Add Visual Studio 2015 to AppVeyor (#5446)

Summary:
This is required to compile on Windows with Visual Studio 2015, which is used for creating the RocksJava releases.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5446

Differential Revision: D18924811

fbshipit-source-id: a183a62e79a2af5aaf59cd08235458a172fe7dcb

.travis.yml

@@ -70,7 +70,7 @@ install:
       CC=gcc-8 && CXX=g++-8;
     fi
   - if [[ "${JOB_NAME}" == cmake* ]] && [ "${TRAVIS_OS_NAME}" == linux ]; then
-      mkdir cmake-dist && curl -sfSL https://github.com/Kitware/CMake/releases/download/v3.14.5/cmake-3.14.5-Linux-x86_64.tar.gz | tar --strip-components=1 -C cmake-dist -xz && export PATH=$PWD/cmake-dist/bin:$PATH;
+      mkdir cmake-dist && curl --silent --fail --show-error --location https://github.com/Kitware/CMake/releases/download/v3.14.5/cmake-3.14.5-Linux-x86_64.tar.gz | tar --strip-components=1 -C cmake-dist -xz && export PATH=$PWD/cmake-dist/bin:$PATH;
     fi
   - if [[ "${JOB_NAME}" == java_test ]]; then
       java -version && echo "JAVA_HOME=${JAVA_HOME}";

CMakeLists.txt

@@ -14,7 +14,7 @@
 #        cd build
 # 3. Run cmake to generate project files for Windows, add more options to enable required third-party libraries.
 #    See thirdparty.inc for more information.
-#        sample command: cmake -G "Visual Studio 15 Win64" -DWITH_GFLAGS=1 -DWITH_SNAPPY=1 -DWITH_JEMALLOC=1 -DWITH_JNI=1 ..
+#        sample command: cmake -G "Visual Studio 15 Win64" -DCMAKE_BUILD_TYPE=Release -DWITH_GFLAGS=1 -DWITH_SNAPPY=1 -DWITH_JEMALLOC=1 -DWITH_JNI=1 ..
 # 4. Then build the project in debug mode (you may want to add /m[:<N>] flag to run msbuild in <N> parallel threads
 #                                          or simply /m to use all avail cores)
 #        msbuild rocksdb.sln
@@ -63,7 +63,12 @@ endif()
 # third-party/folly is only validated to work on Linux and Windows for now.
 # So only turn it on there by default.
 if(CMAKE_SYSTEM_NAME MATCHES "Linux" OR CMAKE_SYSTEM_NAME MATCHES "Windows")
-  option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" ON)
+  if(MSVC AND MSVC_VERSION LESS 1910)
+    # Folly does not compile with MSVC older than VS2017
+    option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" OFF)
+  else()
+    option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" ON)
+  endif()
 else()
   option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" OFF)
 endif()
@@ -174,7 +179,7 @@ else()
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -W -Wextra -Wall")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wsign-compare -Wshadow -Wno-unused-parameter -Wno-unused-variable -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers -Wno-strict-aliasing")
   if(MINGW)
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-format")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-format -fno-asynchronous-unwind-tables")
     add_definitions(-D_POSIX_C_SOURCE=1)
   endif()
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
@@ -885,7 +890,9 @@ if(NOT WIN32 OR ROCKSDB_INSTALL_ON_WINDOWS)
 endif()
 
 option(WITH_TESTS "build with tests" ON)
-if(WITH_TESTS)
+# For test libraries, utilities, and exes that are build iff WITH_TESTS=ON and
+# in Debug mode. Add test only code that is not #ifdefed for Release here.
+if(WITH_TESTS AND CMAKE_BUILD_TYPE STREQUAL "Debug")
   add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
   set(TESTS
         cache/cache_test.cc
@@ -1040,24 +1047,6 @@ if(WITH_TESTS)
     list(APPEND TESTS third-party/folly/folly/synchronization/test/DistributedMutexTest.cpp)
   endif()
 
-  set(BENCHMARKS
-    cache/cache_bench.cc
-    memtable/memtablerep_bench.cc
-    db/range_del_aggregator_bench.cc
-    tools/db_bench.cc
-    table/table_reader_bench.cc
-    util/filter_bench.cc
-    utilities/persistent_cache/hash_table_bench.cc)
-  add_library(testharness OBJECT test_util/testharness.cc)
-  foreach(sourcefile ${BENCHMARKS})
-    get_filename_component(exename ${sourcefile} NAME_WE)
-    add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile}
-      $<TARGET_OBJECTS:testharness>)
-    target_link_libraries(${exename}${ARTIFACT_SUFFIX} gtest ${LIBS})
-  endforeach(sourcefile ${BENCHMARKS})
-
-  # For test util library that is build only in DEBUG mode
-  # and linked to tests. Add test only code that is not #ifdefed for Release here.
   set(TESTUTIL_SOURCE
       db/db_test_util.cc
       monitoring/thread_status_updater_debug.cc
@@ -1065,7 +1054,6 @@ if(WITH_TESTS)
       test_util/fault_injection_test_env.cc
       utilities/cassandra/test_utils.cc
   )
-  # test utilities are only build in debug
   enable_testing()
   add_custom_target(check COMMAND ${CMAKE_CTEST_COMMAND})
   set(TESTUTILLIB testutillib${ARTIFACT_SUFFIX})
@@ -1080,9 +1068,7 @@ if(WITH_TESTS)
         EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
         )
 
-  # Tests are excluded from Release builds
   set(TEST_EXES ${TESTS})
-
   foreach(sourcefile ${TEST_EXES})
       get_filename_component(exename ${sourcefile} NAME_WE)
       add_executable(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX} ${sourcefile}
@@ -1119,6 +1105,29 @@ if(WITH_TESTS)
   endforeach(sourcefile ${C_TEST_EXES})
 endif()
 
+option(WITH_BENCHMARK_TOOLS "build with benchmarks" ON)
+if(WITH_BENCHMARK_TOOLS)
+  if(NOT TARGET gtest)
+    add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
+  endif()
+  set(BENCHMARKS
+        cache/cache_bench.cc
+        memtable/memtablerep_bench.cc
+        db/range_del_aggregator_bench.cc
+        tools/db_bench.cc
+        table/table_reader_bench.cc
+        util/filter_bench.cc
+        utilities/persistent_cache/hash_table_bench.cc
+  )
+  add_library(testharness OBJECT test_util/testharness.cc)
+  foreach(sourcefile ${BENCHMARKS})
+    get_filename_component(exename ${sourcefile} NAME_WE)
+    add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile}
+      $<TARGET_OBJECTS:testharness>)
+    target_link_libraries(${exename}${ARTIFACT_SUFFIX} gtest ${LIBS})
+  endforeach(sourcefile ${BENCHMARKS})
+endif()
+
 option(WITH_TOOLS "build with tools" ON)
 if(WITH_TOOLS)
   add_subdirectory(tools)

HISTORY.md

@@ -1,4 +1,28 @@
 # Rocksdb Change Log
+## Unreleased
+
+## 6.6.4 (1/31/2020)
+### Bug Fixes
+* Fixed issue #6316 that can cause a corruption of the MANIFEST file in the middle when writing to it fails due to no disk space.
+
+## 6.6.3 (01/24/2020)
+### Bug Fixes
+* Fix a bug that can cause write threads to hang when a slowdown/stall happens and there is a mix of writers with WriteOptions::no_slowdown set/unset.
+
+## 6.6.2 (01/13/2020)
+### Bug Fixes
+* Fixed a bug where non-L0 compaction input files were not considered to compute the `creation_time` of new compaction outputs.
+
+## 6.6.1 (01/02/2020)
+### Bug Fixes
+* Fix a bug in WriteBatchWithIndex::MultiGetFromBatchAndDB, which is called by Transaction::MultiGet, that causes due to stale pointer access when the number of keys is > 32
+* Fixed two performance issues related to memtable history trimming. First, a new SuperVersion is now created only if some memtables were actually trimmed. Second, trimming is only scheduled if there is at least one flushed memtable that is kept in memory for the purposes of transaction conflict checking.
+* BlobDB no longer updates the SST to blob file mapping upon failed compactions.
+* Fix a bug in which a snapshot read through an iterator could be affected by a DeleteRange after the snapshot (#6062).
+* Fixed a bug where BlobDB was comparing the `ColumnFamilyHandle` pointers themselves instead of only the column family IDs when checking whether an API call uses the default column family or not.
+* Delete superversions in BackgroundCallPurge.
+* Fix use-after-free and double-deleting files in BackgroundCallPurge().
+
 ## 6.6.0 (11/25/2019)
 ### Bug Fixes
 * Fix data corruption casued by output of intra-L0 compaction on ingested file not being placed in correct order in L0.
@@ -19,8 +43,9 @@
 * A batched MultiGet API (DB::MultiGet()) that supports retrieving keys from multiple column families.
 * Full and partitioned filters in the block-based table use an improved Bloom filter implementation, enabled with format_version 5 (or above) because previous releases cannot read this filter. This replacement is faster and more accurate, especially for high bits per key or millions of keys in a single (full) filter. For example, the new Bloom filter has the same false postive rate at 9.55 bits per key as the old one at 10 bits per key, and a lower false positive rate at 16 bits per key than the old one at 100 bits per key.
 * Added AVX2 instructions to USE_SSE builds to accelerate the new Bloom filter and XXH3-based hash function on compatible x86_64 platforms (Haswell and later, ~2014).
-* Support options.ttl with options.max_open_files = -1. File's oldest ancester time will be written to manifest. If it is availalbe, this information will be used instead of creation_time in table properties.
+* Support options.ttl or options.periodic_compaction_seconds with options.max_open_files = -1. File's oldest ancester time and file creation time will be written to manifest. If it is availalbe, this information will be used instead of creation_time and file_creation_time in table properties.
 * Setting options.ttl for universal compaction now has the same meaning as setting periodic_compaction_seconds.
+* SstFileMetaData also returns file creation time and oldest ancester time.
 * The `sst_dump` command line tool `recompress` command now displays how many blocks were compressed and how many were not, in particular how many were not compressed because the compression ratio was not met (12.5% threshold for GoodCompressionRatio), as seen in the `number.block.not_compressed` counter stat since version 6.0.0.
 * The block cache usage is now takes into account the overhead of metadata per each entry. This results into more accurate managment of memory. A side-effect of this feature is that less items are fit into the block cache of the same size, which would result to higher cache miss rates. This can be remedied by increasing the block cache size or passing kDontChargeCacheMetadata to its constuctor to restore the old behavior.
 * When using BlobDB, a mapping is maintained and persisted in the MANIFEST between each SST file and the oldest non-TTL blob file it references.

Makefile

@@ -1103,16 +1103,21 @@ unity_test: db/db_test.o db/db_test_util.o $(TESTHARNESS) $(TOOLLIBOBJECTS) unit
 rocksdb.h rocksdb.cc: build_tools/amalgamate.py Makefile $(LIB_SOURCES) unity.cc
 	build_tools/amalgamate.py -I. -i./include unity.cc -x include/rocksdb/c.h -H rocksdb.h -o rocksdb.cc
 
-clean: clean-ext-libraries-all clean-rocks
+clean: clean-ext-libraries-all clean-rocks clean-rocksjava
 
-clean-not-downloaded: clean-ext-libraries-bin clean-rocks
+clean-not-downloaded: clean-ext-libraries-bin clean-rocks clean-not-downloaded-rocksjava
 
 clean-rocks:
 	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(LIBRARY) $(SHARED)
 	rm -rf $(CLEAN_FILES) ios-x86 ios-arm scan_build_report
 	$(FIND) . -name "*.[oda]" -exec rm -f {} \;
 	$(FIND) . -type f -regex ".*\.\(\(gcda\)\|\(gcno\)\)" -exec rm {} \;
-	cd java; $(MAKE) clean
+
+clean-rocksjava:
+	cd java && $(MAKE) clean
+
+clean-not-downloaded-rocksjava:
+	cd java && $(MAKE) clean-not-downloaded
 
 clean-ext-libraries-all:
 	rm -rf bzip2* snappy* zlib* lz4* zstd*
@@ -1806,7 +1811,7 @@ endif
 libz.a:
 	-rm -rf zlib-$(ZLIB_VER)
 ifeq (,$(wildcard ./zlib-$(ZLIB_VER).tar.gz))
-	curl --output zlib-$(ZLIB_VER).tar.gz -L ${ZLIB_DOWNLOAD_BASE}/zlib-$(ZLIB_VER).tar.gz
+	curl --fail --output zlib-$(ZLIB_VER).tar.gz --location ${ZLIB_DOWNLOAD_BASE}/zlib-$(ZLIB_VER).tar.gz
 endif
 	ZLIB_SHA256_ACTUAL=`$(SHA256_CMD) zlib-$(ZLIB_VER).tar.gz | cut -d ' ' -f 1`; \
 	if [ "$(ZLIB_SHA256)" != "$$ZLIB_SHA256_ACTUAL" ]; then \
@@ -1820,7 +1825,7 @@ endif
 libbz2.a:
 	-rm -rf bzip2-$(BZIP2_VER)
 ifeq (,$(wildcard ./bzip2-$(BZIP2_VER).tar.gz))
-	curl --output bzip2-$(BZIP2_VER).tar.gz -L ${CURL_SSL_OPTS} ${BZIP2_DOWNLOAD_BASE}/bzip2-$(BZIP2_VER).tar.gz
+	curl --fail --output bzip2-$(BZIP2_VER).tar.gz --location ${CURL_SSL_OPTS} ${BZIP2_DOWNLOAD_BASE}/bzip2-$(BZIP2_VER).tar.gz
 endif
 	BZIP2_SHA256_ACTUAL=`$(SHA256_CMD) bzip2-$(BZIP2_VER).tar.gz | cut -d ' ' -f 1`; \
 	if [ "$(BZIP2_SHA256)" != "$$BZIP2_SHA256_ACTUAL" ]; then \
@@ -1834,7 +1839,7 @@ endif
 libsnappy.a:
 	-rm -rf snappy-$(SNAPPY_VER)
 ifeq (,$(wildcard ./snappy-$(SNAPPY_VER).tar.gz))
-	curl --output snappy-$(SNAPPY_VER).tar.gz -L ${CURL_SSL_OPTS} ${SNAPPY_DOWNLOAD_BASE}/$(SNAPPY_VER).tar.gz
+	curl --fail --output snappy-$(SNAPPY_VER).tar.gz --location ${CURL_SSL_OPTS} ${SNAPPY_DOWNLOAD_BASE}/$(SNAPPY_VER).tar.gz
 endif
 	SNAPPY_SHA256_ACTUAL=`$(SHA256_CMD) snappy-$(SNAPPY_VER).tar.gz | cut -d ' ' -f 1`; \
 	if [ "$(SNAPPY_SHA256)" != "$$SNAPPY_SHA256_ACTUAL" ]; then \
@@ -1849,7 +1854,7 @@ endif
 liblz4.a:
 	-rm -rf lz4-$(LZ4_VER)
 ifeq (,$(wildcard ./lz4-$(LZ4_VER).tar.gz))
-	curl --output lz4-$(LZ4_VER).tar.gz -L ${CURL_SSL_OPTS} ${LZ4_DOWNLOAD_BASE}/v$(LZ4_VER).tar.gz
+	curl --fail --output lz4-$(LZ4_VER).tar.gz --location ${CURL_SSL_OPTS} ${LZ4_DOWNLOAD_BASE}/v$(LZ4_VER).tar.gz
 endif
 	LZ4_SHA256_ACTUAL=`$(SHA256_CMD) lz4-$(LZ4_VER).tar.gz | cut -d ' ' -f 1`; \
 	if [ "$(LZ4_SHA256)" != "$$LZ4_SHA256_ACTUAL" ]; then \
@@ -1863,7 +1868,7 @@ endif
 libzstd.a:
 	-rm -rf zstd-$(ZSTD_VER)
 ifeq (,$(wildcard ./zstd-$(ZSTD_VER).tar.gz))
-	curl --output zstd-$(ZSTD_VER).tar.gz -L ${CURL_SSL_OPTS} ${ZSTD_DOWNLOAD_BASE}/v$(ZSTD_VER).tar.gz
+	curl --fail --output zstd-$(ZSTD_VER).tar.gz --location ${CURL_SSL_OPTS} ${ZSTD_DOWNLOAD_BASE}/v$(ZSTD_VER).tar.gz
 endif
 	ZSTD_SHA256_ACTUAL=`$(SHA256_CMD) zstd-$(ZSTD_VER).tar.gz | cut -d ' ' -f 1`; \
 	if [ "$(ZSTD_SHA256)" != "$$ZSTD_SHA256_ACTUAL" ]; then \
@@ -1933,35 +1938,35 @@ rocksdbjavastaticreleasedocker: rocksdbjavastatic rocksdbjavastaticdockerx86 roc
 
 rocksdbjavastaticdockerx86:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x86-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos6_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_x86-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos6_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerx86_64:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x64-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos6_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_x64-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos6_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerppc64le:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_ppc64le-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos7_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_ppc64le-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos7_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerarm64v8:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_arm64v8-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos7_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_arm64v8-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos7_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerx86musl:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x86-musl-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_x86-musl-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerx86_64musl:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x64-musl-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_x64-musl-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerppc64lemusl:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_ppc64le-musl-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_ppc64le-musl-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerarm64v8musl:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_arm64v8-musl-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_arm64v8-musl-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticpublish: rocksdbjavastaticrelease rocksdbjavastaticpublishcentral
 

appveyor.yml

@@ -17,41 +17,49 @@ environment:
   ZSTD_INCLUDE: $(ZSTD_HOME)\lib;$(ZSTD_HOME)\lib\dictBuilder
   ZSTD_LIB_DEBUG: $(ZSTD_HOME)\build\VS2010\bin\x64_Debug\libzstd_static.lib
   ZSTD_LIB_RELEASE: $(ZSTD_HOME)\build\VS2010\bin\x64_Release\libzstd_static.lib
+  matrix:
+    - APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2015
+      CMAKE_GENERATOR: Visual Studio 14 Win64
+      DEV_ENV: C:\Program Files (x86)\Microsoft Visual Studio 14.0\Common7\IDE\devenv.com
+    - APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2017
+      CMAKE_GENERATOR: Visual Studio 15 Win64
+      DEV_ENV: C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\IDE\devenv.com
 
 install:
   - md %THIRDPARTY_HOME%
   - echo "Building Snappy dependency..."
   - cd %THIRDPARTY_HOME%
-  - curl -fsSL -o snappy-1.1.7.zip https://github.com/google/snappy/archive/1.1.7.zip
+  - curl --fail --silent --show-error --output snappy-1.1.7.zip --location https://github.com/google/snappy/archive/1.1.7.zip
   - unzip snappy-1.1.7.zip
   - cd snappy-1.1.7
   - mkdir build
   - cd build
-  - cmake -DCMAKE_GENERATOR_PLATFORM=x64 ..
+  - cmake -G "%CMAKE_GENERATOR%" ..
   - msbuild Snappy.sln /p:Configuration=Debug /p:Platform=x64
   - msbuild Snappy.sln /p:Configuration=Release /p:Platform=x64
   - echo "Building LZ4 dependency..."
   - cd %THIRDPARTY_HOME%
-  - curl -fsSL -o lz4-1.8.3.zip https://github.com/lz4/lz4/archive/v1.8.3.zip
+  - curl --fail --silent --show-error --output lz4-1.8.3.zip --location https://github.com/lz4/lz4/archive/v1.8.3.zip
   - unzip lz4-1.8.3.zip
   - cd lz4-1.8.3\visual\VS2010
-  - ps: $CMD="C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\IDE\devenv.com"; & $CMD lz4.sln /upgrade
+  - ps: $CMD="$Env:DEV_ENV"; & $CMD lz4.sln /upgrade
   - msbuild lz4.sln /p:Configuration=Debug /p:Platform=x64
   - msbuild lz4.sln /p:Configuration=Release /p:Platform=x64
   - echo "Building ZStd dependency..."
   - cd %THIRDPARTY_HOME%
-  - curl -fsSL -o zstd-1.4.0.zip https://github.com/facebook/zstd/archive/v1.4.0.zip
+  - curl --fail --silent --show-error --output zstd-1.4.0.zip --location https://github.com/facebook/zstd/archive/v1.4.0.zip
   - unzip zstd-1.4.0.zip
   - cd zstd-1.4.0\build\VS2010
-  - ps: $CMD="C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\IDE\devenv.com"; & $CMD zstd.sln /upgrade
+  - ps: $CMD="$Env:DEV_ENV"; & $CMD zstd.sln /upgrade
   - msbuild zstd.sln /p:Configuration=Debug /p:Platform=x64
   - msbuild zstd.sln /p:Configuration=Release /p:Platform=x64
 
 before_build:
   - md %APPVEYOR_BUILD_FOLDER%\build
   - cd %APPVEYOR_BUILD_FOLDER%\build
-  - cmake -G "Visual Studio 15 Win64" -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DLZ4=1 -DZSTD=1 -DXPRESS=1 -DJNI=1 ..
+  - cmake -G "%CMAKE_GENERATOR%" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DLZ4=1 -DZSTD=1 -DXPRESS=1 -DJNI=1 ..
   - cd ..
+
 build:
   project: build\rocksdb.sln
   parallel: true

build_tools/precommit_checker.py

@@ -1,4 +1,4 @@
-#!/usr/local/fbcode/gcc-4.9-glibc-2.20-fb/bin/python2.7
+#!/usr/bin/env python2.7
 # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
 
 from __future__ import absolute_import

build_tools/regression_build_test.sh

@@ -378,7 +378,7 @@ function send_to_ods {
     echo >&2 "ERROR: Key $key doesn't have a value."
     return
   fi
-  curl -s "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build$git_br&key=$key&value=$value" \
+  curl --silent "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build$git_br&key=$key&value=$value" \
     --connect-timeout 60
 }
 

build_tools/rocksdb-lego-determinator

@@ -778,7 +778,7 @@ run_regression()
 
