Disable error as warning

Add one more #include<functional>
Add some include<functional>
2026-11-06 11:02:36 -05:00 · 2019-11-05 11:03:59 -08:00 · 2019-11-05 11:03:58 -08:00 · 2019-10-31 14:29:25 -07:00 · 2019-10-31 14:29:24 -07:00 · 2017-07-17 22:22:06 -07:00
735 changed files with 35055 additions and 16465 deletions
@@ -42,6 +42,7 @@ unity.a
 tags
 rocksdb_dump
 rocksdb_undump
+db_test2

 java/out
 java/target
@@ -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 && make clean && OPT="-DTRAVIS -DROCKSDB_LITE" V=1 make -j4 check
+  - 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 static_lib

 notifications:
    email:
@@ -13,7 +13,8 @@
 #        cd build
 # 3. Run cmake to generate project files for Windows, add more options to enable required third-party libraries.
 #    See thirdparty.inc for more information.
-#        sample command: cmake -G "Visual Studio 12 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 ..
+#        sample command: cmake -G "Visual Studio 12 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 -DJNI=1 ..
+#        OR for VS Studio 15 cmake -G "Visual Studio 14 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 -DJNI=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
@@ -29,10 +30,9 @@ cmake_minimum_required(VERSION 2.6)
 project(rocksdb)

 include(${CMAKE_CURRENT_SOURCE_DIR}/thirdparty.inc)
-
-execute_process(COMMAND $ENV{COMSPEC} " /C date /T" OUTPUT_VARIABLE DATE)
-execute_process(COMMAND $ENV{COMSPEC} " /C time /T" OUTPUT_VARIABLE TIME)
-string(REGEX REPLACE "(..)/(..)/..(..).*" "\\1/\\2/\\3" DATE ${DATE})
+execute_process(COMMAND powershell -Command "Get-Date -format MM_dd_yyyy" OUTPUT_VARIABLE DATE)
+execute_process(COMMAND powershell -Command "Get-Date -format HH:mm:ss" OUTPUT_VARIABLE TIME)
+string(REGEX REPLACE "(..)_(..)_..(..).*" "\\1/\\2/\\3" DATE ${DATE})
 string(REGEX REPLACE "(..):(.....).*" " \\1:\\2" TIME ${TIME})
 string(CONCAT GIT_DATE_TIME ${DATE} ${TIME})

@@ -58,7 +58,8 @@ 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 /wd4127 /wd4244 /wd4267 /wd4800 /wd4804 /wd4996")
+
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /W3 /WX /wd4127 /wd4800 /wd4996 /wd4351")

 # 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
@@ -97,6 +98,7 @@ add_subdirectory(third-party/gtest-1.7.0/fused-src/gtest)

 # Main library source code
 set(SOURCES
+        db/auto_roll_logger.cc
        db/builder.cc
        db/c.cc
        db/column_family.cc
@@ -112,6 +114,7 @@ set(SOURCES
        db/db_impl_debug.cc
        db/db_impl_experimental.cc
        db/db_impl_readonly.cc
+        db/db_info_dumper.cc
        db/db_iter.cc
        db/event_helpers.cc
        db/experimental.cc
@@ -143,9 +146,12 @@ set(SOURCES
        db/write_batch_base.cc
        db/write_controller.cc
        db/write_thread.cc
+        db/xfunc_test_points.cc
        memtable/hash_cuckoo_rep.cc
        memtable/hash_linklist_rep.cc
        memtable/hash_skiplist_rep.cc
+        memtable/skiplistrep.cc
+        memtable/vectorrep.cc
        port/stack_trace.cc
        port/win/env_win.cc
        port/win/port_win.cc
@@ -171,7 +177,6 @@ set(SOURCES
        table/merger.cc
        table/sst_file_writer.cc
        table/meta_blocks.cc
-        table/mock_table.cc
        table/plain_table_builder.cc
        table/plain_table_factory.cc
        table/plain_table_index.cc
@@ -180,27 +185,29 @@ set(SOURCES
        table/table_properties.cc
        table/two_level_iterator.cc
        tools/sst_dump_tool.cc
+        tools/db_bench_tool.cc
        tools/dump/db_dump_tool.cc
        util/arena.cc
-        util/auto_roll_logger.cc
        util/bloom.cc
        util/build_version.cc
        util/cache.cc
        util/coding.cc
        util/compaction_job_stats_impl.cc
        util/comparator.cc
+        util/concurrent_arena.cc
        util/crc32c.cc
-        util/db_info_dumper.cc
-        util/delete_scheduler_impl.cc
+        util/delete_scheduler.cc
        util/dynamic_bloom.cc
        util/env.cc
        util/env_hdfs.cc
        util/event_logger.cc
        util/file_util.cc
        util/file_reader_writer.cc
+        util/sst_file_manager_impl.cc
        util/filter_policy.cc
        util/hash.cc
        util/histogram.cc
+        util/histogram_windowing.cc
        util/instrumented_mutex.cc
        util/iostats_context.cc
        tools/ldb_cmd.cc
@@ -208,7 +215,6 @@ set(SOURCES
        util/logging.cc
        util/log_buffer.cc
        util/memenv.cc
-        util/mock_env.cc
        util/murmurhash.cc
        util/mutable_cf_options.cc
        util/options.cc
@@ -218,8 +224,8 @@ set(SOURCES
        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/statistics.cc
        util/status.cc
@@ -233,14 +239,16 @@ set(SOURCES
        util/thread_status_updater.cc
        util/thread_status_util.cc
        util/thread_status_util_debug.cc
-        util/vectorrep.cc
+        util/transaction_test_util.cc
        util/xfunc.cc
        util/xxhash.cc
        utilities/backupable/backupable_db.cc
        utilities/checkpoint/checkpoint.cc
+        utilities/compaction_filters/remove_emptyvalue_compactionfilter.cc
        utilities/document/document_db.cc
        utilities/document/json_document.cc
        utilities/document/json_document_builder.cc
+        utilities/env_mirror.cc
        utilities/flashcache/flashcache.cc
        utilities/geodb/geodb_impl.cc
        utilities/leveldb_options/leveldb_options.cc
@@ -249,6 +257,7 @@ set(SOURCES
        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
@@ -269,6 +278,8 @@ set(SOURCES
 # and linked to tests. Add test only code that is not #ifdefed for Release here.
 set(TESTUTIL_SOURCE
    db/db_test_util.cc
+    table/mock_table.cc
+    util/mock_env.cc
    util/thread_status_updater_debug.cc
 )

@@ -281,8 +292,19 @@ set_target_properties(rocksdb${ARTIFACT_SUFFIX} PROPERTIES COMPILE_FLAGS "-DROCK
 add_dependencies(rocksdb${ARTIFACT_SUFFIX} GenerateBuildVersion)
 target_link_libraries(rocksdb${ARTIFACT_SUFFIX} ${LIBS})

+if (DEFINED JNI)
+  if (${JNI} EQUAL 1)
+    message(STATUS "JNI library is enabled")
+    add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/java)
+  else()
+    message(STATUS "JNI library is disabled")
+  endif()
+else()
+  message(STATUS "JNI library is disabled")
+endif()
+
 set(APPS
-        db/db_bench.cc
+        tools/db_bench.cc
        db/memtablerep_bench.cc
        table/table_reader_bench.cc
        tools/db_stress.cc
@@ -297,40 +319,48 @@ set(APPS
 set(C_TESTS db/c_test.c)

 set(TESTS
+        db/auto_roll_logger_test.cc
        db/column_family_test.cc
        db/compact_files_test.cc
+        db/compaction_iterator_test.cc
        db/compaction_job_test.cc
        db/compaction_job_stats_test.cc
        db/compaction_picker_test.cc
        db/comparator_db_test.cc
        db/corruption_test.cc
        db/cuckoo_table_db_test.cc
-        db/db_iter_test.cc
-        db/db_test.cc
        db/db_compaction_filter_test.cc
        db/db_compaction_test.cc
        db/db_dynamic_level_test.cc
        db/db_inplace_update_test.cc
+        db/db_iter_test.cc
        db/db_log_iter_test.cc
+        db/db_properties_test.cc
+        db/db_table_properties_test.cc
+        db/db_tailing_iter_test.cc
+        db/db_test.cc
+        db/db_test2.cc
+        db/db_block_cache_test.cc
        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
+        db/repair_test.cc
        db/skiplist_test.cc
        db/table_properties_collector_test.cc
        db/version_builder_test.cc
@@ -354,11 +384,11 @@ set(TESTS
        tools/sst_dump_test.cc
        util/arena_test.cc
        util/autovector_test.cc
-        util/auto_roll_logger_test.cc
        util/bloom_test.cc
        util/cache_test.cc
        util/coding_test.cc
        util/crc32c_test.cc
+        util/delete_scheduler_test.cc
        util/dynamic_bloom_test.cc
        util/env_test.cc
        util/event_logger_test.cc
@@ -366,8 +396,10 @@ set(TESTS
        util/file_reader_writer_test.cc
        util/heap_test.cc
        util/histogram_test.cc
+        util/iostats_context_test.cc
        util/memenv_test.cc
        util/mock_env_test.cc
+        util/options_settable_test.cc
        util/options_test.cc
        util/rate_limiter_test.cc
        util/slice_transform_test.cc
@@ -380,6 +412,7 @@ set(TESTS
        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
@@ -0,0 +1,9 @@
+# RocksDB default options change log
+## 4.7.0 (4/8/2016)
+* options.write_buffer_size changes from 4MB to 64MB
+* options.target_file_size_base changes from 2MB to 64MB
+* options.max_bytes_for_level_base changes from 10MB to 256MB
+* options.soft_pending_compaction_bytes_limit changes from 0 (disabled) to 64GB
+* options.hard_pending_compaction_bytes_limit changes from 0 (disabled) to 256GB
+* table_cache_numshardbits changes from 4 to 6
+* max_file_opening_threads changes from 1 to 16 
@@ -1,16 +1,63 @@
 # Rocksdb Change Log
+## 4.7.0 (4/8/2016)
+### Public API Change
+* rename options compaction_measure_io_stats to report_bg_io_stats and include flush too.
+* Change some default options. Now default options will optimize for server-workloads. Also enable slowdown and full stop triggers for pending compaction bytes. These changes may cause sub-optimal performance or significant increase of resource usage. To avoid these risks, users can open existing RocksDB with options extracted from RocksDB option files. See https://github.com/facebook/rocksdb/wiki/RocksDB-Options-File for how to use RocksDB option files. Or you can call Options.OldDefaults() to recover old defaults. DEFAULT_OPTIONS_HISTORY.md will track change history of default options.

-## Unreleased
+## 4.6.0 (3/10/2016)
+### Public API Changes
+* Change default of BlockBasedTableOptions.format_version to 2. It means default DB created by 4.6 or up cannot be opened by RocksDB version 3.9 or earlier.
+* Added strict_capacity_limit option to NewLRUCache. If the flag is set to true, insert to cache will fail if no enough capacity can be free. Signiture of Cache::Insert() is updated accordingly.
+* Tickers [NUMBER_DB_NEXT, NUMBER_DB_PREV, NUMBER_DB_NEXT_FOUND, NUMBER_DB_PREV_FOUND, ITER_BYTES_READ] are not updated immediately. The are updated when the Iterator is deleted.  
+* Add monotonically increasing counter (DB property "rocksdb.current-super-version-number") that increments upon any change to the LSM tree.
 ### 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 CompactionPri::kMinOverlappingRatio, a compaction picking mode friendly to write amplification.
+* Deprecate Iterator::IsKeyPinned() and replace it with Iterator::GetProperty() with prop_name="rocksdb.iterator.is.key.pinned"
+
+## 4.5.0 (2/5/2016)
+### Public API Changes
+* Add a new perf context level between kEnableCount and kEnableTime. Level 2 now does not include timers for mutexes.
+* Statistics of mutex operation durations will not be measured by default. If you want to have them enabled, you need to set Statistics::stats_level_ to kAll.
+* DBOptions::delete_scheduler and NewDeleteScheduler() are removed, please use DBOptions::sst_file_manager and NewSstFileManager() instead
+
+### New Features
+* ldb tool now supports operations to non-default column families.
+* Add kPersistedTier to ReadTier.  This option allows Get and MultiGet to read only the persited data and skip mem-tables if writes were done with disableWAL = true.
+* Add DBOptions::sst_file_manager. Use NewSstFileManager() in include/rocksdb/sst_file_manager.h to create a SstFileManager that can be used to track the total size of SST files and control the SST files deletion rate.  
+
+## 4.4.0 (1/14/2016)
+### Public API Changes
+* Change names in CompactionPri and add a new one.
+* Deprecate options.soft_rate_limit and add options.soft_pending_compaction_bytes_limit.
+* If options.max_write_buffer_number > 3, writes will be slowed down when writing to the last write buffer to delay a full stop.
+* Introduce CompactionJobInfo::compaction_reason, this field include the reason to trigger the compaction.
+* After slow down is triggered, if estimated pending compaction bytes keep increasing, slowdown more.
+* Increase default options.delayed_write_rate to 2MB/s.
+* Added a new parameter --path to ldb tool. --path accepts the name of either MANIFEST, SST or a WAL file. Either --db or --path can be used when calling ldb.
+
+## 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.
@@ -198,7 +245,7 @@
 * Support Multiple DB paths in universal style compactions
 * Add feature of storing plain table index and bloom filter in SST file.
 * CompactRange() will never output compacted files to level 0. This used to be the case when all the compaction input files were at level 0.
-* Added iterate_upper_bound to define the extent upto which the forward iterator will return entries. This will prevent iterating over delete markers and overwritten entries for edge cases where you want to break out the iterator anyways. This may improve perfomance in case there are a large number of delete markers or overwritten entries.
+* Added iterate_upper_bound to define the extent upto which the forward iterator will return entries. This will prevent iterating over delete markers and overwritten entries for edge cases where you want to break out the iterator anyways. This may improve performance in case there are a large number of delete markers or overwritten entries.

 ### Public API changes
 * DBOptions.db_paths now is a vector of a DBPath structure which indicates both of path and target size
@@ -21,7 +21,7 @@ depend on gflags. You will need to have gflags installed to run `make all`. This
 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
+(-march=native or the equivalent). If you want to build a portable binary, add 'PORTABLE=1' before
 your make commands, like this: `PORTABLE=1 make static_lib`

 ## Dependencies
@@ -0,0 +1,11 @@
+This is the list of all known third-party language bindings for RocksDB. If something is missing, please open a pull request to add it.
+
+* Java - https://github.com/facebook/rocksdb/tree/master/java
+* Python - http://pyrocksdb.readthedocs.org/en/latest/
+* Perl - https://metacpan.org/pod/RocksDB
+* Node.js - https://npmjs.org/package/rocksdb
+* Go - https://github.com/tecbot/gorocksdb
+* Ruby - http://rubygems.org/gems/rocksdb-ruby
+* Haskell - https://hackage.haskell.org/package/rocksdb-haskell
+* PHP - https://github.com/Photonios/rocksdb-php
+* C# - https://github.com/warrenfalk/rocksdb-sharp
@@ -2,7 +2,7 @@ BSD License

 For rocksdb software

-Copyright (c) 2014, Facebook, Inc.
+Copyright (c) 2011-present, Facebook, Inc.
 All rights reserved.
 ---------------------------------------------------------------------

@@ -73,10 +73,19 @@ ifeq ($(MAKECMDGOALS),rocksdbjavastatic)
 	DEBUG_LEVEL=0
 endif

+ifeq ($(MAKECMDGOALS),rocksdbjavastaticrelease)
+	DEBUG_LEVEL=0
+endif
+
+ifeq ($(MAKECMDGOALS),rocksdbjavastaticpublish)
+	DEBUG_LEVEL=0
+endif
+
 # compile with -O2 if debug level is not 2
 ifneq ($(DEBUG_LEVEL), 2)
 OPT += -O2 -fno-omit-frame-pointer
-ifneq ($(MACHINE),ppc64) # ppc64 doesn't support -momit-leaf-frame-pointer
+# Skip for archs that don't support -momit-leaf-frame-pointer
+ifeq (,$(shell $(CXX) -fsyntax-only -momit-leaf-frame-pointer -xc /dev/null 2>&1))
 OPT += -momit-leaf-frame-pointer
 endif
 endif
@@ -135,6 +144,9 @@ else
 OPT += -DNDEBUG
 endif

+ifeq ($(PLATFORM), OS_SOLARIS)
+	PLATFORM_CXXFLAGS += -D _GLIBCXX_USE_C99
+endif
 ifneq ($(filter -DROCKSDB_LITE,$(OPT)),)
 	# found
 	CFLAGS += -fno-exceptions
@@ -162,6 +174,15 @@ else
 	pg = -pg
 endif

+# USAN doesn't work well with jemalloc. If we're compiling with USAN, we should use regular malloc.
+ifdef COMPILE_WITH_UBSAN
+	DISABLE_JEMALLOC=1
+	EXEC_LDFLAGS += -fsanitize=undefined
+	PLATFORM_CCFLAGS += -fsanitize=undefined
+	PLATFORM_CXXFLAGS += -fsanitize=undefined
+endif
+
+
 ifndef DISABLE_JEMALLOC
 	EXEC_LDFLAGS := $(JEMALLOC_LIB) $(EXEC_LDFLAGS)
 	PLATFORM_CXXFLAGS += $(JEMALLOC_INCLUDE)
@@ -179,10 +200,6 @@ default: all
 WARNING_FLAGS = -W -Wextra -Wall -Wsign-compare -Wshadow \
  -Wno-unused-parameter

-ifndef DISABLE_WARNING_AS_ERROR
-	WARNING_FLAGS += -Werror
-endif
-
 CFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CCFLAGS) $(OPT)
 CXXFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CXXFLAGS) $(OPT) -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers

@@ -229,8 +246,12 @@ VALGRIND_VER := $(join $(VALGRIND_VER),valgrind)

 VALGRIND_OPTS = --error-exitcode=$(VALGRIND_ERROR) --leak-check=full

+BENCHTOOLOBJECTS = $(BENCH_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)
+
 TESTS = \
 	db_test \
+	db_test2 \
+	db_block_cache_test \
 	db_iter_test \
 	db_log_iter_test \
 	db_compaction_filter_test \
@@ -240,6 +261,7 @@ TESTS = \
 	db_tailing_iter_test \
 	db_universal_compaction_test \
 	db_wal_test \
+	db_properties_test \
 	db_table_properties_test \
 	block_hash_index_test \
 	autovector_test \
@@ -266,6 +288,7 @@ TESTS = \
 	block_based_filter_block_test \
 	full_filter_block_test \
 	histogram_test \
+	inlineskiplist_test \
 	log_test \
 	manual_compaction_test \
 	memenv_test \
@@ -275,6 +298,7 @@ TESTS = \
 	memory_test \
 	merge_test \
 	merger_test \
+	options_file_test \
 	redis_test \
 	reduce_levels_test \
 	plain_table_db_test \
@@ -302,6 +326,8 @@ TESTS = \
 	rate_limiter_test \
 	delete_scheduler_test \
 	options_test \
+	options_settable_test \
+	options_util_test \
 	event_logger_test \
 	cuckoo_table_builder_test \
 	cuckoo_table_reader_test \
@@ -321,7 +347,8 @@ TESTS = \
 	compact_on_deletion_collector_test \
 	compaction_job_stats_test \
 	transaction_test \
-	ldb_cmd_test
+	ldb_cmd_test \
+	iostats_context_test

 SUBSET :=  $(shell echo $(TESTS) |sed s/^.*$(ROCKSDBTESTS_START)/$(ROCKSDBTESTS_START)/)

@@ -335,6 +362,7 @@ TOOLS = \
 	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

 # if user didn't config LIBNAME, set the default
@@ -390,7 +418,7 @@ $(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
@@ -448,7 +476,7 @@ test_names = \
      -e '/^(\s*)(\S+)/; !$$1 and do {$$p=$$2; break};'			\
      -e 'print qq! $$p$$2!'

-ifeq ($(MAKECMDGOALS),check)
+ifneq (,$(filter check parallel_check,$(MAKECMDGOALS)),)
 # Use /dev/shm if it has the sticky bit set (otherwise, /tmp),
 # and create a randomly-named rocksdb.XXXX directory therein.
 # We'll use that directory in the "make check" rules.
@@ -578,7 +606,7 @@ ldb_tests: ldb
 crash_test: whitebox_crash_test blackbox_crash_test

 blackbox_crash_test: db_stress
-	python -u tools/db_crashtest.py --simple blackbox 
+	python -u tools/db_crashtest.py --simple blackbox
 	python -u tools/db_crashtest.py blackbox

 whitebox_crash_test: db_stress
@@ -595,8 +623,18 @@ asan_crash_test:
 	COMPILE_WITH_ASAN=1 $(MAKE) crash_test
 	$(MAKE) clean

+ubsan_check:
+	$(MAKE) clean
+	COMPILE_WITH_UBSAN=1 $(MAKE) check -j32
+	$(MAKE) clean
+
+ubsan_crash_test:
+	$(MAKE) clean
+	COMPILE_WITH_UBSAN=1 $(MAKE) crash_test
+	$(MAKE) clean
+
 valgrind_check: $(TESTS)
-	for t in $(filter-out skiplist_test,$(TESTS)); do \
+	for t in $(filter-out %skiplist_test options_settable_test,$(TESTS)); do \
 		$(VALGRIND_VER) $(VALGRIND_OPTS) ./$$t; \
 		ret_code=$$?; \
 		if [ $$ret_code -ne 0 ]; then \
@@ -604,6 +642,46 @@ valgrind_check: $(TESTS)
 		fi; \
 	done

+
+ifneq ($(PAR_TEST),)
+parloop:
+	ret_bad=0;							\
+	for t in $(PAR_TEST); do		\
+		echo "===== Running $$t in parallel $(NUM_PAR)";\
+		if [ $(db_test) -eq 1 ]; then \
+			seq $(J) | v="$$t" parallel --gnu 's=$(TMPD)/rdb-{};  export TEST_TMPDIR=$$s;' \
+				'timeout 2m ./db_test --gtest_filter=$$v >> $$s/log-{} 2>1'; \
+		else\
+			seq $(J) | v="./$$t" parallel --gnu 's=$(TMPD)/rdb-{};' \
+			     'export TEST_TMPDIR=$$s; timeout 10m $$v >> $$s/log-{} 2>1'; \
+		fi; \
+		ret_code=$$?; \
+		if [ $$ret_code -ne 0 ]; then \
+			ret_bad=$$ret_code; \
+			echo $$t exited with $$ret_code; \
+		fi; \
+	done; \
+	exit $$ret_bad;
+endif
+
+parallel_check: $(TESTS)
+	$(AM_V_GEN)if test "$(J)" > 1                                  \
+	    && (parallel --gnu --help 2>/dev/null) |                    \
+	        grep -q 'GNU Parallel';                                 \
+	then                                                            \
+	    echo Running in parallel $(J);			\
+	else                                                            \
+	    echo "Need to have GNU Parallel and J > 1"; exit 1;		\
+	fi;								\
+	ret_bad=0;							\
+	echo $(J);\
+	echo Test Dir: $(TMPD); \
+        seq $(J) | parallel --gnu 's=$(TMPD)/rdb-{}; rm -rf $$s; mkdir $$s'; \
+	$(MAKE)  PAR_TEST="$(shell $(test_names))" TMPD=$(TMPD) \
+		J=$(J) db_test=1 parloop; \
+	$(MAKE) PAR_TEST="$(filter-out db_test, $(TESTS))" \
+		TMPD=$(TMPD) J=$(J) db_test=0 parloop;
+
 analyze: clean
 	$(CLANG_SCAN_BUILD) --use-analyzer=$(CLANG_ANALYZER) \
 		--use-c++=$(CXX) --use-cc=$(CC) --status-bugs \
@@ -640,7 +718,7 @@ clean:

 tags:
 	ctags * -R
-	cscope -b `find . -name '*.cc'` `find . -name '*.h'`
+	cscope -b `find . -name '*.cc'` `find . -name '*.h'` `find . -name '*.c'`

 format:
 	build_tools/format-diff.sh
@@ -655,7 +733,7 @@ $(LIBRARY): $(LIBOBJECTS)
 	$(AM_V_AR)rm -f $@
 	$(AM_V_at)$(AR) $(ARFLAGS) $@ $(LIBOBJECTS)

-db_bench: db/db_bench.o $(LIBOBJECTS) $(TESTUTIL)
+db_bench: tools/db_bench.o $(BENCHTOOLOBJECTS)
 	$(AM_LINK)

 cache_bench: util/cache_bench.o $(LIBOBJECTS) $(TESTUTIL)
@@ -730,6 +808,12 @@ slice_transform_test: util/slice_transform_test.o $(LIBOBJECTS) $(TESTHARNESS)
 db_test: db/db_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+db_test2: db/db_test2.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+db_block_cache_test: db/db_block_cache_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_log_iter_test: db/db_log_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -757,6 +841,9 @@ db_universal_compaction_test: db/db_universal_compaction_test.o db/db_test_util.
 db_wal_test: db/db_wal_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+db_properties_test: db/db_properties_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)

@@ -793,6 +880,9 @@ json_document_test: utilities/document/json_document_test.o $(LIBOBJECTS) $(TEST
 spatial_db_test: utilities/spatialdb/spatial_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+env_mirror_test: utilities/env_mirror_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 ttl_test: utilities/ttl/ttl_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -853,6 +943,9 @@ 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)

@@ -892,6 +985,9 @@ merge_test: db/merge_test.o $(LIBOBJECTS) $(TESTHARNESS)
 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)

@@ -913,7 +1009,7 @@ cuckoo_table_reader_test: table/cuckoo_table_reader_test.o $(LIBOBJECTS) $(TESTH
 cuckoo_table_db_test: db/cuckoo_table_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-listener_test: db/listener_test.o $(LIBOBJECTS) $(TESTHARNESS)
+listener_test: db/listener_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 thread_list_test: util/thread_list_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -925,6 +1021,12 @@ compact_files_test: db/compact_files_test.o $(LIBOBJECTS) $(TESTHARNESS)
 options_test: util/options_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+options_settable_test: util/options_settable_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)

@@ -946,7 +1048,7 @@ manual_compaction_test: db/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
 filelock_test: util/filelock_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-auto_roll_logger_test: util/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
+auto_roll_logger_test: db/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 memtable_list_test: db/memtable_list_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -964,12 +1066,18 @@ transaction_test: utilities/transactions/transaction_test.o $(LIBOBJECTS) $(TEST
 sst_dump: tools/sst_dump.o $(LIBOBJECTS)
 	$(AM_LINK)

+repair_test: db/repair_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 ldb_cmd_test: tools/ldb_cmd_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 ldb: tools/ldb.o $(LIBOBJECTS)
 	$(AM_LINK)

+iostats_context_test: util/iostats_context_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_V_CCLD)$(CXX) $^ $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+
 #-------------------------------------------------
 # make install related stuff
 INSTALL_PATH ?= /usr/local
@@ -1012,7 +1120,11 @@ install: install-static
 # ---------------------------------------------------------------------------

 JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/linux
-ARCH := $(shell getconf LONG_BIT)
+ifeq ($(PLATFORM), OS_SOLARIS)
+	ARCH := $(shell isainfo -b)
+else
+	ARCH := $(shell getconf LONG_BIT)
+endif
 ROCKSDBJNILIB = librocksdbjni-linux$(ARCH).so
 ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux$(ARCH).jar
 ROCKSDB_JAR_ALL = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar
@@ -1020,14 +1132,19 @@ ROCKSDB_JAVADOCS_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PA
 ROCKSDB_SOURCES_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-sources.jar

 ifeq ($(PLATFORM), OS_MACOSX)
-ROCKSDBJNILIB = librocksdbjni-osx.jnilib
-ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar
+	ROCKSDBJNILIB = librocksdbjni-osx.jnilib
+	ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar
 ifneq ("$(wildcard $(JAVA_HOME)/include/darwin)","")
 	JAVA_INCLUDE = -I$(JAVA_HOME)/include -I $(JAVA_HOME)/include/darwin
 else
 	JAVA_INCLUDE = -I/System/Library/Frameworks/JavaVM.framework/Headers/
 endif
 endif
+ifeq ($(PLATFORM), OS_SOLARIS)
+	ROCKSDBJNILIB = librocksdbjni-solaris$(ARCH).so
+	ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-solaris$(ARCH).jar
+	JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/solaris
+endif

 libz.a:
 	-rm -rf zlib-1.2.8
@@ -1121,11 +1238,10 @@ jtest: rocksdbjava
 jdb_bench:
 	cd java;$(MAKE) db_bench;

-commit-prereq:
-	$(MAKE) clean && $(MAKE) all check;
+commit_prereq: build_tools/rocksdb-lego-determinator \
+               build_tools/precommit_checker.py
+	J=$(J) build_tools/precommit_checker.py unit unit_481 clang_unit tsan asan ubsan lite
 	$(MAKE) clean && $(MAKE) jclean && $(MAKE) rocksdbjava;
-	$(MAKE) clean && USE_CLANG=1 $(MAKE) all;
-	$(MAKE) clean && OPT=-DROCKSDB_LITE $(MAKE) static_lib;

 xfunc:
 	for xftest in $(XFUNC_TESTS); do \
@@ -37,3 +37,22 @@ Check out our RocksDB Protobuf merge operator at: https://github.com/vladb38/roc

 ## Santanader UK/Cloudera Profession Services
 Check out their blog post: http://blog.cloudera.com/blog/2015/08/inside-santanders-near-real-time-data-ingest-architecture/
+
+## Airbnb
+Airbnb is using RocksDB as a storage engine for their personalized search service. You can learn more about it here: https://www.youtube.com/watch?v=ASQ6XMtogMs
+
+## Pinterest
+Pinterest's Object Retrieval System uses RocksDB for storage: https://www.youtube.com/watch?v=MtFEVEs_2Vo
+
+## Smyte
+[Smyte](https://www.smyte.com/) uses RocksDB as the storage layer for their core key-value storage, high-performance counters and time-windowed HyperLogLog services.
+
+## Rakuten Marketing
+[Rakuten Marketing](https://marketing.rakuten.com/) uses RocksDB as the disk cache layer for the real-time bidding service in their Performance DSP.
+
+## VWO, Wingify
+[VWO's](https://vwo.com/) Smart Code checker and URL helper uses RocksDB to store all the URLs where VWO's Smart Code is installed.
+
+## quasardb
+[quasardb](https://www.quasardb.net) is a high-performance, distributed, transactional key-value database that integrates well with in-memory analytics engines such as Apache Spark. 
+quasardb uses a heavily tuned RocksDB as its persistence layer.
@@ -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\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
+
@@ -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\rocksdb.sln
-  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
@@ -10,13 +10,184 @@ class FacebookArcanistConfiguration extends ArcanistConfiguration {
                                 ArcanistBaseWorkflow $workflow,
                                 $error_code) {
    if ($command == 'diff' && !$workflow->isRawDiffSource()) {
-      $this->maybePushToJenkins($workflow);
+      $this->startTestsInJenkins($workflow);
+      $this->startTestsInSandcastle($workflow);
    }
  }

