Release 4.3.1

Summary:
table_test is failing because we are creating a temp InternalComparator

14:27:28 [ RUN      ] BlockBasedTableTest.NoopTransformSeek
14:27:28 pure virtual method called
14:27:28 terminate called without an active exception
14:27:28 /bin/sh: line 7: 2346261 Aborted                 (core dumped) ./$t

Test Plan: make table_test -j64 && ./table_test --gtest_filter="BlockBasedTableTest.NoopTransformSeek"

Reviewers: igor, sdong, anthony, rven

Reviewed By: rven

Subscribers: dhruba

Differential Revision: https://reviews.facebook.net/D52671

.travis.yml

@@ -34,7 +34,7 @@ before_script:
 # as EnvPosixTest::AllocateTest expects within the Travis OpenVZ environment.
 script:
   - if [[ "${TRAVIS_OS_NAME}" == 'linux' ]]; then OPT=-DTRAVIS CLANG_FORMAT_DIFF=/tmp/clang-format-diff.py make format || true; fi
-  - OPT=-DTRAVIS V=1 make -j4 check && OPT=-DTRAVIS V=1 make clean jclean rocksdbjava jtest
+  - OPT=-DTRAVIS V=1 make -j4 check && OPT=-DTRAVIS V=1 make clean jclean rocksdbjava jtest && make clean && OPT="-DTRAVIS -DROCKSDB_LITE" V=1 make -j4 check
 
 notifications:
     email:

CMakeLists.txt

@@ -14,10 +14,16 @@
 # 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 12 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 ..
-# 4. Then build the project in debug mode (you may want to add /m:<N> flag to run msbuild in <N> parallel threads)
-#        msbuild ALL_BUILD.vcxproj
+#        OR for VS Studio 15 cmake -G "Visual Studio 14 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=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 ot use all avail cores)
+#        msbuild rocksdb.sln
+#
+#        rocksdb.sln build features exclusions of test only code in Release. If you build ALL_BUILD then everything
+#        will be attempted but test only code does not build in Release mode.
+#
 # 5. And release mode (/m[:<N>] is also supported)
-#        msbuild ALL_BUILD.vcxproj /p:Configuration=Release
+#        msbuild rocksdb.sln /p:Configuration=Release
 #
 
 cmake_minimum_required(VERSION 2.6)
@@ -31,7 +37,14 @@ string(REGEX REPLACE "(..)/(..)/..(..).*" "\\1/\\2/\\3" DATE ${DATE})
 string(REGEX REPLACE "(..):(.....).*" " \\1:\\2" TIME ${TIME})
 string(CONCAT GIT_DATE_TIME ${DATE} ${TIME})
 
-execute_process(COMMAND $ENV{COMSPEC} " /C git rev-parse HEAD 2>nil" OUTPUT_VARIABLE GIT_SHA)
+find_package(Git)
+
+if (GIT_FOUND AND EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/.git")
+    execute_process(COMMAND $ENV{COMSPEC} /C ${GIT_EXECUTABLE} -C ${CMAKE_CURRENT_SOURCE_DIR} rev-parse HEAD OUTPUT_VARIABLE GIT_SHA)
+else()
+    set(GIT_SHA 0)
+endif()
+
 string(REGEX REPLACE "[^0-9a-f]+" "" GIT_SHA ${GIT_SHA})
 
 set(BUILD_VERSION_CC ${CMAKE_CURRENT_SOURCE_DIR}/util/build_version.cc)
@@ -46,7 +59,8 @@ add_custom_command(OUTPUT ${BUILD_VERSION_CC}
 add_custom_target(GenerateBuildVersion DEPENDS ${BUILD_VERSION_CC})
 
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zi /nologo  /EHsc /GS /Gd /GR /GF /fp:precise /Zc:wchar_t /Zc:forScope /errorReport:queue")
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /W3 /WX /wd4018 /wd4100 /wd4101 /wd4127 /wd4189 /wd4200 /wd4244 /wd4267 /wd4296 /wd4305 /wd4307 /wd4309 /wd4512 /wd4701 /wd4702 /wd4800 /wd4804 /wd4996")
+
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /W3 /WX /wd4127 /wd4244 /wd4267 /wd4800 /wd4996")
 
 # Used to run CI build and tests so we can run faster
 set(OPTIMIZE_DEBUG_DEFAULT 0)        # Debug build is unoptimized by default use -DOPTDBG=1 to optimize
@@ -83,6 +97,7 @@ set(LIBS ${ROCKSDB_LIBS} ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
 
 add_subdirectory(third-party/gtest-1.7.0/fused-src/gtest)
 
+# Main library source code
 set(SOURCES
         db/builder.cc
         db/c.cc
@@ -130,6 +145,9 @@ set(SOURCES
         db/write_batch_base.cc
         db/write_controller.cc
         db/write_thread.cc
+        memtable/hash_cuckoo_rep.cc
+        memtable/hash_linklist_rep.cc
+        memtable/hash_skiplist_rep.cc
         port/stack_trace.cc
         port/win/env_win.cc
         port/win/port_win.cc
@@ -163,6 +181,7 @@ set(SOURCES
         table/plain_table_reader.cc
         table/table_properties.cc
         table/two_level_iterator.cc
+        tools/sst_dump_tool.cc
         tools/dump/db_dump_tool.cc
         util/arena.cc
         util/auto_roll_logger.cc
@@ -175,7 +194,6 @@ set(SOURCES
         util/crc32c.cc
         util/db_info_dumper.cc
         util/delete_scheduler_impl.cc
-        util/db_test_util.cc
         util/dynamic_bloom.cc
         util/env.cc
         util/env_hdfs.cc
@@ -184,14 +202,11 @@ set(SOURCES
         util/file_reader_writer.cc
         util/filter_policy.cc
         util/hash.cc
-        util/hash_cuckoo_rep.cc
-        util/hash_linklist_rep.cc
-        util/hash_skiplist_rep.cc
         util/histogram.cc
         util/instrumented_mutex.cc
         util/iostats_context.cc
-        util/ldb_cmd.cc
-        util/ldb_tool.cc
+        tools/ldb_cmd.cc
+        tools/ldb_tool.cc
         util/logging.cc
         util/log_buffer.cc
         util/memenv.cc
@@ -202,12 +217,13 @@ set(SOURCES
         util/options_builder.cc
         util/options_helper.cc
         util/options_parser.cc
+        util/options_sanity_check.cc
         util/perf_context.cc
         util/perf_level.cc
+        util/random.cc
         util/rate_limiter.cc
         util/skiplistrep.cc
         util/slice.cc
-        util/sst_dump_tool.cc
         util/statistics.cc
         util/status.cc
         util/status_message.cc
@@ -218,7 +234,6 @@ set(SOURCES
         util/thread_local.cc
         util/thread_status_impl.cc
         util/thread_status_updater.cc
-        util/thread_status_updater_debug.cc
         util/thread_status_util.cc
         util/thread_status_util_debug.cc
         util/vectorrep.cc
@@ -232,10 +247,12 @@ set(SOURCES
         utilities/flashcache/flashcache.cc
         utilities/geodb/geodb_impl.cc
         utilities/leveldb_options/leveldb_options.cc
+        utilities/memory/memory_util.cc
         utilities/merge_operators/string_append/stringappend.cc
         utilities/merge_operators/string_append/stringappend2.cc
         utilities/merge_operators/put.cc
         utilities/merge_operators/uint64add.cc
+        utilities/options/options_util.cc
         utilities/redis/redis_lists.cc
         utilities/spatialdb/spatial_db.cc
         utilities/table_properties_collectors/compact_on_deletion_collector.cc
@@ -252,6 +269,13 @@ set(SOURCES
         utilities/write_batch_with_index/write_batch_with_index_internal.cc
 )
 
+# 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
+    util/thread_status_updater_debug.cc
+)
+
 add_library(rocksdblib${ARTIFACT_SUFFIX} ${SOURCES})
 set_target_properties(rocksdblib${ARTIFACT_SUFFIX} PROPERTIES COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/rocksdblib${ARTIFACT_SUFFIX}.pdb")
 add_dependencies(rocksdblib${ARTIFACT_SUFFIX} GenerateBuildVersion)
@@ -266,6 +290,7 @@ set(APPS
         db/memtablerep_bench.cc
         table/table_reader_bench.cc
         tools/db_stress.cc
+        tools/write_stress.cc
         tools/db_repl_stress.cc
         tools/sst_dump.cc
         tools/dump/rocksdb_dump.cc
@@ -294,17 +319,21 @@ set(TESTS
         db/db_universal_compaction_test.cc
         db/db_wal_test.cc
         db/db_tailing_iter_test.cc
+        db/db_table_properties_test.cc
         db/dbformat_test.cc
         db/deletefile_test.cc
         db/fault_injection_test.cc
         db/file_indexer_test.cc
         db/filename_test.cc
         db/flush_job_test.cc
+        db/inlineskiplist_test.cc
         db/listener_test.cc
         db/log_test.cc
+        db/manual_compaction_test.cc
         db/memtable_list_test.cc
         db/merge_test.cc
         db/merge_helper_test.cc
+        db/options_file_test.cc
         db/perf_context_test.cc
         db/plain_table_db_test.cc
         db/prefix_test.cc
@@ -326,7 +355,9 @@ set(TESTS
         table/merger_test.cc
         table/table_test.cc
         tools/db_sanity_test.cc
+        tools/ldb_cmd_test.cc
         tools/reduce_levels_test.cc
+        tools/sst_dump_test.cc
         util/arena_test.cc
         util/autovector_test.cc
         util/auto_roll_logger_test.cc
@@ -341,14 +372,11 @@ set(TESTS
         util/file_reader_writer_test.cc
         util/heap_test.cc
         util/histogram_test.cc
-        util/ldb_cmd_test.cc
-        util/manual_compaction_test.cc
         util/memenv_test.cc
         util/mock_env_test.cc
         util/options_test.cc
         util/rate_limiter_test.cc
         util/slice_transform_test.cc
-        util/sst_dump_test.cc
         util/thread_list_test.cc
         util/thread_local_test.cc
         utilities/backupable/backupable_db_test.cc
@@ -356,7 +384,9 @@ set(TESTS
         utilities/document/document_db_test.cc
         utilities/document/json_document_test.cc
         utilities/geodb/geodb_test.cc
+        utilities/memory/memory_test.cc
         utilities/merge_operators/string_append/stringappend_test.cc
+        utilities/options/options_util_test.cc
         utilities/redis/redis_lists_test.cc
         utilities/spatialdb/spatial_db_test.cc
         utilities/table_properties_collectors/compact_on_deletion_collector_test.cc
@@ -366,7 +396,7 @@ set(TESTS
         utilities/write_batch_with_index/write_batch_with_index_test.cc
 )
 
-set(EXES ${APPS} ${TESTS})
+set(EXES ${APPS})
 
 foreach(sourcefile ${EXES})
     string(REPLACE ".cc" "" exename ${sourcefile})
@@ -375,12 +405,42 @@ foreach(sourcefile ${EXES})
     target_link_libraries(${exename}${ARTIFACT_SUFFIX} ${LIBS})
 endforeach(sourcefile ${EXES})
 
-# C executables must link to a shared object
-set(C_EXES ${C_TESTS})
+# test utilities are only build in debug
+set(TESTUTILLIB testutillib${ARTIFACT_SUFFIX})
+add_library(${TESTUTILLIB} STATIC ${TESTUTIL_SOURCE})
+set_target_properties(${TESTUTILLIB} PROPERTIES COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/testutillib${ARTIFACT_SUFFIX}.pdb")
+set_target_properties(${TESTUTILLIB}
+      PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
+      )
 
-foreach(sourcefile ${C_EXES})
+# Tests are excluded from Release builds
+set(TEST_EXES ${TESTS})
+
+foreach(sourcefile ${TEST_EXES})
+    string(REPLACE ".cc" "" exename ${sourcefile})
+    string(REGEX REPLACE "^((.+)/)+" "" exename ${exename})
+    add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile})
+    set_target_properties(${exename}${ARTIFACT_SUFFIX}
+      PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
+      )
+    target_link_libraries(${exename}${ARTIFACT_SUFFIX} ${LIBS} testutillib${ARTIFACT_SUFFIX})
+endforeach(sourcefile ${TEST_EXES})
+
+# C executables must link to a shared object
+set(C_TEST_EXES ${C_TESTS})
+
+foreach(sourcefile ${C_TEST_EXES})
     string(REPLACE ".c" "" exename ${sourcefile})
     string(REGEX REPLACE "^((.+)/)+" "" exename ${exename})
     add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile})
-    target_link_libraries(${exename}${ARTIFACT_SUFFIX} rocksdb${ARTIFACT_SUFFIX})
-endforeach(sourcefile ${C_TESTS})
+    set_target_properties(${exename}${ARTIFACT_SUFFIX}
+      PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
+      )
+    target_link_libraries(${exename}${ARTIFACT_SUFFIX} rocksdb${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX})
+endforeach(sourcefile ${C_TEST_EXES})

HISTORY.md

@@ -1,5 +1,28 @@
 # Rocksdb Change Log
 
+## 4.3.0 (12/8/2015)
+### New Features
+* CompactionFilter has new member function called IgnoreSnapshots which allows CompactionFilter to be called even if there are snapshots later than the key.
+* RocksDB will now persist options under the same directory as the RocksDB database on successful DB::Open, CreateColumnFamily, DropColumnFamily, and SetOptions.
+* Introduce LoadLatestOptions() in rocksdb/utilities/options_util.h.  This function can construct the latest DBOptions / ColumnFamilyOptions used by the specified RocksDB intance.
+* Introduce CheckOptionsCompatibility() in rocksdb/utilities/options_util.h.  This function checks whether the input set of options is able to open the specified DB successfully.
+
+### Public API Changes
+* When options.db_write_buffer_size triggers, only the column family with the largest column family size will be flushed, not all the column families.
+
+## 4.2.0 (11/9/2015)
+### New Features
+* Introduce CreateLoggerFromOptions(), this function create a Logger for provided DBOptions. 
+* Add GetAggregatedIntProperty(), which returns the sum of the GetIntProperty of all the column families. 
+* Add MemoryUtil in rocksdb/utilities/memory.h.  It currently offers a way to get the memory usage by type from a list rocksdb instances.
+
+### Public API Changes
+* CompactionFilter::Context includes information of Column Family ID
+* The need-compaction hint given by TablePropertiesCollector::NeedCompact() will be persistent and recoverable after DB recovery. This introduces a breaking format change. If you use this experimental feature, including NewCompactOnDeletionCollectorFactory() in the new version, you may not be able to directly downgrade the DB back to version 4.0 or lower.
+* TablePropertiesCollectorFactory::CreateTablePropertiesCollector() now takes an option Context, containing the information of column family ID for the file being written.
+* Remove DefaultCompactionFilterFactory.
+
+
 ## 4.1.0 (10/8/2015)
 ### New Features
 * Added single delete operation as a more efficient way to delete keys that have not been overwritten.

INSTALL.md

@@ -1,19 +1,24 @@
 ## Compilation
 
+**Important**: If you plan to run RocksDB in production, don't compile using default 
+`make` or `make all`. That will compile RocksDB in debug mode, which is much slower
+than release mode.
+
 RocksDB's library should be able to compile without any dependency installed,
 although we recommend installing some compression libraries (see below).
 We do depend on newer gcc/clang with C++11 support.
 
 There are few options when compiling RocksDB:
 
-* [recommended] `make static_lib` will compile librocksdb.a, RocksDB static library.
+* [recommended] `make static_lib` will compile librocksdb.a, RocksDB static library. Compiles static library in release mode.
 
-* `make shared_lib` will compile librocksdb.so, RocksDB shared library.
+* `make shared_lib` will compile librocksdb.so, RocksDB shared library. Compiles shared library in release mode.
 
-* `make check` will compile and run all the unit tests
+* `make check` will compile and run all the unit tests. `make check` will compile RocksDB in debug mode.
 
 * `make all` will compile our static library, and all our tools and unit tests. Our tools
-depend on gflags. You will need to have gflags installed to run `make all`.
+depend on gflags. You will need to have gflags installed to run `make all`. This will compile RocksDB in debug mode. Don't
+use binaries compiled by `make all` in production.
 
 * By default the binary we produce is optimized for the platform you're compiling on
 (-march=native). If you want to build a portable binary, add 'PORTABLE=1' before
@@ -82,4 +87,5 @@ your make commands, like this: `PORTABLE=1 make static_lib`
   * Run: `TARGET_OS=IOS make static_lib`. When building the project which uses rocksdb iOS library, make sure to define two important pre-processing macros: `ROCKSDB_LITE` and `IOS_CROSS_COMPILE`.
 
 * **Windows**:
+  * For building with MS Visual Studio 13 you will need Update 4 installed.
   * Read and follow the instructions at CMakeLists.txt

Makefile

@@ -33,12 +33,22 @@ quoted_perl_command = $(subst ','\'',$(perl_command))
 # with debug level 0. To compile with level 0, run `make shared_lib`,
 # `make install-shared`, `make static_lib`, `make install-static` or
 # `make install`
-DEBUG_LEVEL=1
+
+# Set the default DEBUG_LEVEL to 1
+DEBUG_LEVEL?=1
 
 ifeq ($(MAKECMDGOALS),dbg)
 	DEBUG_LEVEL=2
 endif
 
+ifeq ($(MAKECMDGOALS),clean)
+	DEBUG_LEVEL=0
+endif
+
+ifeq ($(MAKECMDGOALS),release)
+	DEBUG_LEVEL=0
+endif
+
 ifeq ($(MAKECMDGOALS),shared_lib)
 	DEBUG_LEVEL=0
 endif
@@ -84,6 +94,8 @@ endif
 ifeq ($(DEBUG_LEVEL),0)
 OPT += -DNDEBUG
 DISABLE_WARNING_AS_ERROR=1
+else
+$(warning Warning: Compiling in debug mode. Don't use the resulting binary in production)
 endif
 
 #-----------------------------------------------
@@ -236,6 +248,7 @@ TESTS = \
 	db_tailing_iter_test \
 	db_universal_compaction_test \
 	db_wal_test \
+	db_table_properties_test \
 	block_hash_index_test \
 	autovector_test \
 	column_family_test \
@@ -261,14 +274,17 @@ TESTS = \
 	block_based_filter_block_test \
 	full_filter_block_test \
 	histogram_test \
+	inlineskiplist_test \
 	log_test \
 	manual_compaction_test \
 	memenv_test \
 	mock_env_test \
 	memtable_list_test \
 	merge_helper_test \
+	memory_test \
 	merge_test \
 	merger_test \
+	options_file_test \
 	redis_test \
 	reduce_levels_test \
 	plain_table_db_test \
@@ -296,6 +312,7 @@ TESTS = \
 	rate_limiter_test \
 	delete_scheduler_test \
 	options_test \
+	options_util_test \
 	event_logger_test \
 	cuckoo_table_builder_test \
 	cuckoo_table_reader_test \
@@ -323,17 +340,23 @@ TOOLS = \
 	sst_dump \
 	db_sanity_test \
 	db_stress \
+	write_stress \
 	ldb \
 	db_repl_stress \
 	rocksdb_dump \
 	rocksdb_undump
 
+# TODO: add back forward_iterator_bench, after making it build in all environemnts.
 BENCHMARKS = db_bench table_reader_bench cache_bench memtablerep_bench
 
-# The library name is configurable since we are maintaining libraries of both
-# debug/release mode.
+# if user didn't config LIBNAME, set the default
 ifeq ($(LIBNAME),)
+# we should only run rocksdb in production with DEBUG_LEVEL 0
+ifeq ($(DEBUG_LEVEL),0)
         LIBNAME=librocksdb
+else
+        LIBNAME=librocksdb_debug
+endif
 endif
 LIBRARY = ${LIBNAME}.a
 
@@ -379,7 +402,7 @@ $(SHARED3): $(SHARED4)
 endif
 
 $(SHARED4):
-	$(CXX) $(PLATFORM_SHARED_LDFLAGS)$(SHARED3) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(LIB_SOURCES) \
+	$(CXX) $(PLATFORM_SHARED_LDFLAGS)$(SHARED3) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(LIB_SOURCES) $(TOOL_SOURCES) \
 		$(LDFLAGS) -o $@
 
 endif  # PLATFORM_SHARED_EXT
@@ -403,7 +426,7 @@ dbg: $(LIBRARY) $(BENCHMARKS) tools $(TESTS)
 # creates static library and programs
 release:
 	$(MAKE) clean
-	OPT="-DNDEBUG -O2" $(MAKE) static_lib tools db_bench
+	DEBUG_LEVEL=0 $(MAKE) static_lib tools db_bench
 
 coverage:
 	$(MAKE) clean
@@ -552,8 +575,10 @@ check: all
 	      echo "===== Running $$t"; ./$$t || exit 1; done;          \
 	fi
 	rm -rf $(TMPD)
+ifeq ($(filter -DROCKSDB_LITE,$(OPT)),)
 	python tools/ldb_test.py
 	sh tools/rocksdb_dump_test.sh
+endif
 
 check_some: $(SUBSET) ldb_tests
 	for t in $(SUBSET); do echo "===== Running $$t"; ./$$t || exit 1; done
@@ -565,12 +590,12 @@ ldb_tests: ldb
 crash_test: whitebox_crash_test blackbox_crash_test
 
 blackbox_crash_test: db_stress
-	python -u tools/db_crashtest.py -s
-	python -u tools/db_crashtest.py
+	python -u tools/db_crashtest.py --simple blackbox 
+	python -u tools/db_crashtest.py blackbox
 
 whitebox_crash_test: db_stress
-	python -u tools/db_crashtest2.py -s
-	python -u tools/db_crashtest2.py
+	python -u tools/db_crashtest.py --simple whitebox
+	python -u tools/db_crashtest.py whitebox
 
 asan_check:
 	$(MAKE) clean
@@ -583,7 +608,7 @@ asan_crash_test:
 	$(MAKE) clean
 
 valgrind_check: $(TESTS)
-	for t in $(filter-out skiplist_test,$(TESTS)); do \
+	for t in $(filter-out %skiplist_test,$(TESTS)); do \
 		$(VALGRIND_VER) $(VALGRIND_OPTS) ./$$t; \
 		ret_code=$$?; \
 		if [ $$ret_code -ne 0 ]; then \
@@ -611,7 +636,7 @@ unity.a: unity.o
 	$(AM_V_at)$(AR) $(ARFLAGS) $@ unity.o
 
 # try compiling db_test with unity
-unity_test: db/db_test.o util/db_test_util.o $(TESTHARNESS) unity.a
+unity_test: db/db_test.o db/db_test_util.o $(TESTHARNESS) unity.a
 	$(AM_LINK)
 	./unity_test
 
@@ -657,6 +682,9 @@ block_hash_index_test: table/block_hash_index_test.o $(LIBOBJECTS) $(TESTHARNESS
 db_stress: tools/db_stress.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)
 
+write_stress: tools/write_stress.o $(LIBOBJECTS) $(TESTUTIL)
+	$(AM_LINK)
+
 db_sanity_test: tools/db_sanity_test.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)
 
@@ -711,34 +739,37 @@ crc32c_test: util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS)
 slice_transform_test: util/slice_transform_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_test: db/db_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_test: db/db_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_log_iter_test: db/db_log_iter_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_log_iter_test: db/db_log_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_compaction_filter_test: db/db_compaction_filter_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_compaction_filter_test: db/db_compaction_filter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_compaction_test: db/db_compaction_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_compaction_test: db/db_compaction_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_dynamic_level_test: db/db_dynamic_level_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_dynamic_level_test: db/db_dynamic_level_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_inplace_update_test: db/db_inplace_update_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_inplace_update_test: db/db_inplace_update_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_tailing_iter_test: db/db_tailing_iter_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_tailing_iter_test: db/db_tailing_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
 db_iter_test: db/db_iter_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_universal_compaction_test: db/db_universal_compaction_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_universal_compaction_test: db/db_universal_compaction_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-db_wal_test: db/db_wal_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_wal_test: db/db_wal_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+db_table_properties_test: db/db_table_properties_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
 log_write_bench: util/log_write_bench.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -834,6 +865,9 @@ table_test: table/table_test.o $(LIBOBJECTS) $(TESTHARNESS)
 block_test: table/block_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+inlineskiplist_test: db/inlineskiplist_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 skiplist_test: db/skiplist_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -864,12 +898,18 @@ write_controller_test: db/write_controller_test.o $(LIBOBJECTS) $(TESTHARNESS)
 merge_helper_test: db/merge_helper_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+memory_test: utilities/memory/memory_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 merge_test: db/merge_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
 merger_test: table/merger_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+options_file_test: db/options_file_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 deletefile_test: db/deletefile_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -903,10 +943,13 @@ compact_files_test: db/compact_files_test.o $(LIBOBJECTS) $(TESTHARNESS)
 options_test: util/options_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+options_util_test: utilities/options/options_util_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 event_logger_test: util/event_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-sst_dump_test: util/sst_dump_test.o $(LIBOBJECTS) $(TESTHARNESS)
+sst_dump_test: tools/sst_dump_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
 memenv_test : util/memenv_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -918,7 +961,7 @@ optimistic_transaction_test: utilities/transactions/optimistic_transaction_test.
 mock_env_test : util/mock_env_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-manual_compaction_test: util/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
+manual_compaction_test: db/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
 filelock_test: util/filelock_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -942,7 +985,7 @@ transaction_test: utilities/transactions/transaction_test.o $(LIBOBJECTS) $(TEST
 sst_dump: tools/sst_dump.o $(LIBOBJECTS)
 	$(AM_LINK)
 
-ldb_cmd_test: util/ldb_cmd_test.o $(LIBOBJECTS) $(TESTHARNESS)
+ldb_cmd_test: tools/ldb_cmd_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
 ldb: tools/ldb.o $(LIBOBJECTS)
@@ -1037,20 +1080,23 @@ liblz4.a:
 	   cd lz4-r127/lib && make CFLAGS='-fPIC' all
 	   cp lz4-r127/lib/liblz4.a .
 
-# A version of each $(LIBOBJECTS) compiled with -fPIC
-java_libobjects = $(patsubst %,jl/%,$(LIBOBJECTS))
-CLEAN_FILES += jl
+# A version of each $(LIBOBJECTS) compiled with -fPIC and a fixed set of static compression libraries
+java_static_libobjects = $(patsubst %,jls/%,$(LIBOBJECTS))
+CLEAN_FILES += jls
 
-$(java_libobjects): jl/%.o: %.cc
-	$(AM_V_CC)mkdir -p $(@D) && $(CXX) $(CXXFLAGS) -fPIC -c $< -o $@ $(COVERAGEFLAGS)
+JAVA_STATIC_FLAGS = -DZLIB -DBZIP2 -DSNAPPY -DLZ4
+JAVA_STATIC_INCLUDES = -I./zlib-1.2.8 -I./bzip2-1.0.6 -I./snappy-1.1.1 -I./lz4-r127/lib
 
-rocksdbjavastatic: $(java_libobjects) libz.a libbz2.a libsnappy.a liblz4.a
+$(java_static_libobjects): jls/%.o: %.cc libz.a libbz2.a libsnappy.a liblz4.a
+	$(AM_V_CC)mkdir -p $(@D) && $(CXX) $(CXXFLAGS) $(JAVA_STATIC_FLAGS) $(JAVA_STATIC_INCLUDES) -fPIC -c $< -o $@ $(COVERAGEFLAGS)
+
+rocksdbjavastatic: $(java_static_libobjects)
 	cd java;$(MAKE) javalib;
 	rm -f ./java/target/$(ROCKSDBJNILIB)
 	$(CXX) $(CXXFLAGS) -I./java/. $(JAVA_INCLUDE) -shared -fPIC \
 	  -o ./java/target/$(ROCKSDBJNILIB) $(JNI_NATIVE_SOURCES) \
-	  $(java_libobjects) $(COVERAGEFLAGS) \
-	  libz.a libbz2.a libsnappy.a liblz4.a $(LDFLAGS)
+	  $(java_static_libobjects) $(COVERAGEFLAGS) \
+	  libz.a libbz2.a libsnappy.a liblz4.a $(JAVA_STATIC_LDFLAGS)
 	cd java/target;strip -S -x $(ROCKSDBJNILIB)
 	cd java;jar -cf target/$(ROCKSDB_JAR) HISTORY*.md
 	cd java/target;jar -uf $(ROCKSDB_JAR) $(ROCKSDBJNILIB)
@@ -1061,7 +1107,7 @@ rocksdbjavastatic: $(java_libobjects) libz.a libbz2.a libsnappy.a liblz4.a
 rocksdbjavastaticrelease: rocksdbjavastatic
 	cd java/crossbuild && vagrant destroy -f && vagrant up linux32 && vagrant halt linux32 && vagrant up linux64 && vagrant halt linux64
 	cd java;jar -cf target/$(ROCKSDB_JAR_ALL) HISTORY*.md
-	cd java;jar -uf target/$(ROCKSDB_JAR_ALL) librocksdbjni-*.so librocksdbjni-*.jnilib
+	cd java/target;jar -uf $(ROCKSDB_JAR_ALL) librocksdbjni-*.so librocksdbjni-*.jnilib
 	cd java/target/classes;jar -uf ../$(ROCKSDB_JAR_ALL) org/rocksdb/*.class org/rocksdb/util/*.class
 
 rocksdbjavastaticpublish: rocksdbjavastaticrelease
@@ -1072,6 +1118,13 @@ rocksdbjavastaticpublish: rocksdbjavastaticrelease
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar -Dclassifier=osx
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar
 
+# A version of each $(LIBOBJECTS) compiled with -fPIC
+java_libobjects = $(patsubst %,jl/%,$(LIBOBJECTS))
+CLEAN_FILES += jl
+
+$(java_libobjects): jl/%.o: %.cc
+	$(AM_V_CC)mkdir -p $(@D) && $(CXX) $(CXXFLAGS) -fPIC -c $< -o $@ $(COVERAGEFLAGS)
+
 rocksdbjava: $(java_libobjects)
 	$(AM_V_GEN)cd java;$(MAKE) javalib;
 	$(AM_V_at)rm -f ./java/target/$(ROCKSDBJNILIB)

appveyor.yml

@@ -2,10 +2,14 @@ version: 1.0.{build}
 before_build:
 - md %APPVEYOR_BUILD_FOLDER%\build
 - cd %APPVEYOR_BUILD_FOLDER%\build
-- cmake -G "Visual Studio 12 Win64" ..
+- cmake -G "Visual Studio 12 Win64" -DOPTDBG=1 ..
 - cd ..
 build:
-  project: build\ALL_BUILD.vcxproj
+  project: build\rocksdb.sln
   parallel: true
   verbosity: minimal
-test: off
+test:
+test_script:
+- ps: build_tools\run_ci_db_test.ps1 -EnableRerun -Run db_test -Exclude DBTest.Randomized,DBTest.FileCreationRandomFailure -Concurrency 18
+- ps: build_tools\run_ci_db_test.ps1 -Run env_test -Concurrency 1
+

appveyordailytests.yml

@@ -1,22 +0,0 @@
-version: 1.0.{build}
-before_build:
-- md %APPVEYOR_BUILD_FOLDER%\build
-- cd %APPVEYOR_BUILD_FOLDER%\build
-- cmake -G "Visual Studio 12 Win64" -DOPTDBG=1 ..
-- cd ..
-build:
-  project: build\ALL_BUILD.vcxproj
-  parallel: true
-  verbosity: minimal
-test:
-test_script:
-- ps: build_tools\run_ci_db_test.ps1
-notifications:
-  - provider: Email
-    to:
-      - svmtrocksdb@microsoft.com
-    subject: "Build {{status}}"
-    message: "{{message}}, {{commitId}}, ..."
-    on_build_success: false
-    on_build_failure: true
-    on_build_status_changed: true

arcanist_util/cpp_linter/FbcodeCppLinter.php

@@ -88,6 +88,9 @@ class FbcodeCppLinter extends ArcanistLinter {
   }
 
   private function getCppLintOutput($path) {
+     if (!array_key_exists($path, $this->rawLintOutput)) {
+       return array();
+     }
     list($output) = $this->rawLintOutput[$path];
 
     $msgs = array();

build_tools/build_detect_platform

@@ -8,6 +8,7 @@
 #   CXX                         C++ Compiler path
 #   PLATFORM_LDFLAGS            Linker flags
 #   JAVA_LDFLAGS                Linker flags for RocksDBJava
+#   JAVA_STATIC_LDFLAGS         Linker flags for RocksDBJava static build
 #   PLATFORM_SHARED_EXT         Extension for shared libraries
 #   PLATFORM_SHARED_LDFLAGS     Flags for building shared library
 #   PLATFORM_SHARED_CFLAGS      Flags for compiling objects for shared library
@@ -44,7 +45,7 @@ fi
 # we depend on C++11
 PLATFORM_CXXFLAGS="-std=c++11"
 # we currently depend on POSIX platform
-COMMON_FLAGS="-DROCKSDB_PLATFORM_POSIX"
+COMMON_FLAGS="-DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX"
 
 # Default to fbcode gcc on internal fb machines
 if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
@@ -181,6 +182,7 @@ esac
 
 PLATFORM_CXXFLAGS="$PLATFORM_CXXFLAGS ${CXXFLAGS}"
 JAVA_LDFLAGS="$PLATFORM_LDFLAGS"
+JAVA_STATIC_LDFLAGS="$PLATFORM_LDFLAGS"
 
 if [ "$CROSS_COMPILE" = "true" -o "$FBCODE_BUILD" = "true" ]; then
     # Cross-compiling; do not try any compilation tests.
@@ -374,6 +376,7 @@ echo "CXX=$CXX" >> "$OUTPUT"
 echo "PLATFORM=$PLATFORM" >> "$OUTPUT"
 echo "PLATFORM_LDFLAGS=$PLATFORM_LDFLAGS" >> "$OUTPUT"
 echo "JAVA_LDFLAGS=$JAVA_LDFLAGS" >> "$OUTPUT"
+echo "JAVA_STATIC_LDFLAGS=$JAVA_STATIC_LDFLAGS" >> "$OUTPUT"
 echo "VALGRIND_VER=$VALGRIND_VER" >> "$OUTPUT"
 echo "PLATFORM_CCFLAGS=$PLATFORM_CCFLAGS" >> "$OUTPUT"
 echo "PLATFORM_CXXFLAGS=$PLATFORM_CXXFLAGS" >> "$OUTPUT"

build_tools/fbcode_config.sh

@@ -43,9 +43,9 @@ if test -z $PIC_BUILD; then
   LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/79d2943e2dd7208a3e0b06cf95e9f85f05fe9e1b/r124/gcc-4.9-glibc-2.20/4230243/lib/liblz4.a"
   CFLAGS+=" -DLZ4"
 
-  ZSTD_REV=8df2d01673ae6afcc8c8d16fec862b2d67ecc1e9
-  ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.1.1/gcc-4.8.1-glibc-2.17/c3f970a/include"
-  ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.1.1/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"
+  ZSTD_REV=810b81b4705def5243e998b54701f3c504e4009e
+  ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/include"
+  ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"
   CFLAGS+=" -DZSTD"
 fi
 
@@ -116,7 +116,7 @@ else
 fi
 
 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
 CXXFLAGS+=" $CFLAGS"
 
 EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB"

build_tools/fbcode_config4.8.1.sh

@@ -5,61 +5,66 @@
 # uses jemalloc
 
 # location of libgcc
-LIBGCC_BASE="/mnt/gvfs/third-party2/libgcc/7712e757d7355cb51292454ee0b7b46a467fdfed/4.8.1/gcc-4.8.1-glibc-2.17/8aac7fc"
+LIBGCC_BASE="/mnt/gvfs/third-party2/libgcc/d00277f4559e261ed0a81f30f23c0ce5564e359e/4.8.1/gcc-4.8.1-glibc-2.17/8aac7fc"
 LIBGCC_INCLUDE="$LIBGCC_BASE/include"
 LIBGCC_LIBS=" -L $LIBGCC_BASE/libs"
 
 # location of glibc
-GLIBC_REV=6e40560b4e0b6d690fd1cf8c7a43ad7452b04cfa
+GLIBC_REV=0600c95b31226b5e535614c590677d87c62d8016
 GLIBC_INCLUDE="/mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.17/gcc-4.8.1-glibc-2.17/99df8fc/include"
 GLIBC_LIBS=" -L /mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.17/gcc-4.8.1-glibc-2.17/99df8fc/lib"
 
 # location of snappy headers and libraries
-SNAPPY_INCLUDE=" -I /mnt/gvfs/third-party2/snappy/aef17f6c0b44b4fe408bd06f67c93701ab0a6ceb/1.0.3/gcc-4.8.1-glibc-2.17/43d84e2/include"
-SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/aef17f6c0b44b4fe408bd06f67c93701ab0a6ceb/1.0.3/gcc-4.8.1-glibc-2.17/43d84e2/lib/libsnappy.a"
+SNAPPY_REV=cbf6f1f209e5bd160bdc5d971744e039f36b1566
+SNAPPY_INCLUDE=" -I /mnt/gvfs/third-party2/snappy/$SNAPPY_REV/1.1.3/gcc-4.8.1-glibc-2.17/c3f970a/include"
+SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/$SNAPPY_REV/1.1.3/gcc-4.8.1-glibc-2.17/c3f970a/lib/libsnappy.a"
 
 # location of zlib headers and libraries
-ZLIB_INCLUDE=" -I /mnt/gvfs/third-party2/zlib/25c6216928b4d77b59ddeca0990ff6fe9ac16b81/1.2.5/gcc-4.8.1-glibc-2.17/c3f970a/include"
-ZLIB_LIBS=" /mnt/gvfs/third-party2/zlib/25c6216928b4d77b59ddeca0990ff6fe9ac16b81/1.2.5/gcc-4.8.1-glibc-2.17/c3f970a/lib/libz.a"
+ZLIB_REV=6d39cb54708049f527e713ad19f2aadb9d3667e8
+ZLIB_INCLUDE=" -I /mnt/gvfs/third-party2/zlib/$ZLIB_REV/1.2.8/gcc-4.8.1-glibc-2.17/c3f970a/include"
+ZLIB_LIBS=" /mnt/gvfs/third-party2/zlib/$ZLIB_REV/1.2.8/gcc-4.8.1-glibc-2.17/c3f970a/lib/libz.a"
 
 # location of bzip headers and libraries
-BZIP_INCLUDE=" -I /mnt/gvfs/third-party2/bzip2/c9ef7629c2aa0024f7a416e87602f06eb88f5eac/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a/include/"
-BZIP_LIBS=" /mnt/gvfs/third-party2/bzip2/c9ef7629c2aa0024f7a416e87602f06eb88f5eac/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a/lib/libbz2.a"
+BZIP_REV=d6c789bfc2ec4c51a63d66df2878926b8158cde8
+BZIP_INCLUDE=" -I /mnt/gvfs/third-party2/bzip2/$BZIP_REV/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a/include/"
+BZIP_LIBS=" /mnt/gvfs/third-party2/bzip2/$BZIP_REV/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a/lib/libbz2.a"
 
-LZ4_REV=065ec7e38fe83329031f6668c43bef83eff5808b
-LZ4_INCLUDE=" -I /mnt/gvfs/third-party2/lz4/$LZ4_REV/r108/gcc-4.8.1-glibc-2.17/c3f970a/include"
-LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/$LZ4_REV/r108/gcc-4.8.1-glibc-2.17/c3f970a/lib/liblz4.a"
+LZ4_REV=6858fac689e0f92e584224d91bdb0e39f6c8320d
+LZ4_INCLUDE=" -I /mnt/gvfs/third-party2/lz4/$LZ4_REV/r131/gcc-4.8.1-glibc-2.17/c3f970a/include"
+LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/$LZ4_REV/r131/gcc-4.8.1-glibc-2.17/c3f970a/lib/liblz4.a"
 
-ZSTD_REV=8df2d01673ae6afcc8c8d16fec862b2d67ecc1e9
-ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.1.1/gcc-4.8.1-glibc-2.17/c3f970a/include"
-ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.1.1/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"
+ZSTD_REV=810b81b4705def5243e998b54701f3c504e4009e
+ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/include"
+ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"
 
 # location of gflags headers and libraries
-GFLAGS_INCLUDE=" -I /mnt/gvfs/third-party2/gflags/1ad047a6e6f6673991918ecadc670868205a243a/1.6/gcc-4.8.1-glibc-2.17/c3f970a/include/"
-GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/1ad047a6e6f6673991918ecadc670868205a243a/1.6/gcc-4.8.1-glibc-2.17/c3f970a/lib/libgflags.a"
+GFLAGS_REV=c7275a4ceae0aca0929e56964a31dafc53c1ee96
+GFLAGS_INCLUDE=" -I /mnt/gvfs/third-party2/gflags/$GFLAGS_REV/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a/include/"
+GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/$GFLAGS_REV/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a/lib/libgflags.a"
 
 # location of jemalloc
-JEMALLOC_INCLUDE=" -I /mnt/gvfs/third-party2/jemalloc/3691c776ac26dd8781e84f8888b6a0fbdbc0a9ed/dev/gcc-4.8.1-glibc-2.17/4d53c6f/include"
-JEMALLOC_LIB="/mnt/gvfs/third-party2/jemalloc/3691c776ac26dd8781e84f8888b6a0fbdbc0a9ed/dev/gcc-4.8.1-glibc-2.17/4d53c6f/lib/libjemalloc.a"
+JEMALLOC_REV=c370265e58c4b6602e798df23335a1e9913dae52
+JEMALLOC_INCLUDE=" -I /mnt/gvfs/third-party2/jemalloc/$JEMALLOC_REV/4.0.3/gcc-4.8.1-glibc-2.17/8d31e51/include"
+JEMALLOC_LIB="/mnt/gvfs/third-party2/jemalloc/$JEMALLOC_REV/4.0.3/gcc-4.8.1-glibc-2.17/8d31e51/lib/libjemalloc.a"
 
 # location of numa
-NUMA_REV=829d10dac0230f99cd7e1778869d2adf3da24b65
+NUMA_REV=ae54a5ed22cdabb1c6446dce4e8ffae5b4446d73
 NUMA_INCLUDE=" -I /mnt/gvfs/third-party2/numa/$NUMA_REV/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a/include/"
 NUMA_LIB=" /mnt/gvfs/third-party2/numa/$NUMA_REV/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a/lib/libnuma.a"
 
 # location of libunwind
-LIBUNWIND_REV=2c060e64064559905d46fd194000d61592087bdc
-LIBUNWIND="/mnt/gvfs/third-party2/libunwind/$LIBUNWIND_REV/1.1/gcc-4.8.1-glibc-2.17/675d945/lib/libunwind.a"
+LIBUNWIND_REV=121f1a75c4414683aea8c70b761bfaf187f7c1a3
+LIBUNWIND="/mnt/gvfs/third-party2/libunwind/$LIBUNWIND_REV/trunk/gcc-4.8.1-glibc-2.17/675d945/lib/libunwind.a"
 
 # use Intel SSE support for checksum calculations
 export USE_SSE=1
 
-BINUTILS="/mnt/gvfs/third-party2/binutils/2aff2e7b474cd3e6ab23495ad1224b7d214b9f8e/2.21.1/centos6-native/da39a3e/bin"
+BINUTILS="/mnt/gvfs/third-party2/binutils/75670d0d8ef4891fd1ec2a7513ef01cd002c823b/2.25/centos6-native/da39a3e/bin"
 AR="$BINUTILS/ar"
 
 DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"
 
-GCC_BASE="/mnt/gvfs/third-party2/gcc/1ec615e23800f0815d474478ba476a0adc3fe788/4.8.1/centos6-native/cc6c9dc"
+GCC_BASE="/mnt/gvfs/third-party2/gcc/c0064002d2609ab649603f769f0bd110bbe48029/4.8.1/centos6-native/cc6c9dc"
 STDLIBS="-L $GCC_BASE/lib64"
 
 if [ -z "$USE_CLANG" ]; then
@@ -72,12 +77,12 @@ if [ -z "$USE_CLANG" ]; then
   CFLAGS+=" -isystem $LIBGCC_INCLUDE"
 else
   # clang 
-  CLANG_BASE="/mnt/gvfs/third-party2/clang/9ab68376f938992c4eb5946ca68f90c3185cffc8/3.4"
+  CLANG_BASE="/mnt/gvfs/third-party2/clang/ab054e9a490a8fd4537c0b6ec56e5c91c0f81c91/3.4"
   CLANG_INCLUDE="$CLANG_BASE/gcc-4.8.1-glibc-2.17/fb0f730/lib/clang/3.4/include"
   CC="$CLANG_BASE/centos6-native/9cefd8a/bin/clang"
   CXX="$CLANG_BASE/centos6-native/9cefd8a/bin/clang++"
 
-  KERNEL_HEADERS_INCLUDE="/mnt/gvfs/third-party2/kernel-headers/a683ed7135276731065a9d76d3016c9731f4e2f9/3.2.18_70_fbk11_00129_gc8882d0/gcc-4.8.1-glibc-2.17/da39a3e/include/"
+  KERNEL_HEADERS_INCLUDE="/mnt/gvfs/third-party2/kernel-headers/1a48835975c66d30e47770ec419758ed3b9ba010/3.10.62-62_fbk17_03959_ge29cc63/gcc-4.8.1-glibc-2.17/da39a3e/include/"
 
   CFLAGS="-B$BINUTILS/gold -nostdinc -nostdlib"
   CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.8.1 "
@@ -91,7 +96,7 @@ else
 fi
 
 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
 CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4 -DZSTD -DNUMA"
 CXXFLAGS+=" $CFLAGS"
 
@@ -104,7 +109,7 @@ PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
 
 EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"
 
-VALGRIND_REV=b2a9f85e4b70cd03abc85a7f3027fbc4cef35bd0
+VALGRIND_REV=af85c56f424cd5edfc2c97588299b44ecdec96bb
 VALGRIND_VER="/mnt/gvfs/third-party2/valgrind/$VALGRIND_REV/3.8.1/gcc-4.8.1-glibc-2.17/c3f970a/bin/"
 
 export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE

build_tools/rocksdb-lego-determinator

@@ -68,6 +68,14 @@ CLANG="USE_CLANG=1"
 LITE="OPT=-DROCKSDB_LITE"
 TSAN="COMPILE_WITH_TSAN=1"
 DISABLE_JEMALLOC="DISABLE_JEMALLOC=1"
+PARSER="'parser':'egrep \'Failure|^#|Abort\''"
+
+ARTIFACTS=" 'artifacts': [
+    {
+        'name':'database',
+        'paths':[ '/dev/shm/rocksdb' ],
+    }
+]"
 
 #
 # A mechanism to disable tests temporarily
@@ -100,13 +108,15 @@ PARALLEL_UNIT_TEST_COMMANDS="[
             {
                 'name':'Build and test RocksDB debug version',
                 'shell':'$DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || cat t/log-*',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
             $CLEANUP_ENV,
             {
                 'name':'Build and test RocksDB debug version under gcc-4.8.1',
                 'shell':'$GCC_481 $DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || cat t/log-*',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -125,7 +135,8 @@ UNIT_TEST_COMMANDS="[
             {
                 'name':'Build and test RocksDB debug version',
                 'shell':'$SHM $DEBUG make J=1 check',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -144,7 +155,8 @@ UNIT_TEST_COMMANDS_481="[
             {
                 'name':'Build and test RocksDB debug version',
                 'shell':'$SHM $GCC_481 $DEBUG make J=1 check',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -163,7 +175,8 @@ CLANG_UNIT_TEST_COMMANDS="[
             {
                 'name':'Build and test RocksDB debug',
                 'shell':'$CLANG $SHM $DEBUG make J=1 check',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -182,7 +195,8 @@ CLANG_ANALYZE_COMMANDS="[
             {
                 'name':'RocksDB build and analyze',
                 'shell':'$CLANG $SHM $DEBUG make J=1 analyze',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -201,7 +215,8 @@ CODE_COV_COMMANDS="[
             {
                 'name':'Build, test and collect code coverage info',
                 'shell':'$SHM $DEBUG make J=1 coverage',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -220,7 +235,8 @@ UNITY_COMMANDS="[
             {
                 'name':'Build, test unity test',
                 'shell':'$SHM $DEBUG V=1 make J=1 unity_test',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -239,7 +255,28 @@ LITE_BUILD_COMMANDS="[
             {
                 'name':'Build RocksDB debug version',
                 'shell':'$LITE $DEBUG make J=1 static_lib',
-                'user':'root'
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB lite tests
+#
+LITE_UNIT_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb Lite Unit Test',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build RocksDB debug version',
+                'shell':'$SHM $LITE $DEBUG make J=1 check',
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -259,19 +296,44 @@ STRESS_CRASH_TEST_COMMANDS="[
             {
                 'name':'Build and run RocksDB debug stress tests',
                 'shell':'$SHM $DEBUG make J=1 db_stress',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
             {
                 'name':'Build and run RocksDB debug crash tests',
                 'timeout': 86400,
                 'shell':'$SHM $DEBUG make J=1 crash_test',
-                'user':'root'
+                'user':'root',
+                $PARSER
             }
         ],
+        $ARTIFACTS,
         $REPORT
     }
 ]"
 