   # parameters: $1 -- key, $2 -- value
   function send_size_to_ods {
-    curl -s "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=rocksdb.build_size.$1&value=$2" \
+    curl --silent "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=rocksdb.build_size.$1&value=$2" \
       --connect-timeout 60
   }
 

cache/lru_cache.h

@@ -133,7 +133,6 @@ struct LRUHandle {
   // Caclculate the memory usage by metadata
   inline size_t CalcTotalCharge(
       CacheMetadataChargePolicy metadata_charge_policy) {
-    assert(key_length);
     size_t meta_charge = 0;
     if (metadata_charge_policy == kFullChargeCacheMetadata) {
 #ifdef ROCKSDB_MALLOC_USABLE_SIZE

db/arena_wrapped_db_iter.cc

@@ -64,7 +64,7 @@ Status ArenaWrappedDBIter::Refresh() {
     arena_.~Arena();
     new (&arena_) Arena();
 
-    SuperVersion* sv = cfd_->GetReferencedSuperVersion(db_impl_->mutex());
+    SuperVersion* sv = cfd_->GetReferencedSuperVersion(db_impl_);
     if (read_callback_) {
       read_callback_->Refresh(latest_seq);
     }

db/column_family.cc

@@ -1080,9 +1080,8 @@ Compaction* ColumnFamilyData::CompactRange(
   return result;
 }
 
-SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(
+SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(DBImpl* db) {
-    InstrumentedMutex* db_mutex) {
+  SuperVersion* sv = GetThreadLocalSuperVersion(db);
-  SuperVersion* sv = GetThreadLocalSuperVersion(db_mutex);
   sv->Ref();
   if (!ReturnThreadLocalSuperVersion(sv)) {
     // This Unref() corresponds to the Ref() in GetThreadLocalSuperVersion()
@@ -1094,8 +1093,7 @@ SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(
   return sv;
 }
 
-SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
+SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(DBImpl* db) {
-    InstrumentedMutex* db_mutex) {
   // The SuperVersion is cached in thread local storage to avoid acquiring
   // mutex when SuperVersion does not change since the last use. When a new
   // SuperVersion is installed, the compaction or flush thread cleans up
@@ -1122,16 +1120,21 @@ SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
 
     if (sv && sv->Unref()) {
       RecordTick(ioptions_.statistics, NUMBER_SUPERVERSION_CLEANUPS);
-      db_mutex->Lock();
+      db->mutex()->Lock();
       // NOTE: underlying resources held by superversion (sst files) might
       // not be released until the next background job.
       sv->Cleanup();
-      sv_to_delete = sv;
+      if (db->immutable_db_options().avoid_unnecessary_blocking_io) {
+        db->AddSuperVersionsToFreeQueue(sv);
+        db->SchedulePurge();
+      } else {
+        sv_to_delete = sv;
+      }
     } else {
-      db_mutex->Lock();
+      db->mutex()->Lock();
     }
     sv = super_version_->Ref();
-    db_mutex->Unlock();
+    db->mutex()->Unlock();
 
     delete sv_to_delete;
   }

db/column_family.h

@@ -430,11 +430,11 @@ class ColumnFamilyData {
   SuperVersion* GetSuperVersion() { return super_version_; }
   // thread-safe
   // Return a already referenced SuperVersion to be used safely.
-  SuperVersion* GetReferencedSuperVersion(InstrumentedMutex* db_mutex);
+  SuperVersion* GetReferencedSuperVersion(DBImpl* db);
   // thread-safe
   // Get SuperVersion stored in thread local storage. If it does not exist,
   // get a reference from a current SuperVersion.
-  SuperVersion* GetThreadLocalSuperVersion(InstrumentedMutex* db_mutex);
+  SuperVersion* GetThreadLocalSuperVersion(DBImpl* db);
   // Try to return SuperVersion back to thread local storage. Retrun true on
   // success and false on failure. It fails when the thread local storage
   // contains anything other than SuperVersion::kSVInUse flag.

db/column_family_test.cc

@@ -67,9 +67,9 @@ class ColumnFamilyTestBase : public testing::Test {
 #ifndef ROCKSDB_LITE
     const char* test_env_uri = getenv("TEST_ENV_URI");
     if (test_env_uri) {
-      Status s = ObjectRegistry::NewInstance()->NewSharedObject(test_env_uri,
+      Env* test_env = nullptr;
-                                                                &env_guard_);
+      Status s = Env::LoadEnv(test_env_uri, &test_env, &env_guard_);
-      base_env = env_guard_.get();
+      base_env = test_env;
       EXPECT_OK(s);
       EXPECT_NE(Env::Default(), base_env);
     }

db/compaction/compaction.cc

@@ -547,11 +547,13 @@ bool Compaction::ShouldFormSubcompactions() const {
 
 uint64_t Compaction::MinInputFileOldestAncesterTime() const {
   uint64_t min_oldest_ancester_time = port::kMaxUint64;
-  for (const auto& file : inputs_[0].files) {
+  for (const auto& level_files : inputs_) {
-    uint64_t oldest_ancester_time = file->TryGetOldestAncesterTime();
+    for (const auto& file : level_files.files) {
-    if (oldest_ancester_time != 0) {
+      uint64_t oldest_ancester_time = file->TryGetOldestAncesterTime();
-      min_oldest_ancester_time =
+      if (oldest_ancester_time != 0) {
-          std::min(min_oldest_ancester_time, oldest_ancester_time);
+        min_oldest_ancester_time =
+            std::min(min_oldest_ancester_time, oldest_ancester_time);
+      }
     }
   }
   return min_oldest_ancester_time;

db/compaction/compaction_job.cc

@@ -1501,6 +1501,7 @@ Status CompactionJob::OpenCompactionOutputFile(
     out.meta.fd = FileDescriptor(file_number,
                                  sub_compact->compaction->output_path_id(), 0);
     out.meta.oldest_ancester_time = oldest_ancester_time;
+    out.meta.file_creation_time = current_time;
     out.finished = false;
     sub_compact->outputs.push_back(out);
   }

db/compaction/compaction_job_test.cc

@@ -184,7 +184,7 @@ class CompactionJobTest : public testing::Test {
     VersionEdit edit;
     edit.AddFile(level, file_number, 0, 10, smallest_key, largest_key,
                  smallest_seqno, largest_seqno, false, oldest_blob_file_number,
-                 kUnknownOldestAncesterTime);
+                 kUnknownOldestAncesterTime, kUnknownFileCreationTime);
 
     mutex_.Lock();
     versions_->LogAndApply(versions_->GetColumnFamilySet()->GetDefault(),

db/compaction/compaction_picker_test.cc

@@ -95,7 +95,7 @@ class CompactionPickerTest : public testing::Test {
         InternalKey(smallest, smallest_seq, kTypeValue),
         InternalKey(largest, largest_seq, kTypeValue), smallest_seq,
         largest_seq, /* marked_for_compact */ false, kInvalidBlobFileNumber,
-        kUnknownOldestAncesterTime);
+        kUnknownOldestAncesterTime, kUnknownFileCreationTime);
     f->compensated_file_size =
         (compensated_file_size != 0) ? compensated_file_size : file_size;
     vstorage_->AddFile(level, f);

db/db_compaction_test.cc

@@ -3576,6 +3576,15 @@ TEST_F(DBCompactionTest, LevelTtlCascadingCompactions) {
         ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
       }
       Flush();
+      // Get the first file's creation time. This will be the oldest file in the
+      // DB. Compactions inolving this file's descendents should keep getting
+      // this time.
+      std::vector<std::vector<FileMetaData>> level_to_files;
+      dbfull()->TEST_GetFilesMetaData(dbfull()->DefaultColumnFamily(),
+                                      &level_to_files);
+      uint64_t oldest_time = level_to_files[0][0].oldest_ancester_time;
+      // Add 1 hour and do another flush.
+      env_->addon_time_.fetch_add(1 * 60 * 60);
       for (int i = 101; i <= 200; ++i) {
         ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
       }
@@ -3583,11 +3592,13 @@ TEST_F(DBCompactionTest, LevelTtlCascadingCompactions) {
       MoveFilesToLevel(6);
       ASSERT_EQ("0,0,0,0,0,0,2", FilesPerLevel());
 
+      env_->addon_time_.fetch_add(1 * 60 * 60);
       // Add two L4 files with key ranges: [1 .. 50], [51 .. 150].
       for (int i = 1; i <= 50; ++i) {
         ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
       }
       Flush();
+      env_->addon_time_.fetch_add(1 * 60 * 60);
       for (int i = 51; i <= 150; ++i) {
         ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
       }
@@ -3595,6 +3606,7 @@ TEST_F(DBCompactionTest, LevelTtlCascadingCompactions) {
       MoveFilesToLevel(4);
       ASSERT_EQ("0,0,0,0,2,0,2", FilesPerLevel());
 
+      env_->addon_time_.fetch_add(1 * 60 * 60);
       // Add one L1 file with key range: [26, 75].
       for (int i = 26; i <= 75; ++i) {
         ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
@@ -3636,6 +3648,10 @@ TEST_F(DBCompactionTest, LevelTtlCascadingCompactions) {
       ASSERT_EQ("1,0,0,0,0,0,1", FilesPerLevel());
       ASSERT_EQ(5, ttl_compactions);
 
+      dbfull()->TEST_GetFilesMetaData(dbfull()->DefaultColumnFamily(),
+                                      &level_to_files);
+      ASSERT_EQ(oldest_time, level_to_files[6][0].oldest_ancester_time);
+
       env_->addon_time_.fetch_add(25 * 60 * 60);
       ASSERT_OK(Put(Key(2), "1"));
       if (if_restart) {
@@ -3657,71 +3673,103 @@ TEST_F(DBCompactionTest, LevelPeriodicCompaction) {
   const int kNumLevelFiles = 2;
   const int kValueSize = 100;
 
-  Options options = CurrentOptions();
+  for (bool if_restart : {false, true}) {
-  options.periodic_compaction_seconds = 48 * 60 * 60;  // 2 days
+    for (bool if_open_all_files : {false, true}) {
-  options.max_open_files = -1;       // needed for ttl compaction
+      Options options = CurrentOptions();
-  env_->time_elapse_only_sleep_ = false;
+      options.periodic_compaction_seconds = 48 * 60 * 60;  // 2 days
-  options.env = env_;
+      if (if_open_all_files) {
+        options.max_open_files = -1;  // needed for ttl compaction
+      } else {
+        options.max_open_files = 20;
+      }
+      // RocksDB sanitize max open files to at least 20. Modify it back.
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
+          "SanitizeOptions::AfterChangeMaxOpenFiles", [&](void* arg) {
+            int* max_open_files = static_cast<int*>(arg);
+            *max_open_files = 0;
+          });
+      // In the case where all files are opened and doing DB restart
+      // forcing the file creation time in manifest file to be 0 to
+      // simulate the case of reading from an old version.
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
+          "VersionEdit::EncodeTo:VarintFileCreationTime", [&](void* arg) {
+            if (if_restart && if_open_all_files) {
+              std::string* encoded_fieled = static_cast<std::string*>(arg);
+              *encoded_fieled = "";
+              PutVarint64(encoded_fieled, 0);
+            }
+          });
 
-  env_->addon_time_.store(0);
+      env_->time_elapse_only_sleep_ = false;
-  DestroyAndReopen(options);
+      options.env = env_;
 
-  int periodic_compactions = 0;
+      env_->addon_time_.store(0);
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      DestroyAndReopen(options);
-      "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
+
-        Compaction* compaction = reinterpret_cast<Compaction*>(arg);
+      int periodic_compactions = 0;
-        auto compaction_reason = compaction->compaction_reason();
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
-        if (compaction_reason == CompactionReason::kPeriodicCompaction) {
+          "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
-          periodic_compactions++;
+            Compaction* compaction = reinterpret_cast<Compaction*>(arg);
+            auto compaction_reason = compaction->compaction_reason();
+            if (compaction_reason == CompactionReason::kPeriodicCompaction) {
+              periodic_compactions++;
+            }
+          });
+      rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+      Random rnd(301);
+      for (int i = 0; i < kNumLevelFiles; ++i) {
+        for (int j = 0; j < kNumKeysPerFile; ++j) {
+          ASSERT_OK(Put(Key(i * kNumKeysPerFile + j),
+                        RandomString(&rnd, kValueSize)));
         }
-      });
+        Flush();
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+      }
+      dbfull()->TEST_WaitForCompact();
 
-  Random rnd(301);
+      ASSERT_EQ("2", FilesPerLevel());
-  for (int i = 0; i < kNumLevelFiles; ++i) {
+      ASSERT_EQ(0, periodic_compactions);
-    for (int j = 0; j < kNumKeysPerFile; ++j) {
+
-      ASSERT_OK(
+      // Add 50 hours and do a write
-          Put(Key(i * kNumKeysPerFile + j), RandomString(&rnd, kValueSize)));
+      env_->addon_time_.fetch_add(50 * 60 * 60);
+      ASSERT_OK(Put("a", "1"));
+      Flush();
+      dbfull()->TEST_WaitForCompact();
+      // Assert that the files stay in the same level
+      ASSERT_EQ("3", FilesPerLevel());
+      // The two old files go through the periodic compaction process
+      ASSERT_EQ(2, periodic_compactions);
+
+      MoveFilesToLevel(1);
+      ASSERT_EQ("0,3", FilesPerLevel());
+
+      // Add another 50 hours and do another write
+      env_->addon_time_.fetch_add(50 * 60 * 60);
+      ASSERT_OK(Put("b", "2"));
+      if (if_restart) {
+        Reopen(options);
+      } else {
+        Flush();
+      }
+      dbfull()->TEST_WaitForCompact();
+      ASSERT_EQ("1,3", FilesPerLevel());
+      // The three old files now go through the periodic compaction process. 2
+      // + 3.
+      ASSERT_EQ(5, periodic_compactions);
+
+      // Add another 50 hours and do another write
+      env_->addon_time_.fetch_add(50 * 60 * 60);
+      ASSERT_OK(Put("c", "3"));
+      Flush();
+      dbfull()->TEST_WaitForCompact();
+      ASSERT_EQ("2,3", FilesPerLevel());
+      // The four old files now go through the periodic compaction process. 5
+      // + 4.
+      ASSERT_EQ(9, periodic_compactions);
+
+      rocksdb::SyncPoint::GetInstance()->DisableProcessing();
     }
-    Flush();
   }
-  dbfull()->TEST_WaitForCompact();
-
-  ASSERT_EQ("2", FilesPerLevel());
-  ASSERT_EQ(0, periodic_compactions);
-
-  // Add 50 hours and do a write
-  env_->addon_time_.fetch_add(50 * 60 * 60);
-  ASSERT_OK(Put("a", "1"));
-  Flush();
-  dbfull()->TEST_WaitForCompact();
-  // Assert that the files stay in the same level
-  ASSERT_EQ("3", FilesPerLevel());
-  // The two old files go through the periodic compaction process
-  ASSERT_EQ(2, periodic_compactions);
-
-  MoveFilesToLevel(1);
-  ASSERT_EQ("0,3", FilesPerLevel());
-
-  // Add another 50 hours and do another write
-  env_->addon_time_.fetch_add(50 * 60 * 60);
-  ASSERT_OK(Put("b", "2"));
-  Flush();
-  dbfull()->TEST_WaitForCompact();
-  ASSERT_EQ("1,3", FilesPerLevel());
-  // The three old files now go through the periodic compaction process. 2 + 3.
-  ASSERT_EQ(5, periodic_compactions);
-
-  // Add another 50 hours and do another write
-  env_->addon_time_.fetch_add(50 * 60 * 60);
-  ASSERT_OK(Put("c", "3"));
-  Flush();
-  dbfull()->TEST_WaitForCompact();
-  ASSERT_EQ("2,3", FilesPerLevel());
-  // The four old files now go through the periodic compaction process. 5 + 4.
-  ASSERT_EQ(9, periodic_compactions);
-
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
 
 TEST_F(DBCompactionTest, LevelPeriodicCompactionWithOldDB) {
@@ -3734,7 +3782,6 @@ TEST_F(DBCompactionTest, LevelPeriodicCompactionWithOldDB) {
   const int kValueSize = 100;
 
   Options options = CurrentOptions();
-  options.max_open_files = -1;  // needed for ttl compaction
   env_->time_elapse_only_sleep_ = false;
   options.env = env_;
 

db/db_impl/db_impl.cc

@@ -879,7 +879,7 @@ Status DBImpl::TablesRangeTombstoneSummary(ColumnFamilyHandle* column_family,
           column_family);
   ColumnFamilyData* cfd = cfh->cfd();
 
-  SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+  SuperVersion* super_version = cfd->GetReferencedSuperVersion(this);
   Version* version = super_version->current;
 
   Status s =
@@ -895,7 +895,6 @@ void DBImpl::ScheduleBgLogWriterClose(JobContext* job_context) {
       AddToLogsToFreeQueue(l);
     }
     job_context->logs_to_free.clear();
-    SchedulePurge();
   }
 }
 
@@ -1322,19 +1321,34 @@ void DBImpl::BackgroundCallPurge() {
     delete log_writer;
     mutex_.Lock();
   }
-  for (const auto& file : purge_files_) {
+  while (!superversions_to_free_queue_.empty()) {
-    const PurgeFileInfo& purge_file = file.second;
+    assert(!superversions_to_free_queue_.empty());
+    SuperVersion* sv = superversions_to_free_queue_.front();
+    superversions_to_free_queue_.pop_front();
+    mutex_.Unlock();
+    delete sv;
+    mutex_.Lock();
+  }
+
+  // Can't use iterator to go over purge_files_ because inside the loop we're
+  // unlocking the mutex that protects purge_files_.
+  while (!purge_files_.empty()) {
+    auto it = purge_files_.begin();
+    // Need to make a copy of the PurgeFilesInfo before unlocking the mutex.
+    PurgeFileInfo purge_file = it->second;
+
     const std::string& fname = purge_file.fname;
     const std::string& dir_to_sync = purge_file.dir_to_sync;
     FileType type = purge_file.type;
     uint64_t number = purge_file.number;
     int job_id = purge_file.job_id;
 
+    purge_files_.erase(it);
+
     mutex_.Unlock();
     DeleteObsoleteFileImpl(job_id, fname, dir_to_sync, type, number);
     mutex_.Lock();
   }
-  purge_files_.clear();
 
   bg_purge_scheduled_--;
 
@@ -1374,10 +1388,14 @@ static void CleanupIteratorState(void* arg1, void* /*arg2*/) {
     state->db->FindObsoleteFiles(&job_context, false, true);
     if (state->background_purge) {
       state->db->ScheduleBgLogWriterClose(&job_context);
+      state->db->AddSuperVersionsToFreeQueue(state->super_version);
+      state->db->SchedulePurge();
     }
     state->mu->Unlock();
 
-    delete state->super_version;
+    if (!state->background_purge) {
+      delete state->super_version;
+    }
     if (job_context.HaveSomethingToDelete()) {
       if (state->background_purge) {
         // PurgeObsoleteFiles here does not delete files. Instead, it adds the
@@ -2452,7 +2470,7 @@ Iterator* DBImpl::NewIterator(const ReadOptions& read_options,
     result = nullptr;
 
 #else
-    SuperVersion* sv = cfd->GetReferencedSuperVersion(&mutex_);
+    SuperVersion* sv = cfd->GetReferencedSuperVersion(this);
     auto iter = new ForwardIterator(this, read_options, cfd, sv);
     result = NewDBIterator(
         env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
@@ -2478,7 +2496,7 @@ ArenaWrappedDBIter* DBImpl::NewIteratorImpl(const ReadOptions& read_options,
                                             ReadCallback* read_callback,
                                             bool allow_blob,
                                             bool allow_refresh) {
-  SuperVersion* sv = cfd->GetReferencedSuperVersion(&mutex_);
+  SuperVersion* sv = cfd->GetReferencedSuperVersion(this);
 
   // Try to generate a DB iterator tree in continuous memory area to be
   // cache friendly. Here is an example of result:
@@ -2557,7 +2575,7 @@ Status DBImpl::NewIterators(
 #else
     for (auto cfh : column_families) {
       auto cfd = reinterpret_cast<ColumnFamilyHandleImpl*>(cfh)->cfd();
-      SuperVersion* sv = cfd->GetReferencedSuperVersion(&mutex_);
+      SuperVersion* sv = cfd->GetReferencedSuperVersion(this);
       auto iter = new ForwardIterator(this, read_options, cfd, sv);
       iterators->push_back(NewDBIterator(
           env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
@@ -2884,7 +2902,7 @@ bool DBImpl::GetAggregatedIntProperty(const Slice& property,
 
 SuperVersion* DBImpl::GetAndRefSuperVersion(ColumnFamilyData* cfd) {
   // TODO(ljin): consider using GetReferencedSuperVersion() directly
-  return cfd->GetThreadLocalSuperVersion(&mutex_);
+  return cfd->GetThreadLocalSuperVersion(this);
 }
 