+  //////////////////////////////////////////////////////////////////////
+  /*  Run tests in sandcastle */
+  function postURL($diffID, $url) {
+    $cmd = 'echo \'{"diff_id": "' . $diffID . '", '
+           . '"name":"click here for sandcastle tests for D' . $diffID . '", '
+           . '"link":"' . $url . '"}\' | '
+           . 'http_proxy=fwdproxy.any.facebook.com:8080 '
+           . 'https_proxy=fwdproxy.any.facebook.com:8080 arc call-conduit '
+           . 'differential.updateunitresults';
+    shell_exec($cmd);
+  }
+
+  function updateTestCommand($diffID, $test, $status) {
+    $cmd = 'echo \'{"diff_id": "' . $diffID . '", '
+           . '"name":"' . $test . '", '
+           . '"result":"' . $status . '"}\' | '
+           . 'http_proxy=fwdproxy.any.facebook.com:8080 '
+           . 'https_proxy=fwdproxy.any.facebook.com:8080 arc call-conduit '
+           . 'differential.updateunitresults';
+    return $cmd;
+  }
+
+  function updateTest($diffID, $test) {
+    shell_exec($this->updateTestCommand($diffID, $test, "waiting"));
+  }
+
+  function getSteps($diffID, $username, $test) {
+    $arcrc_content = exec("cat ~/.arcrc | gzip -f | base64 -w0");
+
+    // Sandcastle machines don't have arc setup. We copy the user certificate
+    // and authenticate using that in sandcastle
+    $setup = array(
+      "name" => "Setup arcrc",
+      "shell" => "echo " . $arcrc_content . " | base64 --decode"
+                 . " | gzip -d > ~/.arcrc",
+      "user" => "root"
+    );
+
+    // arc demands certain permission on its config
+    $fix_permission = array(
+      "name" => "Fix environment",
+      "shell" => "chmod 600 ~/.arcrc",
+      "user" => "root"
+    );
+
+    // fbcode is a sub-repo. We cannot patch until we add it to ignore otherwise
+    // git thinks it is uncommited change
+    $fix_git_ignore = array(
+      "name" => "Fix git ignore",
+      "shell" => "echo fbcode >> .git/info/exclude",
+      "user" => "root"
+    );
+
+    // Patch the code (keep your fingures crossed)
+    $patch = array(
+      "name" => "Patch " . $diffID,
+      "shell" => "HTTPS_PROXY=fwdproxy:8080 arc --arcrc-file ~/.arcrc "
+                  . "patch --diff " . $diffID,
+      "user" => "root"
+    );
+
+    // Clean up the user arc config we are using
+    $cleanup = array(
+      "name" => "Arc cleanup",
+      "shell" => "rm -f ~/.arcrc",
+      "user" => "root"
+    );
+
+    // Construct the steps in the order of execution
+    $steps[] = $setup;
+    $steps[] = $fix_permission;
+    $steps[] = $fix_git_ignore;
+    $steps[] = $patch;
+
+    // Run the actual command
+    $this->updateTest($diffID, $test);
+    $cmd = $this->updateTestCommand($diffID, $test, "running") . ";"
+           . "(./build_tools/precommit_checker.py " . $test
+           . "&& "
+           . $this->updateTestCommand($diffID, $test, "pass") . ")"
+           . "|| " . $this->updateTestCommand($diffID, $test, "fail")
+           . "; cat /tmp/precommit-check.log"
+           . "; for f in `ls t/log-*`; do echo \$f; cat \$f; done";
+
+    $run_test = array(
+      "name" => "Run " . $test,
+      "shell" => $cmd,
+      "user" => "root",
+    );
+
+    $steps[] = $run_test;
+    $steps[] = $cleanup;
+
+    return $steps;
+  }
+
+  function startTestsInSandcastle($workflow) {
+    // extract information we need from workflow or CLI
+    $diffID = $workflow->getDiffId();
+    $username = exec("whoami");
+
+    if ($diffID == null || $username == null) {
+      // there is no diff and we can't extract username
+      // we cannot schedule sandcasstle job
+      return;
+    }
+
+    if (strcmp(getenv("ROCKSDB_CHECK_ALL"), 1) == 0) {
+      // extract all tests from the CI definition
+      $output = file_get_contents("build_tools/rocksdb-lego-determinator");
+      preg_match_all('/[ ]{2}([a-zA-Z0-9_]+)[\)]{1}/', $output, $matches);
+      $tests = $matches[1];
+    } else {
+      // manually list of tests we want to run in sandcastle
+      $tests = array(
+        "unit", "unit_481", "clang_unit", "tsan", "asan", "lite", "valgrind"
+      );
+    }
+
+    // construct a job definition for each test and add it to the master plan
+    foreach ($tests as $test) {
+      $arg[] = array(
+        "name" => "RocksDB diff " . $diffID . " test " . $test,
+        "steps" => $this->getSteps($diffID, $username, $test)
+      );
+    }
+
+    // we cannot submit the parallel execution master plan to sandcastle
+    // we need supply the job plan as a determinator
+    // so we construct a small job that will spit out the master job plan
+    // which sandcastle will parse and execute
+    // Why compress ? Otherwise we run over the max string size.
+    $cmd = "echo " . base64_encode(json_encode($arg))
+           . " | gzip -f | base64 -w0";
+    $arg_encoded = shell_exec($cmd);
+
+    $command = array(
+      "name" => "Run diff " . $diffID . "for user " . $username,
+      "steps" => array()
+    );
+
+    $command["steps"][] = array(
+      "name" => "Generate determinator",
+      "shell" => "echo " . $arg_encoded . " | base64 --decode | gzip -d"
+                 . " | base64 --decode",
+      "determinator" => true,
+      "user" => "root"
+    );
+
+    // submit to sandcastle
+    $url = 'https://interngraph.intern.facebook.com/sandcastle/generate?'
+            .'command=SandcastleUniversalCommand'
+            .'&vcs=rocksdb-git&revision=origin%2Fmaster&type=lego'
+            .'&user=krad&alias=rocksdb-precommit'
+            .'&command-args=' . urlencode(json_encode($command));
+
+    $cmd = 'https_proxy= HTTPS_PROXY= curl -s -k -F app=659387027470559 '
+            . '-F token=AeO_3f2Ya3TujjnxGD4 "' . $url . '"';
+
+    $output = shell_exec($cmd);
+
+    // extract sandcastle URL from the response
+    preg_match('/url": "(.+)"/', $output, $sandcastle_url);
+
+    echo "\nSandcastle URL: " . $sandcastle_url[1] . "\n";
+
+    // Ask phabricator to display it on the diff UI
+    $this->postURL($diffID, $sandcastle_url[1]);
+  }
+
  //////////////////////////////////////////////////////////////////////
  /* Send off builds to jenkins */
-  function maybePushToJenkins($workflow) {
+  function startTestsInJenkins($workflow) {
    $diffID = $workflow->getDiffID();
    if ($diffID === null) {
      return;
@@ -31,5 +202,4 @@ class FacebookArcanistConfiguration extends ArcanistConfiguration {
               ."buildWithParameters?token=AUTH&DIFF_ID=$diffID";
    system("curl --noproxy '*' \"$url\" > /dev/null 2>&1");
  }
-
 }
@@ -1,5 +1,5 @@
 #!/usr/bin/python
-# Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+# Copyright (c) 2011-present, 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.
@@ -52,12 +52,7 @@ if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
    FBCODE_BUILD="true"
    # If we're compiling with TSAN we need pic build
    PIC_BUILD=$COMPILE_WITH_TSAN
-    if [ -z "$ROCKSDB_FBCODE_BUILD_WITH_481" ]; then
-      source "$PWD/build_tools/fbcode_config.sh"
-    else
-      # we need this to build with MySQL. Don't use for other purposes.
-      source "$PWD/build_tools/fbcode_config4.8.1.sh"
-    fi
+    source "$PWD/build_tools/fbcode_config.sh"
 fi

 # Delete existing output, if it exists
@@ -189,17 +184,19 @@ if [ "$CROSS_COMPILE" = "true" -o "$FBCODE_BUILD" = "true" ]; then
    # Also don't need any compilation tests if compiling on fbcode
    true
 else
-    # Test whether fallocate is available
-    $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
-      #include <fcntl.h>
-      #include <linux/falloc.h>
-      int main() {
-	int fd = open("/dev/null", 0);
-  fallocate(fd, FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE, 0, 1024);
-      }
+    if ! test $ROCKSDB_DISABLE_FALLOCATE; then
+        # Test whether fallocate is available
+        $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
+          #include <fcntl.h>
+          #include <linux/falloc.h>
+          int main() {
+      int fd = open("/dev/null", 0);
+      fallocate(fd, FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE, 0, 1024);
+          }
 EOF
-    if [ "$?" = 0 ]; then
-        COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_FALLOCATE_PRESENT"
+        if [ "$?" = 0 ]; then
+            COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_FALLOCATE_PRESENT"
+        fi
    fi

    # Test whether Snappy library is installed
@@ -359,7 +356,15 @@ if test "$USE_SSE"; then
  # if Intel SSE instruction set is supported, set USE_SSE=1
  COMMON_FLAGS="$COMMON_FLAGS -msse -msse4.2 "
 elif test -z "$PORTABLE"; then
-  COMMON_FLAGS="$COMMON_FLAGS -march=native "
+  if test -n "`echo $TARGET_ARCHITECTURE | grep ^ppc64`"; then
+    # Tune for this POWER processor, treating '+' models as base models
+    POWER=`LD_SHOW_AUXV=1 /bin/true | grep AT_PLATFORM | grep -E -o power[0-9]+`
+    COMMON_FLAGS="$COMMON_FLAGS -mcpu=$POWER -mtune=$POWER "
+  elif test -n "`echo $TARGET_ARCHITECTURE | grep ^s390x`"; then
+    COMMON_FLAGS="$COMMON_FLAGS -march=z10 "
+  else
+    COMMON_FLAGS="$COMMON_FLAGS -march=native "
+  fi
 fi

 PLATFORM_CCFLAGS="$PLATFORM_CCFLAGS $COMMON_FLAGS"
@@ -0,0 +1,19 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+GCC_BASE=/mnt/gvfs/third-party2/gcc/7331085db891a2ef4a88a48a751d834e8d68f4cb/7.x/centos7-native/b2ef2b6
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/963d9aeda70cc4779885b1277484fe7544a04e3e/9.0.0/platform007/9e92d53/
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/6ace84e956873d53638c738b6f65f3f469cca74c/7.x/platform007/5620abc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/192b0f42d63dcf6210d6ceae387b49af049e6e0c/2.26/platform007/f259413
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/7f9bdaada18f59bc27ec2b0871eb8a6144343aef/1.1.3/platform007/ca4da3d
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/2d9f0b9a4274cc21f61272a9e89bdb859bce8f1f/1.2.8/platform007/ca4da3d
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/dc49a21c5fceec6456a7a28a94dcd16690af1337/1.0.6/platform007/ca4da3d
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/0f607f8fc442ea7d6b876931b1898bb573d5e5da/1.9.1/platform007/ca4da3d
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/ca22bc441a4eb709e9e0b1f9fec9750fed7b31c5/1.4.x/platform007/15a3614
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/0b9929d2588991c65a57168bf88aff2db87c5d48/2.2.0/platform007/ca4da3d
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/c26f08f47ac35fc31da2633b7da92d6b863246eb/master/platform007/c26c002
+NUMA_BASE=/mnt/gvfs/third-party2/numa/3f3fb57a5ccc5fd21c66416c0b83e0aa76a05376/2.0.11/platform007/ca4da3d
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/40c73d874898b386a71847f1b99115d93822d11f/1.4/platform007/6f3e0a9
+TBB_BASE=/mnt/gvfs/third-party2/tbb/4ce8e8dba77cdbd81b75d6f0c32fd7a1b76a11ec/2018_U5/platform007/ca4da3d
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/fb251ecd2f5ae16f8671f7014c246e52a748fe0b/fb/platform007/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/ab9f09bba370e7066cafd4eb59752db93f2e8312/2.29.1/platform007/15a3614
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d42d152a15636529b0861ec493927200ebebca8e/3.15.0/platform007/ca4da3d
+LUA_BASE=/mnt/gvfs/third-party2/lua/f0cd714433206d5139df61659eb7b28b1dea6683/5.3.4/platform007/5007832
@@ -0,0 +1,16 @@
+GCC_BASE=/mnt/gvfs/third-party2/gcc/ebc96bc2fb751b5a0300b8d91a95bdf24ac1d88b/4.8.1/centos6-native/cc6c9dc
+CLANG_BASE=/mnt/gvfs/third-party2/clang/fc904e50a9266b9d7b98cae1993afa0c5aae1440/3.7.1/centos6-native/9d9ecb9/
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/f97108c12512b3b0789ac4515d836bdb1eae1142/4.8.1/gcc-4.8.1-glibc-2.17/8aac7fc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/f5484f168c0e4d19823d41df052c5870c6e575a4/2.17/gcc-4.8.1-glibc-2.17/99df8fc
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/cbf6f1f209e5bd160bdc5d971744e039f36b1566/1.1.3/gcc-4.8.1-glibc-2.17/c3f970a
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/6d39cb54708049f527e713ad19f2aadb9d3667e8/1.2.8/gcc-4.8.1-glibc-2.17/c3f970a
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/2ddd45f0853bfc8bb1c27f0f447236a1a26c338a/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/6858fac689e0f92e584224d91bdb0e39f6c8320d/r131/gcc-4.8.1-glibc-2.17/c3f970a
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/cb6c4880fcb4fee471574ba6af63a3882155a16a/0.5.1/gcc-4.8.1-glibc-2.17/c3f970a
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/c7275a4ceae0aca0929e56964a31dafc53c1ee96/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/d9acac8a9a2c9378eb696e22ffa8bd0e97d9206b/master/gcc-4.8.1-glibc-2.17/8d31e51
+NUMA_BASE=/mnt/gvfs/third-party2/numa/ae54a5ed22cdabb1c6446dce4e8ffae5b4446d73/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/41bfa3759ce52c071f5fd547ec9ecd2522929f0a/trunk/gcc-4.8.1-glibc-2.17/675d945
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/1a48835975c66d30e47770ec419758ed3b9ba010/3.10.62-62_fbk17_03959_ge29cc63/gcc-4.8.1-glibc-2.17/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/71454c53fffcb716a0beb9a90047aff7fb5c984a/2.26/centos6-native/da39a3e
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/af85c56f424cd5edfc2c97588299b44ecdec96bb/3.8.1/gcc-4.8.1-glibc-2.17/c3f970a
@@ -6,128 +6,147 @@
 # Environment variables that change the behavior of this script:
 # PIC_BUILD -- if true, it will only take pic versions of libraries from fbcode. libraries that don't have pic variant will not be included

+
+BASEDIR=`dirname $BASH_SOURCE`
+source "$BASEDIR/dependencies.sh"
+
 CFLAGS=""

-# location of libgcc
-LIBGCC_BASE="/mnt/gvfs/third-party2/libgcc/0473c80518a10d6efcbe24c5eeca3fb4ec9b519c/4.9.x/gcc-4.9-glibc-2.20/e1a7e4e"
-LIBGCC_INCLUDE="$LIBGCC_BASE/include"
-LIBGCC_LIBS=" -L $LIBGCC_BASE/libs"
+# libgcc
+LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/7.3.0"
+LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"

-# location of glibc
-GLIBC_REV=7397bed99280af5d9543439cdb7d018af7542720
-GLIBC_INCLUDE="/mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.20/gcc-4.9-glibc-2.20/99df8fc/include"
-GLIBC_LIBS=" -L /mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.20/gcc-4.9-glibc-2.20/99df8fc/lib"
-
-SNAPPY_INCLUDE=" -I /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/include/"
+# glibc
+GLIBC_INCLUDE="$GLIBC_BASE/include"
+GLIBC_LIBS=" -L $GLIBC_BASE/lib"

+# snappy
+SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
 if test -z $PIC_BUILD; then
-  SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/lib/libsnappy.a"
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"
 else
-  SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/lib/libsnappy_pic.a"
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy_pic.a"
 fi
-
 CFLAGS+=" -DSNAPPY"

 if test -z $PIC_BUILD; then
  # location of zlib headers and libraries
-  ZLIB_INCLUDE=" -I /mnt/gvfs/third-party2/zlib/feb983d9667f4cf5e9da07ce75abc824764b67a1/1.2.8/gcc-4.9-glibc-2.20/4230243/include/"
-  ZLIB_LIBS=" /mnt/gvfs/third-party2/zlib/feb983d9667f4cf5e9da07ce75abc824764b67a1/1.2.8/gcc-4.9-glibc-2.20/4230243/lib/libz.a"
+  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
+  ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"
  CFLAGS+=" -DZLIB"

  # location of bzip headers and libraries
-  BZIP_INCLUDE=" -I /mnt/gvfs/third-party2/bzip2/af004cceebb2dfd173ca29933ea5915e727aad2f/1.0.6/gcc-4.9-glibc-2.20/4230243/include/"
-  BZIP_LIBS=" /mnt/gvfs/third-party2/bzip2/af004cceebb2dfd173ca29933ea5915e727aad2f/1.0.6/gcc-4.9-glibc-2.20/4230243/lib/libbz2.a"
+  BZIP_INCLUDE=" -I $BZIP2_BASE/include/"
+  BZIP_LIBS=" $BZIP2_BASE/lib/libbz2.a"
  CFLAGS+=" -DBZIP2"

-  LZ4_INCLUDE=" -I /mnt/gvfs/third-party2/lz4/79d2943e2dd7208a3e0b06cf95e9f85f05fe9e1b/r124/gcc-4.9-glibc-2.20/4230243/include/"
-  LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/79d2943e2dd7208a3e0b06cf95e9f85f05fe9e1b/r124/gcc-4.9-glibc-2.20/4230243/lib/liblz4.a"
+  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
+  LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
  CFLAGS+=" -DLZ4"
-
-  ZSTD_REV=8df2d01673ae6afcc8c8d16fec862b2d67ecc1e9
-  ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$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"
-  CFLAGS+=" -DZSTD"
 fi

+ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
+if test -z $PIC_BUILD; then
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd.a"
+else
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd_pic.a"
+fi
+CFLAGS+=" -DZSTD"
+
 # location of gflags headers and libraries
-GFLAGS_INCLUDE=" -I /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/include/"
+GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
 if test -z $PIC_BUILD; then
-  GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/lib/libgflags.a"
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"
 else
-  GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/lib/libgflags_pic.a"
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags_pic.a"
 fi
-CFLAGS+=" -DGFLAGS=google"
+CFLAGS+=" -DGFLAGS=gflags"

 # location of jemalloc
-JEMALLOC_INCLUDE=" -I /mnt/gvfs/third-party2/jemalloc/bcd68e5e419efa4e61b9486d6854564d6d75a0b5/3.6.0/gcc-4.9-glibc-2.20/2aafc78/include/"
-JEMALLOC_LIB=" /mnt/gvfs/third-party2/jemalloc/bcd68e5e419efa4e61b9486d6854564d6d75a0b5/3.6.0/gcc-4.9-glibc-2.20/2aafc78/lib/libjemalloc.a"
+JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
+JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc.a"

 if test -z $PIC_BUILD; then
  # location of numa
-  NUMA_INCLUDE=" -I /mnt/gvfs/third-party2/numa/bbefc39ecbf31d0ca184168eb613ef8d397790ee/2.0.8/gcc-4.9-glibc-2.20/4230243/include/"
-  NUMA_LIB=" /mnt/gvfs/third-party2/numa/bbefc39ecbf31d0ca184168eb613ef8d397790ee/2.0.8/gcc-4.9-glibc-2.20/4230243/lib/libnuma.a"
+  NUMA_INCLUDE=" -I $NUMA_BASE/include/"
+  NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"
  CFLAGS+=" -DNUMA"

  # location of libunwind
-  LIBUNWIND="/mnt/gvfs/third-party2/libunwind/1de3b75e0afedfe5585b231bbb340ec7a1542335/1.1/gcc-4.9-glibc-2.20/34235e8/lib/libunwind.a"
+  LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
 fi

+# location of TBB
+TBB_INCLUDE=" -isystem $TBB_BASE/include/"
+if test -z $PIC_BUILD; then
+  TBB_LIBS="$TBB_BASE/lib/libtbb.a"
+else
+  TBB_LIBS="$TBB_BASE/lib/libtbb_pic.a"
+fi
+CFLAGS+=" -DTBB"
+
 # use Intel SSE support for checksum calculations
 export USE_SSE=1

-BINUTILS="/mnt/gvfs/third-party2/binutils/0b6ad0c88ddd903333a48ae8bff134efac468e4a/2.25/centos6-native/da39a3e/bin"
+BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"

-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE"

-GCC_BASE="/mnt/gvfs/third-party2/gcc/1c67a0b88f64d4d9ced0382d141c76aaa7d62fba/4.9.x/centos6-native/1317bc4"
 STDLIBS="-L $GCC_BASE/lib64"

-CLANG_BASE="/mnt/gvfs/third-party2/clang/d81444dd214df3d2466734de45bb264a0486acc3/dev"
-CLANG_BIN="$CLANG_BASE/centos6-native/af4b1a0/bin"
+CLANG_BIN="$CLANG_BASE/bin"
+CLANG_LIB="$CLANG_BASE/lib"
+CLANG_SRC="$CLANG_BASE/../../src"
+
 CLANG_ANALYZER="$CLANG_BIN/clang++"
-CLANG_SCAN_BUILD="$CLANG_BASE/src/clang/tools/scan-build/scan-build"
+CLANG_SCAN_BUILD="$CLANG_SRC/llvm/tools/clang/tools/scan-build/bin/scan-build"

 if [ -z "$USE_CLANG" ]; then
  # gcc
  CC="$GCC_BASE/bin/gcc"
  CXX="$GCC_BASE/bin/g++"
-  
+
  CFLAGS+=" -B$BINUTILS/gold"
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  CFLAGS+=" -isystem $GLIBC_INCLUDE"
+  JEMALLOC=1
 else
-  # clang 
-  CLANG_INCLUDE="$CLANG_BASE/gcc-4.9-glibc-2.20/74c386f/lib/clang/dev/include/"
+  # clang
+  CLANG_INCLUDE="$CLANG_LIB/clang/stable/include"
  CC="$CLANG_BIN/clang"
  CXX="$CLANG_BIN/clang++"

-  KERNEL_HEADERS_INCLUDE="/mnt/gvfs/third-party2/kernel-headers/ffd14f660a43c4b92717986b1bba66722ef089d0/3.2.18_70_fbk11_00129_gc8882d0/gcc-4.9-glibc-2.20/da39a3e/include"
+  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"

  CFLAGS+=" -B$BINUTILS/gold -nostdinc -nostdlib"
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.9.x "
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.9.x/x86_64-facebook-linux "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x/x86_64-facebook-linux "
  CFLAGS+=" -isystem $GLIBC_INCLUDE"
  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
  CFLAGS+=" -isystem $CLANG_INCLUDE"
  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE/linux "
  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE "
+  CFLAGS+=" -Wno-expansion-to-defined "
  CXXFLAGS="-nostdinc++"
 fi

 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
 CXXFLAGS+=" $CFLAGS"

-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB"
-EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/gcc-4.9-glibc-2.20/lib/ld.so"
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
+EXEC_LDFLAGS+=" -B$BINUTILS/gold"
+EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
 EXEC_LDFLAGS+=" $LIBUNWIND"
-EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/gcc-4.9-glibc-2.20/lib"
+EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform007/lib"
+# required by libtbb
+EXEC_LDFLAGS+=" -ldl"

 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"

-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"

-VALGRIND_VER="/mnt/gvfs/third-party2/valgrind/6c45ef049cbf11c2df593addb712cd891049e737/3.10.0/gcc-4.9-glibc-2.20/4230243/bin/"
+VALGRIND_VER="$VALGRIND_BASE/bin/"

-export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD
+export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD 
@@ -4,80 +4,77 @@
 # fbcode settings.  It uses the latest g++ compiler and also
 # uses jemalloc

+BASEDIR=`dirname $BASH_SOURCE`
+source "$BASEDIR/dependencies_4.8.1.sh"
+
 # location of libgcc
-LIBGCC_BASE="/mnt/gvfs/third-party2/libgcc/7712e757d7355cb51292454ee0b7b46a467fdfed/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_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"
+GLIBC_INCLUDE="$GLIBC_BASE/include"
+GLIBC_LIBS=" -L $GLIBC_BASE/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_INCLUDE=" -I $SNAPPY_BASE/include"
+SNAPPY_LIBS=" $SNAPPY_BASE/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_INCLUDE=" -I $ZLIB_BASE/include"
+ZLIB_LIBS=" $ZLIB_BASE/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"
+BZIP2_INCLUDE=" -I $BZIP2_BASE/include/"
+BZIP2_LIBS=" $BZIP2_BASE/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_INCLUDE=" -I $LZ4_BASE/include"
+LZ4_LIBS=" $LZ4_BASE/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_INCLUDE=" -I $ZSTD_BASE/include"
+ZSTD_LIBS=" $ZSTD_BASE/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_INCLUDE=" -I $GFLAGS_BASE/include/"
+GFLAGS_LIBS=" $GFLAGS_BASE/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_INCLUDE=" -I $JEMALLOC_BASE/include"
+JEMALLOC_LIB="$JEMALLOC_BASE/lib/libjemalloc.a"

 # location of numa
-NUMA_REV=829d10dac0230f99cd7e1778869d2adf3da24b65
-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"
+NUMA_INCLUDE=" -I $NUMA_BASE/include/"
+NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"

 # location of libunwind
-LIBUNWIND_REV=2c060e64064559905d46fd194000d61592087bdc
-LIBUNWIND="/mnt/gvfs/third-party2/libunwind/$LIBUNWIND_REV/1.1/gcc-4.8.1-glibc-2.17/675d945/lib/libunwind.a"
+LIBUNWIND="$LIBUNWIND_BASE/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="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"

-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP2_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"

-GCC_BASE="/mnt/gvfs/third-party2/gcc/1ec615e23800f0815d474478ba476a0adc3fe788/4.8.1/centos6-native/cc6c9dc"
 STDLIBS="-L $GCC_BASE/lib64"

 if [ -z "$USE_CLANG" ]; then
  # gcc
  CC="$GCC_BASE/bin/gcc"
  CXX="$GCC_BASE/bin/g++"
-  
+
  CFLAGS="-B$BINUTILS/gold -m64 -mtune=generic"
  CFLAGS+=" -isystem $GLIBC_INCLUDE"
  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
 else
-  # clang 
-  CLANG_BASE="/mnt/gvfs/third-party2/clang/9ab68376f938992c4eb5946ca68f90c3185cffc8/3.4"
-  CLANG_INCLUDE="$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++"
+  # clang
+  CLANG_BIN="$CLANG_BASE/bin"
+  CLANG_LIB="$CLANG_BASE/lib"
+  CLANG_INCLUDE="$CLANG_LIB/clang/*/include"
+  CC="$CLANG_BIN/clang"
+  CXX="$CLANG_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="$KERNEL_HEADERS_BASE/include/"

  CFLAGS="-B$BINUTILS/gold -nostdinc -nostdlib"
  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.8.1 "
@@ -95,16 +92,15 @@ CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PR
 CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4 -DZSTD -DNUMA"
 CXXFLAGS+=" $CFLAGS"

-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB"
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP2_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB"
 EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/gcc-4.8.1-glibc-2.17/lib/ld.so"
 EXEC_LDFLAGS+=" $LIBUNWIND"
 EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/gcc-4.8.1-glibc-2.17/lib"

 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"

-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP2_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"

-VALGRIND_REV=b2a9f85e4b70cd03abc85a7f3027fbc4cef35bd0
-VALGRIND_VER="/mnt/gvfs/third-party2/valgrind/$VALGRIND_REV/3.8.1/gcc-4.8.1-glibc-2.17/c3f970a/bin/"
+VALGRIND_VER="$VALGRIND_BASE/bin/"

 export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE
@@ -1,5 +1,5 @@
 #!/bin/bash
-#  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+#  Copyright (c) 2011-present, 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.
@@ -10,7 +10,7 @@ then
  GIT="git"
 fi

-# Print out the colored progress info so that it can be brainlessly 
+# Print out the colored progress info so that it can be brainlessly
 # distinguished by users.
 function title() {
  echo -e "\033[1;32m$*\033[0m"
@@ -0,0 +1,198 @@
+#!/usr/local/fbcode/gcc-4.8.1-glibc-2.17-fb/bin/python2.7
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+from __future__ import unicode_literals
+import argparse
+import commands
+import subprocess
+import sys
+import re
+import os
+import time
+
+#
+# Simple logger
+#
+
+
+class Log:
+
+    LOG_FILE = "/tmp/precommit-check.log"
+
+    def __init__(self):
+        self.filename = Log.LOG_FILE
+        self.f = open(self.filename, 'w+', 0)
+
+    def caption(self, str):
+        line = "\n##### %s #####\n" % str
+        if self.f:
+            self.f.write("%s \n" % line)
+        else:
+            print(line)
+
+    def error(self, str):
+        data = "\n\n##### ERROR ##### %s" % str
+        if self.f:
+            self.f.write("%s \n" % data)
+        else:
+            print(data)
+
+    def log(self, str):
+        if self.f:
+            self.f.write("%s \n" % str)
+        else:
+            print(str)
+
+#
+# Shell Environment
+#
+
+
+class Env(object):
+
+    def __init__(self, tests):
+        self.tests = tests
+        self.log = Log()
+
+    def shell(self, cmd, path=os.getcwd()):
+        if path:
+            os.chdir(path)
+
+        self.log.log("==== shell session ===========================")
+        self.log.log("%s> %s" % (path, cmd))
+        status = subprocess.call("cd %s; %s" % (path, cmd), shell=True,
+                                 stdout=self.log.f, stderr=self.log.f)
+        self.log.log("status = %s" % status)
+        self.log.log("============================================== \n\n")
+        return status
+
+    def GetOutput(self, cmd, path=os.getcwd()):
+        if path:
+            os.chdir(path)
+
+        self.log.log("==== shell session ===========================")
+        self.log.log("%s> %s" % (path, cmd))
+        status, out = commands.getstatusoutput(cmd)
+        self.log.log("status = %s" % status)
+        self.log.log("out = %s" % out)
+        self.log.log("============================================== \n\n")
+        return status, out
+
+#
+# Pre-commit checker
+#
+
+
+class PreCommitChecker(Env):
+
+    def __init__(self, tests):
+        Env.__init__(self, tests)
+
+    #
+    #   Get commands for a given job from the determinator file
+    #
+    def get_commands(self, test):
+        status, out = self.GetOutput(
+            "build_tools/rocksdb-lego-determinator %s" % test, ".")
+        return status, out
+
+    #
+    # Run a specific CI job
+    #
+    def run_test(self, test):
+        self.log.caption("Running test %s locally" % test)
+
+        # get commands for the CI job determinator
+        status, cmds = self.get_commands(test)
+        if status != 0:
+            self.log.error("Error getting commands for test %s" % test)
+            return False
+
+        # Parse the JSON to extract the commands to run
+        cmds = re.findall("'shell':'([^\']*)'", cmds)
+
+        if len(cmds) == 0:
+            self.log.log("No commands found")
+            return False
+
+        # Run commands
+        for cmd in cmds:
+            # Replace J=<..> with the local environment variable
+            if "J" in os.environ:
+                cmd = cmd.replace("J=1", "J=%s" % os.environ["J"])
+                cmd = cmd.replace("make ", "make -j%s " % os.environ["J"])
+            # Run the command
+            status = self.shell(cmd, ".")
+            if status != 0:
+                self.log.error("Error running command %s for test %s"
+                               % (cmd, test))
+                return False
+
+        return True
+
+    #
+    # Run specified CI jobs
+    #
+    def run_tests(self):
+        if not self.tests:
+            self.log.error("Invalid args. Please provide tests")
+            return False
+
+        self.print_separator()
+        self.print_row("TEST", "RESULT")
+        self.print_separator()
+
+        for test in self.tests:
+            start_time = time.time()
+            self.print_test(test)
+            result = self.run_test(test)
+            elapsed_min = (time.time() - start_time) / 60
+            if not result:
+                self.log.error("Error running test %s" % test)
+                self.print_result("FAIL (%dm)" % elapsed_min)
+                return False
+            self.print_result("PASS (%dm)" % elapsed_min)
+
+        self.print_separator()
+        return True
+
+    #
+    # Print a line
+    #
+    def print_separator(self):
+        print("".ljust(60, "-"))
+
+    #
+    # Print two colums
+    #
+    def print_row(self, c0, c1):
+        print("%s%s" % (c0.ljust(40), c1.ljust(20)))
+
+    def print_test(self, test):
+        print(test.ljust(40), end="")
+        sys.stdout.flush()
+
+    def print_result(self, result):
+        print(result.ljust(20))
+
+#
+# Main
+#
+parser = argparse.ArgumentParser(description='RocksDB pre-commit checker.')
+
+# <test ....>
+parser.add_argument('test', nargs='+',
+                    help='CI test(s) to run. e.g: unit punit asan tsan ubsan')
+
+print("Please follow log %s" % Log.LOG_FILE)
+
+args = parser.parse_args()
+checker = PreCommitChecker(args.test)
+
+if not checker.run_tests():
+    print("Error running tests. Please check log file %s" % Log.LOG_FILE)
+    sys.exit(1)
+
+sys.exit(0)
@@ -243,7 +243,7 @@ make release
    --bloom_bits=10 \
    --num=$((NUM / 4)) \
    --reads=$((NUM / 4)) \
-    --writes_per_second=1000 \
+    --benchmark_write_rate_limit=$(( 110 * 1024 )) \
    --write_buffer_size=100000000 \
    --cache_size=6442450944 \
    --cache_numshardbits=6 \
@@ -329,7 +329,7 @@ common_in_mem_args="--db=/dev/shm/rocksdb \
    --use_existing_db=1 \
    --duration=600 \
    --threads=32 \
-    --writes_per_second=81920 > ${STAT_FILE}.readwhilewriting_in_ram
+    --benchmark_write_rate_limit=9502720 > ${STAT_FILE}.readwhilewriting_in_ram

 # Seekrandomwhilewriting
 ./db_bench \
@@ -342,7 +342,7 @@ common_in_mem_args="--db=/dev/shm/rocksdb \
    --use_tailing_iterator=1 \
    --duration=600 \
    --threads=32 \
-    --writes_per_second=81920 > ${STAT_FILE}.seekwhilewriting_in_ram
+    --benchmark_write_rate_limit=9502720 > ${STAT_FILE}.seekwhilewriting_in_ram

 # measure fillseq with bunch of column families
 ./db_bench \
@@ -67,8 +67,9 @@ ASAN="COMPILE_WITH_ASAN=1"
 CLANG="USE_CLANG=1"
 LITE="OPT=-DROCKSDB_LITE"
 TSAN="COMPILE_WITH_TSAN=1"
+UBSAN="COMPILE_WITH_UBSAN=1"
 DISABLE_JEMALLOC="DISABLE_JEMALLOC=1"
-PARSER="'parser':'egrep \'Failure|^#|Abort\''"
+PARSER="'parser':'egrep \'Failure|^#|Abort|Expected|Actual|GoogleTestFailure|^==\''"

 ARTIFACTS=" 'artifacts': [
    {
@@ -376,6 +377,48 @@ ASAN_CRASH_TEST_COMMANDS="[
    }
 ]"