+# RocksDB write stress test.
+# We run on disk device on purpose (i.e. no $SHM)
+# because we want to add some randomness to fsync commands
+WRITE_STRESS_COMMANDS="[
+    {
+        'name':'Rocksdb Write Stress Test',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build and run RocksDB write stress tests',
+                'shell':'make write_stress && python tools/write_stress_runner.py --runtime_sec=3600 --db=/tmp/rocksdb_write_stress',
+                'user':'root',
+                $PARSER
+            }
+        ],
+        'artifacts': [{'name': 'database', 'paths': ['/tmp/rocksdb_write_stress']}],
+        $REPORT
+    }
+]"
+
+
 #
 # RocksDB test under address sanitizer
 #
@@ -284,7 +346,8 @@ ASAN_TEST_COMMANDS="[
             {
                 'name':'Test RocksDB debug under ASAN',
                 'shell':'set -o pipefail && $SHM $ASAN $DEBUG make J=1 asan_check |& /usr/facebook/ops/scripts/asan_symbolize.py -d',
-                'user':'root'
+                'user':'root',
+                $PARSER
             }
         ],
         $REPORT
@@ -305,7 +368,8 @@ ASAN_CRASH_TEST_COMMANDS="[
                 'name':'Build and run RocksDB debug asan_crash_test',
                 'timeout': 86400,
                 'shell':'$SHM $DEBUG make J=1 asan_crash_test',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -324,7 +388,8 @@ VALGRIND_TEST_COMMANDS="[
             {
                 'name':'Run RocksDB debug unit tests',
                 'shell':'$DISABLE_JEMALLOC $SHM $DEBUG make valgrind_check',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -343,7 +408,8 @@ TSAN_UNIT_TEST_COMMANDS="[
             {
                 'name':'Run RocksDB debug unit test',
                 'shell':'set -o pipefail && $SHM $DEBUG $TSAN make J=1 check',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -364,13 +430,16 @@ TSAN_CRASH_TEST_COMMANDS="[
                 'name':'Compile and run',
                 'timeout': 86400,
                 'shell':'set -o pipefail && $SHM $DEBUG $TSAN make J=1 crash_test',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
     }
 ]"
 
+TSAN_CRASH_TEST_COMMANDS=$DISABLE_COMMANDS
+
 #
 # RocksDB format compatible
 #
@@ -425,7 +494,8 @@ FORMAT_COMPATIBLE_COMMANDS="[
             {
                 'name':'Run RocksDB debug unit test',
                 'shell':'build_tools/rocksdb-lego-determinator run_format_compatible',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -457,7 +527,8 @@ NO_COMPRESSION_COMMANDS="[
             {
                 'name':'Run RocksDB debug unit test',
                 'shell':'build_tools/rocksdb-lego-determinator run_no_compression',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -513,7 +584,8 @@ REGRESSION_COMMANDS="[
             {
                 'name':'Make and run script',
                 'shell':'build_tools/rocksdb-lego-determinator run_regression',
-                'user':'root'
+                'user':'root',
+                $PARSER
             },
         ],
         $REPORT
@@ -545,9 +617,15 @@ case $1 in
   lite)
     echo $LITE_BUILD_COMMANDS
     ;;
+  lite_test)
+    echo $LITE_UNIT_TEST_COMMANDS
+    ;;
   stress_crash)
     echo $STRESS_CRASH_TEST_COMMANDS
     ;;
+  write_stress)
+    echo $WRITE_STRESS_COMMANDS
+    ;;
   asan)
     echo $ASAN_TEST_COMMANDS
     ;;

build_tools/run_ci_db_test.ps1

@@ -5,10 +5,15 @@
 # Run the script from the enlistment
 Param(
   [switch]$EnableJE = $false,  # Use je executable
+  [switch]$EnableRerun = $false, # Rerun failed tests sequentially at the end
   [string]$WorkFolder = "",  # Direct tests to use that folder
   [int]$Limit = -1, # -1 means run all otherwise limit for testing purposes
   [string]$Exclude = "", # Expect a comma separated list, no spaces
-  [string]$Run = "db_test"  # Run db_test|tests
+  [string]$Run = "db_test",  # Run db_test|tests|testname1,testname2...
+   # Number of async tasks that would run concurrently. Recommend a number below 64.
+   # However, CPU utlization really depends on the storage media. Recommend ram based disk.
+   # a value of 1 will run everything serially
+  [int]$Concurrency = 16
 )
 
 # Folders and commands must be fullpath to run assuming
@@ -49,7 +54,7 @@ if($EnableJE) {
 }
 
 Write-Output "Root: $RootFolder, WorkFolder: $WorkFolder"
-Write-Output "Binaries: $BinariesFolder exe: $db_test"
+Write-Output "Binaries: $BinariesFolder db_test: $db_test"
 
 #Exclusions that we do not want to run
 $ExcludeTests = New-Object System.Collections.Generic.HashSet[string]
@@ -57,7 +62,7 @@ $ExcludeTests = New-Object System.Collections.Generic.HashSet[string]
 
 if($Exclude -ne "") {
     Write-Host "Exclude: $Exclude"
-    $l = $Exclude -split ','
+    $l = $Exclude -split ' '
     ForEach($t in $l) { $ExcludeTests.Add($t) | Out-Null }
 }
 
@@ -109,38 +114,69 @@ function Normalize-DbTests($HashTable) {
 
         $test_log = $test -replace '[\./]','_'
         $test_log += ".log"
+        $log_path = -join ($LogFolder, $test_log)
 
         # Add to a hashtable
-        $HashTable.Add($test, $test_log);
+        $HashTable.Add($test, $log_path);
       }
     }
 }
 
+# The function removes trailing .exe siffix if any,
+# creates a name for the log file
+function MakeAndAdd([string]$token, $HashTable) {
+    $test_name = $token -replace '.exe$', ''
+    $log_name =  -join ($test_name, ".log")
+    $log_path = -join ($LogFolder, $log_name)
+    if(!$ExcludeTests.Contains($test_name)) {
+        $HashTable.Add($test_name, $log_path)
+    } else {
+        Write-Warning "Test $test_name is excluded"
+    }
+}
+
 # The function scans build\Debug folder to discover
 # Test executables. It then populates a table with
 # Test executable name -> Log file
-function Discover-TestBinaries($HashTable) {
+function Discover-TestBinaries([string]$Pattern, $HashTable) {
 
     $Exclusions = @("db_test*", "db_sanity_test*")
-    $p = -join ($BinariesFolder, "*_test*.exe")
+
+    $p = -join ($BinariesFolder, $pattern)
+
+    Write-Host "Path: $p"
 
     dir -Path $p -Exclude $Exclusions | ForEach-Object {
-       $t = ($_.Name) -replace '.exe$', ''
-       $test_log = -join ($t, ".log")
-       $HashTable.Add($t, $test_log)
+       MakeAndAdd -token ($_.Name) -HashTable $HashTable
     }
 }
 
-$TestToLog = [ordered]@{}
+$TestsToRun = [ordered]@{}
 
 if($Run -ceq "db_test") {
-    Normalize-DbTests -HashTable $TestToLog
+    Normalize-DbTests -HashTable $TestsToRun
 } elseif($Run -ceq "tests") {
-    Discover-TestBinaries -HashTable $TestToLog
+    if($EnableJE) {
+        $pattern = "*_test_je.exe"
+    } else {
+        $pattern = "*_test.exe"
+    }
+    Discover-TestBinaries -Pattern $pattern -HashTable $TestsToRun
+} else {
+
+    $test_list = $Run -split ' '
+
+    ForEach($t in $test_list) {
+       MakeAndAdd -token $t -HashTable $TestsToRun
+    }
 }
 
+$NumTestsToStart = $TestsToRun.Count
+if($Limit -ge 0 -and $NumTestsToStart -gt $Limit) {
+    $NumTestsToStart = $Limit
+}
 
-Write-Host "Attempting to start: " ($TestToLog.Count) " tests"
+Write-Host "Attempting to start: $NumTestsToStart tests"
 
 # Invoke a test with a filter and redirect all output
 $InvokeTestCase = {
@@ -154,93 +190,123 @@ $InvokeTestAsync = {
     &$exe > $log 2>&1
 }
 
-$jobs = @()
-$JobToLog = @{}
+# Hash that contains tests to rerun if any failed
+# Those tests will be rerun sequentially
+$Rerun = [ordered]@{}
 # Test limiting factor here
 $count = 0
-
-ForEach($k in $TestToLog.keys) {
-
-    Write-Host "Starting $k"
-    $log_path = -join ($LogFolder, ($TestToLog.$k))
-
-    if($Run -ceq "db_test") {
-        $job = Start-Job -Name $k -ScriptBlock $InvokeTestCase -ArgumentList @($db_test,$k,$log_path)
-    } else {
-        [string]$Exe =  -Join ($BinariesFolder, $k)
-        $job = Start-Job -Name $k -ScriptBlock $InvokeTestAsync -ArgumentList @($exe,$log_path)
-    }
-
-    $JobToLog.Add($job, $log_path)
-
-    # Limiting trial runs
-    if(($Limit -gt 0) -and (++$count -ge $Limit)) {
-         break
-    }
-}
-
+# Overall status
 [bool]$success = $true;
 
-# Wait for all to finish and get the results
-while($JobToLog.Count -gt 0) {
-
+function RunJobs($TestToLog, [int]$ConcurrencyVal, [bool]$AddForRerun)
+{
+    # Array to wait for any of the running jobs
     $jobs = @()
-    foreach($k in $JobToLog.Keys) { $jobs += $k }
+    # Hash JobToLog
+    $JobToLog = @{}
 
-<#
-    if(!$success) {
-        break
-    }
-#>
+    # Wait for all to finish and get the results
+    while(($JobToLog.Count -gt 0) -or
+          ($TestToLog.Count -gt 0)) {
 
-    $completed = Wait-Job -Job $jobs -Any
-    $log = $JobToLog[$completed]
-    $JobToLog.Remove($completed)
+        # Make sure we have maximum concurrent jobs running if anything
+        # and the $Limit either not set or allows to proceed
+        while(($JobToLog.Count -lt $ConcurrencyVal) -and
+              (($TestToLog.Count -gt 0) -and
+              (($Limit -lt 0) -or ($count -lt $Limit)))) {
 
-    $message = -join @($completed.Name, " State: ", ($completed.State))
 
-    $log_content = @(Get-Content $log)
-
-    if($completed.State -ne "Completed") {
-        $success = $false
-        Write-Warning $message
-        $log_content | Write-Warning
-    } else {
-        # Scan the log. If we find PASSED and no occurence of FAILED
-        # then it is a success
-        [bool]$pass_found = $false
-        ForEach($l in $log_content) {
-
-            if(($l -match "^\[\s+FAILED") -or
-               ($l -match "Assertion failed:")) {
-                $pass_found = $false
+            # We only need the first key
+            foreach($key in $TestToLog.keys) {
+                $k = $key
                 break
             }
 
-            if(($l -match "^\[\s+PASSED") -or
-               ($l -match " : PASSED$") -or
-                ($l -match "^PASSED") -or
-                ($l -match "Passed all tests!") ) {
-                $pass_found = $true
+            Write-Host "Starting $k"
+            $log_path = ($TestToLog.$k)
+
+            if($Run -ceq "db_test") {
+              $job = Start-Job -Name $k -ScriptBlock $InvokeTestCase -ArgumentList @($db_test,$k,$log_path)
+            } else {
+              [string]$Exe =  -Join ($BinariesFolder, $k)
+               $job = Start-Job -Name $k -ScriptBlock $InvokeTestAsync -ArgumentList @($exe,$log_path)
             }
+
+            $JobToLog.Add($job, $log_path)
+            $TestToLog.Remove($k)
+
+            ++$count
         }
 
-        if(!$pass_found) {
-            $success = $false;
+        if($JobToLog.Count -lt 1) {
+          break
+        }
+
+        $jobs = @()
+        foreach($k in $JobToLog.Keys) { $jobs += $k }
+
+        $completed = Wait-Job -Job $jobs -Any
+        $log = $JobToLog[$completed]
+        $JobToLog.Remove($completed)
+
+        $message = -join @($completed.Name, " State: ", ($completed.State))
+
+        $log_content = @(Get-Content $log)
+
+        if($completed.State -ne "Completed") {
+            $success = $false
             Write-Warning $message
             $log_content | Write-Warning
         } else {
-            Write-Host $message
-        }
-    }
+            # Scan the log. If we find PASSED and no occurence of FAILED
+            # then it is a success
+            [bool]$pass_found = $false
+            ForEach($l in $log_content) {
 
-    # Remove cached job info from the system
-    # Should be no output
-    Receive-Job -Job $completed | Out-Null
+                if(($l -match "^\[\s+FAILED") -or
+                   ($l -match "Assertion failed:")) {
+                    $pass_found = $false
+                    break
+                }
+
+                if(($l -match "^\[\s+PASSED") -or
+                   ($l -match " : PASSED$") -or
+                    ($l -match "^PASS$") -or   # Special c_test case
+                    ($l -match "Passed all tests!") ) {
+                    $pass_found = $true
+                }
+            }
+
+            if(!$pass_found) {
+                $success = $false;
+                Write-Warning $message
+                $log_content | Write-Warning
+                if($AddForRerun) {
+                    $Rerun.Add($completed.Name, $log)
+                }
+            } else {
+                Write-Host $message
+            }
+        }
+
+        # Remove cached job info from the system
+        # Should be no output
+        Receive-Job -Job $completed | Out-Null
+    }
+}
+
+RunJobs -TestToLog $TestsToRun -ConcurrencyVal $Concurrency -AddForRerun $EnableRerun
+
+if($Rerun.Count -gt 0) {
+    Write-Host "Rerunning " ($Rerun.Count) " tests sequentially"
+    $success = $true
+    $count = 0
+    RunJobs -TestToLog $Rerun -ConcurrencyVal 1 -AddForRerun $false
 }
 
 Get-Date
 
+
 if(!$success) {
 # This does not succeed killing off jobs quick
 # So we simply exit

db/builder.cc

@@ -26,6 +26,7 @@
 #include "rocksdb/options.h"
 #include "rocksdb/table.h"
 #include "table/block_based_table_builder.h"
+#include "table/internal_iterator.h"
 #include "util/file_reader_writer.h"
 #include "util/iostats_context_imp.h"
 #include "util/stop_watch.h"
@@ -40,22 +41,25 @@ TableBuilder* NewTableBuilder(
     const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,
-    WritableFileWriter* file, const CompressionType compression_type,
+    uint32_t column_family_id, WritableFileWriter* file,
+    const CompressionType compression_type,
     const CompressionOptions& compression_opts, const bool skip_filters) {
   return ioptions.table_factory->NewTableBuilder(
       TableBuilderOptions(ioptions, internal_comparator,
                           int_tbl_prop_collector_factories, compression_type,
                           compression_opts, skip_filters),
-      file);
+      column_family_id, file);
 }
 
 Status BuildTable(
     const std::string& dbname, Env* env, const ImmutableCFOptions& ioptions,
-    const EnvOptions& env_options, TableCache* table_cache, Iterator* iter,
-    FileMetaData* meta, const InternalKeyComparator& internal_comparator,
+    const EnvOptions& env_options, TableCache* table_cache,
+    InternalIterator* iter, FileMetaData* meta,
+    const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,
-    std::vector<SequenceNumber> snapshots, const CompressionType compression,
+    uint32_t column_family_id, std::vector<SequenceNumber> snapshots,
+    const CompressionType compression,
     const CompressionOptions& compression_opts, bool paranoid_file_checks,
     InternalStats* internal_stats, const Env::IOPriority io_priority,
     TableProperties* table_properties) {
@@ -72,7 +76,7 @@ Status BuildTable(
     unique_ptr<WritableFileWriter> file_writer;
     {
       unique_ptr<WritableFile> file;
-      s = env->NewWritableFile(fname, &file, env_options);
+      s = NewWritableFile(env, fname, &file, env_options);
       if (!s.ok()) {
         return s;
       }
@@ -82,7 +86,7 @@ Status BuildTable(
 
       builder = NewTableBuilder(
           ioptions, internal_comparator, int_tbl_prop_collector_factories,
-          file_writer.get(), compression, compression_opts);
+          column_family_id, file_writer.get(), compression, compression_opts);
     }
 
     MergeHelper merge(env, internal_comparator.user_comparator(),
@@ -139,7 +143,7 @@ Status BuildTable(
 
     if (s.ok() && !empty) {
       // Verify that the table is usable
-      std::unique_ptr<Iterator> it(table_cache->NewIterator(
+      std::unique_ptr<InternalIterator> it(table_cache->NewIterator(
           ReadOptions(), env_options, internal_comparator, meta->fd, nullptr,
           (internal_stats == nullptr) ? nullptr
                                       : internal_stats->GetFileReadHist(0),

db/builder.h

@@ -31,13 +31,15 @@ class VersionEdit;
 class TableBuilder;
 class WritableFileWriter;
 class InternalStats;
+class InternalIterator;
 
 TableBuilder* NewTableBuilder(
     const ImmutableCFOptions& options,
     const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,
-    WritableFileWriter* file, const CompressionType compression_type,
+    uint32_t column_family_id, WritableFileWriter* file,
+    const CompressionType compression_type,
     const CompressionOptions& compression_opts,
     const bool skip_filters = false);
 
@@ -48,11 +50,13 @@ TableBuilder* NewTableBuilder(
 // zero, and no Table file will be produced.
 extern Status BuildTable(
     const std::string& dbname, Env* env, const ImmutableCFOptions& options,
-    const EnvOptions& env_options, TableCache* table_cache, Iterator* iter,
-    FileMetaData* meta, const InternalKeyComparator& internal_comparator,
+    const EnvOptions& env_options, TableCache* table_cache,
+    InternalIterator* iter, FileMetaData* meta,
+    const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,
-    std::vector<SequenceNumber> snapshots, const CompressionType compression,
+    uint32_t column_family_id, std::vector<SequenceNumber> snapshots,
+    const CompressionType compression,
     const CompressionOptions& compression_opts, bool paranoid_file_checks,
     InternalStats* internal_stats,
     const Env::IOPriority io_priority = Env::IO_HIGH,

db/c.cc

@@ -1288,6 +1288,11 @@ void rocksdb_block_based_options_set_cache_index_and_filter_blocks(
   options->rep.cache_index_and_filter_blocks = v;
 }
 
+void rocksdb_block_based_options_set_skip_table_builder_flush(
+    rocksdb_block_based_table_options_t* options, unsigned char v) {
+  options->rep.skip_table_builder_flush = v;
+}
+
 void rocksdb_options_set_block_based_table_factory(
     rocksdb_options_t *opt,
     rocksdb_block_based_table_options_t* table_options) {
@@ -1687,6 +1692,11 @@ void rocksdb_options_set_keep_log_file_num(rocksdb_options_t* opt, size_t v) {
   opt->rep.keep_log_file_num = v;
 }
 
+void rocksdb_options_set_recycle_log_file_num(rocksdb_options_t* opt,
+                                              size_t v) {
+  opt->rep.recycle_log_file_num = v;
+}
+
 void rocksdb_options_set_soft_rate_limit(rocksdb_options_t* opt, double v) {
   opt->rep.soft_rate_limit = v;
 }

db/column_family.cc

@@ -27,9 +27,9 @@
 #include "db/version_set.h"
 #include "db/write_controller.h"
 #include "db/writebuffer.h"
+#include "memtable/hash_skiplist_rep.h"
 #include "util/autovector.h"
 #include "util/compression.h"
-#include "util/hash_skiplist_rep.h"
 #include "util/options_helper.h"
 #include "util/thread_status_util.h"
 #include "util/xfunc.h"
@@ -175,6 +175,12 @@ ColumnFamilyOptions SanitizeOptions(const DBOptions& db_options,
     result.level0_stop_writes_trigger = std::numeric_limits<int>::max();
   }
 
+  if (result.level0_file_num_compaction_trigger == 0) {
+    Warn(db_options.info_log.get(),
+         "level0_file_num_compaction_trigger cannot be 0");
+    result.level0_file_num_compaction_trigger = 1;
+  }
+
   if (result.level0_stop_writes_trigger <
           result.level0_slowdown_writes_trigger ||
       result.level0_slowdown_writes_trigger <

db/column_family_test.cc

@@ -23,8 +23,6 @@
 #include "util/sync_point.h"
 #include "utilities/merge_operators.h"
 
-#if !(defined NDEBUG) || !defined(OS_WIN)
-
 namespace rocksdb {
 
 namespace {
@@ -59,6 +57,7 @@ class ColumnFamilyTest : public testing::Test {
     env_ = new EnvCounter(Env::Default());
     dbname_ = test::TmpDir() + "/column_family_test";
     db_options_.create_if_missing = true;
+    db_options_.fail_if_options_file_error = true;
     db_options_.env = env_;
     DestroyDB(dbname_, Options(db_options_, column_family_options_));
   }
@@ -69,7 +68,9 @@ class ColumnFamilyTest : public testing::Test {
 
   void Close() {
     for (auto h : handles_) {
-      delete h;
+      if (h) {
+        delete h;
+      }
     }
     handles_.clear();
     names_.clear();
@@ -721,7 +722,7 @@ TEST_F(ColumnFamilyTest, DifferentWriteBufferSizes) {
   two.max_write_buffer_number = 10;
   two.min_write_buffer_number_to_merge = 3;
   two.max_write_buffer_number_to_maintain = 2;
-  three.write_buffer_size = 4096 * 22 + 2048;
+  three.write_buffer_size = 4096 * 22;
   three.arena_block_size = 4096;
   three.max_write_buffer_number = 10;
   three.min_write_buffer_number_to_merge = 4;
@@ -746,15 +747,15 @@ TEST_F(ColumnFamilyTest, DifferentWriteBufferSizes) {
   env_->SleepForMicroseconds(micros_wait_for_flush);
   AssertNumberOfImmutableMemtables({0, 1, 2, 0});
   AssertCountLiveLogFiles(4);
-  PutRandomData(3, 91, 990);
+  PutRandomData(3, 93, 990);
   env_->SleepForMicroseconds(micros_wait_for_flush);
   AssertNumberOfImmutableMemtables({0, 1, 2, 1});
   AssertCountLiveLogFiles(5);
-  PutRandomData(3, 90, 990);
+  PutRandomData(3, 88, 990);
   env_->SleepForMicroseconds(micros_wait_for_flush);
   AssertNumberOfImmutableMemtables({0, 1, 2, 2});
   AssertCountLiveLogFiles(6);
-  PutRandomData(3, 90, 990);
+  PutRandomData(3, 88, 990);
   env_->SleepForMicroseconds(micros_wait_for_flush);
   AssertNumberOfImmutableMemtables({0, 1, 2, 3});
   AssertCountLiveLogFiles(7);
@@ -766,11 +767,11 @@ TEST_F(ColumnFamilyTest, DifferentWriteBufferSizes) {
   WaitForFlush(2);
   AssertNumberOfImmutableMemtables({0, 1, 0, 3});
   AssertCountLiveLogFiles(9);
-  PutRandomData(3, 90, 990);
+  PutRandomData(3, 88, 990);
   WaitForFlush(3);
   AssertNumberOfImmutableMemtables({0, 1, 0, 0});
   AssertCountLiveLogFiles(10);
-  PutRandomData(3, 90, 990);
+  PutRandomData(3, 88, 990);
   env_->SleepForMicroseconds(micros_wait_for_flush);
   AssertNumberOfImmutableMemtables({0, 1, 0, 1});
   AssertCountLiveLogFiles(11);
@@ -778,9 +779,9 @@ TEST_F(ColumnFamilyTest, DifferentWriteBufferSizes) {
   WaitForFlush(1);
   AssertNumberOfImmutableMemtables({0, 0, 0, 1});
   AssertCountLiveLogFiles(5);
-  PutRandomData(3, 90 * 3, 990);
+  PutRandomData(3, 88 * 3, 990);
   WaitForFlush(3);
-  PutRandomData(3, 90 * 4, 990);
+  PutRandomData(3, 88 * 4, 990);
   WaitForFlush(3);
   AssertNumberOfImmutableMemtables({0, 0, 0, 0});
   AssertCountLiveLogFiles(12);
@@ -1261,14 +1262,93 @@ TEST_F(ColumnFamilyTest, FlushAndDropRaceCondition) {
   rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
 
+#ifndef ROCKSDB_LITE
+// skipped as persisting options is not supported in ROCKSDB_LITE
+namespace {
+std::atomic<int> test_stage(0);
+const int kMainThreadStartPersistingOptionsFile = 1;
+const int kChildThreadFinishDroppingColumnFamily = 2;
+const int kChildThreadWaitingMainThreadPersistOptions = 3;
+void DropSingleColumnFamily(ColumnFamilyTest* cf_test, int cf_id,
+                            std::vector<Comparator*>* comparators) {
+  while (test_stage < kMainThreadStartPersistingOptionsFile) {
+    Env::Default()->SleepForMicroseconds(100);
+  }
+  cf_test->DropColumnFamilies({cf_id});
+  if ((*comparators)[cf_id]) {
+    delete (*comparators)[cf_id];
+    (*comparators)[cf_id] = nullptr;
+  }
+  test_stage = kChildThreadFinishDroppingColumnFamily;
+}
+}  // namespace
+
+TEST_F(ColumnFamilyTest, CreateAndDropRace) {
+  const int kCfCount = 5;
+  std::vector<ColumnFamilyOptions> cf_opts;
+  std::vector<Comparator*> comparators;
+  for (int i = 0; i < kCfCount; ++i) {
+    cf_opts.emplace_back();
+    comparators.push_back(new test::SimpleSuffixReverseComparator());
+    cf_opts.back().comparator = comparators.back();
+  }
+  db_options_.create_if_missing = true;
+  db_options_.create_missing_column_families = true;
+
+  auto main_thread_id = std::this_thread::get_id();
+
+  rocksdb::SyncPoint::GetInstance()->SetCallBack("PersistRocksDBOptions:start",
+                                                 [&](void* arg) {
+    auto current_thread_id = std::this_thread::get_id();
+    // If it's the main thread hitting this sync-point, then it
+    // will be blocked until some other thread update the test_stage.
+    if (main_thread_id == current_thread_id) {
+      test_stage = kMainThreadStartPersistingOptionsFile;
+      while (test_stage < kChildThreadFinishDroppingColumnFamily) {
+        Env::Default()->SleepForMicroseconds(100);
+      }
+    }
+  });
+
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "WriteThread::EnterUnbatched:Wait", [&](void* arg) {
+        // This means a thread doing DropColumnFamily() is waiting for
+        // other thread to finish persisting options.
+        // In such case, we update the test_stage to unblock the main thread.
+        test_stage = kChildThreadWaitingMainThreadPersistOptions;
+
+        // Note that based on the test setting, this must not be the
+        // main thread.
+        ASSERT_NE(main_thread_id, std::this_thread::get_id());
+      });
+
+  // Create a database with four column families
+  Open({"default", "one", "two", "three"},
+       {cf_opts[0], cf_opts[1], cf_opts[2], cf_opts[3]});
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // Start a thread that will drop the first column family
+  // and its comparator
+  std::thread drop_cf_thread(DropSingleColumnFamily, this, 1, &comparators);
+
+  DropColumnFamilies({2});
+
+  drop_cf_thread.join();
+  Close();
+  Destroy();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  for (auto* comparator : comparators) {
+    if (comparator) {
+      delete comparator;
+    }
+  }
+}
+#endif  // !ROCKSDB_LITE
+
 }  // namespace rocksdb
-#endif
 
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
-#else
-  return 0;
-#endif
 }

db/compact_files_test.cc

@@ -12,6 +12,7 @@
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "util/string_util.h"
+#include "util/sync_point.h"
 #include "util/testharness.h"
 
 namespace rocksdb {
@@ -53,6 +54,62 @@ class FlushedFileCollector : public EventListener {
   std::mutex mutex_;
 };
 
+TEST_F(CompactFilesTest, L0ConflictsFiles) {
+  Options options;
+  // to trigger compaction more easily
+  const int kWriteBufferSize = 10000;
+  const int kLevel0Trigger = 2;
+  options.create_if_missing = true;
+  options.compaction_style = kCompactionStyleLevel;
+  // Small slowdown and stop trigger for experimental purpose.
+  options.level0_slowdown_writes_trigger = 20;
+  options.level0_stop_writes_trigger = 20;
+  options.level0_stop_writes_trigger = 20;
+  options.write_buffer_size = kWriteBufferSize;
+  options.level0_file_num_compaction_trigger = kLevel0Trigger;
+  options.compression = kNoCompression;
+
+  DB* db = nullptr;
+  DestroyDB(db_name_, options);
+  Status s = DB::Open(options, db_name_, &db);
+  assert(s.ok());
+  assert(db);
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      {"CompactFilesImpl:0", "BackgroundCallCompaction:0"},
+      {"BackgroundCallCompaction:1", "CompactFilesImpl:1"},
+  });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // create couple files
+  // Background compaction starts and waits in BackgroundCallCompaction:0
+  for (int i = 0; i < kLevel0Trigger * 4; ++i) {
+    db->Put(WriteOptions(), ToString(i), "");
+    db->Put(WriteOptions(), ToString(100 - i), "");
+    db->Flush(FlushOptions());
+  }
+
+  rocksdb::ColumnFamilyMetaData meta;
+  db->GetColumnFamilyMetaData(&meta);
+  std::string file1;
+  for (auto& file : meta.levels[0].files) {
+    ASSERT_EQ(0, meta.levels[0].level);
+    if (file1 == "") {
+      file1 = file.db_path + "/" + file.name;
+    } else {
+      std::string file2 = file.db_path + "/" + file.name;
+      // Another thread starts a compact files and creates an L0 compaction
+      // The background compaction then notices that there is an L0 compaction
+      // already in progress and doesn't do an L0 compaction
+      // Once the background compaction finishes, the compact files finishes
+      ASSERT_OK(
+          db->CompactFiles(rocksdb::CompactionOptions(), {file1, file2}, 0));
+      break;
+    }
+  }
+  delete db;
+}
+
 TEST_F(CompactFilesTest, ObsoleteFiles) {
   Options options;
   // to trigger compaction more easily

db/compaction.cc

@@ -439,6 +439,7 @@ std::unique_ptr<CompactionFilter> Compaction::CreateCompactionFilter() const {
   CompactionFilter::Context context;
   context.is_full_compaction = is_full_compaction_;
   context.is_manual_compaction = is_manual_compaction_;
+  context.column_family_id = cfd_->GetID();
   return cfd_->ioptions()->compaction_filter_factory->CreateCompactionFilter(
       context);
 }

db/compaction.h

@@ -210,6 +210,14 @@ class Compaction {
       int output_level, VersionStorageInfo* vstorage,
       const std::vector<CompactionInputFiles>& inputs);
 
+  TablePropertiesCollection GetOutputTableProperties() const {
+    return output_table_properties_;
+  }
+
+  void SetOutputTableProperties(TablePropertiesCollection tp) {
+    output_table_properties_ = std::move(tp);
+  }
+
  private:
   // mark (or clear) all files that are being compacted
   void MarkFilesBeingCompacted(bool mark_as_compacted);
@@ -273,6 +281,9 @@ class Compaction {
 
   // Does input compression match the output compression?
   bool InputCompressionMatchesOutput() const;
+
+  // table properties of output files
+  TablePropertiesCollection output_table_properties_;
 };
 
 // Utility function

db/compaction_iterator.cc

@@ -6,11 +6,12 @@
 //  of patent rights can be found in the PATENTS file in the same directory.
 
 #include "db/compaction_iterator.h"
+#include "table/internal_iterator.h"
 
 namespace rocksdb {
 
 CompactionIterator::CompactionIterator(
-    Iterator* input, const Comparator* cmp, MergeHelper* merge_helper,
+    InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
     SequenceNumber last_sequence, std::vector<SequenceNumber>* snapshots,
     Env* env, bool expect_valid_internal_key, Compaction* compaction,
     const CompactionFilter* compaction_filter, LogBuffer* log_buffer)
@@ -41,6 +42,11 @@ CompactionIterator::CompactionIterator(
     earliest_snapshot_ = snapshots_->at(0);
     latest_snapshot_ = snapshots_->back();
   }
+  if (compaction_filter_ != nullptr && compaction_filter_->IgnoreSnapshots()) {
+    ignore_snapshots_ = true;
+  } else {
+    ignore_snapshots_ = false;
+  }
 }
 
 void CompactionIterator::ResetRecordCounts() {
@@ -139,7 +145,8 @@ void CompactionIterator::NextFromInput() {
       current_user_key_snapshot_ = 0;
       // apply the compaction filter to the first occurrence of the user key
       if (compaction_filter_ != nullptr && ikey_.type == kTypeValue &&
-          (visible_at_tip_ || ikey_.sequence > latest_snapshot_)) {
+          (visible_at_tip_ || ikey_.sequence > latest_snapshot_ ||
+           ignore_snapshots_)) {
         // If the user has specified a compaction filter and the sequence
         // number is greater than any external snapshot, then invoke the
         // filter. If the return value of the compaction filter is true,
@@ -169,6 +176,9 @@ void CompactionIterator::NextFromInput() {
     } else {
       // Update the current key to reflect the new sequence number/type without
       // copying the user key.
+      // TODO(rven): Compaction filter does not process keys in this path
+      // Need to have the compaction filter process multiple versions
+      // if we have versions on both sides of a snapshot
       current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
       key_ = current_key_.GetKey();
       ikey_.user_key = current_key_.GetUserKey();

db/compaction_iterator.h

@@ -37,7 +37,7 @@ struct CompactionIteratorStats {
 
 class CompactionIterator {
  public:
-  CompactionIterator(Iterator* input, const Comparator* cmp,
+  CompactionIterator(InternalIterator* input, const Comparator* cmp,
                      MergeHelper* merge_helper, SequenceNumber last_sequence,
                      std::vector<SequenceNumber>* snapshots, Env* env,
                      bool expect_valid_internal_key,
@@ -84,7 +84,7 @@ class CompactionIterator {
   inline SequenceNumber findEarliestVisibleSnapshot(
       SequenceNumber in, SequenceNumber* prev_snapshot);
 
-  Iterator* input_;
+  InternalIterator* input_;
   const Comparator* cmp_;
   MergeHelper* merge_helper_;
   const std::vector<SequenceNumber>* snapshots_;
@@ -98,6 +98,7 @@ class CompactionIterator {
   SequenceNumber visible_at_tip_;
   SequenceNumber earliest_snapshot_;
   SequenceNumber latest_snapshot_;
+  bool ignore_snapshots_;
 
   // State
   //

db/compaction_job.cc

@@ -77,6 +77,7 @@ struct CompactionJob::SubcompactionState {
   struct Output {
     FileMetaData meta;
     bool finished;
+    std::shared_ptr<const TableProperties> table_properties;
   };
 
   // State kept for output being generated
@@ -487,6 +488,16 @@ Status CompactionJob::Run() {
     }
   }
 
+  TablePropertiesCollection tp;
+  for (const auto& state : compact_->sub_compact_states) {
+    for (const auto& output : state.outputs) {
+      auto fn = TableFileName(db_options_.db_paths, output.meta.fd.GetNumber(),
+                              output.meta.fd.GetPathId());
+      tp[fn] = output.table_properties;
+    }
+  }
+  compact_->compaction->SetOutputTableProperties(std::move(tp));
+
   // Finish up all book-keeping to unify the subcompaction results
   AggregateStatistics();
   UpdateCompactionStats();
@@ -574,7 +585,7 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options,
 
 void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   assert(sub_compact != nullptr);
-  std::unique_ptr<Iterator> input(
+  std::unique_ptr<InternalIterator> input(
       versions_->MakeInputIterator(sub_compact->compaction));
 
   AutoThreadOperationStageUpdater stage_updater(
@@ -590,10 +601,10 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   if (measure_io_stats_) {
     prev_perf_level = GetPerfLevel();
     SetPerfLevel(PerfLevel::kEnableTime);
-    prev_write_nanos = iostats_context.write_nanos;
-    prev_fsync_nanos = iostats_context.fsync_nanos;
-    prev_range_sync_nanos = iostats_context.range_sync_nanos;
-    prev_prepare_write_nanos = iostats_context.prepare_write_nanos;
+    prev_write_nanos = IOSTATS(write_nanos);
+    prev_fsync_nanos = IOSTATS(fsync_nanos);
+    prev_range_sync_nanos = IOSTATS(range_sync_nanos);
+    prev_prepare_write_nanos = IOSTATS(prepare_write_nanos);
   }
 
   ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
@@ -717,13 +728,13 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
 
   if (measure_io_stats_) {
     sub_compact->compaction_job_stats.file_write_nanos +=
-        iostats_context.write_nanos - prev_write_nanos;
+        IOSTATS(write_nanos) - prev_write_nanos;
     sub_compact->compaction_job_stats.file_fsync_nanos +=
-        iostats_context.fsync_nanos - prev_fsync_nanos;
+        IOSTATS(fsync_nanos) - prev_fsync_nanos;
     sub_compact->compaction_job_stats.file_range_sync_nanos +=
-        iostats_context.range_sync_nanos - prev_range_sync_nanos;
+        IOSTATS(range_sync_nanos) - prev_range_sync_nanos;
     sub_compact->compaction_job_stats.file_prepare_write_nanos +=
-        iostats_context.prepare_write_nanos - prev_prepare_write_nanos;
+        IOSTATS(prepare_write_nanos) - prev_prepare_write_nanos;
     if (prev_perf_level != PerfLevel::kEnableTime) {
       SetPerfLevel(prev_perf_level);
     }
@@ -800,7 +811,7 @@ Status CompactionJob::FinishCompactionOutputFile(
   if (s.ok() && current_entries > 0) {
     // Verify that the table is usable
     ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
-    Iterator* iter = cfd->table_cache()->NewIterator(
+    InternalIterator* iter = cfd->table_cache()->NewIterator(
         ReadOptions(), env_options_, cfd->internal_comparator(), meta->fd,
         nullptr, cfd->internal_stats()->GetFileReadHist(
                      compact_->compaction->output_level()),
@@ -814,7 +825,10 @@ Status CompactionJob::FinishCompactionOutputFile(
 
     delete iter;
     if (s.ok()) {
-      TableFileCreationInfo info(sub_compact->builder->GetTableProperties());
+      auto tp = sub_compact->builder->GetTableProperties();
+      sub_compact->current_output()->table_properties =
+          std::make_shared<TableProperties>(tp);
+      TableFileCreationInfo info(std::move(tp));
       info.db_name = dbname_;
       info.cf_name = cfd->GetName();
       info.file_path =
@@ -897,7 +911,7 @@ Status CompactionJob::OpenCompactionOutputFile(
   unique_ptr<WritableFile> writable_file;
   std::string fname = TableFileName(db_options_.db_paths, file_number,
                                     sub_compact->compaction->output_path_id());
-  Status s = env_->NewWritableFile(fname, &writable_file, env_options_);
+  Status s = NewWritableFile(env_, fname, &writable_file, env_options_);
   if (!s.ok()) {
     Log(InfoLogLevel::ERROR_LEVEL, db_options_.info_log,
         "[%s] [JOB %d] OpenCompactionOutputFiles for table #%" PRIu64
@@ -927,8 +941,8 @@ Status CompactionJob::OpenCompactionOutputFile(
       cfd->ioptions()->optimize_filters_for_hits && bottommost_level_;
   sub_compact->builder.reset(NewTableBuilder(
       *cfd->ioptions(), cfd->internal_comparator(),
-      cfd->int_tbl_prop_collector_factories(), sub_compact->outfile.get(),
-      sub_compact->compaction->output_compression(),
+      cfd->int_tbl_prop_collector_factories(), cfd->GetID(),
+      sub_compact->outfile.get(), sub_compact->compaction->output_compression(),
       cfd->ioptions()->compression_opts, skip_filters));
   LogFlush(db_options_.info_log);
   return s;

db/compaction_job.h

@@ -35,9 +35,9 @@
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/transaction_log.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/autovector.h"
 #include "util/event_logger.h"
-#include "util/scoped_arena_iterator.h"
 #include "util/stop_watch.h"
 #include "util/thread_local.h"
 

db/compaction_job_stats_test.cc

@@ -27,6 +27,7 @@
 #include "db/job_context.h"
 #include "db/version_set.h"
 #include "db/write_batch_internal.h"
+#include "memtable/hash_linklist_rep.h"
 #include "port/stack_trace.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/compaction_filter.h"
@@ -47,14 +48,13 @@
 #include "table/block_based_table_factory.h"
 #include "table/mock_table.h"
 #include "table/plain_table_factory.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/compression.h"
 #include "util/hash.h"
-#include "util/hash_linklist_rep.h"
 #include "util/logging.h"
 #include "util/mock_env.h"
 #include "util/mutexlock.h"
 #include "util/rate_limiter.h"
-#include "util/scoped_arena_iterator.h"
 #include "util/statistics.h"
 #include "util/string_util.h"
 #include "util/sync_point.h"
@@ -64,7 +64,7 @@
 #include "util/xfunc.h"
 #include "utilities/merge_operators.h"
 
-#if !defined(IOS_CROSS_COMPILE) && (!defined(NDEBUG) || !defined(OS_WIN))
+#if !defined(IOS_CROSS_COMPILE)
 #ifndef ROCKSDB_LITE
 namespace rocksdb {
 

db/compaction_job_test.cc

@@ -3,6 +3,8 @@
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
 
+#ifndef ROCKSDB_LITE
+
 #include <algorithm>
 #include <map>
 #include <string>
@@ -196,7 +198,7 @@ class CompactionJobTest : public testing::Test {
     unique_ptr<WritableFileWriter> file_writer(
         new WritableFileWriter(std::move(file), env_options_));
     {
-      log::Writer log(std::move(file_writer));
+      log::Writer log(std::move(file_writer), 0, false);
       std::string record;
       new_db.EncodeTo(&record);
       s = log.AddRecord(record);
@@ -695,3 +697,14 @@ int main(int argc, char** argv) {
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
 }
+
+#else
+#include <stdio.h>
+
+int main(int argc, char** argv) {
+  fprintf(stderr,
+          "SKIPPED as CompactionJobStats is not supported in ROCKSDB_LITE\n");
+  return 0;
+}
+
+#endif  // ROCKSDB_LITE

db/compaction_picker.cc

@@ -255,16 +255,30 @@ Compaction* CompactionPicker::FormCompaction(
     const CompactionOptions& compact_options,
     const std::vector<CompactionInputFiles>& input_files, int output_level,
     VersionStorageInfo* vstorage, const MutableCFOptions& mutable_cf_options,
-    uint32_t output_path_id) const {
+    uint32_t output_path_id) {
   uint64_t max_grandparent_overlap_bytes =
       output_level + 1 < vstorage->num_levels() ?
           mutable_cf_options.MaxGrandParentOverlapBytes(output_level + 1) :
           std::numeric_limits<uint64_t>::max();
   assert(input_files.size());
-  return new Compaction(
+
+  // TODO(rven ): we might be able to run concurrent level 0 compaction
+  // if the key ranges of the two compactions do not overlap, but for now
+  // we do not allow it.
+  if ((input_files[0].level == 0) && !level0_compactions_in_progress_.empty()) {
+    return nullptr;
+  }
+  auto c = new Compaction(
       vstorage, mutable_cf_options, input_files, output_level,
       compact_options.output_file_size_limit, max_grandparent_overlap_bytes,
       output_path_id, compact_options.compression, /* grandparents */ {}, true);
+
+  // If it's level 0 compaction, make sure we don't execute any other level 0
+  // compactions in parallel
+  if ((c != nullptr) && (input_files[0].level == 0)) {
+    level0_compactions_in_progress_.insert(c);
+  }
+  return c;
 }
 
 Status CompactionPicker::GetCompactionInputsFromFileNumbers(
@@ -1234,8 +1248,9 @@ Compaction* UniversalCompactionPicker::PickCompaction(
   std::vector<SortedRun> sorted_runs =
       CalculateSortedRuns(*vstorage, ioptions_);
 
-  if (sorted_runs.size() <
-      (unsigned int)mutable_cf_options.level0_file_num_compaction_trigger) {
+  if (sorted_runs.size() == 0 ||
+      sorted_runs.size() <
+          (unsigned int)mutable_cf_options.level0_file_num_compaction_trigger) {
     LogToBuffer(log_buffer, "[%s] Universal: nothing to do\n", cf_name.c_str());
     return nullptr;
   }

db/compaction_picker.h

@@ -94,7 +94,7 @@ class CompactionPicker {
       const CompactionOptions& compact_options,
       const std::vector<CompactionInputFiles>& input_files, int output_level,
       VersionStorageInfo* vstorage, const MutableCFOptions& mutable_cf_options,
-      uint32_t output_path_id) const;
+      uint32_t output_path_id);
 
   // Converts a set of compaction input file numbers into
   // a list of CompactionInputFiles.

db/comparator_db_test.cc

@@ -7,10 +7,11 @@
 #include <map>
 #include <string>
 
+#include "memtable/stl_wrappers.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "util/hash.h"
-#include "util/stl_wrappers.h"
+#include "util/kv_map.h"
 #include "util/string_util.h"
 #include "util/testharness.h"
 #include "util/testutil.h"

db/corruption_test.cc

@@ -232,8 +232,16 @@ class CorruptionTest : public testing::Test {
 
   // Return the value to associate with the specified key
   Slice Value(int k, std::string* storage) {
-    Random r(k);
-    return test::RandomString(&r, kValueSize, storage);
+    if (k == 0) {
+      // Ugh.  Random seed of 0 used to produce no entropy.  This code
+      // preserves the implementation that was in place when all of the
+      // magic values in this file were picked.
+      *storage = std::string(kValueSize, ' ');
+      return Slice(*storage);
+    } else {
+      Random r(k);
+      return test::RandomString(&r, kValueSize, storage);
+    }
   }
 };
 

db/db_bench.cc

@@ -315,7 +315,7 @@ DEFINE_int32(compaction_style, (int32_t) rocksdb::Options().compaction_style,
              "style of compaction: level-based vs universal");
 
 static rocksdb::CompactionPri FLAGS_compaction_pri_e;
-DEFINE_int32(compaction_pri, (int32_t)rocksdb::Options().compaction_style,
+DEFINE_int32(compaction_pri, (int32_t)rocksdb::Options().compaction_pri,
              "priority of files to compaction: by size or by data age");
 
 DEFINE_int32(universal_size_ratio, 0,
@@ -373,6 +373,15 @@ DEFINE_int32(new_table_reader_for_compaction_inputs, true,
 
 DEFINE_int32(compaction_readahead_size, 0, "Compaction readahead size");
 
+DEFINE_int32(random_access_max_buffer_size, 1024 * 1024,
+             "Maximum windows randomaccess buffer size");
+
+DEFINE_int32(writable_file_max_buffer_size, 1024 * 1024,
+             "Maximum write buffer for Writable File");
+
+DEFINE_int32(skip_table_builder_flush, false, "Skip flushing block in "
+             "table builder ");
+
 DEFINE_int32(bloom_bits, -1, "Bloom filter bits per key. Negative means"
              " use default settings.");
 DEFINE_int32(memtable_bloom_bits, 0, "Bloom filter bits per key for memtable. "
@@ -486,6 +495,7 @@ DEFINE_int32(deletepercent, 2, "Percentage of deletes out of reads/writes/"
 DEFINE_uint64(delete_obsolete_files_period_micros, 0,
               "Ignored. Left here for backward compatibility");
 
+#ifndef ROCKSDB_LITE
 DEFINE_bool(optimistic_transaction_db, false,
             "Open a OptimisticTransactionDB instance. "
             "Required for randomtransaction benchmark.");
@@ -509,6 +519,7 @@ DEFINE_int32(transaction_sleep, 0,
 DEFINE_uint64(transaction_lock_timeout, 100,
               "If using a transaction_db, specifies the lock wait timeout in"
               " milliseconds before failing a transaction waiting on a lock");
+#endif  // ROCKSDB_LITE
 
 DEFINE_bool(compaction_measure_io_stats, false,
             "Measure times spents on I/Os while in compactions. ");
@@ -624,7 +635,7 @@ DEFINE_double(soft_rate_limit, 0.0, "");
 
 DEFINE_double(hard_rate_limit, 0.0, "DEPRECATED");
 
-DEFINE_uint64(hard_pending_compaction_bytes_limit, 128u * 1024 * 1024 * 1024,
+DEFINE_uint64(hard_pending_compaction_bytes_limit, 128ull * 1024 * 1024 * 1024,
               "Stop writes if pending compaction bytes exceed this number");
 
 DEFINE_uint64(delayed_write_rate, 2097152u,
@@ -645,7 +656,9 @@ DEFINE_int32(max_grandparent_overlap_factor, 10, "Control maximum bytes of "
              "overlaps in grandparent (i.e., level+2) before we stop building a"
              " single file in a level->level+1 compaction.");
 
+#ifndef ROCKSDB_LITE
 DEFINE_bool(readonly, false, "Run read only benchmarks.");
+#endif  // ROCKSDB_LITE
 
 DEFINE_bool(disable_auto_compactions, false, "Do not auto trigger compactions");
 