 // REQUIRED: this function should only be called on the write thread or if the
@@ -2902,11 +2920,19 @@ SuperVersion* DBImpl::GetAndRefSuperVersion(uint32_t column_family_id) {
 void DBImpl::CleanupSuperVersion(SuperVersion* sv) {
   // Release SuperVersion
   if (sv->Unref()) {
+    bool defer_purge =
+            immutable_db_options().avoid_unnecessary_blocking_io;
     {
       InstrumentedMutexLock l(&mutex_);
       sv->Cleanup();
+      if (defer_purge) {
+        AddSuperVersionsToFreeQueue(sv);
+        SchedulePurge();
+      }
+    }
+    if (!defer_purge) {
+      delete sv;
     }
-    delete sv;
     RecordTick(stats_, NUMBER_SUPERVERSION_CLEANUPS);
   }
   RecordTick(stats_, NUMBER_SUPERVERSION_RELEASES);
@@ -3912,7 +3938,7 @@ Status DBImpl::IngestExternalFiles(
     start_file_number += args[i - 1].external_files.size();
     auto* cfd =
         static_cast<ColumnFamilyHandleImpl*>(args[i].column_family)->cfd();
-    SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+    SuperVersion* super_version = cfd->GetReferencedSuperVersion(this);
     exec_results[i].second = ingestion_jobs[i].Prepare(
         args[i].external_files, start_file_number, super_version);
     exec_results[i].first = true;
@@ -3923,7 +3949,7 @@ Status DBImpl::IngestExternalFiles(
   {
     auto* cfd =
         static_cast<ColumnFamilyHandleImpl*>(args[0].column_family)->cfd();
-    SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+    SuperVersion* super_version = cfd->GetReferencedSuperVersion(this);
     exec_results[0].second = ingestion_jobs[0].Prepare(
         args[0].external_files, next_file_number, super_version);
     exec_results[0].first = true;
@@ -4192,7 +4218,7 @@ Status DBImpl::CreateColumnFamilyWithImport(
   dummy_sv_ctx.Clean();
 
   if (status.ok()) {
-    SuperVersion* sv = cfd->GetReferencedSuperVersion(&mutex_);
+    SuperVersion* sv = cfd->GetReferencedSuperVersion(this);
     status = import_job.Prepare(next_file_number, sv);
     CleanupSuperVersion(sv);
   }
@@ -4269,7 +4295,7 @@ Status DBImpl::VerifyChecksum(const ReadOptions& read_options) {
   }
   std::vector<SuperVersion*> sv_list;
   for (auto cfd : cfd_list) {
-    sv_list.push_back(cfd->GetReferencedSuperVersion(&mutex_));
+    sv_list.push_back(cfd->GetReferencedSuperVersion(this));
   }
   for (auto& sv : sv_list) {
     VersionStorageInfo* vstorage = sv->current->storage_info();
@@ -4294,14 +4320,23 @@ Status DBImpl::VerifyChecksum(const ReadOptions& read_options) {
       break;
     }
   }
+  bool defer_purge =
+          immutable_db_options().avoid_unnecessary_blocking_io;
   {
     InstrumentedMutexLock l(&mutex_);
     for (auto sv : sv_list) {
       if (sv && sv->Unref()) {
         sv->Cleanup();
-        delete sv;
+        if (defer_purge) {
+          AddSuperVersionsToFreeQueue(sv);
+        } else {
+          delete sv;
+        }
       }
     }
+    if (defer_purge) {
+      SchedulePurge();
+    }
     for (auto cfd : cfd_list) {
       cfd->Unref();
     }

db/db_impl/db_impl.h

@@ -798,6 +798,10 @@ class DBImpl : public DB {
     logs_to_free_queue_.push_back(log_writer);
   }
 
+  void AddSuperVersionsToFreeQueue(SuperVersion* sv) {
+    superversions_to_free_queue_.push_back(sv);
+  }
+
   void SetSnapshotChecker(SnapshotChecker* snapshot_checker);
 
   // Fill JobContext with snapshot information needed by flush and compaction.
@@ -1109,6 +1113,8 @@ class DBImpl : public DB {
       bool read_only = false, bool error_if_log_file_exist = false,
       bool error_if_data_exists_in_logs = false);
 
+  virtual bool OwnTablesAndLogs() const { return true; }
+
  private:
   friend class DB;
   friend class ErrorHandler;
@@ -1193,7 +1199,7 @@ class DBImpl : public DB {
   };
 
   // PurgeFileInfo is a structure to hold information of files to be deleted in
-  // purge_queue_
+  // purge_files_
   struct PurgeFileInfo {
     std::string fname;
     std::string dir_to_sync;
@@ -1888,6 +1894,7 @@ class DBImpl : public DB {
 
   // A queue to store log writers to close
   std::deque<log::Writer*> logs_to_free_queue_;
+  std::deque<SuperVersion*> superversions_to_free_queue_;
   int unscheduled_flushes_;
   int unscheduled_compactions_;
 

db/db_impl/db_impl_compaction_flush.cc

@@ -659,7 +659,7 @@ Status DBImpl::CompactRange(const CompactRangeOptions& options,
     // one/both sides of the interval are unbounded. But it requires more
     // changes to RangesOverlapWithMemtables.
     Range range(*begin, *end);
-    SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+    SuperVersion* super_version = cfd->GetReferencedSuperVersion(this);
     cfd->RangesOverlapWithMemtables({range}, super_version, &flush_needed);
     CleanupSuperVersion(super_version);
   }
@@ -1257,7 +1257,7 @@ Status DBImpl::ReFitLevel(ColumnFamilyData* cfd, int level, int target_level) {
                    f->fd.GetFileSize(), f->smallest, f->largest,
                    f->fd.smallest_seqno, f->fd.largest_seqno,
                    f->marked_for_compaction, f->oldest_blob_file_number,
-                   f->oldest_ancester_time);
+                   f->oldest_ancester_time, f->file_creation_time);
     }
     ROCKS_LOG_DEBUG(immutable_db_options_.info_log,
                     "[%s] Apply version edit:\n%s", cfd->GetName().c_str(),
@@ -2672,7 +2672,8 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
                            f->fd.GetPathId(), f->fd.GetFileSize(), f->smallest,
                            f->largest, f->fd.smallest_seqno,
                            f->fd.largest_seqno, f->marked_for_compaction,
-                           f->oldest_blob_file_number, f->oldest_ancester_time);
+                           f->oldest_blob_file_number, f->oldest_ancester_time,
+                           f->file_creation_time);
 
         ROCKS_LOG_BUFFER(
             log_buffer,

db/db_impl/db_impl_experimental.cc

@@ -129,7 +129,7 @@ Status DBImpl::PromoteL0(ColumnFamilyHandle* column_family, int target_level) {
                    f->fd.GetFileSize(), f->smallest, f->largest,
                    f->fd.smallest_seqno, f->fd.largest_seqno,
                    f->marked_for_compaction, f->oldest_blob_file_number,
-                   f->oldest_ancester_time);
+                   f->oldest_ancester_time, f->file_creation_time);
     }
 
     status = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),

db/db_impl/db_impl_files.cc

@@ -385,6 +385,7 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
     w->Close();
   }
 
+  bool own_files = OwnTablesAndLogs();
   std::unordered_set<uint64_t> files_to_del;
   for (const auto& candidate_file : candidate_files) {
     const std::string& to_delete = candidate_file.file_name;
@@ -484,6 +485,12 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
     }
 #endif  // !ROCKSDB_LITE
 
+    // If I do not own these files, e.g. secondary instance with max_open_files
+    // = -1, then no need to delete or schedule delete these files since they
+    // will be removed by their owner, e.g. the primary instance.
+    if (!own_files) {
+      continue;
+    }
     Status file_deletion_status;
     if (schedule_only) {
       InstrumentedMutexLock guard_lock(&mutex_);
@@ -495,7 +502,6 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
 
   {
     // After purging obsolete files, remove them from files_grabbed_for_purge_.
-    // Use a temporary vector to perform bulk deletion via swap.
     InstrumentedMutexLock guard_lock(&mutex_);
     autovector<uint64_t> to_be_removed;
     for (auto fn : files_grabbed_for_purge_) {

db/db_impl/db_impl_open.cc

@@ -1226,7 +1226,7 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
                   meta.fd.GetFileSize(), meta.smallest, meta.largest,
                   meta.fd.smallest_seqno, meta.fd.largest_seqno,
                   meta.marked_for_compaction, meta.oldest_blob_file_number,
-                  meta.oldest_ancester_time);
+                  meta.oldest_ancester_time, meta.file_creation_time);
   }
 
   InternalStats::CompactionStats stats(CompactionReason::kFlush, 1);

db/db_impl/db_impl_secondary.cc

@@ -11,6 +11,7 @@
 #include "db/merge_context.h"
 #include "logging/auto_roll_logger.h"
 #include "monitoring/perf_context_imp.h"
+#include "util/cast_util.h"
 
 namespace rocksdb {
 
@@ -405,7 +406,7 @@ ArenaWrappedDBIter* DBImplSecondary::NewIteratorImpl(
     const ReadOptions& read_options, ColumnFamilyData* cfd,
     SequenceNumber snapshot, ReadCallback* read_callback) {
   assert(nullptr != cfd);
-  SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+  SuperVersion* super_version = cfd->GetReferencedSuperVersion(this);
   auto db_iter = NewArenaWrappedDbIterator(
       env_, read_options, *cfd->ioptions(), super_version->mutable_cf_options,
       snapshot,
@@ -497,45 +498,61 @@ Status DBImplSecondary::TryCatchUpWithPrimary() {
   // read the manifest and apply new changes to the secondary instance
   std::unordered_set<ColumnFamilyData*> cfds_changed;
   JobContext job_context(0, true /*create_superversion*/);
-  InstrumentedMutexLock lock_guard(&mutex_);
+  {
-  s = static_cast<ReactiveVersionSet*>(versions_.get())
+    InstrumentedMutexLock lock_guard(&mutex_);
-          ->ReadAndApply(&mutex_, &manifest_reader_, &cfds_changed);
+    s = static_cast_with_check<ReactiveVersionSet>(versions_.get())
+            ->ReadAndApply(&mutex_, &manifest_reader_, &cfds_changed);
 
-  ROCKS_LOG_INFO(immutable_db_options_.info_log, "Last sequence is %" PRIu64,
+    ROCKS_LOG_INFO(immutable_db_options_.info_log, "Last sequence is %" PRIu64,
-                 static_cast<uint64_t>(versions_->LastSequence()));
+                   static_cast<uint64_t>(versions_->LastSequence()));
-  for (ColumnFamilyData* cfd : cfds_changed) {
+    for (ColumnFamilyData* cfd : cfds_changed) {
-    if (cfd->IsDropped()) {
+      if (cfd->IsDropped()) {
-      ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "[%s] is dropped\n",
+        ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "[%s] is dropped\n",
-                      cfd->GetName().c_str());
+                        cfd->GetName().c_str());
-      continue;
+        continue;
+      }
+      VersionStorageInfo::LevelSummaryStorage tmp;
+      ROCKS_LOG_DEBUG(immutable_db_options_.info_log,
+                      "[%s] Level summary: %s\n", cfd->GetName().c_str(),
+                      cfd->current()->storage_info()->LevelSummary(&tmp));
     }
-    VersionStorageInfo::LevelSummaryStorage tmp;
-    ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "[%s] Level summary: %s\n",
-                    cfd->GetName().c_str(),
-                    cfd->current()->storage_info()->LevelSummary(&tmp));
-  }
 
-  // list wal_dir to discover new WALs and apply new changes to the secondary
+    // list wal_dir to discover new WALs and apply new changes to the secondary
-  // instance
+    // instance
-  if (s.ok()) {
+    if (s.ok()) {
-    s = FindAndRecoverLogFiles(&cfds_changed, &job_context);
+      s = FindAndRecoverLogFiles(&cfds_changed, &job_context);
-  }
+    }
-  if (s.IsPathNotFound()) {
+    if (s.IsPathNotFound()) {
-    ROCKS_LOG_INFO(immutable_db_options_.info_log,
+      ROCKS_LOG_INFO(
-                   "Secondary tries to read WAL, but WAL file(s) have already "
+          immutable_db_options_.info_log,
-                   "been purged by primary.");
+          "Secondary tries to read WAL, but WAL file(s) have already "
-    s = Status::OK();
+          "been purged by primary.");
-  }
+      s = Status::OK();
-  if (s.ok()) {
+    }
-    for (auto cfd : cfds_changed) {
+    if (s.ok()) {
-      cfd->imm()->RemoveOldMemTables(cfd->GetLogNumber(),
+      for (auto cfd : cfds_changed) {
-                                     &job_context.memtables_to_free);
+        cfd->imm()->RemoveOldMemTables(cfd->GetLogNumber(),
-      auto& sv_context = job_context.superversion_contexts.back();
+                                       &job_context.memtables_to_free);
-      cfd->InstallSuperVersion(&sv_context, &mutex_);
+        auto& sv_context = job_context.superversion_contexts.back();
-      sv_context.NewSuperVersion();
+        cfd->InstallSuperVersion(&sv_context, &mutex_);
+        sv_context.NewSuperVersion();
+      }
     }
-    job_context.Clean();
   }
+  job_context.Clean();
+
+  // Cleanup unused, obsolete files.
+  JobContext purge_files_job_context(0);
+  {
+    InstrumentedMutexLock lock_guard(&mutex_);
+    // Currently, secondary instance does not own the database files, thus it
+    // is unnecessary for the secondary to force full scan.
+    FindObsoleteFiles(&purge_files_job_context, /*force=*/false);
+  }
+  if (purge_files_job_context.HaveSomethingToDelete()) {
+    PurgeObsoleteFiles(purge_files_job_context);
+  }
+  purge_files_job_context.Clean();
   return s;
 }
 

db/db_impl/db_impl_secondary.h

@@ -172,6 +172,24 @@ class DBImplSecondary : public DBImpl {
     return Status::NotSupported("Not supported operation in secondary mode.");
   }
 
+  using DBImpl::SetDBOptions;
+  Status SetDBOptions(const std::unordered_map<std::string, std::string>&
+                      /*options_map*/) override {
+    // Currently not supported because changing certain options may cause
+    // flush/compaction.
+    return Status::NotSupported("Not supported operation in secondary mode.");
+  }
+
+  using DBImpl::SetOptions;
+  Status SetOptions(
+      ColumnFamilyHandle* /*cfd*/,
+      const std::unordered_map<std::string, std::string>& /*options_map*/)
+      override {
+    // Currently not supported because changing certain options may cause
+    // flush/compaction and/or write to MANIFEST.
+    return Status::NotSupported("Not supported operation in secondary mode.");
+  }
+
   using DBImpl::SyncWAL;
   Status SyncWAL() override {
     return Status::NotSupported("Not supported operation in secondary mode.");
@@ -269,6 +287,14 @@ class DBImplSecondary : public DBImpl {
     return s;
   }
 
+  bool OwnTablesAndLogs() const override {
+    // Currently, the secondary instance does not own the database files. It
+    // simply opens the files of the primary instance and tracks their file
+    // descriptors until they become obsolete. In the future, the secondary may
+    // create links to database files. OwnTablesAndLogs will return true then.
+    return false;
+  }
+
  private:
   friend class DB;
 

db/db_impl/db_impl_write.cc

@@ -1516,12 +1516,14 @@ Status DBImpl::TrimMemtableHistory(WriteContext* context) {
   for (auto& cfd : cfds) {
     autovector<MemTable*> to_delete;
     cfd->imm()->TrimHistory(&to_delete, cfd->mem()->ApproximateMemoryUsage());
-    for (auto m : to_delete) {
+    if (!to_delete.empty()) {
-      delete m;
+      for (auto m : to_delete) {
+        delete m;
+      }
+      context->superversion_context.NewSuperVersion();
+      assert(context->superversion_context.new_superversion.get() != nullptr);
+      cfd->InstallSuperVersion(&context->superversion_context, &mutex_);
     }
-    context->superversion_context.NewSuperVersion();
-    assert(context->superversion_context.new_superversion.get() != nullptr);
-    cfd->InstallSuperVersion(&context->superversion_context, &mutex_);
 
     if (cfd->Unref()) {
       delete cfd;

db/db_impl/db_secondary_test.cc

@@ -195,6 +195,90 @@ TEST_F(DBSecondaryTest, OpenAsSecondary) {
   verify_db_func("new_foo_value", "new_bar_value");
 }
 
+namespace {
+class TraceFileEnv : public EnvWrapper {
+ public:
+  explicit TraceFileEnv(Env* target) : EnvWrapper(target) {}
+  Status NewRandomAccessFile(const std::string& f,
+                             std::unique_ptr<RandomAccessFile>* r,
+                             const EnvOptions& env_options) override {
+    class TracedRandomAccessFile : public RandomAccessFile {
+     public:
+      TracedRandomAccessFile(std::unique_ptr<RandomAccessFile>&& target,
+                             std::atomic<int>& counter)
+          : target_(std::move(target)), files_closed_(counter) {}
+      ~TracedRandomAccessFile() override {
+        files_closed_.fetch_add(1, std::memory_order_relaxed);
+      }
+      Status Read(uint64_t offset, size_t n, Slice* result,
+                  char* scratch) const override {
+        return target_->Read(offset, n, result, scratch);
+      }
+
+     private:
+      std::unique_ptr<RandomAccessFile> target_;
+      std::atomic<int>& files_closed_;
+    };
+    Status s = target()->NewRandomAccessFile(f, r, env_options);
+    if (s.ok()) {
+      r->reset(new TracedRandomAccessFile(std::move(*r), files_closed_));
+    }
+    return s;
+  }
+
+  int files_closed() const {
+    return files_closed_.load(std::memory_order_relaxed);
+  }
+
+ private:
+  std::atomic<int> files_closed_{0};
+};
+}  // namespace
+
+TEST_F(DBSecondaryTest, SecondaryCloseFiles) {
+  Options options;
+  options.env = env_;
+  options.max_open_files = 1;
+  options.disable_auto_compactions = true;
+  Reopen(options);
+  Options options1;
+  std::unique_ptr<Env> traced_env(new TraceFileEnv(env_));
+  options1.env = traced_env.get();
+  OpenSecondary(options1);
+
+  static const auto verify_db = [&]() {
+    std::unique_ptr<Iterator> iter1(dbfull()->NewIterator(ReadOptions()));
+    std::unique_ptr<Iterator> iter2(db_secondary_->NewIterator(ReadOptions()));
+    for (iter1->SeekToFirst(), iter2->SeekToFirst();
+         iter1->Valid() && iter2->Valid(); iter1->Next(), iter2->Next()) {
+      ASSERT_EQ(iter1->key(), iter2->key());
+      ASSERT_EQ(iter1->value(), iter2->value());
+    }
+    ASSERT_FALSE(iter1->Valid());
+    ASSERT_FALSE(iter2->Valid());
+  };
+
+  ASSERT_OK(Put("a", "value"));
+  ASSERT_OK(Put("c", "value"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  verify_db();
+
+  ASSERT_OK(Put("b", "value"));
+  ASSERT_OK(Put("d", "value"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  verify_db();
+
+  ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  ASSERT_EQ(2, static_cast<TraceFileEnv*>(traced_env.get())->files_closed());
+
+  Status s = db_secondary_->SetDBOptions({{"max_open_files", "-1"}});
+  ASSERT_TRUE(s.IsNotSupported());
+  CloseSecondary();
+}
+
 TEST_F(DBSecondaryTest, OpenAsSecondaryWALTailing) {
   Options options;
   options.env = env_;

db/db_range_del_test.cc

@@ -1427,6 +1427,47 @@ TEST_F(DBRangeDelTest, SnapshotPreventsDroppedKeys) {
   db_->ReleaseSnapshot(snapshot);
 }
 
+TEST_F(DBRangeDelTest, SnapshotPreventsDroppedKeysInImmMemTables) {
+  const int kFileBytes = 1 << 20;
+
+  Options options = CurrentOptions();
+  options.compression = kNoCompression;
+  options.disable_auto_compactions = true;
+  options.target_file_size_base = kFileBytes;
+  Reopen(options);
+
+  // block flush thread -> pin immtables in memory
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->LoadDependency({
+      {"SnapshotPreventsDroppedKeysInImmMemTables:AfterNewIterator",
+       "DBImpl::BGWorkFlush"},
+  });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put(Key(0), "a"));
+  std::unique_ptr<const Snapshot, std::function<void(const Snapshot*)>>
+      snapshot(db_->GetSnapshot(),
+               [this](const Snapshot* s) { db_->ReleaseSnapshot(s); });
+
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), Key(0),
+                             Key(10)));
+
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable());
+
+  ReadOptions read_opts;
+  read_opts.snapshot = snapshot.get();
+  std::unique_ptr<Iterator> iter(db_->NewIterator(read_opts));
+
+  TEST_SYNC_POINT("SnapshotPreventsDroppedKeysInImmMemTables:AfterNewIterator");
+
+  iter->SeekToFirst();
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(Key(0), iter->key());
+
+  iter->Next();
+  ASSERT_FALSE(iter->Valid());
+}
+
 TEST_F(DBRangeDelTest, RangeTombstoneWrittenToMinimalSsts) {
   // Adapted from
   // https://github.com/cockroachdb/cockroach/blob/de8b3ea603dd1592d9dc26443c2cc92c356fbc2f/pkg/storage/engine/rocksdb_test.go#L1267-L1398.

db/db_test.cc

@@ -1022,7 +1022,8 @@ TEST_F(DBTest, FailMoreDbPaths) {
 
 void CheckColumnFamilyMeta(
     const ColumnFamilyMetaData& cf_meta,
-    const std::vector<std::vector<FileMetaData>>& files_by_level) {
+    const std::vector<std::vector<FileMetaData>>& files_by_level,
+    uint64_t start_time, uint64_t end_time) {
   ASSERT_EQ(cf_meta.name, kDefaultColumnFamilyName);
   ASSERT_EQ(cf_meta.levels.size(), files_by_level.size());
 
@@ -1060,6 +1061,14 @@ void CheckColumnFamilyMeta(
                 file_meta_from_files.largest.user_key().ToString());
       ASSERT_EQ(file_meta_from_cf.oldest_blob_file_number,
                 file_meta_from_files.oldest_blob_file_number);
+      ASSERT_EQ(file_meta_from_cf.oldest_ancester_time,
+                file_meta_from_files.oldest_ancester_time);
+      ASSERT_EQ(file_meta_from_cf.file_creation_time,
+                file_meta_from_files.file_creation_time);
+      ASSERT_GE(file_meta_from_cf.file_creation_time, start_time);
+      ASSERT_LE(file_meta_from_cf.file_creation_time, end_time);
+      ASSERT_GE(file_meta_from_cf.oldest_ancester_time, start_time);
+      ASSERT_LE(file_meta_from_cf.oldest_ancester_time, end_time);
     }
 
