
       std::unique_ptr<VersionEdit> wal_deletion;
       if (vset->db_options()->track_and_verify_wals_in_manifest) {
-        const auto& wals = vset->GetWalSet().GetWals();
-        if (!wals.empty() && min_wal_number_to_keep > wals.begin()->first) {
+        if (!vset->db_options()->allow_2pc) {
+          min_wal_number_to_keep =
+              PrecomputeMinLogNumberToKeepNon2PC(vset, *cfd, edit_list);
+        }
+        if (min_wal_number_to_keep >
+            vset->GetWalSet().GetMinWalNumberToKeep()) {
           wal_deletion.reset(new VersionEdit);
           wal_deletion->DeleteWalsBefore(min_wal_number_to_keep);
           edit_list.push_back(wal_deletion.get());
@@ -674,20 +675,11 @@ void MemTableList::RemoveMemTablesOrRestoreFlags(
 }
 
 uint64_t MemTableList::PrecomputeMinLogContainingPrepSection(
-    const autovector<MemTable*>& memtables_to_flush) {
+    const std::unordered_set<MemTable*>* memtables_to_flush) {
   uint64_t min_log = 0;
 
   for (auto& m : current_->memlist_) {
-    // Assume the list is very short, we can live with O(m*n). We can optimize
-    // if the performance has some problem.
-    bool should_skip = false;
-    for (MemTable* m_to_flush : memtables_to_flush) {
-      if (m == m_to_flush) {
-        should_skip = true;
-        break;
-      }
-    }
-    if (should_skip) {
+    if (memtables_to_flush && memtables_to_flush->count(m)) {
       continue;
     }
 
@@ -707,7 +699,8 @@ Status InstallMemtableAtomicFlushResults(
     const autovector<ColumnFamilyData*>& cfds,
     const autovector<const MutableCFOptions*>& mutable_cf_options_list,
     const autovector<const autovector<MemTable*>*>& mems_list, VersionSet* vset,
-    InstrumentedMutex* mu, const autovector<FileMetaData*>& file_metas,
+    LogsWithPrepTracker* prep_tracker, InstrumentedMutex* mu,
+    const autovector<FileMetaData*>& file_metas,
     autovector<MemTable*>* to_delete, FSDirectory* db_directory,
     LogBuffer* log_buffer) {
   AutoThreadOperationStageUpdater stage_updater(
@@ -752,19 +745,20 @@ Status InstallMemtableAtomicFlushResults(
     edit_lists.emplace_back(edits);
   }
 
-  // TODO(cc): after https://github.com/facebook/rocksdb/pull/7570, handle 2pc
-  // here.
+  WalNumber min_wal_number_to_keep = 0;
+  if (vset->db_options()->allow_2pc) {
+    min_wal_number_to_keep = PrecomputeMinLogNumberToKeep2PC(
+        vset, cfds, edit_lists, mems_list, prep_tracker);
+    edit_lists.back().back()->SetMinLogNumberToKeep(min_wal_number_to_keep);
+  }
+
   std::unique_ptr<VersionEdit> wal_deletion;
   if (vset->db_options()->track_and_verify_wals_in_manifest) {
-    uint64_t min_wal_number_to_keep =
-        PrecomputeMinLogNumberToKeepNon2PC(vset, *cfds[0], edit_lists[0]);
-    for (size_t i = 1; i < cfds.size(); i++) {
-      min_wal_number_to_keep = std::min(
-          min_wal_number_to_keep,
-          PrecomputeMinLogNumberToKeepNon2PC(vset, *cfds[i], edit_lists[i]));
+    if (!vset->db_options()->allow_2pc) {
+      min_wal_number_to_keep =
+          PrecomputeMinLogNumberToKeepNon2PC(vset, cfds, edit_lists);
     }
-    const auto& wals = vset->GetWalSet().GetWals();
-    if (!wals.empty() && min_wal_number_to_keep > wals.begin()->first) {
+    if (min_wal_number_to_keep > vset->GetWalSet().GetMinWalNumberToKeep()) {
       wal_deletion.reset(new VersionEdit);
       wal_deletion->DeleteWalsBefore(min_wal_number_to_keep);
       edit_lists.back().push_back(wal_deletion.get());

db/memtable_list.h

@@ -76,7 +76,7 @@ class MemTableListVersion {
   }
 
   void MultiGet(const ReadOptions& read_options, MultiGetRange* range,
-                ReadCallback* callback, bool* is_blob);
+                ReadCallback* callback);
 
   // Returns all the merge operands corresponding to the key by searching all
   // memtables starting from the most recent one.
@@ -138,8 +138,8 @@ class MemTableListVersion {
       const autovector<ColumnFamilyData*>& cfds,
       const autovector<const MutableCFOptions*>& mutable_cf_options_list,
       const autovector<const autovector<MemTable*>*>& mems_list,
-      VersionSet* vset, InstrumentedMutex* mu,
-      const autovector<FileMetaData*>& file_meta,
+      VersionSet* vset, LogsWithPrepTracker* prep_tracker,
+      InstrumentedMutex* mu, const autovector<FileMetaData*>& file_meta,
       autovector<MemTable*>* to_delete, FSDirectory* db_directory,
       LogBuffer* log_buffer);
 
@@ -251,7 +251,7 @@ class MemTableList {
 
   // Returns the earliest memtables that needs to be flushed. The returned
   // memtables are guaranteed to be in the ascending order of created time.
-  void PickMemtablesToFlush(const uint64_t* max_memtable_id,
+  void PickMemtablesToFlush(uint64_t max_memtable_id,
                             autovector<MemTable*>* mems);
 
   // Reset status of the given memtable list back to pending state so that
@@ -335,7 +335,7 @@ class MemTableList {
   // Returns the min log containing the prep section after memtables listsed in
   // `memtables_to_flush` are flushed and their status is persisted in manifest.
   uint64_t PrecomputeMinLogContainingPrepSection(
-      const autovector<MemTable*>& memtables_to_flush);
+      const std::unordered_set<MemTable*>* memtables_to_flush = nullptr);
 
   uint64_t GetEarliestMemTableID() const {
     auto& memlist = current_->memlist_;
@@ -381,8 +381,8 @@ class MemTableList {
       const autovector<ColumnFamilyData*>& cfds,
       const autovector<const MutableCFOptions*>& mutable_cf_options_list,
       const autovector<const autovector<MemTable*>*>& mems_list,
-      VersionSet* vset, InstrumentedMutex* mu,
-      const autovector<FileMetaData*>& file_meta,
+      VersionSet* vset, LogsWithPrepTracker* prep_tracker,
+      InstrumentedMutex* mu, const autovector<FileMetaData*>& file_meta,
       autovector<MemTable*>* to_delete, FSDirectory* db_directory,
       LogBuffer* log_buffer);
 
@@ -431,7 +431,8 @@ extern Status InstallMemtableAtomicFlushResults(
     const autovector<ColumnFamilyData*>& cfds,
     const autovector<const MutableCFOptions*>& mutable_cf_options_list,
     const autovector<const autovector<MemTable*>*>& mems_list, VersionSet* vset,
-    InstrumentedMutex* mu, const autovector<FileMetaData*>& file_meta,
+    LogsWithPrepTracker* prep_tracker, InstrumentedMutex* mu,
+    const autovector<FileMetaData*>& file_meta,
     autovector<MemTable*>* to_delete, FSDirectory* db_directory,
     LogBuffer* log_buffer);
 }  // namespace ROCKSDB_NAMESPACE

db/memtable_list_test.cc

@@ -185,8 +185,9 @@ class MemTableListTest : public testing::Test {
     InstrumentedMutex mutex;
     InstrumentedMutexLock l(&mutex);
     return InstallMemtableAtomicFlushResults(
-        &lists, cfds, mutable_cf_options_list, mems_list, &versions, &mutex,
-        file_meta_ptrs, to_delete, nullptr, &log_buffer);
+        &lists, cfds, mutable_cf_options_list, mems_list, &versions,
+        nullptr /* prep_tracker */, &mutex, file_meta_ptrs, to_delete, nullptr,
+        &log_buffer);
   }
 };
 
@@ -199,7 +200,7 @@ TEST_F(MemTableListTest, Empty) {
   ASSERT_FALSE(list.IsFlushPending());
 
   autovector<MemTable*> mems;
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &mems);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &mems);
   ASSERT_EQ(0, mems.size());
 
   autovector<MemTable*> to_delete;
@@ -242,10 +243,10 @@ TEST_F(MemTableListTest, GetTest) {
   mem->Ref();
 
   // Write some keys to this memtable.
-  mem->Add(++seq, kTypeDeletion, "key1", "");
-  mem->Add(++seq, kTypeValue, "key2", "value2");
-  mem->Add(++seq, kTypeValue, "key1", "value1");
-  mem->Add(++seq, kTypeValue, "key2", "value2.2");
+  ASSERT_OK(mem->Add(++seq, kTypeDeletion, "key1", ""));
+  ASSERT_OK(mem->Add(++seq, kTypeValue, "key2", "value2"));
+  ASSERT_OK(mem->Add(++seq, kTypeValue, "key1", "value1"));
+  ASSERT_OK(mem->Add(++seq, kTypeValue, "key2", "value2.2"));
 
   // Fetch the newly written keys
   merge_context.Clear();
@@ -283,8 +284,8 @@ TEST_F(MemTableListTest, GetTest) {
                                 kMaxSequenceNumber, 0 /* column_family_id */);
   mem2->Ref();
 
-  mem2->Add(++seq, kTypeDeletion, "key1", "");
-  mem2->Add(++seq, kTypeValue, "key2", "value2.3");
+  ASSERT_OK(mem2->Add(++seq, kTypeDeletion, "key1", ""));
+  ASSERT_OK(mem2->Add(++seq, kTypeValue, "key2", "value2.3"));
 
   // Add second memtable to list
   list.Add(mem2, &to_delete);
@@ -359,9 +360,9 @@ TEST_F(MemTableListTest, GetFromHistoryTest) {
   mem->Ref();
 
   // Write some keys to this memtable.
-  mem->Add(++seq, kTypeDeletion, "key1", "");
-  mem->Add(++seq, kTypeValue, "key2", "value2");
-  mem->Add(++seq, kTypeValue, "key2", "value2.2");
+  ASSERT_OK(mem->Add(++seq, kTypeDeletion, "key1", ""));
+  ASSERT_OK(mem->Add(++seq, kTypeValue, "key2", "value2"));
+  ASSERT_OK(mem->Add(++seq, kTypeValue, "key2", "value2.2"));
 
   // Fetch the newly written keys
   merge_context.Clear();
@@ -399,7 +400,7 @@ TEST_F(MemTableListTest, GetFromHistoryTest) {
   // Flush this memtable from the list.
   // (It will then be a part of the memtable history).
   autovector<MemTable*> to_flush;
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush);
   ASSERT_EQ(1, to_flush.size());
 
   MutableCFOptions mutable_cf_options(options);
@@ -443,15 +444,15 @@ TEST_F(MemTableListTest, GetFromHistoryTest) {
                                 kMaxSequenceNumber, 0 /* column_family_id */);
   mem2->Ref();
 
-  mem2->Add(++seq, kTypeDeletion, "key1", "");
-  mem2->Add(++seq, kTypeValue, "key3", "value3");
+  ASSERT_OK(mem2->Add(++seq, kTypeDeletion, "key1", ""));
+  ASSERT_OK(mem2->Add(++seq, kTypeValue, "key3", "value3"));
 
   // Add second memtable to list
   list.Add(mem2, &to_delete);
   ASSERT_EQ(0, to_delete.size());
 
   to_flush.clear();
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush);
   ASSERT_EQ(1, to_flush.size());
 
   // Flush second memtable
@@ -554,11 +555,11 @@ TEST_F(MemTableListTest, FlushPendingTest) {
     std::string value;
     MergeContext merge_context;
 
-    mem->Add(++seq, kTypeValue, "key1", ToString(i));
-    mem->Add(++seq, kTypeValue, "keyN" + ToString(i), "valueN");
-    mem->Add(++seq, kTypeValue, "keyX" + ToString(i), "value");
-    mem->Add(++seq, kTypeValue, "keyM" + ToString(i), "valueM");
-    mem->Add(++seq, kTypeDeletion, "keyX" + ToString(i), "");
+    ASSERT_OK(mem->Add(++seq, kTypeValue, "key1", ToString(i)));
+    ASSERT_OK(mem->Add(++seq, kTypeValue, "keyN" + ToString(i), "valueN"));
+    ASSERT_OK(mem->Add(++seq, kTypeValue, "keyX" + ToString(i), "value"));
+    ASSERT_OK(mem->Add(++seq, kTypeValue, "keyM" + ToString(i), "valueM"));
+    ASSERT_OK(mem->Add(++seq, kTypeDeletion, "keyX" + ToString(i), ""));
 
     tables.push_back(mem);
   }
@@ -567,7 +568,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   ASSERT_FALSE(list.IsFlushPending());
   ASSERT_FALSE(list.imm_flush_needed.load(std::memory_order_acquire));
   autovector<MemTable*> to_flush;
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush);
   ASSERT_EQ(0, to_flush.size());
 
   // Request a flush even though there is nothing to flush
@@ -576,7 +577,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   ASSERT_FALSE(list.imm_flush_needed.load(std::memory_order_acquire));
 
   // Attempt to 'flush' to clear request for flush
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush);
   ASSERT_EQ(0, to_flush.size());
   ASSERT_FALSE(list.IsFlushPending());
   ASSERT_FALSE(list.imm_flush_needed.load(std::memory_order_acquire));
@@ -600,7 +601,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   ASSERT_TRUE(list.imm_flush_needed.load(std::memory_order_acquire));
 
   // Pick tables to flush
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush);
   ASSERT_EQ(2, to_flush.size());
   ASSERT_EQ(2, list.NumNotFlushed());
   ASSERT_FALSE(list.IsFlushPending());
@@ -621,7 +622,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   ASSERT_EQ(0, to_delete.size());
 
   // Pick tables to flush
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush);
   ASSERT_EQ(3, to_flush.size());
   ASSERT_EQ(3, list.NumNotFlushed());
   ASSERT_FALSE(list.IsFlushPending());
@@ -629,7 +630,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
 
   // Pick tables to flush again
   autovector<MemTable*> to_flush2;
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush2);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush2);
   ASSERT_EQ(0, to_flush2.size());
   ASSERT_EQ(3, list.NumNotFlushed());
   ASSERT_FALSE(list.IsFlushPending());
@@ -647,7 +648,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   ASSERT_TRUE(list.imm_flush_needed.load(std::memory_order_acquire));
 
   // Pick tables to flush again
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush2);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush2);
   ASSERT_EQ(1, to_flush2.size());
   ASSERT_EQ(4, list.NumNotFlushed());
   ASSERT_FALSE(list.IsFlushPending());
@@ -668,7 +669,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   ASSERT_EQ(0, to_delete.size());
 
   // Pick tables to flush
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush);
   // Should pick 4 of 5 since 1 table has been picked in to_flush2
   ASSERT_EQ(4, to_flush.size());
   ASSERT_EQ(5, list.NumNotFlushed());
@@ -677,7 +678,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
 
   // Pick tables to flush again
   autovector<MemTable*> to_flush3;
-  list.PickMemtablesToFlush(nullptr /* memtable_id */, &to_flush3);
+  list.PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */, &to_flush3);
   ASSERT_EQ(0, to_flush3.size());  // nothing not in progress of being flushed
   ASSERT_EQ(5, list.NumNotFlushed());
   ASSERT_FALSE(list.IsFlushPending());
@@ -738,7 +739,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   autovector<MemTable*> to_flush4;
   list.FlushRequested();
   ASSERT_TRUE(list.HasFlushRequested());
-  list.PickMemtablesToFlush(&memtable_id, &to_flush4);
+  list.PickMemtablesToFlush(memtable_id, &to_flush4);
   ASSERT_TRUE(to_flush4.empty());
   ASSERT_EQ(1, list.NumNotFlushed());
   ASSERT_TRUE(list.imm_flush_needed.load(std::memory_order_acquire));
@@ -749,7 +750,7 @@ TEST_F(MemTableListTest, FlushPendingTest) {
   // equal to 5. Therefore, only tables[5] will be selected.
   memtable_id = 5;
   list.FlushRequested();
-  list.PickMemtablesToFlush(&memtable_id, &to_flush4);
+  list.PickMemtablesToFlush(memtable_id, &to_flush4);
   ASSERT_EQ(1, static_cast<int>(to_flush4.size()));
   ASSERT_EQ(1, list.NumNotFlushed());
   ASSERT_FALSE(list.imm_flush_needed.load(std::memory_order_acquire));
@@ -823,11 +824,11 @@ TEST_F(MemTableListTest, AtomicFlusTest) {
 
       std::string value;
 