@@ -983,7 +996,9 @@ static void AppendWithSpace(std::string* str, Slice msg) {
 struct DBWithColumnFamilies {
   std::vector<ColumnFamilyHandle*> cfh;
   DB* db;
+#ifndef ROCKSDB_LITE
   OptimisticTransactionDB* opt_txn_db;
+#endif  // ROCKSDB_LITE
   std::atomic<size_t> num_created;  // Need to be updated after all the
                                     // new entries in cfh are set.
   size_t num_hot;  // Number of column families to be queried at each moment.
@@ -991,7 +1006,12 @@ struct DBWithColumnFamilies {
                    // Column families will be created and used to be queried.
   port::Mutex create_cf_mutex;  // Only one thread can execute CreateNewCf()
 
-  DBWithColumnFamilies() : db(nullptr), opt_txn_db(nullptr) {
+  DBWithColumnFamilies()
+      : db(nullptr)
+#ifndef ROCKSDB_LITE
+        , opt_txn_db(nullptr)
+#endif  // ROCKSDB_LITE
+  {
     cfh.clear();
     num_created = 0;
     num_hot = 0;
@@ -1000,7 +1020,9 @@ struct DBWithColumnFamilies {
   DBWithColumnFamilies(const DBWithColumnFamilies& other)
       : cfh(other.cfh),
         db(other.db),
+#ifndef ROCKSDB_LITE
         opt_txn_db(other.opt_txn_db),
+#endif  // ROCKSDB_LITE
         num_created(other.num_created.load()),
         num_hot(other.num_hot) {}
 
@@ -1008,13 +1030,18 @@ struct DBWithColumnFamilies {
     std::for_each(cfh.begin(), cfh.end(),
                   [](ColumnFamilyHandle* cfhi) { delete cfhi; });
     cfh.clear();
+#ifndef ROCKSDB_LITE
     if (opt_txn_db) {
       delete opt_txn_db;
       opt_txn_db = nullptr;
     } else {
       delete db;
+      db = nullptr;
     }
+#else
+    delete db;
     db = nullptr;
+#endif  // ROCKSDB_LITE
   }
 
   ColumnFamilyHandle* GetCfh(int64_t rand_num) {
@@ -1940,9 +1967,11 @@ class Benchmark {
         method = &Benchmark::Compress;
       } else if (name == "uncompress") {
         method = &Benchmark::Uncompress;
+#ifndef ROCKSDB_LITE
       } else if (name == "randomtransaction") {
         method = &Benchmark::RandomTransaction;
         post_process_method = &Benchmark::RandomTransactionVerify;
+#endif  // ROCKSDB_LITE
       } else if (name == "randomreplacekeys") {
         fresh_db = true;
         method = &Benchmark::RandomReplaceKeys;
@@ -2275,6 +2304,8 @@ class Benchmark {
     options.new_table_reader_for_compaction_inputs =
         FLAGS_new_table_reader_for_compaction_inputs;
     options.compaction_readahead_size = FLAGS_compaction_readahead_size;
+    options.random_access_max_buffer_size = FLAGS_random_access_max_buffer_size;
+    options.writable_file_max_buffer_size = FLAGS_writable_file_max_buffer_size;
     options.statistics = dbstats;
     if (FLAGS_enable_io_prio) {
       FLAGS_env->LowerThreadPoolIOPriority(Env::LOW);
@@ -2417,6 +2448,8 @@ class Benchmark {
       block_based_options.block_size = FLAGS_block_size;
       block_based_options.block_restart_interval = FLAGS_block_restart_interval;
       block_based_options.filter_policy = filter_policy_;
+      block_based_options.skip_table_builder_flush =
+          FLAGS_skip_table_builder_flush;
       block_based_options.format_version = 2;
       options.table_factory.reset(
           NewBlockBasedTableFactory(block_based_options));
@@ -2518,10 +2551,12 @@ class Benchmark {
           NewGenericRateLimiter(FLAGS_rate_limiter_bytes_per_sec));
     }
 
+#ifndef ROCKSDB_LITE
     if (FLAGS_readonly && FLAGS_transaction_db) {
       fprintf(stderr, "Cannot use readonly flag with transaction_db\n");
       exit(1);
     }
+#endif  // ROCKSDB_LITE
 
     if (FLAGS_num_multi_db <= 1) {
       OpenDb(options, FLAGS_db, &db_);
@@ -2554,6 +2589,7 @@ class Benchmark {
         column_families.push_back(ColumnFamilyDescriptor(
               ColumnFamilyName(i), ColumnFamilyOptions(options)));
       }
+#ifndef ROCKSDB_LITE
       if (FLAGS_readonly) {
         s = DB::OpenForReadOnly(options, db_name, column_families,
             &db->cfh, &db->db);
@@ -2574,9 +2610,13 @@ class Benchmark {
       } else {
         s = DB::Open(options, db_name, column_families, &db->cfh, &db->db);
       }
+#else
+      s = DB::Open(options, db_name, column_families, &db->cfh, &db->db);
+#endif  // ROCKSDB_LITE
       db->cfh.resize(FLAGS_num_column_families);
       db->num_created = num_hot;
       db->num_hot = num_hot;
+#ifndef ROCKSDB_LITE
     } else if (FLAGS_readonly) {
       s = DB::OpenForReadOnly(options, db_name, &db->db);
     } else if (FLAGS_optimistic_transaction_db) {
@@ -2591,7 +2631,7 @@ class Benchmark {
       if (s.ok()) {
         db->db = ptr;
       }
-
+#endif  // ROCKSDB_LITE
     } else {
       s = DB::Open(options, db_name, &db->db);
     }
@@ -3636,6 +3676,7 @@ class Benchmark {
     }
   }
 
+#ifndef ROCKSDB_LITE
   // This benchmark stress tests Transactions.  For a given --duration (or
   // total number of --writes, a Transaction will perform a read-modify-write
   // to increment the value of a key in each of N(--transaction-sets) sets of
@@ -3868,6 +3909,7 @@ class Benchmark {
 
     fprintf(stdout, "RandomTransactionVerify Success!\n");
   }
+#endif  // ROCKSDB_LITE
 
   // Writes and deletes random keys without overwriting keys.
   //

db/db_compaction_filter_test.cc

@@ -7,8 +7,8 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
 
 namespace rocksdb {
 
@@ -47,6 +47,24 @@ class DeleteFilter : public CompactionFilter {
   virtual const char* Name() const override { return "DeleteFilter"; }
 };
 
+class DeleteISFilter : public CompactionFilter {
+ public:
+  virtual bool Filter(int level, const Slice& key, const Slice& value,
+                      std::string* new_value,
+                      bool* value_changed) const override {
+    cfilter_count++;
+    int i = std::stoi(key.ToString());
+    if (i > 5 && i <= 105) {
+      return true;
+    }
+    return false;
+  }
+
+  virtual bool IgnoreSnapshots() const override { return true; }
+
+  virtual const char* Name() const override { return "DeleteFilter"; }
+};
+
 class DelayFilter : public CompactionFilter {
  public:
   explicit DelayFilter(DBTestBase* d) : db_test(d) {}
@@ -97,8 +115,11 @@ class ChangeFilter : public CompactionFilter {
 
 class KeepFilterFactory : public CompactionFilterFactory {
  public:
-  explicit KeepFilterFactory(bool check_context = false)
-      : check_context_(check_context) {}
+  explicit KeepFilterFactory(bool check_context = false,
+                             bool check_context_cf_id = false)
+      : check_context_(check_context),
+        check_context_cf_id_(check_context_cf_id),
+        compaction_filter_created_(false) {}
 
   virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
       const CompactionFilter::Context& context) override {
@@ -106,13 +127,22 @@ class KeepFilterFactory : public CompactionFilterFactory {
       EXPECT_EQ(expect_full_compaction_.load(), context.is_full_compaction);
       EXPECT_EQ(expect_manual_compaction_.load(), context.is_manual_compaction);
     }
+    if (check_context_cf_id_) {
+      EXPECT_EQ(expect_cf_id_.load(), context.column_family_id);
+    }
+    compaction_filter_created_ = true;
     return std::unique_ptr<CompactionFilter>(new KeepFilter());
   }
 
+  bool compaction_filter_created() const { return compaction_filter_created_; }
+
   virtual const char* Name() const override { return "KeepFilterFactory"; }
   bool check_context_;
+  bool check_context_cf_id_;
   std::atomic_bool expect_full_compaction_;
   std::atomic_bool expect_manual_compaction_;
+  std::atomic<uint32_t> expect_cf_id_;
+  bool compaction_filter_created_;
 };
 
 class DeleteFilterFactory : public CompactionFilterFactory {
@@ -129,6 +159,21 @@ class DeleteFilterFactory : public CompactionFilterFactory {
   virtual const char* Name() const override { return "DeleteFilterFactory"; }
 };
 
+// Delete Filter Factory which ignores snapshots
+class DeleteISFilterFactory : public CompactionFilterFactory {
+ public:
+  virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
+      const CompactionFilter::Context& context) override {
+    if (context.is_manual_compaction) {
+      return std::unique_ptr<CompactionFilter>(new DeleteISFilter());
+    } else {
+      return std::unique_ptr<CompactionFilter>(nullptr);
+    }
+  }
+
+  virtual const char* Name() const override { return "DeleteFilterFactory"; }
+};
+
 class DelayFilterFactory : public CompactionFilterFactory {
  public:
   explicit DelayFilterFactory(DBTestBase* d) : db_test(d) {}
@@ -174,6 +219,7 @@ class ChangeFilterFactory : public CompactionFilterFactory {
   virtual const char* Name() const override { return "ChangeFilterFactory"; }
 };
 
+#ifndef ROCKSDB_LITE
 TEST_F(DBTestCompactionFilter, CompactionFilter) {
   Options options = CurrentOptions();
   options.max_open_files = -1;
@@ -216,7 +262,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilter) {
   Arena arena;
   {
     ScopedArenaIterator iter(
-        dbfull()->TEST_NewInternalIterator(&arena, handles_[1]));
+        dbfull()->NewInternalIterator(&arena, handles_[1]));
     iter->SeekToFirst();
     ASSERT_OK(iter->status());
     while (iter->Valid()) {
@@ -304,7 +350,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilter) {
   count = 0;
   {
     ScopedArenaIterator iter(
-        dbfull()->TEST_NewInternalIterator(&arena, handles_[1]));
+        dbfull()->NewInternalIterator(&arena, handles_[1]));
     iter->SeekToFirst();
     ASSERT_OK(iter->status());
     while (iter->Valid()) {
@@ -350,6 +396,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilterDeletesAll) {
 
   delete itr;
 }
+#endif  // ROCKSDB_LITE
 
 TEST_F(DBTestCompactionFilter, CompactionFilterWithValueChange) {
   do {
@@ -481,8 +528,9 @@ TEST_F(DBTestCompactionFilter, CompactionFilterWithMergeOperator) {
   ASSERT_EQ(newvalue, four);
 }
 
+#ifndef ROCKSDB_LITE
 TEST_F(DBTestCompactionFilter, CompactionFilterContextManual) {
-  KeepFilterFactory* filter = new KeepFilterFactory();
+  KeepFilterFactory* filter = new KeepFilterFactory(true, true);
 
   Options options = CurrentOptions();
   options.compaction_style = kCompactionStyleUniversal;
@@ -504,22 +552,24 @@ TEST_F(DBTestCompactionFilter, CompactionFilterContextManual) {
     // be triggered.
     num_keys_per_file /= 2;
   }
+  dbfull()->TEST_WaitForCompact();
 
   // Force a manual compaction
   cfilter_count = 0;
   filter->expect_manual_compaction_.store(true);
-  filter->expect_full_compaction_.store(false);  // Manual compaction always
-                                                 // set this flag.
+  filter->expect_full_compaction_.store(true);
+  filter->expect_cf_id_.store(0);
   dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr);
   ASSERT_EQ(cfilter_count, 700);
   ASSERT_EQ(NumSortedRuns(0), 1);
+  ASSERT_TRUE(filter->compaction_filter_created());
 
   // Verify total number of keys is correct after manual compaction.
   {
     int count = 0;
     int total = 0;
     Arena arena;
-    ScopedArenaIterator iter(dbfull()->TEST_NewInternalIterator(&arena));
+    ScopedArenaIterator iter(dbfull()->NewInternalIterator(&arena));
     iter->SeekToFirst();
     ASSERT_OK(iter->status());
     while (iter->Valid()) {
@@ -536,7 +586,38 @@ TEST_F(DBTestCompactionFilter, CompactionFilterContextManual) {
     ASSERT_EQ(count, 1);
   }
 }
+#endif  // ROCKSDB_LITE
 
+TEST_F(DBTestCompactionFilter, CompactionFilterContextCfId) {
+  KeepFilterFactory* filter = new KeepFilterFactory(false, true);
+  filter->expect_cf_id_.store(1);
+
+  Options options = CurrentOptions();
+  options.compaction_filter_factory.reset(filter);
+  options.compression = kNoCompression;
+  options.level0_file_num_compaction_trigger = 2;
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  int num_keys_per_file = 400;
+  for (int j = 0; j < 3; j++) {
+    // Write several keys.
+    const std::string value(10, 'x');
+    for (int i = 0; i < num_keys_per_file; i++) {
+      char key[100];
+      snprintf(key, sizeof(key), "B%08d%02d", i, j);
+      Put(1, key, value);
+    }
+    Flush(1);
+    // Make sure next file is much smaller so automatic compaction will not
+    // be triggered.
+    num_keys_per_file /= 2;
+  }
+  dbfull()->TEST_WaitForCompact();
+
+  ASSERT_TRUE(filter->compaction_filter_created());
+}
+
+#ifndef ROCKSDB_LITE
 // Compaction filters should only be applied to records that are newer than the
 // latest snapshot. This test inserts records and applies a delete filter.
 TEST_F(DBTestCompactionFilter, CompactionFilterSnapshot) {
@@ -573,14 +654,73 @@ TEST_F(DBTestCompactionFilter, CompactionFilterSnapshot) {
   ASSERT_EQ(0U, CountLiveFiles());
 }
 
+// Compaction filters should only be applied to records that are newer than the
+// latest snapshot. However, if the compaction filter asks to ignore snapshots
+// records newer than the snapshot will also be processed
+TEST_F(DBTestCompactionFilter, CompactionFilterIgnoreSnapshot) {
+  std::string five = ToString(5);
+  Options options;
+  options.compaction_filter_factory = std::make_shared<DeleteISFilterFactory>();
+  options.disable_auto_compactions = true;
+  options.create_if_missing = true;
+  options = CurrentOptions(options);
+  DestroyAndReopen(options);
+
+  // Put some data.
+  const Snapshot* snapshot = nullptr;
+  for (int table = 0; table < 4; ++table) {
+    for (int i = 0; i < 10; ++i) {
+      Put(ToString(table * 100 + i), "val");
+    }
+    Flush();
+
+    if (table == 0) {
+      snapshot = db_->GetSnapshot();
+    }
+  }
+  assert(snapshot != nullptr);
+
+  cfilter_count = 0;
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  // The filter should delete 40 records.
+  ASSERT_EQ(40U, cfilter_count);
+
+  {
+    // Scan the entire database as of the snapshot to ensure
+    // that nothing is left
+    ReadOptions read_options;
+    read_options.snapshot = snapshot;
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    iter->SeekToFirst();
+    int count = 0;
+    while (iter->Valid()) {
+      count++;
+      iter->Next();
+    }
+    ASSERT_EQ(count, 6);
+    read_options.snapshot = 0;
+    std::unique_ptr<Iterator> iter1(db_->NewIterator(read_options));
+    iter1->SeekToFirst();
+    count = 0;
+    while (iter1->Valid()) {
+      count++;
+      iter1->Next();
+    }
+    // We have deleted 10 keys from 40 using the compaction filter
+    //  Keys 6-9 before the snapshot and 100-105 after the snapshot
+    ASSERT_EQ(count, 30);
+  }
+
+  // Release the snapshot and compact again -> now all records should be
+  // removed.
+  db_->ReleaseSnapshot(snapshot);
+}
+#endif  // ROCKSDB_LITE
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
   rocksdb::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
-#else
-  return 0;
-#endif
 }

db/db_compaction_test.cc

@@ -7,15 +7,14 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/experimental.h"
-#include "util/db_test_util.h"
 #include "util/sync_point.h"
 namespace rocksdb {
 
 // SYNC_POINT is not supported in released Windows mode.
-#if !(defined NDEBUG) || !defined(OS_WIN)
-
+#if !defined(ROCKSDB_LITE)
 
 class DBCompactionTest : public DBTestBase {
  public:
@@ -240,9 +239,12 @@ TEST_F(DBCompactionTest, SkipStatsUpdateTest) {
   env_->random_file_open_counter_.store(0);
   Reopen(options);
 
-  // As stats-update is disabled, we expect a very low
-  // number of random file open.
-  ASSERT_LT(env_->random_file_open_counter_.load(), 5);
+  // As stats-update is disabled, we expect a very low number of
+  // random file open.
+  // Note that this number must be changed accordingly if we change
+  // the number of files needed to be opened in the DB::Open process.
+  const int kMaxFileOpenCount = 10;
+  ASSERT_LT(env_->random_file_open_counter_.load(), kMaxFileOpenCount);
 
   // Repeat the reopen process, but this time we enable
   // stats-update.
@@ -252,7 +254,7 @@ TEST_F(DBCompactionTest, SkipStatsUpdateTest) {
 
   // Since we do a normal stats update on db-open, there
   // will be more random open files.
-  ASSERT_GT(env_->random_file_open_counter_.load(), 5);
+  ASSERT_GT(env_->random_file_open_counter_.load(), kMaxFileOpenCount);
 }
 
 TEST_F(DBCompactionTest, TestTableReaderForCompaction) {
@@ -459,12 +461,15 @@ TEST_F(DBCompactionTest, DisableStatsUpdateReopen) {
 
 
 TEST_P(DBCompactionTestWithParam, CompactionTrigger) {
+  const int kNumKeysPerFile = 100;
+
   Options options;
   options.write_buffer_size = 110 << 10;  // 110KB
   options.arena_block_size = 4 << 10;
   options.num_levels = 3;
   options.level0_file_num_compaction_trigger = 3;
   options.max_subcompactions = max_subcompactions_;
+  options.memtable_factory.reset(new SpecialSkipListFactory(kNumKeysPerFile));
   options = CurrentOptions(options);
   CreateAndReopenWithCF({"pikachu"}, options);
 
@@ -474,20 +479,24 @@ TEST_P(DBCompactionTestWithParam, CompactionTrigger) {
        num++) {
     std::vector<std::string> values;
     // Write 100KB (100 values, each 1K)
-    for (int i = 0; i < 100; i++) {
+    for (int i = 0; i < kNumKeysPerFile; i++) {
       values.push_back(RandomString(&rnd, 990));
       ASSERT_OK(Put(1, Key(i), values[i]));
     }
+    // put extra key to trigger flush
+    ASSERT_OK(Put(1, "", ""));
     dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
     ASSERT_EQ(NumTableFilesAtLevel(0, 1), num + 1);
   }
 
   // generate one more file in level-0, and should trigger level-0 compaction
   std::vector<std::string> values;
-  for (int i = 0; i < 100; i++) {
+  for (int i = 0; i < kNumKeysPerFile; i++) {
     values.push_back(RandomString(&rnd, 990));
     ASSERT_OK(Put(1, Key(i), values[i]));
   }
+  // put extra key to trigger flush
+  ASSERT_OK(Put(1, "", ""));
   dbfull()->TEST_WaitForCompact();
 
   ASSERT_EQ(NumTableFilesAtLevel(0, 1), 0);
@@ -847,6 +856,8 @@ TEST_P(DBCompactionTestWithParam, LevelCompactionThirdPath) {
   options.db_paths.emplace_back(dbname_, 500 * 1024);
   options.db_paths.emplace_back(dbname_ + "_2", 4 * 1024 * 1024);
   options.db_paths.emplace_back(dbname_ + "_3", 1024 * 1024 * 1024);
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));
   options.compaction_style = kCompactionStyleLevel;
   options.write_buffer_size = 110 << 10;  // 110KB
   options.arena_block_size = 4 << 10;
@@ -962,6 +973,8 @@ TEST_P(DBCompactionTestWithParam, LevelCompactionPathUse) {
   options.db_paths.emplace_back(dbname_, 500 * 1024);
   options.db_paths.emplace_back(dbname_ + "_2", 4 * 1024 * 1024);
   options.db_paths.emplace_back(dbname_ + "_3", 1024 * 1024 * 1024);
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));
   options.compaction_style = kCompactionStyleLevel;
   options.write_buffer_size = 110 << 10;  // 110KB
   options.arena_block_size = 4 << 10;
@@ -1228,6 +1241,7 @@ TEST_F(DBCompactionTest, L0_CompactionBug_Issue44_b) {
 TEST_P(DBCompactionTestWithParam, ManualCompaction) {
   Options options = CurrentOptions();
   options.max_subcompactions = max_subcompactions_;
+  options.statistics = rocksdb::CreateDBStatistics();
   CreateAndReopenWithCF({"pikachu"}, options);
 
   // iter - 0 with 7 levels
@@ -1259,7 +1273,14 @@ TEST_P(DBCompactionTestWithParam, ManualCompaction) {
     // Compact all
     MakeTables(1, "a", "z", 1);
     ASSERT_EQ("1,0,2", FilesPerLevel(1));
+
+    uint64_t prev_block_cache_add =
+        options.statistics->getTickerCount(BLOCK_CACHE_ADD);
     db_->CompactRange(CompactRangeOptions(), handles_[1], nullptr, nullptr);
+    // Verify manual compaction doesn't fill block cache
+    ASSERT_EQ(prev_block_cache_add,
+              options.statistics->getTickerCount(BLOCK_CACHE_ADD));
+
     ASSERT_EQ("0,0,1", FilesPerLevel(1));
 
     if (iter == 0) {
@@ -1267,6 +1288,7 @@ TEST_P(DBCompactionTestWithParam, ManualCompaction) {
       options.max_background_flushes = 0;
       options.num_levels = 3;
       options.create_if_missing = true;
+      options.statistics = rocksdb::CreateDBStatistics();
       DestroyAndReopen(options);
       CreateAndReopenWithCF({"pikachu"}, options);
     }
@@ -1593,6 +1615,8 @@ TEST_P(DBCompactionTestWithParam, CompressLevelCompaction) {
     return;
   }
   Options options = CurrentOptions();
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));
   options.compaction_style = kCompactionStyleLevel;
   options.write_buffer_size = 110 << 10;  // 110KB
   options.arena_block_size = 4 << 10;
@@ -1844,11 +1868,11 @@ TEST_P(DBCompactionTestWithParam, ForceBottommostLevelCompaction) {
 
 INSTANTIATE_TEST_CASE_P(DBCompactionTestWithParam, DBCompactionTestWithParam,
                         ::testing::Values(1, 4));
-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif // !defined(ROCKSDB_LITE)
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
   rocksdb::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();

db/db_dynamic_level_test.cc

@@ -10,10 +10,10 @@
 // Introduction of SyncPoint effectively disabled building and running this test
 // in Release build.
 // which is a pity, it is a good test
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
 
+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
 
 namespace rocksdb {
 class DBTestDynamicLevel : public DBTestBase {
@@ -484,10 +484,10 @@ TEST_F(DBTestDynamicLevel, MigrateToDynamicLevelMaxBytesBase) {
 }
 }  // namespace rocksdb
 
-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif  // !defined(ROCKSDB_LITE)
 
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
   rocksdb::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();

db/db_impl.h

@@ -36,11 +36,11 @@
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/transaction_log.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/autovector.h"
 #include "util/event_logger.h"
 #include "util/hash.h"
 #include "util/instrumented_mutex.h"
-#include "util/scoped_arena_iterator.h"
 #include "util/stop_watch.h"
 #include "util/thread_local.h"
 
@@ -122,6 +122,9 @@ class DBImpl : public DB {
   using DB::GetIntProperty;
   virtual bool GetIntProperty(ColumnFamilyHandle* column_family,
                               const Slice& property, uint64_t* value) override;
+  using DB::GetAggregatedIntProperty;
+  virtual bool GetAggregatedIntProperty(const Slice& property,
+                                        uint64_t* aggregated_value) override;
   using DB::GetApproximateSizes;
   virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
                                    const Range* range, int n, uint64_t* sizes,
@@ -141,6 +144,9 @@ class DBImpl : public DB {
   virtual Status PauseBackgroundWork() override;
   virtual Status ContinueBackgroundWork() override;
 
+  virtual Status EnableAutoCompaction(
+      const std::vector<ColumnFamilyHandle*>& column_family_handles) override;
+
   using DB::SetOptions;
   Status SetOptions(
       ColumnFamilyHandle* column_family,
@@ -248,7 +254,13 @@ class DBImpl : public DB {
                              const Slice* begin, const Slice* end,
                              bool disallow_trivial_move = false);
 
-#ifndef ROCKSDB_LITE
+  // Return an internal iterator over the current state of the database.
+  // The keys of this iterator are internal keys (see format.h).
+  // The returned iterator should be deleted when no longer needed.
+  InternalIterator* NewInternalIterator(
+      Arena* arena, ColumnFamilyHandle* column_family = nullptr);
+
+#ifndef NDEBUG
   // Extra methods (for testing) that are not in the public DB interface
   // Implemented in db_impl_debug.cc
 
@@ -266,12 +278,6 @@ class DBImpl : public DB {
   // Wait for any compaction
   Status TEST_WaitForCompact();
 
-  // Return an internal iterator over the current state of the database.
-  // The keys of this iterator are internal keys (see format.h).
-  // The returned iterator should be deleted when no longer needed.
-  Iterator* TEST_NewInternalIterator(
-      Arena* arena, ColumnFamilyHandle* column_family = nullptr);
-
   // Return the maximum overlapping data (in bytes) at next level for any
   // file at a level >= 1.
   int64_t TEST_MaxNextLevelOverlappingBytes(ColumnFamilyHandle* column_family =
@@ -305,7 +311,17 @@ class DBImpl : public DB {
 
   uint64_t TEST_LogfileNumber();
 
-#endif  // ROCKSDB_LITE
+  // Returns column family name to ImmutableCFOptions map.
+  Status TEST_GetAllImmutableCFOptions(
+      std::unordered_map<std::string, const ImmutableCFOptions*>* iopts_map);
+
+  Cache* TEST_table_cache() { return table_cache_.get(); }
+
+  const WriteController& TEST_write_controler() const {
+    return write_controller_;
+  }
+
+#endif  // NDEBUG
 
   // Returns the list of live files in 'live' and the list
   // of all files in the filesystem in 'candidate_files'.
@@ -363,6 +379,20 @@ class DBImpl : public DB {
   // Same as above, should called without mutex held and not on write thread.
   ColumnFamilyHandle* GetColumnFamilyHandleUnlocked(uint32_t column_family_id);
 
+  // Returns the number of currently running flushes.
+  // REQUIREMENT: mutex_ must be held when calling this function.
+  int num_running_flushes() {
+    mutex_.AssertHeld();
+    return num_running_flushes_;
+  }
+
+  // Returns the number of currently running compactions.
+  // REQUIREMENT: mutex_ must be held when calling this function.
+  int num_running_compactions() {
+    mutex_.AssertHeld();
+    return num_running_compactions_;
+  }
+
  protected:
   Env* const env_;
   const std::string dbname_;
@@ -370,12 +400,25 @@ class DBImpl : public DB {
   const DBOptions db_options_;
   Statistics* stats_;
 
-  Iterator* NewInternalIterator(const ReadOptions&, ColumnFamilyData* cfd,
-                                SuperVersion* super_version, Arena* arena);
+  InternalIterator* NewInternalIterator(const ReadOptions&,
+                                        ColumnFamilyData* cfd,
+                                        SuperVersion* super_version,
+                                        Arena* arena);
+
+  // Except in DB::Open(), WriteOptionsFile can only be called when:
+  // 1. WriteThread::Writer::EnterUnbatched() is used.
+  // 2. db_mutex is held
+  Status WriteOptionsFile();
+
+  // The following two functions can only be called when:
+  // 1. WriteThread::Writer::EnterUnbatched() is used.
+  // 2. db_mutex is NOT held
+  Status RenameTempFileToOptionsFile(const std::string& file_name);
+  Status DeleteObsoleteOptionsFiles();
 
   void NotifyOnFlushCompleted(ColumnFamilyData* cfd, FileMetaData* file_meta,
                               const MutableCFOptions& mutable_cf_options,
-                              int job_id);
+                              int job_id, TableProperties prop);
 
   void NotifyOnCompactionCompleted(ColumnFamilyData* cfd,
                                    Compaction *c, const Status &st,
@@ -527,8 +570,13 @@ class DBImpl : public DB {
   // Lock over the persistent DB state.  Non-nullptr iff successfully acquired.
   FileLock* db_lock_;
 
+  // The mutex for options file related operations.
+  // NOTE: should never acquire options_file_mutex_ and mutex_ at the
+  //       same time.
+  InstrumentedMutex options_files_mutex_;
   // State below is protected by mutex_
   InstrumentedMutex mutex_;
+
   std::atomic<bool> shutting_down_;
   // This condition variable is signaled on these conditions:
   // * whenever bg_compaction_scheduled_ goes down to 0
@@ -540,6 +588,8 @@ class DBImpl : public DB {
   // * whenever there is an error in background flush or compaction
   InstrumentedCondVar bg_cv_;
   uint64_t logfile_number_;
+  std::deque<uint64_t>
+      log_recycle_files;  // a list of log files that we can recycle
   bool log_dir_synced_;
   bool log_empty_;
   ColumnFamilyHandleImpl* default_cf_handle_;
@@ -685,6 +735,9 @@ class DBImpl : public DB {
   // count how many background compactions are running or have been scheduled
   int bg_compaction_scheduled_;
 
+  // stores the number of compactions are currently running
+  int num_running_compactions_;
+
   // If non-zero, MaybeScheduleFlushOrCompaction() will only schedule manual
   // compactions (if manual_compaction_ is not null). This mechanism enables
   // manual compactions to wait until all other compactions are finished.
@@ -693,6 +746,9 @@ class DBImpl : public DB {
   // number of background memtable flush jobs, submitted to the HIGH pool
   int bg_flush_scheduled_;
 
+  // stores the number of flushes are currently running
+  int num_running_flushes_;
+
   // Information for a manual compaction
   struct ManualCompaction {
     ColumnFamilyData* cfd;
@@ -788,6 +844,10 @@ class DBImpl : public DB {
   virtual Status GetPropertiesOfAllTables(ColumnFamilyHandle* column_family,
                                           TablePropertiesCollection* props)
       override;
+  virtual Status GetPropertiesOfTablesInRange(
+      ColumnFamilyHandle* column_family, const Range* range, std::size_t n,
+      TablePropertiesCollection* props) override;
+
 #endif  // ROCKSDB_LITE
 
   // Function that Get and KeyMayExist call with no_io true or false
@@ -796,9 +856,10 @@ class DBImpl : public DB {
                  const Slice& key, std::string* value,
                  bool* value_found = nullptr);
 
-  bool GetIntPropertyInternal(ColumnFamilyHandle* column_family,
+  bool GetIntPropertyInternal(ColumnFamilyData* cfd,
                               DBPropertyType property_type,
-                              bool need_out_of_mutex, uint64_t* value);
+                              bool need_out_of_mutex, bool is_locked,
+                              uint64_t* value);
 };
 
 // Sanitize db options.  The caller should delete result.info_log if

db/db_impl_debug.cc

@@ -7,7 +7,7 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
-#ifndef ROCKSDB_LITE
+#ifndef NDEBUG
 
 #include "db/db_impl.h"
 #include "util/thread_status_updater.h"
@@ -19,23 +19,6 @@ uint64_t DBImpl::TEST_GetLevel0TotalSize() {
   return default_cf_handle_->cfd()->current()->storage_info()->NumLevelBytes(0);
 }
 
-Iterator* DBImpl::TEST_NewInternalIterator(Arena* arena,
-                                           ColumnFamilyHandle* column_family) {
-  ColumnFamilyData* cfd;
-  if (column_family == nullptr) {
-    cfd = default_cf_handle_->cfd();
-  } else {
-    auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
-    cfd = cfh->cfd();
-  }
-
-  mutex_.Lock();
-  SuperVersion* super_version = cfd->GetSuperVersion()->Ref();
-  mutex_.Unlock();
-  ReadOptions roptions;
-  return NewInternalIterator(roptions, cfd, super_version, arena);
-}
-
 int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes(
     ColumnFamilyHandle* column_family) {
   ColumnFamilyData* cfd;
@@ -152,5 +135,24 @@ uint64_t DBImpl::TEST_LogfileNumber() {
   return logfile_number_;
 }
 
+Status DBImpl::TEST_GetAllImmutableCFOptions(
+    std::unordered_map<std::string, const ImmutableCFOptions*>* iopts_map) {
+  std::vector<std::string> cf_names;
+  std::vector<const ImmutableCFOptions*> iopts;
+  {
+    InstrumentedMutexLock l(&mutex_);
+    for (auto cfd : *versions_->GetColumnFamilySet()) {
+      cf_names.push_back(cfd->GetName());
+      iopts.push_back(cfd->ioptions());
+    }
+  }
+  iopts_map->clear();
+  for (size_t i = 0; i < cf_names.size(); ++i) {
+    iopts_map->insert({cf_names[i], iopts[i]});
+  }
+
+  return Status::OK();
+}
+
 }  // namespace rocksdb
-#endif  // ROCKSDB_LITE
+#endif  // NDEBUG

db/db_inplace_update_test.cc

@@ -6,8 +6,8 @@
 // Copyright (c) 2011 The LevelDB Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
 
 namespace rocksdb {
 

db/db_iter.cc

@@ -15,11 +15,12 @@
 
 #include "db/filename.h"
 #include "db/dbformat.h"
+#include "port/port.h"
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
 #include "rocksdb/iterator.h"
 #include "rocksdb/merge_operator.h"
-#include "port/port.h"
+#include "table/internal_iterator.h"
 #include "util/arena.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
@@ -58,10 +59,11 @@ class DBIter: public Iterator {
     kReverse
   };
 
-  DBIter(Env* env, const ImmutableCFOptions& ioptions,
-         const Comparator* cmp, Iterator* iter, SequenceNumber s,
-         bool arena_mode, uint64_t max_sequential_skip_in_iterations,
-         const Slice* iterate_upper_bound = nullptr)
+  DBIter(Env* env, const ImmutableCFOptions& ioptions, const Comparator* cmp,
+         InternalIterator* iter, SequenceNumber s, bool arena_mode,
+         uint64_t max_sequential_skip_in_iterations,
+         const Slice* iterate_upper_bound = nullptr,
+         bool prefix_same_as_start = false)
       : arena_mode_(arena_mode),
         env_(env),
         logger_(ioptions.info_log),
@@ -73,7 +75,8 @@ class DBIter: public Iterator {
         valid_(false),
         current_entry_is_merged_(false),
         statistics_(ioptions.statistics),
-        iterate_upper_bound_(iterate_upper_bound) {
+        iterate_upper_bound_(iterate_upper_bound),
+        prefix_same_as_start_(prefix_same_as_start) {
     RecordTick(statistics_, NO_ITERATORS);
     prefix_extractor_ = ioptions.prefix_extractor;
     max_skip_ = max_sequential_skip_in_iterations;
@@ -83,10 +86,10 @@ class DBIter: public Iterator {
     if (!arena_mode_) {
       delete iter_;
     } else {
-      iter_->~Iterator();
+      iter_->~InternalIterator();
     }
   }
-  virtual void SetIter(Iterator* iter) {
+  virtual void SetIter(InternalIterator* iter) {
     assert(iter_ == nullptr);
     iter_ = iter;
   }
@@ -142,7 +145,7 @@ class DBIter: public Iterator {
   Logger* logger_;
   const Comparator* const user_comparator_;
   const MergeOperator* const user_merge_operator_;
-  Iterator* iter_;
+  InternalIterator* iter_;
   SequenceNumber const sequence_;
 
   Status status_;
@@ -154,6 +157,8 @@ class DBIter: public Iterator {
   Statistics* statistics_;
   uint64_t max_skip_;
   const Slice* iterate_upper_bound_;
+  IterKey prefix_start_;
+  bool prefix_same_as_start_;
 
   // No copying allowed
   DBIter(const DBIter&);
@@ -181,9 +186,18 @@ void DBIter::Next() {
     if (!iter_->Valid()) {
       iter_->SeekToFirst();
     }
+  } else if (iter_->Valid() && !current_entry_is_merged_) {
+    // If the current value is not a merge, the iter position is the
+    // current key, which is already returned. We can safely issue a
+    // Next() without checking the current key.
+    // If the current key is a merge, very likely iter already points
+    // to the next internal position.
+    iter_->Next();
+    PERF_COUNTER_ADD(internal_key_skipped_count, 1);
   }
 
-  // If the current value is merged, we might already hit end of iter_
+  // Now we point to the next internal position, for both of merge and
+  // not merge cases.
   if (!iter_->Valid()) {
     valid_ = false;
     return;
@@ -196,6 +210,11 @@ void DBIter::Next() {
       RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
     }
   }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_ &&
+      prefix_extractor_->Transform(saved_key_.GetKey())
+              .compare(prefix_start_.GetKey()) != 0) {
+    valid_ = false;
+  }
 }
 
 // PRE: saved_key_ has the current user key if skipping
@@ -366,6 +385,11 @@ void DBIter::Prev() {
       RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
     }
   }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_ &&
+      prefix_extractor_->Transform(saved_key_.GetKey())
+              .compare(prefix_start_.GetKey()) != 0) {
+    valid_ = false;
+  }
 }
 
 void DBIter::ReverseToBackward() {
@@ -667,6 +691,9 @@ void DBIter::Seek(const Slice& target) {
   } else {
     valid_ = false;
   }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
+    prefix_start_.SetKey(prefix_extractor_->Transform(target));
+  }
 }
 
 void DBIter::SeekToFirst() {
@@ -695,6 +722,9 @@ void DBIter::SeekToFirst() {
   } else {
     valid_ = false;
   }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
+    prefix_start_.SetKey(prefix_extractor_->Transform(saved_key_.GetKey()));
+  }
 }
 
 void DBIter::SeekToLast() {
@@ -740,24 +770,28 @@ void DBIter::SeekToLast() {
       RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
     }
   }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
+    prefix_start_.SetKey(prefix_extractor_->Transform(saved_key_.GetKey()));
+  }
 }
 
 Iterator* NewDBIterator(Env* env, const ImmutableCFOptions& ioptions,
                         const Comparator* user_key_comparator,
-                        Iterator* internal_iter,
+                        InternalIterator* internal_iter,
                         const SequenceNumber& sequence,
                         uint64_t max_sequential_skip_in_iterations,
-                        const Slice* iterate_upper_bound) {
+                        const Slice* iterate_upper_bound,
+                        bool prefix_same_as_start) {
   return new DBIter(env, ioptions, user_key_comparator, internal_iter, sequence,
                     false, max_sequential_skip_in_iterations,
-                    iterate_upper_bound);
+                    iterate_upper_bound, prefix_same_as_start);
 }
 
 ArenaWrappedDBIter::~ArenaWrappedDBIter() { db_iter_->~DBIter(); }
 
 void ArenaWrappedDBIter::SetDBIter(DBIter* iter) { db_iter_ = iter; }
 
-void ArenaWrappedDBIter::SetIterUnderDBIter(Iterator* iter) {
+void ArenaWrappedDBIter::SetIterUnderDBIter(InternalIterator* iter) {
   static_cast<DBIter*>(db_iter_)->SetIter(iter);
 }
 
@@ -779,16 +813,16 @@ void ArenaWrappedDBIter::RegisterCleanup(CleanupFunction function, void* arg1,
 
 ArenaWrappedDBIter* NewArenaWrappedDbIterator(
     Env* env, const ImmutableCFOptions& ioptions,
-    const Comparator* user_key_comparator,
-    const SequenceNumber& sequence,
+    const Comparator* user_key_comparator, const SequenceNumber& sequence,
     uint64_t max_sequential_skip_in_iterations,
-    const Slice* iterate_upper_bound) {
+    const Slice* iterate_upper_bound, bool prefix_same_as_start) {
   ArenaWrappedDBIter* iter = new ArenaWrappedDBIter();
   Arena* arena = iter->GetArena();
   auto mem = arena->AllocateAligned(sizeof(DBIter));
-  DBIter* db_iter = new (mem) DBIter(env, ioptions, user_key_comparator,
-      nullptr, sequence, true, max_sequential_skip_in_iterations,
-      iterate_upper_bound);
+  DBIter* db_iter =
+      new (mem) DBIter(env, ioptions, user_key_comparator, nullptr, sequence,
+                       true, max_sequential_skip_in_iterations,
+                       iterate_upper_bound, prefix_same_as_start);
 
   iter->SetDBIter(db_iter);
 

db/db_iter.h

@@ -10,6 +10,7 @@
 #pragma once
 #include <stdint.h>
 #include "rocksdb/db.h"
+#include "rocksdb/iterator.h"
 #include "db/dbformat.h"
 #include "util/arena.h"
 #include "util/autovector.h"
@@ -18,18 +19,18 @@ namespace rocksdb {
 
 class Arena;
 class DBIter;
+class InternalIterator;
 
 // Return a new iterator that converts internal keys (yielded by
 // "*internal_iter") that were live at the specified "sequence" number
 // into appropriate user keys.
-extern Iterator* NewDBIterator(
-    Env* env,
-    const ImmutableCFOptions& options,
-    const Comparator *user_key_comparator,
-    Iterator* internal_iter,
-    const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations,
-    const Slice* iterate_upper_bound = nullptr);
+extern Iterator* NewDBIterator(Env* env, const ImmutableCFOptions& options,
+                               const Comparator* user_key_comparator,
+                               InternalIterator* internal_iter,
+                               const SequenceNumber& sequence,
+                               uint64_t max_sequential_skip_in_iterations,
+                               const Slice* iterate_upper_bound = nullptr,
+                               bool prefix_same_as_start = false);
 
 // A wrapper iterator which wraps DB Iterator and the arena, with which the DB
 // iterator is supposed be allocated. This class is used as an entry point of
@@ -50,7 +51,7 @@ class ArenaWrappedDBIter : public Iterator {
 
   // Set the internal iterator wrapped inside the DB Iterator. Usually it is
   // a merging iterator.
-  virtual void SetIterUnderDBIter(Iterator* iter);
+  virtual void SetIterUnderDBIter(InternalIterator* iter);
   virtual bool Valid() const override;
   virtual void SeekToFirst() override;
   virtual void SeekToLast() override;
@@ -60,6 +61,7 @@ class ArenaWrappedDBIter : public Iterator {
   virtual Slice key() const override;
   virtual Slice value() const override;
   virtual Status status() const override;
+
   void RegisterCleanup(CleanupFunction function, void* arg1, void* arg2);
 
  private:
@@ -70,8 +72,9 @@ class ArenaWrappedDBIter : public Iterator {
 // Generate the arena wrapped iterator class.
 extern ArenaWrappedDBIter* NewArenaWrappedDbIterator(
     Env* env, const ImmutableCFOptions& options,
-    const Comparator* user_key_comparator,
-    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
-    const Slice* iterate_upper_bound = nullptr);
+    const Comparator* user_key_comparator, const SequenceNumber& sequence,
+    uint64_t max_sequential_skip_in_iterations,
+    const Slice* iterate_upper_bound = nullptr,
+    bool prefix_same_as_start = false);
 
 }  // namespace rocksdb

db/db_iter_test.cc

@@ -29,7 +29,7 @@ static uint64_t TestGetTickerCount(const Options& options,
   return options.statistics->getTickerCount(ticker_type);
 }
 
-class TestIterator : public Iterator {
+class TestIterator : public InternalIterator {
  public:
   explicit TestIterator(const Comparator* comparator)
       : initialized_(false),
@@ -1864,11 +1864,12 @@ class DBIterWithMergeIterTest : public testing::Test {
     internal_iter2_->Add("d", kTypeValue, "7", 3u);
     internal_iter2_->Finish();
 
-    std::vector<Iterator*> child_iters;
+    std::vector<InternalIterator*> child_iters;
     child_iters.push_back(internal_iter1_);
     child_iters.push_back(internal_iter2_);
     InternalKeyComparator icomp(BytewiseComparator());
-    Iterator* merge_iter = NewMergingIterator(&icomp_, &child_iters[0], 2u);
+    InternalIterator* merge_iter =
+        NewMergingIterator(&icomp_, &child_iters[0], 2u);
 
     db_iter_.reset(NewDBIterator(env_, ImmutableCFOptions(options_),
                                  BytewiseComparator(), merge_iter,
@@ -1942,8 +1943,6 @@ TEST_F(DBIterWithMergeIterTest, InnerMergeIterator2) {
   ASSERT_EQ(db_iter_->value().ToString(), "4");
 }
 
-#if !(defined NDEBUG) || !defined(OS_WIN)
-
 TEST_F(DBIterWithMergeIterTest, InnerMergeIteratorDataRace1) {
   // Test Prev() when one child iterator is at its end but more rows
   // are added.
@@ -2294,7 +2293,6 @@ TEST_F(DBIterWithMergeIterTest, InnerMergeIteratorDataRace8) {
 
   rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
-#endif // #if !(defined NDEBUG) || !defined(OS_WIN)
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/db_log_iter_test.cc

@@ -10,10 +10,10 @@
 // Introduction of SyncPoint effectively disabled building and running this test
 // in Release build.
 // which is a pity, it is a good test
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
 
+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
 
 namespace rocksdb {
 
@@ -277,10 +277,10 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorBlobs) {
 }
 }  // namespace rocksdb
 
-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif  // !defined(ROCKSDB_LITE)
 