+#
+# RocksDB test under undefined behavior sanitizer
+#
+UBSAN_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb Unit Test under UBSAN',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Test RocksDB debug under UBSAN',
+                'shell':'set -o pipefail && $SHM $UBSAN $DEBUG make J=1 ubsan_check',
+                'user':'root',
+                $PARSER
+            }
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB crash testing under udnefined behavior sanitizer
+#
+UBSAN_CRASH_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb crash test under UBSAN',
+        'oncall':'$ONCALL',
+        'timeout': 86400,
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build and run RocksDB debug ubsan_crash_test',
+                'timeout': 86400,
+                'shell':'$SHM $DEBUG make J=1 ubsan_crash_test',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
 #
 # RocksDB unit test under valgrind
 #
@@ -462,9 +505,9 @@ run_format_compatible()
  if [ -e "build_detect_platform" ]
  then
    sed "s/tcmalloc/nothingnothingnothing/g" build_detect_platform > $TEST_TMPDIR/temp_build_file
-    rm -rf build_detect_platform 
+    rm -rf build_detect_platform
    cp $TEST_TMPDIR/temp_build_file build_detect_platform
-    chmod +x build_detect_platform 
+    chmod +x build_detect_platform
  fi

  make ldb -j32
@@ -632,6 +675,12 @@ case $1 in
  asan_crash)
    echo $ASAN_CRASH_TEST_COMMANDS
    ;;
+  ubsan)
+    echo $UBSAN_TEST_COMMANDS
+    ;;
+  ubsan_crash)
+    echo $UBSAN_CRASH_TEST_COMMANDS
+    ;;
  valgrind)
    echo $VALGRIND_TEST_COMMANDS
    ;;
@@ -1,17 +1,19 @@
 # This script enables you running RocksDB tests by running
-# All the tests in paralell and utilizing all the cores
+# All the tests in parallel and utilizing all the cores
 # For db_test the script first lists and parses the tests
 # and then fires them up in parallel using async PS Job functionality
 # Run the script from the enlistment
 Param(
  [switch]$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|db_test2|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.
-  [int]$Concurrency = 62
+   # a value of 1 will run everything serially
+  [int]$Concurrency = 16
 )

 # Folders and commands must be fullpath to run assuming
@@ -44,15 +46,25 @@ if($WorkFolder -eq "") {
  $Env:TEST_TMPDIR = $WorkFolder
 }

+Write-Output "Root: $RootFolder, WorkFolder: $WorkFolder"
+
 # Use JEMALLOC executables
-if($EnableJE) {
-    $db_test = -Join ($BinariesFolder, "db_test_je.exe")
-} else {
-    $db_test = -Join ($BinariesFolder, "db_test.exe")
+if($Run -ceq "db_test" -or
+   $Run -ceq "db_test2" ) {
+
+   $file_name = $Run
+
+   if($EnableJE) {
+     $file_name += "_je"
+   }
+
+   $file_name += ".exe"
+
+   $db_test = -Join ($BinariesFolder, $file_name)
+
+   Write-Output "Binaries: $BinariesFolder db_test: $db_test"
 }

-Write-Output "Root: $RootFolder, WorkFolder: $WorkFolder"
-Write-Output "Binaries: $BinariesFolder exe: $db_test"

 #Exclusions that we do not want to run
 $ExcludeTests = New-Object System.Collections.Generic.HashSet[string]
@@ -60,7 +72,7 @@ $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 }
 }

@@ -112,48 +124,70 @@ 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) {

    $Exclusions = @("db_test*", "db_sanity_test*")
-    if($EnableJE) {
-        $p = -join ($BinariesFolder, "*_test_je.exe")
-    } else {
-        $p = -join ($BinariesFolder, "*_test.exe")
-    }
+
+    $p = -join ($BinariesFolder, $pattern)
+
+    Write-Host "Path: $p"

    dir -Path $p -Exclude $Exclusions | ForEach-Object {
-       $t = ($_.Name) -replace '.exe$', ''
-       if($ExcludeTests.Contains($t)) {
-            continue
-       }
-       $test_log = -join ($t, ".log")
-       $HashTable.Add($t, $test_log)
+       MakeAndAdd -token ($_.Name) -HashTable $HashTable
    }
 }

-$TestToLog = [ordered]@{}
+$TestsToRun = [ordered]@{}

-if($Run -ceq "db_test") {
-    Normalize-DbTests -HashTable $TestToLog
+if($Run -ceq "db_test" -or
+   $Run -ceq "db_test2") {
+    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 {
-    Write-Warning "Invalid -Run option value"
-    exit 2
+
+    $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 = {
@@ -167,101 +201,124 @@ $InvokeTestAsync = {
    &$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
 $count = 0
-
+# Overall status
 [bool]$success = $true;

-# Wait for all to finish and get the results
-while(($JobToLog.Count -gt 0) -or
-      ($TestToLog.Count -gt 0)) {
-
-    # Make sure we have maximum concurrent jobs running if anything
-    # and the $Limit either not set or allows to proceed
-    while(($JobToLog.Count -lt $Concurrency) -and
-          (($TestToLog.Count -gt 0) -and
-          (($Limit -lt 0) -or ($count -lt $Limit)))) {
-
-
-        # We only need the first key
-        foreach($key in $TestToLog.keys) {
-            $k = $key
-            break
-        }
-
-        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)
-        $TestToLog.Remove($k)
-
-        ++$count
-    }
-
-    if($JobToLog.Count -lt 1) {
-      break
-    }
-
+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 = @{}

-    $completed = Wait-Job -Job $jobs -Any
-    $log = $JobToLog[$completed]
-    $JobToLog.Remove($completed)
+    # Wait for all to finish and get the results
+    while(($JobToLog.Count -gt 0) -or
+          ($TestToLog.Count -gt 0)) {

-    $message = -join @($completed.Name, " State: ", ($completed.State))
+        # 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)))) {

-    $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" -or
+               $Run -ceq "db_test2") {
+              $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 $message
-        }
-    }
+            # Scan the log. If we find PASSED and no occurrence of FAILED
+            # then it is a success
+            [bool]$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(!$success) {
 # This does not succeed killing off jobs quick
 # So we simply exit
@@ -270,4 +327,4 @@ if(!$success) {
    exit 12345
 }

- 
+ 
@@ -0,0 +1,127 @@
+#!/bin/sh
+#
+# Update dependencies.sh file with the latest avaliable versions
+
+BASEDIR=$(dirname $0)
+OUTPUT=""
+
+function log_variable()
+{
+  echo "$1=${!1}" >> "$OUTPUT"
+}
+
+
+TP2_LATEST="/mnt/vol/engshare/fbcode/third-party2"
+## $1 => lib name
+## $2 => lib version (if not provided, will try to pick latest)
+## $3 => platform (if not provided, will try to pick latest gcc)
+##
+## get_lib_base will set a variable named ${LIB_NAME}_BASE to the lib location
+function get_lib_base()
+{
+  local lib_name=$1
+  local lib_version=$2
+  local lib_platform=$3
+
+  local result="$TP2_LATEST/$lib_name/"
+  
+  # Lib Version
+  if [ -z "$lib_version" ] || [ "$lib_version" = "LATEST" ]; then
+    # version is not provided, use latest
+    result=`ls -dr1v $result/*/ | head -n1`
+  else
+    result="$result/$lib_version/"
+  fi
+  
+  # Lib Platform
+  if [ -z "$lib_platform" ]; then
+    # platform is not provided, use latest gcc
+    result=`ls -dr1v $result/gcc-*[^fb]/ | head -n1`
+  else
+    result="$result/$lib_platform/"
+  fi
+  
+  result=`ls -1d $result/*/ | head -n1`
+  
+  # lib_name => LIB_NAME_BASE
+  local __res_var=${lib_name^^}"_BASE"
+  __res_var=`echo $__res_var | tr - _`
+  # LIB_NAME_BASE=$result
+  eval $__res_var=`readlink -f $result`
+  
+  log_variable $__res_var
+}
+
+###########################################################
+#                   4.9.x dependencies                    #
+###########################################################
+
+OUTPUT="$BASEDIR/dependencies.sh"
+
+rm -f "$OUTPUT"
+touch "$OUTPUT"
+
+echo "Writing dependencies to $OUTPUT"
+
+# Compilers locations
+GCC_BASE=`readlink -f $TP2_LATEST/gcc/4.9.x/centos6-native/*/`
+CLANG_BASE=`ls -d1 /mnt/gvfs/third-party2/clang/fc904e50a9266b9d7b98cae1993afa0c5aae1440/3.7.1/centos6-native/*/ | head -n1`
+
+log_variable GCC_BASE
+log_variable CLANG_BASE
+
+# Libraries locations
+get_lib_base libgcc     4.9.x
+get_lib_base glibc      2.20 
+get_lib_base snappy     LATEST
+get_lib_base zlib       LATEST
+get_lib_base bzip2      LATEST
+get_lib_base lz4        LATEST
+get_lib_base zstd       LATEST
+get_lib_base gflags     LATEST
+get_lib_base jemalloc   LATEST
+get_lib_base numa       LATEST
+get_lib_base libunwind  LATEST
+
+get_lib_base kernel-headers LATEST 
+get_lib_base binutils   LATEST centos6-native 
+get_lib_base valgrind   LATEST
+
+git diff $OUTPUT
+
+###########################################################
+#                   4.8.1 dependencies                    #
+###########################################################
+
+OUTPUT="$BASEDIR/dependencies_4.8.1.sh"
+
+rm -f "$OUTPUT"
+touch "$OUTPUT"
+
+echo "Writing 4.8.1 dependencies to $OUTPUT"
+
+# Compilers locations
+GCC_BASE=`readlink -f $TP2_LATEST/gcc/4.8.1/centos6-native/*/`
+CLANG_BASE=`ls -d1 /mnt/gvfs/third-party2/clang/fc904e50a9266b9d7b98cae1993afa0c5aae1440/3.7.1/centos6-native/*/ | head -n1`
+
+log_variable GCC_BASE
+log_variable CLANG_BASE
+
+# Libraries locations
+get_lib_base libgcc     4.8.1  gcc-4.8.1-glibc-2.17
+get_lib_base glibc      2.17   gcc-4.8.1-glibc-2.17  
+get_lib_base snappy     LATEST gcc-4.8.1-glibc-2.17
+get_lib_base zlib       LATEST gcc-4.8.1-glibc-2.17
+get_lib_base bzip2      LATEST gcc-4.8.1-glibc-2.17
+get_lib_base lz4        LATEST gcc-4.8.1-glibc-2.17
+get_lib_base zstd       LATEST gcc-4.8.1-glibc-2.17
+get_lib_base gflags     LATEST gcc-4.8.1-glibc-2.17
+get_lib_base jemalloc   LATEST gcc-4.8.1-glibc-2.17
+get_lib_base numa       LATEST gcc-4.8.1-glibc-2.17
+get_lib_base libunwind  LATEST gcc-4.8.1-glibc-2.17
+
+get_lib_base kernel-headers LATEST gcc-4.8.1-glibc-2.17 
+get_lib_base binutils   LATEST centos6-native 
+get_lib_base valgrind   3.8.1  gcc-4.8.1-glibc-2.17
+
+git diff $OUTPUT
@@ -1,9 +1,9 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
 //
-#include "util/auto_roll_logger.h"
+#include "db/auto_roll_logger.h"
 #include "util/mutexlock.h"

 using namespace std;
@@ -12,7 +12,9 @@ namespace rocksdb {

 // -- AutoRollLogger
 Status AutoRollLogger::ResetLogger() {
+  TEST_SYNC_POINT("AutoRollLogger::ResetLogger:BeforeNewLogger");
  status_ = env_->NewLogger(log_fname_, &logger_);
+  TEST_SYNC_POINT("AutoRollLogger::ResetLogger:AfterNewLogger");

  if (!status_.ok()) {
    return status_;
@@ -32,8 +34,16 @@ Status AutoRollLogger::ResetLogger() {
 }

 void AutoRollLogger::RollLogFile() {
-  std::string old_fname = OldInfoLogFileName(
-      dbname_, env_->NowMicros(), db_absolute_path_, db_log_dir_);
+  // This function is called when log is rotating. Two rotations
+  // can happen quickly (NowMicro returns same value). To not overwrite
+  // previous log file we increment by one micro second and try again.
+  uint64_t now = env_->NowMicros();
+  std::string old_fname;
+  do {
+    old_fname = OldInfoLogFileName(
+      dbname_, now, db_absolute_path_, db_log_dir_);
+    now++;
+  } while (env_->FileExists(old_fname).ok());
  env_->RenameFile(log_fname_, old_fname);
 }

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -13,6 +13,8 @@
 #include "db/filename.h"
 #include "port/port.h"
 #include "port/util_logger.h"
+#include "util/sync_point.h"
+#include "util/mutexlock.h"

 namespace rocksdb {

@@ -53,11 +55,26 @@ class AutoRollLogger : public Logger {
    return status_;
  }

-  size_t GetLogFileSize() const override { return logger_->GetLogFileSize(); }
+  size_t GetLogFileSize() const override {
+    std::shared_ptr<Logger> logger;
+    {
+      MutexLock l(&mutex_);
+      // pin down the current logger_ instance before releasing the mutex.
+      logger = logger_;
+    }
+    return logger->GetLogFileSize();
+  }

  void Flush() override {
-    if (logger_) {
-      logger_->Flush();
+    std::shared_ptr<Logger> logger;
+    {
+      MutexLock l(&mutex_);
+      // pin down the current logger_ instance before releasing the mutex.
+      logger = logger_;
+    }
+    TEST_SYNC_POINT("AutoRollLogger::Flush:PinnedLogger");
+    if (logger) {
+      logger->Flush();
    }
  }

@@ -101,7 +118,7 @@ class AutoRollLogger : public Logger {
  uint64_t ctime_;
  uint64_t cached_now_access_count;
  uint64_t call_NowMicros_every_N_records_;
-  port::Mutex mutex_;
+  mutable port::Mutex mutex_;
 };

 // Facade to craete logger automatically
@@ -1,17 +1,20 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
 //
 #include <string>
+#include <thread>
 #include <vector>
 #include <cmath>
 #include <iostream>
 #include <fstream>
 #include <iterator>
 #include <algorithm>
+#include "db/auto_roll_logger.h"
+#include "port/port.h"
+#include "util/sync_point.h"
 #include "util/testharness.h"
-#include "util/auto_roll_logger.h"
 #include "rocksdb/db.h"
 #include <sys/stat.h>
 #include <errno.h>
@@ -260,7 +263,60 @@ TEST_F(AutoRollLoggerTest, CreateLoggerFromOptions) {
      auto_roll_logger, options.log_file_time_to_roll,
      kSampleMessage + ":CreateLoggerFromOptions - both");
 }
-#endif
+
+TEST_F(AutoRollLoggerTest, LogFlushWhileRolling) {
+  DBOptions options;
+  shared_ptr<Logger> logger;
+
+  InitTestDb();
+  options.max_log_file_size = 1024 * 5;
+  ASSERT_OK(CreateLoggerFromOptions(kTestDir, options, &logger));
+  AutoRollLogger* auto_roll_logger =
+      dynamic_cast<AutoRollLogger*>(logger.get());
+  ASSERT_TRUE(auto_roll_logger);
+  std::thread flush_thread;
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      // Need to pin the old logger before beginning the roll, as rolling grabs
+      // the mutex, which would prevent us from accessing the old logger.
+      {"AutoRollLogger::Flush:PinnedLogger",
+       "AutoRollLoggerTest::LogFlushWhileRolling:PreRollAndPostThreadInit"},
+      // Need to finish the flush thread init before this callback because the
+      // callback accesses flush_thread.get_id() in order to apply certain sync
+      // points only to the flush thread.
+      {"AutoRollLoggerTest::LogFlushWhileRolling:PreRollAndPostThreadInit",
+       "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallbackBegin"},
+      // Need to reset logger at this point in Flush() to exercise a race
+      // condition case, which is executing the flush with the pinned (old)
+      // logger after the roll has cut over to a new logger.
+      {"AutoRollLoggerTest::LogFlushWhileRolling:FlushCallback1",
+       "AutoRollLogger::ResetLogger:BeforeNewLogger"},
+      {"AutoRollLogger::ResetLogger:AfterNewLogger",
+       "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallback2"},
+  });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "PosixLogger::Flush:BeginCallback", [&](void* arg) {
+        TEST_SYNC_POINT(
+            "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallbackBegin");
+        if (std::this_thread::get_id() == flush_thread.get_id()) {
+          TEST_SYNC_POINT(
+              "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallback1");
+          TEST_SYNC_POINT(
+              "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallback2");
+        }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  flush_thread = std::thread([&]() { auto_roll_logger->Flush(); });
+  TEST_SYNC_POINT(
+      "AutoRollLoggerTest::LogFlushWhileRolling:PreRollAndPostThreadInit");
+  RollLogFileBySizeTest(auto_roll_logger, options.max_log_file_size,
+                        kSampleMessage + ":LogFlushWhileRolling");
+  flush_thread.join();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+#endif  // OS_WIN

 TEST_F(AutoRollLoggerTest, InfoLogLevel) {
  InitTestDb();
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -41,13 +41,17 @@ TableBuilder* NewTableBuilder(
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    uint32_t column_family_id, WritableFileWriter* file,
-    const CompressionType compression_type,
+    uint32_t column_family_id, const std::string& column_family_name,
+    WritableFileWriter* file, const CompressionType compression_type,
    const CompressionOptions& compression_opts, const bool skip_filters) {
+  assert((column_family_id ==
+          TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
+         column_family_name.empty());
  return ioptions.table_factory->NewTableBuilder(
      TableBuilderOptions(ioptions, internal_comparator,
                          int_tbl_prop_collector_factories, compression_type,
-                          compression_opts, skip_filters),
+                          compression_opts, skip_filters,
+                          column_family_name),
      column_family_id, file);
 }

@@ -58,11 +62,16 @@ Status BuildTable(
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    uint32_t column_family_id, std::vector<SequenceNumber> snapshots,
+    uint32_t column_family_id, const std::string& column_family_name,
+    std::vector<SequenceNumber> snapshots,
+    SequenceNumber earliest_write_conflict_snapshot,
    const CompressionType compression,
    const CompressionOptions& compression_opts, bool paranoid_file_checks,
    InternalStats* internal_stats, const Env::IOPriority io_priority,
-    TableProperties* table_properties) {
+    TableProperties* table_properties, int level) {
+  assert((column_family_id ==
+          TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
+         column_family_name.empty());
  // Reports the IOStats for flush for every following bytes.
  const size_t kReportFlushIOStatsEvery = 1048576;
  Status s;
@@ -86,7 +95,8 @@ Status BuildTable(

      builder = NewTableBuilder(
          ioptions, internal_comparator, int_tbl_prop_collector_factories,
-          column_family_id, file_writer.get(), compression, compression_opts);
+          column_family_id, column_family_name, file_writer.get(), compression,
+          compression_opts);
    }

    MergeHelper merge(env, internal_comparator.user_comparator(),
@@ -96,7 +106,8 @@ Status BuildTable(
                      snapshots.empty() ? 0 : snapshots.back());

    CompactionIterator c_iter(iter, internal_comparator.user_comparator(),
-                              &merge, kMaxSequenceNumber, &snapshots, env,
+                              &merge, kMaxSequenceNumber, &snapshots,
+                              earliest_write_conflict_snapshot, env,
                              true /* internal key corruption is not ok */);
    c_iter.SeekToFirst();
    for (; c_iter.Valid(); c_iter.Next()) {
@@ -147,7 +158,8 @@ Status BuildTable(
          ReadOptions(), env_options, internal_comparator, meta->fd, nullptr,
          (internal_stats == nullptr) ? nullptr
                                      : internal_stats->GetFileReadHist(0),
-          false));
+          false /* for_compaction */, nullptr /* arena */,
+          false /* skip_filter */, level));
      s = it->status();
      if (s.ok() && paranoid_file_checks) {
        for (it->SeekToFirst(); it->Valid(); it->Next()) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -33,13 +33,16 @@ class WritableFileWriter;
 class InternalStats;
 class InternalIterator;

+// @param column_family_name Name of the column family that is also identified
+//    by column_family_id, or empty string if unknown. It must outlive the
+//    TableBuilder returned by this function.
 TableBuilder* NewTableBuilder(
    const ImmutableCFOptions& options,
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    uint32_t column_family_id, WritableFileWriter* file,
-    const CompressionType compression_type,
+    uint32_t column_family_id, const std::string& column_family_name,
+    WritableFileWriter* file, const CompressionType compression_type,
    const CompressionOptions& compression_opts,
    const bool skip_filters = false);

@@ -48,6 +51,9 @@ TableBuilder* NewTableBuilder(
 // *meta will be filled with metadata about the generated table.
 // If no data is present in *iter, meta->file_size will be set to
 // zero, and no Table file will be produced.
+//
+// @param column_family_name Name of the column family that is also identified
+//    by column_family_id, or empty string if unknown.
 extern Status BuildTable(
    const std::string& dbname, Env* env, const ImmutableCFOptions& options,
    const EnvOptions& env_options, TableCache* table_cache,
@@ -55,11 +61,13 @@ extern Status BuildTable(
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    uint32_t column_family_id, std::vector<SequenceNumber> snapshots,
+    uint32_t column_family_id, const std::string& column_family_name,
+    std::vector<SequenceNumber> snapshots,
+    SequenceNumber earliest_write_conflict_snapshot,
    const CompressionType compression,
    const CompressionOptions& compression_opts, bool paranoid_file_checks,
    InternalStats* internal_stats,
    const Env::IOPriority io_priority = Env::IO_HIGH,
-    TableProperties* table_properties = nullptr);
+    TableProperties* table_properties = nullptr, int level = -1);

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -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_pin_l0_filter_and_index_blocks_in_cache(
+    rocksdb_block_based_table_options_t* options, unsigned char v) {
+  options->rep.pin_l0_filter_and_index_blocks_in_cache = 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;
@@ -1766,6 +1771,11 @@ void rocksdb_options_set_memtable_prefix_bloom_probes(
  opt->rep.memtable_prefix_bloom_probes = v;
 }

+void rocksdb_options_set_memtable_prefix_bloom_huge_page_tlb_size(
+    rocksdb_options_t* opt, size_t v) {
+  opt->rep.memtable_prefix_bloom_huge_page_tlb_size = v;
+}
+
 void rocksdb_options_set_hash_skip_list_rep(
    rocksdb_options_t *opt, size_t bucket_count,
    int32_t skiplist_height, int32_t skiplist_branching_factor) {
@@ -1965,7 +1975,7 @@ void rocksdb_filterpolicy_destroy(rocksdb_filterpolicy_t* filter) {
  delete filter;
 }

-rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom(int bits_per_key) {
+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom_format(int bits_per_key, bool original_format) {
  // Make a rocksdb_filterpolicy_t, but override all of its methods so
  // they delegate to a NewBloomFilterPolicy() instead of user
  // supplied C functions.
@@ -1983,13 +1993,21 @@ rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom(int bits_per_key) {
    static void DoNothing(void*) { }
  };
  Wrapper* wrapper = new Wrapper;
-  wrapper->rep_ = NewBloomFilterPolicy(bits_per_key);
+  wrapper->rep_ = NewBloomFilterPolicy(bits_per_key, original_format);
  wrapper->state_ = nullptr;
  wrapper->delete_filter_ = nullptr;
  wrapper->destructor_ = &Wrapper::DoNothing;
  return wrapper;
 }

+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom_full(int bits_per_key) {
+  return rocksdb_filterpolicy_create_bloom_format(bits_per_key, false);
+}
+
+rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom(int bits_per_key) {
+  return rocksdb_filterpolicy_create_bloom_format(bits_per_key, true);
+}
+
 rocksdb_mergeoperator_t* rocksdb_mergeoperator_create(
    void* state, void (*destructor)(void*),
    char* (*full_merge)(void*, const char* key, size_t key_length,
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -68,6 +68,20 @@ const std::string& ColumnFamilyHandleImpl::GetName() const {
  return cfd()->GetName();
 }

+Status ColumnFamilyHandleImpl::GetDescriptor(ColumnFamilyDescriptor* desc) {
+#ifndef ROCKSDB_LITE
+  // accessing mutable cf-options requires db mutex.
+  InstrumentedMutexLock l(mutex_);
+  *desc = ColumnFamilyDescriptor(
+      cfd()->GetName(),
+      BuildColumnFamilyOptions(*cfd()->options(),
+                               *cfd()->GetLatestMutableCFOptions()));
+  return Status::OK();
+#else
+  return Status::NotSupported();
+#endif  // !ROCKSDB_LITE
+}
+
 const Comparator* ColumnFamilyHandleImpl::user_comparator() const {
  return cfd()->user_comparator();
 }
@@ -110,18 +124,33 @@ Status CheckCompressionSupported(const ColumnFamilyOptions& cf_options) {
  return Status::OK();
 }

+Status CheckConcurrentWritesSupported(const ColumnFamilyOptions& cf_options) {
+  if (cf_options.inplace_update_support) {
+    return Status::InvalidArgument(
+        "In-place memtable updates (inplace_update_support) is not compatible "
+        "with concurrent writes (allow_concurrent_memtable_write)");
+  }
+  if (cf_options.filter_deletes) {
+    return Status::InvalidArgument(
+        "Delete filtering (filter_deletes) is not compatible with concurrent "
+        "memtable writes (allow_concurrent_memtable_writes)");
+  }
+  if (!cf_options.memtable_factory->IsInsertConcurrentlySupported()) {
+    return Status::InvalidArgument(
+        "Memtable doesn't concurrent writes (allow_concurrent_memtable_write)");
+  }
+  return Status::OK();
+}
+
 ColumnFamilyOptions SanitizeOptions(const DBOptions& db_options,
                                    const InternalKeyComparator* icmp,
                                    const ColumnFamilyOptions& src) {
  ColumnFamilyOptions result = src;
  result.comparator = icmp;
-#ifdef OS_MACOSX
-  // TODO(icanadi) make write_buffer_size uint64_t instead of size_t
-  ClipToRange(&result.write_buffer_size, ((size_t)64) << 10, ((size_t)1) << 30);
-#else
-  ClipToRange(&result.write_buffer_size,
-              ((size_t)64) << 10, ((size_t)64) << 30);
-#endif
+  size_t clamp_max = std::conditional<
+      sizeof(size_t) == 4, std::integral_constant<size_t, 0xffffffff>,
+      std::integral_constant<size_t, 64ull << 30>>::type::value;
+  ClipToRange(&result.write_buffer_size, ((size_t)64) << 10, clamp_max);
  // if user sets arena_block_size, we trust user to use this value. Otherwise,
  // calculate a proper value from writer_buffer_size;
  if (result.arena_block_size <= 0) {
@@ -175,6 +204,12 @@ 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 <
@@ -205,6 +240,17 @@ ColumnFamilyOptions SanitizeOptions(const DBOptions& db_options,
         result.level0_slowdown_writes_trigger,
         result.level0_file_num_compaction_trigger);
  }
+
+  if (result.soft_pending_compaction_bytes_limit == 0) {
+    result.soft_pending_compaction_bytes_limit =
+        result.hard_pending_compaction_bytes_limit;
+  } else if (result.hard_pending_compaction_bytes_limit > 0 &&
+             result.soft_pending_compaction_bytes_limit >
+                 result.hard_pending_compaction_bytes_limit) {
+    result.soft_pending_compaction_bytes_limit =
+        result.hard_pending_compaction_bytes_limit;
+  }
+
  if (result.level_compaction_dynamic_level_bytes) {
    if (result.compaction_style != kCompactionStyleLevel ||
        db_options.db_paths.size() > 1U) {
@@ -309,7 +355,8 @@ ColumnFamilyData::ColumnFamilyData(
      log_number_(0),
      column_family_set_(column_family_set),
      pending_flush_(false),
-      pending_compaction_(false) {
+      pending_compaction_(false),
+      prev_compaction_needed_bytes_(0) {
  Ref();

  // Convert user defined table properties collector factories to internal ones.
@@ -427,13 +474,81 @@ void ColumnFamilyData::SetDropped() {
  column_family_set_->RemoveColumnFamily(this);
 }

+const double kSlowdownRatio = 1.2;
+
+namespace {
+std::unique_ptr<WriteControllerToken> SetupDelay(
+    uint64_t max_write_rate, WriteController* write_controller,
+    uint64_t compaction_needed_bytes, uint64_t prev_compaction_neeed_bytes,
+    bool auto_comapctions_disabled) {
+  const uint64_t kMinWriteRate = 1024u;  // Minimum write rate 1KB/s.
+
+  uint64_t write_rate = write_controller->delayed_write_rate();
+
+  if (auto_comapctions_disabled) {
+    // When auto compaction is disabled, always use the value user gave.
+    write_rate = max_write_rate;
+  } else if (write_controller->NeedsDelay() && max_write_rate > kMinWriteRate) {
+    // If user gives rate less than kMinWriteRate, don't adjust it.
+    //
+    // If already delayed, need to adjust based on previous compaction debt.
+    // When there are two or more column families require delay, we always
+    // increase or reduce write rate based on information for one single
+    // column family. It is likely to be OK but we can improve if there is a
+    // problem.
+    // Ignore compaction_needed_bytes = 0 case because compaction_needed_bytes
+    // is only available in level-based compaction
+    //
+    // If the compaction debt stays the same as previously, we also further slow
+    // down. It usually means a mem table is full. It's mainly for the case
+    // where both of flush and compaction are much slower than the speed we
+    // insert to mem tables, so we need to actively slow down before we get
+    // feedback signal from compaction and flushes to avoid the full stop
+    // because of hitting the max write buffer number.
+    if (prev_compaction_neeed_bytes > 0 &&
+        prev_compaction_neeed_bytes <= compaction_needed_bytes) {
+      write_rate = static_cast<uint64_t>(static_cast<double>(write_rate) /
+                                         kSlowdownRatio);
+      if (write_rate < kMinWriteRate) {
+        write_rate = kMinWriteRate;
+      }
+    } else if (prev_compaction_neeed_bytes > compaction_needed_bytes) {
+      // We are speeding up by ratio of kSlowdownRatio when we have paid
+      // compaction debt. But we'll never speed up to faster than the write rate
+      // given by users.
+      write_rate = static_cast<uint64_t>(static_cast<double>(write_rate) *
+                                         kSlowdownRatio);
+      if (write_rate > max_write_rate) {
+        write_rate = max_write_rate;
+      }
+    }
+  }
+  return write_controller->GetDelayToken(write_rate);
+}
+
+int GetL0ThresholdSpeedupCompaction(int level0_file_num_compaction_trigger,
+                                    int level0_slowdown_writes_trigger) {
+  // SanitizeOptions() ensures it.
+  assert(level0_file_num_compaction_trigger <= level0_slowdown_writes_trigger);
+
+  // 1/4 of the way between L0 compaction trigger threshold and slowdown
+  // condition.
+  // Or twice as compaction trigger, if it is smaller.
+  return std::min(level0_file_num_compaction_trigger * 2,
+                  level0_file_num_compaction_trigger +
+                      (level0_slowdown_writes_trigger -
+                       level0_file_num_compaction_trigger) /
+                          4);
+}
+}  // namespace
+
 void ColumnFamilyData::RecalculateWriteStallConditions(
      const MutableCFOptions& mutable_cf_options) {
  if (current_ != nullptr) {
    auto* vstorage = current_->storage_info();
-    const double score = vstorage->max_compaction_score();
-    const int max_level = vstorage->max_compaction_score_level();
    auto write_controller = column_family_set_->write_controller_;
+    uint64_t compaction_needed_bytes =
+        vstorage->estimated_compaction_needed_bytes();

    if (imm()->NumNotFlushed() >= mutable_cf_options.max_write_buffer_number) {
      write_controller_token_ = write_controller->GetStopToken();
@@ -455,37 +570,88 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
          "[%s] Stopping writes because we have %d level-0 files",
          name_.c_str(), vstorage->l0_delay_trigger_count());
    } else if (mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
-               vstorage->estimated_compaction_needed_bytes() >=
+               compaction_needed_bytes >=
                   mutable_cf_options.hard_pending_compaction_bytes_limit) {
      write_controller_token_ = write_controller->GetStopToken();
      internal_stats_->AddCFStats(
          InternalStats::HARD_PENDING_COMPACTION_BYTES_LIMIT, 1);
      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
-          "[%s] Stopping writes because estimated pending compaction "
-          "bytes exceed %" PRIu64,
-          name_.c_str(), vstorage->estimated_compaction_needed_bytes());
+          "[%s] Stopping writes because of estimated pending compaction "
+          "bytes %" PRIu64,
+          name_.c_str(), compaction_needed_bytes);
+    } else if (mutable_cf_options.max_write_buffer_number > 3 &&
+               imm()->NumNotFlushed() >=
+                   mutable_cf_options.max_write_buffer_number - 1) {
+      write_controller_token_ =
+          SetupDelay(ioptions_.delayed_write_rate, write_controller,
+                     compaction_needed_bytes, prev_compaction_needed_bytes_,
+                     mutable_cf_options.disable_auto_compactions);
+      internal_stats_->AddCFStats(InternalStats::MEMTABLE_SLOWDOWN, 1);
+      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+          "[%s] Stalling writes because we have %d immutable memtables "
+          "(waiting for flush), max_write_buffer_number is set to %d "
+          "rate %" PRIu64,
+          name_.c_str(), imm()->NumNotFlushed(),
+          mutable_cf_options.max_write_buffer_number,
+          write_controller->delayed_write_rate());
    } else if (mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
               vstorage->l0_delay_trigger_count() >=
                   mutable_cf_options.level0_slowdown_writes_trigger) {
-      write_controller_token_ = write_controller->GetDelayToken();
+      write_controller_token_ =
+          SetupDelay(ioptions_.delayed_write_rate, write_controller,
+                     compaction_needed_bytes, prev_compaction_needed_bytes_,
+                     mutable_cf_options.disable_auto_compactions);
      internal_stats_->AddCFStats(InternalStats::LEVEL0_SLOWDOWN_TOTAL, 1);
      if (compaction_picker_->IsLevel0CompactionInProgress()) {
        internal_stats_->AddCFStats(
            InternalStats::LEVEL0_SLOWDOWN_WITH_COMPACTION, 1);
      }
      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
-          "[%s] Stalling writes because we have %d level-0 files",
-          name_.c_str(), vstorage->l0_delay_trigger_count());
-    } else if (mutable_cf_options.soft_rate_limit > 0.0 &&
-               score > mutable_cf_options.soft_rate_limit) {
-      write_controller_token_ = write_controller->GetDelayToken();
-      internal_stats_->RecordLevelNSlowdown(max_level, true);
+          "[%s] Stalling writes because we have %d level-0 files "
+          "rate %" PRIu64,
+          name_.c_str(), vstorage->l0_delay_trigger_count(),
+          write_controller->delayed_write_rate());
+    } else if (mutable_cf_options.soft_pending_compaction_bytes_limit > 0 &&
+               vstorage->estimated_compaction_needed_bytes() >=
+                   mutable_cf_options.soft_pending_compaction_bytes_limit) {
+      write_controller_token_ =
+          SetupDelay(ioptions_.delayed_write_rate, write_controller,
+                     compaction_needed_bytes, prev_compaction_needed_bytes_,
+                     mutable_cf_options.disable_auto_compactions);
+      internal_stats_->AddCFStats(
+          InternalStats::SOFT_PENDING_COMPACTION_BYTES_LIMIT, 1);
      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
-          "[%s] Stalling writes because we hit soft limit on level %d",
-          name_.c_str(), max_level);
+          "[%s] Stalling writes because of estimated pending compaction "
+          "bytes %" PRIu64 " rate %" PRIu64,
+          name_.c_str(), vstorage->estimated_compaction_needed_bytes(),
+          write_controller->delayed_write_rate());
+    } else if (vstorage->l0_delay_trigger_count() >=
+               GetL0ThresholdSpeedupCompaction(
+                   mutable_cf_options.level0_file_num_compaction_trigger,
+                   mutable_cf_options.level0_slowdown_writes_trigger)) {
+      write_controller_token_ = write_controller->GetCompactionPressureToken();
+      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+          "[%s] Increasing compaction threads because we have %d level-0 "
+          "files ",
+          name_.c_str(), vstorage->l0_delay_trigger_count());
+    } else if (vstorage->estimated_compaction_needed_bytes() >=
+               mutable_cf_options.soft_pending_compaction_bytes_limit / 4) {
+      // Increase compaction threads if bytes needed for compaction exceeds
+      // 1/4 of threshold for slowing down.
+      // If soft pending compaction byte limit is not set, always speed up
+      // compaction.
+      write_controller_token_ = write_controller->GetCompactionPressureToken();
+      if (mutable_cf_options.soft_pending_compaction_bytes_limit > 0) {
+        Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+            "[%s] Increasing compaction threads because of estimated pending "
+            "compaction "
+            "bytes %" PRIu64,
+            name_.c_str(), vstorage->estimated_compaction_needed_bytes());
+      }
    } else {
      write_controller_token_.reset();
    }
+    prev_compaction_needed_bytes_ = compaction_needed_bytes;
  }
 }