     ASSERT_EQ(level_meta_from_cf.size, level_size);
@@ -1113,6 +1122,11 @@ TEST_F(DBTest, MetaDataTest) {
   Options options = CurrentOptions();
   options.create_if_missing = true;
   options.disable_auto_compactions = true;
+
+  int64_t temp_time = 0;
+  options.env->GetCurrentTime(&temp_time);
+  uint64_t start_time = static_cast<uint64_t>(temp_time);
+
   DestroyAndReopen(options);
 
   Random rnd(301);
@@ -1139,9 +1153,12 @@ TEST_F(DBTest, MetaDataTest) {
   std::vector<std::vector<FileMetaData>> files_by_level;
   dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(), &files_by_level);
 
+  options.env->GetCurrentTime(&temp_time);
+  uint64_t end_time = static_cast<uint64_t>(temp_time);
+
   ColumnFamilyMetaData cf_meta;
   db_->GetColumnFamilyMetaData(&cf_meta);
-  CheckColumnFamilyMeta(cf_meta, files_by_level);
+  CheckColumnFamilyMeta(cf_meta, files_by_level, start_time, end_time);
 
   std::vector<LiveFileMetaData> live_file_meta;
   db_->GetLiveFilesMetaData(&live_file_meta);
@@ -6420,6 +6437,12 @@ TEST_F(DBTest, CreationTimeOfOldestFile) {
           }
         }
       });
+  // Set file creation time in manifest all to 0.
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "FileMetaData::FileMetaData", [&](void* arg) {
+        FileMetaData* meta = static_cast<FileMetaData*>(arg);
+        meta->file_creation_time = 0;
+      });
   rocksdb::SyncPoint::GetInstance()->EnableProcessing();
 
   Random rnd(301);
@@ -6431,7 +6454,7 @@ TEST_F(DBTest, CreationTimeOfOldestFile) {
     Flush();
   }
 
-  // At this point there should be 2 files, oen with file_creation_time = 0 and
+  // At this point there should be 2 files, one with file_creation_time = 0 and
   // the other non-zero. GetCreationTimeOfOldestFile API should return 0.
   uint64_t creation_time;
   Status s1 = dbfull()->GetCreationTimeOfOldestFile(&creation_time);

db/db_test2.cc

@@ -4210,6 +4210,42 @@ TEST_F(DBTest2, SeekFileRangeDeleteTail) {
   }
   db_->ReleaseSnapshot(s1);
 }
+
+TEST_F(DBTest2, BackgroundPurgeTest) {
+  Options options = CurrentOptions();
+  options.write_buffer_manager = std::make_shared<rocksdb::WriteBufferManager>(1 << 20);
+  options.avoid_unnecessary_blocking_io = true;
+  DestroyAndReopen(options);
+  size_t base_value = options.write_buffer_manager->memory_usage();
+
+  ASSERT_OK(Put("a", "a"));
+  Iterator* iter = db_->NewIterator(ReadOptions());
+  ASSERT_OK(Flush());
+  size_t value = options.write_buffer_manager->memory_usage();
+  ASSERT_GT(value, base_value);
+
+  db_->GetEnv()->SetBackgroundThreads(1, Env::Priority::HIGH);
+  test::SleepingBackgroundTask sleeping_task_after;
+  db_->GetEnv()->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                          &sleeping_task_after, Env::Priority::HIGH);
+  delete iter;
+
+  Env::Default()->SleepForMicroseconds(100000);
+  value = options.write_buffer_manager->memory_usage();
+  ASSERT_GT(value, base_value);
+
+  sleeping_task_after.WakeUp();
+  sleeping_task_after.WaitUntilDone();
+
+  test::SleepingBackgroundTask sleeping_task_after2;
+  db_->GetEnv()->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                          &sleeping_task_after2, Env::Priority::HIGH);
+  sleeping_task_after2.WakeUp();
+  sleeping_task_after2.WaitUntilDone();
+
+  value = options.write_buffer_manager->memory_usage();
+  ASSERT_EQ(base_value, value);
+}
 }  // namespace rocksdb
 
 #ifdef ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS

db/db_test_util.cc

@@ -53,9 +53,9 @@ DBTestBase::DBTestBase(const std::string path)
 #ifndef ROCKSDB_LITE
   const char* test_env_uri = getenv("TEST_ENV_URI");
   if (test_env_uri) {
-    Status s = ObjectRegistry::NewInstance()->NewSharedObject(test_env_uri,
+    Env* test_env = nullptr;
-                                                              &env_guard_);
+    Status s = Env::LoadEnv(test_env_uri, &test_env, &env_guard_);
-    base_env = env_guard_.get();
+    base_env = test_env;
     EXPECT_OK(s);
     EXPECT_NE(Env::Default(), base_env);
   }

db/db_write_test.cc

@@ -39,6 +39,97 @@ TEST_P(DBWriteTest, SyncAndDisableWAL) {
   ASSERT_TRUE(dbfull()->Write(write_options, &batch).IsInvalidArgument());
 }
 
+TEST_P(DBWriteTest, WriteThreadHangOnWriteStall) {
+  Options options = GetOptions();
+  options.level0_stop_writes_trigger = options.level0_slowdown_writes_trigger = 4;
+  std::vector<port::Thread> threads;
+  std::atomic<int> thread_num(0);
+  port::Mutex mutex;
+  port::CondVar cv(&mutex);
+
+  Reopen(options);
+
+  std::function<void()> write_slowdown_func = [&]() {
+    int a = thread_num.fetch_add(1);
+    std::string key = "foo" + std::to_string(a);
+    WriteOptions wo;
+    wo.no_slowdown = false;
+    dbfull()->Put(wo, key, "bar");
+  };
+  std::function<void()> write_no_slowdown_func = [&]() {
+    int a = thread_num.fetch_add(1);
+    std::string key = "foo" + std::to_string(a);
+    WriteOptions wo;
+    wo.no_slowdown = true;
+    dbfull()->Put(wo, key, "bar");
+  };
+  std::function<void(void *)> unblock_main_thread_func = [&](void *) {
+    mutex.Lock();
+    cv.SignalAll();
+    mutex.Unlock();
+  };
+
+  // Create 3 L0 files and schedule 4th without waiting
+  Put("foo" + std::to_string(thread_num.fetch_add(1)), "bar");
+  Flush();
+  Put("foo" + std::to_string(thread_num.fetch_add(1)), "bar");
+  Flush();
+  Put("foo" + std::to_string(thread_num.fetch_add(1)), "bar");
+  Flush();
+  Put("foo" + std::to_string(thread_num.fetch_add(1)), "bar");
+
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+    "WriteThread::JoinBatchGroup:Start", unblock_main_thread_func);
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+    {{"DBWriteTest::WriteThreadHangOnWriteStall:1",
+      "DBImpl::BackgroundCallFlush:start"},
+     {"DBWriteTest::WriteThreadHangOnWriteStall:2",
+      "DBImpl::WriteImpl:BeforeLeaderEnters"},
+     // Make compaction start wait for the write stall to be detected and
+     // implemented by a write group leader
+     {"DBWriteTest::WriteThreadHangOnWriteStall:3",
+      "BackgroundCallCompaction:0"}});
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // Schedule creation of 4th L0 file without waiting. This will seal the
+  // memtable and then wait for a sync point before writing the file. We need
+  // to do it this way because SwitchMemtable() needs to enter the
+  // write_thread
+  FlushOptions fopt;
+  fopt.wait = false;
+  dbfull()->Flush(fopt);
+
+  // Create a mix of slowdown/no_slowdown write threads
+  mutex.Lock();
+  // First leader
+  threads.emplace_back(write_slowdown_func);
+  cv.Wait();
+  // Second leader. Will stall writes
+  threads.emplace_back(write_slowdown_func);
+  cv.Wait();
+  threads.emplace_back(write_no_slowdown_func);
+  cv.Wait();
+  threads.emplace_back(write_slowdown_func);
+  cv.Wait();
+  threads.emplace_back(write_no_slowdown_func);
+  cv.Wait();
+  threads.emplace_back(write_slowdown_func);
+  cv.Wait();
+  mutex.Unlock();
+
+  TEST_SYNC_POINT("DBWriteTest::WriteThreadHangOnWriteStall:1");
+  dbfull()->TEST_WaitForFlushMemTable(nullptr);
+  // This would have triggered a write stall. Unblock the write group leader
+  TEST_SYNC_POINT("DBWriteTest::WriteThreadHangOnWriteStall:2");
+  // The leader is going to create missing newer links. When the leader finishes,
+  // the next leader is going to delay writes and fail writers with no_slowdown
+
+  TEST_SYNC_POINT("DBWriteTest::WriteThreadHangOnWriteStall:3");
+  for (auto& t : threads) {
+    t.join();
+  }
+}
+
 TEST_P(DBWriteTest, IOErrorOnWALWritePropagateToWriteThreadFollower) {
   constexpr int kNumThreads = 5;
   std::unique_ptr<FaultInjectionTestEnv> mock_env(

db/error_handler.cc

@@ -166,12 +166,6 @@ Status ErrorHandler::SetBGError(const Status& bg_err, BackgroundErrorReason reas
     return Status::OK();
   }
 
-  // Check if recovery is currently in progress. If it is, we will save this
-  // error so we can check it at the end to see if recovery succeeded or not
-  if (recovery_in_prog_ && recovery_error_.ok()) {
-    recovery_error_ = bg_err;
-  }
-
   bool paranoid = db_options_.paranoid_checks;
   Status::Severity sev = Status::Severity::kFatalError;
   Status new_bg_err;
@@ -204,10 +198,15 @@ Status ErrorHandler::SetBGError(const Status& bg_err, BackgroundErrorReason reas
 
   new_bg_err = Status(bg_err, sev);
 
+  // Check if recovery is currently in progress. If it is, we will save this
+  // error so we can check it at the end to see if recovery succeeded or not
+  if (recovery_in_prog_ && recovery_error_.ok()) {
+    recovery_error_ = new_bg_err;
+  }
+
   bool auto_recovery = auto_recovery_;
   if (new_bg_err.severity() >= Status::Severity::kFatalError && auto_recovery) {
     auto_recovery = false;
-    ;
   }
 
   // Allow some error specific overrides

db/error_handler_test.cc

@@ -22,6 +22,21 @@ namespace rocksdb {
 class DBErrorHandlingTest : public DBTestBase {
  public:
   DBErrorHandlingTest() : DBTestBase("/db_error_handling_test") {}
+
+  std::string GetManifestNameFromLiveFiles() {
+    std::vector<std::string> live_files;
+    uint64_t manifest_size;
+
+    dbfull()->GetLiveFiles(live_files, &manifest_size, false);
+    for (auto& file : live_files) {
+      uint64_t num = 0;
+      FileType type;
+      if (ParseFileName(file, &num, &type) && type == kDescriptorFile) {
+        return file;
+      }
+    }
+    return "";
+  }
 };
 
 class DBErrorHandlingEnv : public EnvWrapper {
@@ -161,6 +176,169 @@ TEST_F(DBErrorHandlingTest, FLushWriteError) {
   Destroy(options);
 }
 
+TEST_F(DBErrorHandlingTest, ManifestWriteError) {
+  std::unique_ptr<FaultInjectionTestEnv> fault_env(
+      new FaultInjectionTestEnv(Env::Default()));
+  std::shared_ptr<ErrorHandlerListener> listener(new ErrorHandlerListener());
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.env = fault_env.get();
+  options.listeners.emplace_back(listener);
+  Status s;
+  std::string old_manifest;
+  std::string new_manifest;
+
+  listener->EnableAutoRecovery(false);
+  DestroyAndReopen(options);
+  old_manifest = GetManifestNameFromLiveFiles();
+
+  Put(Key(0), "val");
+  Flush();
+  Put(Key(1), "val");
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::LogAndApply:WriteManifest", [&](void *) {
+    fault_env->SetFilesystemActive(false, Status::NoSpace("Out of space"));
+  });
+  SyncPoint::GetInstance()->EnableProcessing();
+  s = Flush();
+  ASSERT_EQ(s.severity(), rocksdb::Status::Severity::kHardError);
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  SyncPoint::GetInstance()->DisableProcessing();
+  fault_env->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(DBErrorHandlingTest, DoubleManifestWriteError) {
+  std::unique_ptr<FaultInjectionTestEnv> fault_env(
+      new FaultInjectionTestEnv(Env::Default()));
+  std::shared_ptr<ErrorHandlerListener> listener(new ErrorHandlerListener());
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.env = fault_env.get();
+  options.listeners.emplace_back(listener);
+  Status s;
+  std::string old_manifest;
+  std::string new_manifest;
+
+  listener->EnableAutoRecovery(false);
+  DestroyAndReopen(options);
+  old_manifest = GetManifestNameFromLiveFiles();
+
+  Put(Key(0), "val");
+  Flush();
+  Put(Key(1), "val");
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::LogAndApply:WriteManifest", [&](void *) {
+    fault_env->SetFilesystemActive(false, Status::NoSpace("Out of space"));
+  });
+  SyncPoint::GetInstance()->EnableProcessing();
+  s = Flush();
+  ASSERT_EQ(s.severity(), rocksdb::Status::Severity::kHardError);
+  fault_env->SetFilesystemActive(true);
+
+  // This Resume() will attempt to create a new manifest file and fail again
+  s = dbfull()->Resume();
+  ASSERT_EQ(s.severity(), rocksdb::Status::Severity::kHardError);
+  fault_env->SetFilesystemActive(true);
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  SyncPoint::GetInstance()->DisableProcessing();
+
+  // A successful Resume() will create a new manifest file
+  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(DBErrorHandlingTest, CompactionManifestWriteError) {
+  std::unique_ptr<FaultInjectionTestEnv> fault_env(
+      new FaultInjectionTestEnv(Env::Default()));
+  std::shared_ptr<ErrorHandlerListener> listener(new ErrorHandlerListener());
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.level0_file_num_compaction_trigger = 2;
+  options.listeners.emplace_back(listener);
+  options.env = fault_env.get();
+  Status s;
+  std::string old_manifest;
+  std::string new_manifest;
+  std::atomic<bool> fail_manifest(false);
+  DestroyAndReopen(options);
+  old_manifest = GetManifestNameFromLiveFiles();
+
+  Put(Key(0), "val");
+  Put(Key(2), "val");
+  s = Flush();
+  ASSERT_EQ(s, Status::OK());
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      // Wait for flush of 2nd L0 file before starting compaction
+      {{"FlushMemTableFinished",
+        "BackgroundCallCompaction:0"},
+      // Wait for compaction to detect manifest write error
+       {"BackgroundCallCompaction:1",
+        "CompactionManifestWriteError:0"},
+      // Make compaction thread wait for error to be cleared
+       {"CompactionManifestWriteError:1",
+        "DBImpl::BackgroundCallCompaction:FoundObsoleteFiles"},
+      // Wait for DB instance to clear bg_error before calling
+      // TEST_WaitForCompact
+       {"SstFileManagerImpl::ClearError",
+        "CompactionManifestWriteError:2"}});
+  // trigger manifest write failure in compaction thread
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "BackgroundCallCompaction:0", [&](void *) {
+      fail_manifest.store(true);
+      });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::LogAndApply:WriteManifest", [&](void *) {
+      if (fail_manifest.load()) {
+        fault_env->SetFilesystemActive(false, Status::NoSpace("Out of space"));
+      }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  Put(Key(1), "val");
+  // This Flush will trigger a compaction, which will fail when appending to
+  // the manifest
+  s = Flush();
+  ASSERT_EQ(s, Status::OK());
+
+  TEST_SYNC_POINT("CompactionManifestWriteError:0");
+  // Clear all errors so when the compaction is retried, it will succeed
+  fault_env->SetFilesystemActive(true);
+  rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
+  TEST_SYNC_POINT("CompactionManifestWriteError:1");
+  TEST_SYNC_POINT("CompactionManifestWriteError:2");
+
+  s = dbfull()->TEST_WaitForCompact();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  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)));
+  ASSERT_EQ("val", Get(Key(2)));
+  Close();
+}
+
 TEST_F(DBErrorHandlingTest, CompactionWriteError) {
   std::unique_ptr<FaultInjectionTestEnv> fault_env(
       new FaultInjectionTestEnv(Env::Default()));

db/external_sst_file_ingestion_job.cc

@@ -246,16 +246,19 @@ Status ExternalSstFileIngestionJob::Run() {
 
     // We use the import time as the ancester time. This is the time the data
     // is written to the database.
-    uint64_t oldest_ancester_time = 0;
     int64_t temp_current_time = 0;
+    uint64_t current_time = kUnknownFileCreationTime;
+    uint64_t oldest_ancester_time = kUnknownOldestAncesterTime;
     if (env_->GetCurrentTime(&temp_current_time).ok()) {
-      oldest_ancester_time = static_cast<uint64_t>(temp_current_time);
+      current_time = oldest_ancester_time =
+          static_cast<uint64_t>(temp_current_time);
     }
 
     edit_.AddFile(f.picked_level, f.fd.GetNumber(), f.fd.GetPathId(),
                   f.fd.GetFileSize(), f.smallest_internal_key,
                   f.largest_internal_key, f.assigned_seqno, f.assigned_seqno,
-                  false, kInvalidBlobFileNumber, oldest_ancester_time);
+                  false, kInvalidBlobFileNumber, oldest_ancester_time,
+                  current_time);
   }
   return status;
 }

db/flush_job.cc

@@ -368,6 +368,7 @@ Status FlushJob::WriteLevel0Table() {
       // It's not clear whether oldest_key_time is always available. In case
       // it is not available, use current_time.
       meta_.oldest_ancester_time = std::min(current_time, oldest_key_time);
+      meta_.file_creation_time = current_time;
 
       s = BuildTable(
           dbname_, db_options_.env, *cfd_->ioptions(), mutable_cf_options_,
@@ -413,7 +414,7 @@ Status FlushJob::WriteLevel0Table() {
                    meta_.fd.GetFileSize(), meta_.smallest, meta_.largest,
                    meta_.fd.smallest_seqno, meta_.fd.largest_seqno,
                    meta_.marked_for_compaction, meta_.oldest_blob_file_number,
-                   meta_.oldest_ancester_time);
+                   meta_.oldest_ancester_time, meta_.file_creation_time);
   }
 #ifndef ROCKSDB_LITE
   // Piggyback FlushJobInfo on the first first flushed memtable.

db/forward_iterator.cc

@@ -239,9 +239,13 @@ void ForwardIterator::SVCleanup(DBImpl* db, SuperVersion* sv,
     db->FindObsoleteFiles(&job_context, false, true);
     if (background_purge_on_iterator_cleanup) {
       db->ScheduleBgLogWriterClose(&job_context);
+      db->AddSuperVersionsToFreeQueue(sv);
+      db->SchedulePurge();
     }
     db->mutex_.Unlock();
-    delete sv;
+    if (!background_purge_on_iterator_cleanup) {
+      delete sv;
+    }
     if (job_context.HaveSomethingToDelete()) {
       db->PurgeObsoleteFiles(job_context, background_purge_on_iterator_cleanup);
     }
@@ -614,7 +618,7 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
   Cleanup(refresh_sv);
   if (refresh_sv) {
     // New
-    sv_ = cfd_->GetReferencedSuperVersion(&(db_->mutex_));
+    sv_ = cfd_->GetReferencedSuperVersion(db_);
   }
   ReadRangeDelAggregator range_del_agg(&cfd_->internal_comparator(),
                                        kMaxSequenceNumber /* upper_bound */);
@@ -668,7 +672,7 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
 void ForwardIterator::RenewIterators() {
   SuperVersion* svnew;
   assert(sv_);
-  svnew = cfd_->GetReferencedSuperVersion(&(db_->mutex_));
+  svnew = cfd_->GetReferencedSuperVersion(db_);
 
   if (mutable_iter_ != nullptr) {
     DeleteIterator(mutable_iter_, true /* is_arena */);

db/import_column_family_job.cc

@@ -135,10 +135,12 @@ Status ImportColumnFamilyJob::Run() {
 
   // We use the import time as the ancester time. This is the time the data
   // is written to the database.
-  uint64_t oldest_ancester_time = 0;
   int64_t temp_current_time = 0;
+  uint64_t oldest_ancester_time = kUnknownOldestAncesterTime;
+  uint64_t current_time = kUnknownOldestAncesterTime;
   if (env_->GetCurrentTime(&temp_current_time).ok()) {
-    oldest_ancester_time = static_cast<uint64_t>(temp_current_time);
+    current_time = oldest_ancester_time =
+        static_cast<uint64_t>(temp_current_time);
   }
 
   for (size_t i = 0; i < files_to_import_.size(); ++i) {
@@ -149,7 +151,7 @@ Status ImportColumnFamilyJob::Run() {
                   f.fd.GetFileSize(), f.smallest_internal_key,
                   f.largest_internal_key, file_metadata.smallest_seqno,
                   file_metadata.largest_seqno, false, kInvalidBlobFileNumber,
-                  oldest_ancester_time);
+                  oldest_ancester_time, current_time);
 
     // If incoming sequence number is higher, update local sequence number.
     if (file_metadata.largest_seqno > versions_->LastSequence()) {

db/memtable.h

@@ -138,7 +138,7 @@ class MemTable {
 
   // As a cheap version of `ApproximateMemoryUsage()`, this function doens't
   // require external synchronization. The value may be less accurate though
-  size_t ApproximateMemoryUsageFast() {
+  size_t ApproximateMemoryUsageFast() const {
     return approximate_memory_usage_.load(std::memory_order_relaxed);
   }
 

db/memtable_list.cc

@@ -186,11 +186,14 @@ Status MemTableListVersion::AddRangeTombstoneIterators(
     const ReadOptions& read_opts, Arena* /*arena*/,
     RangeDelAggregator* range_del_agg) {
   assert(range_del_agg != nullptr);
+  // Except for snapshot read, using kMaxSequenceNumber is OK because these
+  // are immutable memtables.
+  SequenceNumber read_seq = read_opts.snapshot != nullptr
+                                ? read_opts.snapshot->GetSequenceNumber()
+                                : kMaxSequenceNumber;
   for (auto& m : memlist_) {
-    // Using kMaxSequenceNumber is OK because these are immutable memtables.
     std::unique_ptr<FragmentedRangeTombstoneIterator> range_del_iter(
-        m->NewRangeTombstoneIterator(read_opts,
+        m->NewRangeTombstoneIterator(read_opts, read_seq));
-                                     kMaxSequenceNumber /* read_seq */));
     range_del_agg->AddTombstones(std::move(range_del_iter));
   }
   return Status::OK();
@@ -277,7 +280,7 @@ void MemTableListVersion::Remove(MemTable* m,
 }
 