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
   rocksdb::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();

db/db_table_properties_test.cc

@@ -0,0 +1,226 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+//
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include <unordered_set>
+#include <vector>
+
+#include "db/db_test_util.h"
+#include "port/stack_trace.h"
+#include "rocksdb/db.h"
+#include "util/testharness.h"
+#include "util/testutil.h"
+
+#ifndef ROCKSDB_LITE
+
+namespace rocksdb {
+
+// A helper function that ensures the table properties returned in
+// `GetPropertiesOfAllTablesTest` is correct.
+// This test assumes entries size is different for each of the tables.
+namespace {
+
+void VerifyTableProperties(DB* db, uint64_t expected_entries_size) {
+  TablePropertiesCollection props;
+  ASSERT_OK(db->GetPropertiesOfAllTables(&props));
+
+  ASSERT_EQ(4U, props.size());
+  std::unordered_set<uint64_t> unique_entries;
+
+  // Indirect test
+  uint64_t sum = 0;
+  for (const auto& item : props) {
+    unique_entries.insert(item.second->num_entries);
+    sum += item.second->num_entries;
+  }
+
+  ASSERT_EQ(props.size(), unique_entries.size());
+  ASSERT_EQ(expected_entries_size, sum);
+}
+}  // namespace
+
+class DBTablePropertiesTest : public DBTestBase {
+ public:
+  DBTablePropertiesTest() : DBTestBase("/db_table_properties_test") {}
+  TablePropertiesCollection TestGetPropertiesOfTablesInRange(
+      std::vector<Range> ranges, std::size_t* num_properties = nullptr,
+      std::size_t* num_files = nullptr);
+};
+
+TEST_F(DBTablePropertiesTest, GetPropertiesOfAllTablesTest) {
+  Options options = CurrentOptions();
+  options.level0_file_num_compaction_trigger = 8;
+  Reopen(options);
+  // Create 4 tables
+  for (int table = 0; table < 4; ++table) {
+    for (int i = 0; i < 10 + table; ++i) {
+      db_->Put(WriteOptions(), ToString(table * 100 + i), "val");
+    }
+    db_->Flush(FlushOptions());
+  }
+
+  // 1. Read table properties directly from file
+  Reopen(options);
+  VerifyTableProperties(db_, 10 + 11 + 12 + 13);
+
+  // 2. Put two tables to table cache and
+  Reopen(options);
+  // fetch key from 1st and 2nd table, which will internally place that table to
+  // the table cache.
+  for (int i = 0; i < 2; ++i) {
+    Get(ToString(i * 100 + 0));
+  }
+
+  VerifyTableProperties(db_, 10 + 11 + 12 + 13);
+
+  // 3. Put all tables to table cache
+  Reopen(options);
+  // fetch key from 1st and 2nd table, which will internally place that table to
+  // the table cache.
+  for (int i = 0; i < 4; ++i) {
+    Get(ToString(i * 100 + 0));
+  }
+  VerifyTableProperties(db_, 10 + 11 + 12 + 13);
+}
+
+TablePropertiesCollection
+DBTablePropertiesTest::TestGetPropertiesOfTablesInRange(
+    std::vector<Range> ranges, std::size_t* num_properties,
+    std::size_t* num_files) {
+
+  // Since we deref zero element in the vector it can not be empty
+  // otherwise we pass an address to some random memory
+  EXPECT_GT(ranges.size(), 0U);
+  // run the query
+  TablePropertiesCollection props;
+  EXPECT_OK(db_->GetPropertiesOfTablesInRange(
+      db_->DefaultColumnFamily(), &ranges[0], ranges.size(), &props));
+
+  // Make sure that we've received properties for those and for those files
+  // only which fall within requested ranges
+  std::vector<LiveFileMetaData> vmd;
+  db_->GetLiveFilesMetaData(&vmd);
+  for (auto& md : vmd) {
+    std::string fn = md.db_path + md.name;
+    bool in_range = false;
+    for (auto& r : ranges) {
+      // smallestkey < limit && largestkey >= start
+      if (r.limit.compare(md.smallestkey) >= 0 &&
+          r.start.compare(md.largestkey) <= 0) {
+        in_range = true;
+        EXPECT_GT(props.count(fn), 0);
+      }
+    }
+    if (!in_range) {
+      EXPECT_EQ(props.count(fn), 0);
+    }
+  }
+
+  if (num_properties) {
+    *num_properties = props.size();
+  }
+
+  if (num_files) {
+    *num_files = vmd.size();
+  }
+  return props;
+}
+
+TEST_F(DBTablePropertiesTest, GetPropertiesOfTablesInRange) {
+  // Fixed random sead
+  Random rnd(301);
+
+  Options options;
+  options.create_if_missing = true;
+  options.write_buffer_size = 4096;
+  options.max_write_buffer_number = 8;
+  options.level0_file_num_compaction_trigger = 2;
+  options.level0_slowdown_writes_trigger = 2;
+  options.level0_stop_writes_trigger = 4;
+  options.target_file_size_base = 2048;
+  options.max_bytes_for_level_base = 10240;
+  options.max_bytes_for_level_multiplier = 4;
+  options.soft_rate_limit = 1.1;
+  options.num_levels = 8;
+
+  DestroyAndReopen(options);
+
+  // build a decent LSM
+  for (int i = 0; i < 10000; i++) {
+    ASSERT_OK(Put(test::RandomKey(&rnd, 5), RandomString(&rnd, 102)));
+  }
+  Flush();
+  db_->PauseBackgroundWork();
+
+  // Ensure that we have at least L0, L1 and L2
+  ASSERT_GT(NumTableFilesAtLevel(0), 0);
+  ASSERT_GT(NumTableFilesAtLevel(1), 0);
+  ASSERT_GT(NumTableFilesAtLevel(2), 0);
+
+  // Query the largest range
+  std::size_t num_properties, num_files;
+  TestGetPropertiesOfTablesInRange(
+      {Range(test::RandomKey(&rnd, 5, test::RandomKeyType::SMALLEST),
+             test::RandomKey(&rnd, 5, test::RandomKeyType::LARGEST))},
+      &num_properties, &num_files);
+  ASSERT_EQ(num_properties, num_files);
+
+  // Query the empty range
+  TestGetPropertiesOfTablesInRange(
+      {Range(test::RandomKey(&rnd, 5, test::RandomKeyType::LARGEST),
+             test::RandomKey(&rnd, 5, test::RandomKeyType::SMALLEST))},
+      &num_properties, &num_files);
+  ASSERT_GT(num_files, 0);
+  ASSERT_EQ(num_properties, 0);
+
+  // Query the middle rangee
+  TestGetPropertiesOfTablesInRange(
+      {Range(test::RandomKey(&rnd, 5, test::RandomKeyType::MIDDLE),
+             test::RandomKey(&rnd, 5, test::RandomKeyType::LARGEST))},
+      &num_properties, &num_files);
+  ASSERT_GT(num_files, 0);
+  ASSERT_GT(num_files, num_properties);
+  ASSERT_GT(num_properties, 0);
+
+  // Query a bunch of random ranges
+  for (int j = 0; j < 100; j++) {
+    // create a bunch of ranges
+    std::vector<std::string> random_keys;
+    // Random returns numbers with zero included
+    // when we pass empty ranges TestGetPropertiesOfTablesInRange()
+    // derefs random memory in the empty ranges[0]
+    // so want to be greater than zero and even since
+    // the below loop requires that random_keys.size() to be even.
+    auto n = 2 * (rnd.Uniform(50) + 1);
+
+    for (uint32_t i = 0; i < n; ++i) {
+      random_keys.push_back(test::RandomKey(&rnd, 5));
+    }
+
+    ASSERT_GT(random_keys.size(), 0U);
+    ASSERT_EQ((random_keys.size() % 2), 0U);
+
+    std::vector<Range> ranges;
+    auto it = random_keys.begin();
+    while (it != random_keys.end()) {
+      ranges.push_back(Range(*it, *(it + 1)));
+      it += 2;
+    }
+
+    TestGetPropertiesOfTablesInRange(std::move(ranges));
+  }
+}
+}  // namespace rocksdb
+
+#endif  // ROCKSDB_LITE
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

db/db_tailing_iter_test.cc

@@ -10,11 +10,11 @@
 // Introduction of SyncPoint effectively disabled building and running this test
 // in Release build.
 // which is a pity, it is a good test
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
 
+#include "db/db_test_util.h"
 #include "db/forward_iterator.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
 
 namespace rocksdb {
 
@@ -140,6 +140,8 @@ TEST_F(DBTestTailingIterator, TailingIteratorTrimSeekToNext) {
   std::unique_ptr<Iterator> iterh(db_->NewIterator(read_options, handles_[1]));
   std::string value(1024, 'a');
   bool file_iters_deleted = false;
+  bool file_iters_renewed_null = false;
+  bool file_iters_renewed_copy = false;
   rocksdb::SyncPoint::GetInstance()->SetCallBack(
       "ForwardIterator::SeekInternal:Return", [&](void* arg) {
         ForwardIterator* fiter = reinterpret_cast<ForwardIterator*>(arg);
@@ -152,6 +154,12 @@ TEST_F(DBTestTailingIterator, TailingIteratorTrimSeekToNext) {
         ASSERT_TRUE(!file_iters_deleted ||
                     fiter->TEST_CheckDeletedIters(&deleted_iters, &num_iters));
       });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "ForwardIterator::RenewIterators:Null",
+      [&](void* arg) { file_iters_renewed_null = true; });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "ForwardIterator::RenewIterators:Copy",
+      [&](void* arg) { file_iters_renewed_copy = true; });
   rocksdb::SyncPoint::GetInstance()->EnableProcessing();
   const int num_records = 1000;
   for (int i = 1; i < num_records; ++i) {
@@ -203,6 +211,8 @@ TEST_F(DBTestTailingIterator, TailingIteratorTrimSeekToNext) {
     ASSERT_LE(num_iters, 1);
     file_iters_deleted = false;
   }
+  ASSERT_TRUE(file_iters_renewed_null);
+  ASSERT_TRUE(file_iters_renewed_copy);
   iter = 0;
   itern = 0;
   iterh = 0;
@@ -646,10 +656,10 @@ TEST_F(DBTestTailingIterator, ManagedTailingIteratorSeekToSame) {
 
 }  // namespace rocksdb
 
-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif  // !defined(ROCKSDB_LITE)
 
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
   rocksdb::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();

db/db_test_util.cc

@@ -7,7 +7,7 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
-#include "util/db_test_util.h"
+#include "db/db_test_util.h"
 
 namespace rocksdb {
 
@@ -18,6 +18,7 @@ SpecialEnv::SpecialEnv(Env* base)
       rnd_(301),
       sleep_counter_(this),
       addon_time_(0),
+      time_elapse_only_sleep_(false),
       no_sleep_(false) {
   delay_sstable_sync_.store(false, std::memory_order_release);
   drop_writes_.store(false, std::memory_order_release);
@@ -38,11 +39,10 @@ SpecialEnv::SpecialEnv(Env* base)
   table_write_callback_ = nullptr;
 }
 
-
-DBTestBase::DBTestBase(const std::string path) : option_config_(kDefault),
-           mem_env_(!getenv("MEM_ENV") ? nullptr :
-                                         new MockEnv(Env::Default())),
-           env_(new SpecialEnv(mem_env_ ? mem_env_ : Env::Default())) {
+DBTestBase::DBTestBase(const std::string path)
+    : option_config_(kDefault),
+      mem_env_(!getenv("MEM_ENV") ? nullptr : new MockEnv(Env::Default())),
+      env_(new SpecialEnv(mem_env_ ? mem_env_ : Env::Default())) {
   env_->SetBackgroundThreads(1, Env::LOW);
   env_->SetBackgroundThreads(1, Env::HIGH);
   dbname_ = test::TmpDir(env_) + path;
@@ -56,15 +56,13 @@ DBTestBase::DBTestBase(const std::string path) : option_config_(kDefault),
   EXPECT_OK(DestroyDB(dbname_, options));
   db_ = nullptr;
   Reopen(options);
+  Random::GetTLSInstance()->Reset(0xdeadbeef);
 }
 
 DBTestBase::~DBTestBase() {
-// SyncPoint is not supported in Released Windows Mode.
-#if !(defined NDEBUG) || !defined(OS_WIN)
   rocksdb::SyncPoint::GetInstance()->DisableProcessing();
   rocksdb::SyncPoint::GetInstance()->LoadDependency({});
   rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
   Close();
   Options options;
   options.db_paths.emplace_back(dbname_, 0);
@@ -79,6 +77,23 @@ DBTestBase::~DBTestBase() {
 // test.  Return false if there are no more configurations to test.
 bool DBTestBase::ChangeOptions(int skip_mask) {
   for (option_config_++; option_config_ < kEnd; option_config_++) {
+#ifdef ROCKSDB_LITE
+    // These options are not supported in ROCKSDB_LITE
+    if (option_config_ == kHashSkipList ||
+        option_config_ == kPlainTableFirstBytePrefix ||
+        option_config_ == kPlainTableCappedPrefix ||
+        option_config_ == kPlainTableCappedPrefixNonMmap ||
+        option_config_ == kPlainTableAllBytesPrefix ||
+        option_config_ == kVectorRep || option_config_ == kHashLinkList ||
+        option_config_ == kHashCuckoo ||
+        option_config_ == kUniversalCompaction ||
+        option_config_ == kUniversalCompactionMultiLevel ||
+        option_config_ == kUniversalSubcompactions ||
+        option_config_ == kFIFOCompaction) {
+      continue;
+    }
+#endif
+
     if ((skip_mask & kSkipDeletesFilterFirst) &&
         option_config_ == kDeletesFilterFirst) {
       continue;
@@ -92,8 +107,7 @@ bool DBTestBase::ChangeOptions(int skip_mask) {
       continue;
     }
     if ((skip_mask & kSkipNoSeekToLast) &&
-        (option_config_ == kHashLinkList ||
-         option_config_ == kHashSkipList)) {;
+        (option_config_ == kHashLinkList || option_config_ == kHashSkipList)) {
       continue;
     }
     if ((skip_mask & kSkipPlainTable) &&
@@ -115,8 +129,7 @@ bool DBTestBase::ChangeOptions(int skip_mask) {
         option_config_ == kFIFOCompaction) {
       continue;
     }
-    if ((skip_mask & kSkipMmapReads) &&
-        option_config_ == kWalDirAndMmapReads) {
+    if ((skip_mask & kSkipMmapReads) && option_config_ == kWalDirAndMmapReads) {
       continue;
     }
     break;
@@ -205,10 +218,10 @@ Options DBTestBase::CurrentOptions(
   BlockBasedTableOptions table_options;
   bool set_block_based_table_factory = true;
   switch (option_config_) {
+#ifndef ROCKSDB_LITE
     case kHashSkipList:
       options.prefix_extractor.reset(NewFixedPrefixTransform(1));
-      options.memtable_factory.reset(
-          NewHashSkipListRepFactory(16));
+      options.memtable_factory.reset(NewHashSkipListRepFactory(16));
       break;
     case kPlainTableFirstBytePrefix:
       options.table_factory.reset(new PlainTableFactory());
@@ -238,6 +251,19 @@ Options DBTestBase::CurrentOptions(
       options.max_sequential_skip_in_iterations = 999999;
       set_block_based_table_factory = false;
       break;
+    case kVectorRep:
+      options.memtable_factory.reset(new VectorRepFactory(100));
+      break;
+    case kHashLinkList:
+      options.prefix_extractor.reset(NewFixedPrefixTransform(1));
+      options.memtable_factory.reset(
+          NewHashLinkListRepFactory(4, 0, 3, true, 4));
+      break;
+    case kHashCuckoo:
+      options.memtable_factory.reset(
+          NewHashCuckooRepFactory(options.write_buffer_size));
+      break;
+#endif  // ROCKSDB_LITE
     case kMergePut:
       options.merge_operator = MergeOperators::CreatePutOperator();
       break;
@@ -274,18 +300,6 @@ Options DBTestBase::CurrentOptions(
     case kDeletesFilterFirst:
       options.filter_deletes = true;
       break;
-    case kVectorRep:
-      options.memtable_factory.reset(new VectorRepFactory(100));
-      break;
-    case kHashLinkList:
-      options.prefix_extractor.reset(NewFixedPrefixTransform(1));
-      options.memtable_factory.reset(
-          NewHashLinkListRepFactory(4, 0, 3, true, 4));
-      break;
-    case kHashCuckoo:
-      options.memtable_factory.reset(
-          NewHashCuckooRepFactory(options.write_buffer_size));
-      break;
     case kUniversalCompaction:
       options.compaction_style = kCompactionStyleUniversal;
       options.num_levels = 1;
@@ -296,7 +310,7 @@ Options DBTestBase::CurrentOptions(
       break;
     case kCompressedBlockCache:
       options.allow_mmap_writes = true;
-      table_options.block_cache_compressed = NewLRUCache(8*1024*1024);
+      table_options.block_cache_compressed = NewLRUCache(8 * 1024 * 1024);
       break;
     case kInfiniteMaxOpenFiles:
       options.max_open_files = -1;
@@ -328,6 +342,10 @@ Options DBTestBase::CurrentOptions(
       options.row_cache = NewLRUCache(1024 * 1024);
       break;
     }
+    case kRecycleLogFiles: {
+      options.recycle_log_file_num = 2;
+      break;
+    }
     case kLevelSubcompactions: {
       options.max_subcompactions = 4;
       break;
@@ -351,11 +369,12 @@ Options DBTestBase::CurrentOptions(
   }
   options.env = env_;
   options.create_if_missing = true;
+  options.fail_if_options_file_error = true;
   return options;
 }
 
 void DBTestBase::CreateColumnFamilies(const std::vector<std::string>& cfs,
-                          const Options& options) {
+                                      const Options& options) {
   ColumnFamilyOptions cf_opts(options);
   size_t cfi = handles_.size();
   handles_.resize(cfi + cfs.size());
@@ -365,7 +384,7 @@ void DBTestBase::CreateColumnFamilies(const std::vector<std::string>& cfs,
 }
 
 void DBTestBase::CreateAndReopenWithCF(const std::vector<std::string>& cfs,
-                           const Options& options) {
+                                       const Options& options) {
   CreateColumnFamilies(cfs, options);
   std::vector<std::string> cfs_plus_default = cfs;
   cfs_plus_default.insert(cfs_plus_default.begin(), kDefaultColumnFamilyName);
@@ -373,18 +392,17 @@ void DBTestBase::CreateAndReopenWithCF(const std::vector<std::string>& cfs,
 }
 
 void DBTestBase::ReopenWithColumnFamilies(const std::vector<std::string>& cfs,
-                              const std::vector<Options>& options) {
+                                          const std::vector<Options>& options) {
   ASSERT_OK(TryReopenWithColumnFamilies(cfs, options));
 }
 
 void DBTestBase::ReopenWithColumnFamilies(const std::vector<std::string>& cfs,
-                              const Options& options) {
+                                          const Options& options) {
   ASSERT_OK(TryReopenWithColumnFamilies(cfs, options));
 }
 
 Status DBTestBase::TryReopenWithColumnFamilies(
-    const std::vector<std::string>& cfs,
-    const std::vector<Options>& options) {
+    const std::vector<std::string>& cfs, const std::vector<Options>& options) {
   Close();
   EXPECT_EQ(cfs.size(), options.size());
   std::vector<ColumnFamilyDescriptor> column_families;
@@ -396,8 +414,7 @@ Status DBTestBase::TryReopenWithColumnFamilies(
 }
 
 Status DBTestBase::TryReopenWithColumnFamilies(
-    const std::vector<std::string>& cfs,
-    const Options& options) {
+    const std::vector<std::string>& cfs, const Options& options) {
   Close();
   std::vector<Options> v_opts(cfs.size(), options);
   return TryReopenWithColumnFamilies(cfs, v_opts);
@@ -454,7 +471,7 @@ Status DBTestBase::Put(const Slice& k, const Slice& v, WriteOptions wo) {
 }
 
 Status DBTestBase::Put(int cf, const Slice& k, const Slice& v,
-           WriteOptions wo) {
+                       WriteOptions wo) {
   if (kMergePut == option_config_) {
     return db_->Merge(wo, handles_[cf], k, v);
   } else {
@@ -493,7 +510,7 @@ std::string DBTestBase::Get(const std::string& k, const Snapshot* snapshot) {
 }
 
 std::string DBTestBase::Get(int cf, const std::string& k,
-                const Snapshot* snapshot) {
+                            const Snapshot* snapshot) {
   ReadOptions options;
   options.verify_checksums = true;
   options.snapshot = snapshot;
@@ -552,9 +569,9 @@ std::string DBTestBase::AllEntriesFor(const Slice& user_key, int cf) {
   Arena arena;
   ScopedArenaIterator iter;
   if (cf == 0) {
-    iter.set(dbfull()->TEST_NewInternalIterator(&arena));
+    iter.set(dbfull()->NewInternalIterator(&arena));
   } else {
-    iter.set(dbfull()->TEST_NewInternalIterator(&arena, handles_[cf]));
+    iter.set(dbfull()->NewInternalIterator(&arena, handles_[cf]));
   }
   InternalKey target(user_key, kMaxSequenceNumber, kTypeValue);
   iter->Seek(target.Encode());
@@ -605,6 +622,7 @@ std::string DBTestBase::AllEntriesFor(const Slice& user_key, int cf) {
   return result;
 }
 
+#ifndef ROCKSDB_LITE
 int DBTestBase::NumSortedRuns(int cf) {
   ColumnFamilyMetaData cf_meta;
   if (cf == 0) {
@@ -631,20 +649,6 @@ uint64_t DBTestBase::TotalSize(int cf) {
   return cf_meta.size;
 }
 
-int DBTestBase::NumTableFilesAtLevel(int level, int cf) {
-  std::string property;
-  if (cf == 0) {
-    // default cfd
-    EXPECT_TRUE(db_->GetProperty(
-        "rocksdb.num-files-at-level" + NumberToString(level), &property));
-  } else {
-    EXPECT_TRUE(db_->GetProperty(
-        handles_[cf], "rocksdb.num-files-at-level" + NumberToString(level),
-        &property));
-  }
-  return atoi(property.c_str());
-}
-
 uint64_t DBTestBase::SizeAtLevel(int level) {
   std::vector<LiveFileMetaData> metadata;
   db_->GetLiveFilesMetaData(&metadata);
@@ -671,6 +675,27 @@ int DBTestBase::TotalLiveFiles(int cf) {
   return num_files;
 }
 
+size_t DBTestBase::CountLiveFiles() {
+  std::vector<LiveFileMetaData> metadata;
+  db_->GetLiveFilesMetaData(&metadata);
+  return metadata.size();
+}
+#endif  // ROCKSDB_LITE
+
+int DBTestBase::NumTableFilesAtLevel(int level, int cf) {
+  std::string property;
+  if (cf == 0) {
+    // default cfd
+    EXPECT_TRUE(db_->GetProperty(
+        "rocksdb.num-files-at-level" + NumberToString(level), &property));
+  } else {
+    EXPECT_TRUE(db_->GetProperty(
+        handles_[cf], "rocksdb.num-files-at-level" + NumberToString(level),
+        &property));
+  }
+  return atoi(property.c_str());
+}
+
 int DBTestBase::TotalTableFiles(int cf, int levels) {
   if (levels == -1) {
     levels = CurrentOptions().num_levels;
@@ -713,12 +738,6 @@ size_t DBTestBase::CountFiles() {
   return files.size() + logfiles.size();
 }
 
-size_t DBTestBase::CountLiveFiles() {
-  std::vector<LiveFileMetaData> metadata;
-  db_->GetLiveFilesMetaData(&metadata);
-  return metadata.size();
-}
-
 uint64_t DBTestBase::Size(const Slice& start, const Slice& limit, int cf) {
   Range r(start, limit);
   uint64_t size;
@@ -731,7 +750,7 @@ uint64_t DBTestBase::Size(const Slice& start, const Slice& limit, int cf) {
 }
 
 void DBTestBase::Compact(int cf, const Slice& start, const Slice& limit,
-             uint32_t target_path_id) {
+                         uint32_t target_path_id) {
   CompactRangeOptions compact_options;
   compact_options.target_path_id = target_path_id;
   ASSERT_OK(db_->CompactRange(compact_options, handles_[cf], &start, &limit));
@@ -748,9 +767,8 @@ void DBTestBase::Compact(const Slice& start, const Slice& limit) {
 
 // Do n memtable compactions, each of which produces an sstable
 // covering the range [small,large].
-void DBTestBase::MakeTables(
-    int n, const std::string& small,
-    const std::string& large, int cf) {
+void DBTestBase::MakeTables(int n, const std::string& small,
+                            const std::string& large, int cf) {
   for (int i = 0; i < n; i++) {
     ASSERT_OK(Put(cf, small, "begin"));
     ASSERT_OK(Put(cf, large, "end"));
@@ -761,8 +779,8 @@ void DBTestBase::MakeTables(
 
 // Prevent pushing of new sstables into deeper levels by adding
 // tables that cover a specified range to all levels.
-void DBTestBase::FillLevels(
-    const std::string& smallest, const std::string& largest, int cf) {
+void DBTestBase::FillLevels(const std::string& smallest,
+                            const std::string& largest, int cf) {
   MakeTables(db_->NumberLevels(handles_[cf]), smallest, largest, cf);
 }
 
@@ -779,7 +797,7 @@ void DBTestBase::MoveFilesToLevel(int level, int cf) {
 void DBTestBase::DumpFileCounts(const char* label) {
   fprintf(stderr, "---\n%s:\n", label);
   fprintf(stderr, "maxoverlap: %" PRIu64 "\n",
-      dbfull()->TEST_MaxNextLevelOverlappingBytes());
+          dbfull()->TEST_MaxNextLevelOverlappingBytes());
   for (int level = 0; level < db_->NumberLevels(); level++) {
     int num = NumTableFilesAtLevel(level);
     if (num > 0) {
@@ -815,7 +833,7 @@ int DBTestBase::GetSstFileCount(std::string path) {
 // this will generate non-overlapping files since it keeps increasing key_idx
 void DBTestBase::GenerateNewFile(int cf, Random* rnd, int* key_idx,
                                  bool nowait) {
-  for (int i = 0; i < 100; i++) {
+  for (int i = 0; i < KNumKeysByGenerateNewFile; i++) {
     ASSERT_OK(Put(cf, Key(*key_idx), RandomString(rnd, (i == 99) ? 1 : 990)));
     (*key_idx)++;
   }
@@ -827,7 +845,7 @@ void DBTestBase::GenerateNewFile(int cf, Random* rnd, int* key_idx,
 
 // this will generate non-overlapping files since it keeps increasing key_idx
 void DBTestBase::GenerateNewFile(Random* rnd, int* key_idx, bool nowait) {
-  for (int i = 0; i < 100; i++) {
+  for (int i = 0; i < KNumKeysByGenerateNewFile; i++) {
     ASSERT_OK(Put(Key(*key_idx), RandomString(rnd, (i == 99) ? 1 : 990)));
     (*key_idx)++;
   }
@@ -837,8 +855,10 @@ void DBTestBase::GenerateNewFile(Random* rnd, int* key_idx, bool nowait) {
   }
 }
 
+const int DBTestBase::kNumKeysByGenerateNewRandomFile = 51;
+
 void DBTestBase::GenerateNewRandomFile(Random* rnd, bool nowait) {
-  for (int i = 0; i < 51; i++) {
+  for (int i = 0; i < kNumKeysByGenerateNewRandomFile; i++) {
     ASSERT_OK(Put("key" + RandomString(rnd, 7), RandomString(rnd, 2000)));
   }
   ASSERT_OK(Put("key" + RandomString(rnd, 7), RandomString(rnd, 200)));
@@ -888,9 +908,10 @@ void DBTestBase::VerifyIterLast(std::string expected_key, int cf) {
 //   sets newValue with delta
 // If previous value is not empty,
 //   updates previous value with 'b' string of previous value size - 1.
-UpdateStatus DBTestBase::updateInPlaceSmallerSize(
-    char* prevValue, uint32_t* prevSize,
-    Slice delta, std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceSmallerSize(char* prevValue,
+                                                  uint32_t* prevSize,
+                                                  Slice delta,
+                                                  std::string* newValue) {
   if (prevValue == nullptr) {
     *newValue = std::string(delta.size(), 'c');
     return UpdateStatus::UPDATED;
@@ -902,9 +923,10 @@ UpdateStatus DBTestBase::updateInPlaceSmallerSize(
   }
 }
 
-UpdateStatus DBTestBase::updateInPlaceSmallerVarintSize(
-    char* prevValue, uint32_t* prevSize,
-    Slice delta, std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceSmallerVarintSize(char* prevValue,
+                                                        uint32_t* prevSize,
+                                                        Slice delta,
+                                                        std::string* newValue) {
   if (prevValue == nullptr) {
     *newValue = std::string(delta.size(), 'c');
     return UpdateStatus::UPDATED;
@@ -916,16 +938,17 @@ UpdateStatus DBTestBase::updateInPlaceSmallerVarintSize(
   }
 }
 
-UpdateStatus DBTestBase::updateInPlaceLargerSize(
-    char* prevValue, uint32_t* prevSize,
-    Slice delta, std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceLargerSize(char* prevValue,
+                                                 uint32_t* prevSize,
+                                                 Slice delta,
+                                                 std::string* newValue) {
   *newValue = std::string(delta.size(), 'c');
   return UpdateStatus::UPDATED;
 }
 
-UpdateStatus DBTestBase::updateInPlaceNoAction(
-    char* prevValue, uint32_t* prevSize,
-    Slice delta, std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceNoAction(char* prevValue,
+                                               uint32_t* prevSize, Slice delta,
+                                               std::string* newValue) {
   return UpdateStatus::UPDATE_FAILED;
 }
 
@@ -934,9 +957,9 @@ void DBTestBase::validateNumberOfEntries(int numValues, int cf) {
   ScopedArenaIterator iter;
   Arena arena;
   if (cf != 0) {
-    iter.set(dbfull()->TEST_NewInternalIterator(&arena, handles_[cf]));
+    iter.set(dbfull()->NewInternalIterator(&arena, handles_[cf]));
   } else {
-    iter.set(dbfull()->TEST_NewInternalIterator(&arena));
+    iter.set(dbfull()->NewInternalIterator(&arena));
   }
   iter->SeekToFirst();
   ASSERT_EQ(iter->status().ok(), true);
@@ -953,9 +976,8 @@ void DBTestBase::validateNumberOfEntries(int numValues, int cf) {
   ASSERT_EQ(0, seq);
 }
 
-void DBTestBase::CopyFile(
-    const std::string& source, const std::string& destination,
-    uint64_t size) {
+void DBTestBase::CopyFile(const std::string& source,
+                          const std::string& destination, uint64_t size) {
   const EnvOptions soptions;
   unique_ptr<SequentialFile> srcfile;
   ASSERT_OK(env_->NewSequentialFile(source, &srcfile, soptions));

db/db_test_util.h

@@ -27,8 +27,10 @@
 #include <vector>
 
 #include "db/db_impl.h"
+#include "db/db_test_util.h"
 #include "db/dbformat.h"
 #include "db/filename.h"
+#include "memtable/hash_linklist_rep.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/convenience.h"
@@ -42,12 +44,11 @@
 #include "table/block_based_table_factory.h"
 #include "table/mock_table.h"
 #include "table/plain_table_factory.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/compression.h"
-#include "util/db_test_util.h"
-#include "util/hash_linklist_rep.h"
 #include "util/mock_env.h"
 #include "util/mutexlock.h"
-#include "util/scoped_arena_iterator.h"
+
 #include "util/string_util.h"
 // SyncPoint is not supported in Released Windows Mode.
 #if !(defined NDEBUG) || !defined(OS_WIN)
@@ -117,6 +118,84 @@ struct OptionsOverride {
 
 }  // namespace anon
 
+// A hacky skip list mem table that triggers flush after number of entries.
+class SpecialMemTableRep : public MemTableRep {
+ public:
+  explicit SpecialMemTableRep(MemTableAllocator* allocator,
+                              MemTableRep* memtable, int num_entries_flush)
+      : MemTableRep(allocator),
+        memtable_(memtable),
+        num_entries_flush_(num_entries_flush),
+        num_entries_(0) {}
+
+  virtual KeyHandle Allocate(const size_t len, char** buf) override {
+    return memtable_->Allocate(len, buf);
+  }
+
+  // Insert key into the list.
+  // REQUIRES: nothing that compares equal to key is currently in the list.
+  virtual void Insert(KeyHandle handle) override {
+    memtable_->Insert(handle);
+    num_entries_++;
+  }
+
+  // Returns true iff an entry that compares equal to key is in the list.
+  virtual bool Contains(const char* key) const override {
+    return memtable_->Contains(key);
+  }
+
+  virtual size_t ApproximateMemoryUsage() override {
+    // Return a high memory usage when number of entries exceeds the threshold
+    // to trigger a flush.
+    return (num_entries_ < num_entries_flush_) ? 0 : 1024 * 1024 * 1024;
+  }
+
+  virtual void Get(const LookupKey& k, void* callback_args,
+                   bool (*callback_func)(void* arg,
+                                         const char* entry)) override {
+    memtable_->Get(k, callback_args, callback_func);
+  }
+
+  uint64_t ApproximateNumEntries(const Slice& start_ikey,
+                                 const Slice& end_ikey) override {
+    return memtable_->ApproximateNumEntries(start_ikey, end_ikey);
+  }
+
+  virtual MemTableRep::Iterator* GetIterator(Arena* arena = nullptr) override {
+    return memtable_->GetIterator(arena);
+  }
+
+  virtual ~SpecialMemTableRep() override {}
+
+ private:
+  unique_ptr<MemTableRep> memtable_;
+  int num_entries_flush_;
+  int num_entries_;
+};
+
+// The factory for the hacky skip list mem table that triggers flush after
+// number of entries exceeds a threshold.
+class SpecialSkipListFactory : public MemTableRepFactory {
+ public:
+  // After number of inserts exceeds `num_entries_flush` in a mem table, trigger
+  // flush.
+  explicit SpecialSkipListFactory(int num_entries_flush)
+      : num_entries_flush_(num_entries_flush) {}
+
+  virtual MemTableRep* CreateMemTableRep(
+      const MemTableRep::KeyComparator& compare, MemTableAllocator* allocator,
+      const SliceTransform* transform, Logger* logger) override {
+    return new SpecialMemTableRep(
+        allocator, factory_.CreateMemTableRep(compare, allocator, transform, 0),
+        num_entries_flush_);
+  }
+  virtual const char* Name() const override { return "SkipListFactory"; }
+
+ private:
+  SkipListFactory factory_;
+  int num_entries_flush_;
+};
+
 // Special Env used to delay background operations
 class SpecialEnv : public EnvWrapper {
  public:
@@ -131,9 +210,7 @@ class SpecialEnv : public EnvWrapper {
 
      public:
       SSTableFile(SpecialEnv* env, unique_ptr<WritableFile>&& base)
-          : env_(env),
-            base_(std::move(base)) {
-      }
+          : env_(env), base_(std::move(base)) {}
       Status Append(const Slice& data) override {
         if (env_->table_write_callback_) {
           (*env_->table_write_callback_)();
@@ -148,9 +225,7 @@ class SpecialEnv : public EnvWrapper {
           return base_->Append(data);
         }
       }
-      Status Truncate(uint64_t size) override {
-        return base_->Truncate(size);
-      }
+      Status Truncate(uint64_t size) override { return base_->Truncate(size); }
       Status Close() override {
 // SyncPoint is not supported in Released Windows Mode.
 #if !(defined NDEBUG) || !defined(OS_WIN)
@@ -180,7 +255,7 @@ class SpecialEnv : public EnvWrapper {
     class ManifestFile : public WritableFile {
      public:
       ManifestFile(SpecialEnv* env, unique_ptr<WritableFile>&& b)
-          : env_(env), base_(std::move(b)) { }
+          : env_(env), base_(std::move(b)) {}
       Status Append(const Slice& data) override {
         if (env_->manifest_write_error_.load(std::memory_order_acquire)) {
           return Status::IOError("simulated writer error");
@@ -283,8 +358,7 @@ class SpecialEnv : public EnvWrapper {
      public:
       CountingFile(unique_ptr<RandomAccessFile>&& target,
                    anon::AtomicCounter* counter)
-          : target_(std::move(target)), counter_(counter) {
-      }
+          : target_(std::move(target)), counter_(counter) {}
       virtual Status Read(uint64_t offset, size_t n, Slice* result,
                           char* scratch) const override {
         counter_->Increment();
@@ -329,18 +403,21 @@ class SpecialEnv : public EnvWrapper {
     return s;
   }
 
-
   virtual void SleepForMicroseconds(int micros) override {
     sleep_counter_.Increment();
-    if (no_sleep_) {
+    if (no_sleep_ || time_elapse_only_sleep_) {
       addon_time_.fetch_add(micros);
-    } else {
+    }
+    if (!no_sleep_) {
       target()->SleepForMicroseconds(micros);
     }
   }
 
   virtual Status GetCurrentTime(int64_t* unix_time) override {
-    Status s = target()->GetCurrentTime(unix_time);
+    Status s;
+    if (!time_elapse_only_sleep_) {
+      s = target()->GetCurrentTime(unix_time);
+    }
     if (s.ok()) {
       *unix_time += addon_time_.load();
     }
@@ -348,11 +425,13 @@ class SpecialEnv : public EnvWrapper {
   }
 
   virtual uint64_t NowNanos() override {
-    return target()->NowNanos() + addon_time_.load() * 1000;
+    return (time_elapse_only_sleep_ ? 0 : target()->NowNanos()) +
+           addon_time_.load() * 1000;
   }
 
   virtual uint64_t NowMicros() override {
-    return target()->NowMicros() + addon_time_.load();
+    return (time_elapse_only_sleep_ ? 0 : target()->NowMicros()) +
+           addon_time_.load();
   }
 
   Random rnd_;
@@ -404,9 +483,12 @@ class SpecialEnv : public EnvWrapper {
   std::function<void()>* table_write_callback_;
 
   std::atomic<int64_t> addon_time_;
+
+  bool time_elapse_only_sleep_;
+
   bool no_sleep_;
 
-  std::atomic<bool> is_wal_sync_thread_safe_ {true};
+  std::atomic<bool> is_wal_sync_thread_safe_{true};
 };
 
 class DBTestBase : public testing::Test {
@@ -442,9 +524,10 @@ class DBTestBase : public testing::Test {
     kFIFOCompaction = 26,
     kOptimizeFiltersForHits = 27,
     kRowCache = 28,
-    kLevelSubcompactions = 29,
-    kUniversalSubcompactions = 30,
-    kEnd = 29
+    kRecycleLogFiles = 29,
+    kLevelSubcompactions = 30,
+    kUniversalSubcompactions = 31,
+    kEnd = 30
   };
   int option_config_;
 
@@ -509,9 +592,7 @@ class DBTestBase : public testing::Test {
       const Options& defaultOptions,
       const anon::OptionsOverride& options_override = anon::OptionsOverride());
 
-  DBImpl* dbfull() {
-    return reinterpret_cast<DBImpl*>(db_);
-  }
+  DBImpl* dbfull() { return reinterpret_cast<DBImpl*>(db_); }
 
   void CreateColumnFamilies(const std::vector<std::string>& cfs,
                             const Options& options);
@@ -525,9 +606,8 @@ class DBTestBase : public testing::Test {
   void ReopenWithColumnFamilies(const std::vector<std::string>& cfs,
                                 const Options& options);
 
-  Status TryReopenWithColumnFamilies(
-      const std::vector<std::string>& cfs,
-      const std::vector<Options>& options);
+  Status TryReopenWithColumnFamilies(const std::vector<std::string>& cfs,
+                                     const std::vector<Options>& options);
 
   Status TryReopenWithColumnFamilies(const std::vector<std::string>& cfs,
                                      const Options& options);
@@ -574,16 +654,20 @@ class DBTestBase : public testing::Test {
 
   std::string AllEntriesFor(const Slice& user_key, int cf = 0);
 
+#ifndef ROCKSDB_LITE
   int NumSortedRuns(int cf = 0);
 
   uint64_t TotalSize(int cf = 0);
 
-  int NumTableFilesAtLevel(int level, int cf = 0);
-
   uint64_t SizeAtLevel(int level);
 
   int TotalLiveFiles(int cf = 0);
 
+  size_t CountLiveFiles();
+#endif  // ROCKSDB_LITE
+
+  int NumTableFilesAtLevel(int level, int cf = 0);
+
   int TotalTableFiles(int cf = 0, int levels = -1);
 
   // Return spread of files per level
@@ -591,8 +675,6 @@ class DBTestBase : public testing::Test {
 
   size_t CountFiles();
 
-  size_t CountLiveFiles();
-
   uint64_t Size(const Slice& start, const Slice& limit, int cf = 0);
 
   void Compact(int cf, const Slice& start, const Slice& limit,
@@ -627,6 +709,9 @@ class DBTestBase : public testing::Test {
 
   void GenerateNewFile(int fd, Random* rnd, int* key_idx, bool nowait = false);
 
+  static const int kNumKeysByGenerateNewRandomFile;
+  static const int KNumKeysByGenerateNewFile = 100;
+
   void GenerateNewRandomFile(Random* rnd, bool nowait = false);
 
   std::string IterStatus(Iterator* iter);
@@ -643,21 +728,21 @@ class DBTestBase : public testing::Test {
   //   sets newValue with delta
   // If previous value is not empty,
   //   updates previous value with 'b' string of previous value size - 1.
-  static UpdateStatus updateInPlaceSmallerSize(
-      char* prevValue, uint32_t* prevSize,
-      Slice delta, std::string* newValue);
+  static UpdateStatus updateInPlaceSmallerSize(char* prevValue,
+                                               uint32_t* prevSize, Slice delta,
+                                               std::string* newValue);
 
-  static UpdateStatus updateInPlaceSmallerVarintSize(
-      char* prevValue, uint32_t* prevSize,
-      Slice delta, std::string* newValue);
+  static UpdateStatus updateInPlaceSmallerVarintSize(char* prevValue,
+                                                     uint32_t* prevSize,
+                                                     Slice delta,
+                                                     std::string* newValue);
 
-  static UpdateStatus updateInPlaceLargerSize(
-      char* prevValue, uint32_t* prevSize,
-      Slice delta, std::string* newValue);
+  static UpdateStatus updateInPlaceLargerSize(char* prevValue,
+                                              uint32_t* prevSize, Slice delta,
+                                              std::string* newValue);
 
-  static UpdateStatus updateInPlaceNoAction(
-      char* prevValue, uint32_t* prevSize,
-      Slice delta, std::string* newValue);
+  static UpdateStatus updateInPlaceNoAction(char* prevValue, uint32_t* prevSize,
+                                            Slice delta, std::string* newValue);
 
   // Utility method to test InplaceUpdate
   void validateNumberOfEntries(int numValues, int cf = 0);

db/db_universal_compaction_test.cc

@@ -7,9 +7,9 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
 #include "util/sync_point.h"
 
 namespace rocksdb {
@@ -112,6 +112,44 @@ class DelayFilterFactory : public CompactionFilterFactory {
 };
 }  // namespace
 
+// Make sure we don't trigger a problem if the trigger conditon is given
+// to be 0, which is invalid.
+TEST_P(DBTestUniversalCompaction, UniversalCompactionSingleSortedRun) {
+  Options options;
+  options = CurrentOptions(options);
+
+  options.compaction_style = kCompactionStyleUniversal;
+  options.num_levels = num_levels_;
+  // Config universal compaction to always compact to one single sorted run.
+  options.level0_file_num_compaction_trigger = 0;
+  options.compaction_options_universal.size_ratio = 10;
+  options.compaction_options_universal.min_merge_width = 2;
+  options.compaction_options_universal.max_size_amplification_percent = 1;
+
+  options.write_buffer_size = 105 << 10;  // 105KB
+  options.arena_block_size = 4 << 10;
+  options.target_file_size_base = 32 << 10;  // 32KB
+  // trigger compaction if there are >= 4 files
+  KeepFilterFactory* filter = new KeepFilterFactory(true);
+  filter->expect_manual_compaction_.store(false);
+  options.compaction_filter_factory.reset(filter);
+
+  DestroyAndReopen(options);
+  ASSERT_EQ(1, db_->GetOptions().level0_file_num_compaction_trigger);
+
+  Random rnd(301);
+  int key_idx = 0;
+
+  filter->expect_full_compaction_.store(true);
+
+  for (int num = 0; num < 16; num++) {
+    // Write 100KB file. And immediately it should be compacted to one file.
+    GenerateNewFile(&rnd, &key_idx);
+    dbfull()->TEST_WaitForCompact();
+    ASSERT_EQ(NumSortedRuns(0), 1);
+  }
+}
+
 // TODO(kailiu) The tests on UniversalCompaction has some issues:
 //  1. A lot of magic numbers ("11" or "12").
 //  2. Made assumption on the memtable flush conditions, which may change from
@@ -119,6 +157,7 @@ class DelayFilterFactory : public CompactionFilterFactory {
 TEST_P(DBTestUniversalCompaction, UniversalCompactionTrigger) {
   Options options;
   options.compaction_style = kCompactionStyleUniversal;
+  options.compaction_options_universal.size_ratio = 5;
   options.num_levels = num_levels_;
   options.write_buffer_size = 105 << 10;  // 105KB
   options.arena_block_size = 4 << 10;
@@ -851,7 +890,10 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionFourPaths) {
   options.db_paths.emplace_back(dbname_ + "_2", 300 * 1024);
   options.db_paths.emplace_back(dbname_ + "_3", 500 * 1024);
   options.db_paths.emplace_back(dbname_ + "_4", 1024 * 1024 * 1024);
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));
   options.compaction_style = kCompactionStyleUniversal;
+  options.compaction_options_universal.size_ratio = 5;
   options.write_buffer_size = 110 << 10;  // 105KB
   options.arena_block_size = 4 << 10;
   options.level0_file_num_compaction_trigger = 2;
@@ -968,13 +1010,15 @@ TEST_P(DBTestUniversalCompaction, IncreaseUniversalCompactionNumLevels) {
   int max_key1 = 200;
   int max_key2 = 600;
   int max_key3 = 800;
+  const int KNumKeysPerFile = 10;
 
   // Stage 1: open a DB with universal compaction, num_levels=1
   Options options = CurrentOptions();
   options.compaction_style = kCompactionStyleUniversal;
   options.num_levels = 1;
-  options.write_buffer_size = 100 << 10;  // 100KB
+  options.write_buffer_size = 200 << 10;  // 200KB
   options.level0_file_num_compaction_trigger = 3;
+  options.memtable_factory.reset(new SpecialSkipListFactory(KNumKeysPerFile));
   options = CurrentOptions(options);
   CreateAndReopenWithCF({"pikachu"}, options);
 
@@ -1047,11 +1091,14 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionSecondPathRatio) {
   options.db_paths.emplace_back(dbname_, 500 * 1024);
   options.db_paths.emplace_back(dbname_ + "_2", 1024 * 1024 * 1024);
   options.compaction_style = kCompactionStyleUniversal;
+  options.compaction_options_universal.size_ratio = 5;
   options.write_buffer_size = 110 << 10;  // 105KB
   options.arena_block_size = 4 * 1024;
   options.arena_block_size = 4 << 10;
   options.level0_file_num_compaction_trigger = 2;
   options.num_levels = 1;
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));
   options = CurrentOptions(options);
 
   std::vector<std::string> filenames;
@@ -1210,10 +1257,10 @@ INSTANTIATE_TEST_CASE_P(DBTestUniversalManualCompactionOutputPathId,
 
 }  // namespace rocksdb
 
-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif  // !defined(ROCKSDB_LITE)
 
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
   rocksdb::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();

db/db_wal_test.cc

@@ -7,11 +7,9 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
-#if !(defined NDEBUG) || !defined(OS_WIN)
 #include "util/sync_point.h"
-#endif
 
 namespace rocksdb {
 class DBWALTest : public DBTestBase {
@@ -70,7 +68,6 @@ TEST_F(DBWALTest, RollLog) {
   } while (ChangeOptions());
 }
 
-#if !(defined NDEBUG) || !defined(OS_WIN)
 TEST_F(DBWALTest, SyncWALNotBlockWrite) {
   Options options = CurrentOptions();
   options.max_write_buffer_number = 4;
@@ -130,15 +127,10 @@ TEST_F(DBWALTest, SyncWALNotWaitWrite) {
   ASSERT_EQ(Get("foo2"), "bar2");
   rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
-#endif
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
   rocksdb::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
-#else
-  return 0;
-#endif
 }

db/event_helpers.cc

@@ -53,7 +53,7 @@ void EventHelpers::LogAndNotifyTableFileCreation(
                 info.table_properties.filter_policy_name;
 
     // user collected properties
-    for (const auto& prop : info.table_properties.user_collected_properties) {
+    for (const auto& prop : info.table_properties.readable_properties) {
       jwriter << prop.first << prop.second;
     }
     jwriter.EndObject();

db/fault_injection_test.cc

@@ -11,8 +11,6 @@
 // the last "sync". It then checks for data loss errors by purposely dropping
 // file data (or entire files) not protected by a "sync".
 