@@ -539,13 +705,12 @@ const int ColumnFamilyData::kCompactAllLevels = -1;
 const int ColumnFamilyData::kCompactToBaseLevel = -2;

 Compaction* ColumnFamilyData::CompactRange(
-    const MutableCFOptions& mutable_cf_options,
-    int input_level, int output_level, uint32_t output_path_id,
-    const InternalKey* begin, const InternalKey* end,
-    InternalKey** compaction_end) {
+    const MutableCFOptions& mutable_cf_options, int input_level,
+    int output_level, uint32_t output_path_id, const InternalKey* begin,
+    const InternalKey* end, InternalKey** compaction_end, bool* conflict) {
  auto* result = compaction_picker_->CompactRange(
      GetName(), mutable_cf_options, current_->storage_info(), input_level,
-      output_level, output_path_id, begin, end, compaction_end);
+      output_level, output_path_id, begin, end, compaction_end, conflict);
  if (result != nullptr) {
    result->SetInputVersion(current_);
  }
@@ -831,13 +996,6 @@ ColumnFamilyHandle* ColumnFamilyMemTablesImpl::GetColumnFamilyHandle() {
  return &handle_;
 }

-void ColumnFamilyMemTablesImpl::CheckMemtableFull() {
-  if (current_ != nullptr && current_->mem()->ShouldScheduleFlush()) {
-    flush_scheduler_->ScheduleFlush(current_);
-    current_->mem()->MarkFlushScheduled();
-  }
-}
-
 uint32_t GetColumnFamilyID(ColumnFamilyHandle* column_family) {
  uint32_t column_family_id = 0;
  if (column_family != nullptr) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -19,12 +19,10 @@
 #include "db/write_controller.h"
 #include "db/table_cache.h"
 #include "db/table_properties_collector.h"
-#include "db/flush_scheduler.h"
 #include "rocksdb/compaction_job_stats.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
-#include "util/instrumented_mutex.h"
 #include "util/mutable_cf_options.h"
 #include "util/thread_local.h"

@@ -44,6 +42,8 @@ class LogBuffer;
 class InstrumentedMutex;
 class InstrumentedMutexLock;

+extern const double kSlowdownRatio;
+
 // ColumnFamilyHandleImpl is the class that clients use to access different
 // column families. It has non-trivial destructor, which gets called when client
 // is done using the column family
@@ -59,6 +59,7 @@ class ColumnFamilyHandleImpl : public ColumnFamilyHandle {

  virtual uint32_t GetID() const override;
  virtual const std::string& GetName() const override;
+  virtual Status GetDescriptor(ColumnFamilyDescriptor* desc) override;

 private:
  ColumnFamilyData* cfd_;
@@ -132,6 +133,9 @@ struct SuperVersion {

 extern Status CheckCompressionSupported(const ColumnFamilyOptions& cf_options);

+extern Status CheckConcurrentWritesSupported(
+    const ColumnFamilyOptions& cf_options);
+
 extern ColumnFamilyOptions SanitizeOptions(const DBOptions& db_options,
                                           const InternalKeyComparator* icmp,
                                           const ColumnFamilyOptions& src);
@@ -156,14 +160,16 @@ class ColumnFamilyData {
  // thread-safe
  const std::string& GetName() const { return name_; }

-  // Ref() can only be called whily holding a DB mutex or during a
-  // single-threaded write.
+  // Ref() can only be called from a context where the caller can guarantee
+  // that ColumnFamilyData is alive (while holding a non-zero ref already,
+  // holding a DB mutex, or as the leader in a write batch group).
  void Ref() { refs_.fetch_add(1, std::memory_order_relaxed); }
-  // will just decrease reference count to 0, but will not delete it. returns
-  // true if the ref count was decreased to zero. in that case, it can be
-  // deleted by the caller immediately, or later, by calling
-  // FreeDeadColumnFamilies()
-  // Unref() can only be called while holding a DB mutex
+
+  // Unref decreases the reference count, but does not handle deletion
+  // when the count goes to 0.  If this method returns true then the
+  // caller should delete the instance immediately, or later, by calling
+  // FreeDeadColumnFamilies().  Unref() can only be called while holding
+  // a DB mutex, or during single-threaded recovery.
  bool Unref() {
    int old_refs = refs_.fetch_sub(1, std::memory_order_relaxed);
    assert(old_refs > 0);
@@ -224,7 +230,7 @@ class ColumnFamilyData {
  MemTable* mem() { return mem_; }
  Version* current() { return current_; }
  Version* dummy_versions() { return dummy_versions_; }
-  void SetCurrent(Version* current);
+  void SetCurrent(Version* _current);
  uint64_t GetNumLiveVersions() const;  // REQUIRE: DB mutex held
  uint64_t GetTotalSstFilesSize() const;  // REQUIRE: DB mutex held
  void SetMemtable(MemTable* new_mem) { mem_ = new_mem; }
@@ -249,11 +255,11 @@ class ColumnFamilyData {
  // A flag to tell a manual compaction's output is base level.
  static const int kCompactToBaseLevel;
  // REQUIRES: DB mutex held
-  Compaction* CompactRange(
-      const MutableCFOptions& mutable_cf_options,
-      int input_level, int output_level, uint32_t output_path_id,
-      const InternalKey* begin, const InternalKey* end,
-      InternalKey** compaction_end);
+  Compaction* CompactRange(const MutableCFOptions& mutable_cf_options,
+                           int input_level, int output_level,
+                           uint32_t output_path_id, const InternalKey* begin,
+                           const InternalKey* end, InternalKey** compaction_end,
+                           bool* manual_conflict);

  CompactionPicker* compaction_picker() { return compaction_picker_.get(); }
  // thread-safe
@@ -305,6 +311,14 @@ class ColumnFamilyData {
  bool pending_flush() { return pending_flush_; }
  bool pending_compaction() { return pending_compaction_; }

+  // Recalculate some small conditions, which are changed only during
+  // compaction, adding new memtable and/or
+  // recalculation of compaction score. These values are used in
+  // DBImpl::MakeRoomForWrite function to decide, if it need to make
+  // a write stall
+  void RecalculateWriteStallConditions(
+      const MutableCFOptions& mutable_cf_options);
+
 private:
  friend class ColumnFamilySet;
  ColumnFamilyData(uint32_t id, const std::string& name,
@@ -314,14 +328,6 @@ class ColumnFamilyData {
                   const DBOptions* db_options, const EnvOptions& env_options,
                   ColumnFamilySet* column_family_set);

-  // Recalculate some small conditions, which are changed only during
-  // compaction, adding new memtable and/or
-  // recalculation of compaction score. These values are used in
-  // DBImpl::MakeRoomForWrite function to decide, if it need to make
-  // a write stall
-  void RecalculateWriteStallConditions(
-      const MutableCFOptions& mutable_cf_options);
-
  uint32_t id_;
  const std::string name_;
  Version* dummy_versions_;  // Head of circular doubly-linked list of versions.
@@ -382,6 +388,8 @@ class ColumnFamilyData {
  // If true --> this ColumnFamily is currently present in
  // DBImpl::compaction_queue_
  bool pending_compaction_;
+
+  uint64_t prev_compaction_needed_bytes_;
 };

 // ColumnFamilySet has interesting thread-safety requirements
@@ -457,6 +465,8 @@ class ColumnFamilySet {
  // Don't call while iterating over ColumnFamilySet
  void FreeDeadColumnFamilies();

+  Cache* get_table_cache() { return table_cache_; }
+
 private:
  friend class ColumnFamilyData;
  // helper function that gets called from cfd destructor
@@ -493,15 +503,18 @@ class ColumnFamilySet {
 // memtables of different column families (specified by ID in the write batch)
 class ColumnFamilyMemTablesImpl : public ColumnFamilyMemTables {
 public:
-  explicit ColumnFamilyMemTablesImpl(ColumnFamilySet* column_family_set,
-                                     FlushScheduler* flush_scheduler)
-      : column_family_set_(column_family_set),
-        current_(nullptr),
-        flush_scheduler_(flush_scheduler) {}
+  explicit ColumnFamilyMemTablesImpl(ColumnFamilySet* column_family_set)
+      : column_family_set_(column_family_set), current_(nullptr) {}
+
+  // Constructs a ColumnFamilyMemTablesImpl equivalent to one constructed
+  // with the arguments used to construct *orig.
+  explicit ColumnFamilyMemTablesImpl(ColumnFamilyMemTablesImpl* orig)
+      : column_family_set_(orig->column_family_set_), current_(nullptr) {}

  // sets current_ to ColumnFamilyData with column_family_id
  // returns false if column family doesn't exist
-  // REQUIRES: under a DB mutex OR from a write thread
+  // REQUIRES: use this function of DBImpl::column_family_memtables_ should be
+  //           under a DB mutex OR from a write thread
  bool Seek(uint32_t column_family_id) override;

  // Returns log number of the selected column family
@@ -509,20 +522,23 @@ class ColumnFamilyMemTablesImpl : public ColumnFamilyMemTables {
  uint64_t GetLogNumber() const override;

  // REQUIRES: Seek() called first
-  // REQUIRES: under a DB mutex OR from a write thread
+  // REQUIRES: use this function of DBImpl::column_family_memtables_ should be
+  //           under a DB mutex OR from a write thread
  virtual MemTable* GetMemTable() const override;

  // Returns column family handle for the selected column family
-  // REQUIRES: under a DB mutex OR from a write thread
+  // REQUIRES: use this function of DBImpl::column_family_memtables_ should be
+  //           under a DB mutex OR from a write thread
  virtual ColumnFamilyHandle* GetColumnFamilyHandle() override;

-  // REQUIRES: under a DB mutex OR from a write thread
-  virtual void CheckMemtableFull() override;
+  // Cannot be called while another thread is calling Seek().
+  // REQUIRES: use this function of DBImpl::column_family_memtables_ should be
+  //           under a DB mutex OR from a write thread
+  virtual ColumnFamilyData* current() override { return current_; }

 private:
  ColumnFamilySet* column_family_set_;
  ColumnFamilyData* current_;
-  FlushScheduler* flush_scheduler_;
  ColumnFamilyHandleInternal handle_;
 };

@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -7,11 +7,13 @@

 #include <mutex>
 #include <string>
+#include <thread>
 #include <vector>

 #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 +55,63 @@ 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;
+    }
+  }
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  delete db;
+}
+
 TEST_F(CompactFilesTest, ObsoleteFiles) {
  Options options;
  // to trigger compaction more easily
@@ -84,9 +143,6 @@ TEST_F(CompactFilesTest, ObsoleteFiles) {
  }

  auto l0_files = collector->GetFlushedFiles();
-  CompactionOptions compact_opt;
-  compact_opt.compression = kNoCompression;
-  compact_opt.output_file_size_limit = kWriteBufferSize * 5;
  ASSERT_OK(db->CompactFiles(CompactionOptions(), l0_files, 1));

  // verify all compaction input files are deleted
@@ -96,6 +152,62 @@ TEST_F(CompactFilesTest, ObsoleteFiles) {
  delete db;
 }

+TEST_F(CompactFilesTest, CapturingPendingFiles) {
+  Options options;
+  options.create_if_missing = true;
+  // Disable RocksDB background compaction.
+  options.compaction_style = kCompactionStyleNone;
+  // Always do full scans for obsolete files (needed to reproduce the issue).
+  options.delete_obsolete_files_period_micros = 0;
+
+  // Add listener.
+  FlushedFileCollector* collector = new FlushedFileCollector();
+  options.listeners.emplace_back(collector);
+
+  DB* db = nullptr;
+  DestroyDB(db_name_, options);
+  Status s = DB::Open(options, db_name_, &db);
+  assert(s.ok());
+  assert(db);
+
+  // Create 5 files.
+  for (int i = 0; i < 5; ++i) {
+    db->Put(WriteOptions(), "key" + ToString(i), "value");
+    db->Flush(FlushOptions());
+  }
+
+  auto l0_files = collector->GetFlushedFiles();
+  EXPECT_EQ(5, l0_files.size());
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      {"CompactFilesImpl:2", "CompactFilesTest.CapturingPendingFiles:0"},
+      {"CompactFilesTest.CapturingPendingFiles:1", "CompactFilesImpl:3"},
+  });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // Start compacting files.
+  std::thread compaction_thread(
+      [&] { EXPECT_OK(db->CompactFiles(CompactionOptions(), l0_files, 1)); });
+
+  // In the meantime flush another file.
+  TEST_SYNC_POINT("CompactFilesTest.CapturingPendingFiles:0");
+  db->Put(WriteOptions(), "key5", "value");
+  db->Flush(FlushOptions());
+  TEST_SYNC_POINT("CompactFilesTest.CapturingPendingFiles:1");
+
+  compaction_thread.join();
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  delete db;
+
+  // Make sure we can reopen the DB.
+  s = DB::Open(options, db_name_, &db);
+  ASSERT_TRUE(s.ok());
+  assert(db);
+  delete db;
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -46,7 +46,7 @@ void Compaction::GetBoundaryKeys(
    Slice* largest_user_key) {
  bool initialized = false;
  const Comparator* ucmp = vstorage->InternalComparator()->user_comparator();
-  for (uint32_t i = 0; i < inputs.size(); ++i) {
+  for (size_t i = 0; i < inputs.size(); ++i) {
    if (inputs[i].files.empty()) {
      continue;
    }
@@ -128,8 +128,8 @@ bool Compaction::TEST_IsBottommostLevel(
 bool Compaction::IsFullCompaction(
    VersionStorageInfo* vstorage,
    const std::vector<CompactionInputFiles>& inputs) {
-  int num_files_in_compaction = 0;
-  int total_num_files = 0;
+  size_t num_files_in_compaction = 0;
+  size_t total_num_files = 0;
  for (int l = 0; l < vstorage->num_levels(); l++) {
    total_num_files += vstorage->NumLevelFiles(l);
  }
@@ -147,7 +147,8 @@ Compaction::Compaction(VersionStorageInfo* vstorage,
                       uint32_t _output_path_id, CompressionType _compression,
                       std::vector<FileMetaData*> _grandparents,
                       bool _manual_compaction, double _score,
-                       bool _deletion_compaction)
+                       bool _deletion_compaction,
+                       CompactionReason _compaction_reason)
    : start_level_(_inputs[0].level),
      output_level_(_output_level),
      max_output_file_size_(_target_file_size),
@@ -161,14 +162,15 @@ Compaction::Compaction(VersionStorageInfo* vstorage,
      deletion_compaction_(_deletion_compaction),
      inputs_(std::move(_inputs)),
      grandparents_(std::move(_grandparents)),
-      grandparent_index_(0),
-      seen_key_(false),
-      overlapped_bytes_(0),
      score_(_score),
      bottommost_level_(IsBottommostLevel(output_level_, vstorage, inputs_)),
      is_full_compaction_(IsFullCompaction(vstorage, inputs_)),
-      is_manual_compaction_(_manual_compaction) {
+      is_manual_compaction_(_manual_compaction),
+      compaction_reason_(_compaction_reason) {
  MarkFilesBeingCompacted(true);
+  if (is_manual_compaction_) {
+    compaction_reason_ = CompactionReason::kManualCompaction;
+  }

 #ifndef NDEBUG
  for (size_t i = 1; i < inputs_.size(); ++i) {
@@ -184,6 +186,9 @@ Compaction::Compaction(VersionStorageInfo* vstorage,
                                &arena_);
    }
  }
+
+  Slice smallest_user_key;
+  GetBoundaryKeys(vstorage, inputs_, &smallest_user_key, &largest_user_key_);
 }

 Compaction::~Compaction() {
@@ -282,36 +287,10 @@ bool Compaction::KeyNotExistsBeyondOutputLevel(
  return true;
 }

-bool Compaction::ShouldStopBefore(const Slice& internal_key) {
-  // Scan to find earliest grandparent file that contains key.
-  const InternalKeyComparator* icmp = &cfd_->internal_comparator();
-  while (grandparent_index_ < grandparents_.size() &&
-      icmp->Compare(internal_key,
-                    grandparents_[grandparent_index_]->largest.Encode()) > 0) {
-    if (seen_key_) {
-      overlapped_bytes_ += grandparents_[grandparent_index_]->fd.GetFileSize();
-    }
-    assert(grandparent_index_ + 1 >= grandparents_.size() ||
-           icmp->Compare(grandparents_[grandparent_index_]->largest.Encode(),
-                         grandparents_[grandparent_index_+1]->smallest.Encode())
-                         < 0);
-    grandparent_index_++;
-  }
-  seen_key_ = true;
-
-  if (overlapped_bytes_ > max_grandparent_overlap_bytes_) {
-    // Too much overlap for current output; start new output
-    overlapped_bytes_ = 0;
-    return true;
-  } else {
-    return false;
-  }
-}
-
 // Mark (or clear) each file that is being compacted
 void Compaction::MarkFilesBeingCompacted(bool mark_as_compacted) {
  for (size_t i = 0; i < num_input_levels(); i++) {
-    for (unsigned int j = 0; j < inputs_[i].size(); j++) {
+    for (size_t j = 0; j < inputs_[i].size(); j++) {
      assert(mark_as_compacted ? !inputs_[i][j]->being_compacted :
                                  inputs_[i][j]->being_compacted);
      inputs_[i][j]->being_compacted = mark_as_compacted;
@@ -371,7 +350,7 @@ int InputSummary(const std::vector<FileMetaData*>& files, char* output,
                 int len) {
  *output = '\0';
  int write = 0;
-  for (unsigned int i = 0; i < files.size(); i++) {
+  for (size_t i = 0; i < files.size(); i++) {
    int sz = len - write;
    int ret;
    char sztxt[16];
@@ -413,7 +392,7 @@ void Compaction::Summary(char* output, int len) {
  snprintf(output + write, len - write, "]");
 }

-uint64_t Compaction::OutputFilePreallocationSize() {
+uint64_t Compaction::OutputFilePreallocationSize() const {
  uint64_t preallocation_size = 0;

  if (cfd_->ioptions()->compaction_style == kCompactionStyleLevel ||
@@ -428,7 +407,7 @@ uint64_t Compaction::OutputFilePreallocationSize() {
  }
  // Over-estimate slightly so we don't end up just barely crossing
  // the threshold
-  return preallocation_size * 1.1;
+  return preallocation_size + (preallocation_size / 10);
 }

 std::unique_ptr<CompactionFilter> Compaction::CreateCompactionFilter() const {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -41,7 +41,8 @@ class Compaction {
             uint32_t output_path_id, CompressionType compression,
             std::vector<FileMetaData*> grandparents,
             bool manual_compaction = false, double score = -1,
-             bool deletion_compaction = false);
+             bool deletion_compaction = false,
+             CompactionReason compaction_reason = CompactionReason::kUnknown);

  // No copying allowed
  Compaction(const Compaction&) = delete;
@@ -103,7 +104,7 @@ class Compaction {
  }

  // Returns the LevelFilesBrief of the specified compaction input level.
-  LevelFilesBrief* input_levels(size_t compaction_input_level) {
+  const LevelFilesBrief* input_levels(size_t compaction_input_level) const {
    return &input_levels_[compaction_input_level];
  }

@@ -131,12 +132,10 @@ class Compaction {
  bool KeyNotExistsBeyondOutputLevel(const Slice& user_key,
                                     std::vector<size_t>* level_ptrs) const;

-  // Returns true iff we should stop building the current output
-  // before processing "internal_key".
-  bool ShouldStopBefore(const Slice& internal_key);
-
  // Clear all files to indicate that they are not being compacted
  // Delete this compaction from the list of running compactions.
+  //
+  // Requirement: DB mutex held
  void ReleaseCompactionFiles(Status status);

  // Returns the summary of the compaction in "output" with maximum "len"
@@ -148,13 +147,13 @@ class Compaction {
  double score() const { return score_; }

  // Is this compaction creating a file in the bottom most level?
-  bool bottommost_level() { return bottommost_level_; }
+  bool bottommost_level() const { return bottommost_level_; }

  // Does this compaction include all sst files?
-  bool is_full_compaction() { return is_full_compaction_; }
+  bool is_full_compaction() const { return is_full_compaction_; }

  // Was this compaction triggered manually by the client?
-  bool is_manual_compaction() { return is_manual_compaction_; }
+  bool is_manual_compaction() const { return is_manual_compaction_; }

  // Used when allow_trivial_move option is set in
  // Universal compaction. If all the input files are
@@ -167,19 +166,21 @@ class Compaction {
  // Used when allow_trivial_move option is set in
  // Universal compaction. Returns true, if the input files
  // are non-overlapping and can be trivially moved.
-  bool is_trivial_move() { return is_trivial_move_; }
+  bool is_trivial_move() const { return is_trivial_move_; }

  // How many total levels are there?
  int number_levels() const { return number_levels_; }

  // Return the MutableCFOptions that should be used throughout the compaction
  // procedure
-  const MutableCFOptions* mutable_cf_options() { return &mutable_cf_options_; }
+  const MutableCFOptions* mutable_cf_options() const {
+    return &mutable_cf_options_;
+  }

  // Returns the size in bytes that the output file should be preallocated to.
  // In level compaction, that is max_file_size_. In universal compaction, that
  // is the sum of all input file sizes.
-  uint64_t OutputFilePreallocationSize();
+  uint64_t OutputFilePreallocationSize() const;

  void SetInputVersion(Version* input_version);

@@ -218,6 +219,18 @@ class Compaction {
    output_table_properties_ = std::move(tp);
  }

+  Slice GetLargestUserKey() const { return largest_user_key_; }
+
+  CompactionReason compaction_reason() { return compaction_reason_; }
+
+  const std::vector<FileMetaData*>& grandparents() const {
+    return grandparents_;
+  }
+
+  uint64_t max_grandparent_overlap_bytes() const {
+    return max_grandparent_overlap_bytes_;
+  }
+
 private:
  // mark (or clear) all files that are being compacted
  void MarkFilesBeingCompacted(bool mark_as_compacted);
@@ -261,10 +274,6 @@ class Compaction {
  // State used to check for number of of overlapping grandparent files
  // (grandparent == "output_level_ + 1")
  std::vector<FileMetaData*> grandparents_;
-  size_t grandparent_index_;   // Index in grandparent_starts_
-  bool seen_key_;              // Some output key has been seen
-  uint64_t overlapped_bytes_;  // Bytes of overlap between current output
-                               // and grandparent files
  const double score_;         // score that was used to pick this compaction.

  // Is this compaction creating a file in the bottom most level?
@@ -284,6 +293,12 @@ class Compaction {

  // table properties of output files
  TablePropertiesCollection output_table_properties_;
+
+  // largest user keys in compaction
+  Slice largest_user_key_;
+
+  // Reason for compaction
+  CompactionReason compaction_reason_;
 };

 // Utility function
@@ -1,6 +1,6 @@
 // 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.
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -13,12 +13,14 @@ namespace rocksdb {
 CompactionIterator::CompactionIterator(
    InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
    SequenceNumber last_sequence, std::vector<SequenceNumber>* snapshots,
-    Env* env, bool expect_valid_internal_key, Compaction* compaction,
+    SequenceNumber earliest_write_conflict_snapshot, Env* env,
+    bool expect_valid_internal_key, const Compaction* compaction,
    const CompactionFilter* compaction_filter, LogBuffer* log_buffer)
    : input_(input),
      cmp_(cmp),
      merge_helper_(merge_helper),
      snapshots_(snapshots),
+      earliest_write_conflict_snapshot_(earliest_write_conflict_snapshot),
      env_(env),
      expect_valid_internal_key_(expect_valid_internal_key),
      compaction_(compaction),
@@ -42,6 +44,11 @@ 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() {
@@ -90,6 +97,11 @@ void CompactionIterator::Next() {
    NextFromInput();
  }

+  if (valid_) {
+    // Record that we've ouputted a record for the current key.
+    has_outputted_key_ = true;
+  }
+
  PrepareOutput();
 }

@@ -136,11 +148,14 @@ void CompactionIterator::NextFromInput() {
      key_ = current_key_.SetKey(key_, &ikey_);
      current_user_key_ = ikey_.user_key;
      has_current_user_key_ = true;
+      has_outputted_key_ = false;
      current_user_key_sequence_ = kMaxSequenceNumber;
      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,
@@ -170,6 +185,9 @@ 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();
@@ -187,31 +205,99 @@ void CompactionIterator::NextFromInput() {
        visible_at_tip_ ? visible_at_tip_ : findEarliestVisibleSnapshot(
                                                ikey_.sequence, &prev_snapshot);

-    if (ikey_.type == kTypeSingleDeletion) {
+    if (clear_and_output_next_key_) {
+      // In the previous iteration we encountered a single delete that we could
+      // not compact out.  We will keep this Put, but can drop it's data.
+      // (See Optimization 3, below.)
+      assert(ikey_.type == kTypeValue);
+      assert(current_user_key_snapshot_ == last_snapshot);
+
+      value_.clear();
+      valid_ = true;
+      clear_and_output_next_key_ = false;
+    } else if (ikey_.type == kTypeSingleDeletion) {
+      // We can compact out a SingleDelete if:
+      // 1) We encounter the corresponding PUT -OR- we know that this key
+      //    doesn't appear past this output level
+      // =AND=
+      // 2) We've already returned a record in this snapshot -OR-
+      //    there are no earlier earliest_write_conflict_snapshot.
+      //
+      // Rule 1 is needed for SingleDelete correctness.  Rule 2 is needed to
+      // allow Transactions to do write-conflict checking (if we compacted away
+      // all keys, then we wouldn't know that a write happened in this
+      // snapshot).  If there is no earlier snapshot, then we know that there
+      // are no active transactions that need to know about any writes.
+      //
+      // Optimization 3:
+      // If we encounter a SingleDelete followed by a PUT and Rule 2 is NOT
+      // true, then we must output a SingleDelete.  In this case, we will decide
+      // to also output the PUT.  While we are compacting less by outputting the
+      // PUT now, hopefully this will lead to better compaction in the future
+      // when Rule 2 is later true (Ie, We are hoping we can later compact out
+      // both the SingleDelete and the Put, while we couldn't if we only
+      // outputted the SingleDelete now).
+      // In this case, we can save space by removing the PUT's value as it will
+      // never be read.
+      //
+      // Deletes and Merges are not supported on the same key that has a
+      // SingleDelete as it is not possible to correctly do any partial
+      // compaction of such a combination of operations.  The result of mixing
+      // those operations for a given key is documented as being undefined.  So
+      // we can choose how to handle such a combinations of operations.  We will
+      // try to compact out as much as we can in these cases.
+
+      // The easiest way to process a SingleDelete during iteration is to peek
+      // ahead at the next key.
      ParsedInternalKey next_ikey;
      input_->Next();