 // return the total memory usage assuming the oldest flushed memtable is dropped
-size_t MemTableListVersion::ApproximateMemoryUsageExcludingLast() {
+size_t MemTableListVersion::ApproximateMemoryUsageExcludingLast() const {
   size_t total_memtable_size = 0;
   for (auto& memtable : memlist_) {
     total_memtable_size += memtable->ApproximateMemoryUsage();
@@ -500,7 +503,7 @@ Status MemTableList::TryInstallMemtableFlushResults(
                            cfd->GetName().c_str(), m->file_number_, mem_id);
           assert(m->file_number_ > 0);
           current_->Remove(m, to_delete);
-          UpdateMemoryUsageExcludingLast();
+          UpdateCachedValuesFromMemTableListVersion();
           ResetTrimHistoryNeeded();
           ++mem_id;
         }
@@ -541,14 +544,14 @@ void MemTableList::Add(MemTable* m, autovector<MemTable*>* to_delete) {
   if (num_flush_not_started_ == 1) {
     imm_flush_needed.store(true, std::memory_order_release);
   }
-  UpdateMemoryUsageExcludingLast();
+  UpdateCachedValuesFromMemTableListVersion();
   ResetTrimHistoryNeeded();
 }
 
 void MemTableList::TrimHistory(autovector<MemTable*>* to_delete, size_t usage) {
   InstallNewVersion();
   current_->TrimHistory(to_delete, usage);
-  UpdateMemoryUsageExcludingLast();
+  UpdateCachedValuesFromMemTableListVersion();
   ResetTrimHistoryNeeded();
 }
 
@@ -563,18 +566,25 @@ size_t MemTableList::ApproximateUnflushedMemTablesMemoryUsage() {
 
 size_t MemTableList::ApproximateMemoryUsage() { return current_memory_usage_; }
 
-size_t MemTableList::ApproximateMemoryUsageExcludingLast() {
+size_t MemTableList::ApproximateMemoryUsageExcludingLast() const {
-  size_t usage =
+  const size_t usage =
       current_memory_usage_excluding_last_.load(std::memory_order_relaxed);
   return usage;
 }
 
-// Update current_memory_usage_excluding_last_, need to call whenever state
+bool MemTableList::HasHistory() const {
-// changes for MemtableListVersion (whenever InstallNewVersion() is called)
+  const bool has_history = current_has_history_.load(std::memory_order_relaxed);
-void MemTableList::UpdateMemoryUsageExcludingLast() {
+  return has_history;
-  size_t total_memtable_size = current_->ApproximateMemoryUsageExcludingLast();
+}
+
+void MemTableList::UpdateCachedValuesFromMemTableListVersion() {
+  const size_t total_memtable_size =
+      current_->ApproximateMemoryUsageExcludingLast();
   current_memory_usage_excluding_last_.store(total_memtable_size,
                                              std::memory_order_relaxed);
+
+  const bool has_history = current_->HasHistory();
+  current_has_history_.store(has_history, std::memory_order_relaxed);
 }
 
 uint64_t MemTableList::ApproximateOldestKeyTime() const {
@@ -704,7 +714,7 @@ Status InstallMemtableAtomicFlushResults(
                          cfds[i]->GetName().c_str(), m->GetFileNumber(),
                          mem_id);
         imm->current_->Remove(m, to_delete);
-        imm->UpdateMemoryUsageExcludingLast();
+        imm->UpdateCachedValuesFromMemTableListVersion();
         imm->ResetTrimHistoryNeeded();
       }
     }
@@ -754,7 +764,7 @@ void MemTableList::RemoveOldMemTables(uint64_t log_number,
     }
   }
 
-  UpdateMemoryUsageExcludingLast();
+  UpdateCachedValuesFromMemTableListVersion();
   ResetTrimHistoryNeeded();
 }
 

db/memtable_list.h

@@ -162,7 +162,11 @@ class MemTableListVersion {
   // excluding the last MemTable in memlist_history_. The reason for excluding
   // the last MemTable is to see if dropping the last MemTable will keep total
   // memory usage above or equal to max_write_buffer_size_to_maintain_
-  size_t ApproximateMemoryUsageExcludingLast();
+  size_t ApproximateMemoryUsageExcludingLast() const;
+
+  // Whether this version contains flushed memtables that are only kept around
+  // for transaction conflict checking.
+  bool HasHistory() const { return !memlist_history_.empty(); }
 
   bool MemtableLimitExceeded(size_t usage);
 
@@ -211,7 +215,8 @@ class MemTableList {
         commit_in_progress_(false),
         flush_requested_(false),
         current_memory_usage_(0),
-        current_memory_usage_excluding_last_(0) {
+        current_memory_usage_excluding_last_(0),
+        current_has_history_(false) {
     current_->Ref();
   }
 
@@ -266,11 +271,16 @@ class MemTableList {
   // Returns an estimate of the number of bytes of data in use.
   size_t ApproximateMemoryUsage();
 
-  // Returns the cached current_memory_usage_excluding_last_ value
+  // Returns the cached current_memory_usage_excluding_last_ value.
-  size_t ApproximateMemoryUsageExcludingLast();
+  size_t ApproximateMemoryUsageExcludingLast() const;
 
-  // Update current_memory_usage_excluding_last_ from MemtableListVersion
+  // Returns the cached current_has_history_ value.
-  void UpdateMemoryUsageExcludingLast();
+  bool HasHistory() const;
+
+  // Updates current_memory_usage_excluding_last_ and current_has_history_
+  // from MemTableListVersion. Must be called whenever InstallNewVersion is
+  // called.
+  void UpdateCachedValuesFromMemTableListVersion();
 
   // `usage` is the current size of the mutable Memtable. When
   // max_write_buffer_size_to_maintain is used, total size of mutable and
@@ -388,7 +398,11 @@ class MemTableList {
   // The current memory usage.
   size_t current_memory_usage_;
 
+  // Cached value of current_->ApproximateMemoryUsageExcludingLast().
   std::atomic<size_t> current_memory_usage_excluding_last_;
+
+  // Cached value of current_->HasHistory().
+  std::atomic<bool> current_has_history_;
 };
 
 // Installs memtable atomic flush results.

db/repair.cc

@@ -577,13 +577,13 @@ class Repairer {
 
       // TODO(opt): separate out into multiple levels
       for (const auto* table : cf_id_and_tables.second) {
-        edit.AddFile(0, table->meta.fd.GetNumber(), table->meta.fd.GetPathId(),
+        edit.AddFile(
-                     table->meta.fd.GetFileSize(), table->meta.smallest,
+            0, table->meta.fd.GetNumber(), table->meta.fd.GetPathId(),
-                     table->meta.largest, table->meta.fd.smallest_seqno,
+            table->meta.fd.GetFileSize(), table->meta.smallest,
-                     table->meta.fd.largest_seqno,
+            table->meta.largest, table->meta.fd.smallest_seqno,
-                     table->meta.marked_for_compaction,
+            table->meta.fd.largest_seqno, table->meta.marked_for_compaction,
-                     table->meta.oldest_blob_file_number,
+            table->meta.oldest_blob_file_number,
-                     table->meta.oldest_ancester_time);
+            table->meta.oldest_ancester_time, table->meta.file_creation_time);
       }
       assert(next_file_number_ > 0);
       vset_.MarkFileNumberUsed(next_file_number_ - 1);

db/version_builder_test.cc

@@ -63,7 +63,7 @@ class VersionBuilderTest : public testing::Test {
         file_number, path_id, file_size, GetInternalKey(smallest, smallest_seq),
         GetInternalKey(largest, largest_seq), smallest_seqno, largest_seqno,
         /* marked_for_compact */ false, kInvalidBlobFileNumber,
-        kUnknownOldestAncesterTime);
+        kUnknownOldestAncesterTime, kUnknownFileCreationTime);
     f->compensated_file_size = file_size;
     f->num_entries = num_entries;
     f->num_deletions = num_deletions;
@@ -114,7 +114,8 @@ TEST_F(VersionBuilderTest, ApplyAndSaveTo) {
   VersionEdit version_edit;
   version_edit.AddFile(2, 666, 0, 100U, GetInternalKey("301"),
                        GetInternalKey("350"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.DeleteFile(3, 27U);
 
   EnvOptions env_options;
@@ -149,7 +150,8 @@ TEST_F(VersionBuilderTest, ApplyAndSaveToDynamic) {
   VersionEdit version_edit;
   version_edit.AddFile(3, 666, 0, 100U, GetInternalKey("301"),
                        GetInternalKey("350"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.DeleteFile(0, 1U);
   version_edit.DeleteFile(0, 88U);
 
@@ -187,7 +189,8 @@ TEST_F(VersionBuilderTest, ApplyAndSaveToDynamic2) {
   VersionEdit version_edit;
   version_edit.AddFile(4, 666, 0, 100U, GetInternalKey("301"),
                        GetInternalKey("350"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.DeleteFile(0, 1U);
   version_edit.DeleteFile(0, 88U);
   version_edit.DeleteFile(4, 6U);
@@ -216,19 +219,24 @@ TEST_F(VersionBuilderTest, ApplyMultipleAndSaveTo) {
   VersionEdit version_edit;
   version_edit.AddFile(2, 666, 0, 100U, GetInternalKey("301"),
                        GetInternalKey("350"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.AddFile(2, 676, 0, 100U, GetInternalKey("401"),
                        GetInternalKey("450"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.AddFile(2, 636, 0, 100U, GetInternalKey("601"),
                        GetInternalKey("650"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.AddFile(2, 616, 0, 100U, GetInternalKey("501"),
                        GetInternalKey("550"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.AddFile(2, 606, 0, 100U, GetInternalKey("701"),
                        GetInternalKey("750"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
 
   EnvOptions env_options;
 
@@ -255,30 +263,37 @@ TEST_F(VersionBuilderTest, ApplyDeleteAndSaveTo) {
   VersionEdit version_edit;
   version_edit.AddFile(2, 666, 0, 100U, GetInternalKey("301"),
                        GetInternalKey("350"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.AddFile(2, 676, 0, 100U, GetInternalKey("401"),
                        GetInternalKey("450"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.AddFile(2, 636, 0, 100U, GetInternalKey("601"),
                        GetInternalKey("650"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.AddFile(2, 616, 0, 100U, GetInternalKey("501"),
                        GetInternalKey("550"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit.AddFile(2, 606, 0, 100U, GetInternalKey("701"),
                        GetInternalKey("750"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_builder.Apply(&version_edit);
 
   VersionEdit version_edit2;
   version_edit.AddFile(2, 808, 0, 100U, GetInternalKey("901"),
                        GetInternalKey("950"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_edit2.DeleteFile(2, 616);
   version_edit2.DeleteFile(2, 636);
   version_edit.AddFile(2, 806, 0, 100U, GetInternalKey("801"),
                        GetInternalKey("850"), 200, 200, false,
-                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+                       kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+                       kUnknownFileCreationTime);
   version_builder.Apply(&version_edit2);
 
   version_builder.SaveTo(&new_vstorage);

db/version_edit.cc

@@ -62,6 +62,7 @@ enum CustomTag : uint32_t {
   kMinLogNumberToKeepHack = 3,
   kOldestBlobFileNumber = 4,
   kOldestAncesterTime = 5,
+  kFileCreationTime = 6,
   kPathId = 65,
 };
 // If this bit for the custom tag is set, opening DB should fail if
@@ -217,6 +218,14 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
     TEST_SYNC_POINT_CALLBACK("VersionEdit::EncodeTo:VarintOldestAncesterTime",
                              &varint_oldest_ancester_time);
     PutLengthPrefixedSlice(dst, Slice(varint_oldest_ancester_time));
+
+    PutVarint32(dst, CustomTag::kFileCreationTime);
+    std::string varint_file_creation_time;
+    PutVarint64(&varint_file_creation_time, f.file_creation_time);
+    TEST_SYNC_POINT_CALLBACK("VersionEdit::EncodeTo:VarintFileCreationTime",
+                             &varint_file_creation_time);
+    PutLengthPrefixedSlice(dst, Slice(varint_file_creation_time));
+
     if (f.fd.GetPathId() != 0) {
       PutVarint32(dst, CustomTag::kPathId);
       char p = static_cast<char>(f.fd.GetPathId());
@@ -335,6 +344,11 @@ const char* VersionEdit::DecodeNewFile4From(Slice* input) {
             return "invalid oldest ancester time";
           }
           break;
+        case kFileCreationTime:
+          if (!GetVarint64(&field, &f.file_creation_time)) {
+            return "invalid file creation time";
+          }
+          break;
         case kNeedCompaction:
           if (field.size() != 1) {
             return "need_compaction field wrong size";
@@ -660,6 +674,8 @@ std::string VersionEdit::DebugString(bool hex_key) const {
     }
     r.append(" oldest_ancester_time:");
     AppendNumberTo(&r, f.oldest_ancester_time);
+    r.append(" file_creation_time:");
+    AppendNumberTo(&r, f.file_creation_time);
   }
   r.append("\n  ColumnFamily: ");
   AppendNumberTo(&r, column_family_);

db/version_edit.h

@@ -26,6 +26,7 @@ class VersionSet;
 constexpr uint64_t kFileNumberMask = 0x3FFFFFFFFFFFFFFF;
 constexpr uint64_t kInvalidBlobFileNumber = 0;
 constexpr uint64_t kUnknownOldestAncesterTime = 0;
+constexpr uint64_t kUnknownFileCreationTime = 0;
 
 extern uint64_t PackFileNumberAndPathId(uint64_t number, uint64_t path_id);
 
@@ -128,7 +129,10 @@ struct FileMetaData {
   // in turn be outputs for compact older SST files. We track the memtable
   // flush timestamp for the oldest SST file that eventaully contribute data
   // to this file. 0 means the information is not available.
-  uint64_t oldest_ancester_time = 0;
+  uint64_t oldest_ancester_time = kUnknownOldestAncesterTime;
+
+  // Unix time when the SST file is created.
+  uint64_t file_creation_time = kUnknownFileCreationTime;
 
   FileMetaData() = default;
 
@@ -136,13 +140,17 @@ struct FileMetaData {
                const InternalKey& smallest_key, const InternalKey& largest_key,
                const SequenceNumber& smallest_seq,
                const SequenceNumber& largest_seq, bool marked_for_compact,
-               uint64_t oldest_blob_file, uint64_t _oldest_ancester_time)
+               uint64_t oldest_blob_file, uint64_t _oldest_ancester_time,
+               uint64_t _file_creation_time)
       : fd(file, file_path_id, file_size, smallest_seq, largest_seq),
         smallest(smallest_key),
         largest(largest_key),
         marked_for_compaction(marked_for_compact),
         oldest_blob_file_number(oldest_blob_file),
-        oldest_ancester_time(_oldest_ancester_time) {}
+        oldest_ancester_time(_oldest_ancester_time),
+        file_creation_time(_file_creation_time) {
+    TEST_SYNC_POINT_CALLBACK("FileMetaData::FileMetaData", this);
+  }
 
   // REQUIRED: Keys must be given to the function in sorted order (it expects
   // the last key to be the largest).
@@ -168,13 +176,23 @@ struct FileMetaData {
   // if table reader is already pinned.
   // 0 means the information is not available.
   uint64_t TryGetOldestAncesterTime() {
-    if (oldest_ancester_time != 0) {
+    if (oldest_ancester_time != kUnknownOldestAncesterTime) {
       return oldest_ancester_time;
     } else if (fd.table_reader != nullptr &&
                fd.table_reader->GetTableProperties() != nullptr) {
       return fd.table_reader->GetTableProperties()->creation_time;
     }
-    return 0;
+    return kUnknownOldestAncesterTime;
+  }
+
+  uint64_t TryGetFileCreationTime() {
+    if (file_creation_time != kUnknownFileCreationTime) {
+      return file_creation_time;
+    } else if (fd.table_reader != nullptr &&
+               fd.table_reader->GetTableProperties() != nullptr) {
+      return fd.table_reader->GetTableProperties()->file_creation_time;
+    }
+    return kUnknownFileCreationTime;
   }
 };
 
@@ -277,14 +295,14 @@ class VersionEdit {
                uint64_t file_size, const InternalKey& smallest,
                const InternalKey& largest, const SequenceNumber& smallest_seqno,
                const SequenceNumber& largest_seqno, bool marked_for_compaction,
-               uint64_t oldest_blob_file_number,
+               uint64_t oldest_blob_file_number, uint64_t oldest_ancester_time,
-               uint64_t oldest_ancester_time) {
+               uint64_t file_creation_time) {
     assert(smallest_seqno <= largest_seqno);
     new_files_.emplace_back(
-        level,
+        level, FileMetaData(file, file_path_id, file_size, smallest, largest,
-        FileMetaData(file, file_path_id, file_size, smallest, largest,
+                            smallest_seqno, largest_seqno,
-                     smallest_seqno, largest_seqno, marked_for_compaction,
+                            marked_for_compaction, oldest_blob_file_number,
-                     oldest_blob_file_number, oldest_ancester_time));
+                            oldest_ancester_time, file_creation_time));
   }
 
   void AddFile(int level, const FileMetaData& f) {

db/version_edit_test.cc

@@ -37,7 +37,7 @@ TEST_F(VersionEditTest, EncodeDecode) {
                  InternalKey("foo", kBig + 500 + i, kTypeValue),
                  InternalKey("zoo", kBig + 600 + i, kTypeDeletion),
                  kBig + 500 + i, kBig + 600 + i, false, kInvalidBlobFileNumber,
-                 888);
+                 888, 678);
     edit.DeleteFile(4, kBig + 700 + i);
   }
 
@@ -55,17 +55,19 @@ TEST_F(VersionEditTest, EncodeDecodeNewFile4) {
   edit.AddFile(3, 300, 3, 100, InternalKey("foo", kBig + 500, kTypeValue),
                InternalKey("zoo", kBig + 600, kTypeDeletion), kBig + 500,
                kBig + 600, true, kInvalidBlobFileNumber,
-               kUnknownOldestAncesterTime);
+               kUnknownOldestAncesterTime, kUnknownFileCreationTime);
   edit.AddFile(4, 301, 3, 100, InternalKey("foo", kBig + 501, kTypeValue),
                InternalKey("zoo", kBig + 601, kTypeDeletion), kBig + 501,
                kBig + 601, false, kInvalidBlobFileNumber,
-               kUnknownOldestAncesterTime);
+               kUnknownOldestAncesterTime, kUnknownFileCreationTime);
   edit.AddFile(5, 302, 0, 100, InternalKey("foo", kBig + 502, kTypeValue),
                InternalKey("zoo", kBig + 602, kTypeDeletion), kBig + 502,
-               kBig + 602, true, kInvalidBlobFileNumber, 666);
+               kBig + 602, true, kInvalidBlobFileNumber, 666, 888);
   edit.AddFile(5, 303, 0, 100, InternalKey("foo", kBig + 503, kTypeBlobIndex),
                InternalKey("zoo", kBig + 603, kTypeBlobIndex), kBig + 503,
-               kBig + 603, true, 1001, kUnknownOldestAncesterTime);
+               kBig + 603, true, 1001, kUnknownOldestAncesterTime,
+               kUnknownFileCreationTime);
+  ;
 
   edit.DeleteFile(4, 700);
 
@@ -104,10 +106,10 @@ TEST_F(VersionEditTest, ForwardCompatibleNewFile4) {
   edit.AddFile(3, 300, 3, 100, InternalKey("foo", kBig + 500, kTypeValue),
                InternalKey("zoo", kBig + 600, kTypeDeletion), kBig + 500,
                kBig + 600, true, kInvalidBlobFileNumber,
-               kUnknownOldestAncesterTime);
+               kUnknownOldestAncesterTime, kUnknownFileCreationTime);
   edit.AddFile(4, 301, 3, 100, InternalKey("foo", kBig + 501, kTypeValue),
                InternalKey("zoo", kBig + 601, kTypeDeletion), kBig + 501,
-               kBig + 601, false, kInvalidBlobFileNumber, 686);
+               kBig + 601, false, kInvalidBlobFileNumber, 686, 868);
   edit.DeleteFile(4, 700);
 
   edit.SetComparatorName("foo");
@@ -154,7 +156,7 @@ TEST_F(VersionEditTest, NewFile4NotSupportedField) {
   edit.AddFile(3, 300, 3, 100, InternalKey("foo", kBig + 500, kTypeValue),
                InternalKey("zoo", kBig + 600, kTypeDeletion), kBig + 500,
                kBig + 600, true, kInvalidBlobFileNumber,
-               kUnknownOldestAncesterTime);
+               kUnknownOldestAncesterTime, kUnknownFileCreationTime);
 
   edit.SetComparatorName("foo");
   edit.SetLogNumber(kBig + 100);
@@ -182,7 +184,8 @@ TEST_F(VersionEditTest, NewFile4NotSupportedField) {
 TEST_F(VersionEditTest, EncodeEmptyFile) {
   VersionEdit edit;
   edit.AddFile(0, 0, 0, 0, InternalKey(), InternalKey(), 0, 0, false,
-               kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+               kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+               kUnknownFileCreationTime);
   std::string buffer;
   ASSERT_TRUE(!edit.EncodeTo(&buffer));
 }

db/version_set.cc

@@ -1459,7 +1459,8 @@ void Version::GetColumnFamilyMetaData(ColumnFamilyMetaData* cf_meta) {
           file->fd.largest_seqno, file->smallest.user_key().ToString(),
           file->largest.user_key().ToString(),
           file->stats.num_reads_sampled.load(std::memory_order_relaxed),
-          file->being_compacted, file->oldest_blob_file_number});
+          file->being_compacted, file->oldest_blob_file_number,
+          file->TryGetOldestAncesterTime(), file->TryGetFileCreationTime()});
       files.back().num_entries = file->num_entries;
       files.back().num_deletions = file->num_deletions;
       level_size += file->fd.GetFileSize();
@@ -1485,10 +1486,9 @@ void Version::GetCreationTimeOfOldestFile(uint64_t* creation_time) {
   for (int level = 0; level < storage_info_.num_non_empty_levels_; level++) {
     for (FileMetaData* meta : storage_info_.LevelFiles(level)) {
       assert(meta->fd.table_reader != nullptr);
-      uint64_t file_creation_time =
+      uint64_t file_creation_time = meta->TryGetFileCreationTime();
-          meta->fd.table_reader->GetTableProperties()->file_creation_time;
+      if (file_creation_time == kUnknownFileCreationTime) {
-      if (file_creation_time == 0) {
+        *creation_time = 0;
-        *creation_time = file_creation_time;
         return;
       }
       if (file_creation_time < oldest_time) {
@@ -2501,8 +2501,7 @@ void VersionStorageInfo::ComputeExpiredTtlFiles(
 void VersionStorageInfo::ComputeFilesMarkedForPeriodicCompaction(
     const ImmutableCFOptions& ioptions,
     const uint64_t periodic_compaction_seconds) {
-  assert(periodic_compaction_seconds > 0 &&
+  assert(periodic_compaction_seconds > 0);
-      periodic_compaction_seconds < port::kMaxUint64);
 
   files_marked_for_periodic_compaction_.clear();
 