-#if !(defined NDEBUG) || !defined(OS_WIN)
-
 #include <map>
 #include <set>
 #include "db/db_impl.h"
@@ -902,13 +900,7 @@ INSTANTIATE_TEST_CASE_P(FaultTest, FaultInjectionTest, ::testing::Bool());
 
 }  // namespace rocksdb
 
-#endif // #if !(defined NDEBUG) || !defined(OS_WIN)
-
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
-#else
-  return 0;
-#endif
 }

db/filename.cc

@@ -21,6 +21,8 @@
 #include "util/file_reader_writer.h"
 #include "util/logging.h"
 #include "util/stop_watch.h"
+#include "util/string_util.h"
+#include "util/sync_point.h"
 
 namespace rocksdb {
 
@@ -46,8 +48,9 @@ static size_t GetInfoLogPrefix(const std::string& path, char* dest, int len) {
         path[i] == '_'){
       dest[write_idx++] = path[i];
     } else {
-      if (i > 0)
+      if (i > 0) {
         dest[write_idx++] = '_';
+      }
     }
     i++;
   }
@@ -145,7 +148,7 @@ std::string LockFileName(const std::string& dbname) {
 }
 
 std::string TempFileName(const std::string& dbname, uint64_t number) {
-  return MakeFileName(dbname, number, "dbtmp");
+  return MakeFileName(dbname, number, kTempFileNameSuffix.c_str());
 }
 
 InfoLogPrefix::InfoLogPrefix(bool has_log_dir,
@@ -185,6 +188,21 @@ std::string OldInfoLogFileName(const std::string& dbname, uint64_t ts,
   return log_dir + "/" + info_log_prefix.buf + ".old." + buf;
 }
 
+std::string OptionsFileName(const std::string& dbname, uint64_t file_num) {
+  char buffer[256];
+  snprintf(buffer, sizeof(buffer), "%s%06" PRIu64,
+           kOptionsFileNamePrefix.c_str(), file_num);
+  return dbname + "/" + buffer;
+}
+
+std::string TempOptionsFileName(const std::string& dbname, uint64_t file_num) {
+  char buffer[256];
+  snprintf(buffer, sizeof(buffer), "%s%06" PRIu64 ".%s",
+           kOptionsFileNamePrefix.c_str(), file_num,
+           kTempFileNameSuffix.c_str());
+  return dbname + "/" + buffer;
+}
+
 std::string MetaDatabaseName(const std::string& dbname, uint64_t number) {
   char buf[100];
   snprintf(buf, sizeof(buf), "/METADB-%llu",
@@ -205,6 +223,8 @@ std::string IdentityFileName(const std::string& dbname) {
 //    dbname/MANIFEST-[0-9]+
 //    dbname/[0-9]+.(log|sst)
 //    dbname/METADB-[0-9]+
+//    dbname/OPTIONS-[0-9]+
+//    dbname/OPTIONS-[0-9]+.dbtmp
 //    Disregards / at the beginning
 bool ParseFileName(const std::string& fname,
                    uint64_t* number,
@@ -267,6 +287,21 @@ bool ParseFileName(const std::string& fname, uint64_t* number,
     }
     *type = kMetaDatabase;
     *number = num;
+  } else if (rest.starts_with(kOptionsFileNamePrefix)) {
+    uint64_t ts_suffix;
+    bool is_temp_file = false;
+    rest.remove_prefix(kOptionsFileNamePrefix.size());
+    const std::string kTempFileNameSuffixWithDot =
+        std::string(".") + kTempFileNameSuffix;
+    if (rest.ends_with(kTempFileNameSuffixWithDot)) {
+      rest.remove_suffix(kTempFileNameSuffixWithDot.size());
+      is_temp_file = true;
+    }
+    if (!ConsumeDecimalNumber(&rest, &ts_suffix)) {
+      return false;
+    }
+    *number = ts_suffix;
+    *type = is_temp_file ? kTempFile : kOptionsFile;
   } else {
     // Avoid strtoull() to keep filename format independent of the
     // current locale
@@ -301,7 +336,7 @@ bool ParseFileName(const std::string& fname, uint64_t* number,
     } else if (suffix == Slice(kRocksDbTFileExt) ||
                suffix == Slice(kLevelDbTFileExt)) {
       *type = kTableFile;
-    } else if (suffix == Slice("dbtmp")) {
+    } else if (suffix == Slice(kTempFileNameSuffix)) {
       *type = kTempFile;
     } else {
       return false;
@@ -322,7 +357,9 @@ Status SetCurrentFile(Env* env, const std::string& dbname,
   std::string tmp = TempFileName(dbname, descriptor_number);
   Status s = WriteStringToFile(env, contents.ToString() + "\n", tmp, true);
   if (s.ok()) {
+    TEST_KILL_RANDOM("SetCurrentFile:0", rocksdb_kill_odds * REDUCE_ODDS2);
     s = env->RenameFile(tmp, CurrentFileName(dbname));
+    TEST_KILL_RANDOM("SetCurrentFile:1", rocksdb_kill_odds * REDUCE_ODDS2);
   }
   if (s.ok()) {
     if (directory_to_fsync != nullptr) {
@@ -351,6 +388,7 @@ Status SetIdentityFile(Env* env, const std::string& dbname) {
 
 Status SyncManifest(Env* env, const DBOptions* db_options,
                     WritableFileWriter* file) {
+  TEST_KILL_RANDOM("SyncManifest:0", rocksdb_kill_odds * REDUCE_ODDS2);
   if (db_options->disableDataSync) {
     return Status::OK();
   } else {

db/filename.h

@@ -36,7 +36,8 @@ enum FileType {
   kTempFile,
   kInfoLogFile,  // Either the current one, or an old one
   kMetaDatabase,
-  kIdentityFile
+  kIdentityFile,
+  kOptionsFile
 };
 
 // Return the name of the log file with the specified number
@@ -114,6 +115,19 @@ extern std::string OldInfoLogFileName(const std::string& dbname, uint64_t ts,
                                       const std::string& db_path = "",
                                       const std::string& log_dir = "");
 
+static const std::string kOptionsFileNamePrefix = "OPTIONS-";
+static const std::string kTempFileNameSuffix = "dbtmp";
+
+// Return a options file name given the "dbname" and file number.
+// Format:  OPTIONS-[number].dbtmp
+extern std::string OptionsFileName(const std::string& dbname,
+                                   uint64_t file_num);
+
+// Return a temp options file name given the "dbname" and file number.
+// Format:  OPTIONS-[number]
+extern std::string TempOptionsFileName(const std::string& dbname,
+                                       uint64_t file_num);
+
 // Return the name to use for a metadatabase. The result will be prefixed with
 // "dbname".
 extern std::string MetaDatabaseName(const std::string& dbname,

db/flush_job.cc

@@ -197,7 +197,7 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
     if (log_buffer_) {
       log_buffer_->FlushBufferToLog();
     }
-    std::vector<Iterator*> memtables;
+    std::vector<InternalIterator*> memtables;
     ReadOptions ro;
     ro.total_order_seek = true;
     Arena arena;
@@ -231,13 +231,15 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
 
       TEST_SYNC_POINT_CALLBACK("FlushJob::WriteLevel0Table:output_compression",
                                &output_compression_);
-      s = BuildTable(
-          dbname_, db_options_.env, *cfd_->ioptions(), env_options_,
-          cfd_->table_cache(), iter.get(), meta, cfd_->internal_comparator(),
-          cfd_->int_tbl_prop_collector_factories(), existing_snapshots_,
-          output_compression_, cfd_->ioptions()->compression_opts,
-          mutable_cf_options_.paranoid_file_checks, cfd_->internal_stats(),
-          Env::IO_HIGH, &info.table_properties);
+      s = BuildTable(dbname_, db_options_.env, *cfd_->ioptions(), env_options_,
+                     cfd_->table_cache(), iter.get(), meta,
+                     cfd_->internal_comparator(),
+                     cfd_->int_tbl_prop_collector_factories(), cfd_->GetID(),
+                     existing_snapshots_, output_compression_,
+                     cfd_->ioptions()->compression_opts,
+                     mutable_cf_options_.paranoid_file_checks,
+                     cfd_->internal_stats(), Env::IO_HIGH, &table_properties_);
+      info.table_properties = table_properties_;
       LogFlush(db_options_.info_log);
     }
     Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,

db/flush_job.h

@@ -27,12 +27,12 @@
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/transaction_log.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/autovector.h"
 #include "util/event_logger.h"
 #include "util/instrumented_mutex.h"
 #include "util/stop_watch.h"
 #include "util/thread_local.h"
-#include "util/scoped_arena_iterator.h"
 #include "db/internal_stats.h"
 #include "db/write_controller.h"
 #include "db/flush_scheduler.h"
@@ -66,6 +66,7 @@ class FlushJob {
   ~FlushJob();
 
   Status Run(FileMetaData* file_meta = nullptr);
+  TableProperties GetTableProperties() const { return table_properties_; }
 
  private:
   void ReportStartedFlush();
@@ -89,6 +90,7 @@ class FlushJob {
   CompressionType output_compression_;
   Statistics* stats_;
   EventLogger* event_logger_;
+  TableProperties table_properties_;
 };
 
 }  // namespace rocksdb

db/flush_job_test.cc

@@ -61,7 +61,7 @@ class FlushJobTest : public testing::Test {
     unique_ptr<WritableFileWriter> file_writer(
         new WritableFileWriter(std::move(file), EnvOptions()));
     {
-      log::Writer log(std::move(file_writer));
+      log::Writer log(std::move(file_writer), 0, false);
       std::string record;
       new_db.EncodeTo(&record);
       s = log.AddRecord(record);

db/forward_iterator.cc

@@ -28,7 +28,7 @@ namespace rocksdb {
 //     iter.SetFileIndex(file_index);
 //     iter.Seek(target);
 //     iter.Next()
-class LevelIterator : public Iterator {
+class LevelIterator : public InternalIterator {
  public:
   LevelIterator(const ColumnFamilyData* const cfd,
       const ReadOptions& read_options,
@@ -113,7 +113,7 @@ class LevelIterator : public Iterator {
   bool valid_;
   uint32_t file_index_;
   Status status_;
-  std::unique_ptr<Iterator> file_iter_;
+  std::unique_ptr<InternalIterator> file_iter_;
 };
 
 ForwardIterator::ForwardIterator(DBImpl* db, const ReadOptions& read_options,
@@ -144,12 +144,29 @@ ForwardIterator::~ForwardIterator() {
   Cleanup(true);
 }
 
+void ForwardIterator::SVCleanup() {
+  if (sv_ != nullptr && sv_->Unref()) {
+    // Job id == 0 means that this is not our background process, but rather
+    // user thread
+    JobContext job_context(0);
+    db_->mutex_.Lock();
+    sv_->Cleanup();
+    db_->FindObsoleteFiles(&job_context, false, true);
+    db_->mutex_.Unlock();
+    delete sv_;
+    if (job_context.HaveSomethingToDelete()) {
+      db_->PurgeObsoleteFiles(job_context);
+    }
+    job_context.Clean();
+  }
+}
+
 void ForwardIterator::Cleanup(bool release_sv) {
   if (mutable_iter_ != nullptr) {
-    mutable_iter_->~Iterator();
+    mutable_iter_->~InternalIterator();
   }
   for (auto* m : imm_iters_) {
-    m->~Iterator();
+    m->~InternalIterator();
   }
   imm_iters_.clear();
   for (auto* f : l0_iters_) {
@@ -162,20 +179,7 @@ void ForwardIterator::Cleanup(bool release_sv) {
   level_iters_.clear();
 
   if (release_sv) {
-    if (sv_ != nullptr && sv_->Unref()) {
-      // Job id == 0 means that this is not our background process, but rather
-      // user thread
-      JobContext job_context(0);
-      db_->mutex_.Lock();
-      sv_->Cleanup();
-      db_->FindObsoleteFiles(&job_context, false, true);
-      db_->mutex_.Unlock();
-      delete sv_;
-      if (job_context.HaveSomethingToDelete()) {
-        db_->PurgeObsoleteFiles(job_context);
-      }
-      job_context.Clean();
-    }
+    SVCleanup();
   }
 }
 
@@ -185,9 +189,10 @@ bool ForwardIterator::Valid() const {
 }
 
 void ForwardIterator::SeekToFirst() {
-  if (sv_ == nullptr ||
-      sv_ ->version_number != cfd_->GetSuperVersionNumber()) {
+  if (sv_ == nullptr) {
     RebuildIterators(true);
+  } else if (sv_->version_number != cfd_->GetSuperVersionNumber()) {
+    RenewIterators();
   } else if (immutable_status_.IsIncomplete()) {
     ResetIncompleteIterators();
   }
@@ -205,9 +210,10 @@ void ForwardIterator::Seek(const Slice& internal_key) {
   if (IsOverUpperBound(internal_key)) {
     valid_ = false;
   }
-  if (sv_ == nullptr ||
-      sv_ ->version_number != cfd_->GetSuperVersionNumber()) {
+  if (sv_ == nullptr) {
     RebuildIterators(true);
+  } else if (sv_->version_number != cfd_->GetSuperVersionNumber()) {
+    RenewIterators();
   } else if (immutable_status_.IsIncomplete()) {
     ResetIncompleteIterators();
   }
@@ -227,7 +233,9 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,
   // an option to turn it off.
   if (seek_to_first || NeedToSeekImmutable(internal_key)) {
     immutable_status_ = Status::OK();
-    if (has_iter_trimmed_for_upper_bound_) {
+    if ((has_iter_trimmed_for_upper_bound_) &&
+        (cfd_->internal_comparator().InternalKeyComparator::Compare(
+             prev_key_.GetKey(), internal_key) > 0)) {
       // Some iterators are trimmed. Need to rebuild.
       RebuildIterators(true);
       // Already seeked mutable iter, so seek again
@@ -393,7 +401,11 @@ void ForwardIterator::Next() {
     std::string current_key = key().ToString();
     Slice old_key(current_key.data(), current_key.size());
 
-    RebuildIterators(true);
+    if (sv_ == nullptr) {
+      RebuildIterators(true);
+    } else {
+      RenewIterators();
+    }
     SeekInternal(old_key, false);
     if (!valid_ || key().compare(old_key) != 0) {
       return;
@@ -484,10 +496,81 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
     l0_iters_.push_back(cfd_->table_cache()->NewIterator(
         read_options_, *cfd_->soptions(), cfd_->internal_comparator(), l0->fd));
   }
+  BuildLevelIterators(vstorage);
+  current_ = nullptr;
+  is_prev_set_ = false;
+}
+
+void ForwardIterator::RenewIterators() {
+  SuperVersion* svnew;
+  assert(sv_);
+  svnew = cfd_->GetReferencedSuperVersion(&(db_->mutex_));
+
+  if (mutable_iter_ != nullptr) {
+    mutable_iter_->~InternalIterator();
+  }
+  for (auto* m : imm_iters_) {
+    m->~InternalIterator();
+  }
+  imm_iters_.clear();
+
+  mutable_iter_ = svnew->mem->NewIterator(read_options_, &arena_);
+  svnew->imm->AddIterators(read_options_, &imm_iters_, &arena_);
+
+  const auto* vstorage = sv_->current->storage_info();
+  const auto& l0_files = vstorage->LevelFiles(0);
+  const auto* vstorage_new = svnew->current->storage_info();
+  const auto& l0_files_new = vstorage_new->LevelFiles(0);
+  uint32_t iold, inew;
+  bool found;
+  std::vector<InternalIterator*> l0_iters_new;
+  l0_iters_new.reserve(l0_files_new.size());
+
+  for (inew = 0; inew < l0_files_new.size(); inew++) {
+    found = false;
+    for (iold = 0; iold < l0_files.size(); iold++) {
+      if (l0_files[iold] == l0_files_new[inew]) {
+        found = true;
+        break;
+      }
+    }
+    if (found) {
+      if (l0_iters_[iold] == nullptr) {
+        l0_iters_new.push_back(nullptr);
+        TEST_SYNC_POINT_CALLBACK("ForwardIterator::RenewIterators:Null", this);
+      } else {
+        l0_iters_new.push_back(l0_iters_[iold]);
+        l0_iters_[iold] = nullptr;
+        TEST_SYNC_POINT_CALLBACK("ForwardIterator::RenewIterators:Copy", this);
+      }
+      continue;
+    }
+    l0_iters_new.push_back(cfd_->table_cache()->NewIterator(
+        read_options_, *cfd_->soptions(), cfd_->internal_comparator(),
+        l0_files_new[inew]->fd));
+  }
+
+  for (auto* f : l0_iters_) {
+    delete f;
+  }
+  l0_iters_.clear();
+  l0_iters_ = l0_iters_new;
+
+  for (auto* l : level_iters_) {
+    delete l;
+  }
+  level_iters_.clear();
+  BuildLevelIterators(vstorage_new);
+  current_ = nullptr;
+  is_prev_set_ = false;
+  SVCleanup();
+  sv_ = svnew;
+}
+
+void ForwardIterator::BuildLevelIterators(const VersionStorageInfo* vstorage) {
   level_iters_.reserve(vstorage->num_levels() - 1);
   for (int32_t level = 1; level < vstorage->num_levels(); ++level) {
     const auto& level_files = vstorage->LevelFiles(level);
-
     if ((level_files.empty()) ||
         ((read_options_.iterate_upper_bound != nullptr) &&
          (user_comparator_->Compare(*read_options_.iterate_upper_bound,
@@ -502,9 +585,6 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
           new LevelIterator(cfd_, read_options_, level_files));
     }
   }
-
-  current_ = nullptr;
-  is_prev_set_ = false;
 }
 
 void ForwardIterator::ResetIncompleteIterators() {

db/forward_iterator.h

@@ -14,6 +14,7 @@
 #include "rocksdb/iterator.h"
 #include "rocksdb/options.h"
 #include "db/dbformat.h"
+#include "table/internal_iterator.h"
 #include "util/arena.h"
 
 namespace rocksdb {
@@ -23,6 +24,7 @@ class Env;
 struct SuperVersion;
 class ColumnFamilyData;
 class LevelIterator;
+class VersionStorageInfo;
 struct FileMetaData;
 
 class MinIterComparator {
@@ -30,16 +32,15 @@ class MinIterComparator {
   explicit MinIterComparator(const Comparator* comparator) :
     comparator_(comparator) {}
 
-  bool operator()(Iterator* a, Iterator* b) {
+  bool operator()(InternalIterator* a, InternalIterator* b) {
     return comparator_->Compare(a->key(), b->key()) > 0;
   }
  private:
   const Comparator* comparator_;
 };
 
-typedef std::priority_queue<Iterator*,
-          std::vector<Iterator*>,
-          MinIterComparator> MinIterHeap;
+typedef std::priority_queue<InternalIterator*, std::vector<InternalIterator*>,
+                            MinIterComparator> MinIterHeap;
 
 /**
  * ForwardIterator is a special type of iterator that only supports Seek()
@@ -48,7 +49,7 @@ typedef std::priority_queue<Iterator*,
  * the iterator. At the current implementation, snapshot is taken at the
  * time Seek() is called. The Next() followed do not see new values after.
  */
-class ForwardIterator : public Iterator {
+class ForwardIterator : public InternalIterator {
  public:
   ForwardIterator(DBImpl* db, const ReadOptions& read_options,
                   ColumnFamilyData* cfd, SuperVersion* current_sv = nullptr);
@@ -74,7 +75,10 @@ class ForwardIterator : public Iterator {
 
  private:
   void Cleanup(bool release_sv);
+  void SVCleanup();
   void RebuildIterators(bool refresh_sv);
+  void RenewIterators();
+  void BuildLevelIterators(const VersionStorageInfo* vstorage);
   void ResetIncompleteIterators();
   void SeekInternal(const Slice& internal_key, bool seek_to_first);
   void UpdateCurrent();
@@ -94,11 +98,11 @@ class ForwardIterator : public Iterator {
   MinIterHeap immutable_min_heap_;
 
   SuperVersion* sv_;
-  Iterator* mutable_iter_;
-  std::vector<Iterator*> imm_iters_;
-  std::vector<Iterator*> l0_iters_;
+  InternalIterator* mutable_iter_;
+  std::vector<InternalIterator*> imm_iters_;
+  std::vector<InternalIterator*> l0_iters_;
   std::vector<LevelIterator*> level_iters_;
-  Iterator* current_;
+  InternalIterator* current_;
   bool valid_;
 
   // Internal iterator status; set only by one of the unsupported methods.

db/forward_iterator_bench.cc

@@ -0,0 +1,374 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#ifndef __STDC_FORMAT_MACROS
+#define __STDC_FORMAT_MACROS
+#endif
+
+#if !defined(GFLAGS) || defined(ROCKSDB_LITE)
+#include <cstdio>
+int main() {
+  fprintf(stderr, "Please install gflags to run rocksdb tools\n");
+  return 1;
+}
+#elif defined(OS_MACOSX) || defined(OS_WIN)
+// Block forward_iterator_bench under MAC and Windows
+int main() { return 0; }
+#else
+#include <gflags/gflags.h>
+#include <semaphore.h>
+#include <atomic>
+#include <bitset>
+#include <chrono>
+#include <climits>
+#include <condition_variable>
+#include <limits>
+#include <mutex>
+#include <queue>
+#include <random>
+#include <thread>
+
+#include "rocksdb/cache.h"
+#include "rocksdb/db.h"
+#include "rocksdb/status.h"
+#include "rocksdb/table.h"
+#include "util/testharness.h"
+
+const int MAX_SHARDS = 100000;
+
+DEFINE_int32(writers, 8, "");
+DEFINE_int32(readers, 8, "");
+DEFINE_int64(rate, 100000, "");
+DEFINE_int64(value_size, 300, "");
+DEFINE_int64(shards, 1000, "");
+DEFINE_int64(memtable_size, 500000000, "");
+DEFINE_int64(block_cache_size, 300000000, "");
+DEFINE_int64(block_size, 65536, "");
+DEFINE_double(runtime, 300.0, "");
+DEFINE_bool(cache_only_first, true, "");
+DEFINE_bool(iterate_upper_bound, true, "");
+
+struct Stats {
+  char pad1[128] __attribute__((__unused__));
+  std::atomic<uint64_t> written{0};
+  char pad2[128] __attribute__((__unused__));
+  std::atomic<uint64_t> read{0};
+  std::atomic<uint64_t> cache_misses{0};
+  char pad3[128] __attribute__((__unused__));
+} stats;
+
+struct Key {
+  Key() {}
+  Key(uint64_t shard_in, uint64_t seqno_in)
+      : shard_be(htobe64(shard_in)), seqno_be(htobe64(seqno_in)) {}
+
+  uint64_t shard() const { return be64toh(shard_be); }
+  uint64_t seqno() const { return be64toh(seqno_be); }
+
+ private:
+  uint64_t shard_be;
+  uint64_t seqno_be;
+} __attribute__((__packed__));
+
+struct Reader;
+struct Writer;
+
+struct ShardState {
+  char pad1[128] __attribute__((__unused__));
+  std::atomic<uint64_t> last_written{0};
+  Writer* writer;
+  Reader* reader;
+  char pad2[128] __attribute__((__unused__));
+  std::atomic<uint64_t> last_read{0};
+  std::unique_ptr<rocksdb::Iterator> it;
+  std::unique_ptr<rocksdb::Iterator> it_cacheonly;
+  Key upper_bound;
+  rocksdb::Slice upper_bound_slice;
+  char pad3[128] __attribute__((__unused__));
+};
+
+struct Reader {
+ public:
+  explicit Reader(std::vector<ShardState>* shard_states, rocksdb::DB* db)
+      : shard_states_(shard_states), db_(db) {
+    sem_init(&sem_, 0, 0);
+    thread_ = std::thread(&Reader::run, this);
+  }
+
+  void run() {
+    while (1) {
+      sem_wait(&sem_);
+      if (done_.load()) {
+        break;
+      }
+
+      uint64_t shard;
+      {
+        std::lock_guard<std::mutex> guard(queue_mutex_);
+        assert(!shards_pending_queue_.empty());
+        shard = shards_pending_queue_.front();
+        shards_pending_queue_.pop();
+        shards_pending_set_.reset(shard);
+      }
+      readOnceFromShard(shard);
+    }
+  }
+
+  void readOnceFromShard(uint64_t shard) {
+    ShardState& state = (*shard_states_)[shard];
+    if (!state.it) {
+      // Initialize iterators
+      rocksdb::ReadOptions options;
+      options.tailing = true;
+      if (FLAGS_iterate_upper_bound) {
+        state.upper_bound = Key(shard, std::numeric_limits<uint64_t>::max());
+        state.upper_bound_slice = rocksdb::Slice(
+            (const char*)&state.upper_bound, sizeof(state.upper_bound));
+        options.iterate_upper_bound = &state.upper_bound_slice;
+      }
+
+      state.it.reset(db_->NewIterator(options));
+
+      if (FLAGS_cache_only_first) {
+        options.read_tier = rocksdb::ReadTier::kBlockCacheTier;
+        state.it_cacheonly.reset(db_->NewIterator(options));
+      }
+    }
+
+    const uint64_t upto = state.last_written.load();
+    for (rocksdb::Iterator* it : {state.it_cacheonly.get(), state.it.get()}) {
+      if (it == nullptr) {
+        continue;
+      }
+      if (state.last_read.load() >= upto) {
+        break;
+      }
+      bool need_seek = true;
+      for (uint64_t seq = state.last_read.load() + 1; seq <= upto; ++seq) {
+        if (need_seek) {
+          Key from(shard, state.last_read.load() + 1);
+          it->Seek(rocksdb::Slice((const char*)&from, sizeof(from)));
+          need_seek = false;
+        } else {
+          it->Next();
+        }
+        if (it->status().IsIncomplete()) {
+          ++::stats.cache_misses;
+          break;
+        }
+        assert(it->Valid());
+        assert(it->key().size() == sizeof(Key));
+        Key key;
+        memcpy(&key, it->key().data(), it->key().size());
+        // fprintf(stderr, "Expecting (%ld, %ld) read (%ld, %ld)\n",
+        //         shard, seq, key.shard(), key.seqno());
+        assert(key.shard() == shard);
+        assert(key.seqno() == seq);
+        state.last_read.store(seq);
+        ++::stats.read;
+      }
+    }
+  }
+
+  void onWrite(uint64_t shard) {
+    {
+      std::lock_guard<std::mutex> guard(queue_mutex_);
+      if (!shards_pending_set_.test(shard)) {
+        shards_pending_queue_.push(shard);
+        shards_pending_set_.set(shard);
+        sem_post(&sem_);
+      }
+    }
+  }
+
+  ~Reader() {
+    done_.store(true);
+    sem_post(&sem_);
+    thread_.join();
+  }
+
+ private:
+  char pad1[128] __attribute__((__unused__));
+  std::vector<ShardState>* shard_states_;
+  rocksdb::DB* db_;
+  std::thread thread_;
+  sem_t sem_;
+  std::mutex queue_mutex_;
+  std::bitset<MAX_SHARDS + 1> shards_pending_set_;
+  std::queue<uint64_t> shards_pending_queue_;
+  std::atomic<bool> done_{false};
+  char pad2[128] __attribute__((__unused__));
+};
+
+struct Writer {
+  explicit Writer(std::vector<ShardState>* shard_states, rocksdb::DB* db)
+      : shard_states_(shard_states), db_(db) {}
+
+  void start() { thread_ = std::thread(&Writer::run, this); }
+
+  void run() {
+    std::queue<std::chrono::steady_clock::time_point> workq;
+    std::chrono::steady_clock::time_point deadline(
+        std::chrono::steady_clock::now() +
+        std::chrono::nanoseconds((uint64_t)(1000000000 * FLAGS_runtime)));
+    std::vector<uint64_t> my_shards;
+    for (int i = 1; i <= FLAGS_shards; ++i) {
+      if ((*shard_states_)[i].writer == this) {
+        my_shards.push_back(i);
+      }
+    }
+
+    std::mt19937 rng{std::random_device()()};
+    std::uniform_int_distribution<int> shard_dist(
+        0, static_cast<int>(my_shards.size()) - 1);
+    std::string value(FLAGS_value_size, '*');
+
+    while (1) {
+      auto now = std::chrono::steady_clock::now();
+      if (FLAGS_runtime >= 0 && now >= deadline) {
+        break;
+      }
+      if (workq.empty()) {
+        for (int i = 0; i < FLAGS_rate; i += FLAGS_writers) {
+          std::chrono::nanoseconds offset(1000000000LL * i / FLAGS_rate);
+          workq.push(now + offset);
+        }
+      }
+      while (!workq.empty() && workq.front() < now) {
+        workq.pop();
+        uint64_t shard = my_shards[shard_dist(rng)];
+        ShardState& state = (*shard_states_)[shard];
+        uint64_t seqno = state.last_written.load() + 1;
+        Key key(shard, seqno);
+        // fprintf(stderr, "Writing (%ld, %ld)\n", shard, seqno);
+        rocksdb::Status status =
+            db_->Put(rocksdb::WriteOptions(),
+                     rocksdb::Slice((const char*)&key, sizeof(key)),
+                     rocksdb::Slice(value));
+        assert(status.ok());
+        state.last_written.store(seqno);
+        state.reader->onWrite(shard);
+        ++::stats.written;
+      }
+      std::this_thread::sleep_for(std::chrono::milliseconds(1));
+    }
+    // fprintf(stderr, "Writer done\n");
+  }
+
+  ~Writer() { thread_.join(); }
+
+ private:
+  char pad1[128] __attribute__((__unused__));
+  std::vector<ShardState>* shard_states_;
+  rocksdb::DB* db_;
+  std::thread thread_;
+  char pad2[128] __attribute__((__unused__));
+};
+
+struct StatsThread {
+  explicit StatsThread(rocksdb::DB* db)
+      : db_(db), thread_(&StatsThread::run, this) {}
+
+  void run() {
+    //    using namespace std::chrono;
+    auto tstart = std::chrono::steady_clock::now(), tlast = tstart;
+    uint64_t wlast = 0, rlast = 0;
+    while (!done_.load()) {
+      {
+        std::unique_lock<std::mutex> lock(cvm_);
+        cv_.wait_for(lock, std::chrono::seconds(1));
+      }
+      auto now = std::chrono::steady_clock::now();
+      double elapsed =
+          std::chrono::duration_cast<std::chrono::duration<double> >(
+              now - tlast).count();
+      uint64_t w = ::stats.written.load();
+      uint64_t r = ::stats.read.load();
+      fprintf(stderr,
+              "%s elapsed %4lds | written %10ld | w/s %10.0f | read %10ld | "
+              "r/s %10.0f | cache misses %10ld\n",
+              db_->GetEnv()->TimeToString(time(nullptr)).c_str(),
+              std::chrono::duration_cast<std::chrono::seconds>(now - tstart)
+                  .count(),
+              w, (w - wlast) / elapsed, r, (r - rlast) / elapsed,
+              ::stats.cache_misses.load());
+      wlast = w;
+      rlast = r;
+      tlast = now;
+    }
+  }
+
+  ~StatsThread() {
+    {
+      std::lock_guard<std::mutex> guard(cvm_);
+      done_.store(true);
+    }
+    cv_.notify_all();
+    thread_.join();
+  }
+
+ private:
+  rocksdb::DB* db_;
+  std::mutex cvm_;
+  std::condition_variable cv_;
+  std::thread thread_;
+  std::atomic<bool> done_{false};
+};
+
+int main(int argc, char** argv) {
+  GFLAGS::ParseCommandLineFlags(&argc, &argv, true);
+
+  std::mt19937 rng{std::random_device()()};
+  rocksdb::Status status;
+  std::string path = rocksdb::test::TmpDir() + "/forward_iterator_test";
+  fprintf(stderr, "db path is %s\n", path.c_str());
+  rocksdb::Options options;
+  options.create_if_missing = true;
+  options.compression = rocksdb::CompressionType::kNoCompression;
+  options.compaction_style = rocksdb::CompactionStyle::kCompactionStyleNone;
+  options.level0_slowdown_writes_trigger = 99999;
+  options.level0_stop_writes_trigger = 99999;
+  options.allow_os_buffer = false;
+  options.write_buffer_size = FLAGS_memtable_size;
+  rocksdb::BlockBasedTableOptions table_options;
+  table_options.block_cache = rocksdb::NewLRUCache(FLAGS_block_cache_size);
+  table_options.block_size = FLAGS_block_size;
+  options.table_factory.reset(
+      rocksdb::NewBlockBasedTableFactory(table_options));
+
+  status = rocksdb::DestroyDB(path, options);
+  assert(status.ok());
+  rocksdb::DB* db_raw;
+  status = rocksdb::DB::Open(options, path, &db_raw);
+  assert(status.ok());
+  std::unique_ptr<rocksdb::DB> db(db_raw);
+
+  std::vector<ShardState> shard_states(FLAGS_shards + 1);
+  std::deque<Reader> readers;
+  while (static_cast<int>(readers.size()) < FLAGS_readers) {
+    readers.emplace_back(&shard_states, db_raw);
+  }
+  std::deque<Writer> writers;
+  while (static_cast<int>(writers.size()) < FLAGS_writers) {
+    writers.emplace_back(&shard_states, db_raw);
+  }
+
+  // Each shard gets a random reader and random writer assigned to it
+  for (int i = 1; i <= FLAGS_shards; ++i) {
+    std::uniform_int_distribution<int> reader_dist(0, FLAGS_readers - 1);
+    std::uniform_int_distribution<int> writer_dist(0, FLAGS_writers - 1);
+    shard_states[i].reader = &readers[reader_dist(rng)];
+    shard_states[i].writer = &writers[writer_dist(rng)];
+  }
+
+  StatsThread stats_thread(db_raw);
+  for (Writer& w : writers) {
+    w.start();
+  }
+
+  writers.clear();
+  readers.clear();
+}
+#endif  // !defined(GFLAGS) || defined(ROCKSDB_LITE)

db/inlineskiplist.h

@@ -0,0 +1,550 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional
+//  grant of patent rights can be found in the PATENTS file in the same
+//  directory.
+//
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.  Use of
+// this source code is governed by a BSD-style license that can be found
+// in the LICENSE file. See the AUTHORS file for names of contributors.
+//
+// InlineSkipList is derived from SkipList (skiplist.h), but it optimizes
+// the memory layout by requiring that the key storage be allocated through
+// the skip list instance.  For the common case of SkipList<const char*,
+// Cmp> this saves 1 pointer per skip list node and gives better cache
+// locality, at the expense of wasted padding from using AllocateAligned
+// instead of Allocate for the keys.  The unused padding will be from
+// 0 to sizeof(void*)-1 bytes, and the space savings are sizeof(void*)
+// bytes, so despite the padding the space used is always less than
+// SkipList<const char*, ..>.
+//
+// Thread safety -------------
+//
+// Writes require external synchronization, most likely a mutex.  Reads
+// require a guarantee that the InlineSkipList will not be destroyed while
+// the read is in progress.  Apart from that, reads progress without any
+// internal locking or synchronization.
+//
+// Invariants:
+//
+// (1) Allocated nodes are never deleted until the InlineSkipList is
+// destroyed.  This is trivially guaranteed by the code since we never
+// delete any skip list nodes.
+//
+// (2) The contents of a Node except for the next/prev pointers are
+// immutable after the Node has been linked into the InlineSkipList.
+// Only Insert() modifies the list, and it is careful to initialize a
+// node and use release-stores to publish the nodes in one or more lists.
+//
+// ... prev vs. next pointer ordering ...
+//
+
+#pragma once
+#include <assert.h>
+#include <stdlib.h>
+#include <atomic>
+#include "port/port.h"
+#include "util/allocator.h"
+#include "util/random.h"
+
+namespace rocksdb {
+
+template <class Comparator>
+class InlineSkipList {
+ private:
+  struct Node;
+
+ public:
+  // Create a new InlineSkipList object that will use "cmp" for comparing
+  // keys, and will allocate memory using "*allocator".  Objects allocated
+  // in the allocator must remain allocated for the lifetime of the
+  // skiplist object.
+  explicit InlineSkipList(Comparator cmp, Allocator* allocator,
+                          int32_t max_height = 12,
+                          int32_t branching_factor = 4);
+
+  // Allocates a key and a skip-list node, returning a pointer to the
+  // key portion of the node.
+  char* AllocateKey(size_t key_size);
+
+  // Inserts a key allocated by AllocateKey, after the actual key value
+  // has been filled in.
+  //
+  // REQUIRES: nothing that compares equal to key is currently in the list.
+  void Insert(const char* key);
+
+  // Returns true iff an entry that compares equal to key is in the list.
+  bool Contains(const char* key) const;
+
+  // Return estimated number of entries smaller than `key`.
+  uint64_t EstimateCount(const char* key) const;
+
+  // Iteration over the contents of a skip list
+  class Iterator {
+   public:
+    // Initialize an iterator over the specified list.
+    // The returned iterator is not valid.
+    explicit Iterator(const InlineSkipList* list);
+
+    // Change the underlying skiplist used for this iterator
+    // This enables us not changing the iterator without deallocating
+    // an old one and then allocating a new one
+    void SetList(const InlineSkipList* list);
+
+    // Returns true iff the iterator is positioned at a valid node.
+    bool Valid() const;
+
+    // Returns the key at the current position.
+    // REQUIRES: Valid()
+    const char* key() const;
+
+    // Advances to the next position.
+    // REQUIRES: Valid()
+    void Next();
+
+    // Advances to the previous position.
+    // REQUIRES: Valid()
+    void Prev();
+
+    // Advance to the first entry with a key >= target
+    void Seek(const char* target);
+
+    // Position at the first entry in list.
+    // Final state of iterator is Valid() iff list is not empty.
+    void SeekToFirst();
+
+    // Position at the last entry in list.
+    // Final state of iterator is Valid() iff list is not empty.
+    void SeekToLast();
+
+   private:
+    const InlineSkipList* list_;
+    Node* node_;
+    // Intentionally copyable
+  };
+
+ private:
+  const uint16_t kMaxHeight_;
+  const uint16_t kBranching_;
+  const uint32_t kScaledInverseBranching_;
+
+  // Immutable after construction
+  Comparator const compare_;
+  Allocator* const allocator_;  // Allocator used for allocations of nodes
+
+  Node* const head_;
+
+  // Modified only by Insert().  Read racily by readers, but stale
+  // values are ok.
+  std::atomic<int> max_height_;  // Height of the entire list
+
+  // Used for optimizing sequential insert patterns.  Tricky.  prev_[i] for
+  // i up to max_height_ is the predecessor of prev_[0] and prev_height_
+  // is the height of prev_[0].  prev_[0] can only be equal to head before
+  // insertion, in which case max_height_ and prev_height_ are 1.
+  Node** prev_;
+  int32_t prev_height_;
+
+  inline int GetMaxHeight() const {
+    return max_height_.load(std::memory_order_relaxed);
+  }
+
+  int RandomHeight();
+
+  Node* AllocateNode(size_t key_size, int height);
+
+  bool Equal(const char* a, const char* b) const {
+    return (compare_(a, b) == 0);
+  }
+
+  // Return true if key is greater than the data stored in "n".  Null n
+  // is considered infinite.
+  bool KeyIsAfterNode(const char* key, Node* n) const;
+
+  // Returns the earliest node with a key >= key.
+  // Return nullptr if there is no such node.
+  Node* FindGreaterOrEqual(const char* key) const;
+
+  // Return the latest node with a key < key.
+  // Return head_ if there is no such node.
+  // Fills prev[level] with pointer to previous node at "level" for every
+  // level in [0..max_height_-1], if prev is non-null.
+  Node* FindLessThan(const char* key, Node** prev = nullptr) const;
+
+  // Return the last node in the list.
+  // Return head_ if list is empty.
+  Node* FindLast() const;
+
+  // No copying allowed
+  InlineSkipList(const InlineSkipList&);
+  InlineSkipList& operator=(const InlineSkipList&);
+};
+
+// Implementation details follow
+
+// The Node data type is more of a pointer into custom-managed memory than
+// a traditional C++ struct.  The key is stored in the bytes immediately
+// after the struct, and the next_ pointers for nodes with height > 1 are
+// stored immediately _before_ the struct.  This avoids the need to include
+// any pointer or sizing data, which reduces per-node memory overheads.
+template <class Comparator>
+struct InlineSkipList<Comparator>::Node {
+  // Stores the height of the node in the memory location normally used for
+  // next_[0].  This is used for passing data from AllocateKey to Insert.
+  void StashHeight(const int height) {
+    assert(sizeof(int) <= sizeof(next_[0]));
+    memcpy(&next_[0], &height, sizeof(int));
+  }
+
+  // Retrieves the value passed to StashHeight.  Undefined after a call
+  // to SetNext or NoBarrier_SetNext.
+  int UnstashHeight() const {
+    int rv;
+    memcpy(&rv, &next_[0], sizeof(int));
+    return rv;
+  }
+
+  const char* Key() const { return reinterpret_cast<const char*>(&next_[1]); }
+
+  // Accessors/mutators for links.  Wrapped in methods so we can add
+  // the appropriate barriers as necessary, and perform the necessary
+  // addressing trickery for storing links below the Node in memory.
+  Node* Next(int n) {
+    assert(n >= 0);
+    // Use an 'acquire load' so that we observe a fully initialized
+    // version of the returned Node.
+    return (next_[-n].load(std::memory_order_acquire));
+  }
+
+  void SetNext(int n, Node* x) {
+    assert(n >= 0);
+    // Use a 'release store' so that anybody who reads through this
+    // pointer observes a fully initialized version of the inserted node.
+    next_[-n].store(x, std::memory_order_release);
+  }
+
+  // No-barrier variants that can be safely used in a few locations.
+  Node* NoBarrier_Next(int n) {
+    assert(n >= 0);
+    return next_[-n].load(std::memory_order_relaxed);
+  }
+
+  void NoBarrier_SetNext(int n, Node* x) {
+    assert(n >= 0);
+    next_[-n].store(x, std::memory_order_relaxed);
+  }
+
+ private:
+  // next_[0] is the lowest level link (level 0).  Higher levels are
+  // stored _earlier_, so level 1 is at next_[-1].
+  std::atomic<Node*> next_[1];
+};
+
+template <class Comparator>
+inline InlineSkipList<Comparator>::Iterator::Iterator(
+    const InlineSkipList* list) {
+  SetList(list);
+}
+
+template <class Comparator>
+inline void InlineSkipList<Comparator>::Iterator::SetList(
+    const InlineSkipList* list) {
+  list_ = list;
+  node_ = nullptr;
+}
+
+template <class Comparator>
+inline bool InlineSkipList<Comparator>::Iterator::Valid() const {
+  return node_ != nullptr;
+}
+
+template <class Comparator>
+inline const char* InlineSkipList<Comparator>::Iterator::key() const {
+  assert(Valid());
+  return node_->Key();
+}
+
+template <class Comparator>
+inline void InlineSkipList<Comparator>::Iterator::Next() {
+  assert(Valid());
+  node_ = node_->Next(0);
+}
+
+template <class Comparator>
+inline void InlineSkipList<Comparator>::Iterator::Prev() {
+  // Instead of using explicit "prev" links, we just search for the
+  // last node that falls before key.
+  assert(Valid());
+  node_ = list_->FindLessThan(node_->Key());
+  if (node_ == list_->head_) {
+    node_ = nullptr;
+  }
+}
+
+template <class Comparator>
+inline void InlineSkipList<Comparator>::Iterator::Seek(const char* target) {
+  node_ = list_->FindGreaterOrEqual(target);
+}
+
+template <class Comparator>
+inline void InlineSkipList<Comparator>::Iterator::SeekToFirst() {
+  node_ = list_->head_->Next(0);
+}
+
+template <class Comparator>
+inline void InlineSkipList<Comparator>::Iterator::SeekToLast() {
+  node_ = list_->FindLast();
+  if (node_ == list_->head_) {
+    node_ = nullptr;
+  }
+}
+
+template <class Comparator>
+int InlineSkipList<Comparator>::RandomHeight() {
+  auto rnd = Random::GetTLSInstance();
+
+  // Increase height with probability 1 in kBranching
+  int height = 1;
+  while (height < kMaxHeight_ && rnd->Next() < kScaledInverseBranching_) {
+    height++;
+  }
+  assert(height > 0);
+  assert(height <= kMaxHeight_);
+  return height;
+}
+
+template <class Comparator>
+bool InlineSkipList<Comparator>::KeyIsAfterNode(const char* key,
+                                                Node* n) const {
+  // nullptr n is considered infinite
+  return (n != nullptr) && (compare_(n->Key(), key) < 0);
+}
+
+template <class Comparator>
+typename InlineSkipList<Comparator>::Node*
+InlineSkipList<Comparator>::FindGreaterOrEqual(const char* key) const {
+  // Note: It looks like we could reduce duplication by implementing
+  // this function as FindLessThan(key)->Next(0), but we wouldn't be able
+  // to exit early on equality and the result wouldn't even be correct.
+  // A concurrent insert might occur after FindLessThan(key) but before
+  // we get a chance to call Next(0).
+  Node* x = head_;
+  int level = GetMaxHeight() - 1;
+  Node* last_bigger = nullptr;
+  while (true) {
+    Node* next = x->Next(level);
+    // Make sure the lists are sorted
+    assert(x == head_ || next == nullptr || KeyIsAfterNode(next->Key(), x));
+    // Make sure we haven't overshot during our search
+    assert(x == head_ || KeyIsAfterNode(key, x));
+    int cmp = (next == nullptr || next == last_bigger)
+                  ? 1
+                  : compare_(next->Key(), key);
+    if (cmp == 0 || (cmp > 0 && level == 0)) {
+      return next;
+    } else if (cmp < 0) {
+      // Keep searching in this list
+      x = next;
+    } else {
+      // Switch to next list, reuse compare_() result
+      last_bigger = next;
+      level--;
+    }
+  }
+}
+
+template <class Comparator>
+typename InlineSkipList<Comparator>::Node*
+InlineSkipList<Comparator>::FindLessThan(const char* key, Node** prev) const {
+  Node* x = head_;
+  int level = GetMaxHeight() - 1;
+  // KeyIsAfter(key, last_not_after) is definitely false
+  Node* last_not_after = nullptr;
+  while (true) {
+    Node* next = x->Next(level);
+    assert(x == head_ || next == nullptr || KeyIsAfterNode(next->Key(), x));
+    assert(x == head_ || KeyIsAfterNode(key, x));
+    if (next != last_not_after && KeyIsAfterNode(key, next)) {
+      // Keep searching in this list
+      x = next;
+    } else {
+      if (prev != nullptr) {
+        prev[level] = x;
+      }
+      if (level == 0) {
+        return x;
+      } else {
+        // Switch to next list, reuse KeyIUsAfterNode() result
+        last_not_after = next;
+        level--;
+      }
+    }
+  }
+}
+
+template <class Comparator>
+typename InlineSkipList<Comparator>::Node*
+InlineSkipList<Comparator>::FindLast() const {
+  Node* x = head_;
+  int level = GetMaxHeight() - 1;
+  while (true) {
+    Node* next = x->Next(level);
+    if (next == nullptr) {
+      if (level == 0) {
+        return x;
+      } else {
+        // Switch to next list
+        level--;
+      }
+    } else {
+      x = next;
+    }
+  }
+}
+
+template <class Comparator>
+uint64_t InlineSkipList<Comparator>::EstimateCount(const char* key) const {
+  uint64_t count = 0;
+
+  Node* x = head_;
+  int level = GetMaxHeight() - 1;
+  while (true) {
+    assert(x == head_ || compare_(x->Key(), key) < 0);
+    Node* next = x->Next(level);
+    if (next == nullptr || compare_(next->Key(), key) >= 0) {
+      if (level == 0) {
+        return count;
+      } else {
+        // Switch to next list
+        count *= kBranching_;
+        level--;
+      }
+    } else {
+      x = next;
+      count++;
+    }
+  }
+}
+
+template <class Comparator>
+InlineSkipList<Comparator>::InlineSkipList(const Comparator cmp,
+                                           Allocator* allocator,
+                                           int32_t max_height,
+                                           int32_t branching_factor)
+    : kMaxHeight_(max_height),
+      kBranching_(branching_factor),
+      kScaledInverseBranching_((Random::kMaxNext + 1) / kBranching_),
+      compare_(cmp),
+      allocator_(allocator),
+      head_(AllocateNode(0, max_height)),
+      max_height_(1),
+      prev_height_(1) {
+  assert(max_height > 0 && kMaxHeight_ == static_cast<uint32_t>(max_height));
+  assert(branching_factor > 0 &&
+         kBranching_ == static_cast<uint32_t>(branching_factor));
+  assert(kScaledInverseBranching_ > 0);
+  // Allocate the prev_ Node* array, directly from the passed-in allocator.
+  // prev_ does not need to be freed, as its life cycle is tied up with
+  // the allocator as a whole.
+  prev_ = reinterpret_cast<Node**>(
+      allocator_->AllocateAligned(sizeof(Node*) * kMaxHeight_));
+  for (int i = 0; i < kMaxHeight_; i++) {
+    head_->SetNext(i, nullptr);
+    prev_[i] = head_;
+  }
+}
+
+template <class Comparator>
+char* InlineSkipList<Comparator>::AllocateKey(size_t key_size) {
+  return const_cast<char*>(AllocateNode(key_size, RandomHeight())->Key());
+}
+
+template <class Comparator>
+typename InlineSkipList<Comparator>::Node*
+InlineSkipList<Comparator>::AllocateNode(size_t key_size, int height) {
+  auto prefix = sizeof(std::atomic<Node*>) * (height - 1);
+
+  // prefix is space for the height - 1 pointers that we store before
+  // the Node instance (next_[-(height - 1) .. -1]).  Node starts at
+  // raw + prefix, and holds the bottom-mode (level 0) skip list pointer
+  // next_[0].  key_size is the bytes for the key, which comes just after
+  // the Node.
+  char* raw = allocator_->AllocateAligned(prefix + sizeof(Node) + key_size);
+  Node* x = reinterpret_cast<Node*>(raw + prefix);
+
+  // Once we've linked the node into the skip list we don't actually need
+  // to know its height, because we can implicitly use the fact that we
+  // traversed into a node at level h to known that h is a valid level
+  // for that node.  We need to convey the height to the Insert step,
+  // however, so that it can perform the proper links.  Since we're not
+  // using the pointers at the moment, StashHeight temporarily borrow
+  // storage from next_[0] for that purpose.
+  x->StashHeight(height);
+  return x;
+}
+
+template <class Comparator>
+void InlineSkipList<Comparator>::Insert(const char* key) {
+  // fast path for sequential insertion
+  if (!KeyIsAfterNode(key, prev_[0]->NoBarrier_Next(0)) &&
+      (prev_[0] == head_ || KeyIsAfterNode(key, prev_[0]))) {
+    assert(prev_[0] != head_ || (prev_height_ == 1 && GetMaxHeight() == 1));
+
+    // Outside of this method prev_[1..max_height_] is the predecessor
+    // of prev_[0], and prev_height_ refers to prev_[0].  Inside Insert
+    // prev_[0..max_height - 1] is the predecessor of key.  Switch from
+    // the external state to the internal
+    for (int i = 1; i < prev_height_; i++) {
+      prev_[i] = prev_[0];
+    }
+  } else {
+    // TODO(opt): we could use a NoBarrier predecessor search as an
+    // optimization for architectures where memory_order_acquire needs
+    // a synchronization instruction.  Doesn't matter on x86
+    FindLessThan(key, prev_);
+  }
+
+  // Our data structure does not allow duplicate insertion
+  assert(prev_[0]->Next(0) == nullptr || !Equal(key, prev_[0]->Next(0)->Key()));
+
+  // Find the Node that we placed before the key in AllocateKey
+  Node* x = reinterpret_cast<Node*>(const_cast<char*>(key)) - 1;
+  int height = x->UnstashHeight();
+  assert(height >= 1 && height <= kMaxHeight_);
+
+  if (height > GetMaxHeight()) {
+    for (int i = GetMaxHeight(); i < height; i++) {
+      prev_[i] = head_;
+    }
+
+    // It is ok to mutate max_height_ without any synchronization
+    // with concurrent readers.  A concurrent reader that observes
+    // the new value of max_height_ will see either the old value of
+    // new level pointers from head_ (nullptr), or a new value set in
+    // the loop below.  In the former case the reader will
+    // immediately drop to the next level since nullptr sorts after all
+    // keys.  In the latter case the reader will use the new node.
+    max_height_.store(height, std::memory_order_relaxed);
+  }
+
+  for (int i = 0; i < height; i++) {
+    // NoBarrier_SetNext() suffices since we will add a barrier when
+    // we publish a pointer to "x" in prev[i].
+    x->NoBarrier_SetNext(i, prev_[i]->NoBarrier_Next(i));
+    prev_[i]->SetNext(i, x);
+  }
+  prev_[0] = x;
+  prev_height_ = height;
+}
+
+template <class Comparator>
+bool InlineSkipList<Comparator>::Contains(const char* key) const {
+  Node* x = FindGreaterOrEqual(key);
+  if (x != nullptr && Equal(key, x->Key())) {
+    return true;
+  } else {
+    return false;
+  }
+}
+
+}  // namespace rocksdb