-      // Check whether the current key is valid, not corrupt and the same
+      // Check whether the next key exists, is not corrupt, and is the same key
      // as the single delete.
      if (input_->Valid() && ParseInternalKey(input_->key(), &next_ikey) &&
          cmp_->Equal(ikey_.user_key, next_ikey.user_key)) {
-        // Mixing single deletes and merges is not supported. Consecutive
-        // single deletes are not valid.
-        if (next_ikey.type != kTypeValue) {
-          assert(false);
-          status_ =
-              Status::InvalidArgument("Put expected after single delete.");
-          break;
-        }
-
-        // Check whether the current key belongs to the same snapshot as the
-        // single delete.
+        // Check whether the next key belongs to the same snapshot as the
+        // SingleDelete.
        if (prev_snapshot == 0 || next_ikey.sequence > prev_snapshot) {
-          // Found the matching value, we can drop the single delete and the
-          // value.
-          ++iter_stats_.num_record_drop_hidden;
-          ++iter_stats_.num_record_drop_obsolete;
-          input_->Next();
+          if (next_ikey.type == kTypeSingleDeletion) {
+            // We encountered two SingleDeletes in a row.  This could be due to
+            // unexpected user input.
+            // Skip the first SingleDelete and let the next iteration decide how
+            // to handle the second SingleDelete
+
+            // First SingleDelete has been skipped since we already called
+            // input_->Next().
+            ++iter_stats_.num_record_drop_obsolete;
+          } else if ((ikey_.sequence <= earliest_write_conflict_snapshot_) ||
+                     has_outputted_key_) {
+            // Found a matching value, we can drop the single delete and the
+            // value.  It is safe to drop both records since we've already
+            // outputted a key in this snapshot, or there is no earlier
+            // snapshot (Rule 2 above).
+
+            // Note: it doesn't matter whether the second key is a Put or if it
+            // is an unexpected Merge or Delete.  We will compact it out
+            // either way.
+            ++iter_stats_.num_record_drop_hidden;
+            ++iter_stats_.num_record_drop_obsolete;
+            // Already called input_->Next() once.  Call it a second time to
+            // skip past the second key.
+            input_->Next();
+          } else {
+            // Found a matching value, but we cannot drop both keys since
+            // there is an earlier snapshot and we need to leave behind a record
+            // to know that a write happened in this snapshot (Rule 2 above).
+            // Clear the value and output the SingleDelete. (The value will be
+            // outputted on the next iteration.)
+            ++iter_stats_.num_record_drop_hidden;
+
+            // Setting valid_ to true will output the current SingleDelete
+            valid_ = true;
+
+            // Set up the Put to be outputted in the next iteration.
+            // (Optimization 3).
+            clear_and_output_next_key_ = true;
+          }
        } else {
          // We hit the next snapshot without hitting a put, so the iterator
          // returns the single delete.
@@ -226,11 +312,14 @@ void CompactionIterator::NextFromInput() {
        // iteration. If the next key is corrupt, we return before the
        // comparison, so the value of has_current_user_key does not matter.
        has_current_user_key_ = false;
-        if (compaction_ != nullptr &&
+        if (compaction_ != nullptr && ikey_.sequence <= earliest_snapshot_ &&
            compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
                                                       &level_ptrs_)) {
+          // Key doesn't exist outside of this range.
+          // Can compact out this SingleDelete.
          ++iter_stats_.num_record_drop_obsolete;
        } else {
+          // Output SingleDelete
          valid_ = true;
        }
      }
@@ -244,6 +333,10 @@ void CompactionIterator::NextFromInput() {
      // same key, then this kv is not visible in any snapshot.
      // Hidden by an newer entry for same user key
      // TODO: why not > ?
+      //
+      // Note: Dropping this key will not affect TransactionDB write-conflict
+      // checking since there has already been a record returned for this key
+      // in this snapshot.
      assert(last_sequence >= current_user_key_sequence_);
      ++iter_stats_.num_record_drop_hidden;  // (A)
      input_->Next();
@@ -262,6 +355,9 @@ void CompactionIterator::NextFromInput() {
      //     smaller sequence numbers will be dropped in the next
      //     few iterations of this loop (by rule (A) above).
      // Therefore this deletion marker is obsolete and can be dropped.
+      //
+      // Note:  Dropping this Delete will not affect TransactionDB
+      // write-conflict checking since it is earlier than any snapshot.
      ++iter_stats_.num_record_drop_obsolete;
      input_->Next();
    } else if (ikey_.type == kTypeMerge) {
@@ -310,9 +406,14 @@ void CompactionIterator::PrepareOutput() {
  // Zeroing out the sequence number leads to better compression.
  // If this is the bottommost level (no files in lower levels)
  // and the earliest snapshot is larger than this seqno
+  // and the userkey differs from the last userkey in compaction
  // then we can squash the seqno to zero.
+
+  // This is safe for TransactionDB write-conflict checking since transactions
+  // only care about sequence number larger than any active snapshots.
  if (bottommost_level_ && valid_ && ikey_.sequence < earliest_snapshot_ &&
-      ikey_.type != kTypeMerge) {
+      ikey_.type != kTypeMerge &&
+      !cmp_->Equal(compaction_->GetLargestUserKey(), ikey_.user_key)) {
    assert(ikey_.type != kTypeDeletion && ikey_.type != kTypeSingleDeletion);
    ikey_.sequence = 0;
    current_key_.UpdateInternalKey(0, ikey_.type);
@@ -1,6 +1,6 @@
 // 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.
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -39,9 +39,10 @@ class CompactionIterator {
 public:
  CompactionIterator(InternalIterator* input, const Comparator* cmp,
                     MergeHelper* merge_helper, SequenceNumber last_sequence,
-                     std::vector<SequenceNumber>* snapshots, Env* env,
+                     std::vector<SequenceNumber>* snapshots,
+                     SequenceNumber earliest_write_conflict_snapshot, Env* env,
                     bool expect_valid_internal_key,
-                     Compaction* compaction = nullptr,
+                     const Compaction* compaction = nullptr,
                     const CompactionFilter* compaction_filter = nullptr,
                     LogBuffer* log_buffer = nullptr);

@@ -88,9 +89,10 @@ class CompactionIterator {
  const Comparator* cmp_;
  MergeHelper* merge_helper_;
  const std::vector<SequenceNumber>* snapshots_;
+  const SequenceNumber earliest_write_conflict_snapshot_;
  Env* env_;
  bool expect_valid_internal_key_;
-  Compaction* compaction_;
+  const Compaction* compaction_;
  const CompactionFilter* compaction_filter_;
  LogBuffer* log_buffer_;
  bool bottommost_level_;
@@ -98,6 +100,7 @@ class CompactionIterator {
  SequenceNumber visible_at_tip_;
  SequenceNumber earliest_snapshot_;
  SequenceNumber latest_snapshot_;
+  bool ignore_snapshots_;

  // State
  //
@@ -124,6 +127,14 @@ class CompactionIterator {
  Slice current_user_key_;
  SequenceNumber current_user_key_sequence_;
  SequenceNumber current_user_key_snapshot_;
+
+  // True if the iterator has already returned a record for the current key.
+  bool has_outputted_key_ = false;
+
+  // truncated the value of the next key and output it without applying any
+  // compaction rules.  This is used for outputting a put after a single delete.
+  bool clear_and_output_next_key_ = false;
+
  MergeOutputIterator merge_out_iter_;
  std::string compaction_filter_value_;
  // "level_ptrs" holds indices that remember which file of an associated
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -20,9 +20,9 @@ class CompactionIteratorTest : public testing::Test {
                                        nullptr, 0U, false, 0));
    iter_.reset(new test::VectorIterator(ks, vs));
    iter_->SeekToFirst();
-    c_iter_.reset(new CompactionIterator(iter_.get(), cmp_, merge_helper_.get(),
-                                         last_sequence, &snapshots_,
-                                         Env::Default(), false));
+    c_iter_.reset(new CompactionIterator(
+        iter_.get(), cmp_, merge_helper_.get(), last_sequence, &snapshots_,
+        kMaxSequenceNumber, Env::Default(), false));
  }

  const Comparator* cmp_;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -51,6 +51,7 @@
 #include "util/iostats_context_imp.h"
 #include "util/log_buffer.h"
 #include "util/logging.h"
+#include "util/sst_file_manager_impl.h"
 #include "util/mutexlock.h"
 #include "util/perf_context_imp.h"
 #include "util/stop_watch.h"
@@ -62,7 +63,7 @@ namespace rocksdb {

 // Maintains state for each sub-compaction
 struct CompactionJob::SubcompactionState {
-  Compaction* compaction;
+  const Compaction* compaction;
  std::unique_ptr<CompactionIterator> c_iter;

  // The boundaries of the key-range this compaction is interested in. No two
@@ -103,6 +104,14 @@ struct CompactionJob::SubcompactionState {
  uint64_t num_output_records;
  CompactionJobStats compaction_job_stats;
  uint64_t approx_size;
+  // An index that used to speed up Compaction::ShouldStopBefore().
+  size_t grandparent_index = 0;
+  // The number of bytes overlapping between the current output and
+  // grandparent files used in Compaction::ShouldStopBefore().
+  uint64_t overlapped_bytes = 0;
+  // A flag determine whether the key has been seen in
+  // Compaction::ShouldStopBefore()
+  bool seen_key = false;

  SubcompactionState(Compaction* c, Slice* _start, Slice* _end,
                     uint64_t size = 0)
@@ -114,7 +123,10 @@ struct CompactionJob::SubcompactionState {
        total_bytes(0),
        num_input_records(0),
        num_output_records(0),
-        approx_size(size) {
+        approx_size(size),
+        grandparent_index(0),
+        overlapped_bytes(0),
+        seen_key(false) {
    assert(compaction != nullptr);
  }

@@ -133,6 +145,9 @@ struct CompactionJob::SubcompactionState {
    num_output_records = std::move(o.num_output_records);
    compaction_job_stats = std::move(o.compaction_job_stats);
    approx_size = std::move(o.approx_size);
+    grandparent_index = std::move(o.grandparent_index);
+    overlapped_bytes = std::move(o.overlapped_bytes);
+    seen_key = std::move(o.seen_key);
    return *this;
  }

@@ -140,6 +155,38 @@ struct CompactionJob::SubcompactionState {
  SubcompactionState(const SubcompactionState&) = delete;

  SubcompactionState& operator=(const SubcompactionState&) = delete;
+
+  // Returns true iff we should stop building the current output
+  // before processing "internal_key".
+  bool ShouldStopBefore(const Slice& internal_key) {
+    const InternalKeyComparator* icmp =
+        &compaction->column_family_data()->internal_comparator();
+    const std::vector<FileMetaData*>& grandparents = compaction->grandparents();
+
+    // Scan to find earliest grandparent file that contains key.
+    while (grandparent_index < grandparents.size() &&
+           icmp->Compare(internal_key,
+                         grandparents[grandparent_index]->largest.Encode()) >
+               0) {
+      if (seen_key) {
+        overlapped_bytes += grandparents[grandparent_index]->fd.GetFileSize();
+      }
+      assert(grandparent_index + 1 >= grandparents.size() ||
+             icmp->Compare(
+                 grandparents[grandparent_index]->largest.Encode(),
+                 grandparents[grandparent_index + 1]->smallest.Encode()) < 0);
+      grandparent_index++;
+    }
+    seen_key = true;
+
+    if (overlapped_bytes > compaction->max_grandparent_overlap_bytes()) {
+      // Too much overlap for current output; start new output
+      overlapped_bytes = 0;
+      return true;
+    }
+
+    return false;
+  }
 };

 // Maintains state for the entire compaction
@@ -211,7 +258,9 @@ CompactionJob::CompactionJob(
    const EnvOptions& env_options, VersionSet* versions,
    std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
    Directory* db_directory, Directory* output_directory, Statistics* stats,
+    InstrumentedMutex* db_mutex, Status* db_bg_error,
    std::vector<SequenceNumber> existing_snapshots,
+    SequenceNumber earliest_write_conflict_snapshot,
    std::shared_ptr<Cache> table_cache, EventLogger* event_logger,
    bool paranoid_file_checks, bool measure_io_stats, const std::string& dbname,
    CompactionJobStats* compaction_job_stats)
@@ -229,13 +278,18 @@ CompactionJob::CompactionJob(
      db_directory_(db_directory),
      output_directory_(output_directory),
      stats_(stats),
+      db_mutex_(db_mutex),
+      db_bg_error_(db_bg_error),
      existing_snapshots_(std::move(existing_snapshots)),
+      earliest_write_conflict_snapshot_(earliest_write_conflict_snapshot),
      table_cache_(std::move(table_cache)),
      event_logger_(event_logger),
      paranoid_file_checks_(paranoid_file_checks),
      measure_io_stats_(measure_io_stats) {
  assert(log_buffer_ != nullptr);
-  ThreadStatusUtil::SetColumnFamily(compact_->compaction->column_family_data());
+  const auto* cfd = compact_->compaction->column_family_data();
+  ThreadStatusUtil::SetColumnFamily(cfd, cfd->ioptions()->env,
+                                    cfd->options()->enable_thread_tracking);
  ThreadStatusUtil::SetThreadOperation(ThreadStatus::OP_COMPACTION);
  ReportStartedCompaction(compaction);
 }
@@ -247,8 +301,9 @@ CompactionJob::~CompactionJob() {

 void CompactionJob::ReportStartedCompaction(
    Compaction* compaction) {
-  ThreadStatusUtil::SetColumnFamily(
-      compact_->compaction->column_family_data());
+  const auto* cfd = compact_->compaction->column_family_data();
+  ThreadStatusUtil::SetColumnFamily(cfd, cfd->ioptions()->env,
+                                    cfd->options()->enable_thread_tracking);

  ThreadStatusUtil::SetThreadOperationProperty(
      ThreadStatus::COMPACTION_JOB_ID,
@@ -332,11 +387,6 @@ struct RangeWithSize {
      : range(a, b), size(s) {}
 };

-bool SliceCompare(const Comparator* cmp, const Slice& a, const Slice& b) {
-  // Returns true if a < b
-  return cmp->Compare(ExtractUserKey(a), ExtractUserKey(b)) < 0;
-}
-
 // Generates a histogram representing potential divisions of key ranges from
 // the input. It adds the starting and/or ending keys of certain input files
 // to the working set and then finds the approximate size of data in between
@@ -345,14 +395,13 @@ bool SliceCompare(const Comparator* cmp, const Slice& a, const Slice& b) {
 void CompactionJob::GenSubcompactionBoundaries() {
  auto* c = compact_->compaction;
  auto* cfd = c->column_family_data();
-  std::set<Slice, std::function<bool(const Slice& a, const Slice& b)> > bounds(
-      std::bind(&SliceCompare, cfd->user_comparator(), std::placeholders::_1,
-                std::placeholders::_2));
+  const Comparator* cfd_comparator = cfd->user_comparator();
+  std::vector<Slice> bounds;
  int start_lvl = c->start_level();
  int out_lvl = c->output_level();

  // Add the starting and/or ending key of certain input files as a potential
-  // boundary (because we're inserting into a set, it avoids duplicates)
+  // boundary
  for (size_t lvl_idx = 0; lvl_idx < c->num_input_levels(); lvl_idx++) {
    int lvl = c->level(lvl_idx);
    if (lvl >= start_lvl && lvl <= out_lvl) {
@@ -360,34 +409,44 @@ void CompactionJob::GenSubcompactionBoundaries() {
      size_t num_files = flevel->num_files;

      if (num_files == 0) {
-        break;
+        continue;
      }

      if (lvl == 0) {
        // For level 0 add the starting and ending key of each file since the
        // files may have greatly differing key ranges (not range-partitioned)
        for (size_t i = 0; i < num_files; i++) {
-          bounds.emplace(flevel->files[i].smallest_key);
-          bounds.emplace(flevel->files[i].largest_key);
+          bounds.emplace_back(flevel->files[i].smallest_key);
+          bounds.emplace_back(flevel->files[i].largest_key);
        }
      } else {
        // For all other levels add the smallest/largest key in the level to
        // encompass the range covered by that level
-        bounds.emplace(flevel->files[0].smallest_key);
-        bounds.emplace(flevel->files[num_files - 1].largest_key);
+        bounds.emplace_back(flevel->files[0].smallest_key);
+        bounds.emplace_back(flevel->files[num_files - 1].largest_key);
        if (lvl == out_lvl) {
          // For the last level include the starting keys of all files since
          // the last level is the largest and probably has the widest key
          // range. Since it's range partitioned, the ending key of one file
          // and the starting key of the next are very close (or identical).
          for (size_t i = 1; i < num_files; i++) {
-            bounds.emplace(flevel->files[i].smallest_key);
+            bounds.emplace_back(flevel->files[i].smallest_key);
          }
        }
      }
    }
  }

+  std::sort(bounds.begin(), bounds.end(),
+    [cfd_comparator] (const Slice& a, const Slice& b) -> bool {
+      return cfd_comparator->Compare(ExtractUserKey(a), ExtractUserKey(b)) < 0;
+    });
+  // Remove duplicated entries from bounds
+  bounds.erase(std::unique(bounds.begin(), bounds.end(),
+    [cfd_comparator] (const Slice& a, const Slice& b) -> bool {
+      return cfd_comparator->Compare(ExtractUserKey(a), ExtractUserKey(b)) == 0;
+    }), bounds.end());
+
  // Combine consecutive pairs of boundaries into ranges with an approximate
  // size of data covered by keys in that range
  uint64_t sum = 0;
@@ -409,9 +468,9 @@ void CompactionJob::GenSubcompactionBoundaries() {

  // Group the ranges into subcompactions
  const double min_file_fill_percent = 4.0 / 5;
-  uint64_t max_output_files = std::ceil(
+  uint64_t max_output_files = static_cast<uint64_t>(std::ceil(
      sum / min_file_fill_percent /
-      cfd->GetCurrentMutableCFOptions()->MaxFileSizeForLevel(out_lvl));
+      cfd->GetCurrentMutableCFOptions()->MaxFileSizeForLevel(out_lvl)));
  uint64_t subcompactions =
      std::min({static_cast<uint64_t>(ranges.size()),
                static_cast<uint64_t>(db_options_.max_subcompactions),
@@ -424,12 +483,12 @@ void CompactionJob::GenSubcompactionBoundaries() {
    // sizes becomes >= the expected mean size of a subcompaction
    sum = 0;
    for (size_t i = 0; i < ranges.size() - 1; i++) {
+      sum += ranges[i].size;
      if (subcompactions == 1) {
        // If there's only one left to schedule then it goes to the end so no
        // need to put an end boundary
-        break;
+        continue;
      }
-      sum += ranges[i].size;
      if (sum >= mean) {
        boundaries_.emplace_back(ExtractUserKey(ranges[i].range.limit));
        sizes_.emplace_back(sum);
@@ -509,18 +568,17 @@ Status CompactionJob::Run() {
  return status;
 }

-Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options,
-                              InstrumentedMutex* db_mutex) {
+Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options) {
  AutoThreadOperationStageUpdater stage_updater(
      ThreadStatus::STAGE_COMPACTION_INSTALL);
-  db_mutex->AssertHeld();
+  db_mutex_->AssertHeld();
  Status status = compact_->status;
  ColumnFamilyData* cfd = compact_->compaction->column_family_data();
  cfd->internal_stats()->AddCompactionStats(
      compact_->compaction->output_level(), compaction_stats_);

  if (status.ok()) {
-    status = InstallCompactionResults(mutable_cf_options, db_mutex);
+    status = InstallCompactionResults(mutable_cf_options);
  }
  VersionStorageInfo::LevelSummaryStorage tmp;
  auto vstorage = cfd->current()->storage_info();
@@ -638,8 +696,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  Status status;
  sub_compact->c_iter.reset(new CompactionIterator(
      input.get(), cfd->user_comparator(), &merge, versions_->LastSequence(),
-      &existing_snapshots_, env_, false, sub_compact->compaction,
-      compaction_filter));
+      &existing_snapshots_, earliest_write_conflict_snapshot_, env_, false,
+      sub_compact->compaction, compaction_filter));
  auto c_iter = sub_compact->c_iter.get();
  c_iter->SeekToFirst();
  const auto& c_iter_stats = c_iter->iter_stats();
@@ -657,7 +715,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
    if (end != nullptr &&
        cfd->user_comparator()->Compare(c_iter->user_key(), *end) >= 0) {
      break;
-    } else if (sub_compact->compaction->ShouldStopBefore(key) &&
+    } else if (sub_compact->ShouldStopBefore(key) &&
               sub_compact->builder != nullptr) {
      status = FinishCompactionOutputFile(input->status(), sub_compact);
      if (!status.ok()) {
@@ -846,13 +904,33 @@ Status CompactionJob::FinishCompactionOutputFile(
          event_logger_, cfd->ioptions()->listeners, meta->fd, info);
    }
  }
+
+  // Report new file to SstFileManagerImpl
+  auto sfm =
+      static_cast<SstFileManagerImpl*>(db_options_.sst_file_manager.get());
+  if (sfm && meta->fd.GetPathId() == 0) {
+    ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
+    auto fn = TableFileName(cfd->ioptions()->db_paths, meta->fd.GetNumber(),
+                            meta->fd.GetPathId());
+    sfm->OnAddFile(fn);
+    if (sfm->IsMaxAllowedSpaceReached()) {
+      InstrumentedMutexLock l(db_mutex_);
+      if (db_bg_error_->ok()) {
+        s = Status::IOError("Max allowed space was reached");
+        *db_bg_error_ = s;
+        TEST_SYNC_POINT(
+            "CompactionJob::FinishCompactionOutputFile:MaxAllowedSpaceReached");
+      }
+    }
+  }
+
  sub_compact->builder.reset();
  return s;
 }

 Status CompactionJob::InstallCompactionResults(
-    const MutableCFOptions& mutable_cf_options, InstrumentedMutex* db_mutex) {
-  db_mutex->AssertHeld();
+    const MutableCFOptions& mutable_cf_options) {
+  db_mutex_->AssertHeld();

  auto* compaction = compact_->compaction;
  // paranoia: verify that the files that we started with
@@ -887,7 +965,7 @@ Status CompactionJob::InstallCompactionResults(
  }
  return versions_->LogAndApply(compaction->column_family_data(),
                                mutable_cf_options, compaction->edit(),
-                                db_mutex, db_directory_);
+                                db_mutex_, db_directory_);
 }

 void CompactionJob::RecordCompactionIOStats() {
@@ -941,7 +1019,7 @@ 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(), cfd->GetID(),
+      cfd->int_tbl_prop_collector_factories(), cfd->GetID(), cfd->GetName(),
      sub_compact->outfile.get(), sub_compact->compaction->output_compression(),
      cfd->ioptions()->compression_opts, skip_filters));
  LogFlush(db_options_.info_log);
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -56,8 +56,10 @@ class CompactionJob {
                const EnvOptions& env_options, VersionSet* versions,
                std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
                Directory* db_directory, Directory* output_directory,
-                Statistics* stats,
+                Statistics* stats, InstrumentedMutex* db_mutex,
+                Status* db_bg_error,
                std::vector<SequenceNumber> existing_snapshots,
+                SequenceNumber earliest_write_conflict_snapshot,
                std::shared_ptr<Cache> table_cache, EventLogger* event_logger,
                bool paranoid_file_checks, bool measure_io_stats,
                const std::string& dbname,
@@ -76,8 +78,7 @@ class CompactionJob {
  Status Run();

  // REQUIRED: mutex held
-  Status Install(const MutableCFOptions& mutable_cf_options,
-                 InstrumentedMutex* db_mutex);
+  Status Install(const MutableCFOptions& mutable_cf_options);

 private:
  struct SubcompactionState;
@@ -94,8 +95,7 @@ class CompactionJob {

  Status FinishCompactionOutputFile(const Status& input_status,
                                    SubcompactionState* sub_compact);
-  Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options,
-                                  InstrumentedMutex* db_mutex);
+  Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options);
  void RecordCompactionIOStats();
  Status OpenCompactionOutputFile(SubcompactionState* sub_compact);
  void CleanupCompaction();
@@ -129,11 +129,19 @@ class CompactionJob {
  Directory* db_directory_;
  Directory* output_directory_;
  Statistics* stats_;
+  InstrumentedMutex* db_mutex_;
+  Status* db_bg_error_;
  // If there were two snapshots with seq numbers s1 and
  // s2 and s1 < s2, and if we find two instances of a key k1 then lies
  // entirely within s1 and s2, then the earlier version of k1 can be safely
  // deleted because that version is not visible in any snapshot.
  std::vector<SequenceNumber> existing_snapshots_;
+
+  // This is the earliest snapshot that could be used for write-conflict
+  // checking by a transaction.  For any user-key newer than this snapshot, we
+  // should make sure not to remove evidence that a write occurred.
+  SequenceNumber earliest_write_conflict_snapshot_;
+
  std::shared_ptr<Cache> table_cache_;

  EventLogger* event_logger_;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -551,8 +551,9 @@ uint64_t EstimatedFileSize(
  const size_t kFooterSize = 512;

  uint64_t data_size =
+    static_cast<uint64_t>(
      num_records * (key_size + value_size * compression_ratio +
-                     kPerKeyOverhead);
+                     kPerKeyOverhead));

  return data_size + kFooterSize
         + num_records * bloom_bits_per_key / 8      // filter block
@@ -659,7 +660,7 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
  options.max_subcompactions = max_subcompactions_;
  options.bytes_per_sync = 512 * 1024;

-  options.compaction_measure_io_stats = true;
+  options.report_bg_io_stats = true;
  for (int test = 0; test < 2; ++test) {
    DestroyAndReopen(options);
    CreateAndReopenWithCF({"pikachu"}, options);
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -162,9 +162,12 @@ class CompactionJobTest : public testing::Test {
        auto key = ToString(i * kMatchingKeys + k);
        auto value = ToString(i * kKeysPerFile + k);
        InternalKey internal_key(key, ++sequence_number, kTypeValue);
+
        // This is how the key will look like once it's written in bottommost
        // file
-        InternalKey bottommost_internal_key(key, 0, kTypeValue);
+        InternalKey bottommost_internal_key(
+            key, (key == "9999") ? sequence_number : 0, kTypeValue);
+
        if (corrupt_id(k)) {
          test::CorruptKeyType(&internal_key);
          test::CorruptKeyType(&bottommost_internal_key);
@@ -217,9 +220,11 @@ class CompactionJobTest : public testing::Test {
    cfd_ = versions_->GetColumnFamilySet()->GetDefault();
  }

-  void RunCompaction(const std::vector<std::vector<FileMetaData*>>& input_files,
-                     const stl_wrappers::KVMap& expected_results,
-                     const std::vector<SequenceNumber>& snapshots = {}) {
+  void RunCompaction(
+      const std::vector<std::vector<FileMetaData*>>& input_files,
+      const stl_wrappers::KVMap& expected_results,
+      const std::vector<SequenceNumber>& snapshots = {},
+      SequenceNumber earliest_write_conflict_snapshot = kMaxSequenceNumber) {
    auto cfd = versions_->GetColumnFamilySet()->GetDefault();

    size_t num_input_files = 0;
@@ -243,11 +248,11 @@ class CompactionJobTest : public testing::Test {
    LogBuffer log_buffer(InfoLogLevel::INFO_LEVEL, db_options_.info_log.get());
    mutex_.Lock();
    EventLogger event_logger(db_options_.info_log.get());
-    CompactionJob compaction_job(0, &compaction, db_options_, env_options_,
-                                 versions_.get(), &shutting_down_, &log_buffer,
-                                 nullptr, nullptr, nullptr, snapshots,
-                                 table_cache_, &event_logger, false, false,
-                                 dbname_, &compaction_job_stats_);
+    CompactionJob compaction_job(
+        0, &compaction, db_options_, env_options_, versions_.get(),
+        &shutting_down_, &log_buffer, nullptr, nullptr, nullptr, &mutex_,
+        &bg_error_, snapshots, earliest_write_conflict_snapshot, table_cache_,
+        &event_logger, false, false, dbname_, &compaction_job_stats_);

    VerifyInitializationOfCompactionJobStats(compaction_job_stats_);

@@ -257,8 +262,7 @@ class CompactionJobTest : public testing::Test {
    s = compaction_job.Run();
    ASSERT_OK(s);
    mutex_.Lock();
-    ASSERT_OK(compaction_job.Install(*cfd->GetLatestMutableCFOptions(),
-                                     &mutex_));
+    ASSERT_OK(compaction_job.Install(*cfd->GetLatestMutableCFOptions()));
    mutex_.Unlock();

    if (expected_results.size() == 0) {
@@ -290,6 +294,7 @@ class CompactionJobTest : public testing::Test {
  ColumnFamilyData* cfd_;
  std::unique_ptr<CompactionFilter> compaction_filter_;
  std::shared_ptr<MergeOperator> merge_op_;
+  Status bg_error_;
 };

 TEST_F(CompactionJobTest, Simple) {
@@ -346,7 +351,7 @@ TEST_F(CompactionJobTest, SimpleOverwrite) {

  auto expected_results =
      mock::MakeMockFile({{KeyStr("a", 0U, kTypeValue), "val2"},
-                          {KeyStr("b", 0U, kTypeValue), "val3"}});
+                          {KeyStr("b", 4U, kTypeValue), "val3"}});

  SetLastSequence(4U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
@@ -399,7 +404,7 @@ TEST_F(CompactionJobTest, SimpleMerge) {

  auto expected_results =
      mock::MakeMockFile({{KeyStr("a", 0U, kTypeValue), "3,4,5"},
-                          {KeyStr("b", 0U, kTypeValue), "1,2"}});
+                          {KeyStr("b", 2U, kTypeValue), "1,2"}});

  SetLastSequence(5U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
@@ -559,34 +564,144 @@ TEST_F(CompactionJobTest, SimpleSingleDelete) {
 TEST_F(CompactionJobTest, SingleDeleteSnapshots) {
  NewDB();

-  auto file1 = mock::MakeMockFile({{KeyStr("A", 12U, kTypeSingleDeletion), ""},
-                                   {KeyStr("a", 12U, kTypeSingleDeletion), ""},
-                                   {KeyStr("b", 21U, kTypeSingleDeletion), ""},
-                                   {KeyStr("c", 22U, kTypeSingleDeletion), ""},
-                                   {KeyStr("d", 9U, kTypeSingleDeletion), ""}});
+  auto file1 = mock::MakeMockFile({
+      {KeyStr("A", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("a", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("b", 21U, kTypeSingleDeletion), ""},
+      {KeyStr("c", 22U, kTypeSingleDeletion), ""},
+      {KeyStr("d", 9U, kTypeSingleDeletion), ""},
+      {KeyStr("f", 21U, kTypeSingleDeletion), ""},
+      {KeyStr("j", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("j", 9U, kTypeSingleDeletion), ""},
+      {KeyStr("k", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("k", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("l", 3U, kTypeSingleDeletion), ""},
+      {KeyStr("l", 2U, kTypeSingleDeletion), ""},
+  });
  AddMockFile(file1);

-  auto file2 = mock::MakeMockFile({{KeyStr("0", 2U, kTypeSingleDeletion), ""},
-                                   {KeyStr("a", 11U, kTypeValue), "val1"},
-                                   {KeyStr("b", 11U, kTypeValue), "val2"},
-                                   {KeyStr("c", 21U, kTypeValue), "val3"},
-                                   {KeyStr("d", 8U, kTypeValue), "val4"},
-                                   {KeyStr("e", 2U, kTypeSingleDeletion), ""}});
+  auto file2 = mock::MakeMockFile({
+      {KeyStr("0", 2U, kTypeSingleDeletion), ""},
+      {KeyStr("a", 11U, kTypeValue), "val1"},
+      {KeyStr("b", 11U, kTypeValue), "val2"},
+      {KeyStr("c", 21U, kTypeValue), "val3"},
+      {KeyStr("d", 8U, kTypeValue), "val4"},
+      {KeyStr("e", 2U, kTypeSingleDeletion), ""},
+      {KeyStr("f", 1U, kTypeValue), "val1"},
+      {KeyStr("g", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("h", 2U, kTypeSingleDeletion), ""},
+      {KeyStr("m", 12U, kTypeValue), "val1"},
+      {KeyStr("m", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("m", 8U, kTypeValue), "val2"},
+  });
  AddMockFile(file2);

-  auto file3 = mock::MakeMockFile({{KeyStr("A", 1U, kTypeValue), "val"},
-                                   {KeyStr("e", 1U, kTypeValue), "val"}});
+  auto file3 = mock::MakeMockFile({
+      {KeyStr("A", 1U, kTypeValue), "val"},
+      {KeyStr("e", 1U, kTypeValue), "val"},
+  });
  AddMockFile(file3, 2);

-  auto expected_results =
-      mock::MakeMockFile({{KeyStr("A", 12U, kTypeSingleDeletion), ""},
-                          {KeyStr("b", 21U, kTypeSingleDeletion), ""},
-                          {KeyStr("b", 11U, kTypeValue), "val2"},
-                          {KeyStr("e", 2U, kTypeSingleDeletion), ""}});
+  auto expected_results = mock::MakeMockFile({
+      {KeyStr("A", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("a", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("a", 11U, kTypeValue), ""},
+      {KeyStr("b", 21U, kTypeSingleDeletion), ""},
+      {KeyStr("b", 11U, kTypeValue), "val2"},
+      {KeyStr("c", 22U, kTypeSingleDeletion), ""},
+      {KeyStr("c", 21U, kTypeValue), ""},
+      {KeyStr("e", 2U, kTypeSingleDeletion), ""},
+      {KeyStr("f", 21U, kTypeSingleDeletion), ""},
+      {KeyStr("f", 1U, kTypeValue), "val1"},
+      {KeyStr("g", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("j", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("k", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("m", 12U, kTypeValue), "val1"},
+      {KeyStr("m", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("m", 8U, kTypeValue), "val2"},
+  });

  SetLastSequence(22U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
-  RunCompaction({files}, expected_results, {10U, 20U});
+  RunCompaction({files}, expected_results, {10U, 20U}, 10U);
+}
+
+TEST_F(CompactionJobTest, EarliestWriteConflictSnapshot) {
+  NewDB();
+
+  // Test multiple snapshots where the earliest snapshot is not a
+  // write-conflic-snapshot.
+
+  auto file1 = mock::MakeMockFile({
+      {KeyStr("A", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("A", 23U, kTypeValue), "val"},
+      {KeyStr("B", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("B", 23U, kTypeValue), "val"},
+      {KeyStr("D", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("G", 32U, kTypeSingleDeletion), ""},
+      {KeyStr("G", 31U, kTypeValue), "val"},
+      {KeyStr("G", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("G", 23U, kTypeValue), "val2"},
+      {KeyStr("H", 31U, kTypeValue), "val"},
+      {KeyStr("H", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("H", 23U, kTypeValue), "val"},
+      {KeyStr("I", 35U, kTypeSingleDeletion), ""},
+      {KeyStr("I", 34U, kTypeValue), "val2"},
+      {KeyStr("I", 33U, kTypeSingleDeletion), ""},
+      {KeyStr("I", 32U, kTypeValue), "val3"},
+      {KeyStr("I", 31U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 34U, kTypeValue), "val"},
+      {KeyStr("J", 33U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 25U, kTypeValue), "val2"},
+      {KeyStr("J", 24U, kTypeSingleDeletion), ""},
+  });
+  AddMockFile(file1);
+
+  auto file2 = mock::MakeMockFile({
+      {KeyStr("A", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("A", 13U, kTypeValue), "val2"},
+      {KeyStr("C", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("C", 13U, kTypeValue), "val"},
+      {KeyStr("E", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("F", 4U, kTypeSingleDeletion), ""},
+      {KeyStr("F", 3U, kTypeValue), "val"},
+      {KeyStr("G", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("G", 13U, kTypeValue), "val3"},
+      {KeyStr("H", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("H", 13U, kTypeValue), "val2"},
+      {KeyStr("I", 13U, kTypeValue), "val4"},
+      {KeyStr("I", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("I", 11U, kTypeValue), "val5"},
+      {KeyStr("J", 15U, kTypeValue), "val3"},
+      {KeyStr("J", 14U, kTypeSingleDeletion), ""},
+  });
+  AddMockFile(file2);
+
+  auto expected_results = mock::MakeMockFile({
+      {KeyStr("A", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("A", 23U, kTypeValue), ""},
+      {KeyStr("B", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("B", 23U, kTypeValue), ""},
+      {KeyStr("D", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("E", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("G", 32U, kTypeSingleDeletion), ""},
+      {KeyStr("G", 31U, kTypeValue), ""},
+      {KeyStr("H", 31U, kTypeValue), "val"},
+      {KeyStr("I", 35U, kTypeSingleDeletion), ""},
+      {KeyStr("I", 34U, kTypeValue), ""},
+      {KeyStr("I", 31U, kTypeSingleDeletion), ""},
+      {KeyStr("I", 13U, kTypeValue), "val4"},
+      {KeyStr("J", 34U, kTypeValue), "val"},
+      {KeyStr("J", 33U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 25U, kTypeValue), "val2"},
+      {KeyStr("J", 24U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 15U, kTypeValue), "val3"},
+      {KeyStr("J", 14U, kTypeSingleDeletion), ""},
+  });
+
+  SetLastSequence(24U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results, {10U, 20U, 30U}, 20U);
 }