-      mem->Add(++seq, kTypeValue, "key1", ToString(i));
-      mem->Add(++seq, kTypeValue, "keyN" + ToString(i), "valueN");
-      mem->Add(++seq, kTypeValue, "keyX" + ToString(i), "value");
-      mem->Add(++seq, kTypeValue, "keyM" + ToString(i), "valueM");
-      mem->Add(++seq, kTypeDeletion, "keyX" + ToString(i), "");
+      ASSERT_OK(mem->Add(++seq, kTypeValue, "key1", ToString(i)));
+      ASSERT_OK(mem->Add(++seq, kTypeValue, "keyN" + ToString(i), "valueN"));
+      ASSERT_OK(mem->Add(++seq, kTypeValue, "keyX" + ToString(i), "value"));
+      ASSERT_OK(mem->Add(++seq, kTypeValue, "keyM" + ToString(i), "valueM"));
+      ASSERT_OK(mem->Add(++seq, kTypeDeletion, "keyX" + ToString(i), ""));
 
       elem.push_back(mem);
     }
@@ -841,7 +842,8 @@ TEST_F(MemTableListTest, AtomicFlusTest) {
     auto* list = lists[i];
     ASSERT_FALSE(list->IsFlushPending());
     ASSERT_FALSE(list->imm_flush_needed.load(std::memory_order_acquire));
-    list->PickMemtablesToFlush(nullptr /* memtable_id */, &flush_candidates[i]);
+    list->PickMemtablesToFlush(port::kMaxUint64 /* memtable_id */,
+                               &flush_candidates[i]);
     ASSERT_EQ(0, flush_candidates[i].size());
   }
   // Request flush even though there is nothing to flush
@@ -871,8 +873,7 @@ TEST_F(MemTableListTest, AtomicFlusTest) {
   // Pick memtables to flush
   for (auto i = 0; i != num_cfs; ++i) {
     flush_candidates[i].clear();
-    lists[i]->PickMemtablesToFlush(&flush_memtable_ids[i],
-                                   &flush_candidates[i]);
+    lists[i]->PickMemtablesToFlush(flush_memtable_ids[i], &flush_candidates[i]);
     ASSERT_EQ(flush_memtable_ids[i] - 0 + 1,
               static_cast<uint64_t>(flush_candidates[i].size()));
   }

db/version_builder.cc

@@ -334,17 +334,23 @@ class VersionBuilder::Rep {
           TEST_SYNC_POINT_CALLBACK("VersionBuilder::CheckConsistency1", &pair);
 #endif
           if (!level_nonzero_cmp_(f1, f2)) {
-            return Status::Corruption("L" + NumberToString(level) +
-                                      " files are not sorted properly");
+            return Status::Corruption(
+                "L" + NumberToString(level) +
+                " files are not sorted properly: files #" +
+                NumberToString(f1->fd.GetNumber()) + ", #" +
+                NumberToString(f2->fd.GetNumber()));
           }
 
           // Make sure there is no overlap in levels > 0
           if (vstorage->InternalComparator()->Compare(f1->largest,
                                                       f2->smallest) >= 0) {
             return Status::Corruption(
-                "L" + NumberToString(level) + " have overlapping ranges " +
-                (f1->largest).DebugString(true) + " vs. " +
-                (f2->smallest).DebugString(true));
+                "L" + NumberToString(level) +
+                " have overlapping ranges: file #" +
+                NumberToString(f1->fd.GetNumber()) +
+                " largest key: " + (f1->largest).DebugString(true) +
+                " vs. file #" + NumberToString(f2->fd.GetNumber()) +
+                " smallest key: " + (f2->smallest).DebugString(true));
           }
         }
       }

db/version_edit.cc

@@ -96,6 +96,7 @@ void VersionEdit::Clear() {
   column_family_name_.clear();
   is_in_atomic_group_ = false;
   remaining_entries_ = 0;
+  full_history_ts_low_.clear();
 }
 
 bool VersionEdit::EncodeTo(std::string* dst) const {
@@ -225,13 +226,17 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
   }
 
   for (const auto& wal_addition : wal_additions_) {
-    PutVarint32(dst, kWalAddition);
-    wal_addition.EncodeTo(dst);
+    PutVarint32(dst, kWalAddition2);
+    std::string encoded;
+    wal_addition.EncodeTo(&encoded);
+    PutLengthPrefixedSlice(dst, encoded);
   }
 
   if (!wal_deletion_.IsEmpty()) {
-    PutVarint32(dst, kWalDeletion);
-    wal_deletion_.EncodeTo(dst);
+    PutVarint32(dst, kWalDeletion2);
+    std::string encoded;
+    wal_deletion_.EncodeTo(&encoded);
+    PutLengthPrefixedSlice(dst, encoded);
   }
 
   // 0 is default and does not need to be explicitly written
@@ -252,6 +257,11 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
     PutVarint32(dst, kInAtomicGroup);
     PutVarint32(dst, remaining_entries_);
   }
+
+  if (HasFullHistoryTsLow()) {
+    PutVarint32(dst, kFullHistoryTsLow);
+    PutLengthPrefixedSlice(dst, full_history_ts_low_);
+  }
   return true;
 }
 
@@ -369,6 +379,11 @@ const char* VersionEdit::DecodeNewFile4From(Slice* input) {
 
 Status VersionEdit::DecodeFrom(const Slice& src) {
   Clear();
+#ifndef NDEBUG
+  bool ignore_ignorable_tags = false;
+  TEST_SYNC_POINT_CALLBACK("VersionEdit::EncodeTo:IgnoreIgnorableTags",
+                           &ignore_ignorable_tags);
+#endif
   Slice input = src;
   const char* msg = nullptr;
   uint32_t tag = 0;
@@ -379,6 +394,11 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
   Slice str;
   InternalKey key;
   while (msg == nullptr && GetVarint32(&input, &tag)) {
+#ifndef NDEBUG
+    if (ignore_ignorable_tags && tag > kTagSafeIgnoreMask) {
+      tag = kTagSafeIgnoreMask;
+    }
+#endif
     switch (tag) {
       case kDbId:
         if (GetLengthPrefixedSlice(&input, &str)) {
@@ -569,6 +589,23 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
         break;
       }
 
+      case kWalAddition2: {
+        Slice encoded;
+        if (!GetLengthPrefixedSlice(&input, &encoded)) {
+          msg = "WalAddition not prefixed by length";
+          break;
+        }
+
+        WalAddition wal_addition;
+        const Status s = wal_addition.DecodeFrom(&encoded);
+        if (!s.ok()) {
+          return s;
+        }
+
+        wal_additions_.emplace_back(std::move(wal_addition));
+        break;
+      }
+
       case kWalDeletion: {
         WalDeletion wal_deletion;
         const Status s = wal_deletion.DecodeFrom(&input);
@@ -580,6 +617,23 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
         break;
       }
 
+      case kWalDeletion2: {
+        Slice encoded;
+        if (!GetLengthPrefixedSlice(&input, &encoded)) {
+          msg = "WalDeletion not prefixed by length";
+          break;
+        }
+
+        WalDeletion wal_deletion;
+        const Status s = wal_deletion.DecodeFrom(&encoded);
+        if (!s.ok()) {
+          return s;
+        }
+
+        wal_deletion_ = std::move(wal_deletion);
+        break;
+      }
+
       case kColumnFamily:
         if (!GetVarint32(&input, &column_family_)) {
           if (!msg) {
@@ -612,6 +666,16 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
         }
         break;
 
+      case kFullHistoryTsLow:
+        if (!GetLengthPrefixedSlice(&input, &str)) {
+          msg = "full_history_ts_low";
+        } else if (str.empty()) {
+          msg = "full_history_ts_low: empty";
+        } else {
+          full_history_ts_low_.assign(str.data(), str.size());
+        }
+        break;
+
       default:
         if (tag & kTagSafeIgnoreMask) {
           // Tag from future which can be safely ignored.
@@ -744,6 +808,10 @@ std::string VersionEdit::DebugString(bool hex_key) const {
     AppendNumberTo(&r, remaining_entries_);
     r.append(" entries remains");
   }
+  if (HasFullHistoryTsLow()) {
+    r.append("\n FullHistoryTsLow: ");
+    r.append(Slice(full_history_ts_low_).ToString(hex_key));
+  }
   r.append("\n}\n");
   return r;
 }
@@ -873,6 +941,10 @@ std::string VersionEdit::DebugJSON(int edit_num, bool hex_key) const {
     jw << "AtomicGroup" << remaining_entries_;
   }
 
+  if (HasFullHistoryTsLow()) {
+    jw << "FullHistoryTsLow" << Slice(full_history_ts_low_).ToString(hex_key);
+  }
+
   jw.EndObject();
 
   return jw.Get();

db/version_edit.h

@@ -61,6 +61,9 @@ enum Tag : uint32_t {
   kBlobFileGarbage,
   kWalAddition,
   kWalDeletion,
+  kFullHistoryTsLow,
+  kWalAddition2,
+  kWalDeletion2,
 };
 
 enum NewFileCustomTag : uint32_t {
@@ -524,6 +527,16 @@ class VersionEdit {
   bool IsInAtomicGroup() const { return is_in_atomic_group_; }
   uint32_t GetRemainingEntries() const { return remaining_entries_; }
 
+  bool HasFullHistoryTsLow() const { return !full_history_ts_low_.empty(); }
+  const std::string& GetFullHistoryTsLow() const {
+    assert(HasFullHistoryTsLow());
+    return full_history_ts_low_;
+  }
+  void SetFullHistoryTsLow(std::string full_history_ts_low) {
+    assert(!full_history_ts_low.empty());
+    full_history_ts_low_ = std::move(full_history_ts_low);
+  }
+
   // return true on success.
   bool EncodeTo(std::string* dst) const;
   Status DecodeFrom(const Slice& src);
@@ -586,6 +599,8 @@ class VersionEdit {
 
   bool is_in_atomic_group_ = false;
   uint32_t remaining_entries_ = 0;
+
+  std::string full_history_ts_low_;
 };
 
 }  // namespace ROCKSDB_NAMESPACE

db/version_edit_handler.cc

@@ -520,6 +520,10 @@ Status VersionEditHandler::ExtractInfoFromVersionEdit(ColumnFamilyData* cfd,
           cfd->user_comparator()->Name(),
           "does not match existing comparator " + edit.comparator_);
     }
+    if (edit.HasFullHistoryTsLow()) {
+      const std::string& new_ts = edit.GetFullHistoryTsLow();
+      cfd->SetFullHistoryTsLow(new_ts);
+    }
   }
 
   if (s.ok()) {

db/version_edit_test.cc

@@ -8,8 +8,10 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/version_edit.h"
+
 #include "test_util/sync_point.h"
 #include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/coding.h"
 #include "util/string_util.h"
 
@@ -322,14 +324,22 @@ TEST_F(VersionEditTest, AddWalEncodeDecode) {
   TestEncodeDecode(edit);
 }
 
+static std::string PrefixEncodedWalAdditionWithLength(
+    const std::string& encoded) {
+  std::string ret;
+  PutVarint32(&ret, Tag::kWalAddition2);
+  PutLengthPrefixedSlice(&ret, encoded);
+  return ret;
+}
+
 TEST_F(VersionEditTest, AddWalDecodeBadLogNumber) {
   std::string encoded;
-  PutVarint32(&encoded, Tag::kWalAddition);
 
   {
     // No log number.
+    std::string encoded_edit = PrefixEncodedWalAdditionWithLength(encoded);
     VersionEdit edit;
-    Status s = edit.DecodeFrom(encoded);
+    Status s = edit.DecodeFrom(encoded_edit);
     ASSERT_TRUE(s.IsCorruption());
     ASSERT_TRUE(s.ToString().find("Error decoding WAL log number") !=
                 std::string::npos)
@@ -343,8 +353,10 @@ TEST_F(VersionEditTest, AddWalDecodeBadLogNumber) {
     unsigned char* ptr = reinterpret_cast<unsigned char*>(&c);
     *ptr = 128;
     encoded.append(1, c);
+
+    std::string encoded_edit = PrefixEncodedWalAdditionWithLength(encoded);
     VersionEdit edit;
-    Status s = edit.DecodeFrom(encoded);
+    Status s = edit.DecodeFrom(encoded_edit);
     ASSERT_TRUE(s.IsCorruption());
     ASSERT_TRUE(s.ToString().find("Error decoding WAL log number") !=
                 std::string::npos)
@@ -356,14 +368,14 @@ TEST_F(VersionEditTest, AddWalDecodeBadTag) {
   constexpr WalNumber kLogNumber = 100;
   constexpr uint64_t kSizeInBytes = 100;
 
-  std::string encoded_without_tag;
-  PutVarint32(&encoded_without_tag, Tag::kWalAddition);
-  PutVarint64(&encoded_without_tag, kLogNumber);
+  std::string encoded;
+  PutVarint64(&encoded, kLogNumber);
 
   {
     // No tag.
+    std::string encoded_edit = PrefixEncodedWalAdditionWithLength(encoded);
     VersionEdit edit;
-    Status s = edit.DecodeFrom(encoded_without_tag);
+    Status s = edit.DecodeFrom(encoded_edit);
     ASSERT_TRUE(s.IsCorruption());
     ASSERT_TRUE(s.ToString().find("Error decoding tag") != std::string::npos)
         << s.ToString();
@@ -371,12 +383,15 @@ TEST_F(VersionEditTest, AddWalDecodeBadTag) {
 
   {
     // Only has size tag, no terminate tag.
-    std::string encoded_with_size = encoded_without_tag;
+    std::string encoded_with_size = encoded;
     PutVarint32(&encoded_with_size,
                 static_cast<uint32_t>(WalAdditionTag::kSyncedSize));
     PutVarint64(&encoded_with_size, kSizeInBytes);
+
+    std::string encoded_edit =
+        PrefixEncodedWalAdditionWithLength(encoded_with_size);
     VersionEdit edit;
-    Status s = edit.DecodeFrom(encoded_with_size);
+    Status s = edit.DecodeFrom(encoded_edit);
     ASSERT_TRUE(s.IsCorruption());
     ASSERT_TRUE(s.ToString().find("Error decoding tag") != std::string::npos)
         << s.ToString();
@@ -384,11 +399,14 @@ TEST_F(VersionEditTest, AddWalDecodeBadTag) {
 
   {
     // Only has terminate tag.
-    std::string encoded_with_terminate = encoded_without_tag;
+    std::string encoded_with_terminate = encoded;
     PutVarint32(&encoded_with_terminate,
                 static_cast<uint32_t>(WalAdditionTag::kTerminate));
+
+    std::string encoded_edit =
+        PrefixEncodedWalAdditionWithLength(encoded_with_terminate);
     VersionEdit edit;
-    ASSERT_OK(edit.DecodeFrom(encoded_with_terminate));
+    ASSERT_OK(edit.DecodeFrom(encoded_edit));
     auto& wal_addition = edit.GetWalAdditions()[0];
     ASSERT_EQ(wal_addition.GetLogNumber(), kLogNumber);
     ASSERT_FALSE(wal_addition.GetMetadata().HasSyncedSize());
@@ -399,15 +417,15 @@ TEST_F(VersionEditTest, AddWalDecodeNoSize) {
   constexpr WalNumber kLogNumber = 100;
 
   std::string encoded;
-  PutVarint32(&encoded, Tag::kWalAddition);
   PutVarint64(&encoded, kLogNumber);
   PutVarint32(&encoded, static_cast<uint32_t>(WalAdditionTag::kSyncedSize));
   // No real size after the size tag.
 