db/inlineskiplist_test.cc

@@ -0,0 +1,391 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+//
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "db/inlineskiplist.h"
+#include <set>
+#include "rocksdb/env.h"
+#include "util/arena.h"
+#include "util/hash.h"
+#include "util/random.h"
+#include "util/testharness.h"
+
+namespace rocksdb {
+
+// Our test skip list stores 8-byte unsigned integers
+typedef uint64_t Key;
+
+static const char* Encode(const uint64_t* key) {
+  return reinterpret_cast<const char*>(key);
+}
+
+static Key Decode(const char* key) {
+  Key rv;
+  memcpy(&rv, key, sizeof(Key));
+  return rv;
+}
+
+struct TestComparator {
+  int operator()(const char* a, const char* b) const {
+    if (Decode(a) < Decode(b)) {
+      return -1;
+    } else if (Decode(a) > Decode(b)) {
+      return +1;
+    } else {
+      return 0;
+    }
+  }
+};
+
+class InlineSkipTest : public testing::Test {};
+
+TEST_F(InlineSkipTest, Empty) {
+  Arena arena;
+  TestComparator cmp;
+  InlineSkipList<TestComparator> list(cmp, &arena);
+  Key key = 10;
+  ASSERT_TRUE(!list.Contains(Encode(&key)));
+
+  InlineSkipList<TestComparator>::Iterator iter(&list);
+  ASSERT_TRUE(!iter.Valid());
+  iter.SeekToFirst();
+  ASSERT_TRUE(!iter.Valid());
+  key = 100;
+  iter.Seek(Encode(&key));
+  ASSERT_TRUE(!iter.Valid());
+  iter.SeekToLast();
+  ASSERT_TRUE(!iter.Valid());
+}
+
+TEST_F(InlineSkipTest, InsertAndLookup) {
+  const int N = 2000;
+  const int R = 5000;
+  Random rnd(1000);
+  std::set<Key> keys;
+  Arena arena;
+  TestComparator cmp;
+  InlineSkipList<TestComparator> list(cmp, &arena);
+  for (int i = 0; i < N; i++) {
+    Key key = rnd.Next() % R;
+    if (keys.insert(key).second) {
+      char* buf = list.AllocateKey(sizeof(Key));
+      memcpy(buf, &key, sizeof(Key));
+      list.Insert(buf);
+    }
+  }
+
+  for (Key i = 0; i < R; i++) {
+    if (list.Contains(Encode(&i))) {
+      ASSERT_EQ(keys.count(i), 1U);
+    } else {
+      ASSERT_EQ(keys.count(i), 0U);
+    }
+  }
+
+  // Simple iterator tests
+  {
+    InlineSkipList<TestComparator>::Iterator iter(&list);
+    ASSERT_TRUE(!iter.Valid());
+
+    uint64_t zero = 0;
+    iter.Seek(Encode(&zero));
+    ASSERT_TRUE(iter.Valid());
+    ASSERT_EQ(*(keys.begin()), Decode(iter.key()));
+
+    iter.SeekToFirst();
+    ASSERT_TRUE(iter.Valid());
+    ASSERT_EQ(*(keys.begin()), Decode(iter.key()));
+
+    iter.SeekToLast();
+    ASSERT_TRUE(iter.Valid());
+    ASSERT_EQ(*(keys.rbegin()), Decode(iter.key()));
+  }
+
+  // Forward iteration test
+  for (Key i = 0; i < R; i++) {
+    InlineSkipList<TestComparator>::Iterator iter(&list);
+    iter.Seek(Encode(&i));
+
+    // Compare against model iterator
+    std::set<Key>::iterator model_iter = keys.lower_bound(i);
+    for (int j = 0; j < 3; j++) {
+      if (model_iter == keys.end()) {
+        ASSERT_TRUE(!iter.Valid());
+        break;
+      } else {
+        ASSERT_TRUE(iter.Valid());
+        ASSERT_EQ(*model_iter, Decode(iter.key()));
+        ++model_iter;
+        iter.Next();
+      }
+    }
+  }
+
+  // Backward iteration test
+  {
+    InlineSkipList<TestComparator>::Iterator iter(&list);
+    iter.SeekToLast();
+
+    // Compare against model iterator
+    for (std::set<Key>::reverse_iterator model_iter = keys.rbegin();
+         model_iter != keys.rend(); ++model_iter) {
+      ASSERT_TRUE(iter.Valid());
+      ASSERT_EQ(*model_iter, Decode(iter.key()));
+      iter.Prev();
+    }
+    ASSERT_TRUE(!iter.Valid());
+  }
+}
+
+// We want to make sure that with a single writer and multiple
+// concurrent readers (with no synchronization other than when a
+// reader's iterator is created), the reader always observes all the
+// data that was present in the skip list when the iterator was
+// constructor.  Because insertions are happening concurrently, we may
+// also observe new values that were inserted since the iterator was
+// constructed, but we should never miss any values that were present
+// at iterator construction time.
+//
+// We generate multi-part keys:
+//     <key,gen,hash>
+// where:
+//     key is in range [0..K-1]
+//     gen is a generation number for key
+//     hash is hash(key,gen)
+//
+// The insertion code picks a random key, sets gen to be 1 + the last
+// generation number inserted for that key, and sets hash to Hash(key,gen).
+//
+// At the beginning of a read, we snapshot the last inserted
+// generation number for each key.  We then iterate, including random
+// calls to Next() and Seek().  For every key we encounter, we
+// check that it is either expected given the initial snapshot or has
+// been concurrently added since the iterator started.
+class ConcurrentTest {
+ private:
+  static const uint32_t K = 4;
+
+  static uint64_t key(Key key) { return (key >> 40); }
+  static uint64_t gen(Key key) { return (key >> 8) & 0xffffffffu; }
+  static uint64_t hash(Key key) { return key & 0xff; }
+
+  static uint64_t HashNumbers(uint64_t k, uint64_t g) {
+    uint64_t data[2] = {k, g};
+    return Hash(reinterpret_cast<char*>(data), sizeof(data), 0);
+  }
+
+  static Key MakeKey(uint64_t k, uint64_t g) {
+    assert(sizeof(Key) == sizeof(uint64_t));
+    assert(k <= K);  // We sometimes pass K to seek to the end of the skiplist
+    assert(g <= 0xffffffffu);
+    return ((k << 40) | (g << 8) | (HashNumbers(k, g) & 0xff));
+  }
+
+  static bool IsValidKey(Key k) {
+    return hash(k) == (HashNumbers(key(k), gen(k)) & 0xff);
+  }
+
+  static Key RandomTarget(Random* rnd) {
+    switch (rnd->Next() % 10) {
+      case 0:
+        // Seek to beginning
+        return MakeKey(0, 0);
+      case 1:
+        // Seek to end
+        return MakeKey(K, 0);
+      default:
+        // Seek to middle
+        return MakeKey(rnd->Next() % K, 0);
+    }
+  }
+
+  // Per-key generation
+  struct State {
+    std::atomic<int> generation[K];
+    void Set(int k, int v) {
+      generation[k].store(v, std::memory_order_release);
+    }
+    int Get(int k) { return generation[k].load(std::memory_order_acquire); }
+
+    State() {
+      for (unsigned int k = 0; k < K; k++) {
+        Set(k, 0);
+      }
+    }
+  };
+
+  // Current state of the test
+  State current_;
+
+  Arena arena_;
+
+  // InlineSkipList is not protected by mu_.  We just use a single writer
+  // thread to modify it.
+  InlineSkipList<TestComparator> list_;
+
+ public:
+  ConcurrentTest() : list_(TestComparator(), &arena_) {}
+
+  // REQUIRES: External synchronization
+  void WriteStep(Random* rnd) {
+    const uint32_t k = rnd->Next() % K;
+    const int g = current_.Get(k) + 1;
+    const Key new_key = MakeKey(k, g);
+    char* buf = list_.AllocateKey(sizeof(Key));
+    memcpy(buf, &new_key, sizeof(Key));
+    list_.Insert(buf);
+    current_.Set(k, g);
+  }
+
+  void ReadStep(Random* rnd) {
+    // Remember the initial committed state of the skiplist.
+    State initial_state;
+    for (unsigned int k = 0; k < K; k++) {
+      initial_state.Set(k, current_.Get(k));
+    }
+
+    Key pos = RandomTarget(rnd);
+    InlineSkipList<TestComparator>::Iterator iter(&list_);
+    iter.Seek(Encode(&pos));
+    while (true) {
+      Key current;
+      if (!iter.Valid()) {
+        current = MakeKey(K, 0);
+      } else {
+        current = Decode(iter.key());
+        ASSERT_TRUE(IsValidKey(current)) << current;
+      }
+      ASSERT_LE(pos, current) << "should not go backwards";
+
+      // Verify that everything in [pos,current) was not present in
+      // initial_state.
+      while (pos < current) {
+        ASSERT_LT(key(pos), K) << pos;
+
+        // Note that generation 0 is never inserted, so it is ok if
+        // <*,0,*> is missing.
+        ASSERT_TRUE((gen(pos) == 0U) ||
+                    (gen(pos) > static_cast<uint64_t>(initial_state.Get(
+                                    static_cast<int>(key(pos))))))
+            << "key: " << key(pos) << "; gen: " << gen(pos)
+            << "; initgen: " << initial_state.Get(static_cast<int>(key(pos)));
+
+        // Advance to next key in the valid key space
+        if (key(pos) < key(current)) {
+          pos = MakeKey(key(pos) + 1, 0);
+        } else {
+          pos = MakeKey(key(pos), gen(pos) + 1);
+        }
+      }
+
+      if (!iter.Valid()) {
+        break;
+      }
+
+      if (rnd->Next() % 2) {
+        iter.Next();
+        pos = MakeKey(key(pos), gen(pos) + 1);
+      } else {
+        Key new_target = RandomTarget(rnd);
+        if (new_target > pos) {
+          pos = new_target;
+          iter.Seek(Encode(&new_target));
+        }
+      }
+    }
+  }
+};
+const uint32_t ConcurrentTest::K;
+
+// Simple test that does single-threaded testing of the ConcurrentTest
+// scaffolding.
+TEST_F(InlineSkipTest, ConcurrentWithoutThreads) {
+  ConcurrentTest test;
+  Random rnd(test::RandomSeed());
+  for (int i = 0; i < 10000; i++) {
+    test.ReadStep(&rnd);
+    test.WriteStep(&rnd);
+  }
+}
+
+class TestState {
+ public:
+  ConcurrentTest t_;
+  int seed_;
+  std::atomic<bool> quit_flag_;
+
+  enum ReaderState { STARTING, RUNNING, DONE };
+
+  explicit TestState(int s)
+      : seed_(s), quit_flag_(false), state_(STARTING), state_cv_(&mu_) {}
+
+  void Wait(ReaderState s) {
+    mu_.Lock();
+    while (state_ != s) {
+      state_cv_.Wait();
+    }
+    mu_.Unlock();
+  }
+
+  void Change(ReaderState s) {
+    mu_.Lock();
+    state_ = s;
+    state_cv_.Signal();
+    mu_.Unlock();
+  }
+
+ private:
+  port::Mutex mu_;
+  ReaderState state_;
+  port::CondVar state_cv_;
+};
+
+static void ConcurrentReader(void* arg) {
+  TestState* state = reinterpret_cast<TestState*>(arg);
+  Random rnd(state->seed_);
+  int64_t reads = 0;
+  state->Change(TestState::RUNNING);
+  while (!state->quit_flag_.load(std::memory_order_acquire)) {
+    state->t_.ReadStep(&rnd);
+    ++reads;
+  }
+  state->Change(TestState::DONE);
+}
+
+static void RunConcurrent(int run) {
+  const int seed = test::RandomSeed() + (run * 100);
+  Random rnd(seed);
+  const int N = 1000;
+  const int kSize = 1000;
+  for (int i = 0; i < N; i++) {
+    if ((i % 100) == 0) {
+      fprintf(stderr, "Run %d of %d\n", i, N);
+    }
+    TestState state(seed + 1);
+    Env::Default()->Schedule(ConcurrentReader, &state);
+    state.Wait(TestState::RUNNING);
+    for (int k = 0; k < kSize; k++) {
+      state.t_.WriteStep(&rnd);
+    }
+    state.quit_flag_.store(true, std::memory_order_release);
+    state.Wait(TestState::DONE);
+  }
+}
+
+TEST_F(InlineSkipTest, Concurrent1) { RunConcurrent(1); }
+TEST_F(InlineSkipTest, Concurrent2) { RunConcurrent(2); }
+TEST_F(InlineSkipTest, Concurrent3) { RunConcurrent(3); }
+TEST_F(InlineSkipTest, Concurrent4) { RunConcurrent(4); }
+TEST_F(InlineSkipTest, Concurrent5) { RunConcurrent(5); }
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

db/internal_stats.cc

@@ -130,6 +130,8 @@ static const std::string aggregated_table_properties =
     "aggregated-table-properties";
 static const std::string aggregated_table_properties_at_level =
     aggregated_table_properties + "-at-level";
+static const std::string num_running_compactions = "num-running-compactions";
+static const std::string num_running_flushes = "num-running-flushes";
 
 const std::string DB::Properties::kNumFilesAtLevelPrefix =
                       rocksdb_prefix + num_files_at_level_prefix;
@@ -143,6 +145,10 @@ const std::string DB::Properties::kMemTableFlushPending =
                       rocksdb_prefix + mem_table_flush_pending;
 const std::string DB::Properties::kCompactionPending =
                       rocksdb_prefix + compaction_pending;
+const std::string DB::Properties::kNumRunningCompactions =
+    rocksdb_prefix + num_running_compactions;
+const std::string DB::Properties::kNumRunningFlushes =
+    rocksdb_prefix + num_running_flushes;
 const std::string DB::Properties::kBackgroundErrors =
                       rocksdb_prefix + background_errors;
 const std::string DB::Properties::kCurSizeActiveMemTable =
@@ -260,6 +266,10 @@ DBPropertyType GetPropertyType(const Slice& property, bool* is_int_property,
     return kTotalSstFilesSize;
   } else if (in == estimate_pending_comp_bytes) {
     return kEstimatePendingCompactionBytes;
+  } else if (in == num_running_flushes) {
+    return kNumRunningFlushes;
+  } else if (in == num_running_compactions) {
+    return kNumRunningCompactions;
   }
   return kUnknown;
 }
@@ -388,11 +398,17 @@ bool InternalStats::GetIntProperty(DBPropertyType property_type,
       // Return number of mem tables that are ready to flush (made immutable)
       *value = (cfd_->imm()->IsFlushPending() ? 1 : 0);
       return true;
+    case kNumRunningFlushes:
+      *value = db->num_running_flushes();
+      return true;
     case kCompactionPending:
       // 1 if the system already determines at least one compaction is needed.
       // 0 otherwise,
       *value = (cfd_->compaction_picker()->NeedsCompaction(vstorage) ? 1 : 0);
       return true;
+    case kNumRunningCompactions:
+      *value = db->num_running_compactions_;
+      return true;
     case kBackgroundErrors:
       // Accumulated number of  errors in background flushes or compactions.
       *value = GetBackgroundErrorCount();

db/internal_stats.h

@@ -38,7 +38,10 @@ enum DBPropertyType : uint32_t {
                                  // in memory that have already been flushed
   kMemtableFlushPending,         // Return 1 if mem table flushing is pending,
                                  // otherwise 0.
+  kNumRunningFlushes,      // Return the number of currently running flushes.
   kCompactionPending,      // Return 1 if a compaction is pending. Otherwise 0.
+  kNumRunningCompactions,  // Return the number of currently running
+                           // compactions.
   kBackgroundErrors,       // Return accumulated background errors encountered.
   kCurSizeActiveMemTable,  // Return current size of the active memtable
   kCurSizeAllMemTables,    // Return current size of unflushed
@@ -49,10 +52,10 @@ enum DBPropertyType : uint32_t {
                                    // memtable.
   kNumEntriesInImmutableMemtable,  // Return sum of number of entries in all
                                    // the immutable mem tables.
-  kNumDeletesInMutableMemtable,    // Return number of entries in the mutable
-                                   // memtable.
-  kNumDeletesInImmutableMemtable,  // Return sum of number of deletes in all
-                                   // the immutable mem tables.
+  kNumDeletesInMutableMemtable,    // Return number of deletion entries in the
+                                   // mutable memtable.
+  kNumDeletesInImmutableMemtable,  // Return the total number of deletion
+                                   // entries in all the immutable mem tables.
   kEstimatedNumKeys,  // Estimated total number of keys in the database.
   kEstimatedUsageByTableReaders,  // Estimated memory by table readers.
   kIsFileDeletionEnabled,         // Equals disable_delete_obsolete_files_,

db/listener_test.cc

@@ -8,6 +8,7 @@
 #include "db/filename.h"
 #include "db/version_set.h"
 #include "db/write_batch_internal.h"
+#include "memtable/hash_linklist_rep.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/db.h"
@@ -22,7 +23,6 @@
 #include "table/block_based_table_factory.h"
 #include "table/plain_table_factory.h"
 #include "util/hash.h"
-#include "util/hash_linklist_rep.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
 #include "util/rate_limiter.h"
@@ -152,6 +152,40 @@ class EventListenerTest : public testing::Test {
   std::vector<ColumnFamilyHandle*> handles_;
 };
 
+struct TestPropertiesCollector : public rocksdb::TablePropertiesCollector {
+  virtual rocksdb::Status AddUserKey(const rocksdb::Slice& key,
+                                     const rocksdb::Slice& value,
+                                     rocksdb::EntryType type,
+                                     rocksdb::SequenceNumber seq,
+                                     uint64_t file_size) override {
+    return Status::OK();
+  }
+  virtual rocksdb::Status Finish(
+      rocksdb::UserCollectedProperties* properties) override {
+    properties->insert({"0", "1"});
+    return Status::OK();
+  }
+
+  virtual const char* Name() const override {
+    return "TestTablePropertiesCollector";
+  }
+
+  rocksdb::UserCollectedProperties GetReadableProperties() const override {
+    rocksdb::UserCollectedProperties ret;
+    ret["2"] = "3";
+    return ret;
+  }
+};
+
+class TestPropertiesCollectorFactory : public TablePropertiesCollectorFactory {
+ public:
+  virtual TablePropertiesCollector* CreateTablePropertiesCollector(
+      TablePropertiesCollectorFactory::Context context) override {
+    return new TestPropertiesCollector;
+  }
+  const char* Name() const override { return "TestTablePropertiesCollector"; }
+};
+
 class TestCompactionListener : public EventListener {
  public:
   void OnCompactionCompleted(DB *db, const CompactionJobInfo& ci) override {
@@ -161,6 +195,16 @@ class TestCompactionListener : public EventListener {
     ASSERT_GT(ci.output_files.size(), 0U);
     ASSERT_EQ(db->GetEnv()->GetThreadID(), ci.thread_id);
     ASSERT_GT(ci.thread_id, 0U);
+
+    for (auto fl : {ci.input_files, ci.output_files}) {
+      for (auto fn : fl) {
+        auto it = ci.table_properties.find(fn);
+        ASSERT_NE(it, ci.table_properties.end());
+        auto tp = it->second;
+        ASSERT_TRUE(tp != nullptr);
+        ASSERT_EQ(tp->user_collected_properties.find("0")->second, "1");
+      }
+    }
   }
 
   std::vector<DB*> compacted_dbs_;
@@ -186,6 +230,8 @@ TEST_F(EventListenerTest, OnSingleDBCompactionTest) {
   options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
   options.level0_file_num_compaction_trigger = kNumL0Files;
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());
 
   TestCompactionListener* listener = new TestCompactionListener();
   options.listeners.emplace_back(listener);
@@ -274,6 +320,8 @@ class TestFlushListener : public EventListener {
     ASSERT_EQ(prev_fc_info_.file_path, info.file_path);
     ASSERT_EQ(db->GetEnv()->GetThreadID(), info.thread_id);
     ASSERT_GT(info.thread_id, 0U);
+    ASSERT_EQ(info.table_properties.user_collected_properties.find("0")->second,
+              "1");
   }
 
   std::vector<std::string> flushed_column_family_names_;
@@ -299,6 +347,8 @@ TEST_F(EventListenerTest, OnSingleDBFlushTest) {
   std::vector<std::string> cf_names = {
       "pikachu", "ilya", "muromec", "dobrynia",
       "nikitich", "alyosha", "popovich"};
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());
   CreateAndReopenWithCF(cf_names, &options);
 
   ASSERT_OK(Put(1, "pikachu", std::string(90000, 'p')));
@@ -330,6 +380,8 @@ TEST_F(EventListenerTest, MultiCF) {
 #endif  // ROCKSDB_USING_THREAD_STATUS
   TestFlushListener* listener = new TestFlushListener(options.env);
   options.listeners.emplace_back(listener);
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());
   std::vector<std::string> cf_names = {
       "pikachu", "ilya", "muromec", "dobrynia",
       "nikitich", "alyosha", "popovich"};
@@ -360,6 +412,8 @@ TEST_F(EventListenerTest, MultiDBMultiListeners) {
 #if ROCKSDB_USING_THREAD_STATUS
   options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());
   std::vector<TestFlushListener*> listeners;
   const int kNumDBs = 5;
   const int kNumListeners = 10;
@@ -454,6 +508,8 @@ TEST_F(EventListenerTest, DisableBGCompaction) {
   options.compaction_style = kCompactionStyleNone;
   options.compression = kNoCompression;
   options.write_buffer_size = 100000;  // Small write buffer
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());
 
   CreateAndReopenWithCF({"pikachu"}, &options);
   ColumnFamilyMetaData cf_meta;
@@ -477,4 +533,3 @@ int main(int argc, char** argv) {
   ::testing::InitGoogleTest(&argc, argv);
   return RUN_ALL_TESTS();
 }
-

db/log_format.h

@@ -22,14 +22,24 @@ enum RecordType {
   // For fragments
   kFirstType = 2,
   kMiddleType = 3,
-  kLastType = 4
+  kLastType = 4,
+
+  // For recycled log files
+  kRecyclableFullType = 5,
+  kRecyclableFirstType = 6,
+  kRecyclableMiddleType = 7,
+  kRecyclableLastType = 8,
 };
-static const int kMaxRecordType = kLastType;
+static const int kMaxRecordType = kRecyclableLastType;
 
 static const unsigned int kBlockSize = 32768;
 
 // Header is checksum (4 bytes), type (1 byte), length (2 bytes).
 static const int kHeaderSize = 4 + 1 + 2;
 
+// Recyclable header is checksum (4 bytes), type (1 byte), log number
+// (4 bytes), length (2 bytes).
+static const int kRecyclableHeaderSize = 4 + 1 + 4 + 2;
+
 }  // namespace log
 }  // namespace rocksdb

db/log_reader.cc

@@ -21,9 +21,12 @@ namespace log {
 Reader::Reporter::~Reporter() {
 }
 
-Reader::Reader(unique_ptr<SequentialFileReader>&& _file, Reporter* reporter,
-               bool checksum, uint64_t initial_offset)
-    : file_(std::move(_file)),
+Reader::Reader(std::shared_ptr<Logger> info_log,
+	       unique_ptr<SequentialFileReader>&& _file,
+               Reporter* reporter, bool checksum, uint64_t initial_offset,
+               uint64_t log_num)
+    : info_log_(info_log),
+      file_(std::move(_file)),
       reporter_(reporter),
       checksum_(checksum),
       backing_store_(new char[kBlockSize]),
@@ -33,7 +36,8 @@ Reader::Reader(unique_ptr<SequentialFileReader>&& _file, Reporter* reporter,
       eof_offset_(0),
       last_record_offset_(0),
       end_of_buffer_offset_(0),
-      initial_offset_(initial_offset) {}
+      initial_offset_(initial_offset),
+      log_number_(log_num) {}
 
 Reader::~Reader() {
   delete[] backing_store_;
@@ -62,8 +66,15 @@ bool Reader::SkipToInitialBlock() {
   return true;
 }
 
+// For kAbsoluteConsistency, on clean shutdown we don't expect any error
+// in the log files.  For other modes, we can ignore only incomplete records
+// in the last log file, which are presumably due to a write in progress
+// during restart (or from log recycling).
+//
+// TODO krad: Evaluate if we need to move to a more strict mode where we
+// restrict the inconsistency to only the last log
 bool Reader::ReadRecord(Slice* record, std::string* scratch,
-                        const bool report_eof_inconsistency) {
+                        WALRecoveryMode wal_recovery_mode) {
   if (last_record_offset_ < initial_offset_) {
     if (!SkipToInitialBlock()) {
       return false;
@@ -80,10 +91,11 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
   Slice fragment;
   while (true) {
     uint64_t physical_record_offset = end_of_buffer_offset_ - buffer_.size();
-    const unsigned int record_type =
-        ReadPhysicalRecord(&fragment, report_eof_inconsistency);
+    size_t drop_size;
+    const unsigned int record_type = ReadPhysicalRecord(&fragment, &drop_size);
     switch (record_type) {
       case kFullType:
+      case kRecyclableFullType:
         if (in_fragmented_record && !scratch->empty()) {
           // Handle bug in earlier versions of log::Writer where
           // it could emit an empty kFirstType record at the tail end
@@ -98,6 +110,7 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
         return true;
 
       case kFirstType:
+      case kRecyclableFirstType:
         if (in_fragmented_record && !scratch->empty()) {
           // Handle bug in earlier versions of log::Writer where
           // it could emit an empty kFirstType record at the tail end
@@ -111,6 +124,7 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
         break;
 
       case kMiddleType:
+      case kRecyclableMiddleType:
         if (!in_fragmented_record) {
           ReportCorruption(fragment.size(),
                            "missing start of fragmented record(1)");
@@ -120,6 +134,7 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
         break;
 
       case kLastType:
+      case kRecyclableLastType:
         if (!in_fragmented_record) {
           ReportCorruption(fragment.size(),
                            "missing start of fragmented record(2)");
@@ -131,9 +146,17 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
         }
         break;
 
+      case kBadHeader:
+        if (wal_recovery_mode == WALRecoveryMode::kAbsoluteConsistency) {
+          // in clean shutdown we don't expect any error in the log files
+          ReportCorruption(drop_size, "truncated header");
+        }
+      // fall-thru
+
       case kEof:
         if (in_fragmented_record) {
-          if (report_eof_inconsistency) {
+          if (wal_recovery_mode == WALRecoveryMode::kAbsoluteConsistency) {
+            // in clean shutdown we don't expect any error in the log files
             ReportCorruption(scratch->size(), "error reading trailing data");
           }
           // This can be caused by the writer dying immediately after
@@ -143,6 +166,23 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
         }
         return false;
 
+      case kOldRecord:
+        if (wal_recovery_mode != WALRecoveryMode::kSkipAnyCorruptedRecords) {
+          // Treat a record from a previous instance of the log as EOF.
+          if (in_fragmented_record) {
+            if (wal_recovery_mode == WALRecoveryMode::kAbsoluteConsistency) {
+              // in clean shutdown we don't expect any error in the log files
+              ReportCorruption(scratch->size(), "error reading trailing data");
+            }
+            // This can be caused by the writer dying immediately after
+            //  writing a physical record but before completing the next; don't
+            //  treat it as a corruption, just ignore the entire logical record.
+            scratch->clear();
+          }
+          return false;
+        }
+      // fall-thru
+
       case kBadRecord:
         if (in_fragmented_record) {
           ReportCorruption(scratch->size(), "error in middle of record");
@@ -244,36 +284,49 @@ void Reader::ReportDrop(size_t bytes, const Status& reason) {
   }
 }
 
-unsigned int Reader::ReadPhysicalRecord(Slice* result,
-                                        const bool report_eof_inconsistency) {
+bool Reader::ReadMore(size_t* drop_size, int *error) {
+  if (!eof_ && !read_error_) {
+    // Last read was a full read, so this is a trailer to skip
+    buffer_.clear();
+    Status status = file_->Read(kBlockSize, &buffer_, backing_store_);
+    end_of_buffer_offset_ += buffer_.size();
+    if (!status.ok()) {
+      buffer_.clear();
+      ReportDrop(kBlockSize, status);
+      read_error_ = true;
+      *error = kEof;
+      return false;
+    } else if (buffer_.size() < (size_t)kBlockSize) {
+      eof_ = true;
+      eof_offset_ = buffer_.size();
+    }
+    return true;
+  } else {
+    // Note that if buffer_ is non-empty, we have a truncated header at the
+    //  end of the file, which can be caused by the writer crashing in the
+    //  middle of writing the header. Unless explicitly requested we don't
+    //  considering this an error, just report EOF.
+    if (buffer_.size()) {
+      *drop_size = buffer_.size();
+      buffer_.clear();
+      *error = kBadHeader;
+      return false;
+    }
+    buffer_.clear();
+    *error = kEof;
+    return false;
+  }
+}
+
+unsigned int Reader::ReadPhysicalRecord(Slice* result, size_t* drop_size) {
   while (true) {
+    // We need at least the minimum header size
     if (buffer_.size() < (size_t)kHeaderSize) {
-      if (!eof_ && !read_error_) {
-        // Last read was a full read, so this is a trailer to skip
-        buffer_.clear();
-        Status status = file_->Read(kBlockSize, &buffer_, backing_store_);
-        end_of_buffer_offset_ += buffer_.size();
-        if (!status.ok()) {
-          buffer_.clear();
-          ReportDrop(kBlockSize, status);
-          read_error_ = true;
-          return kEof;
-        } else if (buffer_.size() < (size_t)kBlockSize) {
-          eof_ = true;
-          eof_offset_ = buffer_.size();
-        }
-        continue;
-      } else {
-        // Note that if buffer_ is non-empty, we have a truncated header at the
-        //  end of the file, which can be caused by the writer crashing in the
-        //  middle of writing the header. Unless explicitly requested we don't
-        //  considering this an error, just report EOF.
-        if (buffer_.size() && report_eof_inconsistency) {
-          ReportCorruption(buffer_.size(), "truncated header");
-        }
-        buffer_.clear();
-        return kEof;
+      int r;
+      if (!ReadMore(drop_size, &r)) {
+	return r;
       }
+      continue;
     }
 
     // Parse the header
@@ -282,18 +335,34 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result,
     const uint32_t b = static_cast<uint32_t>(header[5]) & 0xff;
     const unsigned int type = header[6];
     const uint32_t length = a | (b << 8);
-    if (kHeaderSize + length > buffer_.size()) {
-      size_t drop_size = buffer_.size();
+    int header_size = kHeaderSize;
+    if (type >= kRecyclableFullType && type <= kRecyclableLastType) {
+      header_size = kRecyclableHeaderSize;
+      // We need enough for the larger header
+      if (buffer_.size() < (size_t)kRecyclableHeaderSize) {
+	int r;
+	if (!ReadMore(drop_size, &r)) {
+          return r;
+        }
+	continue;
+      }
+      const uint32_t log_num = DecodeFixed32(header + 7);
+      if (log_num != log_number_) {
+        return kOldRecord;
+      }
+    }
+    if (header_size + length > buffer_.size()) {
+      *drop_size = buffer_.size();
       buffer_.clear();
       if (!eof_) {
-        ReportCorruption(drop_size, "bad record length");
+        ReportCorruption(*drop_size, "bad record length");
         return kBadRecord;
       }
       // If the end of the file has been reached without reading |length| bytes
       // of payload, assume the writer died in the middle of writing the record.
       // Don't report a corruption unless requested.
-      if (drop_size && report_eof_inconsistency) {
-        ReportCorruption(drop_size, "truncated header");
+      if (*drop_size) {
+        return kBadHeader;
       }
       return kEof;
     }
@@ -311,29 +380,29 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result,
     // Check crc
     if (checksum_) {
       uint32_t expected_crc = crc32c::Unmask(DecodeFixed32(header));
-      uint32_t actual_crc = crc32c::Value(header + 6, 1 + length);
+      uint32_t actual_crc = crc32c::Value(header + 6, length + header_size - 6);
       if (actual_crc != expected_crc) {
         // Drop the rest of the buffer since "length" itself may have
         // been corrupted and if we trust it, we could find some
         // fragment of a real log record that just happens to look
         // like a valid log record.
-        size_t drop_size = buffer_.size();
+        *drop_size = buffer_.size();
         buffer_.clear();
-        ReportCorruption(drop_size, "checksum mismatch");
+        ReportCorruption(*drop_size, "checksum mismatch");
         return kBadRecord;
       }
     }
 
-    buffer_.remove_prefix(kHeaderSize + length);
+    buffer_.remove_prefix(header_size + length);
 
     // Skip physical record that started before initial_offset_
-    if (end_of_buffer_offset_ - buffer_.size() - kHeaderSize - length <
+    if (end_of_buffer_offset_ - buffer_.size() - header_size - length <
         initial_offset_) {
       result->clear();
       return kBadRecord;
     }
 
-    *result = Slice(header + kHeaderSize, length);
+    *result = Slice(header + header_size, length);
     return type;
   }
 }

db/log_reader.h

@@ -14,10 +14,12 @@
 #include "db/log_format.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
+#include "rocksdb/options.h"
 
 namespace rocksdb {
 
 class SequentialFileReader;
+class Logger;
 using std::unique_ptr;
 
 namespace log {
@@ -51,8 +53,10 @@ class Reader {
   //
   // The Reader will start reading at the first record located at physical
   // position >= initial_offset within the file.
-  Reader(unique_ptr<SequentialFileReader>&& file, Reporter* reporter,
-         bool checksum, uint64_t initial_offset);
+  Reader(std::shared_ptr<Logger> info_log,
+	 unique_ptr<SequentialFileReader>&& file,
+         Reporter* reporter, bool checksum, uint64_t initial_offset,
+         uint64_t log_num);
 
   ~Reader();
 
@@ -62,7 +66,8 @@ class Reader {
   // will only be valid until the next mutating operation on this
   // reader or the next mutation to *scratch.
   bool ReadRecord(Slice* record, std::string* scratch,
-                  bool report_eof_inconsistency = false);
+                  WALRecoveryMode wal_recovery_mode =
+                      WALRecoveryMode::kTolerateCorruptedTailRecords);
 
   // Returns the physical offset of the last record returned by ReadRecord.
   //
@@ -84,6 +89,7 @@ class Reader {
   SequentialFileReader* file() { return file_.get(); }
 
  private:
+  std::shared_ptr<Logger> info_log_;
   const unique_ptr<SequentialFileReader> file_;
   Reporter* const reporter_;
   bool const checksum_;
@@ -104,6 +110,9 @@ class Reader {
   // Offset at which to start looking for the first record to return
   uint64_t const initial_offset_;
 
+  // which log number this is
+  uint64_t const log_number_;
+
   // Extend record types with the following special values
   enum {
     kEof = kMaxRecordType + 1,
@@ -112,7 +121,11 @@ class Reader {
     // * The record has an invalid CRC (ReadPhysicalRecord reports a drop)
     // * The record is a 0-length record (No drop is reported)
     // * The record is below constructor's initial_offset (No drop is reported)
-    kBadRecord = kMaxRecordType + 2
+    kBadRecord = kMaxRecordType + 2,
+    // Returned when we fail to read a valid header.
+    kBadHeader = kMaxRecordType + 3,
+    // Returned when we read an old record from a previous user of the log.
+    kOldRecord = kMaxRecordType + 4,
   };
 
   // Skips all blocks that are completely before "initial_offset_".
@@ -121,8 +134,10 @@ class Reader {
   bool SkipToInitialBlock();
 
   // Return type, or one of the preceding special values
-  unsigned int ReadPhysicalRecord(Slice* result,
-                                  bool report_eof_inconsistency = false);
+  unsigned int ReadPhysicalRecord(Slice* result, size_t* drop_size);
+
+  // Read some more
+  bool ReadMore(size_t* drop_size, int *error);
 
   // Reports dropped bytes to the reporter.
   // buffer_ must be updated to remove the dropped bytes prior to invocation.

db/log_test.cc

@@ -43,7 +43,7 @@ static std::string RandomSkewedString(int i, Random* rnd) {
   return BigString(NumberString(i), rnd->Skewed(17));
 }
 
-class LogTest : public testing::Test {
+class LogTest : public ::testing::TestWithParam<int> {
  private:
   class StringSource : public SequentialFile {
    public:
@@ -153,19 +153,26 @@ class LogTest : public testing::Test {
 
   // Record metadata for testing initial offset functionality
   static size_t initial_offset_record_sizes_[];
-  static uint64_t initial_offset_last_record_offsets_[];
+  uint64_t initial_offset_last_record_offsets_[4];
 
  public:
   LogTest()
       : reader_contents_(),
-        dest_holder_(
-            test::GetWritableFileWriter(
-              new test::StringSink(&reader_contents_))),
+        dest_holder_(test::GetWritableFileWriter(
+            new test::StringSink(&reader_contents_))),
         source_holder_(
             test::GetSequentialFileReader(new StringSource(reader_contents_))),
-        writer_(std::move(dest_holder_)),
-        reader_(std::move(source_holder_), &report_, true /*checksum*/,
-                0 /*initial_offset*/) {}
+        writer_(std::move(dest_holder_), 123, GetParam()),
+        reader_(NULL, std::move(source_holder_), &report_, true /*checksum*/,
+                0 /*initial_offset*/, 123) {
+    int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+    initial_offset_last_record_offsets_[0] = 0;
+    initial_offset_last_record_offsets_[1] = header_size + 10000;
+    initial_offset_last_record_offsets_[2] = 2 * (header_size + 10000);
+    initial_offset_last_record_offsets_[3] = 2 * (header_size + 10000) +
+                                             (2 * log::kBlockSize - 1000) +
+                                             3 * header_size;
+  }
 
   void Write(const std::string& msg) {
     writer_.AddRecord(Slice(msg));
@@ -175,10 +182,11 @@ class LogTest : public testing::Test {
     return dest_contents().size();
   }
 
-  std::string Read(const bool report_eof_inconsistency = false) {
+  std::string Read(const WALRecoveryMode wal_recovery_mode =
+                       WALRecoveryMode::kTolerateCorruptedTailRecords) {
     std::string scratch;
     Slice record;
-    if (reader_.ReadRecord(&record, &scratch, report_eof_inconsistency)) {
+    if (reader_.ReadRecord(&record, &scratch, wal_recovery_mode)) {
       return record.ToString();
     } else {
       return "EOF";
@@ -200,9 +208,11 @@ class LogTest : public testing::Test {
     dest->Drop(bytes);
   }
 
-  void FixChecksum(int header_offset, int len) {
+  void FixChecksum(int header_offset, int len, bool recyclable) {
     // Compute crc of type/len/data
-    uint32_t crc = crc32c::Value(&dest_contents()[header_offset+6], 1 + len);
+    int header_size = recyclable ? kRecyclableHeaderSize : kHeaderSize;
+    uint32_t crc = crc32c::Value(&dest_contents()[header_offset + 6],
+                                 header_size - 6 + len);
     crc = crc32c::Mask(crc);
     EncodeFixed32(&dest_contents()[header_offset], crc);
   }
@@ -259,8 +269,8 @@ class LogTest : public testing::Test {
     unique_ptr<SequentialFileReader> file_reader(
         test::GetSequentialFileReader(new StringSource(reader_contents_)));
     unique_ptr<Reader> offset_reader(
-        new Reader(std::move(file_reader), &report_, true /*checksum*/,
-                   WrittenBytes() + offset_past_end));
+        new Reader(NULL, std::move(file_reader), &report_,
+                   true /*checksum*/, WrittenBytes() + offset_past_end, 123));
     Slice record;
     std::string scratch;
     ASSERT_TRUE(!offset_reader->ReadRecord(&record, &scratch));
@@ -271,8 +281,9 @@ class LogTest : public testing::Test {
     WriteInitialOffsetLog();
     unique_ptr<SequentialFileReader> file_reader(
         test::GetSequentialFileReader(new StringSource(reader_contents_)));
-    unique_ptr<Reader> offset_reader(new Reader(
-        std::move(file_reader), &report_, true /*checksum*/, initial_offset));
+    unique_ptr<Reader> offset_reader(
+        new Reader(NULL, std::move(file_reader), &report_,
+                   true /*checksum*/, initial_offset, 123));
     Slice record;
     std::string scratch;
     ASSERT_TRUE(offset_reader->ReadRecord(&record, &scratch));
@@ -291,16 +302,9 @@ size_t LogTest::initial_offset_record_sizes_[] =
      2 * log::kBlockSize - 1000,  // Span three blocks
      1};
 
-uint64_t LogTest::initial_offset_last_record_offsets_[] =
-    {0,
-     kHeaderSize + 10000,
-     2 * (kHeaderSize + 10000),
-     2 * (kHeaderSize + 10000) +
-         (2 * log::kBlockSize - 1000) + 3 * kHeaderSize};
+TEST_P(LogTest, Empty) { ASSERT_EQ("EOF", Read()); }
 
-TEST_F(LogTest, Empty) { ASSERT_EQ("EOF", Read()); }
-
-TEST_F(LogTest, ReadWrite) {
+TEST_P(LogTest, ReadWrite) {
   Write("foo");
   Write("bar");
   Write("");
@@ -313,7 +317,7 @@ TEST_F(LogTest, ReadWrite) {
   ASSERT_EQ("EOF", Read());  // Make sure reads at eof work
 }
 
-TEST_F(LogTest, ManyBlocks) {
+TEST_P(LogTest, ManyBlocks) {
   for (int i = 0; i < 100000; i++) {
     Write(NumberString(i));
   }
@@ -323,7 +327,7 @@ TEST_F(LogTest, ManyBlocks) {
   ASSERT_EQ("EOF", Read());
 }
 
-TEST_F(LogTest, Fragmentation) {
+TEST_P(LogTest, Fragmentation) {
   Write("small");
   Write(BigString("medium", 50000));
   Write(BigString("large", 100000));
@@ -333,11 +337,12 @@ TEST_F(LogTest, Fragmentation) {
   ASSERT_EQ("EOF", Read());
 }
 
-TEST_F(LogTest, MarginalTrailer) {
+TEST_P(LogTest, MarginalTrailer) {
   // Make a trailer that is exactly the same length as an empty record.
-  const int n = kBlockSize - 2*kHeaderSize;
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int n = kBlockSize - 2 * header_size;
   Write(BigString("foo", n));
-  ASSERT_EQ((unsigned int)(kBlockSize - kHeaderSize), WrittenBytes());
+  ASSERT_EQ((unsigned int)(kBlockSize - header_size), WrittenBytes());
   Write("");
   Write("bar");
   ASSERT_EQ(BigString("foo", n), Read());
@@ -346,11 +351,12 @@ TEST_F(LogTest, MarginalTrailer) {
   ASSERT_EQ("EOF", Read());
 }
 
-TEST_F(LogTest, MarginalTrailer2) {
+TEST_P(LogTest, MarginalTrailer2) {
   // Make a trailer that is exactly the same length as an empty record.
-  const int n = kBlockSize - 2*kHeaderSize;
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int n = kBlockSize - 2 * header_size;
   Write(BigString("foo", n));
-  ASSERT_EQ((unsigned int)(kBlockSize - kHeaderSize), WrittenBytes());
+  ASSERT_EQ((unsigned int)(kBlockSize - header_size), WrittenBytes());
   Write("bar");
   ASSERT_EQ(BigString("foo", n), Read());
   ASSERT_EQ("bar", Read());
@@ -359,10 +365,11 @@ TEST_F(LogTest, MarginalTrailer2) {
   ASSERT_EQ("", ReportMessage());
 }
 
-TEST_F(LogTest, ShortTrailer) {
-  const int n = kBlockSize - 2*kHeaderSize + 4;
+TEST_P(LogTest, ShortTrailer) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int n = kBlockSize - 2 * header_size + 4;
   Write(BigString("foo", n));
-  ASSERT_EQ((unsigned int)(kBlockSize - kHeaderSize + 4), WrittenBytes());
+  ASSERT_EQ((unsigned int)(kBlockSize - header_size + 4), WrittenBytes());
   Write("");
   Write("bar");
   ASSERT_EQ(BigString("foo", n), Read());
@@ -371,15 +378,16 @@ TEST_F(LogTest, ShortTrailer) {
   ASSERT_EQ("EOF", Read());
 }
 
-TEST_F(LogTest, AlignedEof) {
-  const int n = kBlockSize - 2*kHeaderSize + 4;
+TEST_P(LogTest, AlignedEof) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int n = kBlockSize - 2 * header_size + 4;
   Write(BigString("foo", n));
-  ASSERT_EQ((unsigned int)(kBlockSize - kHeaderSize + 4), WrittenBytes());
+  ASSERT_EQ((unsigned int)(kBlockSize - header_size + 4), WrittenBytes());
   ASSERT_EQ(BigString("foo", n), Read());
   ASSERT_EQ("EOF", Read());
 }
 
-TEST_F(LogTest, RandomRead) {
+TEST_P(LogTest, RandomRead) {
   const int N = 500;
   Random write_rnd(301);
   for (int i = 0; i < N; i++) {
@@ -394,7 +402,7 @@ TEST_F(LogTest, RandomRead) {
 
 // Tests of all the error paths in log_reader.cc follow:
 
-TEST_F(LogTest, ReadError) {
+TEST_P(LogTest, ReadError) {
   Write("foo");
   ForceError();
   ASSERT_EQ("EOF", Read());
@@ -402,17 +410,17 @@ TEST_F(LogTest, ReadError) {
   ASSERT_EQ("OK", MatchError("read error"));
 }
 
-TEST_F(LogTest, BadRecordType) {
+TEST_P(LogTest, BadRecordType) {
   Write("foo");
   // Type is stored in header[6]
   IncrementByte(6, 100);
-  FixChecksum(0, 3);
+  FixChecksum(0, 3, false);
   ASSERT_EQ("EOF", Read());
   ASSERT_EQ(3U, DroppedBytes());
   ASSERT_EQ("OK", MatchError("unknown record type"));
 }
 
-TEST_F(LogTest, TruncatedTrailingRecordIsIgnored) {
+TEST_P(LogTest, TruncatedTrailingRecordIsIgnored) {
   Write("foo");
   ShrinkSize(4);   // Drop all payload as well as a header byte
   ASSERT_EQ("EOF", Read());
@@ -421,17 +429,18 @@ TEST_F(LogTest, TruncatedTrailingRecordIsIgnored) {
   ASSERT_EQ("", ReportMessage());
 }
 
-TEST_F(LogTest, TruncatedTrailingRecordIsNotIgnored) {
+TEST_P(LogTest, TruncatedTrailingRecordIsNotIgnored) {
   Write("foo");
   ShrinkSize(4);  // Drop all payload as well as a header byte
-  ASSERT_EQ("EOF", Read(/*report_eof_inconsistency*/ true));
+  ASSERT_EQ("EOF", Read(WALRecoveryMode::kAbsoluteConsistency));
   // Truncated last record is ignored, not treated as an error
   ASSERT_GT(DroppedBytes(), 0U);
   ASSERT_EQ("OK", MatchError("Corruption: truncated header"));
 }
 
-TEST_F(LogTest, BadLength) {
-  const int kPayloadSize = kBlockSize - kHeaderSize;
+TEST_P(LogTest, BadLength) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int kPayloadSize = kBlockSize - header_size;
   Write(BigString("bar", kPayloadSize));
   Write("foo");
   // Least significant size byte is stored in header[4].
@@ -441,7 +450,7 @@ TEST_F(LogTest, BadLength) {
   ASSERT_EQ("OK", MatchError("bad record length"));
 }
 
-TEST_F(LogTest, BadLengthAtEndIsIgnored) {
+TEST_P(LogTest, BadLengthAtEndIsIgnored) {
   Write("foo");
   ShrinkSize(1);
   ASSERT_EQ("EOF", Read());
@@ -449,63 +458,63 @@ TEST_F(LogTest, BadLengthAtEndIsIgnored) {
   ASSERT_EQ("", ReportMessage());
 }
 
-TEST_F(LogTest, BadLengthAtEndIsNotIgnored) {
+TEST_P(LogTest, BadLengthAtEndIsNotIgnored) {
   Write("foo");
   ShrinkSize(1);
-  ASSERT_EQ("EOF", Read(/*report_eof_inconsistency=*/true));
+  ASSERT_EQ("EOF", Read(WALRecoveryMode::kAbsoluteConsistency));
   ASSERT_GT(DroppedBytes(), 0U);
   ASSERT_EQ("OK", MatchError("Corruption: truncated header"));
 }
 
-TEST_F(LogTest, ChecksumMismatch) {
-  Write("foo");
-  IncrementByte(0, 10);
+TEST_P(LogTest, ChecksumMismatch) {
+  Write("foooooo");
+  IncrementByte(0, 14);
   ASSERT_EQ("EOF", Read());
-  ASSERT_EQ(10U, DroppedBytes());
+  ASSERT_EQ(14U + 4 * !!GetParam(), DroppedBytes());
   ASSERT_EQ("OK", MatchError("checksum mismatch"));
 }
 
-TEST_F(LogTest, UnexpectedMiddleType) {
+TEST_P(LogTest, UnexpectedMiddleType) {
   Write("foo");
-  SetByte(6, kMiddleType);
-  FixChecksum(0, 3);
+  SetByte(6, GetParam() ? kRecyclableMiddleType : kMiddleType);
+  FixChecksum(0, 3, !!GetParam());
   ASSERT_EQ("EOF", Read());
   ASSERT_EQ(3U, DroppedBytes());
   ASSERT_EQ("OK", MatchError("missing start"));
 }
 