 TEST_F(CompactionJobTest, SingleDeleteZeroSeq) {
@@ -604,7 +719,7 @@ TEST_F(CompactionJobTest, SingleDeleteZeroSeq) {
  AddMockFile(file2);

  auto expected_results = mock::MakeMockFile({
-      {KeyStr("dummy", 0U, kTypeValue), "val2"},
+      {KeyStr("dummy", 5U, kTypeValue), "val2"},
  });

  SetLastSequence(22U);
@@ -616,9 +731,22 @@ TEST_F(CompactionJobTest, MultiSingleDelete) {
  // Tests three scenarios involving multiple single delete/put pairs:
  //
  // A: Put Snapshot SDel Put SDel -> Put Snapshot SDel
-  // B: Put SDel Put SDel -> (Removed)
+  // B: Snapshot Put SDel Put SDel Snapshot -> Snapshot SDel Snapshot
  // C: SDel Put SDel Snapshot Put -> Snapshot Put
-  // D: (Put) SDel Snapshot Put SDel -> (Put) SDel Snapshot
+  // D: (Put) SDel Snapshot Put SDel -> (Put) SDel Snapshot SDel
+  // E: Put SDel Snapshot Put SDel -> Snapshot SDel
+  // F: Put SDel Put Sdel Snapshot -> removed
+  // G: Snapshot SDel Put SDel Put -> Snapshot Put SDel
+  // H: (Put) Put SDel Put Sdel Snapshot -> Removed
+  // I: (Put) Snapshot Put SDel Put SDel -> SDel
+  // J: Put Put SDel Put SDel SDel Snapshot Put Put SDel SDel Put
+  //      -> Snapshot Put
+  // K: SDel SDel Put SDel Put Put Snapshot SDel Put SDel SDel Put SDel
+  //      -> Snapshot Put Snapshot SDel
+  // L: SDel Put Del Put SDel Snapshot Del Put Del SDel Put SDel
+  //      -> Snapshot SDel
+  // M: (Put) SDel Put Del Put SDel Snapshot Put Del SDel Put SDel Del
+  //      -> SDel Snapshot Del
  NewDB();

  auto file1 = mock::MakeMockFile({
@@ -630,6 +758,34 @@ TEST_F(CompactionJobTest, MultiSingleDelete) {
      {KeyStr("C", 14U, kTypeValue), "val3"},
      {KeyStr("D", 12U, kTypeSingleDeletion), ""},
      {KeyStr("D", 11U, kTypeValue), "val4"},
+      {KeyStr("G", 15U, kTypeValue), "val"},
+      {KeyStr("G", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("G", 13U, kTypeValue), "val"},
+      {KeyStr("I", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("I", 13U, kTypeValue), "val"},
+      {KeyStr("J", 15U, kTypeValue), "val"},
+      {KeyStr("J", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 13U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 12U, kTypeValue), "val"},
+      {KeyStr("J", 11U, kTypeValue), "val"},
+      {KeyStr("K", 16U, kTypeSingleDeletion), ""},
+      {KeyStr("K", 15U, kTypeValue), "val1"},
+      {KeyStr("K", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("K", 13U, kTypeSingleDeletion), ""},
+      {KeyStr("K", 12U, kTypeValue), "val2"},
+      {KeyStr("K", 11U, kTypeSingleDeletion), ""},
+      {KeyStr("L", 16U, kTypeSingleDeletion), ""},
+      {KeyStr("L", 15U, kTypeValue), "val"},
+      {KeyStr("L", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("L", 13U, kTypeDeletion), ""},
+      {KeyStr("L", 12U, kTypeValue), "val"},
+      {KeyStr("L", 11U, kTypeDeletion), ""},
+      {KeyStr("M", 16U, kTypeDeletion), ""},
+      {KeyStr("M", 15U, kTypeSingleDeletion), ""},
+      {KeyStr("M", 14U, kTypeValue), "val"},
+      {KeyStr("M", 13U, kTypeSingleDeletion), ""},
+      {KeyStr("M", 12U, kTypeDeletion), ""},
+      {KeyStr("M", 11U, kTypeValue), "val"},
  });
  AddMockFile(file1);

@@ -641,24 +797,87 @@ TEST_F(CompactionJobTest, MultiSingleDelete) {
      {KeyStr("C", 9U, kTypeValue), "val6"},
      {KeyStr("C", 8U, kTypeSingleDeletion), ""},
      {KeyStr("D", 10U, kTypeSingleDeletion), ""},
+      {KeyStr("E", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("E", 11U, kTypeValue), "val"},
+      {KeyStr("E", 5U, kTypeSingleDeletion), ""},
+      {KeyStr("E", 4U, kTypeValue), "val"},
+      {KeyStr("F", 6U, kTypeSingleDeletion), ""},
+      {KeyStr("F", 5U, kTypeValue), "val"},
+      {KeyStr("F", 4U, kTypeSingleDeletion), ""},
+      {KeyStr("F", 3U, kTypeValue), "val"},
+      {KeyStr("G", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("H", 6U, kTypeSingleDeletion), ""},
+      {KeyStr("H", 5U, kTypeValue), "val"},
+      {KeyStr("H", 4U, kTypeSingleDeletion), ""},
+      {KeyStr("H", 3U, kTypeValue), "val"},
+      {KeyStr("I", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("I", 11U, kTypeValue), "val"},
+      {KeyStr("J", 6U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 5U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 4U, kTypeValue), "val"},
+      {KeyStr("J", 3U, kTypeSingleDeletion), ""},
+      {KeyStr("J", 2U, kTypeValue), "val"},
+      {KeyStr("K", 8U, kTypeValue), "val3"},
+      {KeyStr("K", 7U, kTypeValue), "val4"},
+      {KeyStr("K", 6U, kTypeSingleDeletion), ""},
+      {KeyStr("K", 5U, kTypeValue), "val5"},
+      {KeyStr("K", 2U, kTypeSingleDeletion), ""},
+      {KeyStr("K", 1U, kTypeSingleDeletion), ""},
+      {KeyStr("L", 5U, kTypeSingleDeletion), ""},
+      {KeyStr("L", 4U, kTypeValue), "val"},
+      {KeyStr("L", 3U, kTypeDeletion), ""},
+      {KeyStr("L", 2U, kTypeValue), "val"},
+      {KeyStr("L", 1U, kTypeSingleDeletion), ""},
+      {KeyStr("M", 10U, kTypeSingleDeletion), ""},
+      {KeyStr("M", 7U, kTypeValue), "val"},
+      {KeyStr("M", 5U, kTypeDeletion), ""},
+      {KeyStr("M", 4U, kTypeValue), "val"},
+      {KeyStr("M", 3U, kTypeSingleDeletion), ""},
  });
  AddMockFile(file2);

  auto file3 = mock::MakeMockFile({
-      {KeyStr("D", 11U, kTypeValue), "val"},
+      {KeyStr("D", 1U, kTypeValue), "val"},
+      {KeyStr("H", 1U, kTypeValue), "val"},
+      {KeyStr("I", 2U, kTypeValue), "val"},
  });
  AddMockFile(file3, 2);

-  auto expected_results = mock::MakeMockFile({
-      {KeyStr("A", 12U, kTypeSingleDeletion), ""},
-      {KeyStr("A", 10U, kTypeValue), "val"},
-      {KeyStr("C", 14U, kTypeValue), "val3"},
-      {KeyStr("D", 10U, kTypeSingleDeletion), ""},
+  auto file4 = mock::MakeMockFile({
+      {KeyStr("M", 1U, kTypeValue), "val"},
  });
+  AddMockFile(file4, 2);
+
+  auto expected_results =
+      mock::MakeMockFile({{KeyStr("A", 14U, kTypeSingleDeletion), ""},
+                          {KeyStr("A", 13U, kTypeValue), ""},
+                          {KeyStr("A", 12U, kTypeSingleDeletion), ""},
+                          {KeyStr("A", 10U, kTypeValue), "val"},
+                          {KeyStr("B", 14U, kTypeSingleDeletion), ""},
+                          {KeyStr("B", 13U, kTypeValue), ""},
+                          {KeyStr("C", 14U, kTypeValue), "val3"},
+                          {KeyStr("D", 12U, kTypeSingleDeletion), ""},
+                          {KeyStr("D", 11U, kTypeValue), ""},
+                          {KeyStr("D", 10U, kTypeSingleDeletion), ""},
+                          {KeyStr("E", 12U, kTypeSingleDeletion), ""},
+                          {KeyStr("E", 11U, kTypeValue), ""},
+                          {KeyStr("G", 15U, kTypeValue), "val"},
+                          {KeyStr("G", 12U, kTypeSingleDeletion), ""},
+                          {KeyStr("I", 14U, kTypeSingleDeletion), ""},
+                          {KeyStr("I", 13U, kTypeValue), ""},
+                          {KeyStr("J", 15U, kTypeValue), "val"},
+                          {KeyStr("K", 16U, kTypeSingleDeletion), ""},
+                          {KeyStr("K", 15U, kTypeValue), ""},
+                          {KeyStr("K", 11U, kTypeSingleDeletion), ""},
+                          {KeyStr("K", 8U, kTypeValue), "val3"},
+                          {KeyStr("L", 16U, kTypeSingleDeletion), ""},
+                          {KeyStr("L", 15U, kTypeValue), ""},
+                          {KeyStr("M", 16U, kTypeDeletion), ""},
+                          {KeyStr("M", 3U, kTypeSingleDeletion), ""}});

  SetLastSequence(22U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
-  RunCompaction({files}, expected_results, {10U});
+  RunCompaction({files}, expected_results, {10U}, 10U);
 }

 // This test documents the behavior where a corrupt key follows a deletion or a
@@ -684,7 +903,7 @@ TEST_F(CompactionJobTest, CorruptionAfterDeletion) {
      mock::MakeMockFile({{test::KeyStr("A", 0U, kTypeValue), "val3"},
                          {test::KeyStr("a", 0U, kTypeValue, true), "val"},
                          {test::KeyStr("b", 0U, kTypeValue, true), "val"},
-                          {test::KeyStr("c", 0U, kTypeValue), "val2"}});
+                          {test::KeyStr("c", 1U, kTypeValue), "val2"}});

  SetLastSequence(6U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -141,7 +141,8 @@ CompactionPicker::~CompactionPicker() {}

 // Delete this compaction from the list of running compactions.
 void CompactionPicker::ReleaseCompactionFiles(Compaction* c, Status status) {
-  if (c->start_level() == 0) {
+  if (c->start_level() == 0 ||
+      ioptions_.compaction_style == kCompactionStyleUniversal) {
    level0_compactions_in_progress_.erase(c);
  }
  if (!status.ok()) {
@@ -243,7 +244,7 @@ bool CompactionPicker::ExpandWhileOverlapping(const std::string& cf_name,
 // Returns true if any one of specified files are being compacted
 bool CompactionPicker::FilesInCompaction(
    const std::vector<FileMetaData*>& files) {
-  for (unsigned int i = 0; i < files.size(); i++) {
+  for (size_t i = 0; i < files.size(); i++) {
    if (files[i]->being_compacted) {
      return true;
    }
@@ -255,16 +256,30 @@ 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(
@@ -441,7 +456,7 @@ Compaction* CompactionPicker::CompactRange(
    const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
    VersionStorageInfo* vstorage, int input_level, int output_level,
    uint32_t output_path_id, const InternalKey* begin, const InternalKey* end,
-    InternalKey** compaction_end) {
+    InternalKey** compaction_end, bool* manual_conflict) {
  // CompactionPickerFIFO has its own implementation of compact range
  assert(ioptions_.compaction_style != kCompactionStyleFIFO);

@@ -467,6 +482,12 @@ Compaction* CompactionPicker::CompactRange(
      return nullptr;
    }

+    if ((start_level == 0) && (!level0_compactions_in_progress_.empty())) {
+      *manual_conflict = true;
+      // Only one level 0 compaction allowed
+      return nullptr;
+    }
+
    std::vector<CompactionInputFiles> inputs(vstorage->num_levels() -
                                             start_level);
    for (int level = start_level; level < vstorage->num_levels(); level++) {
@@ -475,13 +496,21 @@ Compaction* CompactionPicker::CompactRange(
      for (FileMetaData* f : vstorage->LevelFiles(level)) {
        files.push_back(f);
      }
+      if (FilesInCompaction(files)) {
+        *manual_conflict = true;
+        return nullptr;
+      }
    }
-    return new Compaction(
+    Compaction* c = new Compaction(
        vstorage, mutable_cf_options, std::move(inputs), output_level,
        mutable_cf_options.MaxFileSizeForLevel(output_level),
        /* max_grandparent_overlap_bytes */ LLONG_MAX, output_path_id,
        GetCompressionType(ioptions_, output_level, 1),
        /* grandparents */ {}, /* is manual */ true);
+    if (start_level == 0) {
+      level0_compactions_in_progress_.insert(c);
+    }
+    return c;
  }

  CompactionInputFiles inputs;
@@ -500,6 +529,13 @@ Compaction* CompactionPicker::CompactRange(
    return nullptr;
  }

+  if ((input_level == 0) && (!level0_compactions_in_progress_.empty())) {
+    // Only one level 0 compaction allowed
+    TEST_SYNC_POINT("CompactionPicker::CompactRange:Conflict");
+    *manual_conflict = true;
+    return nullptr;
+  }
+
  // Avoid compacting too much in one shot in case the range is large.
  // But we cannot do this for level-0 since level-0 files can overlap
  // and we must not pick one file and drop another older file if the
@@ -522,9 +558,10 @@ Compaction* CompactionPicker::CompactRange(
  assert(output_path_id < static_cast<uint32_t>(ioptions_.db_paths.size()));

  if (ExpandWhileOverlapping(cf_name, vstorage, &inputs) == false) {
-    // manual compaction is currently single-threaded, so it should never
+    // manual compaction is now multi-threaded, so it can
    // happen that ExpandWhileOverlapping fails
-    assert(false);
+    // we handle it higher in RunManualCompaction
+    *manual_conflict = true;
    return nullptr;
  }

@@ -543,9 +580,10 @@ Compaction* CompactionPicker::CompactRange(
    int parent_index = -1;
    if (!SetupOtherInputs(cf_name, mutable_cf_options, vstorage, &inputs,
                          &output_level_inputs, &parent_index, -1)) {
-      // manual compaction is currently single-threaded, so it should never
+      // manual compaction is now multi-threaded, so it can
      // happen that SetupOtherInputs fails
-      assert(false);
+      // we handle it higher in RunManualCompaction
+      *manual_conflict = true;
      return nullptr;
    }
  }
@@ -554,6 +592,12 @@ Compaction* CompactionPicker::CompactRange(
  if (!output_level_inputs.empty()) {
    compaction_inputs.push_back(output_level_inputs);
  }
+  for (size_t i = 0; i < compaction_inputs.size(); i++) {
+    if (FilesInCompaction(compaction_inputs[i].files)) {
+      *manual_conflict = true;
+      return nullptr;
+    }
+  }

  std::vector<FileMetaData*> grandparents;
  GetGrandparents(vstorage, inputs, output_level_inputs, &grandparents);
@@ -566,6 +610,20 @@ Compaction* CompactionPicker::CompactRange(
      std::move(grandparents), /* is manual compaction */ true);

  TEST_SYNC_POINT_CALLBACK("CompactionPicker::CompactRange:Return", compaction);
+  if (input_level == 0) {
+    level0_compactions_in_progress_.insert(compaction);
+  }
+
+  // Creating a compaction influences the compaction score because the score
+  // takes running compactions into account (by skipping files that are already
+  // being compacted). Since we just changed compaction score, we recalculate it
+  // here
+  {  // this piece of code recomputes compaction score
+    CompactionOptionsFIFO dummy_compaction_options_fifo;
+    vstorage->ComputeCompactionScore(mutable_cf_options,
+                                     dummy_compaction_options_fifo);
+  }
+
  return compaction;
 }

@@ -870,22 +928,40 @@ Compaction* LevelCompactionPicker::PickCompaction(
  int base_index = -1;
  CompactionInputFiles inputs;
  double score = 0;
+  CompactionReason compaction_reason = CompactionReason::kUnknown;

  // Find the compactions by size on all levels.
+  bool skipped_l0 = false;
  for (int i = 0; i < NumberLevels() - 1; i++) {
    score = vstorage->CompactionScore(i);
    level = vstorage->CompactionScoreLevel(i);
    assert(i == 0 || score <= vstorage->CompactionScore(i - 1));
    if (score >= 1) {
+      if (skipped_l0 && level == vstorage->base_level()) {
+        // If L0->base_level compaction is pending, don't schedule further
+        // compaction from base level. Otherwise L0->base_level compaction
+        // may starve.
+        continue;
+      }
      output_level = (level == 0) ? vstorage->base_level() : level + 1;
      if (PickCompactionBySize(vstorage, level, output_level, &inputs,
                               &parent_index, &base_index) &&
          ExpandWhileOverlapping(cf_name, vstorage, &inputs)) {
        // found the compaction!
+        if (level == 0) {
+          // L0 score = `num L0 files` / `level0_file_num_compaction_trigger`
+          compaction_reason = CompactionReason::kLevelL0FilesNum;
+        } else {
+          // L1+ score = `Level files size` / `MaxBytesForLevel`
+          compaction_reason = CompactionReason::kLevelMaxLevelSize;
+        }
        break;
      } else {
        // didn't find the compaction, clear the inputs
        inputs.clear();
+        if (level == 0) {
+          skipped_l0 = true;
+        }
      }
    }
  }
@@ -898,6 +974,9 @@ Compaction* LevelCompactionPicker::PickCompaction(
    parent_index = base_index = -1;
    PickFilesMarkedForCompactionExperimental(cf_name, vstorage, &inputs, &level,
                                             &output_level);
+    if (!inputs.empty()) {
+      compaction_reason = CompactionReason::kFilesMarkedForCompaction;
+    }
  }
  if (inputs.empty()) {
    return nullptr;
@@ -948,7 +1027,8 @@ Compaction* LevelCompactionPicker::PickCompaction(
      mutable_cf_options.MaxGrandParentOverlapBytes(level),
      GetPathId(ioptions_, mutable_cf_options, output_level),
      GetCompressionType(ioptions_, output_level, vstorage->base_level()),
-      std::move(grandparents), is_manual, score);
+      std::move(grandparents), is_manual, score,
+      false /* deletion_compaction */, compaction_reason);

  // If it's level 0 compaction, make sure we don't execute any other level 0
  // compactions in parallel
@@ -1019,6 +1099,7 @@ bool LevelCompactionPicker::PickCompactionBySize(VersionStorageInfo* vstorage,
  // could be made better by looking at key-ranges that are
  // being compacted at level 0.
  if (level == 0 && !level0_compactions_in_progress_.empty()) {
+    TEST_SYNC_POINT("LevelCompactionPicker::PickCompactionBySize:0");
    return false;
  }

@@ -1095,18 +1176,19 @@ void UniversalCompactionPicker::SortedRun::Dump(char* out_buf,
 }

 void UniversalCompactionPicker::SortedRun::DumpSizeInfo(
-    char* out_buf, size_t out_buf_size, int sorted_run_count) const {
+    char* out_buf, size_t out_buf_size, size_t sorted_run_count) const {
  if (level == 0) {
    assert(file != nullptr);
    snprintf(out_buf, out_buf_size,
-             "file %" PRIu64
-             "[%d] "
+             "file %" PRIu64 "[%" ROCKSDB_PRIszt
+             "] "
             "with size %" PRIu64 " (compensated size %" PRIu64 ")",
             file->fd.GetNumber(), sorted_run_count, file->fd.GetFileSize(),
             file->compensated_file_size);
  } else {
    snprintf(out_buf, out_buf_size,
-             "level %d[%d] "
+             "level %d[%" ROCKSDB_PRIszt
+             "] "
             "with size %" PRIu64 " (compensated size %" PRIu64 ")",
             level, sorted_run_count, size, compensated_file_size);
  }
@@ -1234,8 +1316,9 @@ 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;
  }
@@ -1387,16 +1470,21 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalReadAmp(

  const SortedRun* sr = nullptr;
  bool done = false;
-  int start_index = 0;
+  size_t start_index = 0;
  unsigned int candidate_count = 0;

  unsigned int max_files_to_compact = std::min(max_merge_width,
                                       max_number_of_files_to_compact);
  min_merge_width = std::max(min_merge_width, 2U);

+  // Caller checks the size before executing this function. This invariant is
+  // important because otherwise we may have a possible integer underflow when
+  // dealing with unsigned types.
+  assert(sorted_runs.size() > 0);
+
  // Considers a candidate file only if it is smaller than the
  // total size accumulated so far.
-  for (unsigned int loop = 0; loop < sorted_runs.size(); loop++) {
+  for (size_t loop = 0; loop < sorted_runs.size(); loop++) {
    candidate_count = 0;

    // Skip files that are already being compacted
@@ -1428,7 +1516,7 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalReadAmp(
    }

    // Check if the succeeding files need compaction.
-    for (unsigned int i = loop + 1;
+    for (size_t i = loop + 1;
         candidate_count < max_files_to_compact && i < sorted_runs.size();
         i++) {
      const SortedRun* succeeding_sr = &sorted_runs[i];
@@ -1470,7 +1558,7 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalReadAmp(
      done = true;
      break;
    } else {
-      for (unsigned int i = loop;
+      for (size_t i = loop;
           i < loop + candidate_count && i < sorted_runs.size(); i++) {
        const SortedRun* skipping_sr = &sorted_runs[i];
        char file_num_buf[256];
@@ -1483,7 +1571,7 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalReadAmp(
  if (!done || candidate_count <= 1) {
    return nullptr;
  }
-  unsigned int first_index_after = start_index + candidate_count;
+  size_t first_index_after = start_index + candidate_count;
  // Compression is enabled if files compacted earlier already reached
  // size ratio of compression.
  bool enable_compression = true;
@@ -1524,7 +1612,7 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalReadAmp(
  for (size_t i = 0; i < inputs.size(); ++i) {
    inputs[i].level = start_level + static_cast<int>(i);
  }
-  for (unsigned int i = start_index; i < first_index_after; i++) {
+  for (size_t i = start_index; i < first_index_after; i++) {
    auto& picking_sr = sorted_runs[i];
    if (picking_sr.level == 0) {
      FileMetaData* picking_file = picking_sr.file;
@@ -1541,11 +1629,18 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalReadAmp(
                file_num_buf);
  }

+  CompactionReason compaction_reason;
+  if (max_number_of_files_to_compact == UINT_MAX) {
+    compaction_reason = CompactionReason::kUniversalSortedRunNum;
+  } else {
+    compaction_reason = CompactionReason::kUniversalSizeRatio;
+  }
  return new Compaction(
      vstorage, mutable_cf_options, std::move(inputs), output_level,
      mutable_cf_options.MaxFileSizeForLevel(output_level), LLONG_MAX, path_id,
      GetCompressionType(ioptions_, start_level, 1, enable_compression),
-      /* grandparents */ {}, /* is manual */ false, score);
+      /* grandparents */ {}, /* is manual */ false, score,
+      false /* deletion_compaction */, compaction_reason);
 }

 // Look at overall size amplification. If size amplification
@@ -1564,11 +1659,11 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalSizeAmp(

  unsigned int candidate_count = 0;
  uint64_t candidate_size = 0;
-  unsigned int start_index = 0;
+  size_t start_index = 0;
  const SortedRun* sr = nullptr;

  // Skip files that are already being compacted
-  for (unsigned int loop = 0; loop < sorted_runs.size() - 1; loop++) {
+  for (size_t loop = 0; loop < sorted_runs.size() - 1; loop++) {
    sr = &sorted_runs[loop];
    if (!sr->being_compacted) {
      start_index = loop;         // Consider this as the first candidate.
@@ -1588,13 +1683,14 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalSizeAmp(
  {
    char file_num_buf[kFormatFileNumberBufSize];
    sr->Dump(file_num_buf, sizeof(file_num_buf), true);
-    LogToBuffer(log_buffer, "[%s] Universal: First candidate %s[%d] %s",
+    LogToBuffer(log_buffer,
+                "[%s] Universal: First candidate %s[%" ROCKSDB_PRIszt "] %s",
                cf_name.c_str(), file_num_buf, start_index,
                " to reduce size amp.\n");
  }

  // keep adding up all the remaining files
-  for (unsigned int loop = start_index; loop < sorted_runs.size() - 1; loop++) {
+  for (size_t loop = start_index; loop < sorted_runs.size() - 1; loop++) {
    sr = &sorted_runs[loop];
    if (sr->being_compacted) {
      char file_num_buf[kFormatFileNumberBufSize];
@@ -1634,7 +1730,7 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalSizeAmp(

  // Estimate total file size
  uint64_t estimated_total_size = 0;
-  for (unsigned int loop = start_index; loop < sorted_runs.size(); loop++) {
+  for (size_t loop = start_index; loop < sorted_runs.size(); loop++) {
    estimated_total_size += sorted_runs[loop].size;
  }
  uint32_t path_id = GetPathId(ioptions_, estimated_total_size);
@@ -1645,7 +1741,7 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalSizeAmp(
    inputs[i].level = start_level + static_cast<int>(i);
  }
  // We always compact all the files, so always compress.
-  for (unsigned int loop = start_index; loop < sorted_runs.size(); loop++) {
+  for (size_t loop = start_index; loop < sorted_runs.size(); loop++) {
    auto& picking_sr = sorted_runs[loop];
    if (picking_sr.level == 0) {
      FileMetaData* f = picking_sr.file;
@@ -1657,7 +1753,7 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalSizeAmp(
      }
    }
    char file_num_buf[256];
-    sr->DumpSizeInfo(file_num_buf, sizeof(file_num_buf), loop);
+    picking_sr.DumpSizeInfo(file_num_buf, sizeof(file_num_buf), loop);
    LogToBuffer(log_buffer, "[%s] Universal: size amp picking %s",
                cf_name.c_str(), file_num_buf);
  }
@@ -1668,7 +1764,9 @@ Compaction* UniversalCompactionPicker::PickCompactionUniversalSizeAmp(
      mutable_cf_options.MaxFileSizeForLevel(vstorage->num_levels() - 1),
      /* max_grandparent_overlap_bytes */ LLONG_MAX, path_id,
      GetCompressionType(ioptions_, vstorage->num_levels() - 1, 1),
-      /* grandparents */ {}, /* is manual */ false, score);
+      /* grandparents */ {}, /* is manual */ false, score,
+      false /* deletion_compaction */,
+      CompactionReason::kUniversalSizeAmplification);
 }

 bool FIFOCompactionPicker::NeedsCompaction(const VersionStorageInfo* vstorage)
@@ -1727,7 +1825,7 @@ Compaction* FIFOCompactionPicker::PickCompaction(
  Compaction* c = new Compaction(
      vstorage, mutable_cf_options, std::move(inputs), 0, 0, 0, 0,
      kNoCompression, {}, /* is manual */ false, vstorage->CompactionScore(0),
-      /* is deletion compaction */ true);
+      /* is deletion compaction */ true, CompactionReason::kFIFOMaxSize);
  level0_compactions_in_progress_.insert(c);
  return c;
 }
@@ -1736,7 +1834,7 @@ Compaction* FIFOCompactionPicker::CompactRange(
    const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
    VersionStorageInfo* vstorage, int input_level, int output_level,
    uint32_t output_path_id, const InternalKey* begin, const InternalKey* end,
-    InternalKey** compaction_end) {
+    InternalKey** compaction_end, bool* manual_conflict) {
  assert(input_level == 0);
  assert(output_level == 0);
  *compaction_end = nullptr;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -60,7 +60,7 @@ class CompactionPicker {
      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
      VersionStorageInfo* vstorage, int input_level, int output_level,
      uint32_t output_path_id, const InternalKey* begin, const InternalKey* end,
-      InternalKey** compaction_end);
+      InternalKey** compaction_end, bool* manual_conflict);

  // The maximum allowed output level.  Default value is NumberLevels() - 1.
  virtual int MaxOutputLevel() const {
@@ -83,6 +83,8 @@ class CompactionPicker {
 #endif  // ROCKSDB_LITE

  // Free up the files that participated in a compaction
+  //
+  // Requirement: DB mutex held
  void ReleaseCompactionFiles(Compaction* c, Status status);

  // Returns true if any one of the specified files are being compacted
@@ -94,7 +96,7 @@ 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.
@@ -251,7 +253,7 @@ class UniversalCompactionPicker : public CompactionPicker {

    // sorted_run_count is added into the string to print
    void DumpSizeInfo(char* out_buf, size_t out_buf_size,
-                      int sorted_run_count) const;
+                      size_t sorted_run_count) const;

    int level;
    // `file` Will be null for level > 0. For level = 0, the sorted run is
@@ -302,7 +304,7 @@ class FIFOCompactionPicker : public CompactionPicker {
      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
      VersionStorageInfo* vstorage, int input_level, int output_level,
      uint32_t output_path_id, const InternalKey* begin, const InternalKey* end,
-      InternalKey** compaction_end) override;
+      InternalKey** compaction_end, bool* manual_conflict) override;

  // The maximum allowed output level.  Always returns 0.
  virtual int MaxOutputLevel() const override {
@@ -329,11 +331,13 @@ class NullCompactionPicker : public CompactionPicker {
  }

  // Always return "nullptr"
-  Compaction* CompactRange(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      VersionStorageInfo* vstorage, int input_level, int output_level,
-      uint32_t output_path_id, const InternalKey* begin, const InternalKey* end,
-      InternalKey** compaction_end) override {
+  Compaction* CompactRange(const std::string& cf_name,
+                           const MutableCFOptions& mutable_cf_options,
+                           VersionStorageInfo* vstorage, int input_level,
+                           int output_level, uint32_t output_path_id,
+                           const InternalKey* begin, const InternalKey* end,
+                           InternalKey** compaction_end,
+                           bool* manual_conflict) override {
    return nullptr;
  }

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -195,6 +195,7 @@ TEST_F(CompactionPickerTest, LevelMaxScore) {
  NewVersionStorage(6, kCompactionStyleLevel);
  mutable_cf_options_.target_file_size_base = 10000000;
  mutable_cf_options_.target_file_size_multiplier = 10;
+  mutable_cf_options_.max_bytes_for_level_base = 10 * 1024 * 1024;
  Add(0, 1U, "150", "200", 1000000000U);
  // Level 1 score 1.2
  Add(1, 66U, "150", "200", 6000000U);
@@ -487,6 +488,90 @@ TEST_F(CompactionPickerTest, NeedsCompactionFIFO) {
 }
 #endif  // ROCKSDB_LITE