   {
     // Without terminate tag.
+    std::string encoded_edit = PrefixEncodedWalAdditionWithLength(encoded);
     VersionEdit edit;
-    Status s = edit.DecodeFrom(encoded);
+    Status s = edit.DecodeFrom(encoded_edit);
     ASSERT_TRUE(s.IsCorruption());
     ASSERT_TRUE(s.ToString().find("Error decoding WAL file size") !=
                 std::string::npos)
@@ -417,8 +435,10 @@ TEST_F(VersionEditTest, AddWalDecodeNoSize) {
   {
     // With terminate tag.
     PutVarint32(&encoded, static_cast<uint32_t>(WalAdditionTag::kTerminate));
+
+    std::string encoded_edit = PrefixEncodedWalAdditionWithLength(encoded);
     VersionEdit edit;
-    Status s = edit.DecodeFrom(encoded);
+    Status s = edit.DecodeFrom(encoded_edit);
     ASSERT_TRUE(s.IsCorruption());
     // The terminate tag is misunderstood as the size.
     ASSERT_TRUE(s.ToString().find("Error decoding tag") != std::string::npos)
@@ -505,6 +525,70 @@ TEST_F(VersionEditTest, DeleteWalDebug) {
   ASSERT_EQ(edit.DebugJSON(4, true), expected_json);
 }
 
+TEST_F(VersionEditTest, FullHistoryTsLow) {
+  VersionEdit edit;
+  ASSERT_FALSE(edit.HasFullHistoryTsLow());
+  std::string ts = test::EncodeInt(0);
+  edit.SetFullHistoryTsLow(ts);
+  TestEncodeDecode(edit);
+}
+
+// Tests that if RocksDB is downgraded, the new types of VersionEdits
+// that have a tag larger than kTagSafeIgnoreMask can be safely ignored.
+TEST_F(VersionEditTest, IgnorableTags) {
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionEdit::EncodeTo:IgnoreIgnorableTags", [&](void* arg) {
+        bool* ignore = static_cast<bool*>(arg);
+        *ignore = true;
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  constexpr uint64_t kPrevLogNumber = 100;
+  constexpr uint64_t kLogNumber = 200;
+  constexpr uint64_t kNextFileNumber = 300;
+  constexpr uint64_t kColumnFamilyId = 400;
+
+  VersionEdit edit;
+  // Add some ignorable entries.
+  for (int i = 0; i < 2; i++) {
+    edit.AddWal(i + 1, WalMetadata(i + 2));
+  }
+  edit.SetDBId("db_id");
+  // Add unignorable entries.
+  edit.SetPrevLogNumber(kPrevLogNumber);
+  edit.SetLogNumber(kLogNumber);
+  // Add more ignorable entries.
+  edit.DeleteWalsBefore(100);
+  // Add unignorable entry.
+  edit.SetNextFile(kNextFileNumber);
+  // Add more ignorable entries.
+  edit.SetFullHistoryTsLow("ts");
+  // Add unignorable entry.
+  edit.SetColumnFamily(kColumnFamilyId);
+
+  std::string encoded;
+  ASSERT_TRUE(edit.EncodeTo(&encoded));
+
+  VersionEdit decoded;
+  ASSERT_OK(decoded.DecodeFrom(encoded));
+
+  // Check that all ignorable entries are ignored.
+  ASSERT_FALSE(decoded.HasDbId());
+  ASSERT_FALSE(decoded.HasFullHistoryTsLow());
+  ASSERT_FALSE(decoded.IsWalAddition());
+  ASSERT_FALSE(decoded.IsWalDeletion());
+  ASSERT_TRUE(decoded.GetWalAdditions().empty());
+  ASSERT_TRUE(decoded.GetWalDeletion().IsEmpty());
+
+  // Check that unignorable entries are still present.
+  ASSERT_EQ(edit.GetPrevLogNumber(), kPrevLogNumber);
+  ASSERT_EQ(edit.GetLogNumber(), kLogNumber);
+  ASSERT_EQ(edit.GetNextFile(), kNextFileNumber);
+  ASSERT_EQ(edit.GetColumnFamily(), kColumnFamilyId);
+
+  SyncPoint::GetInstance()->DisableProcessing();
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/version_set.cc

@@ -1791,9 +1791,8 @@ Version::Version(ColumnFamilyData* column_family_data, VersionSet* vset,
       io_tracer_(io_tracer) {}
 
 Status Version::GetBlob(const ReadOptions& read_options, const Slice& user_key,
+                        const Slice& blob_index_slice,
                         PinnableSlice* value) const {
-  assert(value);
-
   if (read_options.read_tier == kBlockCacheTier) {
     return Status::Incomplete("Cannot read blob: no disk I/O allowed");
   }
@@ -1801,12 +1800,20 @@ Status Version::GetBlob(const ReadOptions& read_options, const Slice& user_key,
   BlobIndex blob_index;
 
   {
-    Status s = blob_index.DecodeFrom(*value);
+    Status s = blob_index.DecodeFrom(blob_index_slice);
     if (!s.ok()) {
       return s;
     }
   }
 
+  return GetBlob(read_options, user_key, blob_index, value);
+}
+
+Status Version::GetBlob(const ReadOptions& read_options, const Slice& user_key,
+                        const BlobIndex& blob_index,
+                        PinnableSlice* value) const {
+  assert(value);
+
   if (blob_index.HasTTL() || blob_index.IsInlined()) {
     return Status::Corruption("Unexpected TTL/inlined blob index");
   }
@@ -1931,9 +1938,20 @@ void Version::Get(const ReadOptions& read_options, const LookupKey& k,
         // TODO: update per-level perfcontext user_key_return_count for kMerge
         break;
       case GetContext::kFound:
+        if (fp.GetHitFileLevel() == 0) {
+          RecordTick(db_statistics_, GET_HIT_L0);
+        } else if (fp.GetHitFileLevel() == 1) {
+          RecordTick(db_statistics_, GET_HIT_L1);
+        } else if (fp.GetHitFileLevel() >= 2) {
+          RecordTick(db_statistics_, GET_HIT_L2_AND_UP);
+        }
+
+        PERF_COUNTER_BY_LEVEL_ADD(user_key_return_count, 1,
+                                  fp.GetHitFileLevel());
+
         if (is_blob_index) {
           if (do_merge && value) {
-            *status = GetBlob(read_options, user_key, value);
+            *status = GetBlob(read_options, user_key, *value, value);
             if (!status->ok()) {
               if (status->IsIncomplete()) {
                 get_context.MarkKeyMayExist();
@@ -1943,15 +1961,6 @@ void Version::Get(const ReadOptions& read_options, const LookupKey& k,
           }
         }
 
-        if (fp.GetHitFileLevel() == 0) {
-          RecordTick(db_statistics_, GET_HIT_L0);
-        } else if (fp.GetHitFileLevel() == 1) {
-          RecordTick(db_statistics_, GET_HIT_L1);
-        } else if (fp.GetHitFileLevel() >= 2) {
-          RecordTick(db_statistics_, GET_HIT_L2_AND_UP);
-        }
-        PERF_COUNTER_BY_LEVEL_ADD(user_key_return_count, 1,
-                                  fp.GetHitFileLevel());
         return;
       case GetContext::kDeleted:
         // Use empty error message for speed
@@ -2001,7 +2010,7 @@ void Version::Get(const ReadOptions& read_options, const LookupKey& k,
 }
 
 void Version::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
-                       ReadCallback* callback, bool* is_blob) {
+                       ReadCallback* callback) {
   PinnedIteratorsManager pinned_iters_mgr;
 
   // Pin blocks that we read to hold merge operands
@@ -2026,7 +2035,7 @@ void Version::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
         iter->ukey_with_ts, iter->value, iter->timestamp, nullptr,
         &(iter->merge_context), true, &iter->max_covering_tombstone_seq,
         this->env_, nullptr, merge_operator_ ? &pinned_iters_mgr : nullptr,
-        callback, is_blob, tracing_mget_id);
+        callback, &iter->is_blob_index, tracing_mget_id);
     // MergeInProgress status, if set, has been transferred to the get_context
     // state, so we set status to ok here. From now on, the iter status will
     // be used for IO errors, and get_context state will be used for any
@@ -2128,10 +2137,27 @@ void Version::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
           } else if (fp.GetHitFileLevel() >= 2) {
             RecordTick(db_statistics_, GET_HIT_L2_AND_UP);
           }
+
           PERF_COUNTER_BY_LEVEL_ADD(user_key_return_count, 1,
                                     fp.GetHitFileLevel());
-          file_range.AddValueSize(iter->value->size());
+
           file_range.MarkKeyDone(iter);
+
+          if (iter->is_blob_index) {
+            if (iter->value) {
+              *status = GetBlob(read_options, iter->ukey_with_ts, *iter->value,
+                                iter->value);
+              if (!status->ok()) {
+                if (status->IsIncomplete()) {
+                  get_context.MarkKeyMayExist();
+                }
+
+                continue;
+              }
+            }
+          }
+
+          file_range.AddValueSize(iter->value->size());
           if (file_range.GetValueSize() > read_options.value_size_soft_limit) {
             s = Status::Aborted();
             break;
@@ -4037,7 +4063,9 @@ Status VersionSet::ProcessManifestWrites(
     }
     for (const auto* cfd : *column_family_set_) {
       assert(curr_state.find(cfd->GetID()) == curr_state.end());
-      curr_state[cfd->GetID()] = {cfd->GetLogNumber()};
+      curr_state.emplace(std::make_pair(
+          cfd->GetID(),
+          MutableCFState(cfd->GetLogNumber(), cfd->GetFullHistoryTsLow())));
     }
 
     for (const auto& wal : wals_.GetWals()) {
@@ -4220,23 +4248,23 @@ Status VersionSet::ProcessManifestWrites(
       // Each version in versions corresponds to a column family.
       // For each column family, update its log number indicating that logs
       // with number smaller than this should be ignored.
-      for (const auto version : versions) {
-        uint64_t max_log_number_in_batch = 0;
-        uint32_t cf_id = version->cfd_->GetID();
-        for (const auto& e : batch_edits) {
-          if (e->has_log_number_ && e->column_family_ == cf_id) {
-            max_log_number_in_batch =
-                std::max(max_log_number_in_batch, e->log_number_);
+      uint64_t last_min_log_number_to_keep = 0;
+      for (const auto& e : batch_edits) {
+        ColumnFamilyData* cfd = nullptr;
+        if (!e->IsColumnFamilyManipulation()) {
+          cfd = column_family_set_->GetColumnFamily(e->column_family_);
+          // e would not have been added to batch_edits if its corresponding
+          // column family is dropped.
+          assert(cfd);
+        }
+        if (cfd) {
+          if (e->has_log_number_ && e->log_number_ > cfd->GetLogNumber()) {
+            cfd->SetLogNumber(e->log_number_);
+          }
+          if (e->HasFullHistoryTsLow()) {
+            cfd->SetFullHistoryTsLow(e->GetFullHistoryTsLow());
           }
         }
-        if (max_log_number_in_batch != 0) {
-          assert(version->cfd_->GetLogNumber() <= max_log_number_in_batch);
-          version->cfd_->SetLogNumber(max_log_number_in_batch);
-        }
-      }
-
-      uint64_t last_min_log_number_to_keep = 0;
-      for (auto& e : batch_edits) {
         if (e->has_min_log_number_to_keep_) {
           last_min_log_number_to_keep =
               std::max(last_min_log_number_to_keep, e->min_log_number_to_keep_);
@@ -4582,6 +4610,10 @@ Status VersionSet::ExtractInfoFromVersionEdit(
           cfd->user_comparator()->Name(),
           "does not match existing comparator " + from_edit.comparator_);
     }
+    if (from_edit.HasFullHistoryTsLow()) {
+      const std::string& new_ts = from_edit.GetFullHistoryTsLow();
+      cfd->SetFullHistoryTsLow(new_ts);
+    }
   }
 
   if (from_edit.has_prev_log_number_) {
@@ -5272,6 +5304,21 @@ Status VersionSet::WriteCurrentStateToManifest(
       assert(iter != curr_state.end());
       uint64_t log_number = iter->second.log_number;
       edit.SetLogNumber(log_number);
+
+      if (cfd->GetID() == 0) {
+        // min_log_number_to_keep is for the whole db, not for specific column family.
+        // So it does not need to be set for every column family, just need to be set once.
+        // Since default CF can never be dropped, we set the min_log to the default CF here.
+        uint64_t min_log = min_log_number_to_keep_2pc();
+        if (min_log != 0) {
+          edit.SetMinLogNumberToKeep(min_log);
+        }
+      }
+
+      const std::string& full_history_ts_low = iter->second.full_history_ts_low;
+      if (!full_history_ts_low.empty()) {
+        edit.SetFullHistoryTsLow(full_history_ts_low);
+      }
       std::string record;
       if (!edit.EncodeTo(&record)) {
         return Status::Corruption(

db/version_set.h

@@ -26,6 +26,7 @@
 #include <memory>
 #include <set>
 #include <string>
+#include <unordered_set>
 #include <utility>
 #include <vector>
 
@@ -59,6 +60,7 @@ namespace log {
 class Writer;
 }
 
+class BlobIndex;
 class Compaction;
 class LogBuffer;
 class LookupKey;
@@ -680,7 +682,19 @@ class Version {
            bool* is_blob = nullptr, bool do_merge = true);
 
   void MultiGet(const ReadOptions&, MultiGetRange* range,
-                ReadCallback* callback = nullptr, bool* is_blob = nullptr);
+                ReadCallback* callback = nullptr);
+
+  // Interprets blob_index_slice as a blob reference, and (assuming the
+  // corresponding blob file is part of this Version) retrieves the blob and
+  // saves it in *value.
+  // REQUIRES: blob_index_slice stores an encoded blob reference
+  Status GetBlob(const ReadOptions& read_options, const Slice& user_key,
+                 const Slice& blob_index_slice, PinnableSlice* value) const;
+
+  // Retrieves a blob using a blob reference and saves it in *value,
+  // assuming the corresponding blob file is part of this Version.
+  Status GetBlob(const ReadOptions& read_options, const Slice& user_key,
+                 const BlobIndex& blob_index, PinnableSlice* value) const;
 
   // Loads some stats information from files. Call without mutex held. It needs
   // to be called before applying the version to the version set.
@@ -777,13 +791,6 @@ class Version {
     return storage_info_.user_comparator_;
   }
 
-  // Interprets *value as a blob reference, and (assuming the corresponding
-  // blob file is part of this Version) retrieves the blob and saves it in
-  // *value, replacing the blob reference.
-  // REQUIRES: *value stores an encoded blob reference
-  Status GetBlob(const ReadOptions& read_options, const Slice& user_key,
-                 PinnableSlice* value) const;
-
   // Returns true if the filter blocks in the specified level will not be
   // checked during read operations. In certain cases (trivial move or preload),
   // the filter block may already be cached, but we still do not access it such
@@ -1120,10 +1127,28 @@ class VersionSet {
     return PreComputeMinLogNumberWithUnflushedData(nullptr);
   }
   // Returns the minimum log number which still has data not flushed to any SST
+  // file.
+  // Empty column families' log number is considered to be
+  // new_log_number_for_empty_cf.
+  uint64_t PreComputeMinLogNumberWithUnflushedData(
+      uint64_t new_log_number_for_empty_cf) const {
+    uint64_t min_log_num = port::kMaxUint64;
+    for (auto cfd : *column_family_set_) {
+      // It's safe to ignore dropped column families here:
+      // cfd->IsDropped() becomes true after the drop is persisted in MANIFEST.
+      uint64_t num =
+          cfd->IsEmpty() ? new_log_number_for_empty_cf : cfd->GetLogNumber();
+      if (min_log_num > num && !cfd->IsDropped()) {
+        min_log_num = num;
+      }
+    }
+    return min_log_num;
+  }
+  // Returns the minimum log number which still has data not flushed to any SST
   // file, except data from `cfd_to_skip`.
   uint64_t PreComputeMinLogNumberWithUnflushedData(
       const ColumnFamilyData* cfd_to_skip) const {
-    uint64_t min_log_num = std::numeric_limits<uint64_t>::max();
+    uint64_t min_log_num = port::kMaxUint64;
     for (auto cfd : *column_family_set_) {
       if (cfd == cfd_to_skip) {
         continue;
@@ -1136,6 +1161,23 @@ class VersionSet {
     }
     return min_log_num;
   }
+  // Returns the minimum log number which still has data not flushed to any SST
+  // file, except data from `cfds_to_skip`.
+  uint64_t PreComputeMinLogNumberWithUnflushedData(
+      const std::unordered_set<const ColumnFamilyData*>& cfds_to_skip) const {
+    uint64_t min_log_num = port::kMaxUint64;
+    for (auto cfd : *column_family_set_) {
+      if (cfds_to_skip.count(cfd)) {
+        continue;
+      }
+      // It's safe to ignore dropped column families here:
+      // cfd->IsDropped() becomes true after the drop is persisted in MANIFEST.
+      if (min_log_num > cfd->GetLogNumber() && !cfd->IsDropped()) {
+        min_log_num = cfd->GetLogNumber();
+      }
+    }
+    return min_log_num;
+  }
 