-TEST_F(LogTest, UnexpectedLastType) {
+TEST_P(LogTest, UnexpectedLastType) {
   Write("foo");
-  SetByte(6, kLastType);
-  FixChecksum(0, 3);
+  SetByte(6, GetParam() ? kRecyclableLastType : kLastType);
+  FixChecksum(0, 3, !!GetParam());
   ASSERT_EQ("EOF", Read());
   ASSERT_EQ(3U, DroppedBytes());
   ASSERT_EQ("OK", MatchError("missing start"));
 }
 
-TEST_F(LogTest, UnexpectedFullType) {
+TEST_P(LogTest, UnexpectedFullType) {
   Write("foo");
   Write("bar");
-  SetByte(6, kFirstType);
-  FixChecksum(0, 3);
+  SetByte(6, GetParam() ? kRecyclableFirstType : kFirstType);
+  FixChecksum(0, 3, !!GetParam());
   ASSERT_EQ("bar", Read());
   ASSERT_EQ("EOF", Read());
   ASSERT_EQ(3U, DroppedBytes());
   ASSERT_EQ("OK", MatchError("partial record without end"));
 }
 
-TEST_F(LogTest, UnexpectedFirstType) {
+TEST_P(LogTest, UnexpectedFirstType) {
   Write("foo");
   Write(BigString("bar", 100000));
-  SetByte(6, kFirstType);
-  FixChecksum(0, 3);
+  SetByte(6, GetParam() ? kRecyclableFirstType : kFirstType);
+  FixChecksum(0, 3, !!GetParam());
   ASSERT_EQ(BigString("bar", 100000), Read());
   ASSERT_EQ("EOF", Read());
   ASSERT_EQ(3U, DroppedBytes());
   ASSERT_EQ("OK", MatchError("partial record without end"));
 }
 
-TEST_F(LogTest, MissingLastIsIgnored) {
+TEST_P(LogTest, MissingLastIsIgnored) {
   Write(BigString("bar", kBlockSize));
   // Remove the LAST block, including header.
   ShrinkSize(14);
@@ -514,16 +523,16 @@ TEST_F(LogTest, MissingLastIsIgnored) {
   ASSERT_EQ(0U, DroppedBytes());
 }
 
-TEST_F(LogTest, MissingLastIsNotIgnored) {
+TEST_P(LogTest, MissingLastIsNotIgnored) {
   Write(BigString("bar", kBlockSize));
   // Remove the LAST block, including header.
   ShrinkSize(14);
-  ASSERT_EQ("EOF", Read(/*report_eof_inconsistency=*/true));
+  ASSERT_EQ("EOF", Read(WALRecoveryMode::kAbsoluteConsistency));
   ASSERT_GT(DroppedBytes(), 0U);
   ASSERT_EQ("OK", MatchError("Corruption: error reading trailing data"));
 }
 
-TEST_F(LogTest, PartialLastIsIgnored) {
+TEST_P(LogTest, PartialLastIsIgnored) {
   Write(BigString("bar", kBlockSize));
   // Cause a bad record length in the LAST block.
   ShrinkSize(1);
@@ -532,18 +541,18 @@ TEST_F(LogTest, PartialLastIsIgnored) {
   ASSERT_EQ(0U, DroppedBytes());
 }
 
-TEST_F(LogTest, PartialLastIsNotIgnored) {
+TEST_P(LogTest, PartialLastIsNotIgnored) {
   Write(BigString("bar", kBlockSize));
   // Cause a bad record length in the LAST block.
   ShrinkSize(1);
-  ASSERT_EQ("EOF", Read(/*report_eof_inconsistency=*/true));
+  ASSERT_EQ("EOF", Read(WALRecoveryMode::kAbsoluteConsistency));
   ASSERT_GT(DroppedBytes(), 0U);
   ASSERT_EQ("OK", MatchError(
                       "Corruption: truncated headerCorruption: "
                       "error reading trailing data"));
 }
 
-TEST_F(LogTest, ErrorJoinsRecords) {
+TEST_P(LogTest, ErrorJoinsRecords) {
   // Consider two fragmented records:
   //    first(R1) last(R1) first(R2) last(R2)
   // where the middle two fragments disappear.  We do not want
@@ -566,46 +575,60 @@ TEST_F(LogTest, ErrorJoinsRecords) {
   ASSERT_GE(dropped, 2 * kBlockSize);
 }
 
-TEST_F(LogTest, ReadStart) { CheckInitialOffsetRecord(0, 0); }
+TEST_P(LogTest, ReadStart) { CheckInitialOffsetRecord(0, 0); }
 
-TEST_F(LogTest, ReadSecondOneOff) { CheckInitialOffsetRecord(1, 1); }
+TEST_P(LogTest, ReadSecondOneOff) { CheckInitialOffsetRecord(1, 1); }
 
-TEST_F(LogTest, ReadSecondTenThousand) { CheckInitialOffsetRecord(10000, 1); }
+TEST_P(LogTest, ReadSecondTenThousand) { CheckInitialOffsetRecord(10000, 1); }
 
-TEST_F(LogTest, ReadSecondStart) { CheckInitialOffsetRecord(10007, 1); }
+TEST_P(LogTest, ReadSecondStart) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  CheckInitialOffsetRecord(10000 + header_size, 1);
+}
 
-TEST_F(LogTest, ReadThirdOneOff) { CheckInitialOffsetRecord(10008, 2); }
+TEST_P(LogTest, ReadThirdOneOff) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  CheckInitialOffsetRecord(10000 + header_size + 1, 2);
+}
 
-TEST_F(LogTest, ReadThirdStart) { CheckInitialOffsetRecord(20014, 2); }
+TEST_P(LogTest, ReadThirdStart) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  CheckInitialOffsetRecord(20000 + 2 * header_size, 2);
+}
 
-TEST_F(LogTest, ReadFourthOneOff) { CheckInitialOffsetRecord(20015, 3); }
+TEST_P(LogTest, ReadFourthOneOff) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  CheckInitialOffsetRecord(20000 + 2 * header_size + 1, 3);
+}
 
-TEST_F(LogTest, ReadFourthFirstBlockTrailer) {
+TEST_P(LogTest, ReadFourthFirstBlockTrailer) {
   CheckInitialOffsetRecord(log::kBlockSize - 4, 3);
 }
 
-TEST_F(LogTest, ReadFourthMiddleBlock) {
+TEST_P(LogTest, ReadFourthMiddleBlock) {
   CheckInitialOffsetRecord(log::kBlockSize + 1, 3);
 }
 
-TEST_F(LogTest, ReadFourthLastBlock) {
+TEST_P(LogTest, ReadFourthLastBlock) {
   CheckInitialOffsetRecord(2 * log::kBlockSize + 1, 3);
 }
 
-TEST_F(LogTest, ReadFourthStart) {
+TEST_P(LogTest, ReadFourthStart) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
   CheckInitialOffsetRecord(
-      2 * (kHeaderSize + 1000) + (2 * log::kBlockSize - 1000) + 3 * kHeaderSize,
+      2 * (header_size + 1000) + (2 * log::kBlockSize - 1000) + 3 * header_size,
       3);
 }
 
-TEST_F(LogTest, ReadEnd) { CheckOffsetPastEndReturnsNoRecords(0); }
+TEST_P(LogTest, ReadEnd) { CheckOffsetPastEndReturnsNoRecords(0); }
 
-TEST_F(LogTest, ReadPastEnd) { CheckOffsetPastEndReturnsNoRecords(5); }
+TEST_P(LogTest, ReadPastEnd) { CheckOffsetPastEndReturnsNoRecords(5); }
 
-TEST_F(LogTest, ClearEofSingleBlock) {
+TEST_P(LogTest, ClearEofSingleBlock) {
   Write("foo");
   Write("bar");
-  ForceEOF(3 + kHeaderSize + 2);
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  ForceEOF(3 + header_size + 2);
   ASSERT_EQ("foo", Read());
   UnmarkEOF();
   ASSERT_EQ("bar", Read());
@@ -617,12 +640,13 @@ TEST_F(LogTest, ClearEofSingleBlock) {
   ASSERT_TRUE(IsEOF());
 }
 
-TEST_F(LogTest, ClearEofMultiBlock) {
+TEST_P(LogTest, ClearEofMultiBlock) {
   size_t num_full_blocks = 5;
-  size_t n = (kBlockSize - kHeaderSize) * num_full_blocks + 25;
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  size_t n = (kBlockSize - header_size) * num_full_blocks + 25;
   Write(BigString("foo", n));
   Write(BigString("bar", n));
-  ForceEOF(n + num_full_blocks * kHeaderSize + 10);
+  ForceEOF(n + num_full_blocks * header_size + header_size + 3);
   ASSERT_EQ(BigString("foo", n), Read());
   ASSERT_TRUE(IsEOF());
   UnmarkEOF();
@@ -634,7 +658,7 @@ TEST_F(LogTest, ClearEofMultiBlock) {
   ASSERT_TRUE(IsEOF());
 }
 
-TEST_F(LogTest, ClearEofError) {
+TEST_P(LogTest, ClearEofError) {
   // If an error occurs during Read() in UnmarkEOF(), the records contained
   // in the buffer should be returned on subsequent calls of ReadRecord()
   // until no more full records are left, whereafter ReadRecord() should return
@@ -652,7 +676,7 @@ TEST_F(LogTest, ClearEofError) {
   ASSERT_EQ("EOF", Read());
 }
 
-TEST_F(LogTest, ClearEofError2) {
+TEST_P(LogTest, ClearEofError2) {
   Write("foo");
   Write("bar");
   UnmarkEOF();
@@ -666,6 +690,8 @@ TEST_F(LogTest, ClearEofError2) {
   ASSERT_EQ("OK", MatchError("read error"));
 }
 
+INSTANTIATE_TEST_CASE_P(bool, LogTest, ::testing::Values(0, 2));
+
 }  // namespace log
 }  // namespace rocksdb
 

db/log_writer.cc

@@ -18,8 +18,12 @@
 namespace rocksdb {
 namespace log {
 
-Writer::Writer(unique_ptr<WritableFileWriter>&& dest)
-    : dest_(std::move(dest)), block_offset_(0) {
+Writer::Writer(unique_ptr<WritableFileWriter>&& dest,
+               uint64_t log_number, bool recycle_log_files)
+    : dest_(std::move(dest)),
+      block_offset_(0),
+      log_number_(log_number),
+      recycle_log_files_(recycle_log_files) {
   for (int i = 0; i <= kMaxRecordType; i++) {
     char t = static_cast<char>(i);
     type_crc_[i] = crc32c::Value(&t, 1);
@@ -33,6 +37,10 @@ Status Writer::AddRecord(const Slice& slice) {
   const char* ptr = slice.data();
   size_t left = slice.size();
 
+  // Header size varies depending on whether we are recycling or not.
+  const int header_size =
+      recycle_log_files_ ? kRecyclableHeaderSize : kHeaderSize;
+
   // Fragment the record if necessary and emit it.  Note that if slice
   // is empty, we still want to iterate once to emit a single
   // zero-length record
@@ -41,32 +49,34 @@ Status Writer::AddRecord(const Slice& slice) {
   do {
     const int leftover = kBlockSize - block_offset_;
     assert(leftover >= 0);
-    if (leftover < kHeaderSize) {
+    if (leftover < header_size) {
       // Switch to a new block
       if (leftover > 0) {
-        // Fill the trailer (literal below relies on kHeaderSize being 7)
-        assert(kHeaderSize == 7);
-        dest_->Append(Slice("\x00\x00\x00\x00\x00\x00", leftover));
+        // Fill the trailer (literal below relies on kHeaderSize and
+        // kRecyclableHeaderSize being <= 11)
+        assert(header_size <= 11);
+        dest_->Append(
+            Slice("\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00", leftover));
       }
       block_offset_ = 0;
     }
 
-    // Invariant: we never leave < kHeaderSize bytes in a block.
-    assert(static_cast<int>(kBlockSize) - block_offset_ >= kHeaderSize);
+    // Invariant: we never leave < header_size bytes in a block.
+    assert(static_cast<int>(kBlockSize) - block_offset_ >= header_size);
 
-    const size_t avail = kBlockSize - block_offset_ - kHeaderSize;
+    const size_t avail = kBlockSize - block_offset_ - header_size;
     const size_t fragment_length = (left < avail) ? left : avail;
 
     RecordType type;
     const bool end = (left == fragment_length);
     if (begin && end) {
-      type = kFullType;
+      type = recycle_log_files_ ? kRecyclableFullType : kFullType;
     } else if (begin) {
-      type = kFirstType;
+      type = recycle_log_files_ ? kRecyclableFirstType : kFirstType;
     } else if (end) {
-      type = kLastType;
+      type = recycle_log_files_ ? kRecyclableLastType : kLastType;
     } else {
-      type = kMiddleType;
+      type = recycle_log_files_ ? kRecyclableMiddleType : kMiddleType;
     }
 
     s = EmitPhysicalRecord(type, ptr, fragment_length);
@@ -79,28 +89,48 @@ Status Writer::AddRecord(const Slice& slice) {
 
 Status Writer::EmitPhysicalRecord(RecordType t, const char* ptr, size_t n) {
   assert(n <= 0xffff);  // Must fit in two bytes
-  assert(block_offset_ + kHeaderSize + n <= kBlockSize);
+
+  size_t header_size;
+  char buf[kRecyclableHeaderSize];
 
   // Format the header
-  char buf[kHeaderSize];
   buf[4] = static_cast<char>(n & 0xff);
   buf[5] = static_cast<char>(n >> 8);
   buf[6] = static_cast<char>(t);
 
+  uint32_t crc = type_crc_[t];
+  if (t < kRecyclableFullType) {
+    // Legacy record format
+    assert(block_offset_ + kHeaderSize + n <= kBlockSize);
+    header_size = kHeaderSize;
+  } else {
+    // Recyclable record format
+    assert(block_offset_ + kRecyclableHeaderSize + n <= kBlockSize);
+    header_size = kRecyclableHeaderSize;
+
+    // Only encode low 32-bits of the 64-bit log number.  This means
+    // we will fail to detect an old record if we recycled a log from
+    // ~4 billion logs ago, but that is effectively impossible, and
+    // even if it were we'dbe far more likely to see a false positive
+    // on the 32-bit CRC.
+    EncodeFixed32(buf + 7, static_cast<uint32_t>(log_number_));
+    crc = crc32c::Extend(crc, buf + 7, 4);
+  }
+
   // Compute the crc of the record type and the payload.
-  uint32_t crc = crc32c::Extend(type_crc_[t], ptr, n);
-  crc = crc32c::Mask(crc);                 // Adjust for storage
+  crc = crc32c::Extend(crc, ptr, n);
+  crc = crc32c::Mask(crc);  // Adjust for storage
   EncodeFixed32(buf, crc);
 
   // Write the header and the payload
-  Status s = dest_->Append(Slice(buf, kHeaderSize));
+  Status s = dest_->Append(Slice(buf, header_size));
   if (s.ok()) {
     s = dest_->Append(Slice(ptr, n));
     if (s.ok()) {
       s = dest_->Flush();
     }
   }
-  block_offset_ += kHeaderSize + n;
+  block_offset_ += header_size + n;
   return s;
 }
 

db/log_writer.h

@@ -43,7 +43,7 @@ namespace log {
  * Data is written out in kBlockSize chunks. If next record does not fit
  * into the space left, the leftover space will be padded with \0.
  *
- * Record format:
+ * Legacy record format:
  *
  * +---------+-----------+-----------+--- ... ---+
  * |CRC (4B) | Size (2B) | Type (1B) | Payload   |
@@ -57,13 +57,23 @@ namespace log {
  *        blocks that are larger than kBlockSize
  * Payload = Byte stream as long as specified by the payload size
  *
+ * Recyclable record format:
+ *
+ * +---------+-----------+-----------+----------------+--- ... ---+
+ * |CRC (4B) | Size (2B) | Type (1B) | Log number (4B)| Payload   |
+ * +---------+-----------+-----------+----------------+--- ... ---+
+ *
+ * Same as above, with the addition of
+ * Log number = 32bit log file number, so that we can distinguish between
+ * records written by the most recent log writer vs a previous one.
  */
 class Writer {
  public:
   // Create a writer that will append data to "*dest".
   // "*dest" must be initially empty.
   // "*dest" must remain live while this Writer is in use.
-  explicit Writer(unique_ptr<WritableFileWriter>&& dest);
+  explicit Writer(unique_ptr<WritableFileWriter>&& dest,
+                  uint64_t log_number, bool recycle_log_files);
   ~Writer();
 
   Status AddRecord(const Slice& slice);
@@ -74,6 +84,8 @@ class Writer {
  private:
   unique_ptr<WritableFileWriter> dest_;
   int block_offset_;       // Current offset in block
+  uint64_t log_number_;
+  bool recycle_log_files_;
 
   // crc32c values for all supported record types.  These are
   // pre-computed to reduce the overhead of computing the crc of the

db/memtable.cc

@@ -21,6 +21,7 @@
 #include "rocksdb/iterator.h"
 #include "rocksdb/merge_operator.h"
 #include "rocksdb/slice_transform.h"
+#include "table/internal_iterator.h"
 #include "table/merger.h"
 #include "util/arena.h"
 #include "util/coding.h"
@@ -202,7 +203,7 @@ const char* EncodeKey(std::string* scratch, const Slice& target) {
   return scratch->data();
 }
 
-class MemTableIterator: public Iterator {
+class MemTableIterator : public InternalIterator {
  public:
   MemTableIterator(
       const MemTable& mem, const ReadOptions& read_options, Arena* arena)
@@ -285,7 +286,8 @@ class MemTableIterator: public Iterator {
   void operator=(const MemTableIterator&);
 };
 
-Iterator* MemTable::NewIterator(const ReadOptions& read_options, Arena* arena) {
+InternalIterator* MemTable::NewIterator(const ReadOptions& read_options,
+                                        Arena* arena) {
   assert(arena != nullptr);
   auto mem = arena->AllocateAligned(sizeof(MemTableIterator));
   return new (mem) MemTableIterator(*this, read_options, arena);

db/memtable.h

@@ -31,6 +31,7 @@ class Mutex;
 class MemTableIterator;
 class MergeContext;
 class WriteBuffer;
+class InternalIterator;
 
 struct MemTableOptions {
   explicit MemTableOptions(
@@ -140,7 +141,7 @@ class MemTable {
   // arena: If not null, the arena needs to be used to allocate the Iterator.
   //        Calling ~Iterator of the iterator will destroy all the states but
   //        those allocated in arena.
-  Iterator* NewIterator(const ReadOptions& read_options, Arena* arena);
+  InternalIterator* NewIterator(const ReadOptions& read_options, Arena* arena);
 
   // Add an entry into memtable that maps key to value at the
   // specified sequence number and with the specified type.

db/memtable_list.cc

@@ -138,9 +138,9 @@ bool MemTableListVersion::GetFromList(std::list<MemTable*>* list,
   return false;
 }
 
-void MemTableListVersion::AddIterators(const ReadOptions& options,
-                                       std::vector<Iterator*>* iterator_list,
-                                       Arena* arena) {
+void MemTableListVersion::AddIterators(
+    const ReadOptions& options, std::vector<InternalIterator*>* iterator_list,
+    Arena* arena) {
   for (auto& m : memlist_) {
     iterator_list->push_back(m->NewIterator(options, arena));
   }

db/memtable_list.h

@@ -74,7 +74,8 @@ class MemTableListVersion {
   }
 
   void AddIterators(const ReadOptions& options,
-                    std::vector<Iterator*>* iterator_list, Arena* arena);
+                    std::vector<InternalIterator*>* iterator_list,
+                    Arena* arena);
 
   void AddIterators(const ReadOptions& options,
                     MergeIteratorBuilder* merge_iter_builder);

db/memtablerep_bench.cc

@@ -132,8 +132,6 @@ DEFINE_int64(seed, 0,
              "Seed base for random number generators. "
              "When 0 it is deterministic.");
 
-static rocksdb::Env* FLAGS_env = rocksdb::Env::Default();
-
 namespace rocksdb {
 
 namespace {
@@ -592,6 +590,7 @@ int main(int argc, char** argv) {
   std::unique_ptr<rocksdb::MemTableRepFactory> factory;
   if (FLAGS_memtablerep == "skiplist") {
     factory.reset(new rocksdb::SkipListFactory);
+#ifndef ROCKSDB_LITE
   } else if (FLAGS_memtablerep == "vector") {
     factory.reset(new rocksdb::VectorRepFactory);
   } else if (FLAGS_memtablerep == "hashskiplist") {
@@ -613,6 +612,7 @@ int main(int argc, char** argv) {
         static_cast<uint32_t>(FLAGS_hash_function_count)));
     options.prefix_extractor.reset(
         rocksdb::NewFixedPrefixTransform(FLAGS_prefix_length));
+#endif  // ROCKSDB_LITE
   } else {
     fprintf(stdout, "Unknown memtablerep: %s\n", FLAGS_memtablerep.c_str());
     exit(1);

db/merge_helper.cc

@@ -12,6 +12,7 @@
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/merge_operator.h"
+#include "table/internal_iterator.h"
 #include "util/perf_context_imp.h"
 #include "util/statistics.h"
 
@@ -56,7 +57,8 @@ Status MergeHelper::TimedFullMerge(const Slice& key, const Slice* value,
 //       keys_ stores the list of keys encountered while merging.
 //       operands_ stores the list of merge operands encountered while merging.
 //       keys_[i] corresponds to operands_[i] for each i.
-Status MergeHelper::MergeUntil(Iterator* iter, const SequenceNumber stop_before,
+Status MergeHelper::MergeUntil(InternalIterator* iter,
+                               const SequenceNumber stop_before,
                                const bool at_bottom) {
   // Get a copy of the internal key, before it's invalidated by iter->Next()
   // Also maintain the list of merge operands seen.

db/merge_helper.h

@@ -22,6 +22,7 @@ class Iterator;
 class Logger;
 class MergeOperator;
 class Statistics;
+class InternalIterator;
 
 class MergeHelper {
  public:
@@ -82,7 +83,8 @@ class MergeHelper {
   //   with asserts removed).
   //
   // REQUIRED: The first key in the input is not corrupted.
-  Status MergeUntil(Iterator* iter, const SequenceNumber stop_before = 0,
+  Status MergeUntil(InternalIterator* iter,
+                    const SequenceNumber stop_before = 0,
                     const bool at_bottom = false);
 
   // Filters a merge operand using the compaction filter specified

db/options_file_test.cc

@@ -0,0 +1,119 @@
+//  Copyright (c) 2015, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#ifndef ROCKSDB_LITE
+#include <string>
+
+#include "db/db_impl.h"
+#include "db/db_test_util.h"
+#include "rocksdb/options.h"
+#include "rocksdb/table.h"
+#include "util/testharness.h"
+
+namespace rocksdb {
+class OptionsFileTest : public testing::Test {
+ public:
+  OptionsFileTest() : dbname_(test::TmpDir() + "/options_file_test") {}
+
+  std::string dbname_;
+};
+
+namespace {
+void UpdateOptionsFiles(DB* db,
+                        std::unordered_set<std::string>* filename_history,
+                        int* options_files_count) {
+  std::vector<std::string> filenames;
+  db->GetEnv()->GetChildren(db->GetName(), &filenames);
+  uint64_t number;
+  FileType type;
+  *options_files_count = 0;
+  for (auto filename : filenames) {
+    if (ParseFileName(filename, &number, &type) && type == kOptionsFile) {
+      filename_history->insert(filename);
+      (*options_files_count)++;
+    }
+  }
+}
+
+// Verify whether the current Options Files are the latest ones.
+void VerifyOptionsFileName(
+    DB* db, const std::unordered_set<std::string>& past_filenames) {
+  std::vector<std::string> filenames;
+  std::unordered_set<std::string> current_filenames;
+  db->GetEnv()->GetChildren(db->GetName(), &filenames);
+  uint64_t number;
+  FileType type;
+  for (auto filename : filenames) {
+    if (ParseFileName(filename, &number, &type) && type == kOptionsFile) {
+      current_filenames.insert(filename);
+    }
+  }
+  for (auto past_filename : past_filenames) {
+    if (current_filenames.find(past_filename) != current_filenames.end()) {
+      continue;
+    }
+    for (auto filename : current_filenames) {
+      ASSERT_GT(filename, past_filename);
+    }
+  }
+}
+}  // namespace
+
+TEST_F(OptionsFileTest, NumberOfOptionsFiles) {
+  const int kReopenCount = 20;
+  Options opt;
+  opt.create_if_missing = true;
+  DestroyDB(dbname_, opt);
+  std::unordered_set<std::string> filename_history;
+  DB* db;
+  for (int i = 0; i < kReopenCount; ++i) {
+    ASSERT_OK(DB::Open(opt, dbname_, &db));
+    int num_options_files = 0;
+    UpdateOptionsFiles(db, &filename_history, &num_options_files);
+    ASSERT_GT(num_options_files, 0);
+    ASSERT_LE(num_options_files, 2);
+    // Make sure we always keep the latest option files.
+    VerifyOptionsFileName(db, filename_history);
+    delete db;
+  }
+}
+
+TEST_F(OptionsFileTest, OptionsFileName) {
+  const uint64_t kOptionsFileNum = 12345;
+  uint64_t number;
+  FileType type;
+
+  auto options_file_name = OptionsFileName("", kOptionsFileNum);
+  ASSERT_TRUE(ParseFileName(options_file_name, &number, &type, nullptr));
+  ASSERT_EQ(type, kOptionsFile);
+  ASSERT_EQ(number, kOptionsFileNum);
+
+  const uint64_t kTempOptionsFileNum = 54352;
+  auto temp_options_file_name = TempOptionsFileName("", kTempOptionsFileNum);
+  ASSERT_TRUE(ParseFileName(temp_options_file_name, &number, &type, nullptr));
+  ASSERT_NE(temp_options_file_name.find(kTempFileNameSuffix),
+            std::string::npos);
+  ASSERT_EQ(type, kTempFile);
+  ASSERT_EQ(number, kTempOptionsFileNum);
+}
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+#if !(defined NDEBUG) || !defined(OS_WIN)
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+#else
+  return 0;
+#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+}
+#else
+
+#include <cstdio>
+
+int main(int argc, char** argv) {
+  printf("Skipped as Options file is not supported in RocksDBLite.\n");
+  return 0;
+}
+#endif  // !ROCKSDB_LITE

db/perf_context_test.cc

@@ -5,6 +5,7 @@
 //
 #include <algorithm>
 #include <iostream>
+#include <thread>
 #include <vector>
 
 #include "rocksdb/db.h"
@@ -444,6 +445,7 @@ void ProfileQueries(bool enabled_time = false) {
   }
 }
 
+#ifndef ROCKSDB_LITE
 TEST_F(PerfContextTest, KeyComparisonCount) {
   SetPerfLevel(kEnableCount);
   ProfileQueries();
@@ -454,6 +456,7 @@ TEST_F(PerfContextTest, KeyComparisonCount) {
   SetPerfLevel(kEnableTime);
   ProfileQueries(true);
 }
+#endif  // ROCKSDB_LITE
 
 // make perf_context_test
 // export ROCKSDB_TESTS=PerfContextTest.SeekKeyComparison
@@ -539,6 +542,49 @@ TEST_F(PerfContextTest, SeekKeyComparison) {
   }
 }
 
+TEST_F(PerfContextTest, DBMutexLockCounter) {
+  SetPerfLevel(kEnableTime);
+  int stats_code[] = {0, static_cast<int>(DB_MUTEX_WAIT_MICROS)};
+  for (int c = 0; c < 2; ++c) {
+    InstrumentedMutex mutex(nullptr, Env::Default(), stats_code[c]);
+    mutex.Lock();
+    std::thread child_thread([&] {
+      SetPerfLevel(kEnableTime);
+      perf_context.Reset();
+      ASSERT_EQ(perf_context.db_mutex_lock_nanos, 0);
+      mutex.Lock();
+      mutex.Unlock();
+      if (stats_code[c] == DB_MUTEX_WAIT_MICROS) {
+        // increment the counter only when it's a DB Mutex
+        ASSERT_GT(perf_context.db_mutex_lock_nanos, 0);
+      } else {
+        ASSERT_EQ(perf_context.db_mutex_lock_nanos, 0);
+      }
+    });
+    Env::Default()->SleepForMicroseconds(100);
+    mutex.Unlock();
+    child_thread.join();
+  }
+}
+
+TEST_F(PerfContextTest, FalseDBMutexWait) {
+  SetPerfLevel(kEnableTime);
+  int stats_code[] = {0, static_cast<int>(DB_MUTEX_WAIT_MICROS)};
+  for (int c = 0; c < 2; ++c) {
+    InstrumentedMutex mutex(nullptr, Env::Default(), stats_code[c]);
+    InstrumentedCondVar lock(&mutex);
+    perf_context.Reset();
+    mutex.Lock();
+    lock.TimedWait(100);
+    mutex.Unlock();
+    if (stats_code[c] == static_cast<int>(DB_MUTEX_WAIT_MICROS)) {
+      // increment the counter only when it's a DB Mutex
+      ASSERT_GT(perf_context.db_condition_wait_nanos, 0);
+    } else {
+      ASSERT_EQ(perf_context.db_condition_wait_nanos, 0);
+    }
+  }
+}
 }
 
 int main(int argc, char** argv) {

db/prefix_test.cc

@@ -20,9 +20,11 @@ int main() {
 #include <gflags/gflags.h>
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
+#include "rocksdb/filter_policy.h"
 #include "rocksdb/perf_context.h"
 #include "rocksdb/slice_transform.h"
 #include "rocksdb/memtablerep.h"
+#include "rocksdb/table.h"
 #include "util/histogram.h"
 #include "util/stop_watch.h"
 #include "util/string_util.h"
@@ -163,6 +165,12 @@ class PrefixTest : public testing::Test {
     options.memtable_prefix_bloom_huge_page_tlb_size =
         FLAGS_memtable_prefix_bloom_huge_page_tlb_size;
 
+    options.prefix_extractor.reset(NewFixedPrefixTransform(8));
+    BlockBasedTableOptions bbto;
+    bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+    bbto.whole_key_filtering = false;
+    options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+
     Status s = DB::Open(options, kDbName,  &db);
     EXPECT_OK(s);
     return std::shared_ptr<DB>(db);
@@ -393,6 +401,55 @@ TEST_F(PrefixTest, TestResult) {
   }
 }
 
+// Show results in prefix
+TEST_F(PrefixTest, PrefixValid) {
+  for (int num_buckets = 1; num_buckets <= 2; num_buckets++) {
+    FirstOption();
+    while (NextOptions(num_buckets)) {
+      std::cout << "*** Mem table: " << options.memtable_factory->Name()
+                << " number of buckets: " << num_buckets << std::endl;
+      DestroyDB(kDbName, Options());
+      auto db = OpenDb();
+      WriteOptions write_options;
+      ReadOptions read_options;
+
+      // Insert keys with common prefix and one key with different
+      Slice v16("v16");
+      Slice v17("v17");
+      Slice v18("v18");
+      Slice v19("v19");
+      PutKey(db.get(), write_options, 12345, 6, v16);
+      PutKey(db.get(), write_options, 12345, 7, v17);
+      PutKey(db.get(), write_options, 12345, 8, v18);
+      PutKey(db.get(), write_options, 12345, 9, v19);
+      PutKey(db.get(), write_options, 12346, 8, v16);
+      db->Flush(FlushOptions());
+      db->Delete(write_options, TestKeyToSlice(TestKey(12346, 8)));
+      db->Flush(FlushOptions());
+      read_options.prefix_same_as_start = true;
+      std::unique_ptr<Iterator> iter(db->NewIterator(read_options));
+      SeekIterator(iter.get(), 12345, 6);
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_TRUE(v16 == iter->value());
+
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_TRUE(v17 == iter->value());
+
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_TRUE(v18 == iter->value());
+
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_TRUE(v19 == iter->value());
+      iter->Next();
+      ASSERT_FALSE(iter->Valid());
+      ASSERT_EQ(kNotFoundResult, Get(db.get(), read_options, 12346, 8));
+    }
+  }
+}
+
 TEST_F(PrefixTest, DynamicPrefixIterator) {
   while (NextOptions(FLAGS_bucket_count)) {
     std::cout << "*** Mem table: " << options.memtable_factory->Name()

db/repair.cc

@@ -81,8 +81,8 @@
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
 #include "rocksdb/immutable_options.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/file_reader_writer.h"
-#include "util/scoped_arena_iterator.h"
 
 namespace rocksdb {
 
@@ -249,8 +249,8 @@ class Repairer {
     // corruptions cause entire commits to be skipped instead of
     // propagating bad information (like overly large sequence
     // numbers).
-    log::Reader reader(std::move(lfile_reader), &reporter,
-                       true /*enable checksum*/, 0 /*initial_offset*/);
+    log::Reader reader(options_.info_log, std::move(lfile_reader), &reporter,
+                       true /*enable checksum*/, 0 /*initial_offset*/, log);
 
     // Read all the records and add to a memtable
     std::string scratch;
@@ -290,10 +290,11 @@ class Repairer {
       ro.total_order_seek = true;
       Arena arena;
       ScopedArenaIterator iter(mem->NewIterator(ro, &arena));
-      status = BuildTable(dbname_, env_, ioptions_, env_options_, table_cache_,
-                          iter.get(), &meta, icmp_,
-                          &int_tbl_prop_collector_factories_, {},
-                          kNoCompression, CompressionOptions(), false, nullptr);
+      status = BuildTable(
+          dbname_, env_, ioptions_, env_options_, table_cache_, iter.get(),
+          &meta, icmp_, &int_tbl_prop_collector_factories_,
+          TablePropertiesCollectorFactory::Context::kUnknownColumnFamily, {},
+          kNoCompression, CompressionOptions(), false, nullptr);
     }
     delete mem->Unref();
     delete cf_mems_default;
@@ -339,7 +340,7 @@ class Repairer {
     t->meta.fd = FileDescriptor(t->meta.fd.GetNumber(), t->meta.fd.GetPathId(),
                                 file_size);
     if (status.ok()) {
-      Iterator* iter = table_cache_->NewIterator(
+      InternalIterator* iter = table_cache_->NewIterator(
           ReadOptions(), env_options_, icmp_, t->meta.fd);
       bool empty = true;
       ParsedInternalKey parsed;
@@ -412,7 +413,7 @@ class Repairer {
     {
       unique_ptr<WritableFileWriter> file_writer(
           new WritableFileWriter(std::move(file), env_options));
-      log::Writer log(std::move(file_writer));
+      log::Writer log(std::move(file_writer), 0, false);
       std::string record;
       edit_->EncodeTo(&record);
       status = log.AddRecord(record);

db/skiplist.h

@@ -107,8 +107,9 @@ class SkipList {
   };
 
  private:
-  const int32_t kMaxHeight_;
-  const int32_t kBranching_;
+  const uint16_t kMaxHeight_;
+  const uint16_t kBranching_;
+  const uint32_t kScaledInverseBranching_;
 
   // Immutable after construction
   Comparator const compare_;
@@ -131,9 +132,6 @@ class SkipList {
     return max_height_.load(std::memory_order_relaxed);
   }
 
-  // Read/written only by Insert().
-  Random rnd_;
-
   Node* NewNode(const Key& key, int height);
   int RandomHeight();
   bool Equal(const Key& a, const Key& b) const { return (compare_(a, b) == 0); }
@@ -264,9 +262,11 @@ inline void SkipList<Key, Comparator>::Iterator::SeekToLast() {
 
 template<typename Key, class Comparator>
 int SkipList<Key, Comparator>::RandomHeight() {
+  auto rnd = Random::GetTLSInstance();
+
   // Increase height with probability 1 in kBranching
   int height = 1;
-  while (height < kMaxHeight_ && ((rnd_.Next() % kBranching_) == 0)) {
+  while (height < kMaxHeight_ && rnd->Next() < kScaledInverseBranching_) {
     height++;
   }
   assert(height > 0);
@@ -391,14 +391,16 @@ SkipList<Key, Comparator>::SkipList(const Comparator cmp, Allocator* allocator,
                                     int32_t branching_factor)
     : kMaxHeight_(max_height),
       kBranching_(branching_factor),
+      kScaledInverseBranching_((Random::kMaxNext + 1) / kBranching_),
       compare_(cmp),
       allocator_(allocator),
       head_(NewNode(0 /* any key will do */, max_height)),
       max_height_(1),
-      prev_height_(1),
-      rnd_(0xdeadbeef) {
-  assert(kMaxHeight_ > 0);
-  assert(kBranching_ > 0);
+      prev_height_(1) {
+  assert(max_height > 0 && kMaxHeight_ == static_cast<uint32_t>(max_height));
+  assert(branching_factor > 0 &&
+         kBranching_ == static_cast<uint32_t>(branching_factor));
+  assert(kScaledInverseBranching_ > 0);
   // Allocate the prev_ Node* array, directly from the passed-in allocator.
   // prev_ does not need to be freed, as its life cycle is tied up with
   // the allocator as a whole.

db/table_cache.cc

@@ -14,6 +14,7 @@
 #include "db/version_edit.h"
 
 #include "rocksdb/statistics.h"
+#include "table/internal_iterator.h"
 #include "table/iterator_wrapper.h"
 #include "table/table_builder.h"
 #include "table/table_reader.h"
@@ -148,13 +149,11 @@ Status TableCache::FindTable(const EnvOptions& env_options,
   return s;
 }
 
-Iterator* TableCache::NewIterator(const ReadOptions& options,
-                                  const EnvOptions& env_options,
-                                  const InternalKeyComparator& icomparator,
-                                  const FileDescriptor& fd,
-                                  TableReader** table_reader_ptr,
-                                  HistogramImpl* file_read_hist,
-                                  bool for_compaction, Arena* arena) {
+InternalIterator* TableCache::NewIterator(
+    const ReadOptions& options, const EnvOptions& env_options,
+    const InternalKeyComparator& icomparator, const FileDescriptor& fd,
+    TableReader** table_reader_ptr, HistogramImpl* file_read_hist,
+    bool for_compaction, Arena* arena) {
   PERF_TIMER_GUARD(new_table_iterator_nanos);
 
   if (table_reader_ptr != nullptr) {
@@ -171,7 +170,7 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,
         env_options, icomparator, fd, /* sequential mode */ true,
         /* record stats */ false, nullptr, &table_reader_unique_ptr);
     if (!s.ok()) {
-      return NewErrorIterator(s, arena);
+      return NewErrorInternalIterator(s, arena);
     }
     table_reader = table_reader_unique_ptr.release();
   } else {
@@ -182,13 +181,13 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,
                     options.read_tier == kBlockCacheTier /* no_io */,
                     !for_compaction /* record read_stats */, file_read_hist);
       if (!s.ok()) {
-        return NewErrorIterator(s, arena);
+        return NewErrorInternalIterator(s, arena);
       }
       table_reader = GetTableReaderFromHandle(handle);
     }
   }
 
-  Iterator* result = table_reader->NewIterator(options, arena);
+  InternalIterator* result = table_reader->NewIterator(options, arena);
 
   if (create_new_table_reader) {
     assert(handle == nullptr);

db/table_cache.h

@@ -29,6 +29,7 @@ class Arena;
 struct FileDescriptor;
 class GetContext;
 class HistogramImpl;
+class InternalIterator;
 
 class TableCache {
  public:
@@ -43,12 +44,12 @@ class TableCache {
   // the returned iterator.  The returned "*tableptr" object is owned by
   // the cache and should not be deleted, and is valid for as long as the
   // returned iterator is live.
-  Iterator* NewIterator(const ReadOptions& options, const EnvOptions& toptions,
-                        const InternalKeyComparator& internal_comparator,
-                        const FileDescriptor& file_fd,
-                        TableReader** table_reader_ptr = nullptr,
-                        HistogramImpl* file_read_hist = nullptr,
-                        bool for_compaction = false, Arena* arena = nullptr);
+  InternalIterator* NewIterator(
+      const ReadOptions& options, const EnvOptions& toptions,
+      const InternalKeyComparator& internal_comparator,
+      const FileDescriptor& file_fd, TableReader** table_reader_ptr = nullptr,
+      HistogramImpl* file_read_hist = nullptr, bool for_compaction = false,
+      Arena* arena = nullptr);
 
   // If a seek to internal key "k" in specified file finds an entry,
   // call (*handle_result)(arg, found_key, found_value) repeatedly until

db/table_properties_collector.h

@@ -41,7 +41,8 @@ class IntTblPropCollectorFactory {
  public:
   virtual ~IntTblPropCollectorFactory() {}
   // has to be thread-safe
-  virtual IntTblPropCollector* CreateIntTblPropCollector() = 0;
+  virtual IntTblPropCollector* CreateIntTblPropCollector(
+      uint32_t column_family_id) = 0;
 
   // The name of the properties collector can be used for debugging purpose.
   virtual const char* Name() const = 0;
@@ -69,7 +70,8 @@ class InternalKeyPropertiesCollector : public IntTblPropCollector {
 class InternalKeyPropertiesCollectorFactory
     : public IntTblPropCollectorFactory {
  public:
-  virtual IntTblPropCollector* CreateIntTblPropCollector() override {
+  virtual IntTblPropCollector* CreateIntTblPropCollector(
+      uint32_t column_family_id) override {
     return new InternalKeyPropertiesCollector();
   }
 
@@ -114,9 +116,12 @@ class UserKeyTablePropertiesCollectorFactory
   explicit UserKeyTablePropertiesCollectorFactory(
       std::shared_ptr<TablePropertiesCollectorFactory> user_collector_factory)
       : user_collector_factory_(user_collector_factory) {}
-  virtual IntTblPropCollector* CreateIntTblPropCollector() override {
+  virtual IntTblPropCollector* CreateIntTblPropCollector(
+      uint32_t column_family_id) override {
+    TablePropertiesCollectorFactory::Context context;
+    context.column_family_id = column_family_id;
     return new UserKeyTablePropertiesCollector(
-        user_collector_factory_->CreateTablePropertiesCollector());
+        user_collector_factory_->CreateTablePropertiesCollector(context));
   }
 
   virtual const char* Name() const override {

db/table_properties_collector_test.cc

@@ -35,6 +35,8 @@ class TablePropertiesTest : public testing::Test,
 
 // Utilities test functions
 namespace {
+static const uint32_t kTestColumnFamilyId = 66;
+
 void MakeBuilder(const Options& options, const ImmutableCFOptions& ioptions,
                  const InternalKeyComparator& internal_comparator,
                  const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
@@ -46,7 +48,8 @@ void MakeBuilder(const Options& options, const ImmutableCFOptions& ioptions,
 
   builder->reset(NewTableBuilder(
       ioptions, internal_comparator, int_tbl_prop_collector_factories,
-      writable->get(), options.compression, options.compression_opts));
+      kTestColumnFamilyId /* column_family_id */, writable->get(),
+      options.compression, options.compression_opts));
 }
 }  // namespace
 
@@ -178,14 +181,17 @@ class RegularKeysStartWithAFactory : public IntTblPropCollectorFactory,
  public:
   explicit RegularKeysStartWithAFactory(bool backward_mode)
       : backward_mode_(backward_mode) {}
-  virtual TablePropertiesCollector* CreateTablePropertiesCollector() override {
+  virtual TablePropertiesCollector* CreateTablePropertiesCollector(
+      TablePropertiesCollectorFactory::Context context) override {
+    EXPECT_EQ(kTestColumnFamilyId, context.column_family_id);
     if (!backward_mode_) {
       return new RegularKeysStartWithA();
     } else {
       return new RegularKeysStartWithABackwardCompatible();
     }
   }
-  virtual IntTblPropCollector* CreateIntTblPropCollector() override {
+  virtual IntTblPropCollector* CreateIntTblPropCollector(
+      uint32_t column_family_id) override {
     return new RegularKeysStartWithAInternal();
   }
   const char* Name() const override { return "RegularKeysStartWithA"; }

db/transaction_log_impl.cc

@@ -262,8 +262,10 @@ Status TransactionLogIteratorImpl::OpenLogReader(const LogFile* logFile) {
     return s;
   }
   assert(file);
-  currentLogReader_.reset(new log::Reader(std::move(file), &reporter_,
-                                          read_options_.verify_checksums_, 0));
+  currentLogReader_.reset(new log::Reader(options_->info_log,
+					  std::move(file), &reporter_,
+                                          read_options_.verify_checksums_, 0,
+                                          logFile->LogNumber()));
   return Status::OK();
 }
 }  //  namespace rocksdb

db/version_builder.cc

@@ -82,6 +82,7 @@ class VersionBuilder::Rep {
   };
 
   const EnvOptions& env_options_;
+  Logger* info_log_;
   TableCache* table_cache_;
   VersionStorageInfo* base_vstorage_;
   LevelState* levels_;
@@ -89,9 +90,10 @@ class VersionBuilder::Rep {
   FileComparator level_nonzero_cmp_;
 
  public:
-  Rep(const EnvOptions& env_options, TableCache* table_cache,
+  Rep(const EnvOptions& env_options, Logger* info_log, TableCache* table_cache,
       VersionStorageInfo* base_vstorage)
       : env_options_(env_options),
+        info_log_(info_log),
         table_cache_(table_cache),
         base_vstorage_(base_vstorage) {
     levels_ = new LevelState[base_vstorage_->num_levels()];
@@ -134,7 +136,10 @@ class VersionBuilder::Rep {
         auto f2 = level_files[i];
         if (level == 0) {
           assert(level_zero_cmp_(f1, f2));
-          assert(f1->largest_seqno > f2->largest_seqno);
+          assert(f1->largest_seqno > f2->largest_seqno ||
+                 // We can have multiple files with seqno = 0 as a result of
+                 // using DB::AddFile()
+                 (f1->largest_seqno == 0 && f2->largest_seqno == 0));
         } else {
           assert(level_nonzero_cmp_(f1, f2));
 
@@ -335,15 +340,16 @@ class VersionBuilder::Rep {
     if (levels_[level].deleted_files.count(f->fd.GetNumber()) > 0) {
       // File is deleted: do nothing
     } else {
-      vstorage->AddFile(level, f);
+      vstorage->AddFile(level, f, info_log_);
     }
   }
 };
 
 VersionBuilder::VersionBuilder(const EnvOptions& env_options,
                                TableCache* table_cache,
-                               VersionStorageInfo* base_vstorage)
-    : rep_(new Rep(env_options, table_cache, base_vstorage)) {}
+                               VersionStorageInfo* base_vstorage,
+                               Logger* info_log)
+    : rep_(new Rep(env_options, info_log, table_cache, base_vstorage)) {}
 VersionBuilder::~VersionBuilder() { delete rep_; }
 void VersionBuilder::CheckConsistency(VersionStorageInfo* vstorage) {
   rep_->CheckConsistency(vstorage);

db/version_builder.h

@@ -24,7 +24,7 @@ class InternalStats;
 class VersionBuilder {
  public:
   VersionBuilder(const EnvOptions& env_options, TableCache* table_cache,
-                 VersionStorageInfo* base_vstorage);
+                 VersionStorageInfo* base_vstorage, Logger* info_log = nullptr);
   ~VersionBuilder();
   void CheckConsistency(VersionStorageInfo* vstorage);
   void CheckConsistencyForDeletes(VersionEdit* edit, uint64_t number,

db/version_edit.cc

@@ -12,6 +12,7 @@
 #include "db/version_set.h"
 #include "util/coding.h"
 #include "util/event_logger.h"
+#include "util/sync_point.h"
 #include "rocksdb/slice.h"
 
 namespace rocksdb {
@@ -32,12 +33,22 @@ enum Tag {
   // these are new formats divergent from open source leveldb
   kNewFile2 = 100,
   kNewFile3 = 102,
+  kNewFile4 = 103,      // 4th (the latest) format version of adding files
   kColumnFamily = 200,  // specify column family for version edit
   kColumnFamilyAdd = 201,
   kColumnFamilyDrop = 202,
   kMaxColumnFamily = 203,
 };
 