+TEST_F(CompactionPickerTest, CompactionPriMinOverlapping1) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.target_file_size_base = 10000000;
+  mutable_cf_options_.target_file_size_multiplier = 10;
+  mutable_cf_options_.max_bytes_for_level_base = 10 * 1024 * 1024;
+  mutable_cf_options_.compaction_pri = kMinOverlappingRatio;
+
+  Add(2, 6U, "150", "179", 50000000U);
+  Add(2, 7U, "180", "220", 50000000U);
+  Add(2, 8U, "321", "400", 50000000U);  // File not overlapping
+  Add(2, 9U, "721", "800", 50000000U);
+
+  Add(3, 26U, "150", "170", 260000000U);
+  Add(3, 27U, "171", "179", 260000000U);
+  Add(3, 28U, "191", "220", 260000000U);
+  Add(3, 29U, "221", "300", 260000000U);
+  Add(3, 30U, "750", "900", 260000000U);
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  // Pick file 8 because it overlaps with 0 files on level 3.
+  ASSERT_EQ(8U, compaction->input(0, 0)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, CompactionPriMinOverlapping2) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.target_file_size_base = 10000000;
+  mutable_cf_options_.target_file_size_multiplier = 10;
+  mutable_cf_options_.max_bytes_for_level_base = 10 * 1024 * 1024;
+  mutable_cf_options_.compaction_pri = kMinOverlappingRatio;
+
+  Add(2, 6U, "150", "175",
+      60000000U);  // Overlaps with file 26, 27, total size 521M
+  Add(2, 7U, "176", "200", 60000000U);  // Overlaps with file 27, 28, total size
+                                        // 520M, the smalelst overlapping
+  Add(2, 8U, "201", "300",
+      60000000U);  // Overlaps with file 28, 29, total size 521M
+
+  Add(3, 26U, "100", "110", 261000000U);
+  Add(3, 26U, "150", "170", 261000000U);
+  Add(3, 27U, "171", "179", 260000000U);
+  Add(3, 28U, "191", "220", 260000000U);
+  Add(3, 29U, "221", "300", 261000000U);
+  Add(3, 30U, "321", "400", 261000000U);
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  // Picking file 7 because overlapping ratio is the biggest.
+  ASSERT_EQ(7U, compaction->input(0, 0)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, CompactionPriMinOverlapping3) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.max_bytes_for_level_base = 10000000;
+  mutable_cf_options_.max_bytes_for_level_multiplier = 10;
+  mutable_cf_options_.compaction_pri = kMinOverlappingRatio;
+
+  // file 7 and 8 over lap with the same file, but file 8 is smaller so
+  // it will be picked.
+  Add(2, 6U, "150", "167", 60000000U);  // Overlaps with file 26, 27
+  Add(2, 7U, "168", "169", 60000000U);  // Overlaps with file 27
+  Add(2, 8U, "201", "300", 61000000U);  // Overlaps with file 28, but the file
+                                        // itself is larger. Should be picked.
+
+  Add(3, 26U, "160", "165", 260000000U);
+  Add(3, 27U, "166", "170", 260000000U);
+  Add(3, 28U, "180", "400", 260000000U);
+  Add(3, 29U, "401", "500", 260000000U);
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  // Picking file 8 because overlapping ratio is the biggest.
+  ASSERT_EQ(8U, compaction->input(0, 0)->fd.GetNumber());
+}
+
 // This test exhibits the bug where we don't properly reset parent_index in
 // PickCompaction()
 TEST_F(CompactionPickerTest, ParentIndexResetBug) {
@@ -580,6 +665,101 @@ TEST_F(CompactionPickerTest, OverlappingUserKeys3) {
  ASSERT_EQ(7U, compaction->input(1, 1)->fd.GetNumber());
 }

+TEST_F(CompactionPickerTest, NotScheduleL1IfL0WithHigherPri1) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_bytes_for_level_base = 900000000U;
+
+  // 6 L0 files, score 3.
+  Add(0, 1U, "000", "400", 1U);
+  Add(0, 2U, "001", "400", 1U, 0, 0);
+  Add(0, 3U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 31U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 32U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 33U, "001", "400", 1000000000U, 0, 0);
+
+  // L1 total size 2GB, score 2.2. If one file being comapcted, score 1.1.
+  Add(1, 4U, "050", "300", 1000000000U, 0, 0);
+  file_map_[4u].first->being_compacted = true;
+  Add(1, 5U, "301", "350", 1000000000U, 0, 0);
+
+  // Output level overlaps with the beginning and the end of the chain
+  Add(2, 6U, "050", "100", 1U);
+  Add(2, 7U, "300", "400", 1U);
+
+  // No compaction should be scheduled, if L0 has higher priority than L1
+  // but L0->L1 compaction is blocked by a file in L1 being compacted.
+  UpdateVersionStorageInfo();
+  ASSERT_EQ(0, vstorage_->CompactionScoreLevel(0));
+  ASSERT_EQ(1, vstorage_->CompactionScoreLevel(1));
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() == nullptr);
+}
+
+TEST_F(CompactionPickerTest, NotScheduleL1IfL0WithHigherPri2) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_bytes_for_level_base = 900000000U;
+
+  // 6 L0 files, score 3.
+  Add(0, 1U, "000", "400", 1U);
+  Add(0, 2U, "001", "400", 1U, 0, 0);
+  Add(0, 3U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 31U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 32U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 33U, "001", "400", 1000000000U, 0, 0);
+
+  // L1 total size 2GB, score 2.2. If one file being comapcted, score 1.1.
+  Add(1, 4U, "050", "300", 1000000000U, 0, 0);
+  Add(1, 5U, "301", "350", 1000000000U, 0, 0);
+
+  // Output level overlaps with the beginning and the end of the chain
+  Add(2, 6U, "050", "100", 1U);
+  Add(2, 7U, "300", "400", 1U);
+
+  // If no file in L1 being compacted, L0->L1 compaction will be scheduled.
+  UpdateVersionStorageInfo();  // being_compacted flag is cleared here.
+  ASSERT_EQ(0, vstorage_->CompactionScoreLevel(0));
+  ASSERT_EQ(1, vstorage_->CompactionScoreLevel(1));
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+}
+
+TEST_F(CompactionPickerTest, NotScheduleL1IfL0WithHigherPri3) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
+  mutable_cf_options_.max_bytes_for_level_base = 900000000U;
+
+  // 6 L0 files, score 3.
+  Add(0, 1U, "000", "400", 1U);
+  Add(0, 2U, "001", "400", 1U, 0, 0);
+  Add(0, 3U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 31U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 32U, "001", "400", 1000000000U, 0, 0);
+  Add(0, 33U, "001", "400", 1000000000U, 0, 0);
+
+  // L1 score more than 6.
+  Add(1, 4U, "050", "300", 1000000000U, 0, 0);
+  file_map_[4u].first->being_compacted = true;
+  Add(1, 5U, "301", "350", 1000000000U, 0, 0);
+  Add(1, 51U, "351", "400", 6000000000U, 0, 0);
+
+  // Output level overlaps with the beginning and the end of the chain
+  Add(2, 6U, "050", "100", 1U);
+  Add(2, 7U, "300", "400", 1U);
+
+  // If score in L1 is larger than L0, L1 compaction goes through despite
+  // there is pending L0 compaction.
+  UpdateVersionStorageInfo();
+  ASSERT_EQ(1, vstorage_->CompactionScoreLevel(0));
+  ASSERT_EQ(0, vstorage_->CompactionScoreLevel(1));
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+}
+
 TEST_F(CompactionPickerTest, EstimateCompactionBytesNeeded1) {
  int num_levels = ioptions_.num_levels;
  ioptions_.level_compaction_dynamic_level_bytes = false;
@@ -1,6 +1,6 @@
 // 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.
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -18,6 +18,13 @@ namespace rocksdb {
 void CancelAllBackgroundWork(DB* db, bool wait) {
  (dynamic_cast<DBImpl*>(db))->CancelAllBackgroundWork(wait);
 }
+
+Status DeleteFilesInRange(DB* db, ColumnFamilyHandle* column_family,
+                          const Slice* begin, const Slice* end) {
+  return (dynamic_cast<DBImpl*>(db))
+      ->DeleteFilesInRange(column_family, begin, end);
+}
+
 }  // namespace rocksdb

 #endif  // ROCKSDB_LITE
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -11,6 +11,7 @@

 #include "rocksdb/db.h"

+#include <inttypes.h>
 #include <errno.h>
 #include <fcntl.h>
 #include <sys/stat.h>
@@ -104,8 +105,8 @@ class CorruptionTest : public testing::Test {
  }

  void Check(int min_expected, int max_expected) {
-    unsigned int next_expected = 0;
-    int missed = 0;
+    uint64_t next_expected = 0;
+    uint64_t missed = 0;
    int bad_keys = 0;
    int bad_values = 0;
    int correct = 0;
@@ -126,7 +127,7 @@ class CorruptionTest : public testing::Test {
        continue;
      }
      missed += (key - next_expected);
-      next_expected = static_cast<unsigned int>(key + 1);
+      next_expected = key + 1;
      if (iter->value() != Value(static_cast<int>(key), &value_space)) {
        bad_values++;
      } else {
@@ -136,8 +137,9 @@ class CorruptionTest : public testing::Test {
    delete iter;

    fprintf(stderr,
-            "expected=%d..%d; got=%d; bad_keys=%d; bad_values=%d; missed=%d\n",
-            min_expected, max_expected, correct, bad_keys, bad_values, missed);
+      "expected=%d..%d; got=%d; bad_keys=%d; bad_values=%d; missed=%llu\n",
+            min_expected, max_expected, correct, bad_keys, bad_values,
+            static_cast<unsigned long long>(missed));
    ASSERT_LE(min_expected, correct);
    ASSERT_GE(max_expected, correct);
  }
@@ -183,7 +185,7 @@ class CorruptionTest : public testing::Test {
    FileType type;
    std::string fname;
    int picked_number = -1;
-    for (unsigned int i = 0; i < filenames.size(); i++) {
+    for (size_t i = 0; i < filenames.size(); i++) {
      if (ParseFileName(filenames[i], &number, &type) &&
          type == filetype &&
          static_cast<int>(number) > picked_number) {  // Pick latest file
@@ -232,8 +234,16 @@ 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);
+    }
  }
 };

@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -0,0 +1,240 @@
+//  Copyright (c) 2011-present, 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 <cstdlib>
+#include "db/db_test_util.h"
+#include "port/stack_trace.h"
+
+namespace rocksdb {
+
+static uint64_t TestGetTickerCount(const Options& options,
+                                   Tickers ticker_type) {
+  return options.statistics->getTickerCount(ticker_type);
+}
+
+class DBBlockCacheTest : public DBTestBase {
+ private:
+  size_t miss_count_ = 0;
+  size_t hit_count_ = 0;
+  size_t insert_count_ = 0;
+  size_t failure_count_ = 0;
+  size_t compressed_miss_count_ = 0;
+  size_t compressed_hit_count_ = 0;
+  size_t compressed_insert_count_ = 0;
+  size_t compressed_failure_count_ = 0;
+
+ public:
+  const size_t kNumBlocks = 10;
+  const size_t kValueSize = 100;
+
+  DBBlockCacheTest() : DBTestBase("/db_block_cache_test") {}
+
+  BlockBasedTableOptions GetTableOptions() {
+    BlockBasedTableOptions table_options;
+    // Set a small enough block size so that each key-value get its own block.
+    table_options.block_size = 1;
+    return table_options;
+  }
+
+  Options GetOptions(const BlockBasedTableOptions& table_options) {
+    Options options = CurrentOptions();
+    options.create_if_missing = true;
+    // options.compression = kNoCompression;
+    options.statistics = rocksdb::CreateDBStatistics();
+    options.table_factory.reset(new BlockBasedTableFactory(table_options));
+    return options;
+  }
+
+  void InitTable(const Options& options) {
+    std::string value(kValueSize, 'a');
+    for (size_t i = 0; i < kNumBlocks; i++) {
+      ASSERT_OK(Put(ToString(i), value.c_str()));
+    }
+  }
+
+  void RecordCacheCounters(const Options& options) {
+    miss_count_ = TestGetTickerCount(options, BLOCK_CACHE_MISS);
+    hit_count_ = TestGetTickerCount(options, BLOCK_CACHE_HIT);
+    insert_count_ = TestGetTickerCount(options, BLOCK_CACHE_ADD);
+    failure_count_ = TestGetTickerCount(options, BLOCK_CACHE_ADD_FAILURES);
+    compressed_miss_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS);
+    compressed_hit_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_HIT);
+    compressed_insert_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD);
+    compressed_failure_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD_FAILURES);
+  }
+
+  void CheckCacheCounters(const Options& options, size_t expected_misses,
+                          size_t expected_hits, size_t expected_inserts,
+                          size_t expected_failures) {
+    size_t new_miss_count = TestGetTickerCount(options, BLOCK_CACHE_MISS);
+    size_t new_hit_count = TestGetTickerCount(options, BLOCK_CACHE_HIT);
+    size_t new_insert_count = TestGetTickerCount(options, BLOCK_CACHE_ADD);
+    size_t new_failure_count =
+        TestGetTickerCount(options, BLOCK_CACHE_ADD_FAILURES);
+    ASSERT_EQ(miss_count_ + expected_misses, new_miss_count);
+    ASSERT_EQ(hit_count_ + expected_hits, new_hit_count);
+    ASSERT_EQ(insert_count_ + expected_inserts, new_insert_count);
+    ASSERT_EQ(failure_count_ + expected_failures, new_failure_count);
+    miss_count_ = new_miss_count;
+    hit_count_ = new_hit_count;
+    insert_count_ = new_insert_count;
+    failure_count_ = new_failure_count;
+  }
+
+  void CheckCompressedCacheCounters(const Options& options,
+                                    size_t expected_misses,
+                                    size_t expected_hits,
+                                    size_t expected_inserts,
+                                    size_t expected_failures) {
+    size_t new_miss_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS);
+    size_t new_hit_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_HIT);
+    size_t new_insert_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD);
+    size_t new_failure_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD_FAILURES);
+    ASSERT_EQ(compressed_miss_count_ + expected_misses, new_miss_count);
+    ASSERT_EQ(compressed_hit_count_ + expected_hits, new_hit_count);
+    ASSERT_EQ(compressed_insert_count_ + expected_inserts, new_insert_count);
+    ASSERT_EQ(compressed_failure_count_ + expected_failures, new_failure_count);
+    compressed_miss_count_ = new_miss_count;
+    compressed_hit_count_ = new_hit_count;
+    compressed_insert_count_ = new_insert_count;
+    compressed_failure_count_ = new_failure_count;
+  }
+};
+
+TEST_F(DBBlockCacheTest, TestWithoutCompressedBlockCache) {
+  ReadOptions read_options;
+  auto table_options = GetTableOptions();
+  auto options = GetOptions(table_options);
+  InitTable(options);
+
+  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  table_options.block_cache = cache;
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  Reopen(options);
+  RecordCacheCounters(options);
+
+  std::vector<std::unique_ptr<Iterator>> iterators(kNumBlocks - 1);
+  Iterator* iter = nullptr;
+
+  // Load blocks into cache.
+  for (size_t i = 0; i < kNumBlocks - 1; i++) {
+    iter = db_->NewIterator(read_options);
+    iter->Seek(ToString(i));
+    ASSERT_OK(iter->status());
+    CheckCacheCounters(options, 1, 0, 1, 0);
+    iterators[i].reset(iter);
+  }
+  size_t usage = cache->GetUsage();
+  ASSERT_LT(0, usage);
+  cache->SetCapacity(usage);
+  ASSERT_EQ(usage, cache->GetPinnedUsage());
+
+  // Test with strict capacity limit.
+  cache->SetStrictCapacityLimit(true);
+  iter = db_->NewIterator(read_options);
+  iter->Seek(ToString(kNumBlocks - 1));
+  ASSERT_TRUE(iter->status().IsIncomplete());
+  CheckCacheCounters(options, 1, 0, 0, 1);
+  delete iter;
+  iter = nullptr;
+
+  // Release interators and access cache again.
+  for (size_t i = 0; i < kNumBlocks - 1; i++) {
+    iterators[i].reset();
+    CheckCacheCounters(options, 0, 0, 0, 0);
+  }
+  ASSERT_EQ(0, cache->GetPinnedUsage());
+  for (size_t i = 0; i < kNumBlocks - 1; i++) {
+    iter = db_->NewIterator(read_options);
+    iter->Seek(ToString(i));
+    ASSERT_OK(iter->status());
+    CheckCacheCounters(options, 0, 1, 0, 0);
+    iterators[i].reset(iter);
+  }
+}
+
+#ifdef SNAPPY
+TEST_F(DBBlockCacheTest, TestWithCompressedBlockCache) {
+  ReadOptions read_options;
+  auto table_options = GetTableOptions();
+  auto options = GetOptions(table_options);
+  options.compression = CompressionType::kSnappyCompression;
+  InitTable(options);
+
+  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  std::shared_ptr<Cache> compressed_cache = NewLRUCache(0, 0, false);
+  table_options.block_cache = cache;
+  table_options.block_cache_compressed = compressed_cache;
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  Reopen(options);
+  RecordCacheCounters(options);
+
+  std::vector<std::unique_ptr<Iterator>> iterators(kNumBlocks - 1);
+  Iterator* iter = nullptr;
+
+  // Load blocks into cache.
+  for (size_t i = 0; i < kNumBlocks - 1; i++) {
+    iter = db_->NewIterator(read_options);
+    iter->Seek(ToString(i));
+    ASSERT_OK(iter->status());
+    CheckCacheCounters(options, 1, 0, 1, 0);
+    CheckCompressedCacheCounters(options, 1, 0, 1, 0);
+    iterators[i].reset(iter);
+  }
+  size_t usage = cache->GetUsage();
+  ASSERT_LT(0, usage);
+  ASSERT_EQ(usage, cache->GetPinnedUsage());
+  size_t compressed_usage = compressed_cache->GetUsage();
+  ASSERT_LT(0, compressed_usage);
+  // Compressed block cache cannot be pinned.
+  ASSERT_EQ(0, compressed_cache->GetPinnedUsage());
+
+  // Set strict capacity limit flag. Now block will only load into compressed
+  // block cache.
+  cache->SetCapacity(usage);
+  cache->SetStrictCapacityLimit(true);
+  ASSERT_EQ(usage, cache->GetPinnedUsage());
+  // compressed_cache->SetCapacity(compressed_usage);
+  compressed_cache->SetCapacity(0);
+  // compressed_cache->SetStrictCapacityLimit(true);
+  iter = db_->NewIterator(read_options);
+  iter->Seek(ToString(kNumBlocks - 1));
+  ASSERT_TRUE(iter->status().IsIncomplete());
+  CheckCacheCounters(options, 1, 0, 0, 1);
+  CheckCompressedCacheCounters(options, 1, 0, 1, 0);
+  delete iter;
+  iter = nullptr;
+
+  // Clear strict capacity limit flag. This time we shall hit compressed block
+  // cache.
+  cache->SetStrictCapacityLimit(false);
+  iter = db_->NewIterator(read_options);
+  iter->Seek(ToString(kNumBlocks - 1));
+  ASSERT_OK(iter->status());
+  CheckCacheCounters(options, 1, 0, 1, 0);
+  CheckCompressedCacheCounters(options, 0, 1, 0, 0);
+  delete iter;
+  iter = nullptr;
+}
+#endif
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -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) {}
@@ -141,6 +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) {}
@@ -335,11 +368,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilter) {
 // entries are deleted. The compaction should create bunch of 'DeleteFile'
 // entries in VersionEdit, but none of the 'AddFile's.
 TEST_F(DBTestCompactionFilter, CompactionFilterDeletesAll) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_filter_factory = std::make_shared<DeleteFilterFactory>();
  options.disable_auto_compactions = true;
  options.create_if_missing = true;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  // put some data
@@ -367,11 +399,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilterDeletesAll) {

 TEST_F(DBTestCompactionFilter, CompactionFilterWithValueChange) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.num_levels = 3;
    options.compaction_filter_factory =
      std::make_shared<ChangeFilterFactory>();
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Write 100K+1 keys, these are written to a few files
@@ -434,8 +465,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilterWithMergeOperator) {
  PutFixed64(&three, 3);
  PutFixed64(&four, 4);

-  Options options;
-  options = CurrentOptions(options);
+  Options options = CurrentOptions();
  options.create_if_missing = true;
  options.merge_operator = MergeOperators::CreateUInt64AddOperator();
  options.num_levels = 3;
@@ -550,7 +580,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilterContextManual) {
      iter->Next();
    }
    ASSERT_EQ(total, 700);
-    ASSERT_EQ(count, 1);
+    ASSERT_EQ(count, 2);
  }
 }
 #endif  // ROCKSDB_LITE
@@ -588,11 +618,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilterContextCfId) {
 // 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) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_filter_factory = std::make_shared<DeleteFilterFactory>();
  options.disable_auto_compactions = true;
  options.create_if_missing = true;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  // Put some data.
@@ -620,6 +649,67 @@ TEST_F(DBTestCompactionFilter, CompactionFilterSnapshot) {
  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
  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 = CurrentOptions();
+  options.compaction_filter_factory = std::make_shared<DeleteISFilterFactory>();
+  options.disable_auto_compactions = true;
+  options.create_if_missing = true;
+  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
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -10,6 +10,7 @@
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/experimental.h"
+#include "rocksdb/utilities/convenience.h"
 #include "util/sync_point.h"
 namespace rocksdb {

@@ -21,11 +22,13 @@ class DBCompactionTest : public DBTestBase {
  DBCompactionTest() : DBTestBase("/db_compaction_test") {}
 };

-class DBCompactionTestWithParam : public DBTestBase,
-                                public testing::WithParamInterface<uint32_t> {
+class DBCompactionTestWithParam
+    : public DBTestBase,
+      public testing::WithParamInterface<std::tuple<uint32_t, bool>> {
 public:
  DBCompactionTestWithParam() : DBTestBase("/db_compaction_test") {
-    max_subcompactions_ = GetParam();
+    max_subcompactions_ = std::get<0>(GetParam());
+    exclusive_manual_compaction_ = std::get<1>(GetParam());
  }

  // Required if inheriting from testing::WithParamInterface<>
@@ -33,6 +36,7 @@ class DBCompactionTestWithParam : public DBTestBase,
  static void TearDownTestCase() {}

  uint32_t max_subcompactions_;
+  bool exclusive_manual_compaction_;
 };

 namespace {
@@ -65,6 +69,32 @@ class OnFileDeletionListener : public EventListener {
  std::string expected_file_name_;
 };

+class FlushedFileCollector : public EventListener {
+ public:
+  FlushedFileCollector() {}
+  ~FlushedFileCollector() {}
+
+  virtual void OnFlushCompleted(DB* db, const FlushJobInfo& info) override {
+    std::lock_guard<std::mutex> lock(mutex_);
+    flushed_files_.push_back(info.file_path);
+  }
+
+  std::vector<std::string> GetFlushedFiles() {
+    std::lock_guard<std::mutex> lock(mutex_);
+    std::vector<std::string> result;
+    for (auto fname : flushed_files_) {
+      result.push_back(fname);
+    }
+    return result;
+  }
+
+  void ClearFlushedFiles() { flushed_files_.clear(); }
+
+ private:
+  std::vector<std::string> flushed_files_;
+  std::mutex mutex_;
+};
+
 static const int kCDTValueSize = 1000;
 static const int kCDTKeysPerBuffer = 4;
 static const int kCDTNumLevels = 8;
@@ -160,7 +190,7 @@ const SstFileMetaData* PickFileRandomly(
      auto result = rand->Uniform(file_id);
      return &(level_meta.files[result]);
    }
-    file_id -= level_meta.files.size();
+    file_id -= static_cast<uint32_t>(level_meta.files.size());
  }
  assert(false);
  return nullptr;
@@ -239,9 +269,12 @@ 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.
@@ -251,12 +284,11 @@ 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) {
-  Options options;
-  options = CurrentOptions(options);
+  Options options = CurrentOptions();
  options.env = env_;
  options.new_table_reader_for_compaction_inputs = true;
  options.max_open_files = 100;
@@ -458,13 +490,15 @@ TEST_F(DBCompactionTest, DisableStatsUpdateReopen) {


 TEST_P(DBCompactionTestWithParam, CompactionTrigger) {
-  Options options;
+  const int kNumKeysPerFile = 100;
+
+  Options options = CurrentOptions();
  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 = CurrentOptions(options);
+  options.memtable_factory.reset(new SpecialSkipListFactory(kNumKeysPerFile));
  CreateAndReopenWithCF({"pikachu"}, options);

  Random rnd(301);
@@ -473,31 +507,131 @@ 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);
  ASSERT_EQ(NumTableFilesAtLevel(1, 1), 1);
 }

+TEST_F(DBCompactionTest, BGCompactionsAllowed) {
+  // Create several column families. Make compaction triggers in all of them
+  // and see number of compactions scheduled to be less than allowed.
+  const int kNumKeysPerFile = 100;
+
+  Options options = CurrentOptions();
+  options.write_buffer_size = 110 << 10;  // 110KB
+  options.arena_block_size = 4 << 10;
+  options.num_levels = 3;
+  // Should speed up compaction when there are 4 files.
+  options.level0_file_num_compaction_trigger = 2;
+  options.level0_slowdown_writes_trigger = 20;
+  options.soft_pending_compaction_bytes_limit = 1 << 30;  // Infinitely large
+  options.base_background_compactions = 1;
+  options.max_background_compactions = 3;
+  options.memtable_factory.reset(new SpecialSkipListFactory(kNumKeysPerFile));
+
+  // Block all threads in thread pool.
+  const size_t kTotalTasks = 4;
+  env_->SetBackgroundThreads(4, Env::LOW);
+  test::SleepingBackgroundTask sleeping_tasks[kTotalTasks];
+  for (size_t i = 0; i < kTotalTasks; i++) {
+    env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                   &sleeping_tasks[i], Env::Priority::LOW);
+    sleeping_tasks[i].WaitUntilSleeping();
+  }
+
+  CreateAndReopenWithCF({"one", "two", "three"}, options);
+
+  Random rnd(301);
+  for (int cf = 0; cf < 4; cf++) {
+    for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+      for (int i = 0; i < kNumKeysPerFile; i++) {
+        ASSERT_OK(Put(cf, Key(i), ""));
+      }
+      // put extra key to trigger flush
+      ASSERT_OK(Put(cf, "", ""));
+      dbfull()->TEST_WaitForFlushMemTable(handles_[cf]);
+      ASSERT_EQ(NumTableFilesAtLevel(0, cf), num + 1);
+    }
+  }
+
+  // Now all column families qualify compaction but only one should be
+  // scheduled, because no column family hits speed up condition.
+  ASSERT_EQ(1, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+
+  // Create two more files for one column family, which triggers speed up
+  // condition, three compactions will be scheduled.
+  for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+    for (int i = 0; i < kNumKeysPerFile; i++) {
+      ASSERT_OK(Put(2, Key(i), ""));
+    }
+    // put extra key to trigger flush
+    ASSERT_OK(Put(2, "", ""));
+    dbfull()->TEST_WaitForFlushMemTable(handles_[2]);
+    ASSERT_EQ(options.level0_file_num_compaction_trigger + num + 1,
+              NumTableFilesAtLevel(0, 2));
+  }
+  ASSERT_EQ(3, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+
+  // Unblock all threads to unblock all compactions.
+  for (size_t i = 0; i < kTotalTasks; i++) {
+    sleeping_tasks[i].WakeUp();
+    sleeping_tasks[i].WaitUntilDone();
+  }
+  dbfull()->TEST_WaitForCompact();
+
+  // Verify number of compactions allowed will come back to 1.
+
+  for (size_t i = 0; i < kTotalTasks; i++) {
+    sleeping_tasks[i].Reset();
+    env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                   &sleeping_tasks[i], Env::Priority::LOW);
+    sleeping_tasks[i].WaitUntilSleeping();
+  }
+  for (int cf = 0; cf < 4; cf++) {
+    for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+      for (int i = 0; i < kNumKeysPerFile; i++) {
+        ASSERT_OK(Put(cf, Key(i), ""));
+      }
+      // put extra key to trigger flush
+      ASSERT_OK(Put(cf, "", ""));
+      dbfull()->TEST_WaitForFlushMemTable(handles_[cf]);
+      ASSERT_EQ(NumTableFilesAtLevel(0, cf), num + 1);
+    }
+  }
+
+  // Now all column families qualify compaction but only one should be
+  // scheduled, because no column family hits speed up condition.
+  ASSERT_EQ(1, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+
+  for (size_t i = 0; i < kTotalTasks; i++) {
+    sleeping_tasks[i].WakeUp();
+    sleeping_tasks[i].WaitUntilDone();
+  }
+}
+
 TEST_P(DBCompactionTestWithParam, CompactionsGenerateMultipleFiles) {
-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;        // Large write buffer
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  CreateAndReopenWithCF({"pikachu"}, options);

  Random rnd(301);
@@ -524,9 +658,8 @@ TEST_P(DBCompactionTestWithParam, CompactionsGenerateMultipleFiles) {

 TEST_F(DBCompactionTest, MinorCompactionsHappen) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.write_buffer_size = 10000;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    const int N = 500;
@@ -550,14 +683,66 @@ TEST_F(DBCompactionTest, MinorCompactionsHappen) {
  } while (ChangeCompactOptions());
 }

+TEST_F(DBCompactionTest, ZeroSeqIdCompaction) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleLevel;
+  options.level0_file_num_compaction_trigger = 3;
+
+  FlushedFileCollector* collector = new FlushedFileCollector();
+  options.listeners.emplace_back(collector);
+
+  // compaction options
+  CompactionOptions compact_opt;
+  compact_opt.compression = kNoCompression;
+  compact_opt.output_file_size_limit = 4096;
+  const size_t key_len =
+    static_cast<size_t>(compact_opt.output_file_size_limit) / 5;
+
+  DestroyAndReopen(options);
+
+  std::vector<const Snapshot*> snaps;
+
+  // create first file and flush to l0
+  for (auto& key : {"1", "2", "3", "3", "3", "3"}) {
+    Put(key, std::string(key_len, 'A'));
+    snaps.push_back(dbfull()->GetSnapshot());
+  }
+  Flush();
+  dbfull()->TEST_WaitForFlushMemTable();
+
+  // create second file and flush to l0
+  for (auto& key : {"3", "4", "5", "6", "7", "8"}) {
+    Put(key, std::string(key_len, 'A'));
+    snaps.push_back(dbfull()->GetSnapshot());
+  }
+  Flush();
+  dbfull()->TEST_WaitForFlushMemTable();
+
+  // move both files down to l1
+  dbfull()->CompactFiles(compact_opt, collector->GetFlushedFiles(), 1);
+
+  // release snap so that first instance of key(3) can have seqId=0
+  for (auto snap : snaps) {
+    dbfull()->ReleaseSnapshot(snap);
+  }
+
+  // create 3 files in l0 so to trigger compaction
+  for (int i = 0; i < options.level0_file_num_compaction_trigger; i++) {
+    Put("2", std::string(1, 'A'));
+    Flush();
+    dbfull()->TEST_WaitForFlushMemTable();
+  }
+
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_OK(Put("", ""));
+}
+
 // Check that writes done during a memtable compaction are recovered
 // if the database is shutdown during the memtable compaction.
 TEST_F(DBCompactionTest, RecoverDuringMemtableCompaction) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.env = env_;
-    options.write_buffer_size = 1000000;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Trigger a long memtable compaction and reopen the database during it
@@ -581,10 +766,9 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveOneFile) {
      [&](void* arg) { trivial_move++; });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  int32_t num_keys = 80;
@@ -607,8 +791,11 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveOneFile) {
  ASSERT_EQ(metadata.size(), 1U);
  LiveFileMetaData level0_file = metadata[0];  // L0 file meta

+  CompactRangeOptions cro;
+  cro.exclusive_manual_compaction = exclusive_manual_compaction_;
+
  // Compaction will initiate a trivial move from L0 to L1
-  dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  dbfull()->CompactRange(cro, nullptr, nullptr);

  // File moved From L0 to L1
  ASSERT_EQ(NumTableFilesAtLevel(0, 0), 0);  // 0 files in L0
@@ -660,7 +847,7 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveNonOverlappingFiles) {

  Random rnd(301);
  std::map<int32_t, std::string> values;
-  for (uint32_t i = 0; i < ranges.size(); i++) {
+  for (size_t i = 0; i < ranges.size(); i++) {
    for (int32_t j = ranges[i].first; j <= ranges[i].second; j++) {
      values[j] = RandomString(&rnd, value_size);
      ASSERT_OK(Put(Key(j), values[j]));
@@ -672,14 +859,17 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveNonOverlappingFiles) {
  ASSERT_EQ(level0_files, ranges.size());    // Multiple files in L0
  ASSERT_EQ(NumTableFilesAtLevel(1, 0), 0);  // No files in L1

+  CompactRangeOptions cro;
+  cro.exclusive_manual_compaction = exclusive_manual_compaction_;
+
  // Since data is non-overlapping we expect compaction to initiate
  // a trivial move
-  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  db_->CompactRange(cro, nullptr, nullptr);
  // We expect that all the files were trivially moved from L0 to L1
  ASSERT_EQ(NumTableFilesAtLevel(0, 0), 0);
  ASSERT_EQ(NumTableFilesAtLevel(1, 0) /* level1_files */, level0_files);

-  for (uint32_t i = 0; i < ranges.size(); i++) {
+  for (size_t i = 0; i < ranges.size(); i++) {
    for (int32_t j = ranges[i].first; j <= ranges[i].second; j++) {
      ASSERT_EQ(Get(Key(j)), values[j]);
    }
@@ -703,7 +893,7 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveNonOverlappingFiles) {
    {500, 560},  // this range overlap with the next one
    {551, 599},
  };
-  for (uint32_t i = 0; i < ranges.size(); i++) {
+  for (size_t i = 0; i < ranges.size(); i++) {
    for (int32_t j = ranges[i].first; j <= ranges[i].second; j++) {
      values[j] = RandomString(&rnd, value_size);
      ASSERT_OK(Put(Key(j), values[j]));
@@ -711,9 +901,9 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveNonOverlappingFiles) {
    ASSERT_OK(Flush());
  }

-  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  db_->CompactRange(cro, nullptr, nullptr);