   // Create an iterator that reads over the compaction inputs for "*c".
   // The caller should delete the iterator when no longer needed.
@@ -1254,6 +1296,11 @@ class VersionSet {
 
   struct MutableCFState {
     uint64_t log_number;
+    std::string full_history_ts_low;
+
+    explicit MutableCFState() = default;
+    explicit MutableCFState(uint64_t _log_number, std::string ts_low)
+        : log_number(_log_number), full_history_ts_low(std::move(ts_low)) {}
   };
 
   // Save current contents to *log
@@ -1303,7 +1350,7 @@ class VersionSet {
   std::string db_id_;
   const ImmutableDBOptions* const db_options_;
   std::atomic<uint64_t> next_file_number_;
-  // Any log number equal or lower than this should be ignored during recovery,
+  // Any WAL number smaller than this should be ignored during recovery,
   // and is qualified for being deleted in 2PC mode. In non-2PC mode, this
   // number is ignored.
   std::atomic<uint64_t> min_log_number_to_keep_2pc_ = {0};

db/version_set_test.cc

@@ -826,6 +826,14 @@ class VersionSetTestBase {
               versions_->GetColumnFamilySet()->NumberOfColumnFamilies());
   }
 
+  void ReopenDB() {
+    versions_.reset(
+        new VersionSet(dbname_, &db_options_, env_options_, table_cache_.get(),
+                       &write_buffer_manager_, &write_controller_,
+                       /*block_cache_tracer=*/nullptr, /*io_tracer=*/nullptr));
+    EXPECT_OK(versions_->Recover(column_families_, false));
+  }
+
   void VerifyManifest(std::string* manifest_path) const {
     assert(manifest_path != nullptr);
     uint64_t manifest_file_number = 0;
@@ -869,6 +877,28 @@ class VersionSetTestBase {
     mutex_.Unlock();
   }
 
+  ColumnFamilyData* CreateColumnFamily(const std::string& cf_name,
+                                       const ColumnFamilyOptions& cf_options) {
+    VersionEdit new_cf;
+    new_cf.AddColumnFamily(cf_name);
+    uint32_t new_id = versions_->GetColumnFamilySet()->GetNextColumnFamilyID();
+    new_cf.SetColumnFamily(new_id);
+    new_cf.SetLogNumber(0);
+    new_cf.SetComparatorName(cf_options.comparator->Name());
+    Status s;
+    mutex_.Lock();
+    s = versions_->LogAndApply(/*column_family_data=*/nullptr,
+                               MutableCFOptions(cf_options), &new_cf, &mutex_,
+                               /*db_directory=*/nullptr,
+                               /*new_descriptor_log=*/false, &cf_options);
+    mutex_.Unlock();
+    EXPECT_OK(s);
+    ColumnFamilyData* cfd =
+        versions_->GetColumnFamilySet()->GetColumnFamily(cf_name);
+    EXPECT_NE(nullptr, cfd);
+    return cfd;
+  }
+
   Env* mem_env_;
   Env* env_;
   std::shared_ptr<Env> env_guard_;
@@ -1667,6 +1697,104 @@ TEST_F(VersionSetTest, AtomicGroupWithWalEdits) {
   }
 }
 
+class VersionSetWithTimestampTest : public VersionSetTest {
+ public:
+  static const std::string kNewCfName;
+
+  explicit VersionSetWithTimestampTest() : VersionSetTest() {}
+
+  void SetUp() override {
+    NewDB();
+    Options options;
+    options.comparator = test::ComparatorWithU64Ts();
+    cfd_ = CreateColumnFamily(kNewCfName, options);
+    EXPECT_NE(nullptr, cfd_);
+    EXPECT_NE(nullptr, cfd_->GetLatestMutableCFOptions());
+    column_families_.emplace_back(kNewCfName, options);
+  }
+
+  void TearDown() override {
+    for (auto* e : edits_) {
+      delete e;
+    }
+    edits_.clear();
+  }
+
+  void GenVersionEditsToSetFullHistoryTsLow(
+      const std::vector<uint64_t>& ts_lbs) {
+    for (const auto ts_lb : ts_lbs) {
+      VersionEdit* edit = new VersionEdit;
+      edit->SetColumnFamily(cfd_->GetID());
+      std::string ts_str = test::EncodeInt(ts_lb);
+      edit->SetFullHistoryTsLow(ts_str);
+      edits_.emplace_back(edit);
+    }
+  }
+
+  void VerifyFullHistoryTsLow(uint64_t expected_ts_low) {
+    std::unique_ptr<VersionSet> vset(
+        new VersionSet(dbname_, &db_options_, env_options_, table_cache_.get(),
+                       &write_buffer_manager_, &write_controller_,
+                       /*block_cache_tracer=*/nullptr, /*io_tracer=*/nullptr));
+    ASSERT_OK(vset->Recover(column_families_, /*read_only=*/false,
+                            /*db_id=*/nullptr));
+    for (auto* cfd : *(vset->GetColumnFamilySet())) {
+      ASSERT_NE(nullptr, cfd);
+      if (cfd->GetName() == kNewCfName) {
+        ASSERT_EQ(test::EncodeInt(expected_ts_low), cfd->GetFullHistoryTsLow());
+      } else {
+        ASSERT_TRUE(cfd->GetFullHistoryTsLow().empty());
+      }
+    }
+  }
+
+  void DoTest(const std::vector<uint64_t>& ts_lbs) {
+    if (ts_lbs.empty()) {
+      return;
+    }
+
+    GenVersionEditsToSetFullHistoryTsLow(ts_lbs);
+
+    Status s;
+    mutex_.Lock();
+    s = versions_->LogAndApply(cfd_, *(cfd_->GetLatestMutableCFOptions()),
+                               edits_, &mutex_);
+    mutex_.Unlock();
+    ASSERT_OK(s);
+    VerifyFullHistoryTsLow(*std::max_element(ts_lbs.begin(), ts_lbs.end()));
+  }
+
+ protected:
+  ColumnFamilyData* cfd_{nullptr};
+  // edits_ must contain and own pointers to heap-alloc VersionEdit objects.
+  autovector<VersionEdit*> edits_;
+};
+
+const std::string VersionSetWithTimestampTest::kNewCfName("new_cf");
+
+TEST_F(VersionSetWithTimestampTest, SetFullHistoryTsLbOnce) {
+  constexpr uint64_t kTsLow = 100;
+  DoTest({kTsLow});
+}
+
+// Simulate the application increasing full_history_ts_low.
+TEST_F(VersionSetWithTimestampTest, IncreaseFullHistoryTsLb) {
+  const std::vector<uint64_t> ts_lbs = {100, 101, 102, 103};
+  DoTest(ts_lbs);
+}
+
+// Simulate the application trying to decrease full_history_ts_low
+// unsuccessfully. If the application calls public API sequentially to
+// decrease the lower bound ts, RocksDB will return an InvalidArgument
+// status before involving VersionSet. Only when multiple threads trying
+// to decrease the lower bound concurrently will this case ever happen. Even
+// so, the lower bound cannot be decreased. The application will be notified
+// via return value of the API.
+TEST_F(VersionSetWithTimestampTest, TryDecreaseFullHistoryTsLb) {
+  const std::vector<uint64_t> ts_lbs = {103, 102, 101, 100};
+  DoTest(ts_lbs);
+}
+
 class VersionSetAtomicGroupTest : public VersionSetTestBase,
                                   public testing::Test {
  public:
@@ -2164,10 +2292,7 @@ TEST_P(VersionSetTestDropOneCF, HandleDroppedColumnFamilyInAtomicGroup) {
   mutex_.Unlock();
   ASSERT_OK(s);
   ASSERT_EQ(1, called);
-  if (cfd_to_drop->Unref()) {
-    delete cfd_to_drop;
-    cfd_to_drop = nullptr;
-  }
+  cfd_to_drop->UnrefAndTryDelete();
 }
 
 INSTANTIATE_TEST_CASE_P(
@@ -2881,6 +3006,27 @@ TEST_F(VersionSetTestMissingFiles, NoFileMissing) {
   }
 }
 
+TEST_F(VersionSetTestMissingFiles, MinLogNumberToKeep2PC) {
+  NewDB();
+
+  SstInfo sst(100, kDefaultColumnFamilyName, "a");
+  std::vector<FileMetaData> file_metas;
+  CreateDummyTableFiles({sst}, &file_metas);
+
+  constexpr WalNumber kMinWalNumberToKeep2PC = 10;
+  VersionEdit edit;
+  edit.AddFile(0, file_metas[0]);
+  edit.SetMinLogNumberToKeep(kMinWalNumberToKeep2PC);
+  ASSERT_OK(LogAndApplyToDefaultCF(edit));
+  ASSERT_EQ(versions_->min_log_number_to_keep_2pc(), kMinWalNumberToKeep2PC);
+
+  for (int i = 0; i < 3; i++) {
+    CreateNewManifest();
+    ReopenDB();
+    ASSERT_EQ(versions_->min_log_number_to_keep_2pc(), kMinWalNumberToKeep2PC);
+  }
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/wal_edit.cc

@@ -105,12 +105,17 @@ std::string WalDeletion::DebugString() const {
 }
 
 Status WalSet::AddWal(const WalAddition& wal) {
+  if (wal.GetLogNumber() < min_wal_number_to_keep_) {
+    // The WAL has been obsolete, ignore it.
+    return Status::OK();
+  }
+
   auto it = wals_.lower_bound(wal.GetLogNumber());
   bool existing = it != wals_.end() && it->first == wal.GetLogNumber();
   if (existing && !wal.GetMetadata().HasSyncedSize()) {
     std::stringstream ss;
     ss << "WAL " << wal.GetLogNumber() << " is created more than once";
-    return Status::Corruption("WalSet", ss.str());
+    return Status::Corruption("WalSet::AddWal", ss.str());
   }
   // If the WAL has synced size, it must >= the previous size.
   if (wal.GetMetadata().HasSyncedSize() && existing &&
@@ -120,7 +125,7 @@ Status WalSet::AddWal(const WalAddition& wal) {
     std::stringstream ss;
     ss << "WAL " << wal.GetLogNumber()
        << " must not have smaller synced size than previous one";
-    return Status::Corruption("WalSet", ss.str());
+    return Status::Corruption("WalSet::AddWal", ss.str());
   }
   if (existing) {
     it->second.SetSyncedSizeInBytes(wal.GetMetadata().GetSyncedSizeInBytes());
@@ -142,11 +147,17 @@ Status WalSet::AddWals(const WalAdditions& wals) {
 }
 
 Status WalSet::DeleteWalsBefore(WalNumber wal) {
-  wals_.erase(wals_.begin(), wals_.lower_bound(wal));
+  if (wal > min_wal_number_to_keep_) {
+    min_wal_number_to_keep_ = wal;
+    wals_.erase(wals_.begin(), wals_.lower_bound(wal));
+  }
   return Status::OK();
 }
 
-void WalSet::Reset() { wals_.clear(); }
+void WalSet::Reset() {
+  wals_.clear();
+  min_wal_number_to_keep_ = 0;
+}
 
 Status WalSet::CheckWals(
     Env* env,

db/wal_edit.h

@@ -141,6 +141,9 @@ class WalSet {
   // Resets the internal state.
   void Reset();
 
+  // WALs with number less than MinWalNumberToKeep should not exist in WalSet.
+  WalNumber GetMinWalNumberToKeep() const { return min_wal_number_to_keep_; }
+
   const std::map<WalNumber, WalMetadata>& GetWals() const { return wals_; }
 
   // Checks whether there are missing or corrupted WALs.
@@ -155,6 +158,9 @@ class WalSet {
 
  private:
   std::map<WalNumber, WalMetadata> wals_;
+  // WAL number < min_wal_number_to_keep_ should not exist in wals_.
+  // It's monotonically increasing, in-memory only, not written to MANIFEST.
+  WalNumber min_wal_number_to_keep_ = 0;
 };
 
 }  // namespace ROCKSDB_NAMESPACE

db/wal_edit_test.cc

@@ -81,6 +81,26 @@ TEST(WalSet, DeleteAllWals) {
   ASSERT_OK(wals.DeleteWalsBefore(kMaxWalNumber + 1));
 }
 
+TEST(WalSet, AddObsoleteWal) {
+  constexpr WalNumber kNumber = 100;
+  WalSet wals;
+  ASSERT_OK(wals.DeleteWalsBefore(kNumber + 1));
+  ASSERT_OK(wals.AddWal(WalAddition(kNumber)));
+  ASSERT_TRUE(wals.GetWals().empty());
+}
+
+TEST(WalSet, MinWalNumberToKeep) {
+  constexpr WalNumber kNumber = 100;
+  WalSet wals;
+  ASSERT_EQ(wals.GetMinWalNumberToKeep(), 0);
+  ASSERT_OK(wals.DeleteWalsBefore(kNumber));
+  ASSERT_EQ(wals.GetMinWalNumberToKeep(), kNumber);
+  ASSERT_OK(wals.DeleteWalsBefore(kNumber - 1));
+  ASSERT_EQ(wals.GetMinWalNumberToKeep(), kNumber);
+  ASSERT_OK(wals.DeleteWalsBefore(kNumber + 1));
+  ASSERT_EQ(wals.GetMinWalNumberToKeep(), kNumber + 1);
+}
+
 class WalSetTest : public DBTestBase {
  public:
   WalSetTest() : DBTestBase("WalSetTest", /* env_do_fsync */ true) {}

db/write_batch.cc

@@ -1425,19 +1425,21 @@ class MemTableInserter : public WriteBatch::Handler {
 
     Status ret_status;
     if (UNLIKELY(!SeekToColumnFamily(column_family_id, &ret_status))) {
-      bool batch_boundry = false;
-      if (rebuilding_trx_ != nullptr) {
+      if (ret_status.ok() && rebuilding_trx_ != nullptr) {
         assert(!write_after_commit_);
         // The CF is probably flushed and hence no need for insert but we still
         // need to keep track of the keys for upcoming rollback/commit.
         ret_status = WriteBatchInternal::Put(rebuilding_trx_, column_family_id,
                                              key, value);
-        assert(ret_status.ok());
-        batch_boundry = IsDuplicateKeySeq(column_family_id, key);
+        if (ret_status.ok()) {
+          MaybeAdvanceSeq(IsDuplicateKeySeq(column_family_id, key));
+        }
+      } else if (ret_status.ok()) {
+        MaybeAdvanceSeq(false /* batch_boundary */);
       }
-      MaybeAdvanceSeq(batch_boundry);
       return ret_status;
     }
+    assert(ret_status.ok());
 