@@ -2513,8 +2512,8 @@ void VersionStorageInfo::ComputeFilesMarkedForPeriodicCompaction(
   }
   const uint64_t current_time = static_cast<uint64_t>(temp_current_time);
 
-  // If periodic_compaction_seconds > current_time, no file possibly qualifies
+  // If periodic_compaction_seconds is larger than current time, periodic
-  // periodic compaction.
+  // compaction can't possibly be triggered.
   if (periodic_compaction_seconds > current_time) {
     return;
   }
@@ -2524,20 +2523,18 @@ void VersionStorageInfo::ComputeFilesMarkedForPeriodicCompaction(
 
   for (int level = 0; level < num_levels(); level++) {
     for (auto f : files_[level]) {
-      if (!f->being_compacted && f->fd.table_reader != nullptr &&
+      if (!f->being_compacted) {
-          f->fd.table_reader->GetTableProperties() != nullptr) {
         // Compute a file's modification time in the following order:
         // 1. Use file_creation_time table property if it is > 0.
         // 2. Use creation_time table property if it is > 0.
         // 3. Use file's mtime metadata if the above two table properties are 0.
         // Don't consider the file at all if the modification time cannot be
         // correctly determined based on the above conditions.
-        uint64_t file_modification_time =
+        uint64_t file_modification_time = f->TryGetFileCreationTime();
-            f->fd.table_reader->GetTableProperties()->file_creation_time;
+        if (file_modification_time == kUnknownFileCreationTime) {
-        if (file_modification_time == 0) {
           file_modification_time = f->TryGetOldestAncesterTime();
         }
-        if (file_modification_time == 0) {
+        if (file_modification_time == kUnknownOldestAncesterTime) {
           auto file_path = TableFileName(ioptions.cf_paths, f->fd.GetNumber(),
                                          f->fd.GetPathId());
           status = ioptions.env->GetFileModificationTime(
@@ -3945,12 +3942,15 @@ Status VersionSet::ProcessManifestWrites(
     for (auto v : versions) {
       delete v;
     }
+    // If manifest append failed for whatever reason, the file could be
+    // corrupted. So we need to force the next version update to start a
+    // new manifest file.
+    descriptor_log_.reset();
     if (new_descriptor_log) {
       ROCKS_LOG_INFO(db_options_->info_log,
                      "Deleting manifest %" PRIu64 " current manifest %" PRIu64
                      "\n",
                      manifest_file_number_, pending_manifest_file_number_);
-      descriptor_log_.reset();
       env_->DeleteFile(
           DescriptorFileName(dbname_, pending_manifest_file_number_));
     }
@@ -4980,7 +4980,7 @@ Status VersionSet::WriteCurrentStateToManifest(log::Writer* log) {
                        f->fd.GetFileSize(), f->smallest, f->largest,
                        f->fd.smallest_seqno, f->fd.largest_seqno,
                        f->marked_for_compaction, f->oldest_blob_file_number,
-                       f->oldest_ancester_time);
+                       f->oldest_ancester_time, f->file_creation_time);
         }
       }
       edit.SetLogNumber(cfd->GetLogNumber());

db/version_set_test.cc

@@ -40,7 +40,7 @@ class GenerateLevelFilesBriefTest : public testing::Test {
         InternalKey(smallest, smallest_seq, kTypeValue),
         InternalKey(largest, largest_seq, kTypeValue), smallest_seq,
         largest_seq, /* marked_for_compact */ false, kInvalidBlobFileNumber,
-        kUnknownOldestAncesterTime);
+        kUnknownOldestAncesterTime, kUnknownFileCreationTime);
     files_.push_back(f);
   }
 
@@ -135,7 +135,7 @@ class VersionStorageInfoTest : public testing::Test {
         file_number, 0, file_size, GetInternalKey(smallest, 0),
         GetInternalKey(largest, 0), /* smallest_seq */ 0, /* largest_seq */ 0,
         /* marked_for_compact */ false, kInvalidBlobFileNumber,
-        kUnknownOldestAncesterTime);
+        kUnknownOldestAncesterTime, kUnknownFileCreationTime);
     f->compensated_file_size = file_size;
     vstorage_.AddFile(level, f);
   }
@@ -146,7 +146,8 @@ class VersionStorageInfoTest : public testing::Test {
     FileMetaData* f = new FileMetaData(
         file_number, 0, file_size, smallest, largest, /* smallest_seq */ 0,
         /* largest_seq */ 0, /* marked_for_compact */ false,
-        kInvalidBlobFileNumber, kUnknownOldestAncesterTime);
+        kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
+        kUnknownFileCreationTime);
     f->compensated_file_size = file_size;
     vstorage_.AddFile(level, f);
   }

db/write_batch.cc

@@ -1784,14 +1784,28 @@ class MemTableInserter : public WriteBatch::Handler {
     // check if memtable_list size exceeds max_write_buffer_size_to_maintain
     if (trim_history_scheduler_ != nullptr) {
       auto* cfd = cf_mems_->current();
-      assert(cfd != nullptr);
+
-      if (cfd->ioptions()->max_write_buffer_size_to_maintain > 0 &&
+      assert(cfd);
-          cfd->mem()->ApproximateMemoryUsageFast() +
+      assert(cfd->ioptions());
-                  cfd->imm()->ApproximateMemoryUsageExcludingLast() >=
+
-              static_cast<size_t>(
+      const size_t size_to_maintain = static_cast<size_t>(
-                  cfd->ioptions()->max_write_buffer_size_to_maintain) &&
+          cfd->ioptions()->max_write_buffer_size_to_maintain);
-          cfd->imm()->MarkTrimHistoryNeeded()) {
+
-        trim_history_scheduler_->ScheduleWork(cfd);
+      if (size_to_maintain > 0) {
+        MemTableList* const imm = cfd->imm();
+        assert(imm);
+
+        if (imm->HasHistory()) {
+          const MemTable* const mem = cfd->mem();
+          assert(mem);
+
+          if (mem->ApproximateMemoryUsageFast() +
+                      imm->ApproximateMemoryUsageExcludingLast() >=
+                  size_to_maintain &&
+              imm->MarkTrimHistoryNeeded()) {
+            trim_history_scheduler_->ScheduleWork(cfd);
+          }
+        }
       }
     }
   }

db/write_thread.cc

@@ -344,6 +344,9 @@ void WriteThread::BeginWriteStall() {
       prev->link_older = w->link_older;
       w->status = Status::Incomplete("Write stall");
       SetState(w, STATE_COMPLETED);
+      if (prev->link_older) {
+        prev->link_older->link_newer = prev;
+      }
       w = prev->link_older;
     } else {
       prev = w;
@@ -355,7 +358,11 @@ void WriteThread::BeginWriteStall() {
 void WriteThread::EndWriteStall() {
   MutexLock lock(&stall_mu_);
 
+  // Unlink write_stall_dummy_ from the write queue. This will unblock
+  // pending write threads to enqueue themselves
   assert(newest_writer_.load(std::memory_order_relaxed) == &write_stall_dummy_);
+  assert(write_stall_dummy_.link_older != nullptr);
+  write_stall_dummy_.link_older->link_newer = write_stall_dummy_.link_newer;
   newest_writer_.exchange(write_stall_dummy_.link_older);
 
   // Wake up writers

file/delete_scheduler.cc

@@ -73,7 +73,8 @@ Status DeleteScheduler::DeleteFile(const std::string& file_path,
   s = MarkAsTrash(file_path, &trash_file);
 
   if (!s.ok()) {
-    ROCKS_LOG_ERROR(info_log_, "Failed to mark %s as trash", file_path.c_str());
+    ROCKS_LOG_ERROR(info_log_, "Failed to mark %s as trash -- %s",
+                    file_path.c_str(), s.ToString().c_str());
     s = env_->DeleteFile(file_path);
     if (s.ok()) {
       sst_file_manager_->OnDeleteFile(file_path);

file/sst_file_manager_impl.cc

@@ -306,6 +306,7 @@ void SstFileManagerImpl::ClearError() {
       // since the ErrorHandler::recovery_in_prog_ flag would be true
       cur_instance_ = error_handler;
       mu_.Unlock();
+      TEST_SYNC_POINT("SstFileManagerImpl::ClearError");
       s = error_handler->RecoverFromBGError();
       mu_.Lock();
       // The DB instance might have been deleted while we were

include/rocksdb/cache.h

@@ -255,7 +255,7 @@ class Cache {
   // Always delete the DB object before calling this method!
   virtual void DisownData(){
       // default implementation is noop
-  };
+  }
 
   // Apply callback to all entries in the cache
   // If thread_safe is true, it will also lock the accesses. Otherwise, it will

include/rocksdb/env.h

@@ -672,7 +672,7 @@ class RandomAccessFile {
   virtual size_t GetUniqueId(char* /*id*/, size_t /*max_size*/) const {
     return 0;  // Default implementation to prevent issues with backwards
                // compatibility.
-  };
+  }
 
   enum AccessPattern { NORMAL, RANDOM, SEQUENTIAL, WILLNEED, DONTNEED };
 
@@ -1414,7 +1414,7 @@ class RandomAccessFileWrapper : public RandomAccessFile {
   }
   size_t GetUniqueId(char* id, size_t max_size) const override {
     return target_->GetUniqueId(id, max_size);
-  };
+  }
   void Hint(AccessPattern pattern) override { target_->Hint(pattern); }
   bool use_direct_io() const override { return target_->use_direct_io(); }
   size_t GetRequiredBufferAlignment() const override {

include/rocksdb/metadata.h

@@ -69,7 +69,8 @@ struct SstFileMetaData {
                   SequenceNumber _smallest_seqno, SequenceNumber _largest_seqno,
                   const std::string& _smallestkey,
                   const std::string& _largestkey, uint64_t _num_reads_sampled,
-                  bool _being_compacted, uint64_t _oldest_blob_file_number)
+                  bool _being_compacted, uint64_t _oldest_blob_file_number,
+                  uint64_t _oldest_ancester_time, uint64_t _file_creation_time)
       : size(_size),
         name(_file_name),
         file_number(_file_number),
@@ -82,7 +83,9 @@ struct SstFileMetaData {
         being_compacted(_being_compacted),
         num_entries(0),
         num_deletions(0),
-        oldest_blob_file_number(_oldest_blob_file_number) {}
+        oldest_blob_file_number(_oldest_blob_file_number),
+        oldest_ancester_time(_oldest_ancester_time),
+        file_creation_time(_file_creation_time) {}
 
   // File size in bytes.
   size_t size;
@@ -105,6 +108,15 @@ struct SstFileMetaData {
 
   uint64_t oldest_blob_file_number;  // The id of the oldest blob file
                                      // referenced by the file.
+  // An SST file may be generated by compactions whose input files may
+  // in turn be generated by earlier compactions. The creation time of the
+  // oldest SST file that is the compaction ancester of this file.
+  // The timestamp is provided Env::GetCurrentTime().
+  // 0 if the information is not available.
+  uint64_t oldest_ancester_time;
+  // Timestamp when the SST file is created, provided by Env::GetCurrentTime().
+  // 0 if the information is not available.
+  uint64_t file_creation_time;
 };
 
 // The full set of metadata associated with each SST file.

include/rocksdb/options.h

@@ -1079,10 +1079,10 @@ struct DBOptions {
   // independently if the process crashes later and tries to recover.
   bool atomic_flush = false;
 
-  // If true, ColumnFamilyHandle's and Iterator's destructors won't delete
+  // If true, working thread may avoid doing unnecessary and long-latency
-  // obsolete files directly and will instead schedule a background job
+  // operation (such as deleting obsolete files directly or deleting memtable)
-  // to do it. Use it if you're destroying iterators or ColumnFamilyHandle-s
+  // and will instead schedule a background job to do it.
-  // from latency-sensitive threads.
+  // Use it if you're latency-sensitive.
   // If set to true, takes precedence over
   // ReadOptions::background_purge_on_iterator_cleanup.
   bool avoid_unnecessary_blocking_io = false;

include/rocksdb/slice.h

@@ -195,8 +195,15 @@ class PinnableSlice : public Slice, public Cleanable {
     }
   }
 
-  void remove_prefix(size_t /*n*/) {
+  void remove_prefix(size_t n) {
-    assert(0);  // Not implemented
+    assert(n <= size());
+    if (pinned_) {
+      data_ += n;
+      size_ -= n;
+    } else {
+      buf_->erase(0, n);
+      PinSelf();
+    }
   }
 
   void Reset() {

include/rocksdb/statistics.h

@@ -523,7 +523,7 @@ class Statistics {
   virtual bool getTickerMap(std::map<std::string, uint64_t>*) const {
     // Do nothing by default
     return false;
-  };
+  }
 
   // Override this function to disable particular histogram collection
   virtual bool HistEnabledForType(uint32_t type) const {

include/rocksdb/table.h

@@ -269,9 +269,9 @@ struct BlockBasedTableOptions {
   // probably use this as it would reduce the index size.
   // This option only affects newly written tables. When reading existing
   // tables, the information about version is read from the footer.
-  // 5 -- Can be read by RocksDB's versions since X.X.X (something after 6.4.6)
+  // 5 -- Can be read by RocksDB's versions since 6.6.0. Full and partitioned
-  // Full and partitioned filters use a generally faster and more accurate
+  // filters use a generally faster and more accurate Bloom filter
-  // Bloom filter implementation, with a different schema.
+  // implementation, with a different schema.
   uint32_t format_version = 2;
 
   // Store index blocks on disk in compressed format. Changing this option to

include/rocksdb/utilities/backupable_db.h

@@ -104,13 +104,13 @@ struct BackupableDBOptions {
   // Default: 4194304
   uint64_t callback_trigger_interval_size;
 
-  // When Open() is called, it will open at most this many of the latest
+  // For BackupEngineReadOnly, Open() will open at most this many of the
-  // non-corrupted backups.
+  // latest non-corrupted backups.
   //
-  // Note setting this to a non-default value prevents old files from being
+  // Note: this setting is ignored (behaves like INT_MAX) for any kind of
-  // deleted in the shared directory, as we can't do proper ref-counting. If
+  // writable BackupEngine because it would inhibit accounting for shared
-  // using this option, make sure to occasionally disable it (by resetting to
+  // files for proper backup deletion, including purging any incompletely
-  // INT_MAX) and run GarbageCollect to clean accumulated stale files.
+  // created backups on creation of a new backup.
   //
   // Default: INT_MAX
   int max_valid_backups_to_open;

include/rocksdb/version.h

@@ -6,7 +6,7 @@
 
 #define ROCKSDB_MAJOR 6
 #define ROCKSDB_MINOR 6
-#define ROCKSDB_PATCH 0
+#define ROCKSDB_PATCH 4
 
 // Do not use these. We made the mistake of declaring macros starting with
 // double underscore. Now we have to live with our choice. We'll deprecate these

java/CMakeLists.txt

@@ -312,17 +312,19 @@ if(NOT EXISTS ${JAVA_TEST_LIBDIR})
   file(MAKE_DIRECTORY mkdir ${JAVA_TEST_LIBDIR})
 endif()
 
-if (DEFINED CUSTOM_REPO_URL)
+if (DEFINED CUSTOM_DEPS_URL)
-  set(SEARCH_REPO_URL ${CUSTOM_REPO_URL}/)
+  set(DEPS_URL ${CUSTOM_DEPS_URL}/)
-  set(CENTRAL_REPO_URL ${CUSTOM_REPO_URL}/)
 else ()
-  set(SEARCH_REPO_URL "http://search.maven.org/remotecontent?filepath=")
+  # This is a URL for artifacts from a "fake" release on pdillinger's fork,
-  set(CENTRAL_REPO_URL "http://central.maven.org/maven2/")
+  # so as not to put binaries in git (ew). We should move to hosting these
+  # under the facebook account on github, or something else more reliable
+  # than maven.org, which has been failing frequently from Travis.
+  set(DEPS_URL "https://github.com/pdillinger/rocksdb/releases/download/v6.6.x-java-deps")
 endif()
 
 if(NOT EXISTS ${JAVA_JUNIT_JAR})
   message("Downloading ${JAVA_JUNIT_JAR}")
-  file(DOWNLOAD ${SEARCH_REPO_URL}junit/junit/4.12/junit-4.12.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
+  file(DOWNLOAD ${DEPS_URL}/junit-4.12.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
   list(GET downloadStatus 0 error_code)
   if(NOT error_code EQUAL 0)
     message(FATAL_ERROR "Failed downloading ${JAVA_JUNIT_JAR}")
@@ -331,7 +333,7 @@ if(NOT EXISTS ${JAVA_JUNIT_JAR})
 endif()
 if(NOT EXISTS ${JAVA_HAMCR_JAR})
   message("Downloading ${JAVA_HAMCR_JAR}")
-  file(DOWNLOAD ${SEARCH_REPO_URL}org/hamcrest/hamcrest-core/1.3/hamcrest-core-1.3.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
+  file(DOWNLOAD ${DEPS_URL}/hamcrest-core-1.3.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
   list(GET downloadStatus 0 error_code)
   if(NOT error_code EQUAL 0)
     message(FATAL_ERROR "Failed downloading ${JAVA_HAMCR_JAR}")
@@ -340,7 +342,7 @@ if(NOT EXISTS ${JAVA_HAMCR_JAR})
 endif()
 if(NOT EXISTS ${JAVA_MOCKITO_JAR})
   message("Downloading ${JAVA_MOCKITO_JAR}")
-  file(DOWNLOAD ${SEARCH_REPO_URL}org/mockito/mockito-all/1.10.19/mockito-all-1.10.19.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
+  file(DOWNLOAD ${DEPS_URL}/mockito-all-1.10.19.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
   list(GET downloadStatus 0 error_code)
   if(NOT error_code EQUAL 0)
     message(FATAL_ERROR "Failed downloading ${JAVA_MOCKITO_JAR}")
@@ -349,7 +351,7 @@ if(NOT EXISTS ${JAVA_MOCKITO_JAR})
 endif()
 if(NOT EXISTS ${JAVA_CGLIB_JAR})
   message("Downloading ${JAVA_CGLIB_JAR}")
-  file(DOWNLOAD ${SEARCH_REPO_URL}cglib/cglib/2.2.2/cglib-2.2.2.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
+  file(DOWNLOAD ${DEPS_URL}/cglib-2.2.2.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
   list(GET downloadStatus 0 error_code)
   if(NOT error_code EQUAL 0)
     message(FATAL_ERROR "Failed downloading ${JAVA_CGLIB_JAR}")
@@ -358,7 +360,7 @@ if(NOT EXISTS ${JAVA_CGLIB_JAR})
 endif()
 if(NOT EXISTS ${JAVA_ASSERTJ_JAR})
   message("Downloading ${JAVA_ASSERTJ_JAR}")
-  file(DOWNLOAD ${CENTRAL_REPO_URL}org/assertj/assertj-core/1.7.1/assertj-core-1.7.1.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
+  file(DOWNLOAD ${DEPS_URL}/assertj-core-1.7.1.jar ${JAVA_TMP_JAR} STATUS downloadStatus)
   list(GET downloadStatus 0 error_code)
   if(NOT error_code EQUAL 0)
     message(FATAL_ERROR "Failed downloading ${JAVA_ASSERTJ_JAR}")

java/Makefile

@@ -213,16 +213,23 @@ ifneq ($(DEBUG_LEVEL),0)
 	JAVAC_ARGS = -Xlint:deprecation -Xlint:unchecked
 endif
 
-SEARCH_REPO_URL?=http://search.maven.org/remotecontent?filepath=
+# This is a URL for artifacts from a "fake" release on pdillinger's fork,
-CENTRAL_REPO_URL?=http://central.maven.org/maven2/
+# so as not to put binaries in git (ew). We should move to hosting these
+# under the facebook account on github, or something else more reliable
+# than maven.org, which has been failing frequently from Travis.
+DEPS_URL?=https://github.com/pdillinger/rocksdb/releases/download/v6.6.x-java-deps
 
-clean:
+clean: clean-not-downloaded clean-downloaded
-	$(AM_V_at)rm -rf include/*
+
-	$(AM_V_at)rm -rf test-libs/
+clean-not-downloaded:
+	$(AM_V_at)rm -rf $(NATIVE_INCLUDE)
 	$(AM_V_at)rm -rf $(OUTPUT)
 	$(AM_V_at)rm -rf $(BENCHMARK_OUTPUT)
 	$(AM_V_at)rm -rf $(SAMPLES_OUTPUT)
 