+enum CustomTag {
+  kTerminate = 1,  // The end of customized fields
+  kNeedCompaction = 2,
+  kPathId = 65,
+};
+// If this bit for the custom tag is set, opening DB should fail if
+// we don't know this field.
+uint32_t kCustomTagNonSafeIgnoreMask = 1 << 6;
+
 uint64_t PackFileNumberAndPathId(uint64_t number, uint64_t path_id) {
   assert(number <= kFileNumberMask);
   return number | (path_id * (kFileNumberMask + 1));
@@ -102,7 +113,11 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
     if (!f.smallest.Valid() || !f.largest.Valid()) {
       return false;
     }
-    if (f.fd.GetPathId() == 0) {
+    bool has_customized_fields = false;
+    if (f.marked_for_compaction) {
+      PutVarint32(dst, kNewFile4);
+      has_customized_fields = true;
+    } else if (f.fd.GetPathId() == 0) {
       // Use older format to make sure user can roll back the build if they
       // don't config multiple DB paths.
       PutVarint32(dst, kNewFile2);
@@ -111,7 +126,8 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
     }
     PutVarint32(dst, new_files_[i].first);  // level
     PutVarint64(dst, f.fd.GetNumber());
-    if (f.fd.GetPathId() != 0) {
+    if (f.fd.GetPathId() != 0 && !has_customized_fields) {
+      // kNewFile3
       PutVarint32(dst, f.fd.GetPathId());
     }
     PutVarint64(dst, f.fd.GetFileSize());
@@ -119,6 +135,48 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
     PutLengthPrefixedSlice(dst, f.largest.Encode());
     PutVarint64(dst, f.smallest_seqno);
     PutVarint64(dst, f.largest_seqno);
+    if (has_customized_fields) {
+      // Customized fields' format:
+      // +-----------------------------+
+      // | 1st field's tag (varint32)  |
+      // +-----------------------------+
+      // | 1st field's size (varint32) |
+      // +-----------------------------+
+      // |    bytes for 1st field      |
+      // |  (based on size decoded)    |
+      // +-----------------------------+
+      // |                             |
+      // |          ......             |
+      // |                             |
+      // +-----------------------------+
+      // | last field's size (varint32)|
+      // +-----------------------------+
+      // |    bytes for last field     |
+      // |  (based on size decoded)    |
+      // +-----------------------------+
+      // | terminating tag (varint32)  |
+      // +-----------------------------+
+      //
+      // Customized encoding for fields:
+      //   tag kPathId: 1 byte as path_id
+      //   tag kNeedCompaction:
+      //        now only can take one char value 1 indicating need-compaction
+      //
+      if (f.fd.GetPathId() != 0) {
+        PutVarint32(dst, CustomTag::kPathId);
+        char p = static_cast<char>(f.fd.GetPathId());
+        PutLengthPrefixedSlice(dst, Slice(&p, 1));
+      }
+      if (f.marked_for_compaction) {
+        PutVarint32(dst, CustomTag::kNeedCompaction);
+        char p = static_cast<char>(1);
+        PutLengthPrefixedSlice(dst, Slice(&p, 1));
+      }
+      TEST_SYNC_POINT_CALLBACK("VersionEdit::EncodeTo:NewFile4:CustomizeFields",
+                               dst);
+
+      PutVarint32(dst, CustomTag::kTerminate);
+    }
   }
 
   // 0 is default and does not need to be explicitly written
@@ -161,6 +219,63 @@ bool VersionEdit::GetLevel(Slice* input, int* level, const char** msg) {
   }
 }
 
+const char* VersionEdit::DecodeNewFile4From(Slice* input) {
+  const char* msg = nullptr;
+  int level;
+  FileMetaData f;
+  uint64_t number;
+  uint32_t path_id = 0;
+  uint64_t file_size;
+  if (GetLevel(input, &level, &msg) && GetVarint64(input, &number) &&
+      GetVarint64(input, &file_size) && GetInternalKey(input, &f.smallest) &&
+      GetInternalKey(input, &f.largest) &&
+      GetVarint64(input, &f.smallest_seqno) &&
+      GetVarint64(input, &f.largest_seqno)) {
+    // See comments in VersionEdit::EncodeTo() for format of customized fields
+    while (true) {
+      uint32_t custom_tag;
+      Slice field;
+      if (!GetVarint32(input, &custom_tag)) {
+        return "new-file4 custom field";
+      }
+      if (custom_tag == kTerminate) {
+        break;
+      }
+      if (!GetLengthPrefixedSlice(input, &field)) {
+        return "new-file4 custom field lenth prefixed slice error";
+      }
+      switch (custom_tag) {
+        case kPathId:
+          if (field.size() != 1) {
+            return "path_id field wrong size";
+          }
+          path_id = field[0];
+          if (path_id > 3) {
+            return "path_id wrong vaue";
+          }
+          break;
+        case kNeedCompaction:
+          if (field.size() != 1) {
+            return "need_compaction field wrong size";
+          }
+          f.marked_for_compaction = (field[0] == 1);
+          break;
+        default:
+          if ((custom_tag & kCustomTagNonSafeIgnoreMask) != 0) {
+            // Should not proceed if cannot understand it
+            return "new-file4 custom field not supported";
+          }
+          break;
+      }
+    }
+  } else {
+    return "new-file4 entry";
+  }
+  f.fd = FileDescriptor(number, path_id, file_size);
+  new_files_.push_back(std::make_pair(level, f));
+  return nullptr;
+}
+
 Status VersionEdit::DecodeFrom(const Slice& src) {
   Clear();
   Slice input = src;
@@ -304,6 +419,11 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
         break;
       }
 
+      case kNewFile4: {
+        msg = DecodeNewFile4From(&input);
+        break;
+      }
+
       case kColumnFamily:
         if (!GetVarint32(&input, &column_family_)) {
           if (!msg) {

db/version_edit.h

@@ -237,6 +237,8 @@ class VersionEdit {
   bool EncodeTo(std::string* dst) const;
   Status DecodeFrom(const Slice& src);
 
+  const char* DecodeNewFile4From(Slice* input);
+
   typedef std::set<std::pair<int, uint64_t>> DeletedFileSet;
 
   const DeletedFileSet& GetDeletedFiles() { return deleted_files_; }

db/version_edit_test.cc

@@ -8,6 +8,7 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/version_edit.h"
+#include "util/sync_point.h"
 #include "util/testharness.h"
 
 namespace rocksdb {
@@ -45,6 +46,121 @@ TEST_F(VersionEditTest, EncodeDecode) {
   TestEncodeDecode(edit);
 }
 
+TEST_F(VersionEditTest, EncodeDecodeNewFile4) {
+  static const uint64_t kBig = 1ull << 50;
+
+  VersionEdit edit;
+  edit.AddFile(3, 300, 3, 100, InternalKey("foo", kBig + 500, kTypeValue),
+               InternalKey("zoo", kBig + 600, kTypeDeletion), kBig + 500,
+               kBig + 600, true);
+  edit.AddFile(4, 301, 3, 100, InternalKey("foo", kBig + 501, kTypeValue),
+               InternalKey("zoo", kBig + 601, kTypeDeletion), kBig + 501,
+               kBig + 601, false);
+  edit.AddFile(5, 302, 0, 100, InternalKey("foo", kBig + 502, kTypeValue),
+               InternalKey("zoo", kBig + 602, kTypeDeletion), kBig + 502,
+               kBig + 602, true);
+
+  edit.DeleteFile(4, 700);
+
+  edit.SetComparatorName("foo");
+  edit.SetLogNumber(kBig + 100);
+  edit.SetNextFile(kBig + 200);
+  edit.SetLastSequence(kBig + 1000);
+  TestEncodeDecode(edit);
+
+  std::string encoded, encoded2;
+  edit.EncodeTo(&encoded);
+  VersionEdit parsed;
+  Status s = parsed.DecodeFrom(encoded);
+  ASSERT_TRUE(s.ok()) << s.ToString();
+  auto& new_files = parsed.GetNewFiles();
+  ASSERT_TRUE(new_files[0].second.marked_for_compaction);
+  ASSERT_TRUE(!new_files[1].second.marked_for_compaction);
+  ASSERT_TRUE(new_files[2].second.marked_for_compaction);
+  ASSERT_EQ(3, new_files[0].second.fd.GetPathId());
+  ASSERT_EQ(3, new_files[1].second.fd.GetPathId());
+  ASSERT_EQ(0, new_files[2].second.fd.GetPathId());
+}
+
+TEST_F(VersionEditTest, ForwardCompatibleNewFile4) {
+  static const uint64_t kBig = 1ull << 50;
+  VersionEdit edit;
+  edit.AddFile(3, 300, 3, 100, InternalKey("foo", kBig + 500, kTypeValue),
+               InternalKey("zoo", kBig + 600, kTypeDeletion), kBig + 500,
+               kBig + 600, true);
+  edit.AddFile(4, 301, 3, 100, InternalKey("foo", kBig + 501, kTypeValue),
+               InternalKey("zoo", kBig + 601, kTypeDeletion), kBig + 501,
+               kBig + 601, false);
+  edit.DeleteFile(4, 700);
+
+  edit.SetComparatorName("foo");
+  edit.SetLogNumber(kBig + 100);
+  edit.SetNextFile(kBig + 200);
+  edit.SetLastSequence(kBig + 1000);
+
+  std::string encoded;
+
+  // Call back function to add extra customized builds.
+  bool first = true;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "VersionEdit::EncodeTo:NewFile4:CustomizeFields", [&](void* arg) {
+        std::string* str = reinterpret_cast<std::string*>(arg);
+        PutVarint32(str, 33);
+        const std::string str1 = "random_string";
+        PutLengthPrefixedSlice(str, str1);
+        if (first) {
+          first = false;
+          PutVarint32(str, 22);
+          const std::string str2 = "s";
+          PutLengthPrefixedSlice(str, str2);
+        }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  edit.EncodeTo(&encoded);
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  VersionEdit parsed;
+  Status s = parsed.DecodeFrom(encoded);
+  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_TRUE(!first);
+  auto& new_files = parsed.GetNewFiles();
+  ASSERT_TRUE(new_files[0].second.marked_for_compaction);
+  ASSERT_TRUE(!new_files[1].second.marked_for_compaction);
+  ASSERT_EQ(3, new_files[0].second.fd.GetPathId());
+  ASSERT_EQ(3, new_files[1].second.fd.GetPathId());
+  ASSERT_EQ(1u, parsed.GetDeletedFiles().size());
+}
+
+TEST_F(VersionEditTest, NewFile4NotSupportedField) {
+  static const uint64_t kBig = 1ull << 50;
+  VersionEdit edit;
+  edit.AddFile(3, 300, 3, 100, InternalKey("foo", kBig + 500, kTypeValue),
+               InternalKey("zoo", kBig + 600, kTypeDeletion), kBig + 500,
+               kBig + 600, true);
+
+  edit.SetComparatorName("foo");
+  edit.SetLogNumber(kBig + 100);
+  edit.SetNextFile(kBig + 200);
+  edit.SetLastSequence(kBig + 1000);
+
+  std::string encoded;
+
+  // Call back function to add extra customized builds.
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "VersionEdit::EncodeTo:NewFile4:CustomizeFields", [&](void* arg) {
+        std::string* str = reinterpret_cast<std::string*>(arg);
+        const std::string str1 = "s";
+        PutLengthPrefixedSlice(str, str1);
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  edit.EncodeTo(&encoded);
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  VersionEdit parsed;
+  Status s = parsed.DecodeFrom(encoded);
+  ASSERT_NOK(s);
+}
+
 TEST_F(VersionEditTest, EncodeEmptyFile) {
   VersionEdit edit;
   edit.AddFile(0, 0, 0, 0, InternalKey(), InternalKey(), 0, 0, false);

db/version_set.cc

@@ -35,6 +35,7 @@
 #include "db/writebuffer.h"
 #include "rocksdb/env.h"
 #include "rocksdb/merge_operator.h"
+#include "table/internal_iterator.h"
 #include "table/table_reader.h"
 #include "table/merger.h"
 #include "table/two_level_iterator.h"
@@ -420,7 +421,7 @@ namespace {
 // is the largest key that occurs in the file, and value() is an
 // 16-byte value containing the file number and file size, both
 // encoded using EncodeFixed64.
-class LevelFileNumIterator : public Iterator {
+class LevelFileNumIterator : public InternalIterator {
  public:
   LevelFileNumIterator(const InternalKeyComparator& icmp,
                        const LevelFilesBrief* flevel)
@@ -488,9 +489,9 @@ class LevelFileIteratorState : public TwoLevelIteratorState {
         file_read_hist_(file_read_hist),
         for_compaction_(for_compaction) {}
 
-  Iterator* NewSecondaryIterator(const Slice& meta_handle) override {
+  InternalIterator* NewSecondaryIterator(const Slice& meta_handle) override {
     if (meta_handle.size() != sizeof(FileDescriptor)) {
-      return NewErrorIterator(
+      return NewErrorInternalIterator(
           Status::Corruption("FileReader invoked with unexpected value"));
     } else {
       const FileDescriptor* fd =
@@ -523,7 +524,7 @@ class BaseReferencedVersionBuilder {
   explicit BaseReferencedVersionBuilder(ColumnFamilyData* cfd)
       : version_builder_(new VersionBuilder(
             cfd->current()->version_set()->env_options(), cfd->table_cache(),
-            cfd->current()->storage_info())),
+            cfd->current()->storage_info(), cfd->ioptions()->info_log)),
         version_(cfd->current()) {
     version_->Ref();
   }
@@ -541,7 +542,7 @@ class BaseReferencedVersionBuilder {
 
 Status Version::GetTableProperties(std::shared_ptr<const TableProperties>* tp,
                                    const FileMetaData* file_meta,
-                                   const std::string* fname) {
+                                   const std::string* fname) const {
   auto table_cache = cfd_->table_cache();
   auto ioptions = cfd_->ioptions();
   Status s = table_cache->GetTableProperties(
@@ -623,6 +624,38 @@ Status Version::GetPropertiesOfAllTables(TablePropertiesCollection* props,
   return Status::OK();
 }
 
+Status Version::GetPropertiesOfTablesInRange(
+    const Range* range, std::size_t n, TablePropertiesCollection* props) const {
+  for (int level = 0; level < storage_info_.num_non_empty_levels(); level++) {
+    for (decltype(n) i = 0; i < n; i++) {
+      // Convert user_key into a corresponding internal key.
+      InternalKey k1(range[i].start, kMaxSequenceNumber, kValueTypeForSeek);
+      InternalKey k2(range[i].limit, kMaxSequenceNumber, kValueTypeForSeek);
+      std::vector<FileMetaData*> files;
+      storage_info_.GetOverlappingInputs(level, &k1, &k2, &files, -1, nullptr,
+                                         false);
+      for (const auto& file_meta : files) {
+        auto fname =
+            TableFileName(vset_->db_options_->db_paths,
+                          file_meta->fd.GetNumber(), file_meta->fd.GetPathId());
+        if (props->count(fname) == 0) {
+          // 1. If the table is already present in table cache, load table
+          // properties from there.
+          std::shared_ptr<const TableProperties> table_properties;
+          Status s = GetTableProperties(&table_properties, file_meta, &fname);
+          if (s.ok()) {
+            props->insert({fname, table_properties});
+          } else {
+            return s;
+          }
+        }
+      }
+    }
+  }
+
+  return Status::OK();
+}
+
 Status Version::GetAggregatedTableProperties(
     std::shared_ptr<const TableProperties>* tp, int level) {
   TablePropertiesCollection props;
@@ -1064,7 +1097,7 @@ void VersionStorageInfo::EstimateCompactionBytesNeeded(
   // We keep doing it to Level 2, 3, etc, until the last level and return the
   // accumulated bytes.
 
-  size_t bytes_compact_to_next_level = 0;
+  uint64_t bytes_compact_to_next_level = 0;
   // Level 0
   bool level0_compact_triggered = false;
   if (static_cast<int>(files_[0].size()) >
@@ -1080,7 +1113,7 @@ void VersionStorageInfo::EstimateCompactionBytesNeeded(
 
   // Level 1 and up.
   for (int level = base_level(); level <= MaxInputLevel(); level++) {
-    size_t level_size = 0;
+    uint64_t level_size = 0;
     for (auto* f : files_[level]) {
       level_size += f->fd.GetFileSize();
     }
@@ -1091,7 +1124,7 @@ void VersionStorageInfo::EstimateCompactionBytesNeeded(
     // Add size added by previous compaction
     level_size += bytes_compact_to_next_level;
     bytes_compact_to_next_level = 0;
-    size_t level_target = MaxBytesForLevel(level);
+    uint64_t level_target = MaxBytesForLevel(level);
     if (level_size > level_target) {
       bytes_compact_to_next_level = level_size - level_target;
       // Simplify to assume the actual compaction fan-out ratio is always
@@ -1229,13 +1262,27 @@ bool CompareCompensatedSizeDescending(const Fsize& first, const Fsize& second) {
 }
 } // anonymous namespace
 
-void VersionStorageInfo::AddFile(int level, FileMetaData* f) {
+void VersionStorageInfo::AddFile(int level, FileMetaData* f, Logger* info_log) {
   auto* level_files = &files_[level];
   // Must not overlap
-  assert(level <= 0 || level_files->empty() ||
-         internal_comparator_->Compare(
-             (*level_files)[level_files->size() - 1]->largest, f->smallest) <
-             0);
+#ifndef NDEBUG
+  if (level > 0 && !level_files->empty() &&
+      internal_comparator_->Compare(
+          (*level_files)[level_files->size() - 1]->largest, f->smallest) >= 0) {
+    auto* f2 = (*level_files)[level_files->size() - 1];
+    if (info_log != nullptr) {
+      Error(info_log, "Adding new file %" PRIu64
+                      " range (%s, %s) to level %d but overlapping "
+                      "with existing file %" PRIu64 " %s %s",
+            f->fd.GetNumber(), f->smallest.DebugString(true).c_str(),
+            f->largest.DebugString(true).c_str(), level, f2->fd.GetNumber(),
+            f2->smallest.DebugString(true).c_str(),
+            f2->largest.DebugString(true).c_str());
+      LogFlush(info_log);
+    }
+    assert(false);
+  }
+#endif
   f->refs++;
   level_files->push_back(f);
 }
@@ -1405,7 +1452,8 @@ bool VersionStorageInfo::OverlapInLevel(int level,
 // The file_index returns a pointer to any file in an overlapping range.
 void VersionStorageInfo::GetOverlappingInputs(
     int level, const InternalKey* begin, const InternalKey* end,
-    std::vector<FileMetaData*>* inputs, int hint_index, int* file_index) {
+    std::vector<FileMetaData*>* inputs, int hint_index, int* file_index,
+    bool expand_range) const {
   if (level >= num_non_empty_levels_) {
     // this level is empty, no overlapping inputs
     return;
@@ -1438,7 +1486,7 @@ void VersionStorageInfo::GetOverlappingInputs(
       // "f" is completely after specified range; skip it
     } else {
       inputs->push_back(files_[level][i-1]);
-      if (level == 0) {
+      if (level == 0 && expand_range) {
         // Level-0 files may overlap each other.  So check if the newly
         // added file has expanded the range.  If so, restart search.
         if (begin != nullptr && user_cmp->Compare(file_start, user_begin) < 0) {
@@ -1464,7 +1512,7 @@ void VersionStorageInfo::GetOverlappingInputs(
 // forwards to find all overlapping files.
 void VersionStorageInfo::GetOverlappingInputsBinarySearch(
     int level, const Slice& user_begin, const Slice& user_end,
-    std::vector<FileMetaData*>* inputs, int hint_index, int* file_index) {
+    std::vector<FileMetaData*>* inputs, int hint_index, int* file_index) const {
   assert(level > 0);
   int min = 0;
   int mid = 0;
@@ -1512,8 +1560,7 @@ void VersionStorageInfo::GetOverlappingInputsBinarySearch(
 // Use FileLevel in searching, make it faster
 void VersionStorageInfo::ExtendOverlappingInputs(
     int level, const Slice& user_begin, const Slice& user_end,
-    std::vector<FileMetaData*>* inputs, unsigned int midIndex) {
-
+    std::vector<FileMetaData*>* inputs, unsigned int midIndex) const {
   const Comparator* user_cmp = user_comparator_;
   const FdWithKeyRange* files = level_files_brief_[level].files;
 #ifndef NDEBUG
@@ -2070,8 +2117,8 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
           "Creating manifest %" PRIu64 "\n", pending_manifest_file_number_);
       unique_ptr<WritableFile> descriptor_file;
       EnvOptions opt_env_opts = env_->OptimizeForManifestWrite(env_options_);
-      s = env_->NewWritableFile(
-          DescriptorFileName(dbname_, pending_manifest_file_number_),
+      s = NewWritableFile(
+          env_, DescriptorFileName(dbname_, pending_manifest_file_number_),
           &descriptor_file, opt_env_opts);
       if (s.ok()) {
         descriptor_file->SetPreallocationBlockSize(
@@ -2079,7 +2126,7 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
 
         unique_ptr<WritableFileWriter> file_writer(
             new WritableFileWriter(std::move(descriptor_file), opt_env_opts));
-        descriptor_log_.reset(new log::Writer(std::move(file_writer)));
+        descriptor_log_.reset(new log::Writer(std::move(file_writer), 0, false));
         s = WriteSnapshot(descriptor_log_.get());
       }
     }
@@ -2098,6 +2145,8 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
               "Unable to Encode VersionEdit:" + e->DebugString(true));
           break;
         }
+        TEST_KILL_RANDOM("VersionSet::LogAndApply:BeforeAddRecord",
+                         rocksdb_kill_odds * REDUCE_ODDS2);
         s = descriptor_log_->AddRecord(record);
         if (!s.ok()) {
           break;
@@ -2341,8 +2390,8 @@ Status VersionSet::Recover(
   {
     VersionSet::LogReporter reporter;
     reporter.status = &s;
-    log::Reader reader(std::move(manifest_file_reader), &reporter,
-                       true /*checksum*/, 0 /*initial_offset*/);
+    log::Reader reader(NULL, std::move(manifest_file_reader), &reporter,
+                       true /*checksum*/, 0 /*initial_offset*/, 0);
     Slice record;
     std::string scratch;
     while (reader.ReadRecord(&record, &scratch) && s.ok()) {
@@ -2594,8 +2643,8 @@ Status VersionSet::ListColumnFamilies(std::vector<std::string>* column_families,
   column_family_names.insert({0, kDefaultColumnFamilyName});
   VersionSet::LogReporter reporter;
   reporter.status = &s;
-  log::Reader reader(std::move(file_reader), &reporter, true /*checksum*/,
-                     0 /*initial_offset*/);
+  log::Reader reader(NULL, std::move(file_reader), &reporter, true /*checksum*/,
+                     0 /*initial_offset*/, 0);
   Slice record;
   std::string scratch;
   while (reader.ReadRecord(&record, &scratch) && s.ok()) {
@@ -2752,8 +2801,8 @@ Status VersionSet::DumpManifest(Options& options, std::string& dscname,
   {
     VersionSet::LogReporter reporter;
     reporter.status = &s;
-    log::Reader reader(std::move(file_reader), &reporter, true /*checksum*/,
-                       0 /*initial_offset*/);
+    log::Reader reader(NULL, std::move(file_reader), &reporter,
+                       true /*checksum*/, 0 /*initial_offset*/, 0);
     Slice record;
     std::string scratch;
     while (reader.ReadRecord(&record, &scratch) && s.ok()) {
@@ -3005,7 +3054,8 @@ bool VersionSet::ManifestContains(uint64_t manifest_file_num,
     }
     file_reader.reset(new SequentialFileReader(std::move(file)));
   }
-  log::Reader reader(std::move(file_reader), nullptr, true /*checksum*/, 0);
+  log::Reader reader(NULL, std::move(file_reader), nullptr,
+                     true /*checksum*/, 0, 0);
   Slice r;
   std::string scratch;
   bool result = false;
@@ -3119,7 +3169,7 @@ uint64_t VersionSet::ApproximateSize(Version* v, const FdWithKeyRange& f,
     // "key" falls in the range for this table.  Add the
     // approximate offset of "key" within the table.
     TableReader* table_reader_ptr;
-    Iterator* iter = v->cfd_->table_cache()->NewIterator(
+    InternalIterator* iter = v->cfd_->table_cache()->NewIterator(
         ReadOptions(), env_options_, v->cfd_->internal_comparator(), f.fd,
         &table_reader_ptr);
     if (table_reader_ptr != nullptr) {
@@ -3166,7 +3216,7 @@ void VersionSet::AddLiveFiles(std::vector<FileDescriptor>* live_list) {
   }
 }
 
-Iterator* VersionSet::MakeInputIterator(Compaction* c) {
+InternalIterator* VersionSet::MakeInputIterator(Compaction* c) {
   auto cfd = c->column_family_data();
   ReadOptions read_options;
   read_options.verify_checksums =
@@ -3182,7 +3232,7 @@ Iterator* VersionSet::MakeInputIterator(Compaction* c) {
   const size_t space = (c->level() == 0 ? c->input_levels(0)->num_files +
                                               c->num_input_levels() - 1
                                         : c->num_input_levels());
-  Iterator** list = new Iterator* [space];
+  InternalIterator** list = new InternalIterator* [space];
   size_t num = 0;
   for (size_t which = 0; which < c->num_input_levels(); which++) {
     if (c->input_levels(which)->num_files != 0) {
@@ -3209,7 +3259,7 @@ Iterator* VersionSet::MakeInputIterator(Compaction* c) {
     }
   }
   assert(num <= space);
-  Iterator* result =
+  InternalIterator* result =
       NewMergingIterator(&c->column_family_data()->internal_comparator(), list,
                          static_cast<int>(num));
   delete[] list;

db/version_set.h

@@ -48,7 +48,7 @@ class Writer;
 }
 
 class Compaction;
-class Iterator;
+class InternalIterator;
 class LogBuffer;
 class LookupKey;
 class MemTable;
@@ -97,7 +97,7 @@ class VersionStorageInfo {
 
   void Reserve(int level, size_t size) { files_[level].reserve(size); }
 
-  void AddFile(int level, FileMetaData* f);
+  void AddFile(int level, FileMetaData* f, Logger* info_log = nullptr);
 
   void SetFinalized();
 
@@ -159,23 +159,26 @@ class VersionStorageInfo {
       int level, const InternalKey* begin,  // nullptr means before all keys
       const InternalKey* end,               // nullptr means after all keys
       std::vector<FileMetaData*>* inputs,
-      int hint_index = -1,         // index of overlap file
-      int* file_index = nullptr);  // return index of overlap file
+      int hint_index = -1,        // index of overlap file
+      int* file_index = nullptr,  // return index of overlap file
+      bool expand_range = true)   // if set, returns files which overlap the
+      const;                      // range and overlap each other. If false,
+                                  // then just files intersecting the range
 
   void GetOverlappingInputsBinarySearch(
       int level,
       const Slice& begin,  // nullptr means before all keys
       const Slice& end,    // nullptr means after all keys
       std::vector<FileMetaData*>* inputs,
-      int hint_index,    // index of overlap file
-      int* file_index);  // return index of overlap file
+      int hint_index,          // index of overlap file
+      int* file_index) const;  // return index of overlap file
 
   void ExtendOverlappingInputs(
       int level,
       const Slice& begin,  // nullptr means before all keys
       const Slice& end,    // nullptr means after all keys
       std::vector<FileMetaData*>* inputs,
-      unsigned int index);  // start extending from this index
+      unsigned int index) const;  // start extending from this index
 
   // Returns true iff some file in the specified level overlaps
   // some part of [*smallest_user_key,*largest_user_key].
@@ -456,15 +459,16 @@ class Version {
   // file-name conversion.
   Status GetTableProperties(std::shared_ptr<const TableProperties>* tp,
                             const FileMetaData* file_meta,
-                            const std::string* fname = nullptr);
+                            const std::string* fname = nullptr) const;
 
   // REQUIRES: lock is held
   // On success, *props will be populated with all SSTables' table properties.
   // The keys of `props` are the sst file name, the values of `props` are the
   // tables' propertis, represented as shared_ptr.
   Status GetPropertiesOfAllTables(TablePropertiesCollection* props);
-
   Status GetPropertiesOfAllTables(TablePropertiesCollection* props, int level);
+  Status GetPropertiesOfTablesInRange(const Range* range, std::size_t n,
+                                      TablePropertiesCollection* props) const;
 
   // REQUIRES: lock is held
   // On success, "tp" will contains the aggregated table property amoug
@@ -502,7 +506,8 @@ class Version {
     return storage_info_.user_comparator_;
   }
 
-  bool PrefixMayMatch(const ReadOptions& read_options, Iterator* level_iter,
+  bool PrefixMayMatch(const ReadOptions& read_options,
+                      InternalIterator* level_iter,
                       const Slice& internal_prefix) const;
 
   // The helper function of UpdateAccumulatedStats, which may fill the missing
@@ -643,7 +648,7 @@ class VersionSet {
 
   // Create an iterator that reads over the compaction inputs for "*c".
   // The caller should delete the iterator when no longer needed.
-  Iterator* MakeInputIterator(Compaction* c);
+  InternalIterator* MakeInputIterator(Compaction* c);
 
   // Add all files listed in any live version to *live.
   void AddLiveFiles(std::vector<FileDescriptor>* live_list);

db/wal_manager.cc

@@ -448,8 +448,8 @@ Status WalManager::ReadFirstLine(const std::string& fname,
   reporter.fname = fname.c_str();
   reporter.status = &status;
   reporter.ignore_error = !db_options_.paranoid_checks;
-  log::Reader reader(std::move(file_reader), &reporter, true /*checksum*/,
-                     0 /*initial_offset*/);
+  log::Reader reader(db_options_.info_log, std::move(file_reader), &reporter,
+                     true /*checksum*/, 0 /*initial_offset*/, *sequence);
   std::string scratch;
   Slice record;
 

db/wal_manager_test.cc

@@ -77,7 +77,7 @@ class WalManagerTest : public testing::Test {
     ASSERT_OK(env_->NewWritableFile(fname, &file, env_options_));
     unique_ptr<WritableFileWriter> file_writer(
         new WritableFileWriter(std::move(file), env_options_));
-    current_log_writer_.reset(new log::Writer(std::move(file_writer)));
+    current_log_writer_.reset(new log::Writer(std::move(file_writer), 0, false));
   }
 
   void CreateArchiveLogs(int num_logs, int entries_per_log) {
@@ -127,7 +127,8 @@ TEST_F(WalManagerTest, ReadFirstRecordCache) {
 
   unique_ptr<WritableFileWriter> file_writer(
       new WritableFileWriter(std::move(file), EnvOptions()));
-  log::Writer writer(std::move(file_writer));
+  log::Writer writer(std::move(file_writer), 1,
+		     db_options_.recycle_log_file_num > 0);
   WriteBatch batch;
   batch.Put("foo", "bar");
   WriteBatchInternal::SetSequence(&batch, 10);

db/write_batch.cc

@@ -42,22 +42,91 @@
 
 namespace rocksdb {
 
+// anon namespace for file-local types
+namespace {
+
+enum ContentFlags : uint32_t {
+  DEFERRED = 1,
+  HAS_PUT = 2,
+  HAS_DELETE = 4,
+  HAS_SINGLE_DELETE = 8,
+  HAS_MERGE = 16,
+};
+
+struct BatchContentClassifier : public WriteBatch::Handler {
+  uint32_t content_flags = 0;
+
+  Status PutCF(uint32_t, const Slice&, const Slice&) override {
+    content_flags |= ContentFlags::HAS_PUT;
+    return Status::OK();
+  }
+
+  Status DeleteCF(uint32_t, const Slice&) override {
+    content_flags |= ContentFlags::HAS_DELETE;
+    return Status::OK();
+  }
+
+  Status SingleDeleteCF(uint32_t, const Slice&) override {
+    content_flags |= ContentFlags::HAS_SINGLE_DELETE;
+    return Status::OK();
+  }
+
+  Status MergeCF(uint32_t, const Slice&, const Slice&) override {
+    content_flags |= ContentFlags::HAS_MERGE;
+    return Status::OK();
+  }
+};
+
+}  // anon namespace
+
 // WriteBatch header has an 8-byte sequence number followed by a 4-byte count.
 static const size_t kHeader = 12;
 
 struct SavePoint {
   size_t size;  // size of rep_
   int count;    // count of elements in rep_
-  SavePoint(size_t s, int c) : size(s), count(c) {}
+  uint32_t content_flags;
 };
 
 struct SavePoints {
   std::stack<SavePoint> stack;
 };
 
-WriteBatch::WriteBatch(size_t reserved_bytes) : save_points_(nullptr) {
+WriteBatch::WriteBatch(size_t reserved_bytes)
+    : save_points_(nullptr), content_flags_(0), rep_() {
   rep_.reserve((reserved_bytes > kHeader) ? reserved_bytes : kHeader);
-  Clear();
+  rep_.resize(kHeader);
+}
+
+WriteBatch::WriteBatch(const std::string& rep)
+    : save_points_(nullptr),
+      content_flags_(ContentFlags::DEFERRED),
+      rep_(rep) {}
+
+WriteBatch::WriteBatch(const WriteBatch& src)
+    : save_points_(src.save_points_),
+      content_flags_(src.content_flags_.load(std::memory_order_relaxed)),
+      rep_(src.rep_) {}
+
+WriteBatch::WriteBatch(WriteBatch&& src)
+    : save_points_(std::move(src.save_points_)),
+      content_flags_(src.content_flags_.load(std::memory_order_relaxed)),
+      rep_(std::move(src.rep_)) {}
+
+WriteBatch& WriteBatch::operator=(const WriteBatch& src) {
+  if (&src != this) {
+    this->~WriteBatch();
+    new (this) WriteBatch(src);
+  }
+  return *this;
+}
+
+WriteBatch& WriteBatch::operator=(WriteBatch&& src) {
+  if (&src != this) {
+    this->~WriteBatch();
+    new (this) WriteBatch(std::move(src));
+  }
+  return *this;
 }
 
 WriteBatch::~WriteBatch() {
@@ -81,6 +150,8 @@ void WriteBatch::Clear() {
   rep_.clear();
   rep_.resize(kHeader);
 
+  content_flags_.store(0, std::memory_order_relaxed);
+
   if (save_points_ != nullptr) {
     while (!save_points_->stack.empty()) {
       save_points_->stack.pop();
@@ -92,6 +163,38 @@ int WriteBatch::Count() const {
   return WriteBatchInternal::Count(this);
 }
 
+uint32_t WriteBatch::ComputeContentFlags() const {
+  auto rv = content_flags_.load(std::memory_order_relaxed);
+  if ((rv & ContentFlags::DEFERRED) != 0) {
+    BatchContentClassifier classifier;
+    Iterate(&classifier);
+    rv = classifier.content_flags;
+
+    // this method is conceptually const, because it is performing a lazy
+    // computation that doesn't affect the abstract state of the batch.
+    // content_flags_ is marked mutable so that we can perform the
+    // following assignment
+    content_flags_.store(rv, std::memory_order_relaxed);
+  }
+  return rv;
+}
+
+bool WriteBatch::HasPut() const {
+  return (ComputeContentFlags() & ContentFlags::HAS_PUT) != 0;
+}
+
+bool WriteBatch::HasDelete() const {
+  return (ComputeContentFlags() & ContentFlags::HAS_DELETE) != 0;
+}
+
+bool WriteBatch::HasSingleDelete() const {
+  return (ComputeContentFlags() & ContentFlags::HAS_SINGLE_DELETE) != 0;
+}
+
+bool WriteBatch::HasMerge() const {
+  return (ComputeContentFlags() & ContentFlags::HAS_MERGE) != 0;
+}
+
 Status ReadRecordFromWriteBatch(Slice* input, char* tag,
                                 uint32_t* column_family, Slice* key,
                                 Slice* value, Slice* blob) {
@@ -169,21 +272,29 @@ Status WriteBatch::Iterate(Handler* handler) const {
     switch (tag) {
       case kTypeColumnFamilyValue:
       case kTypeValue:
+        assert(content_flags_.load(std::memory_order_relaxed) &
+               (ContentFlags::DEFERRED | ContentFlags::HAS_PUT));
         s = handler->PutCF(column_family, key, value);
         found++;
         break;
       case kTypeColumnFamilyDeletion:
       case kTypeDeletion:
+        assert(content_flags_.load(std::memory_order_relaxed) &
+               (ContentFlags::DEFERRED | ContentFlags::HAS_DELETE));
         s = handler->DeleteCF(column_family, key);
         found++;
         break;
       case kTypeColumnFamilySingleDeletion:
       case kTypeSingleDeletion:
+        assert(content_flags_.load(std::memory_order_relaxed) &
+               (ContentFlags::DEFERRED | ContentFlags::HAS_SINGLE_DELETE));
         s = handler->SingleDeleteCF(column_family, key);
         found++;
         break;
       case kTypeColumnFamilyMerge:
       case kTypeMerge:
+        assert(content_flags_.load(std::memory_order_relaxed) &
+               (ContentFlags::DEFERRED | ContentFlags::HAS_MERGE));
         s = handler->MergeCF(column_family, key, value);
         found++;
         break;
@@ -233,6 +344,9 @@ void WriteBatchInternal::Put(WriteBatch* b, uint32_t column_family_id,
   }
   PutLengthPrefixedSlice(&b->rep_, key);
   PutLengthPrefixedSlice(&b->rep_, value);
+  b->content_flags_.store(
+      b->content_flags_.load(std::memory_order_relaxed) | ContentFlags::HAS_PUT,
+      std::memory_order_relaxed);
 }
 
 void WriteBatch::Put(ColumnFamilyHandle* column_family, const Slice& key,
@@ -251,6 +365,9 @@ void WriteBatchInternal::Put(WriteBatch* b, uint32_t column_family_id,
   }
   PutLengthPrefixedSliceParts(&b->rep_, key);
   PutLengthPrefixedSliceParts(&b->rep_, value);
+  b->content_flags_.store(
+      b->content_flags_.load(std::memory_order_relaxed) | ContentFlags::HAS_PUT,
+      std::memory_order_relaxed);
 }
 
 void WriteBatch::Put(ColumnFamilyHandle* column_family, const SliceParts& key,
@@ -268,6 +385,9 @@ void WriteBatchInternal::Delete(WriteBatch* b, uint32_t column_family_id,
     PutVarint32(&b->rep_, column_family_id);
   }
   PutLengthPrefixedSlice(&b->rep_, key);
+  b->content_flags_.store(b->content_flags_.load(std::memory_order_relaxed) |
+                              ContentFlags::HAS_DELETE,
+                          std::memory_order_relaxed);
 }
 
 void WriteBatch::Delete(ColumnFamilyHandle* column_family, const Slice& key) {
@@ -284,6 +404,9 @@ void WriteBatchInternal::Delete(WriteBatch* b, uint32_t column_family_id,
     PutVarint32(&b->rep_, column_family_id);
   }
   PutLengthPrefixedSliceParts(&b->rep_, key);
+  b->content_flags_.store(b->content_flags_.load(std::memory_order_relaxed) |
+                              ContentFlags::HAS_DELETE,
+                          std::memory_order_relaxed);
 }
 
 void WriteBatch::Delete(ColumnFamilyHandle* column_family,
@@ -301,6 +424,9 @@ void WriteBatchInternal::SingleDelete(WriteBatch* b, uint32_t column_family_id,
     PutVarint32(&b->rep_, column_family_id);
   }
   PutLengthPrefixedSlice(&b->rep_, key);
+  b->content_flags_.store(b->content_flags_.load(std::memory_order_relaxed) |
+                              ContentFlags::HAS_SINGLE_DELETE,
+                          std::memory_order_relaxed);
 }
 
 void WriteBatch::SingleDelete(ColumnFamilyHandle* column_family,
@@ -318,6 +444,9 @@ void WriteBatchInternal::SingleDelete(WriteBatch* b, uint32_t column_family_id,
     PutVarint32(&b->rep_, column_family_id);
   }
   PutLengthPrefixedSliceParts(&b->rep_, key);
+  b->content_flags_.store(b->content_flags_.load(std::memory_order_relaxed) |
+                              ContentFlags::HAS_SINGLE_DELETE,
+                          std::memory_order_relaxed);
 }
 
 void WriteBatch::SingleDelete(ColumnFamilyHandle* column_family,
@@ -336,6 +465,9 @@ void WriteBatchInternal::Merge(WriteBatch* b, uint32_t column_family_id,
   }
   PutLengthPrefixedSlice(&b->rep_, key);
   PutLengthPrefixedSlice(&b->rep_, value);
+  b->content_flags_.store(b->content_flags_.load(std::memory_order_relaxed) |
+                              ContentFlags::HAS_MERGE,
+                          std::memory_order_relaxed);
 }
 
 void WriteBatch::Merge(ColumnFamilyHandle* column_family, const Slice& key,
@@ -355,6 +487,9 @@ void WriteBatchInternal::Merge(WriteBatch* b, uint32_t column_family_id,
   }
   PutLengthPrefixedSliceParts(&b->rep_, key);
   PutLengthPrefixedSliceParts(&b->rep_, value);
+  b->content_flags_.store(b->content_flags_.load(std::memory_order_relaxed) |
+                              ContentFlags::HAS_MERGE,
+                          std::memory_order_relaxed);
 }
 
 void WriteBatch::Merge(ColumnFamilyHandle* column_family,
@@ -374,7 +509,8 @@ void WriteBatch::SetSavePoint() {
     save_points_ = new SavePoints();
   }
   // Record length and count of current batch of writes.
-  save_points_->stack.push(SavePoint(GetDataSize(), Count()));
+  save_points_->stack.push(SavePoint{
+      GetDataSize(), Count(), content_flags_.load(std::memory_order_relaxed)});
 }
 
 Status WriteBatch::RollbackToSavePoint() {
@@ -387,6 +523,7 @@ Status WriteBatch::RollbackToSavePoint() {
   save_points_->stack.pop();
 
   assert(savepoint.size <= rep_.size());
+  assert(savepoint.count <= Count());
 
   if (savepoint.size == rep_.size()) {
     // No changes to rollback
@@ -396,6 +533,7 @@ Status WriteBatch::RollbackToSavePoint() {
   } else {
     rep_.resize(savepoint.size);
     WriteBatchInternal::SetCount(this, savepoint.count);
+    content_flags_.store(savepoint.content_flags, std::memory_order_relaxed);
   }
 
   return Status::OK();
@@ -453,6 +591,7 @@ class MemTableInserter : public WriteBatch::Handler {
     }
     return true;
   }
+
   virtual Status PutCF(uint32_t column_family_id, const Slice& key,
                        const Slice& value) override {
     Status seek_status;
@@ -509,8 +648,8 @@ class MemTableInserter : public WriteBatch::Handler {
     return Status::OK();
   }
 
-  virtual Status DeleteCF(uint32_t column_family_id,
-                          const Slice& key) override {
+  Status DeleteImpl(uint32_t column_family_id, const Slice& key,
+                    ValueType delete_type) {
     Status seek_status;
     if (!SeekToColumnFamily(column_family_id, &seek_status)) {
       ++sequence_;
@@ -533,40 +672,20 @@ class MemTableInserter : public WriteBatch::Handler {
         return Status::OK();
       }
     }
-    mem->Add(sequence_, kTypeDeletion, key, Slice());
+    mem->Add(sequence_, delete_type, key, Slice());
     sequence_++;
     cf_mems_->CheckMemtableFull();
     return Status::OK();
   }
 
+  virtual Status DeleteCF(uint32_t column_family_id,
+                          const Slice& key) override {
+    return DeleteImpl(column_family_id, key, kTypeDeletion);
+  }
+
   virtual Status SingleDeleteCF(uint32_t column_family_id,
                                 const Slice& key) override {
-    Status seek_status;
-    if (!SeekToColumnFamily(column_family_id, &seek_status)) {
-      ++sequence_;
-      return seek_status;
-    }
-    MemTable* mem = cf_mems_->GetMemTable();
-    auto* moptions = mem->GetMemTableOptions();
-    if (!dont_filter_deletes_ && moptions->filter_deletes) {
-      SnapshotImpl read_from_snapshot;
-      read_from_snapshot.number_ = sequence_;
-      ReadOptions ropts;
-      ropts.snapshot = &read_from_snapshot;
-      std::string value;
-      auto cf_handle = cf_mems_->GetColumnFamilyHandle();
-      if (cf_handle == nullptr) {
-        cf_handle = db_->DefaultColumnFamily();
-      }
-      if (!db_->KeyMayExist(ropts, cf_handle, key, &value)) {
-        RecordTick(moptions->statistics, NUMBER_FILTERED_DELETES);
-        return Status::OK();
-      }
-    }
-    mem->Add(sequence_, kTypeSingleDeletion, key, Slice());
-    sequence_++;
-    cf_mems_->CheckMemtableFull();
-    return Status::OK();
+    return DeleteImpl(column_family_id, key, kTypeSingleDeletion);
   }
 
   virtual Status MergeCF(uint32_t column_family_id, const Slice& key,
@@ -653,29 +772,57 @@ class MemTableInserter : public WriteBatch::Handler {
 
 // This function can only be called in these conditions:
 // 1) During Recovery()
-// 2) during Write(), in a single-threaded write thread
-// The reason is that it calles ColumnFamilyMemTablesImpl::Seek(), which needs
-// to be called from a single-threaded write thread (or while holding DB mutex)
-Status WriteBatchInternal::InsertInto(const WriteBatch* b,
+// 2) During Write(), in a single-threaded write thread
+// The reason is that it calls memtables->Seek(), which has a stateful cache
+Status WriteBatchInternal::InsertInto(const autovector<WriteBatch*>& batches,
+                                      SequenceNumber sequence,
                                       ColumnFamilyMemTables* memtables,
                                       bool ignore_missing_column_families,
                                       uint64_t log_number, DB* db,
                                       const bool dont_filter_deletes) {
-  MemTableInserter inserter(WriteBatchInternal::Sequence(b), memtables,
+  MemTableInserter inserter(sequence, memtables, ignore_missing_column_families,
+                            log_number, db, dont_filter_deletes);
+  Status rv = Status::OK();
+  for (size_t i = 0; i < batches.size() && rv.ok(); ++i) {
+    rv = batches[i]->Iterate(&inserter);
+  }
+  return rv;
+}
+
+Status WriteBatchInternal::InsertInto(const WriteBatch* batch,
+                                      ColumnFamilyMemTables* memtables,
+                                      bool ignore_missing_column_families,
+                                      uint64_t log_number, DB* db,
+                                      const bool dont_filter_deletes) {
+  MemTableInserter inserter(WriteBatchInternal::Sequence(batch), memtables,
                             ignore_missing_column_families, log_number, db,
                             dont_filter_deletes);
-  return b->Iterate(&inserter);
+  return batch->Iterate(&inserter);
 }
 
 void WriteBatchInternal::SetContents(WriteBatch* b, const Slice& contents) {
   assert(contents.size() >= kHeader);
   b->rep_.assign(contents.data(), contents.size());
+  b->content_flags_.store(ContentFlags::DEFERRED, std::memory_order_relaxed);
 }
 
 void WriteBatchInternal::Append(WriteBatch* dst, const WriteBatch* src) {
   SetCount(dst, Count(dst) + Count(src));
   assert(src->rep_.size() >= kHeader);
   dst->rep_.append(src->rep_.data() + kHeader, src->rep_.size() - kHeader);
+  dst->content_flags_.store(
+      dst->content_flags_.load(std::memory_order_relaxed) |
+          src->content_flags_.load(std::memory_order_relaxed),
+      std::memory_order_relaxed);
+}
+
+size_t WriteBatchInternal::AppendedByteSize(size_t leftByteSize,
+                                            size_t rightByteSize) {
+  if (leftByteSize == 0 || rightByteSize == 0) {
+    return leftByteSize + rightByteSize;
+  } else {
+    return leftByteSize + rightByteSize - kHeader;
+  }
 }
 
 }  // namespace rocksdb

db/write_batch_internal.h

@@ -12,6 +12,7 @@
 #include "rocksdb/write_batch.h"
 #include "rocksdb/db.h"
 #include "rocksdb/options.h"
+#include "util/autovector.h"
 
 namespace rocksdb {
 
@@ -112,17 +113,28 @@ class WriteBatchInternal {
 
   static void SetContents(WriteBatch* batch, const Slice& contents);
 
-  // Inserts batch entries into memtable
-  // If dont_filter_deletes is false AND options.filter_deletes is true,
-  // then --> Drops deletes in batch if db->KeyMayExist returns false
-  // If ignore_missing_column_families == true. WriteBatch referencing
-  // non-existing column family should be ignored.
-  // However, if ignore_missing_column_families == false, any WriteBatch
-  // referencing non-existing column family will return a InvalidArgument()
-  // failure.
+  // Inserts batches[i] into memtable, for i in 0..num_batches-1 inclusive.
+  //
+  // If dont_filter_deletes is false AND options.filter_deletes is true
+  // AND db->KeyMayExist is false, then a Delete won't modify the memtable.
+  //
+  // If ignore_missing_column_families == true. WriteBatch
+  // referencing non-existing column family will be ignored.
+  // If ignore_missing_column_families == false, processing of the
+  // batches will be stopped if a reference is found to a non-existing
+  // column family and InvalidArgument() will be returned.  The writes
+  // in batches may be only partially applied at that point.
   //
   // If log_number is non-zero, the memtable will be updated only if
-  // memtables->GetLogNumber() >= log_number
+  // memtables->GetLogNumber() >= log_number.
+  static Status InsertInto(const autovector<WriteBatch*>& batches,
+                           SequenceNumber sequence,
+                           ColumnFamilyMemTables* memtables,
+                           bool ignore_missing_column_families = false,
+                           uint64_t log_number = 0, DB* db = nullptr,
+                           const bool dont_filter_deletes = true);
+
+  // Convenience form of InsertInto when you have only one batch
   static Status InsertInto(const WriteBatch* batch,
                            ColumnFamilyMemTables* memtables,
                            bool ignore_missing_column_families = false,
@@ -130,6 +142,10 @@ class WriteBatchInternal {
                            const bool dont_filter_deletes = true);
 
   static void Append(WriteBatch* dst, const WriteBatch* src);
+
+  // Returns the byte size of appending a WriteBatch with ByteSize
+  // leftByteSize and a WriteBatch with ByteSize rightByteSize
+  static size_t AppendedByteSize(size_t leftByteSize, size_t rightByteSize);
 };
 
 }  // namespace rocksdb