     MemTable* mem = cf_mems_->GetMemTable();
     auto* moptions = mem->GetImmutableMemTableOptions();
@@ -1445,23 +1447,17 @@ class MemTableInserter : public WriteBatch::Handler {
     // any kind of transactions including the ones that use seq_per_batch
     assert(!seq_per_batch_ || !moptions->inplace_update_support);
     if (!moptions->inplace_update_support) {
-      bool mem_res =
+      ret_status =
           mem->Add(sequence_, value_type, key, value,
                    concurrent_memtable_writes_, get_post_process_info(mem),
                    hint_per_batch_ ? &GetHintMap()[mem] : nullptr);
-      if (UNLIKELY(!mem_res)) {
-        assert(seq_per_batch_);
-        ret_status = Status::TryAgain("key+seq exists");
-        const bool BATCH_BOUNDRY = true;
-        MaybeAdvanceSeq(BATCH_BOUNDRY);
-      }
     } else if (moptions->inplace_callback == nullptr) {
       assert(!concurrent_memtable_writes_);
-      mem->Update(sequence_, key, value);
+      ret_status = mem->Update(sequence_, key, value);
     } else {
       assert(!concurrent_memtable_writes_);
-      if (mem->UpdateCallback(sequence_, key, value)) {
-      } else {
+      ret_status = mem->UpdateCallback(sequence_, key, value);
+      if (ret_status.IsNotFound()) {
         // key not found in memtable. Do sst get, update, add
         SnapshotImpl read_from_snapshot;
         read_from_snapshot.number_ = sequence_;
@@ -1475,50 +1471,69 @@ class MemTableInserter : public WriteBatch::Handler {
         std::string merged_value;
 
         auto cf_handle = cf_mems_->GetColumnFamilyHandle();
-        Status s = Status::NotSupported();
+        Status get_status = Status::NotSupported();
         if (db_ != nullptr && recovering_log_number_ == 0) {
           if (cf_handle == nullptr) {
             cf_handle = db_->DefaultColumnFamily();
           }
-          s = db_->Get(ropts, cf_handle, key, &prev_value);
+          get_status = db_->Get(ropts, cf_handle, key, &prev_value);
+        }
+        // Intentionally overwrites the `NotFound` in `ret_status`.
+        if (!get_status.ok() && !get_status.IsNotFound()) {
+          ret_status = get_status;
+        } else {
+          ret_status = Status::OK();
         }
 
-        char* prev_buffer = const_cast<char*>(prev_value.c_str());
-        uint32_t prev_size = static_cast<uint32_t>(prev_value.size());
-        auto status = moptions->inplace_callback(s.ok() ? prev_buffer : nullptr,
-                                                 s.ok() ? &prev_size : nullptr,
-                                                 value, &merged_value);
-        if (status == UpdateStatus::UPDATED_INPLACE) {
-          // prev_value is updated in-place with final value.
-          bool mem_res __attribute__((__unused__));
-          mem_res = mem->Add(
-              sequence_, value_type, key, Slice(prev_buffer, prev_size));
-          assert(mem_res);
-          RecordTick(moptions->statistics, NUMBER_KEYS_WRITTEN);
-        } else if (status == UpdateStatus::UPDATED) {
-          // merged_value contains the final value.
-          bool mem_res __attribute__((__unused__));
-          mem_res =
-              mem->Add(sequence_, value_type, key, Slice(merged_value));
-          assert(mem_res);
-          RecordTick(moptions->statistics, NUMBER_KEYS_WRITTEN);
+        if (ret_status.ok()) {
+          UpdateStatus update_status;
+          char* prev_buffer = const_cast<char*>(prev_value.c_str());
+          uint32_t prev_size = static_cast<uint32_t>(prev_value.size());
+          if (get_status.ok()) {
+            update_status = moptions->inplace_callback(prev_buffer, &prev_size,
+                                                       value, &merged_value);
+          } else {
+            update_status = moptions->inplace_callback(
+                nullptr /* existing_value */, nullptr /* existing_value_size */,
+                value, &merged_value);
+          }
+          if (update_status == UpdateStatus::UPDATED_INPLACE) {
+            assert(get_status.ok());
+            // prev_value is updated in-place with final value.
+            ret_status = mem->Add(sequence_, value_type, key,
+                                  Slice(prev_buffer, prev_size));
+            if (ret_status.ok()) {
+              RecordTick(moptions->statistics, NUMBER_KEYS_WRITTEN);
+            }
+          } else if (update_status == UpdateStatus::UPDATED) {
+            // merged_value contains the final value.
+            ret_status =
+                mem->Add(sequence_, value_type, key, Slice(merged_value));
+            if (ret_status.ok()) {
+              RecordTick(moptions->statistics, NUMBER_KEYS_WRITTEN);
+            }
+          }
         }
       }
     }
+    if (UNLIKELY(ret_status.IsTryAgain())) {
+      assert(seq_per_batch_);
+      const bool kBatchBoundary = true;
+      MaybeAdvanceSeq(kBatchBoundary);
+    } else if (ret_status.ok()) {
+      MaybeAdvanceSeq();
+      CheckMemtableFull();
+    }
     // optimize for non-recovery mode
-    if (UNLIKELY(!ret_status.IsTryAgain() && rebuilding_trx_ != nullptr)) {
+    // If `ret_status` is `TryAgain` then the next (successful) try will add
+    // the key to the rebuilding transaction object. If `ret_status` is
+    // another non-OK `Status`, then the `rebuilding_trx_` will be thrown
+    // away. So we only need to add to it when `ret_status.ok()`.
+    if (UNLIKELY(ret_status.ok() && rebuilding_trx_ != nullptr)) {
       assert(!write_after_commit_);
-      // If the ret_status is TryAgain then let the next try to add the ky to
-      // the rebuilding transaction object.
       ret_status = WriteBatchInternal::Put(rebuilding_trx_, column_family_id,
                                            key, value);
-      assert(ret_status.ok());
     }
-    // Since all Puts are logged in transaction logs (if enabled), always bump
-    // sequence number. Even if the update eventually fails and does not result
-    // in memtable add/update.
-    MaybeAdvanceSeq();
-    CheckMemtableFull();
     return ret_status;
   }
 
@@ -1531,18 +1546,18 @@ class MemTableInserter : public WriteBatch::Handler {
                     const Slice& value, ValueType delete_type) {
     Status ret_status;
     MemTable* mem = cf_mems_->GetMemTable();
-    bool mem_res =
+    ret_status =
         mem->Add(sequence_, delete_type, key, value,
                  concurrent_memtable_writes_, get_post_process_info(mem),
                  hint_per_batch_ ? &GetHintMap()[mem] : nullptr);
-    if (UNLIKELY(!mem_res)) {
+    if (UNLIKELY(ret_status.IsTryAgain())) {
       assert(seq_per_batch_);
-      ret_status = Status::TryAgain("key+seq exists");
-      const bool BATCH_BOUNDRY = true;
-      MaybeAdvanceSeq(BATCH_BOUNDRY);
+      const bool kBatchBoundary = true;
+      MaybeAdvanceSeq(kBatchBoundary);
+    } else if (ret_status.ok()) {
+      MaybeAdvanceSeq();
+      CheckMemtableFull();
     }
-    MaybeAdvanceSeq();
-    CheckMemtableFull();
     return ret_status;
   }
 
@@ -1555,17 +1570,18 @@ class MemTableInserter : public WriteBatch::Handler {
 
     Status ret_status;
     if (UNLIKELY(!SeekToColumnFamily(column_family_id, &ret_status))) {
-      bool batch_boundry = false;
-      if (rebuilding_trx_ != nullptr) {
+      if (ret_status.ok() && rebuilding_trx_ != nullptr) {
         assert(!write_after_commit_);
         // The CF is probably flushed and hence no need for insert but we still
         // need to keep track of the keys for upcoming rollback/commit.
         ret_status =
             WriteBatchInternal::Delete(rebuilding_trx_, column_family_id, key);
-        assert(ret_status.ok());
-        batch_boundry = IsDuplicateKeySeq(column_family_id, key);
+        if (ret_status.ok()) {
+          MaybeAdvanceSeq(IsDuplicateKeySeq(column_family_id, key));
+        }
+      } else if (ret_status.ok()) {
+        MaybeAdvanceSeq(false /* batch_boundary */);
       }
-      MaybeAdvanceSeq(batch_boundry);
       return ret_status;
     }
 
@@ -1578,10 +1594,12 @@ class MemTableInserter : public WriteBatch::Handler {
         (0 == ts_sz) ? kTypeDeletion : kTypeDeletionWithTimestamp;
     ret_status = DeleteImpl(column_family_id, key, Slice(), delete_type);
     // optimize for non-recovery mode
-    if (UNLIKELY(!ret_status.IsTryAgain() && rebuilding_trx_ != nullptr)) {
+    // If `ret_status` is `TryAgain` then the next (successful) try will add
+    // the key to the rebuilding transaction object. If `ret_status` is
+    // another non-OK `Status`, then the `rebuilding_trx_` will be thrown
+    // away. So we only need to add to it when `ret_status.ok()`.
+    if (UNLIKELY(ret_status.ok() && rebuilding_trx_ != nullptr)) {
       assert(!write_after_commit_);
-      // If the ret_status is TryAgain then let the next try to add the ky to
-      // the rebuilding transaction object.
       ret_status =
           WriteBatchInternal::Delete(rebuilding_trx_, column_family_id, key);
     }
@@ -1598,27 +1616,31 @@ class MemTableInserter : public WriteBatch::Handler {
 
     Status ret_status;
     if (UNLIKELY(!SeekToColumnFamily(column_family_id, &ret_status))) {
-      bool batch_boundry = false;
-      if (rebuilding_trx_ != nullptr) {
+      if (ret_status.ok() && rebuilding_trx_ != nullptr) {
         assert(!write_after_commit_);
         // The CF is probably flushed and hence no need for insert but we still
         // need to keep track of the keys for upcoming rollback/commit.
         ret_status = WriteBatchInternal::SingleDelete(rebuilding_trx_,
                                                       column_family_id, key);
-        assert(ret_status.ok());
-        batch_boundry = IsDuplicateKeySeq(column_family_id, key);
+        if (ret_status.ok()) {
+          MaybeAdvanceSeq(IsDuplicateKeySeq(column_family_id, key));
+        }
+      } else if (ret_status.ok()) {
+        MaybeAdvanceSeq(false /* batch_boundary */);
       }
-      MaybeAdvanceSeq(batch_boundry);
       return ret_status;
     }
+    assert(ret_status.ok());
 
     ret_status =
         DeleteImpl(column_family_id, key, Slice(), kTypeSingleDeletion);
     // optimize for non-recovery mode
-    if (UNLIKELY(!ret_status.IsTryAgain() && rebuilding_trx_ != nullptr)) {
+    // If `ret_status` is `TryAgain` then the next (successful) try will add
+    // the key to the rebuilding transaction object. If `ret_status` is
+    // another non-OK `Status`, then the `rebuilding_trx_` will be thrown
+    // away. So we only need to add to it when `ret_status.ok()`.
+    if (UNLIKELY(ret_status.ok() && rebuilding_trx_ != nullptr)) {
       assert(!write_after_commit_);
-      // If the ret_status is TryAgain then let the next try to add the ky to
-      // the rebuilding transaction object.
       ret_status = WriteBatchInternal::SingleDelete(rebuilding_trx_,
                                                     column_family_id, key);
     }
@@ -1636,21 +1658,22 @@ class MemTableInserter : public WriteBatch::Handler {
 
     Status ret_status;
     if (UNLIKELY(!SeekToColumnFamily(column_family_id, &ret_status))) {
-      bool batch_boundry = false;
-      if (rebuilding_trx_ != nullptr) {
+      if (ret_status.ok() && rebuilding_trx_ != nullptr) {
         assert(!write_after_commit_);
         // The CF is probably flushed and hence no need for insert but we still
         // need to keep track of the keys for upcoming rollback/commit.
         ret_status = WriteBatchInternal::DeleteRange(
             rebuilding_trx_, column_family_id, begin_key, end_key);
-        assert(ret_status.ok());
-        // TODO(myabandeh): when transactional DeleteRange support is added,
-        // check if end_key must also be added.
-        batch_boundry = IsDuplicateKeySeq(column_family_id, begin_key);
+        if (ret_status.ok()) {
+          MaybeAdvanceSeq(IsDuplicateKeySeq(column_family_id, begin_key));
+        }
+      } else if (ret_status.ok()) {
+        MaybeAdvanceSeq(false /* batch_boundary */);
       }
-      MaybeAdvanceSeq(batch_boundry);
       return ret_status;
     }
+    assert(ret_status.ok());
+
     if (db_ != nullptr) {
       auto cf_handle = cf_mems_->GetColumnFamilyHandle();
       if (cf_handle == nullptr) {
@@ -1659,6 +1682,8 @@ class MemTableInserter : public WriteBatch::Handler {
       auto* cfd =
           static_cast_with_check<ColumnFamilyHandleImpl>(cf_handle)->cfd();
       if (!cfd->is_delete_range_supported()) {
+        // TODO(ajkr): refactor `SeekToColumnFamily()` so it returns a `Status`.
+        ret_status.PermitUncheckedError();
         return Status::NotSupported(
             std::string("DeleteRange not supported for table type ") +
             cfd->ioptions()->table_factory->Name() + " in CF " +
@@ -1666,10 +1691,14 @@ class MemTableInserter : public WriteBatch::Handler {
       }
       int cmp = cfd->user_comparator()->Compare(begin_key, end_key);
       if (cmp > 0) {
+        // TODO(ajkr): refactor `SeekToColumnFamily()` so it returns a `Status`.
+        ret_status.PermitUncheckedError();
         // It's an empty range where endpoints appear mistaken. Don't bother
         // applying it to the DB, and return an error to the user.
         return Status::InvalidArgument("end key comes before start key");
       } else if (cmp == 0) {
+        // TODO(ajkr): refactor `SeekToColumnFamily()` so it returns a `Status`.
+        ret_status.PermitUncheckedError();
         // It's an empty range. Don't bother applying it to the DB.
         return Status::OK();
       }
@@ -1678,10 +1707,12 @@ class MemTableInserter : public WriteBatch::Handler {
     ret_status =
         DeleteImpl(column_family_id, begin_key, end_key, kTypeRangeDeletion);
     // optimize for non-recovery mode
+    // If `ret_status` is `TryAgain` then the next (successful) try will add
+    // the key to the rebuilding transaction object. If `ret_status` is
+    // another non-OK `Status`, then the `rebuilding_trx_` will be thrown
+    // away. So we only need to add to it when `ret_status.ok()`.
     if (UNLIKELY(!ret_status.IsTryAgain() && rebuilding_trx_ != nullptr)) {
       assert(!write_after_commit_);
-      // If the ret_status is TryAgain then let the next try to add the ky to
-      // the rebuilding transaction object.
       ret_status = WriteBatchInternal::DeleteRange(
           rebuilding_trx_, column_family_id, begin_key, end_key);
     }
@@ -1699,22 +1730,28 @@ class MemTableInserter : public WriteBatch::Handler {
 
     Status ret_status;
     if (UNLIKELY(!SeekToColumnFamily(column_family_id, &ret_status))) {
-      bool batch_boundry = false;
-      if (rebuilding_trx_ != nullptr) {
+      if (ret_status.ok() && rebuilding_trx_ != nullptr) {
         assert(!write_after_commit_);
         // The CF is probably flushed and hence no need for insert but we still
         // need to keep track of the keys for upcoming rollback/commit.
         ret_status = WriteBatchInternal::Merge(rebuilding_trx_,
                                                column_family_id, key, value);
-        assert(ret_status.ok());
-        batch_boundry = IsDuplicateKeySeq(column_family_id, key);
+        if (ret_status.ok()) {
+          MaybeAdvanceSeq(IsDuplicateKeySeq(column_family_id, key));
+        }
+      } else if (ret_status.ok()) {
+        MaybeAdvanceSeq(false /* batch_boundary */);
       }
-      MaybeAdvanceSeq(batch_boundry);
       return ret_status;
     }
+    assert(ret_status.ok());
 