+clean-downloaded:
+	$(AM_V_at)rm -rf $(JAVA_TEST_LIBDIR)
+
 
 javadocs: java
 	$(AM_V_GEN)mkdir -p $(JAVADOC)
@@ -279,11 +286,11 @@ optimistic_transaction_sample: java
 
 resolve_test_deps:
 	test -d "$(JAVA_TEST_LIBDIR)" || mkdir -p "$(JAVA_TEST_LIBDIR)"
-	test -s "$(JAVA_JUNIT_JAR)" || cp $(MVN_LOCAL)/junit/junit/4.12/junit-4.12.jar $(JAVA_TEST_LIBDIR) || curl -k -L -o $(JAVA_JUNIT_JAR) $(SEARCH_REPO_URL)junit/junit/4.12/junit-4.12.jar
+	test -s "$(JAVA_JUNIT_JAR)" || cp $(MVN_LOCAL)/junit/junit/4.12/junit-4.12.jar $(JAVA_TEST_LIBDIR) || curl --fail --insecure --output $(JAVA_JUNIT_JAR) --location $(DEPS_URL)/junit-4.12.jar
-	test -s "$(JAVA_HAMCR_JAR)" || cp $(MVN_LOCAL)/org/hamcrest/hamcrest-core/1.3/hamcrest-core-1.3.jar $(JAVA_TEST_LIBDIR) || curl -k -L -o $(JAVA_HAMCR_JAR) $(SEARCH_REPO_URL)org/hamcrest/hamcrest-core/1.3/hamcrest-core-1.3.jar
+	test -s "$(JAVA_HAMCR_JAR)" || cp $(MVN_LOCAL)/org/hamcrest/hamcrest-core/1.3/hamcrest-core-1.3.jar $(JAVA_TEST_LIBDIR) || curl --fail --insecure --output $(JAVA_HAMCR_JAR) --location $(DEPS_URL)/hamcrest-core-1.3.jar
-	test -s "$(JAVA_MOCKITO_JAR)" || cp $(MVN_LOCAL)/org/mockito/mockito-all/1.10.19/mockito-all-1.10.19.jar $(JAVA_TEST_LIBDIR) || curl -k -L -o "$(JAVA_MOCKITO_JAR)" $(SEARCH_REPO_URL)org/mockito/mockito-all/1.10.19/mockito-all-1.10.19.jar
+	test -s "$(JAVA_MOCKITO_JAR)" || cp $(MVN_LOCAL)/org/mockito/mockito-all/1.10.19/mockito-all-1.10.19.jar $(JAVA_TEST_LIBDIR) || curl --fail --insecure --output "$(JAVA_MOCKITO_JAR)" --location $(DEPS_URL)/mockito-all-1.10.19.jar
-	test -s "$(JAVA_CGLIB_JAR)" || cp $(MVN_LOCAL)/cglib/cglib/2.2.2/cglib-2.2.2.jar $(JAVA_TEST_LIBDIR) || curl -k -L -o "$(JAVA_CGLIB_JAR)" $(SEARCH_REPO_URL)cglib/cglib/2.2.2/cglib-2.2.2.jar
+	test -s "$(JAVA_CGLIB_JAR)" || cp $(MVN_LOCAL)/cglib/cglib/2.2.2/cglib-2.2.2.jar $(JAVA_TEST_LIBDIR) || curl --fail --insecure --output "$(JAVA_CGLIB_JAR)" --location $(DEPS_URL)/cglib-2.2.2.jar
-	test -s "$(JAVA_ASSERTJ_JAR)" || cp $(MVN_LOCAL)/org/assertj/assertj-core/1.7.1/assertj-core-1.7.1.jar $(JAVA_TEST_LIBDIR) || curl -k -L -o "$(JAVA_ASSERTJ_JAR)" $(CENTRAL_REPO_URL)org/assertj/assertj-core/1.7.1/assertj-core-1.7.1.jar
+	test -s "$(JAVA_ASSERTJ_JAR)" || cp $(MVN_LOCAL)/org/assertj/assertj-core/1.7.1/assertj-core-1.7.1.jar $(JAVA_TEST_LIBDIR) || curl --fail --insecure --output "$(JAVA_ASSERTJ_JAR)" --location $(DEPS_URL)/assertj-core-1.7.1.jar
 
 java_test: java resolve_test_deps
 	$(AM_V_GEN)mkdir -p $(TEST_CLASSES)

java/rocksjni/options.cc

@@ -1247,9 +1247,51 @@ jint Java_org_rocksdb_Options_statsDumpPeriodSec(
  */
 void Java_org_rocksdb_Options_setStatsDumpPeriodSec(
     JNIEnv*, jobject, jlong jhandle,
-    jint stats_dump_period_sec) {
+    jint jstats_dump_period_sec) {
   reinterpret_cast<rocksdb::Options*>(jhandle)->stats_dump_period_sec =
-      static_cast<int>(stats_dump_period_sec);
+      static_cast<unsigned int>(jstats_dump_period_sec);
+}
+
+/*
+ * Class:     org_rocksdb_Options
+ * Method:    statsPersistPeriodSec
+ * Signature: (J)I
+ */
+jint Java_org_rocksdb_Options_statsPersistPeriodSec(
+    JNIEnv*, jobject, jlong jhandle) {
+  return reinterpret_cast<rocksdb::Options*>(jhandle)->stats_persist_period_sec;
+}
+
+/*
+ * Class:     org_rocksdb_Options
+ * Method:    setStatsPersistPeriodSec
+ * Signature: (JI)V
+ */
+void Java_org_rocksdb_Options_setStatsPersistPeriodSec(
+    JNIEnv*, jobject, jlong jhandle, jint jstats_persist_period_sec) {
+  reinterpret_cast<rocksdb::Options*>(jhandle)->stats_persist_period_sec =
+      static_cast<unsigned int>(jstats_persist_period_sec);
+}
+
+/*
+ * Class:     org_rocksdb_Options
+ * Method:    statsHistoryBufferSize
+ * Signature: (J)J
+ */
+jlong Java_org_rocksdb_Options_statsHistoryBufferSize(
+    JNIEnv*, jobject, jlong jhandle) {
+  return reinterpret_cast<rocksdb::Options*>(jhandle)->stats_history_buffer_size;
+}
+
+/*
+ * Class:     org_rocksdb_Options
+ * Method:    setStatsHistoryBufferSize
+ * Signature: (JJ)V
+ */
+void Java_org_rocksdb_Options_setStatsHistoryBufferSize(
+    JNIEnv*, jobject, jlong jhandle, jlong jstats_history_buffer_size) {
+  reinterpret_cast<rocksdb::Options*>(jhandle)->stats_history_buffer_size =
+      static_cast<size_t>(jstats_history_buffer_size);
 }
 
 /*
@@ -1481,6 +1523,28 @@ jlong Java_org_rocksdb_Options_walBytesPerSync(
   return static_cast<jlong>(opt->wal_bytes_per_sync);
 }
 
+/*
+ * Class:     org_rocksdb_Options
+ * Method:    setStrictBytesPerSync
+ * Signature: (JZ)V
+ */
+void Java_org_rocksdb_Options_setStrictBytesPerSync(
+    JNIEnv*, jobject, jlong jhandle, jboolean jstrict_bytes_per_sync) {
+  reinterpret_cast<rocksdb::Options*>(jhandle)->strict_bytes_per_sync =
+      jstrict_bytes_per_sync == JNI_TRUE;
+}
+
+/*
+ * Class:     org_rocksdb_Options
+ * Method:    strictBytesPerSync
+ * Signature: (J)Z
+ */
+jboolean Java_org_rocksdb_Options_strictBytesPerSync(
+    JNIEnv*, jobject, jlong jhandle) {
+  auto* opt = reinterpret_cast<rocksdb::Options*>(jhandle);
+  return static_cast<jboolean>(opt->strict_bytes_per_sync);
+}
+
 /*
  * Class:     org_rocksdb_Options
  * Method:    setEnableThreadTracking
@@ -5441,9 +5505,9 @@ jboolean Java_org_rocksdb_DBOptions_isFdCloseOnExec(
  * Signature: (JI)V
  */
 void Java_org_rocksdb_DBOptions_setStatsDumpPeriodSec(
-    JNIEnv*, jobject, jlong jhandle, jint stats_dump_period_sec) {
+    JNIEnv*, jobject, jlong jhandle, jint jstats_dump_period_sec) {
   reinterpret_cast<rocksdb::DBOptions*>(jhandle)->stats_dump_period_sec =
-      static_cast<int>(stats_dump_period_sec);
+      static_cast<unsigned int>(jstats_dump_period_sec);
 }
 
 /*
@@ -5456,6 +5520,48 @@ jint Java_org_rocksdb_DBOptions_statsDumpPeriodSec(
   return reinterpret_cast<rocksdb::DBOptions*>(jhandle)->stats_dump_period_sec;
 }
 
+/*
+ * Class:     org_rocksdb_DBOptions
+ * Method:    setStatsPersistPeriodSec
+ * Signature: (JI)V
+ */
+void Java_org_rocksdb_DBOptions_setStatsPersistPeriodSec(
+    JNIEnv*, jobject, jlong jhandle, jint jstats_persist_period_sec) {
+  reinterpret_cast<rocksdb::DBOptions*>(jhandle)->stats_persist_period_sec =
+      static_cast<unsigned int>(jstats_persist_period_sec);
+}
+
+/*
+ * Class:     org_rocksdb_DBOptions
+ * Method:    statsPersistPeriodSec
+ * Signature: (J)I
+ */
+jint Java_org_rocksdb_DBOptions_statsPersistPeriodSec(
+    JNIEnv*, jobject, jlong jhandle) {
+  return reinterpret_cast<rocksdb::DBOptions*>(jhandle)->stats_persist_period_sec;
+}
+
+/*
+ * Class:     org_rocksdb_DBOptions
+ * Method:    setStatsHistoryBufferSize
+ * Signature: (JJ)V
+ */
+void Java_org_rocksdb_DBOptions_setStatsHistoryBufferSize(
+    JNIEnv*, jobject, jlong jhandle, jlong jstats_history_buffer_size) {
+  reinterpret_cast<rocksdb::DBOptions*>(jhandle)->stats_history_buffer_size =
+      static_cast<size_t>(jstats_history_buffer_size);
+}
+
+/*
+ * Class:     org_rocksdb_DBOptions
+ * Method:    statsHistoryBufferSize
+ * Signature: (J)J
+ */
+jlong Java_org_rocksdb_DBOptions_statsHistoryBufferSize(
+    JNIEnv*, jobject, jlong jhandle) {
+  return reinterpret_cast<rocksdb::DBOptions*>(jhandle)->stats_history_buffer_size;
+}
+
 /*
  * Class:     org_rocksdb_DBOptions
  * Method:    setAdviseRandomOnOpen
@@ -5694,6 +5800,28 @@ jlong Java_org_rocksdb_DBOptions_walBytesPerSync(
   return static_cast<jlong>(opt->wal_bytes_per_sync);
 }
 
+/*
+ * Class:     org_rocksdb_DBOptions
+ * Method:    setStrictBytesPerSync
+ * Signature: (JZ)V
+ */
+void Java_org_rocksdb_DBOptions_setStrictBytesPerSync(
+    JNIEnv*, jobject, jlong jhandle, jboolean jstrict_bytes_per_sync) {
+  reinterpret_cast<rocksdb::DBOptions*>(jhandle)->strict_bytes_per_sync =
+      jstrict_bytes_per_sync == JNI_TRUE;
+}
+
+/*
+ * Class:     org_rocksdb_DBOptions
+ * Method:    strictBytesPerSync
+ * Signature: (J)Z
+ */
+jboolean Java_org_rocksdb_DBOptions_strictBytesPerSync(
+    JNIEnv*, jobject, jlong jhandle) {
+  return static_cast<jboolean>(
+      reinterpret_cast<rocksdb::DBOptions*>(jhandle)->strict_bytes_per_sync);
+}
+
 /*
  * Class:     org_rocksdb_DBOptions
  * Method:    setDelayedWriteRate

java/src/main/java/org/rocksdb/DBOptions.java

@@ -660,6 +660,34 @@ public class DBOptions extends RocksObject
     return statsDumpPeriodSec(nativeHandle_);
   }
 
+  @Override
+  public DBOptions setStatsPersistPeriodSec(
+      final int statsPersistPeriodSec) {
+    assert(isOwningHandle());
+    setStatsPersistPeriodSec(nativeHandle_, statsPersistPeriodSec);
+    return this;
+  }
+
+  @Override
+  public int statsPersistPeriodSec() {
+    assert(isOwningHandle());
+    return statsPersistPeriodSec(nativeHandle_);
+  }
+
+  @Override
+  public DBOptions setStatsHistoryBufferSize(
+      final long statsHistoryBufferSize) {
+    assert(isOwningHandle());
+    setStatsHistoryBufferSize(nativeHandle_, statsHistoryBufferSize);
+    return this;
+  }
+
+  @Override
+  public long statsHistoryBufferSize() {
+    assert(isOwningHandle());
+    return statsHistoryBufferSize(nativeHandle_);
+  }
+
   @Override
   public DBOptions setAdviseRandomOnOpen(
       final boolean adviseRandomOnOpen) {
@@ -807,6 +835,19 @@ public class DBOptions extends RocksObject
     return walBytesPerSync(nativeHandle_);
   }
 
+  @Override
+  public DBOptions setStrictBytesPerSync(final boolean strictBytesPerSync) {
+    assert(isOwningHandle());
+    setStrictBytesPerSync(nativeHandle_, strictBytesPerSync);
+    return this;
+  }
+
+  @Override
+  public boolean strictBytesPerSync() {
+    assert(isOwningHandle());
+    return strictBytesPerSync(nativeHandle_);
+  }
+
   //TODO(AR) NOW
 //  @Override
 //  public DBOptions setListeners(final List<EventListener> listeners) {
@@ -1239,6 +1280,14 @@ public class DBOptions extends RocksObject
   private native void setStatsDumpPeriodSec(
       long handle, int statsDumpPeriodSec);
   private native int statsDumpPeriodSec(long handle);
+  private native void setStatsPersistPeriodSec(
+      final long handle, final int statsPersistPeriodSec);
+  private native int statsPersistPeriodSec(
+      final long handle);
+  private native void setStatsHistoryBufferSize(
+      final long handle, final long statsHistoryBufferSize);
+  private native long statsHistoryBufferSize(
+      final long handle);
   private native void setAdviseRandomOnOpen(
       long handle, boolean adviseRandomOnOpen);
   private native boolean adviseRandomOnOpen(long handle);
@@ -1270,6 +1319,10 @@ public class DBOptions extends RocksObject
   private native long bytesPerSync(long handle);
   private native void setWalBytesPerSync(long handle, long walBytesPerSync);
   private native long walBytesPerSync(long handle);
+  private native void setStrictBytesPerSync(
+      final long handle, final boolean strictBytesPerSync);
+  private native boolean strictBytesPerSync(
+      final long handle);
   private native void setEnableThreadTracking(long handle,
       boolean enableThreadTracking);
   private native boolean enableThreadTracking(long handle);

java/src/main/java/org/rocksdb/Env.java

@@ -13,6 +13,10 @@ import java.util.List;
  */
 public abstract class Env extends RocksObject {
 
+  static {
+    RocksDB.loadLibrary();
+  }
+
   private static final Env DEFAULT_ENV = new RocksEnv(getDefaultEnvInternal());
   static {
     /**

java/src/main/java/org/rocksdb/MutableDBOptions.java

@@ -89,9 +89,12 @@ public class MutableDBOptions extends AbstractMutableOptions {
     max_total_wal_size(ValueType.LONG),
     delete_obsolete_files_period_micros(ValueType.LONG),
     stats_dump_period_sec(ValueType.INT),
+    stats_persist_period_sec(ValueType.INT),
+    stats_history_buffer_size(ValueType.LONG),
     max_open_files(ValueType.INT),
     bytes_per_sync(ValueType.LONG),
     wal_bytes_per_sync(ValueType.LONG),
+    strict_bytes_per_sync(ValueType.BOOLEAN),
     compaction_readahead_size(ValueType.LONG);
 
     private final ValueType valueType;
@@ -240,6 +243,28 @@ public class MutableDBOptions extends AbstractMutableOptions {
       return getInt(DBOption.stats_dump_period_sec);
     }
 
+    @Override
+    public MutableDBOptionsBuilder setStatsPersistPeriodSec(
+        final int statsPersistPeriodSec) {
+      return setInt(DBOption.stats_persist_period_sec, statsPersistPeriodSec);
+    }
+
+    @Override
+    public int statsPersistPeriodSec() {
+      return getInt(DBOption.stats_persist_period_sec);
+    }
+
+    @Override
+    public MutableDBOptionsBuilder setStatsHistoryBufferSize(
+        final long statsHistoryBufferSize) {
+      return setLong(DBOption.stats_history_buffer_size, statsHistoryBufferSize);
+    }
+
+    @Override
+    public long statsHistoryBufferSize() {
+      return getLong(DBOption.stats_history_buffer_size);
+    }
+
     @Override
     public MutableDBOptionsBuilder setMaxOpenFiles(final int maxOpenFiles) {
       return setInt(DBOption.max_open_files, maxOpenFiles);
@@ -271,6 +296,17 @@ public class MutableDBOptions extends AbstractMutableOptions {
       return getLong(DBOption.wal_bytes_per_sync);
     }
 
+    @Override
+    public MutableDBOptionsBuilder setStrictBytesPerSync(
+        final boolean strictBytesPerSync) {
+      return setBoolean(DBOption.strict_bytes_per_sync, strictBytesPerSync);
+    }
+
+    @Override
+    public boolean strictBytesPerSync() {
+      return getBoolean(DBOption.strict_bytes_per_sync);
+    }
+
     @Override
     public MutableDBOptionsBuilder setCompactionReadaheadSize(
         final long compactionReadaheadSize) {

java/src/main/java/org/rocksdb/MutableDBOptionsInterface.java

@@ -237,6 +237,44 @@ public interface MutableDBOptionsInterface<T extends MutableDBOptionsInterface<T
    */
   int statsDumpPeriodSec();
 
+  /**
+   * If not zero, dump rocksdb.stats to RocksDB every
+   * {@code statsPersistPeriodSec}
+   *
+   * Default: 600
+   *
+   * @param statsPersistPeriodSec time interval in seconds.
+   * @return the instance of the current object.
+   */
+  T setStatsPersistPeriodSec(int statsPersistPeriodSec);
+
+  /**
+   * If not zero, dump rocksdb.stats to RocksDB every
+   * {@code statsPersistPeriodSec}
+   *
+   * @return time interval in seconds.
+   */
+  int statsPersistPeriodSec();
+
+  /**
+   * If not zero, periodically take stats snapshots and store in memory, the
+   * memory size for stats snapshots is capped at {@code statsHistoryBufferSize}
+   *
+   * Default: 1MB
+   *
+   * @param statsHistoryBufferSize the size of the buffer.
+   * @return the instance of the current object.
+   */
+  T setStatsHistoryBufferSize(long statsHistoryBufferSize);
+
+  /**
+   * If not zero, periodically take stats snapshots and store in memory, the
+   * memory size for stats snapshots is capped at {@code statsHistoryBufferSize}
+   *
+   * @return the size of the buffer.
+   */
+  long statsHistoryBufferSize();
+
   /**
    * Number of open files that can be used by the DB.  You may need to
    * increase this if your database has a large working set. Value -1 means
@@ -303,6 +341,42 @@ public interface MutableDBOptionsInterface<T extends MutableDBOptionsInterface<T
    */
   long walBytesPerSync();
 
+  /**
+   * When true, guarantees WAL files have at most {@link #walBytesPerSync()}
+   * bytes submitted for writeback at any given time, and SST files have at most
+   * {@link #bytesPerSync()} bytes pending writeback at any given time. This
+   * can be used to handle cases where processing speed exceeds I/O speed
+   * during file generation, which can lead to a huge sync when the file is
+   * finished, even with {@link #bytesPerSync()} / {@link #walBytesPerSync()}
+   * properly configured.
+   *
+   * - If `sync_file_range` is supported it achieves this by waiting for any
+   *   prior `sync_file_range`s to finish before proceeding. In this way,
+   *   processing (compression, etc.) can proceed uninhibited in the gap
+   *   between `sync_file_range`s, and we block only when I/O falls
+   *   behind.
+   * - Otherwise the `WritableFile::Sync` method is used. Note this mechanism
+   *   always blocks, thus preventing the interleaving of I/O and processing.
+   *
+   * Note: Enabling this option does not provide any additional persistence
+   * guarantees, as it may use `sync_file_range`, which does not write out
+   * metadata.
+   *
+   * Default: false
+   *
+   * @param strictBytesPerSync the bytes per sync
+   * @return the instance of the current object.
+   */
+  T setStrictBytesPerSync(boolean strictBytesPerSync);
+
+  /**
+   * Return the strict byte limit per sync.
+   *
+   * See {@link #setStrictBytesPerSync(boolean)}
+   *
+   * @return the limit in bytes.
+   */
+  boolean strictBytesPerSync();
 
   /**
    * If non-zero, we perform bigger reads when doing compaction. If you're

java/src/main/java/org/rocksdb/Options.java

@@ -739,6 +739,34 @@ public class Options extends RocksObject
     return this;
   }
 
+  @Override
+  public Options setStatsPersistPeriodSec(
+      final int statsPersistPeriodSec) {
+    assert(isOwningHandle());
+    setStatsPersistPeriodSec(nativeHandle_, statsPersistPeriodSec);
+    return this;
+  }
+
+  @Override
+  public int statsPersistPeriodSec() {
+    assert(isOwningHandle());
+    return statsPersistPeriodSec(nativeHandle_);
+  }
+
+  @Override
+  public Options setStatsHistoryBufferSize(
+      final long statsHistoryBufferSize) {
+    assert(isOwningHandle());
+    setStatsHistoryBufferSize(nativeHandle_, statsHistoryBufferSize);
+    return this;
+  }
+
+  @Override
+  public long statsHistoryBufferSize() {
+    assert(isOwningHandle());
+    return statsHistoryBufferSize(nativeHandle_);
+  }
+
   @Override
   public boolean adviseRandomOnOpen() {
     return adviseRandomOnOpen(nativeHandle_);
@@ -883,6 +911,19 @@ public class Options extends RocksObject
     return walBytesPerSync(nativeHandle_);
   }
 
+  @Override
+  public Options setStrictBytesPerSync(final boolean strictBytesPerSync) {
+    assert(isOwningHandle());
+    setStrictBytesPerSync(nativeHandle_, strictBytesPerSync);
+    return this;
+  }
+
+  @Override
+  public boolean strictBytesPerSync() {
+    assert(isOwningHandle());
+    return strictBytesPerSync(nativeHandle_);
+  }
+
   @Override
   public Options setEnableThreadTracking(final boolean enableThreadTracking) {
     assert(isOwningHandle());
@@ -1858,6 +1899,14 @@ public class Options extends RocksObject
   private native void setStatsDumpPeriodSec(
       long handle, int statsDumpPeriodSec);
   private native int statsDumpPeriodSec(long handle);
+  private native void setStatsPersistPeriodSec(
+      final long handle, final int statsPersistPeriodSec);
+  private native int statsPersistPeriodSec(
+      final long handle);
+  private native void setStatsHistoryBufferSize(
+      final long handle, final long statsHistoryBufferSize);
+  private native long statsHistoryBufferSize(
+      final long handle);
   private native void setAdviseRandomOnOpen(
       long handle, boolean adviseRandomOnOpen);
   private native boolean adviseRandomOnOpen(long handle);
@@ -1889,6 +1938,10 @@ public class Options extends RocksObject
   private native long bytesPerSync(long handle);
   private native void setWalBytesPerSync(long handle, long walBytesPerSync);
   private native long walBytesPerSync(long handle);
+  private native void setStrictBytesPerSync(
+      final long handle, final boolean strictBytesPerSync);
+  private native boolean strictBytesPerSync(
+      final long handle);
   private native void setEnableThreadTracking(long handle,
       boolean enableThreadTracking);
   private native boolean enableThreadTracking(long handle);

java/src/test/java/org/rocksdb/DBOptionsTest.java

@@ -406,6 +406,24 @@ public class DBOptionsTest {
     }
   }
 
+  @Test
+  public void statsPersistPeriodSec() {
+    try (final DBOptions opt = new DBOptions()) {
+      final int intValue = rand.nextInt();
+      opt.setStatsPersistPeriodSec(intValue);
+      assertThat(opt.statsPersistPeriodSec()).isEqualTo(intValue);
+    }
+  }
+
+  @Test
+  public void statsHistoryBufferSize() {
+    try (final DBOptions opt = new DBOptions()) {
+      final long longValue = rand.nextLong();
+      opt.setStatsHistoryBufferSize(longValue);
+      assertThat(opt.statsHistoryBufferSize()).isEqualTo(longValue);
+    }
+  }
+
   @Test
   public void adviseRandomOnOpen() {
     try(final DBOptions opt = new DBOptions()) {
@@ -516,6 +534,15 @@ public class DBOptionsTest {
     }
   }
 
+  @Test
+  public void strictBytesPerSync() {
+    try (final DBOptions opt = new DBOptions()) {
+      assertThat(opt.strictBytesPerSync()).isFalse();
+      opt.setStrictBytesPerSync(true);
+      assertThat(opt.strictBytesPerSync()).isTrue();
+    }
+  }
+
   @Test
   public void enableThreadTracking() {
     try (final DBOptions opt = new DBOptions()) {

java/src/test/java/org/rocksdb/MutableDBOptionsTest.java

@@ -56,21 +56,22 @@ public class MutableDBOptionsTest {
   }
 
   @Test
-  public void mutableColumnFamilyOptions_toString() {
+  public void mutableDBOptions_toString() {
     final String str = MutableDBOptions
         .builder()
         .setMaxOpenFiles(99)
         .setDelayedWriteRate(789)
         .setAvoidFlushDuringShutdown(true)
+        .setStrictBytesPerSync(true)
         .build()
         .toString();
 
     assertThat(str).isEqualTo("max_open_files=99;delayed_write_rate=789;"
-        + "avoid_flush_during_shutdown=true");
+        + "avoid_flush_during_shutdown=true;strict_bytes_per_sync=true");
   }
 
   @Test
-  public void mutableColumnFamilyOptions_parse() {
+  public void mutableDBOptions_parse() {
     final String str = "max_open_files=99;delayed_write_rate=789;"
         + "avoid_flush_during_shutdown=true";
 

java/src/test/java/org/rocksdb/OptionsTest.java

@@ -625,6 +625,24 @@ public class OptionsTest {
     }
   }
 
+  @Test
+  public void statsPersistPeriodSec() {
+    try (final Options opt = new Options()) {
+      final int intValue = rand.nextInt();
+      opt.setStatsPersistPeriodSec(intValue);
+      assertThat(opt.statsPersistPeriodSec()).isEqualTo(intValue);
+    }
+  }
+
+  @Test
+  public void statsHistoryBufferSize() {
+    try (final Options opt = new Options()) {
+      final long longValue = rand.nextLong();
+      opt.setStatsHistoryBufferSize(longValue);
+      assertThat(opt.statsHistoryBufferSize()).isEqualTo(longValue);
+    }
+  }
+
   @Test
   public void adviseRandomOnOpen() {
     try (final Options opt = new Options()) {
@@ -735,6 +753,15 @@ public class OptionsTest {
     }
   }
 
+  @Test
+  public void strictBytesPerSync() {
+    try (final Options opt = new Options()) {
+      assertThat(opt.strictBytesPerSync()).isFalse();
+      opt.setStrictBytesPerSync(true);
+      assertThat(opt.strictBytesPerSync()).isTrue();
+    }
+  }
+
   @Test
   public void enableThreadTracking() {
     try (final Options opt = new Options()) {

port/win/io_win.cc

@@ -1025,7 +1025,12 @@ Status WinRandomRWFile::Close() {
 //////////////////////////////////////////////////////////////////////////
 /// WinMemoryMappedBufer
 WinMemoryMappedBuffer::~WinMemoryMappedBuffer() {
-  BOOL ret = FALSE;
+  BOOL ret
+#if defined(_MSC_VER)
+    = FALSE;
+#else
+    __attribute__((__unused__));
+#endif
   if (base_ != nullptr) {
     ret = ::UnmapViewOfFile(base_);
     assert(ret);

port/win/win_thread.cc

@@ -138,7 +138,12 @@ void WindowsThread::join() {
       "WaitForSingleObjectFailed: thread join");
   }
 
-  BOOL rc;
+  BOOL rc
+#if defined(_MSC_VER)
+    = FALSE;
+#else
+    __attribute__((__unused__));
+#endif
   rc = CloseHandle(reinterpret_cast<HANDLE>(data_->handle_));
   assert(rc != 0);
   data_->handle_ = 0;

table/block_based/filter_policy.cc

@@ -443,7 +443,8 @@ void BloomFilterPolicy::CreateFilter(const Slice* keys, int n,
                                      std::string* dst) const {
   // We should ideally only be using this deprecated interface for
   // appropriately constructed BloomFilterPolicy
-  assert(mode_ == kDeprecatedBlock);
+  // FIXME disabled because of bug in C interface; see issue #6129
+  //assert(mode_ == kDeprecatedBlock);
 
   // Compute bloom filter size (in both bits and bytes)
   uint32_t bits = static_cast<uint32_t>(n * whole_bits_per_key_);

util/channel.h

@@ -60,7 +60,7 @@ class channel {
 
  private:
   std::condition_variable cv_;
-  std::mutex lock_;
+  mutable std::mutex lock_;
   std::queue<T> buffer_;
   bool eof_;
 };

util/status.cc

@@ -21,7 +21,12 @@ const char* Status::CopyState(const char* state) {
 #ifdef OS_WIN
   const size_t cch = std::strlen(state) + 1;  // +1 for the null terminator
   char* result = new char[cch];
-  errno_t ret;
+  errno_t ret
+#if defined(_MSC_VER)
+    ;
+#else
+    __attribute__((__unused__));
+#endif
   ret = strncpy_s(result, cch, state, cch - 1);
   result[cch - 1] = '\0';
   assert(ret == 0);

utilities/blob_db/blob_db.h

@@ -87,7 +87,7 @@ class BlobDB : public StackableDB {
   virtual Status Put(const WriteOptions& options,
                      ColumnFamilyHandle* column_family, const Slice& key,
                      const Slice& value) override {
-    if (column_family != DefaultColumnFamily()) {
+    if (column_family->GetID() != DefaultColumnFamily()->GetID()) {
       return Status::NotSupported(
           "Blob DB doesn't support non-default column family.");
     }
@@ -98,7 +98,7 @@ class BlobDB : public StackableDB {
   virtual Status Delete(const WriteOptions& options,
                         ColumnFamilyHandle* column_family,
                         const Slice& key) override {
-    if (column_family != DefaultColumnFamily()) {
+    if (column_family->GetID() != DefaultColumnFamily()->GetID()) {
       return Status::NotSupported(
           "Blob DB doesn't support non-default column family.");
     }
@@ -111,7 +111,7 @@ class BlobDB : public StackableDB {
   virtual Status PutWithTTL(const WriteOptions& options,
                             ColumnFamilyHandle* column_family, const Slice& key,
                             const Slice& value, uint64_t ttl) {
-    if (column_family != DefaultColumnFamily()) {
+    if (column_family->GetID() != DefaultColumnFamily()->GetID()) {
       return Status::NotSupported(
           "Blob DB doesn't support non-default column family.");
     }
@@ -125,7 +125,7 @@ class BlobDB : public StackableDB {
   virtual Status PutUntil(const WriteOptions& options,
                           ColumnFamilyHandle* column_family, const Slice& key,
                           const Slice& value, uint64_t expiration) {
-    if (column_family != DefaultColumnFamily()) {
+    if (column_family->GetID() != DefaultColumnFamily()->GetID()) {
       return Status::NotSupported(
           "Blob DB doesn't support non-default column family.");
     }
@@ -157,7 +157,7 @@ class BlobDB : public StackableDB {
       const std::vector<Slice>& keys,
       std::vector<std::string>* values) override {
     for (auto column_family : column_families) {
-      if (column_family != DefaultColumnFamily()) {
+      if (column_family->GetID() != DefaultColumnFamily()->GetID()) {
         return std::vector<Status>(
             column_families.size(),
             Status::NotSupported(
@@ -197,7 +197,7 @@ class BlobDB : public StackableDB {
   virtual Iterator* NewIterator(const ReadOptions& options) override = 0;
   virtual Iterator* NewIterator(const ReadOptions& options,
                                 ColumnFamilyHandle* column_family) override {
-    if (column_family != DefaultColumnFamily()) {
+    if (column_family->GetID() != DefaultColumnFamily()->GetID()) {
       // Blob DB doesn't support non-default column family.
       return nullptr;
     }
@@ -217,7 +217,7 @@ class BlobDB : public StackableDB {
       const int output_path_id = -1,
       std::vector<std::string>* const output_file_names = nullptr,
       CompactionJobInfo* compaction_job_info = nullptr) override {
-    if (column_family != DefaultColumnFamily()) {
+    if (column_family->GetID() != DefaultColumnFamily()->GetID()) {
       return Status::NotSupported(
           "Blob DB doesn't support non-default column family.");
     }

utilities/blob_db/blob_db_impl.cc

@@ -459,6 +459,10 @@ void BlobDBImpl::ProcessFlushJobInfo(const FlushJobInfo& info) {
 void BlobDBImpl::ProcessCompactionJobInfo(const CompactionJobInfo& info) {
   assert(bdb_options_.enable_garbage_collection);
 
+  if (!info.status.ok()) {
+    return;
+  }
+
   // Note: the same SST file may appear in both the input and the output
   // file list in case of a trivial move. We process the inputs first
   // to ensure the blob file still has a link after processing all updates.
@@ -975,7 +979,7 @@ Status BlobDBImpl::PutBlobValue(const WriteOptions& /*options*/,
     }
     if (s.ok()) {
       assert(blob_file != nullptr);
-      assert(blob_file->compression() == bdb_options_.compression);
+      assert(blob_file->GetCompressionType() == bdb_options_.compression);
       s = AppendBlob(blob_file, headerbuf, key, value_compressed, expiration,
                      &index_entry);
     }
@@ -1364,7 +1368,7 @@ Status BlobDBImpl::GetRawBlobFromFile(const Slice& key, uint64_t file_number,
     blob_file = it->second;
   }
 
-  *compression_type = blob_file->compression();
+  *compression_type = blob_file->GetCompressionType();
 
   // takes locks when called
   std::shared_ptr<RandomAccessFileReader> reader;
@@ -1462,7 +1466,7 @@ Status BlobDBImpl::Get(const ReadOptions& read_options,
 Status BlobDBImpl::GetImpl(const ReadOptions& read_options,
                            ColumnFamilyHandle* column_family, const Slice& key,
                            PinnableSlice* value, uint64_t* expiration) {
-  if (column_family != DefaultColumnFamily()) {
+  if (column_family->GetID() != DefaultColumnFamily()->GetID()) {
     return Status::NotSupported(
         "Blob DB doesn't support non-default column family.");
   }

utilities/blob_db/blob_db_test.cc

@@ -1780,6 +1780,36 @@ TEST_F(BlobDBTest, MaintainBlobFileToSstMapping) {
     ASSERT_EQ(obsolete_files[0]->BlobFileNumber(), 1);
   }
 
+  // Simulate a failed compaction. No mappings should be updated.
+  {
+    CompactionJobInfo info{};
+    info.input_file_infos.emplace_back(CompactionFileInfo{1, 7, 2});
+    info.input_file_infos.emplace_back(CompactionFileInfo{2, 22, 5});
+    info.output_file_infos.emplace_back(CompactionFileInfo{2, 25, 3});
+    info.status = Status::Corruption();
+
+    blob_db_impl()->TEST_ProcessCompactionJobInfo(info);
+
+    const std::vector<std::unordered_set<uint64_t>> expected_sst_files{
+        {}, {7}, {3, 8, 23}, {4, 9}, {5, 10, 22}};
+    const std::vector<bool> expected_obsolete{true, false, false, false, false};
+    for (size_t i = 0; i < 5; ++i) {
+      const auto &blob_file = blob_files[i];
+      ASSERT_EQ(blob_file->GetLinkedSstFiles(), expected_sst_files[i]);
+      ASSERT_EQ(blob_file->Obsolete(), expected_obsolete[i]);
+    }
+
+    auto live_imm_files = blob_db_impl()->TEST_GetLiveImmNonTTLFiles();
+    ASSERT_EQ(live_imm_files.size(), 4);
+    for (size_t i = 0; i < 4; ++i) {
+      ASSERT_EQ(live_imm_files[i]->BlobFileNumber(), i + 2);
+    }
+
+    auto obsolete_files = blob_db_impl()->TEST_GetObsoleteFiles();
+    ASSERT_EQ(obsolete_files.size(), 1);
+    ASSERT_EQ(obsolete_files[0]->BlobFileNumber(), 1);
+  }
+
   // Simulate another compaction. Blob file 2 loses all its linked SSTs
   // but since it got marked immutable at sequence number 300 which hasn't
   // been flushed yet, it cannot be marked obsolete at this point.

utilities/blob_db/blob_file.cc

@@ -54,7 +54,7 @@ BlobFile::~BlobFile() {
   }
 }
 
-uint32_t BlobFile::column_family_id() const { return column_family_id_; }
+uint32_t BlobFile::GetColumnFamilyId() const { return column_family_id_; }
 
 std::string BlobFile::PathName() const {
   return BlobFileName(path_to_dir_, file_number_);

utilities/blob_db/blob_file.h

@@ -116,7 +116,7 @@ class BlobFile {
 
   ~BlobFile();
 
-  uint32_t column_family_id() const;
+  uint32_t GetColumnFamilyId() const;
 
   // Returns log file's absolute pathname.
   std::string PathName() const;
@@ -204,7 +204,7 @@ class BlobFile {
 
   void SetHasTTL(bool has_ttl) { has_ttl_ = has_ttl; }
 
-  CompressionType compression() const { return compression_; }
+  CompressionType GetCompressionType() const { return compression_; }
 
   std::shared_ptr<Writer> GetWriter() const { return log_writer_; }
 

utilities/transactions/transaction_test.cc

@@ -2823,6 +2823,104 @@ TEST_P(TransactionTest, MultiGetBatchedTest) {
   }
 }
 
+// This test calls WriteBatchWithIndex::MultiGetFromBatchAndDB with a large
+// number of keys, i.e greater than MultiGetContext::MAX_BATCH_SIZE, which is
+// is 32. This forces autovector allocations in the MultiGet code paths
+// to use std::vector in addition to stack allocations. The MultiGet keys
+// includes Merges, which are handled specially in MultiGetFromBatchAndDB by
+// allocating an autovector of MergeContexts
+TEST_P(TransactionTest, MultiGetLargeBatchedTest) {
+  WriteOptions write_options;
+  ReadOptions read_options, snapshot_read_options;
+  string value;
+  Status s;
+
+  ColumnFamilyHandle* cf;
+  ColumnFamilyOptions cf_options;
+
+  std::vector<std::string> key_str;
+  for (int i = 0; i < 100; ++i) {
+    key_str.emplace_back(std::to_string(i));
+  }
+  // Create a new column families
+  s = db->CreateColumnFamily(cf_options, "CF", &cf);
+  ASSERT_OK(s);
+
+  delete cf;
+  delete db;
+  db = nullptr;
+
+  // open DB with three column families
+  std::vector<ColumnFamilyDescriptor> column_families;
+  // have to open default column family
+  column_families.push_back(
+      ColumnFamilyDescriptor(kDefaultColumnFamilyName, ColumnFamilyOptions()));
+  // open the new column families
+  cf_options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  column_families.push_back(ColumnFamilyDescriptor("CF", cf_options));
+
+  std::vector<ColumnFamilyHandle*> handles;
+
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  ASSERT_OK(ReOpenNoDelete(column_families, &handles));
+  assert(db != nullptr);
+
+  // Write some data to the db
+  WriteBatch batch;
+  for (int i = 0; i < 3 * MultiGetContext::MAX_BATCH_SIZE; ++i) {
+    std::string val = "val" + std::to_string(i);
+    batch.Put(handles[1], key_str[i], val);
+  }
+  s = db->Write(write_options, &batch);
+  ASSERT_OK(s);
+
+  WriteBatchWithIndex wb;
+  // Write some data to the db
+  s = wb.Delete(handles[1], std::to_string(1));
+  ASSERT_OK(s);
+  s = wb.Put(handles[1], std::to_string(2), "new_val" + std::to_string(2));
+  ASSERT_OK(s);
+  // Write a lot of merges so when we call MultiGetFromBatchAndDB later on,
+  // it is forced to use std::vector in rocksdb::autovector to allocate
+  // MergeContexts. The number of merges needs to be >
+  // MultiGetContext::MAX_BATCH_SIZE
+  for (int i = 8; i < MultiGetContext::MAX_BATCH_SIZE + 24; ++i) {
+    s = wb.Merge(handles[1], std::to_string(i), "merge");
+    ASSERT_OK(s);
+  }
+
+  // MultiGet a lot of keys in order to force std::vector reallocations
+  std::vector<Slice> keys;
+  for (int i = 0; i < MultiGetContext::MAX_BATCH_SIZE + 32; ++i) {
+    keys.emplace_back(key_str[i]);
+  }
+  std::vector<PinnableSlice> values(keys.size());
+  std::vector<Status> statuses(keys.size());
+
+  wb.MultiGetFromBatchAndDB(db, snapshot_read_options, handles[1], keys.size(), keys.data(),
+                values.data(), statuses.data(), false);
+  for (size_t i =0; i < keys.size(); ++i) {
+    if (i == 1) {
+      ASSERT_TRUE(statuses[1].IsNotFound());
+    } else if (i == 2) {
+      ASSERT_TRUE(statuses[2].ok());
+      ASSERT_EQ(values[2], "new_val" + std::to_string(2));
+    } else if (i >= 8 && i < 56) {
+      ASSERT_TRUE(statuses[i].ok());
+      ASSERT_EQ(values[i], "val" + std::to_string(i) + ",merge");
+    } else {
+      ASSERT_TRUE(statuses[i].ok());
+      if (values[i] != "val" + std::to_string(i)) {
+        ASSERT_EQ(values[i], "val" + std::to_string(i));
+      }
+    }
+  }
+
+  for (auto handle : handles) {
+    delete handle;
+  }
+}
+
 TEST_P(TransactionTest, ColumnFamiliesTest2) {
   WriteOptions write_options;
   ReadOptions read_options, snapshot_read_options;

utilities/write_batch_with_index/write_batch_with_index.cc

@@ -1004,10 +1004,13 @@ void WriteBatchWithIndex::MultiGetFromBatchAndDB(
     assert(result == WriteBatchWithIndexInternal::Result::kMergeInProgress ||
            result == WriteBatchWithIndexInternal::Result::kNotFound);
     key_context.emplace_back(column_family, keys[i], &values[i], &statuses[i]);
-    sorted_keys.emplace_back(&key_context.back());
     merges.emplace_back(result, std::move(merge_context));
   }
 
+  for (KeyContext& key : key_context) {
+    sorted_keys.emplace_back(&key);
+  }
+
   // Did not find key in batch OR could not resolve Merges.  Try DB.
   static_cast_with_check<DBImpl, DB>(db->GetRootDB())
       ->PrepareMultiGetKeys(key_context.size(), sorted_input, &sorted_keys);