     MemTable* mem = cf_mems_->GetMemTable();
     auto* moptions = mem->GetImmutableMemTableOptions();
+    if (moptions->merge_operator == nullptr) {
+      return Status::InvalidArgument(
+          "Merge requires `ColumnFamilyOptions::merge_operator != nullptr`");
+    }
     bool perform_merge = false;
     assert(!concurrent_memtable_writes_ ||
            moptions->max_successive_merges == 0);
@@ -1752,60 +1789,60 @@ class MemTableInserter : public WriteBatch::Handler {
       if (cf_handle == nullptr) {
         cf_handle = db_->DefaultColumnFamily();
       }
-      db_->Get(read_options, cf_handle, key, &get_value);
-      Slice get_value_slice = Slice(get_value);
-
-      // 2) Apply this merge
-      auto merge_operator = moptions->merge_operator;
-      assert(merge_operator);
-
-      std::string new_value;
-
-      Status merge_status = MergeHelper::TimedFullMerge(
-          merge_operator, key, &get_value_slice, {value}, &new_value,
-          moptions->info_log, moptions->statistics, Env::Default());
-
-      if (!merge_status.ok()) {
-        // Failed to merge!
-        // Store the delta in memtable
+      Status get_status = db_->Get(read_options, cf_handle, key, &get_value);
+      if (!get_status.ok()) {
+        // Failed to read a key we know exists. Store the delta in memtable.
         perform_merge = false;
       } else {
-        // 3) Add value to memtable
-        assert(!concurrent_memtable_writes_);
-        bool mem_res = mem->Add(sequence_, kTypeValue, key, new_value);
-        if (UNLIKELY(!mem_res)) {
-          assert(seq_per_batch_);
-          ret_status = Status::TryAgain("key+seq exists");
-          const bool BATCH_BOUNDRY = true;
-          MaybeAdvanceSeq(BATCH_BOUNDRY);
+        Slice get_value_slice = Slice(get_value);
+
+        // 2) Apply this merge
+        auto merge_operator = moptions->merge_operator;
+        assert(merge_operator);
+
+        std::string new_value;
+
+        Status merge_status = MergeHelper::TimedFullMerge(
+            merge_operator, key, &get_value_slice, {value}, &new_value,
+            moptions->info_log, moptions->statistics, Env::Default());
+
+        if (!merge_status.ok()) {
+          // Failed to merge!
+          // Store the delta in memtable
+          perform_merge = false;
+        } else {
+          // 3) Add value to memtable
+          assert(!concurrent_memtable_writes_);
+          ret_status = mem->Add(sequence_, kTypeValue, key, new_value);
         }
       }
     }
 
     if (!perform_merge) {
-      // Add merge operator to memtable
-      bool mem_res =
+      // Add merge operand to memtable
+      ret_status =
           mem->Add(sequence_, kTypeMerge, key, value,
                    concurrent_memtable_writes_, get_post_process_info(mem));
-      if (UNLIKELY(!mem_res)) {
-        assert(seq_per_batch_);
-        ret_status = Status::TryAgain("key+seq exists");
-        const bool BATCH_BOUNDRY = true;
-        MaybeAdvanceSeq(BATCH_BOUNDRY);
-      }
     }
 
+    if (UNLIKELY(ret_status.IsTryAgain())) {
+      assert(seq_per_batch_);
+      const bool kBatchBoundary = true;
+      MaybeAdvanceSeq(kBatchBoundary);
+    } else if (ret_status.ok()) {
+      MaybeAdvanceSeq();
+      CheckMemtableFull();
+    }
     // optimize for non-recovery mode
-    if (UNLIKELY(!ret_status.IsTryAgain() && rebuilding_trx_ != nullptr)) {
+    // If `ret_status` is `TryAgain` then the next (successful) try will add
+    // the key to the rebuilding transaction object. If `ret_status` is
+    // another non-OK `Status`, then the `rebuilding_trx_` will be thrown
+    // away. So we only need to add to it when `ret_status.ok()`.
+    if (UNLIKELY(ret_status.ok() && rebuilding_trx_ != nullptr)) {
       assert(!write_after_commit_);
-      // If the ret_status is TryAgain then let the next try to add the ky to
-      // the rebuilding transaction object.
       ret_status = WriteBatchInternal::Merge(rebuilding_trx_, column_family_id,
                                              key, value);
-      assert(ret_status.ok());
     }
-    MaybeAdvanceSeq();
-    CheckMemtableFull();
     return ret_status;
   }
 

db/write_batch_test.cc

@@ -7,12 +7,13 @@
 // 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 "rocksdb/db.h"
-
 #include <memory>
+
 #include "db/column_family.h"
+#include "db/db_test_util.h"
 #include "db/memtable.h"
 #include "db/write_batch_internal.h"
+#include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/utilities/write_batch_with_index.h"
@@ -23,11 +24,15 @@
 
 namespace ROCKSDB_NAMESPACE {
 
-static std::string PrintContents(WriteBatch* b) {
+static std::string PrintContents(WriteBatch* b,
+                                 bool merge_operator_supported = true) {
   InternalKeyComparator cmp(BytewiseComparator());
   auto factory = std::make_shared<SkipListFactory>();
   Options options;
   options.memtable_factory = factory;
+  if (merge_operator_supported) {
+    options.merge_operator.reset(new TestPutOperator());
+  }
   ImmutableCFOptions ioptions(options);
   WriteBufferManager wb(options.db_write_buffer_size);
   MemTable* mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
@@ -59,6 +64,7 @@ static std::string PrintContents(WriteBatch* b) {
     if (iter == nullptr) {
       continue;
     }
+    EXPECT_OK(iter->status());
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
       ParsedInternalKey ikey;
       ikey.clear();
@@ -112,16 +118,19 @@ static std::string PrintContents(WriteBatch* b) {
       state.append("@");
       state.append(NumberToString(ikey.sequence));
     }
+    EXPECT_OK(iter->status());
   }
-  EXPECT_EQ(b->HasPut(), put_count > 0);
-  EXPECT_EQ(b->HasDelete(), delete_count > 0);
-  EXPECT_EQ(b->HasSingleDelete(), single_delete_count > 0);
-  EXPECT_EQ(b->HasDeleteRange(), delete_range_count > 0);
-  EXPECT_EQ(b->HasMerge(), merge_count > 0);
-  if (!s.ok()) {
+  if (s.ok()) {
+    EXPECT_EQ(b->HasPut(), put_count > 0);
+    EXPECT_EQ(b->HasDelete(), delete_count > 0);
+    EXPECT_EQ(b->HasSingleDelete(), single_delete_count > 0);
+    EXPECT_EQ(b->HasDeleteRange(), delete_range_count > 0);
+    EXPECT_EQ(b->HasMerge(), merge_count > 0);
+    if (count != WriteBatchInternal::Count(b)) {
+      state.append("CountMismatch()");
+    }
+  } else {
     state.append(s.ToString());
-  } else if (count != WriteBatchInternal::Count(b)) {
-    state.append("CountMismatch()");
   }
   delete mem->Unref();
   return state;
@@ -354,6 +363,16 @@ TEST_F(WriteBatchTest, MergeNotImplemented) {
   ASSERT_OK(batch.Iterate(&handler));
 }
 
+TEST_F(WriteBatchTest, MergeWithoutOperatorInsertionFailure) {
+  WriteBatch batch;
+  ASSERT_OK(batch.Merge(Slice("foo"), Slice("bar")));
+  ASSERT_EQ(1u, batch.Count());
+  ASSERT_EQ(
+      "Invalid argument: Merge requires `ColumnFamilyOptions::merge_operator "
+      "!= nullptr`",
+      PrintContents(&batch, false /* merge_operator_supported */));
+}
+
 TEST_F(WriteBatchTest, Blob) {
   WriteBatch batch;
   ASSERT_OK(batch.Put(Slice("k1"), Slice("v1")));

db/write_thread.cc

@@ -208,6 +208,7 @@ uint8_t WriteThread::AwaitState(Writer* w, uint8_t goal_mask,
 }
 
 void WriteThread::SetState(Writer* w, uint8_t new_state) {
+  assert(w);
   auto state = w->state.load(std::memory_order_acquire);
   if (state == STATE_LOCKED_WAITING ||
       !w->state.compare_exchange_strong(state, new_state)) {

db_stress_tool/db_stress_common.cc

@@ -16,10 +16,10 @@
 #include "util/file_checksum_helper.h"
 #include "util/xxhash.h"
 
-ROCKSDB_NAMESPACE::DbStressEnvWrapper* db_stress_env = nullptr;
+ROCKSDB_NAMESPACE::Env* db_stress_env = nullptr;
 #ifndef NDEBUG
 // If non-null, injects read error at a rate specified by the
-// read_fault_one_in flag
+// read_fault_one_in or write_fault_one_in flag
 std::shared_ptr<ROCKSDB_NAMESPACE::FaultInjectionTestFS> fault_fs_guard;
 #endif // NDEBUG
 enum ROCKSDB_NAMESPACE::CompressionType compression_type_e =

db_stress_tool/db_stress_common.h

@@ -252,7 +252,7 @@ const int kRandomValueMaxFactor = 3;
 const int kValueMaxLen = 100;
 
 // wrapped posix or hdfs environment
-extern ROCKSDB_NAMESPACE::DbStressEnvWrapper* db_stress_env;
+extern ROCKSDB_NAMESPACE::Env* db_stress_env;
 #ifndef NDEBUG
 namespace ROCKSDB_NAMESPACE {
 class FaultInjectionTestFS;

db_stress_tool/db_stress_gflags.cc

@@ -757,4 +757,7 @@ DEFINE_string(file_checksum_impl, "none",
               "Name of an implementation for file_checksum_gen_factory, or "
               "\"none\" for null.");
 
+DEFINE_int32(write_fault_one_in, 0,
+             "On non-zero, enables fault injection on write");
+
 #endif  // GFLAGS

db_stress_tool/db_stress_shared_state.h

@@ -29,6 +29,7 @@ DECLARE_bool(test_batches_snapshots);
 DECLARE_int32(compaction_thread_pool_adjust_interval);
 DECLARE_int32(continuous_verification_interval);
 DECLARE_int32(read_fault_one_in);
+DECLARE_int32(write_fault_one_in);
 
 namespace ROCKSDB_NAMESPACE {
 class StressTest;

db_stress_tool/db_stress_test_base.cc

@@ -15,6 +15,7 @@
 #include "db_stress_tool/db_stress_table_properties_collector.h"
 #include "rocksdb/convenience.h"
 #include "rocksdb/sst_file_manager.h"
+#include "rocksdb/types.h"
 #include "util/cast_util.h"
 #include "utilities/fault_injection_fs.h"
 
@@ -525,6 +526,16 @@ void StressTest::OperateDb(ThreadState* thread) {
     fault_fs_guard->SetThreadLocalReadErrorContext(thread->shared->GetSeed(),
                                             FLAGS_read_fault_one_in);
   }
+  if (FLAGS_write_fault_one_in) {
+    IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
+    error_msg.SetRetryable(true);
+    std::vector<FileType> types;
+    types.push_back(FileType::kTableFile);
+    types.push_back(FileType::kDescriptorFile);
+    types.push_back(FileType::kCurrentFile);
+    fault_fs_guard->SetRandomWriteError(
+        thread->shared->GetSeed(), FLAGS_write_fault_one_in, error_msg, types);
+  }
 #endif // NDEBUG
   thread->stats.Start();
   for (int open_cnt = 0; open_cnt <= FLAGS_reopen; ++open_cnt) {
@@ -618,7 +629,8 @@ void StressTest::OperateDb(ThreadState* thread) {
 
 #ifndef ROCKSDB_LITE
       // Verify GetLiveFiles with a 1 in N chance.
-      if (thread->rand.OneInOpt(FLAGS_get_live_files_one_in)) {
+      if (thread->rand.OneInOpt(FLAGS_get_live_files_one_in) &&
+          !FLAGS_write_fault_one_in) {
         Status status = VerifyGetLiveFiles();
         if (!status.ok()) {
           VerificationAbort(shared, "VerifyGetLiveFiles status not OK", status);
@@ -1460,7 +1472,7 @@ Status StressTest::TestCheckpoint(ThreadState* thread,
       FLAGS_db + "/.checkpoint" + ToString(thread->tid);
   Options tmp_opts(options_);
   tmp_opts.listeners.clear();
-  tmp_opts.env = db_stress_env->target();
+  tmp_opts.env = db_stress_env;
 
   DestroyDB(checkpoint_dir, tmp_opts);
 
@@ -1952,6 +1964,7 @@ void StressTest::PrintEnv() const {
   fprintf(stdout, "Use dynamic level         : %d\n",
           static_cast<int>(FLAGS_level_compaction_dynamic_level_bytes));
   fprintf(stdout, "Read fault one in         : %d\n", FLAGS_read_fault_one_in);
+  fprintf(stdout, "Write fault one in        : %d\n", FLAGS_write_fault_one_in);
   fprintf(stdout, "Sync fault injection      : %d\n", FLAGS_sync_fault_injection);
   fprintf(stdout, "Best efforts recovery     : %d\n",
           static_cast<int>(FLAGS_best_efforts_recovery));

db_stress_tool/db_stress_tool.cc

@@ -97,20 +97,42 @@ int db_stress_tool(int argc, char** argv) {
   }
 
 #ifndef NDEBUG
-  if (FLAGS_read_fault_one_in || FLAGS_sync_fault_injection) {
+  if (FLAGS_read_fault_one_in || FLAGS_sync_fault_injection ||
+      FLAGS_write_fault_one_in) {
     FaultInjectionTestFS* fs =
         new FaultInjectionTestFS(raw_env->GetFileSystem());
     fault_fs_guard.reset(fs);
-    fault_fs_guard->SetFilesystemDirectWritable(true);
+    if (FLAGS_write_fault_one_in) {
+      fault_fs_guard->SetFilesystemDirectWritable(false);
+    } else {
+      fault_fs_guard->SetFilesystemDirectWritable(true);
+    }
     fault_env_guard =
         std::make_shared<CompositeEnvWrapper>(raw_env, fault_fs_guard);
     raw_env = fault_env_guard.get();
   }
+  if (FLAGS_write_fault_one_in) {
+    SyncPoint::GetInstance()->SetCallBack(
+        "BuildTable:BeforeFinishBuildTable",
+        [&](void*) { fault_fs_guard->EnableWriteErrorInjection(); });
+    SyncPoint::GetInstance()->EnableProcessing();
+  }
 #endif
 
   env_wrapper_guard = std::make_shared<DbStressEnvWrapper>(raw_env);
   db_stress_env = env_wrapper_guard.get();
 
+#ifndef NDEBUG
+  if (FLAGS_write_fault_one_in) {
+    // In the write injection case, we need to use the FS interface and returns
+    // the IOStatus with different error and flags. Therefore,
+    // DbStressEnvWrapper cannot be used which will swallow the FS
+    // implementations. We should directly use the raw_env which is the
+    // CompositeEnvWrapper of env and fault_fs.
+    db_stress_env = raw_env;
+  }
+#endif
+
   FLAGS_rep_factory = StringToRepFactory(FLAGS_memtablerep.c_str());
 
   // The number of background threads should be at least as much the

env/fs_posix.cc

@@ -766,9 +766,9 @@ class PosixFileSystem : public FileSystem {
     // closed, all locks the process holds for that *file* are released
     const auto it_success = locked_files.insert({fname, lhi});
     if (it_success.second == false) {
+      LockHoldingInfo prev_info = it_success.first->second;
       mutex_locked_files.Unlock();
       errno = ENOLCK;
-      LockHoldingInfo& prev_info = it_success.first->second;
       // Note that the thread ID printed is the same one as the one in
       // posix logger, but posix logger prints it hex format.
       return IOError("lock hold by current process, acquire time " +

fuzz/Makefile

@@ -0,0 +1,61 @@
+# Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+# This source code is licensed under both the GPLv2 (found in the
+# COPYING file in the root directory) and Apache 2.0 License
+# (found in the LICENSE.Apache file in the root directory).
+
+ROOT_DIR = $(abspath $(shell pwd)/../)
+
+include $(ROOT_DIR)/make_config.mk
+
+PROTOBUF_CFLAGS = `pkg-config --cflags protobuf`
+PROTOBUF_LDFLAGS = `pkg-config --libs protobuf`
+
+PROTOBUF_MUTATOR_CFLAGS = `pkg-config --cflags libprotobuf-mutator`
+PROTOBUF_MUTATOR_LDFLAGS = `pkg-config --libs libprotobuf-mutator`
+
+ROCKSDB_INCLUDE_DIR = $(ROOT_DIR)/include
+ROCKSDB_LIB_DIR = $(ROOT_DIR)
+
+PROTO_IN = $(ROOT_DIR)/fuzz/proto
+PROTO_OUT = $(ROOT_DIR)/fuzz/proto/gen
+
+ifneq ($(FUZZ_ENV), ossfuzz)
+CC = clang++
+CCFLAGS += -Wall -fsanitize=address,fuzzer
+CFLAGS += $(PLATFORM_CXXFLAGS) $(PROTOBUF_CFLAGS) $(PROTOBUF_MUTATOR_CFLAGS) -I$(PROTO_OUT) -I$(ROCKSDB_INCLUDE_DIR) -I$(ROCKSDB_LIB_DIR)
+LDFLAGS += $(PLATFORM_LDFLAGS) $(PROTOBUF_LDFLAGS) $(PROTOBUF_MUTATOR_LDFLAGS) -L$(ROCKSDB_LIB_DIR) -lrocksdb
+else
+# OSS-Fuzz sets various environment flags that are used for compilation.
+# These environment flags depend on which type of sanitizer build is being
+# used, however, an ASan build would set the environment flags as follows:
+# CFLAGS="-O1 -fno-omit-frame-pointer -gline-tables-only \
+         -DFUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION -fsanitize=address \
+         -fsanitize-address-use-after-scope -fsanitize=fuzzer-no-link"
+# CXXFLAGS="-O1 -fno-omit-frame-pointer -gline-tables-only \
+           -DFUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION -fsanitize=address \
+           -fsanitize-address-use-after-scope -fsanitize=fuzzer-no-link \
+           -stdlib=libc++"
+# LIB_FUZZING_ENGINE="-fsanitize=fuzzer"
+CC = $(CXX)
+CCFLAGS = $(CXXFLAGS)
+CFLAGS += $(PROTOBUF_CFLAGS) $(PROTOBUF_MUTATOR_CFLAGS) -I$(PROTO_OUT) -I$(ROCKSDB_INCLUDE_DIR) -I$(ROCKSDB_LIB_DIR)
+LDFLAGS += $(PLATFORM_LDFLAGS) $(LIB_FUZZING_ENGINE) $(PROTOBUF_MUTATOR_LDFLAGS) $(PROTOBUF_LDFLAGS) -L$(ROCKSDB_LIB_DIR) -lrocksdb
+endif
+
+.PHONY: gen_proto
+
+gen_proto:
+	mkdir -p $(PROTO_OUT)
+	protoc \
+		--proto_path=$(PROTO_IN) \
+		--cpp_out=$(PROTO_OUT) \
+		$(PROTO_IN)/*.proto
+
+db_fuzzer: db_fuzzer.cc
+	$(CC) $(CCFLAGS) -o db_fuzzer db_fuzzer.cc $(CFLAGS) $(LDFLAGS)
+
+db_map_fuzzer: gen_proto db_map_fuzzer.cc proto/gen/db_operation.pb.cc
+	$(CC) $(CCFLAGS) -o db_map_fuzzer db_map_fuzzer.cc proto/gen/db_operation.pb.cc $(CFLAGS) $(LDFLAGS)
+
+sst_file_writer_fuzzer: gen_proto sst_file_writer_fuzzer.cc proto/gen/db_operation.pb.cc
+	$(CC) $(CCFLAGS) -o sst_file_writer_fuzzer sst_file_writer_fuzzer.cc proto/gen/db_operation.pb.cc $(CFLAGS) $(LDFLAGS)

fuzz/README.md

@@ -0,0 +1,160 @@
+# Fuzzing RocksDB
+
+## Overview
+
+This directory contains [fuzz tests](https://en.wikipedia.org/wiki/Fuzzing) for RocksDB.
+RocksDB testing infrastructure currently includes unit tests and [stress tests](https://github.com/facebook/rocksdb/wiki/Stress-test),
+we hope fuzz testing can catch more bugs.
+
+## Prerequisite
+
+We use [LLVM libFuzzer](http://llvm.org/docs/LibFuzzer.html) as the fuzzying engine,
+so make sure you have [clang](https://clang.llvm.org/get_started.html) as your compiler.
+
+Some tests rely on [structure aware fuzzing](https://github.com/google/fuzzing/blob/master/docs/structure-aware-fuzzing.md).
+We use [protobuf](https://developers.google.com/protocol-buffers) to define structured input to the fuzzer,
+and use [libprotobuf-mutator](https://github.com/google/libprotobuf-mutator) as the custom libFuzzer mutator.
+So make sure you have protobuf and libprotobuf-mutator installed, and make sure `pkg-config` can find them.
+
+## Example
+
+This example shows you how to do structure aware fuzzing to `rocksdb::SstFileWriter`.
+
+After walking through the steps to create the fuzzer, we'll introduce a bug into `rocksdb::SstFileWriter::Put`,
+then show that the fuzzer can catch the bug.
+
+### Design the test
+
+We want the fuzzing engine to automatically generate a list of database operations,
+then we apply these operations to `SstFileWriter` in sequence,
+finally, after the SST file is generated, we use `SstFileReader` to check the file's checksum.
+
+### Define input
+
+We define the database operations in protobuf, each operation has a type of operation and a key value pair,
+see [proto/db_operation.proto](proto/db_operation.proto) for details.
+
+### Define tests with the input
+
+In [sst_file_writer_fuzzer.cc](sst_file_writer_fuzzer.cc),
+we define the tests to be run on the generated input:
+
+```
+DEFINE_PROTO_FUZZER(DBOperations& input) {
+  // apply the operations to SstFileWriter and use SstFileReader to verify checksum.
+  // ...
+}
+```
+
+`SstFileWriter` requires the keys of the operations to be unique and be in ascending order,
+but the fuzzing engine generates the input randomly, so we need to process the generated input before
+passing it to `DEFINE_PROTO_FUZZER`, this is accomplished by registering a post processor:
+
+```
+protobuf_mutator::libfuzzer::PostProcessorRegistration<DBOperations>
+```
+
+### Compile and link the fuzzer
+
+In the rocksdb root directory, compile rocksdb library by `make static_lib`.
+
+Go to the `fuzz` directory,
+run `make sst_file_writer_fuzzer` to generate the fuzzer,
+it will compile rocksdb static library, generate protobuf, then compile and link `sst_file_writer_fuzzer`.
+
+### Introduce a bug
+
+Manually introduce a bug to `SstFileWriter::Put`:
+
+```
+diff --git a/table/sst_file_writer.cc b/table/sst_file_writer.cc
+index ab1ee7c4e..c7da9ffa0 100644
+--- a/table/sst_file_writer.cc
++++ b/table/sst_file_writer.cc
+@@ -277,6 +277,11 @@ Status SstFileWriter::Add(const Slice& user_key, const Slice& value) {
+ }
+
+ Status SstFileWriter::Put(const Slice& user_key, const Slice& value) {
++  if (user_key.starts_with("!")) {
++    if (value.ends_with("!")) {
++      return Status::Corruption("bomb");
++    }
++  }
+   return rep_->Add(user_key, value, ValueType::kTypeValue);
+ }
+```
+
+The bug is that for `Put`, if `user_key` starts with `!` and `value` ends with `!`, then corrupt.
+
+### Run fuzz testing to catch the bug
+
+Run the fuzzer by `time ./sst_file_writer_fuzzer`.
+
+Here is the output on my machine:
+
+```
+Corruption: bomb
+==59680== ERROR: libFuzzer: deadly signal
+    #0 0x109487315 in __sanitizer_print_stack_trace+0x35 (libclang_rt.asan_osx_dynamic.dylib:x86_64+0x4d315)
+    #1 0x108d63f18 in fuzzer::PrintStackTrace() FuzzerUtil.cpp:205
+    #2 0x108d47613 in fuzzer::Fuzzer::CrashCallback() FuzzerLoop.cpp:232
+    #3 0x7fff6af535fc in _sigtramp+0x1c (libsystem_platform.dylib:x86_64+0x35fc)
+    #4 0x7ffee720f3ef  (<unknown module>)
+    #5 0x7fff6ae29807 in abort+0x77 (libsystem_c.dylib:x86_64+0x7f807)
+    #6 0x108cf1c4c in TestOneProtoInput(DBOperations&)+0x113c (sst_file_writer_fuzzer:x86_64+0x100302c4c)
+    #7 0x108cf09be in LLVMFuzzerTestOneInput+0x16e (sst_file_writer_fuzzer:x86_64+0x1003019be)
+    #8 0x108d48ce0 in fuzzer::Fuzzer::ExecuteCallback(unsigned char const*, unsigned long) FuzzerLoop.cpp:556
+    #9 0x108d48425 in fuzzer::Fuzzer::RunOne(unsigned char const*, unsigned long, bool, fuzzer::InputInfo*, bool*) FuzzerLoop.cpp:470
+    #10 0x108d4a626 in fuzzer::Fuzzer::MutateAndTestOne() FuzzerLoop.cpp:698
+    #11 0x108d4b325 in fuzzer::Fuzzer::Loop(std::__1::vector<fuzzer::SizedFile, fuzzer::fuzzer_allocator<fuzzer::SizedFile> >&) FuzzerLoop.cpp:830
+    #12 0x108d37fcd in fuzzer::FuzzerDriver(int*, char***, int (*)(unsigned char const*, unsigned long)) FuzzerDriver.cpp:829
+    #13 0x108d652b2 in main FuzzerMain.cpp:19
+    #14 0x7fff6ad5acc8 in start+0x0 (libdyld.dylib:x86_64+0x1acc8)
+
+NOTE: libFuzzer has rudimentary signal handlers.
+      Combine libFuzzer with AddressSanitizer or similar for better crash reports.
+SUMMARY: libFuzzer: deadly signal
+MS: 7 Custom-CustomCrossOver-InsertByte-Custom-ChangeBit-Custom-CustomCrossOver-; base unit: 90863b4d83c3f994bba0a417d0c2ee3b68f9e795
+0x6f,0x70,0x65,0x72,0x61,0x74,0x69,0x6f,0x6e,0x73,0x20,0x7b,0xa,0x20,0x20,0x6b,0x65,0x79,0x3a,0x20,0x22,0x21,0x22,0xa,0x20,0x20,0x76,0x61,0x6c,0x75,0x65,0x3a,0x20,0x22,0x21,0x22,0xa,0x20,0x20,0x74,0x79,0x70,0x65,0x3a,0x20,0x50,0x55,0x54,0xa,0x7d,0xa,0x6f,0x70,0x65,0x72,0x61,0x74,0x69,0x6f,0x6e,0x73,0x20,0x7b,0xa,0x20,0x20,0x6b,0x65,0x79,0x3a,0x20,0x22,0x2b,0x22,0xa,0x20,0x20,0x74,0x79,0x70,0x65,0x3a,0x20,0x50,0x55,0x54,0xa,0x7d,0xa,0x6f,0x70,0x65,0x72,0x61,0x74,0x69,0x6f,0x6e,0x73,0x20,0x7b,0xa,0x20,0x20,0x6b,0x65,0x79,0x3a,0x20,0x22,0x2e,0x22,0xa,0x20,0x20,0x74,0x79,0x70,0x65,0x3a,0x20,0x50,0x55,0x54,0xa,0x7d,0xa,0x6f,0x70,0x65,0x72,0x61,0x74,0x69,0x6f,0x6e,0x73,0x20,0x7b,0xa,0x20,0x20,0x6b,0x65,0x79,0x3a,0x20,0x22,0x5c,0x32,0x35,0x33,0x22,0xa,0x20,0x20,0x74,0x79,0x70,0x65,0x3a,0x20,0x50,0x55,0x54,0xa,0x7d,0xa,
+operations {\x0a  key: \"!\"\x0a  value: \"!\"\x0a  type: PUT\x0a}\x0aoperations {\x0a  key: \"+\"\x0a  type: PUT\x0a}\x0aoperations {\x0a  key: \".\"\x0a  type: PUT\x0a}\x0aoperations {\x0a  key: \"\\253\"\x0a  type: PUT\x0a}\x0a
+artifact_prefix='./'; Test unit written to ./crash-a1460be302d09b548e61787178d9edaa40aea467
+Base64: b3BlcmF0aW9ucyB7CiAga2V5OiAiISIKICB2YWx1ZTogIiEiCiAgdHlwZTogUFVUCn0Kb3BlcmF0aW9ucyB7CiAga2V5OiAiKyIKICB0eXBlOiBQVVQKfQpvcGVyYXRpb25zIHsKICBrZXk6ICIuIgogIHR5cGU6IFBVVAp9Cm9wZXJhdGlvbnMgewogIGtleTogIlwyNTMiCiAgdHlwZTogUFVUCn0K
+./sst_file_writer_fuzzer  5.97s user 4.40s system 64% cpu 16.195 total
+```
+
+Within 6 seconds, it catches the bug.
+
+The input that triggers the bug is persisted in `./crash-a1460be302d09b548e61787178d9edaa40aea467`:
+
+```
+$ cat ./crash-a1460be302d09b548e61787178d9edaa40aea467
+operations {
+  key: "!"
+  value: "!"
+  type: PUT
+}
+operations {
+  key: "+"
+  type: PUT
+}
+operations {
+  key: "."
+  type: PUT
+}
+operations {
+  key: "\253"
+  type: PUT
+}
+```
+
+### Reproduce the crash to debug
+
+The above crash can be reproduced by `./sst_file_writer_fuzzer ./crash-a1460be302d09b548e61787178d9edaa40aea467`,
+so you can debug the crash.
+
+## Future Work
+
+According to [OSS-Fuzz](https://github.com/google/oss-fuzz),
+`as of June 2020, OSS-Fuzz has found over 20,000 bugs in 300 open source projects.`
+
+RocksDB can join OSS-Fuzz together with other open source projects such as sqlite.

fuzz/db_fuzzer.cc

@@ -0,0 +1,159 @@
+#include <fuzzer/FuzzedDataProvider.h>
+
+#include "rocksdb/db.h"
+
+enum OperationType {
+  kPut,
+  kGet,
+  kDelete,
+  kGetProperty,
+  kIterator,
+  kSnapshot,
+  kOpenClose,
+  kColumn,
+  kCompactRange,
+  kSeekForPrev,
+  OP_COUNT
+};
+
+constexpr char db_path[] = "/tmp/testdb";
+
+// Fuzzes DB operations by doing interpretations on the data. Both the
+// sequence of API calls to be called on the DB as well as the arguments
+// to each of these APIs are interpreted by way of the data buffer.
+// The operations that the fuzzer supports are given by the OperationType
+// enum. The goal is to capture sanitizer bugs, so the code should be
+// compiled with a given sanitizer (ASan, UBSan, MSan).
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t* data, size_t size) {
+  rocksdb::DB* db;
+  rocksdb::Options options;
+  options.create_if_missing = true;
+  rocksdb::Status status = rocksdb::DB::Open(options, db_path, &db);
+  if (!status.ok()) {
+    return 0;
+  }
+  FuzzedDataProvider fuzzed_data(data, size);
+
+  // perform a sequence of calls on our db instance
+  int max_iter = static_cast<int>(data[0]);
+  for (int i = 0; i < max_iter && i < size; i++) {
+    OperationType op = static_cast<OperationType>(data[i] % OP_COUNT);
+
+    switch (op) {
+      case kPut: {
+        std::string key = fuzzed_data.ConsumeRandomLengthString();
+        std::string val = fuzzed_data.ConsumeRandomLengthString();
+        db->Put(rocksdb::WriteOptions(), key, val);
+        break;
+      }
+      case kGet: {
+        std::string key = fuzzed_data.ConsumeRandomLengthString();
+        std::string value;
+        db->Get(rocksdb::ReadOptions(), key, &value);
+        break;
+      }
+      case kDelete: {
+        std::string key = fuzzed_data.ConsumeRandomLengthString();
+        db->Delete(rocksdb::WriteOptions(), key);
+        break;
+      }
+      case kGetProperty: {
+        std::string prop;
+        std::string property_name = fuzzed_data.ConsumeRandomLengthString();
+        db->GetProperty(property_name, &prop);
+        break;
+      }
+      case kIterator: {
+        rocksdb::Iterator* it = db->NewIterator(rocksdb::ReadOptions());
+        for (it->SeekToFirst(); it->Valid(); it->Next()) {
+        }
+        delete it;
+        break;
+      }
+      case kSnapshot: {
+        rocksdb::ReadOptions snapshot_options;
+        snapshot_options.snapshot = db->GetSnapshot();
+        rocksdb::Iterator* it = db->NewIterator(snapshot_options);
+        db->ReleaseSnapshot(snapshot_options.snapshot);
+        delete it;
+        break;
+      }
+      case kOpenClose: {
+        db->Close();
+        delete db;
+        status = rocksdb::DB::Open(options, db_path, &db);
+        if (!status.ok()) {
+          rocksdb::DestroyDB(db_path, options);
+          return 0;
+        }
+
+        break;
+      }
+      case kColumn: {
+        rocksdb::ColumnFamilyHandle* cf;
+        rocksdb::Status s;
+        s = db->CreateColumnFamily(rocksdb::ColumnFamilyOptions(), "new_cf",
+                                   &cf);
+        s = db->DestroyColumnFamilyHandle(cf);
+        db->Close();
+        delete db;
+
+        // open DB with two column families
+        std::vector<rocksdb::ColumnFamilyDescriptor> column_families;
+        // have to open default column family
+        column_families.push_back(rocksdb::ColumnFamilyDescriptor(
+            rocksdb::kDefaultColumnFamilyName, rocksdb::ColumnFamilyOptions()));
+        // open the new one, too
+        column_families.push_back(rocksdb::ColumnFamilyDescriptor(
+            "new_cf", rocksdb::ColumnFamilyOptions()));
+        std::vector<rocksdb::ColumnFamilyHandle*> handles;
+        s = rocksdb::DB::Open(rocksdb::DBOptions(), db_path, column_families,
+                              &handles, &db);
+
+        if (s.ok()) {
+          std::string key1 = fuzzed_data.ConsumeRandomLengthString();
+          std::string val1 = fuzzed_data.ConsumeRandomLengthString();
+          std::string key2 = fuzzed_data.ConsumeRandomLengthString();
+          s = db->Put(rocksdb::WriteOptions(), handles[1], key1, val1);
+          std::string value;
+          s = db->Get(rocksdb::ReadOptions(), handles[1], key2, &value);
+          s = db->DropColumnFamily(handles[1]);
+          for (auto handle : handles) {
+            s = db->DestroyColumnFamilyHandle(handle);
+          }
+        } else {
+          status = rocksdb::DB::Open(options, db_path, &db);
+          if (!status.ok()) {
+            // At this point there is no saving to do. So we exit
+            rocksdb::DestroyDB(db_path, rocksdb::Options());
+            return 0;
+          }
+        }
+        break;
+      }
+      case kCompactRange: {
+        std::string slice_start = fuzzed_data.ConsumeRandomLengthString();
+        std::string slice_end = fuzzed_data.ConsumeRandomLengthString();
+
+        rocksdb::Slice begin(slice_start);
+        rocksdb::Slice end(slice_end);
+        rocksdb::CompactRangeOptions options;
+        rocksdb::Status s = db->CompactRange(options, &begin, &end);
+        break;
+      }
+      case kSeekForPrev: {
+        std::string key = fuzzed_data.ConsumeRandomLengthString();
+        auto iter = db->NewIterator(rocksdb::ReadOptions());
+        iter->SeekForPrev(key);
+        delete iter;
+        break;
+      }
+    }
+  }
+
+  // Cleanup DB
+  db->Close();
+  delete db;
+  rocksdb::DestroyDB(db_path, options);
+  return 0;
+}

fuzz/db_map_fuzzer.cc

@@ -0,0 +1,103 @@
+// Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+// This source code is licensed under both the GPLv2 (found in the
+// COPYING file in the root directory) and Apache 2.0 License
+// (found in the LICENSE.Apache file in the root directory).
+
+#include <algorithm>
+#include <iostream>
+#include <map>
+#include <string>
+
+#include "proto/gen/db_operation.pb.h"
+#include "rocksdb/db.h"
+#include "rocksdb/file_system.h"
+#include "src/libfuzzer/libfuzzer_macro.h"
+#include "util.h"
+
+protobuf_mutator::libfuzzer::PostProcessorRegistration<DBOperations> reg = {
+    [](DBOperations* input, unsigned int /* seed */) {
+      const rocksdb::Comparator* comparator = rocksdb::BytewiseComparator();
+      auto ops = input->mutable_operations();
+      // Make sure begin <= end for DELETE_RANGE.
+      for (DBOperation& op : *ops) {
+        if (op.type() == OpType::DELETE_RANGE) {
+          auto begin = op.key();
+          auto end = op.value();
+          if (comparator->Compare(begin, end) > 0) {
+            std::swap(begin, end);
+            op.set_key(begin);
+            op.set_value(end);
+          }
+        }
+      }
+    }};
+
+// Execute randomly generated operations on both a DB and a std::map,
+// then reopen the DB and make sure that iterating the DB produces the
+// same key-value pairs as iterating through the std::map.
+DEFINE_PROTO_FUZZER(DBOperations& input) {
+  if (input.operations().empty()) {
+    return;
+  }
+
+  const std::string kDbPath = "/tmp/db_map_fuzzer_test";
+  auto fs = rocksdb::FileSystem::Default();
+  if (fs->FileExists(kDbPath, rocksdb::IOOptions(), /*dbg=*/nullptr).ok()) {
+    std::cerr << "db path " << kDbPath << " already exists" << std::endl;
+    abort();
+  }
+
+  std::map<std::string, std::string> kv;
+  rocksdb::DB* db = nullptr;
+  rocksdb::Options options;
+  options.create_if_missing = true;
+  CHECK_OK(rocksdb::DB::Open(options, kDbPath, &db));
+
+  for (const DBOperation& op : input.operations()) {
+    switch (op.type()) {
+      case OpType::PUT: {
+        CHECK_OK(db->Put(rocksdb::WriteOptions(), op.key(), op.value()));
+        kv[op.key()] = op.value();
+        break;
+      }
+      case OpType::MERGE: {
+        break;
+      }
+      case OpType::DELETE: {
+        CHECK_OK(db->Delete(rocksdb::WriteOptions(), op.key()));
+        kv.erase(op.key());
+        break;
+      }
+      case OpType::DELETE_RANGE: {
+        // [op.key(), op.value()) corresponds to [begin, end).
+        CHECK_OK(db->DeleteRange(rocksdb::WriteOptions(),
+                                 db->DefaultColumnFamily(), op.key(),
+                                 op.value()));
+        kv.erase(kv.lower_bound(op.key()), kv.lower_bound(op.value()));
+        break;
+      }
+      default: {
+        std::cerr << "Unsupported operation" << static_cast<int>(op.type());
+        return;
+      }
+    }
+  }
+  CHECK_OK(db->Close());
+  delete db;
+  db = nullptr;
+
+  CHECK_OK(rocksdb::DB::Open(options, kDbPath, &db));
+  auto kv_it = kv.begin();
+  rocksdb::Iterator* it = db->NewIterator(rocksdb::ReadOptions());
+  for (it->SeekToFirst(); it->Valid(); it->Next(), kv_it++) {
+    CHECK_TRUE(kv_it != kv.end());
+    CHECK_EQ(it->key().ToString(), kv_it->first);
+    CHECK_EQ(it->value().ToString(), kv_it->second);
+  }
+  CHECK_TRUE(kv_it == kv.end());
+  delete it;
+
+  CHECK_OK(db->Close());
+  delete db;
+  CHECK_OK(rocksdb::DestroyDB(kDbPath, options));
+}

fuzz/proto/db_operation.proto

@@ -0,0 +1,28 @@
+// Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+// This source code is licensed under both the GPLv2 (found in the
+// COPYING file in the root directory) and Apache 2.0 License
+// (found in the LICENSE.Apache file in the root directory).
+
+// Defines database operations.
+// Each operation is a key-value pair and an operation type.
+
+syntax = "proto2";
+
+enum OpType {
+  PUT = 0;
+  MERGE = 1;
+  DELETE = 2;
+  DELETE_RANGE = 3;
+}
+
+message DBOperation {
+  required string key = 1;
+  // value is ignored for DELETE.
+  // [key, value] is the range for DELETE_RANGE.
+  optional string value = 2;
+  required OpType type = 3;
+}
+
+message DBOperations {
+  repeated DBOperation operations = 1;
+}

fuzz/sst_file_writer_fuzzer.cc

@@ -0,0 +1,188 @@
+// Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+// This source code is licensed under both the GPLv2 (found in the
+// COPYING file in the root directory) and Apache 2.0 License
+// (found in the LICENSE.Apache file in the root directory).
+
+#include <algorithm>
+#include <iostream>
+#include <memory>
+#include <string>
+
+#include "proto/gen/db_operation.pb.h"
+#include "rocksdb/file_system.h"
+#include "rocksdb/sst_file_writer.h"
+#include "src/libfuzzer/libfuzzer_macro.h"
+#include "table/table_reader.h"
+#include "util.h"
+
+using namespace ROCKSDB_NAMESPACE;
+
+// Keys in SST file writer operations must be unique and in ascending order.
+// For each DBOperation generated by the fuzzer, this function is called on
+// it to deduplicate and sort the keys in the DBOperations.
+protobuf_mutator::libfuzzer::PostProcessorRegistration<DBOperations> reg = {
+    [](DBOperations* input, unsigned int /* seed */) {
+      const Comparator* comparator = BytewiseComparator();
+      auto ops = input->mutable_operations();
+
+      // Make sure begin <= end for DELETE_RANGE.
+      for (DBOperation& op : *ops) {
+        if (op.type() == OpType::DELETE_RANGE) {
+          auto begin = op.key();
+          auto end = op.value();
+          if (comparator->Compare(begin, end) > 0) {
+            std::swap(begin, end);
+            op.set_key(begin);
+            op.set_value(end);
+          }
+        }
+      }
+
+      std::sort(ops->begin(), ops->end(),
+                [&comparator](const DBOperation& a, const DBOperation& b) {
+                  return comparator->Compare(a.key(), b.key()) < 0;
+                });
+
+      auto last = std::unique(
+          ops->begin(), ops->end(),
+          [&comparator](const DBOperation& a, const DBOperation& b) {
+            return comparator->Compare(a.key(), b.key()) == 0;
+          });
+      ops->erase(last, ops->end());
+    }};
+
+TableReader* NewTableReader(const std::string& sst_file_path,
+                            const Options& options,
+                            const EnvOptions& env_options,
+                            const ImmutableCFOptions& cf_ioptions) {
+  // This code block is similar to SstFileReader::Open.
+
+  uint64_t file_size = 0;
+  std::unique_ptr<RandomAccessFile> file;
+  std::unique_ptr<RandomAccessFileReader> file_reader;
+  std::unique_ptr<TableReader> table_reader;
+  Status s = options.env->GetFileSize(sst_file_path, &file_size);
+  if (s.ok()) {
+    s = options.env->NewRandomAccessFile(sst_file_path, &file, env_options);
+  }
+  if (s.ok()) {
+    file_reader.reset(new RandomAccessFileReader(
+        NewLegacyRandomAccessFileWrapper(file), sst_file_path));
+  }
+  if (s.ok()) {
+    TableReaderOptions t_opt(cf_ioptions, /*prefix_extractor=*/nullptr,
+                             env_options, cf_ioptions.internal_comparator);
+    t_opt.largest_seqno = kMaxSequenceNumber;
+    s = options.table_factory->NewTableReader(t_opt, std::move(file_reader),
+                                              file_size, &table_reader,
+                                              /*prefetch=*/false);
+  }
+  if (!s.ok()) {
+    std::cerr << "Failed to create TableReader for " << sst_file_path << ": "
+              << s.ToString() << std::endl;
+    abort();
+  }
+  return table_reader.release();
+}
+
+ValueType ToValueType(OpType op_type) {
+  switch (op_type) {
+    case OpType::PUT:
+      return ValueType::kTypeValue;
+    case OpType::MERGE:
+      return ValueType::kTypeMerge;
+    case OpType::DELETE:
+      return ValueType::kTypeDeletion;
+    case OpType::DELETE_RANGE:
+      return ValueType::kTypeRangeDeletion;
+    default:
+      std::cerr << "Unknown operation type " << static_cast<int>(op_type)
+                << std::endl;
+      abort();
+  }
+}
+
+// Fuzzes DB operations as input, let SstFileWriter generate a SST file
+// according to the operations, then let TableReader read and check all the
+// key-value pairs from the generated SST file.
+DEFINE_PROTO_FUZZER(DBOperations& input) {
+  if (input.operations().empty()) {
+    return;
+  }
+
+  std::string sstfile;
+  {
+    auto fs = FileSystem::Default();
+    std::string dir;
+    IOOptions opt;
+    CHECK_OK(fs->GetTestDirectory(opt, &dir, nullptr));
+    sstfile = dir + "/SstFileWriterFuzzer.sst";
+  }
+
+  Options options;
+  EnvOptions env_options(options);
+  ImmutableCFOptions cf_ioptions(options);
+
+  // Generate sst file.
+  SstFileWriter writer(env_options, options);
+  CHECK_OK(writer.Open(sstfile));
+  for (const DBOperation& op : input.operations()) {
+    switch (op.type()) {
+      case OpType::PUT: {
+        CHECK_OK(writer.Put(op.key(), op.value()));
+        break;
+      }
+      case OpType::MERGE: {
+        CHECK_OK(writer.Merge(op.key(), op.value()));
+        break;
+      }
+      case OpType::DELETE: {
+        CHECK_OK(writer.Delete(op.key()));
+        break;
+      }
+      case OpType::DELETE_RANGE: {
+        CHECK_OK(writer.DeleteRange(op.key(), op.value()));
+        break;
+      }
+      default: {
+        std::cerr << "Unsupported operation" << static_cast<int>(op.type())
+                  << std::endl;
+        abort();
+      }
+    }
+  }
+  ExternalSstFileInfo info;
+  CHECK_OK(writer.Finish(&info));
+
+  // Iterate and verify key-value pairs.
+  std::unique_ptr<TableReader> table_reader(
+      NewTableReader(sstfile, options, env_options, cf_ioptions));
+  ReadOptions roptions;
+  CHECK_OK(table_reader->VerifyChecksum(roptions,
+                                        TableReaderCaller::kUncategorized));
+  std::unique_ptr<InternalIterator> it(
+      table_reader->NewIterator(roptions, /*prefix_extractor=*/nullptr,
+                                /*arena=*/nullptr, /*skip_filters=*/true,
+                                TableReaderCaller::kUncategorized));
+  it->SeekToFirst();
+  for (const DBOperation& op : input.operations()) {
+    if (op.type() == OpType::DELETE_RANGE) {
+      // InternalIterator cannot iterate over DELETE_RANGE entries.
+      continue;
+    }
+    CHECK_TRUE(it->Valid());
+    ParsedInternalKey ikey;
+    CHECK_OK(ParseInternalKey(it->key(), &ikey, /*log_err_key=*/true));
+    CHECK_EQ(ikey.user_key.ToString(), op.key());
+    CHECK_EQ(ikey.sequence, 0);
+    CHECK_EQ(ikey.type, ToValueType(op.type()));
+    if (op.type() != OpType::DELETE) {
+      CHECK_EQ(op.value(), it->value().ToString());
+    }
+    it->Next();
+  }
+  CHECK_TRUE(!it->Valid());
+
+  // Delete sst file.
+  remove(sstfile.c_str());
+}

fuzz/util.h

@@ -0,0 +1,23 @@
+#pragma once
+
+#define CHECK_OK(expression)                       \
+  do {                                             \
+    auto status = (expression);                    \
+    if (!status.ok()) {                            \
+      std::cerr << status.ToString() << std::endl; \
+      abort();                                     \
+    }                                              \
+  } while (0)
+
+#define CHECK_EQ(a, b)                                                      \
+  if (a != b) {                                                             \
+    std::cerr << "(" << #a << "=" << a << ") != (" << #b << "=" << b << ")" \
+              << std::endl;                                                 \
+    abort();                                                                \
+  }
+
+#define CHECK_TRUE(cond)                                      \
+  if (!(cond)) {                                              \
+    std::cerr << "\"" << #cond << "\" is false" << std::endl; \
+    abort();                                                  \
+  }

include/rocksdb/advanced_options.h

@@ -675,7 +675,11 @@ struct AdvancedColumnFamilyOptions {
   // Dynamically changeable through SetOptions() API
   bool report_bg_io_stats = false;
 
-  // Files older than TTL will go through the compaction process.
+  // Files containing updates older than TTL will go through the compaction
+  // process. This usually happens in a cascading way so that those entries
+  // will be compacted to bottommost level/file.
+  // The feature is used to remove stale entries that have been deleted or
+  // updated from the file system.
   // Pre-req: This needs max_open_files to be set to -1.
   // In Level: Non-bottom-level files older than TTL will go through the
   //           compation process.
@@ -695,6 +699,9 @@ struct AdvancedColumnFamilyOptions {
 
   // Files older than this value will be picked up for compaction, and
   // re-written to the same level as they were before.
+  // One main use of the feature is to make sure a file goes through compaction
+  // filters periodically. Users can also use the feature to clear up SST
+  // files using old format.
   //
   // A file's age is computed by looking at file_creation_time or creation_time
   // table properties in order, if they have valid non-zero values; if not, the