-  for (uint32_t i = 0; i < ranges.size(); i++) {
+  for (size_t i = 0; i < ranges.size(); i++) {
    for (int32_t j = ranges[i].first; j <= ranges[i].second; j++) {
      ASSERT_EQ(Get(Key(j)), values[j]);
    }
@@ -767,6 +957,7 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveTargetLevel) {
  CompactRangeOptions compact_options;
  compact_options.change_level = true;
  compact_options.target_level = 6;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
  ASSERT_OK(db_->CompactRange(compact_options, nullptr, nullptr));
  // 2 files in L6
  ASSERT_EQ("0,0,0,0,0,0,2", FilesPerLevel(0));
@@ -782,6 +973,382 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveTargetLevel) {
  }
 }

+TEST_P(DBCompactionTestWithParam, ManualCompactionPartial) {
+  int32_t trivial_move = 0;
+  int32_t non_trivial_move = 0;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BackgroundCompaction:TrivialMove",
+      [&](void* arg) { trivial_move++; });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BackgroundCompaction:NonTrivial",
+      [&](void* arg) { non_trivial_move++; });
+  bool first = true;
+  bool second = true;
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBCompaction::ManualPartial:4", "DBCompaction::ManualPartial:1"},
+       {"DBCompaction::ManualPartial:2", "DBCompaction::ManualPartial:3"}});
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
+        if (first) {
+          TEST_SYNC_POINT("DBCompaction::ManualPartial:4");
+          first = false;
+          TEST_SYNC_POINT("DBCompaction::ManualPartial:3");
+        } else if (second) {
+          TEST_SYNC_POINT("DBCompaction::ManualPartial:2");
+        }
+      });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  Options options = CurrentOptions();
+  options.write_buffer_size = 10 * 1024 * 1024;
+  options.num_levels = 7;
+  options.max_subcompactions = max_subcompactions_;
+  options.level0_file_num_compaction_trigger = 3;
+  options.max_background_compactions = 3;
+  options.target_file_size_base = 1 << 23;  // 8 MB
+
+  DestroyAndReopen(options);
+  int32_t value_size = 10 * 1024;  // 10 KB
+
+  // Add 2 non-overlapping files
+  Random rnd(301);
+  std::map<int32_t, std::string> values;
+
+  // file 1 [0 => 100]
+  for (int32_t i = 0; i < 100; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // file 2 [100 => 300]
+  for (int32_t i = 100; i < 300; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // 2 files in L0
+  ASSERT_EQ("2", FilesPerLevel(0));
+  CompactRangeOptions compact_options;
+  compact_options.change_level = true;
+  compact_options.target_level = 6;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
+  // Trivial move the two non-overlapping files to level 6
+  ASSERT_OK(db_->CompactRange(compact_options, nullptr, nullptr));
+  // 2 files in L6
+  ASSERT_EQ("0,0,0,0,0,0,2", FilesPerLevel(0));
+
+  ASSERT_EQ(trivial_move, 1);
+  ASSERT_EQ(non_trivial_move, 0);
+
+  // file 3 [ 0 => 200]
+  for (int32_t i = 0; i < 200; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // 1 files in L0
+  ASSERT_EQ("1,0,0,0,0,0,2", FilesPerLevel(0));
+  ASSERT_OK(dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr, false));
+  ASSERT_OK(dbfull()->TEST_CompactRange(1, nullptr, nullptr, nullptr, false));
+  ASSERT_OK(dbfull()->TEST_CompactRange(2, nullptr, nullptr, nullptr, false));
+  ASSERT_OK(dbfull()->TEST_CompactRange(3, nullptr, nullptr, nullptr, false));
+  ASSERT_OK(dbfull()->TEST_CompactRange(4, nullptr, nullptr, nullptr, false));
+  // 2 files in L6, 1 file in L5
+  ASSERT_EQ("0,0,0,0,0,1,2", FilesPerLevel(0));
+
+  ASSERT_EQ(trivial_move, 6);
+  ASSERT_EQ(non_trivial_move, 0);
+
+  std::thread threads([&] {
+    compact_options.change_level = false;
+    compact_options.exclusive_manual_compaction = false;
+    std::string begin_string = Key(0);
+    std::string end_string = Key(199);
+    Slice begin(begin_string);
+    Slice end(end_string);
+    ASSERT_OK(db_->CompactRange(compact_options, &begin, &end));
+  });
+
+  TEST_SYNC_POINT("DBCompaction::ManualPartial:1");
+  // file 4 [300 => 400)
+  for (int32_t i = 300; i <= 400; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // file 5 [400 => 500)
+  for (int32_t i = 400; i <= 500; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // file 6 [500 => 600)
+  for (int32_t i = 500; i <= 600; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // 3 files in L0
+  ASSERT_EQ("3,0,0,0,0,1,2", FilesPerLevel(0));
+  // 1 file in L6, 1 file in L1
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ("0,1,0,0,0,0,1", FilesPerLevel(0));
+  threads.join();
+
+  for (int32_t i = 0; i < 600; i++) {
+    ASSERT_EQ(Get(Key(i)), values[i]);
+  }
+}
+
+TEST_F(DBCompactionTest, ManualPartialFill) {
+  int32_t trivial_move = 0;
+  int32_t non_trivial_move = 0;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BackgroundCompaction:TrivialMove",
+      [&](void* arg) { trivial_move++; });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BackgroundCompaction:NonTrivial",
+      [&](void* arg) { non_trivial_move++; });
+  bool first = true;
+  bool second = true;
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBCompaction::PartialFill:4", "DBCompaction::PartialFill:1"},
+       {"DBCompaction::PartialFill:2", "DBCompaction::PartialFill:3"}});
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
+        if (first) {
+          TEST_SYNC_POINT("DBCompaction::PartialFill:4");
+          first = false;
+          TEST_SYNC_POINT("DBCompaction::PartialFill:3");
+        } else if (second) {
+        }
+      });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  Options options = CurrentOptions();
+  options.write_buffer_size = 10 * 1024 * 1024;
+  options.max_bytes_for_level_multiplier = 2;
+  options.num_levels = 4;
+  options.level0_file_num_compaction_trigger = 3;
+  options.max_background_compactions = 3;
+
+  DestroyAndReopen(options);
+  int32_t value_size = 10 * 1024;  // 10 KB
+
+  // Add 2 non-overlapping files
+  Random rnd(301);
+  std::map<int32_t, std::string> values;
+
+  // file 1 [0 => 100]
+  for (int32_t i = 0; i < 100; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // file 2 [100 => 300]
+  for (int32_t i = 100; i < 300; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // 2 files in L0
+  ASSERT_EQ("2", FilesPerLevel(0));
+  CompactRangeOptions compact_options;
+  compact_options.change_level = true;
+  compact_options.target_level = 2;
+  ASSERT_OK(db_->CompactRange(compact_options, nullptr, nullptr));
+  // 2 files in L2
+  ASSERT_EQ("0,0,2", FilesPerLevel(0));
+
+  ASSERT_EQ(trivial_move, 1);
+  ASSERT_EQ(non_trivial_move, 0);
+
+  // file 3 [ 0 => 200]
+  for (int32_t i = 0; i < 200; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // 2 files in L2, 1 in L0
+  ASSERT_EQ("1,0,2", FilesPerLevel(0));
+  ASSERT_OK(dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr, false));
+  // 2 files in L2, 1 in L1
+  ASSERT_EQ("0,1,2", FilesPerLevel(0));
+
+  ASSERT_EQ(trivial_move, 2);
+  ASSERT_EQ(non_trivial_move, 0);
+
+  std::thread threads([&] {
+    compact_options.change_level = false;
+    compact_options.exclusive_manual_compaction = false;
+    std::string begin_string = Key(0);
+    std::string end_string = Key(199);
+    Slice begin(begin_string);
+    Slice end(end_string);
+    ASSERT_OK(db_->CompactRange(compact_options, &begin, &end));
+  });
+
+  TEST_SYNC_POINT("DBCompaction::PartialFill:1");
+  // Many files 4 [300 => 4300)
+  for (int32_t i = 0; i <= 5; i++) {
+    for (int32_t j = 300; j < 4300; j++) {
+      if (j == 2300) {
+        ASSERT_OK(Flush());
+        dbfull()->TEST_WaitForFlushMemTable();
+      }
+      values[j] = RandomString(&rnd, value_size);
+      ASSERT_OK(Put(Key(j), values[j]));
+    }
+  }
+
+  // Verify level sizes
+  uint64_t target_size = 4 * options.max_bytes_for_level_base;
+  for (int32_t i = 1; i < options.num_levels; i++) {
+    ASSERT_LE(SizeAtLevel(i), target_size);
+    target_size *= options.max_bytes_for_level_multiplier;
+  }
+
+  TEST_SYNC_POINT("DBCompaction::PartialFill:2");
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_WaitForCompact();
+  threads.join();
+
+  for (int32_t i = 0; i < 4300; i++) {
+    ASSERT_EQ(Get(Key(i)), values[i]);
+  }
+}
+
+TEST_F(DBCompactionTest, DeleteFileRange) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 10 * 1024 * 1024;
+  options.max_bytes_for_level_multiplier = 2;
+  options.num_levels = 4;
+  options.level0_file_num_compaction_trigger = 3;
+  options.max_background_compactions = 3;
+
+  DestroyAndReopen(options);
+  int32_t value_size = 10 * 1024;  // 10 KB
+
+  // Add 2 non-overlapping files
+  Random rnd(301);
+  std::map<int32_t, std::string> values;
+
+  // file 1 [0 => 100]
+  for (int32_t i = 0; i < 100; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // file 2 [100 => 300]
+  for (int32_t i = 100; i < 300; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // 2 files in L0
+  ASSERT_EQ("2", FilesPerLevel(0));
+  CompactRangeOptions compact_options;
+  compact_options.change_level = true;
+  compact_options.target_level = 2;
+  ASSERT_OK(db_->CompactRange(compact_options, nullptr, nullptr));
+  // 2 files in L2
+  ASSERT_EQ("0,0,2", FilesPerLevel(0));
+
+  // file 3 [ 0 => 200]
+  for (int32_t i = 0; i < 200; i++) {
+    values[i] = RandomString(&rnd, value_size);
+    ASSERT_OK(Put(Key(i), values[i]));
+  }
+  ASSERT_OK(Flush());
+
+  // Many files 4 [300 => 4300)
+  for (int32_t i = 0; i <= 5; i++) {
+    for (int32_t j = 300; j < 4300; j++) {
+      if (j == 2300) {
+        ASSERT_OK(Flush());
+        dbfull()->TEST_WaitForFlushMemTable();
+      }
+      values[j] = RandomString(&rnd, value_size);
+      ASSERT_OK(Put(Key(j), values[j]));
+    }
+  }
+  ASSERT_OK(Flush());
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_WaitForCompact();
+
+  // Verify level sizes
+  uint64_t target_size = 4 * options.max_bytes_for_level_base;
+  for (int32_t i = 1; i < options.num_levels; i++) {
+    ASSERT_LE(SizeAtLevel(i), target_size);
+    target_size *= options.max_bytes_for_level_multiplier;
+  }
+
+  size_t old_num_files = CountFiles();
+  std::string begin_string = Key(1000);
+  std::string end_string = Key(2000);
+  Slice begin(begin_string);
+  Slice end(end_string);
+  ASSERT_OK(DeleteFilesInRange(db_, db_->DefaultColumnFamily(), &begin, &end));
+
+  int32_t deleted_count = 0;
+  for (int32_t i = 0; i < 4300; i++) {
+    if (i < 1000 || i > 2000) {
+      ASSERT_EQ(Get(Key(i)), values[i]);
+    } else {
+      ReadOptions roptions;
+      std::string result;
+      Status s = db_->Get(roptions, Key(i), &result);
+      ASSERT_TRUE(s.IsNotFound() || s.ok());
+      if (s.IsNotFound()) {
+        deleted_count++;
+      }
+    }
+  }
+  ASSERT_GT(deleted_count, 0);
+  begin_string = Key(5000);
+  end_string = Key(6000);
+  Slice begin1(begin_string);
+  Slice end1(end_string);
+  // Try deleting files in range which contain no keys
+  ASSERT_OK(
+      DeleteFilesInRange(db_, db_->DefaultColumnFamily(), &begin1, &end1));
+
+  // Push data from level 0 to level 1 to force all data to be deleted
+  // Note that we don't delete level 0 files
+  compact_options.change_level = true;
+  compact_options.target_level = 1;
+  ASSERT_OK(dbfull()->TEST_CompactRange(0, nullptr, nullptr));
+
+  ASSERT_OK(
+      DeleteFilesInRange(db_, db_->DefaultColumnFamily(), nullptr, nullptr));
+
+  int32_t deleted_count2 = 0;
+  for (int32_t i = 0; i < 4300; i++) {
+    ReadOptions roptions;
+    std::string result;
+    Status s = db_->Get(roptions, Key(i), &result);
+    ASSERT_TRUE(s.IsNotFound());
+    deleted_count2++;
+  }
+  ASSERT_GT(deleted_count2, deleted_count);
+  size_t new_num_files = CountFiles();
+  ASSERT_GT(old_num_files, new_num_files);
+}
+
 TEST_P(DBCompactionTestWithParam, TrivialMoveToLastLevelWithFiles) {
  int32_t trivial_move = 0;
  int32_t non_trivial_move = 0;
@@ -793,10 +1360,9 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveToLastLevelWithFiles) {
      [&](void* arg) { non_trivial_move++; });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  int32_t value_size = 10 * 1024;  // 10 KB
@@ -815,6 +1381,7 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveToLastLevelWithFiles) {
  CompactRangeOptions compact_options;
  compact_options.change_level = true;
  compact_options.target_level = 3;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
  ASSERT_OK(db_->CompactRange(compact_options, nullptr, nullptr));
  ASSERT_EQ("0,0,0,1", FilesPerLevel(0));
  ASSERT_EQ(trivial_move, 1);
@@ -828,8 +1395,10 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveToLastLevelWithFiles) {
  ASSERT_OK(Flush());

  ASSERT_EQ("1,0,0,1", FilesPerLevel(0));
+  CompactRangeOptions cro;
+  cro.exclusive_manual_compaction = exclusive_manual_compaction_;
  // Compaction will do L0=>L1 L1=>L2 L2=>L3 (3 trivial moves)
-  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
  ASSERT_EQ("0,0,0,2", FilesPerLevel(0));
  ASSERT_EQ(trivial_move, 4);
  ASSERT_EQ(non_trivial_move, 0);
@@ -846,6 +1415,8 @@ 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;
@@ -961,6 +1532,8 @@ 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;
@@ -1078,7 +1651,7 @@ TEST_P(DBCompactionTestWithParam, ConvertCompactionStyle) {
  int max_key_universal_insert = 600;

  // Stage 1: generate a db with level compaction
-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 110 << 10;  // 110KB
  options.arena_block_size = 4 << 10;
  options.num_levels = 4;
@@ -1088,7 +1661,6 @@ TEST_P(DBCompactionTestWithParam, ConvertCompactionStyle) {
  options.target_file_size_base = 200 << 10;  // 200KB
  options.target_file_size_multiplier = 1;
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  CreateAndReopenWithCF({"pikachu"}, options);

  for (int i = 0; i <= max_key_level_insert; i++) {
@@ -1129,6 +1701,7 @@ TEST_P(DBCompactionTestWithParam, ConvertCompactionStyle) {
  compact_options.target_level = 0;
  compact_options.bottommost_level_compaction =
      BottommostLevelCompaction::kForce;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
  dbfull()->CompactRange(compact_options, handles_[1], nullptr, nullptr);

  // Only 1 file in L0
@@ -1262,7 +1835,9 @@ TEST_P(DBCompactionTestWithParam, ManualCompaction) {

    uint64_t prev_block_cache_add =
        options.statistics->getTickerCount(BLOCK_CACHE_ADD);
-    db_->CompactRange(CompactRangeOptions(), handles_[1], nullptr, nullptr);
+    CompactRangeOptions cro;
+    cro.exclusive_manual_compaction = exclusive_manual_compaction_;
+    db_->CompactRange(cro, handles_[1], nullptr, nullptr);
    // Verify manual compaction doesn't fill block cache
    ASSERT_EQ(prev_block_cache_add,
              options.statistics->getTickerCount(BLOCK_CACHE_ADD));
@@ -1341,6 +1916,7 @@ TEST_P(DBCompactionTestWithParam, ManualLevelCompactionOutputPathId) {
    ASSERT_EQ(1, GetSstFileCount(options.db_paths[0].path));
    CompactRangeOptions compact_options;
    compact_options.target_path_id = 1;
+    compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
    db_->CompactRange(compact_options, handles_[1], nullptr, nullptr);

    ASSERT_EQ("0,1", FilesPerLevel(1));
@@ -1409,7 +1985,7 @@ TEST_P(DBCompactionTestWithParam, DISABLED_CompactFilesOnLevelCompaction) {
    std::set<std::string> overlapping_file_names;
    std::vector<std::string> compaction_input_file_names;
    for (int f = 0; f < file_picked; ++f) {
-      int level;
+      int level = 0;
      auto file_meta = PickFileRandomly(cf_meta, &rnd, &level);
      compaction_input_file_names.push_back(file_meta->name);
      GetOverlappingFileNumbersForLevelCompaction(
@@ -1601,6 +2177,8 @@ 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;
@@ -1707,6 +2285,25 @@ TEST_P(DBCompactionTestWithParam, CompressLevelCompaction) {
  Destroy(options);
 }

+TEST_F(DBCompactionTest, SanitizeCompactionOptionsTest) {
+  Options options = CurrentOptions();
+  options.max_background_compactions = 5;
+  options.soft_pending_compaction_bytes_limit = 0;
+  options.hard_pending_compaction_bytes_limit = 100;
+  options.create_if_missing = true;
+  DestroyAndReopen(options);
+  ASSERT_EQ(5, db_->GetOptions().base_background_compactions);
+  ASSERT_EQ(100, db_->GetOptions().soft_pending_compaction_bytes_limit);
+
+  options.base_background_compactions = 4;
+  options.max_background_compactions = 3;
+  options.soft_pending_compaction_bytes_limit = 200;
+  options.hard_pending_compaction_bytes_limit = 150;
+  DestroyAndReopen(options);
+  ASSERT_EQ(3, db_->GetOptions().base_background_compactions);
+  ASSERT_EQ(150, db_->GetOptions().soft_pending_compaction_bytes_limit);
+}
+
 // This tests for a bug that could cause two level0 compactions running
 // concurrently
 // TODO(aekmekji): Make sure that the reason this fails when run with
@@ -1778,10 +2375,9 @@ TEST_P(DBCompactionTestWithParam, ForceBottommostLevelCompaction) {
      [&](void* arg) { non_trivial_move++; });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  int32_t value_size = 10 * 1024;  // 10 KB
@@ -1851,7 +2447,61 @@ TEST_P(DBCompactionTestWithParam, ForceBottommostLevelCompaction) {
 }

 INSTANTIATE_TEST_CASE_P(DBCompactionTestWithParam, DBCompactionTestWithParam,
-                        ::testing::Values(1, 4));
+                        ::testing::Values(std::make_tuple(1, true),
+                                          std::make_tuple(1, false),
+                                          std::make_tuple(4, true),
+                                          std::make_tuple(4, false)));
+
+class CompactionPriTest : public DBTestBase,
+                          public testing::WithParamInterface<uint32_t> {
+ public:
+  CompactionPriTest() : DBTestBase("/compaction_pri_test") {
+    compaction_pri_ = GetParam();
+  }
+
+  // Required if inheriting from testing::WithParamInterface<>
+  static void SetUpTestCase() {}
+  static void TearDownTestCase() {}
+
+  uint32_t compaction_pri_;
+};
+
+TEST_P(CompactionPriTest, Test) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 16 * 1024;
+  options.compaction_pri = static_cast<CompactionPri>(compaction_pri_);
+  options.hard_pending_compaction_bytes_limit = 256 * 1024;
+  options.max_bytes_for_level_base = 64 * 1024;
+  options.max_bytes_for_level_multiplier = 4;
+  options.compression = kNoCompression;
+
+  DestroyAndReopen(options);
+
+  Random rnd(301);
+  const int kNKeys = 5000;
+  int keys[kNKeys];
+  for (int i = 0; i < kNKeys; i++) {
+    keys[i] = i;
+  }
+  std::random_shuffle(std::begin(keys), std::end(keys));
+
+  for (int i = 0; i < kNKeys; i++) {
+    ASSERT_OK(Put(Key(keys[i]), RandomString(&rnd, 102)));
+  }
+
+  dbfull()->TEST_WaitForCompact();
+  for (int i = 0; i < kNKeys; i++) {
+    ASSERT_NE("NOT_FOUND", Get(Key(i)));
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(
+    CompactionPriTest, CompactionPriTest,
+    ::testing::Values(CompactionPri::kByCompensatedSize,
+                      CompactionPri::kOldestLargestSeqFirst,
+                      CompactionPri::kOldestSmallestSeqFirst,
+                      CompactionPri::kMinOverlappingRatio));
+
 #endif // !defined(ROCKSDB_LITE)
 }  // namespace rocksdb

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -144,6 +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,
@@ -185,6 +188,8 @@ class DBImpl : public DB {
      const TransactionLogIterator::ReadOptions&
          read_options = TransactionLogIterator::ReadOptions()) override;
  virtual Status DeleteFile(std::string name) override;
+  Status DeleteFilesInRange(ColumnFamilyHandle* column_family,
+                            const Slice* begin, const Slice* end);

  virtual void GetLiveFilesMetaData(
      std::vector<LiveFileMetaData>* metadata) override;
@@ -223,13 +228,30 @@ class DBImpl : public DB {
                                                   bool include_history);

  // For a given key, check to see if there are any records for this key
-  // in the memtables, including memtable history.
-
-  // On success, *seq will contain the sequence number for the
-  // latest such change or kMaxSequenceNumber if no records were present.
-  // Returns OK on success, other status on error reading memtables.
-  Status GetLatestSequenceForKeyFromMemtable(SuperVersion* sv, const Slice& key,
-                                             SequenceNumber* seq);
+  // in the memtables, including memtable history.  If cache_only is false,
+  // SST files will also be checked.
+  //
+  // If a key is found, *found_record_for_key will be set to true and
+  // *seq will will be set to the stored sequence number for the latest
+  // operation on this key or kMaxSequenceNumber if unknown.
+  // If no key is found, *found_record_for_key will be set to false.
+  //
+  // Note: If cache_only=false, it is possible for *seq to be set to 0 if
+  // the sequence number has been cleared from the record.  If the caller is
+  // holding an active db snapshot, we know the missing sequence must be less
+  // than the snapshot's sequence number (sequence numbers are only cleared
+  // when there are no earlier active snapshots).
+  //
+  // If NotFound is returned and found_record_for_key is set to false, then no
+  // record for this key was found.  If the caller is holding an active db
+  // snapshot, we know that no key could have existing after this snapshot
+  // (since we do not compact keys that have an earlier snapshot).
+  //
+  // Returns OK or NotFound on success,
+  // other status on unexpected error.
+  Status GetLatestSequenceForKey(SuperVersion* sv, const Slice& key,
+                                 bool cache_only, SequenceNumber* seq,
+                                 bool* found_record_for_key);

  using DB::AddFile;
  virtual Status AddFile(ColumnFamilyHandle* column_family,
@@ -240,6 +262,12 @@ class DBImpl : public DB {

 #endif  // ROCKSDB_LITE

+  // Similar to GetSnapshot(), but also lets the db know that this snapshot
+  // will be used for transaction write-conflict checking.  The DB can then
+  // make sure not to compact any keys that would prevent a write-conflict from
+  // being detected.
+  const Snapshot* GetSnapshotForWriteConflictBoundary();
+
  // checks if all live files exist on file system and that their file sizes
  // match to our in-memory records
  virtual Status CheckConsistency();
@@ -249,6 +277,7 @@ class DBImpl : public DB {
  Status RunManualCompaction(ColumnFamilyData* cfd, int input_level,
                             int output_level, uint32_t output_path_id,
                             const Slice* begin, const Slice* end,
+                             bool exclusive,
                             bool disallow_trivial_move = false);

  // Return an internal iterator over the current state of the database.
@@ -314,8 +343,14 @@ class DBImpl : public DB {

  Cache* TEST_table_cache() { return table_cache_.get(); }

+  WriteController& TEST_write_controler() { return write_controller_; }
+
 #endif  // NDEBUG

+  // Return maximum background compaction alowed to be scheduled based on
+  // compaction status.
+  int BGCompactionsAllowed() const;
+
  // Returns the list of live files in 'live' and the list
  // of all files in the filesystem in 'candidate_files'.
  // If force == false and the last call was less than
@@ -398,6 +433,17 @@ class DBImpl : public DB {
                                        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, TableProperties prop);
@@ -448,10 +494,11 @@ class DBImpl : public DB {

  // Background process needs to call
  //     auto x = CaptureCurrentFileNumberInPendingOutputs()
+  //     auto file_num = versions_->NewFileNumber();
  //     <do something>
  //     ReleaseFileNumberFromPendingOutputs(x)
-  // This will protect any temporary files created while <do something> is
-  // executing from being deleted.
+  // This will protect any file with number `file_num` or greater from being
+  // deleted while <do something> is running.
  // -----------
  // This function will capture current file number and append it to
  // pending_outputs_. This will prevent any background process to delete any
@@ -476,7 +523,7 @@ class DBImpl : public DB {
                         SequenceNumber* max_sequence, bool read_only);

  // The following two methods are used to flush a memtable to
-  // storage. The first one is used atdatabase RecoveryTime (when the
+  // storage. The first one is used at database RecoveryTime (when the
  // database is opened) and is heavyweight because it holds the mutex
  // for the entire period. The second method WriteLevel0Table supports
  // concurrent flush memtables to storage.
@@ -513,12 +560,13 @@ class DBImpl : public DB {
  void MaybeScheduleFlushOrCompaction();
  void SchedulePendingFlush(ColumnFamilyData* cfd);
  void SchedulePendingCompaction(ColumnFamilyData* cfd);
-  static void BGWorkCompaction(void* db);
+  static void BGWorkCompaction(void* arg);
  static void BGWorkFlush(void* db);
-  void BackgroundCallCompaction();
+  static void UnscheduleCallback(void* arg);
+  void BackgroundCallCompaction(void* arg);
  void BackgroundCallFlush();
  Status BackgroundCompaction(bool* madeProgress, JobContext* job_context,
-                              LogBuffer* log_buffer);
+                              LogBuffer* log_buffer, void* m = 0);
  Status BackgroundFlush(bool* madeProgress, JobContext* job_context,
                         LogBuffer* log_buffer);

@@ -546,18 +594,25 @@ class DBImpl : public DB {
  // helper function to call after some of the logs_ were synced
  void MarkLogsSynced(uint64_t up_to, bool synced_dir, const Status& status);

+  const Snapshot* GetSnapshotImpl(bool is_write_conflict_boundary);
+
  // table_cache_ provides its own synchronization
  std::shared_ptr<Cache> table_cache_;

  // 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
-  // * if bg_manual_only_ > 0, whenever a compaction finishes, even if it hasn't
+  // * if AnyManualCompaction, whenever a compaction finishes, even if it hasn't
  // made any progress
  // * whenever a compaction made any progress
  // * whenever bg_flush_scheduled_ value decreases (i.e. whenever a flush is
@@ -715,11 +770,6 @@ class DBImpl : public DB {
  // 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.
-  int bg_manual_only_;
-
  // number of background memtable flush jobs, submitted to the HIGH pool
  int bg_flush_scheduled_;

@@ -732,15 +782,25 @@ class DBImpl : public DB {
    int input_level;
    int output_level;
    uint32_t output_path_id;
-    bool done;
    Status status;
+    bool done;
    bool in_progress;             // compaction request being processed?
+    bool incomplete;              // only part of requested range compacted
+    bool exclusive;               // current behavior of only one manual
+    bool disallow_trivial_move;   // Force actual compaction to run
    const InternalKey* begin;     // nullptr means beginning of key range
    const InternalKey* end;       // nullptr means end of key range
+    InternalKey* manual_end;      // how far we are compacting
    InternalKey tmp_storage;      // Used to keep track of compaction progress
-    bool disallow_trivial_move;   // Force actual compaction to run
+    InternalKey tmp_storage1;     // Used to keep track of compaction progress
+    Compaction* compaction;
+  };
+  std::deque<ManualCompaction*> manual_compaction_dequeue_;
+
+  struct CompactionArg {
+    DBImpl* db;
+    ManualCompaction* m;
  };
-  ManualCompaction* manual_compaction_;

  // Have we encountered a background error in paranoid mode?
  Status bg_error_;
@@ -763,7 +823,10 @@ class DBImpl : public DB {
  // they're unique
  std::atomic<int> next_job_id_;

-  bool flush_on_destroy_; // Used when disableWAL is true.
+  // A flag indicating whether the current rocksdb database has any
+  // data that is not yet persisted into either WAL or SST file.
+  // Used when disableWAL is true.
+  bool has_unpersisted_data_;

  static const int KEEP_LOG_FILE_NUM = 1000;
  // MSVC version 1800 still does not have constexpr for ::max()
@@ -781,9 +844,12 @@ class DBImpl : public DB {
  // Unified interface for logging events
  EventLogger event_logger_;

-  // A value of >0 temporarily disables scheduling of background work
+  // A value of > 0 temporarily disables scheduling of background work
  int bg_work_paused_;

+  // A value of > 0 temporarily disables scheduling of background compaction
+  int bg_compaction_paused_;
+
  // Guard against multiple concurrent refitting
  bool refitting_level_;

@@ -834,9 +900,16 @@ class DBImpl : public DB {
                 bool* value_found = nullptr);

  bool GetIntPropertyInternal(ColumnFamilyData* cfd,
-                              DBPropertyType property_type,
-                              bool need_out_of_mutex, bool is_locked,
-                              uint64_t* value);
+                              const DBPropertyInfo& property_info,
+                              bool is_locked, uint64_t* value);
+
+  bool HasPendingManualCompaction();
+  bool HasExclusiveManualCompaction();
+  void AddManualCompaction(ManualCompaction* m);
+  void RemoveManualCompaction(ManualCompaction* m);
+  bool ShouldntRunManualCompaction(ManualCompaction* m);
+  bool HaveManualCompaction(ColumnFamilyData* cfd);
+  bool MCOverlap(ManualCompaction* m, ManualCompaction* m1);
 };

 // Sanitize db options.  The caller should delete result.info_log if
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -70,7 +70,7 @@ Status DBImpl::TEST_CompactRange(int level, const Slice* begin,
       cfd->ioptions()->compaction_style == kCompactionStyleFIFO)
          ? level
          : level + 1;
-  return RunManualCompaction(cfd, level, output_level, 0, begin, end,
+  return RunManualCompaction(cfd, level, output_level, 0, begin, end, true,
                             disallow_trivial_move);
 }

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -54,10 +54,11 @@ Iterator* DBImplReadOnly::NewIterator(const ReadOptions& read_options,
  auto db_iter = NewArenaWrappedDbIterator(
      env_, *cfd->ioptions(), cfd->user_comparator(),
      (read_options.snapshot != nullptr
-           ? reinterpret_cast<const SnapshotImpl*>(
-                read_options.snapshot)->number_
+           ? reinterpret_cast<const SnapshotImpl*>(read_options.snapshot)
+                 ->number_
           : latest_snapshot),
-      super_version->mutable_cf_options.max_sequential_skip_in_iterations);
+      super_version->mutable_cf_options.max_sequential_skip_in_iterations,
+      super_version->version_number);
  auto internal_iter = NewInternalIterator(
      read_options, cfd, super_version, db_iter->GetArena());
  db_iter->SetIterUnderDBIter(internal_iter);
@@ -81,10 +82,11 @@ Status DBImplReadOnly::NewIterators(
    auto* db_iter = NewArenaWrappedDbIterator(
        env_, *cfd->ioptions(), cfd->user_comparator(),
        (read_options.snapshot != nullptr
-            ? reinterpret_cast<const SnapshotImpl*>(
-                  read_options.snapshot)->number_
-            : latest_snapshot),
-        sv->mutable_cf_options.max_sequential_skip_in_iterations);
+             ? reinterpret_cast<const SnapshotImpl*>(read_options.snapshot)
+                   ->number_
+             : latest_snapshot),
+        sv->mutable_cf_options.max_sequential_skip_in_iterations,
+        sv->version_number);
    auto* internal_iter = NewInternalIterator(
        read_options, cfd, sv, db_iter->GetArena());
    db_iter->SetIterUnderDBIter(internal_iter);
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -7,6 +7,8 @@
 #define __STDC_FORMAT_MACROS
 #endif

+#include "db/db_info_dumper.h"
+
 #include <inttypes.h>
 #include <stdio.h>
 #include <string>
@@ -16,7 +18,6 @@
 #include "db/filename.h"
 #include "rocksdb/options.h"
 #include "rocksdb/env.h"
-#include "util/db_info_dumper.h"

 namespace rocksdb {

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -18,12 +18,11 @@ class DBTestInPlaceUpdate : public DBTestBase {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdate) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;
    options.env = env_;
    options.write_buffer_size = 100000;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller size
@@ -41,12 +40,11 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdate) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateLargeNewValue) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;
    options.env = env_;
    options.write_buffer_size = 100000;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of larger size
@@ -64,7 +62,7 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateLargeNewValue) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerSize) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;

@@ -72,7 +70,6 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerSize) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceSmallerSize;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller size
@@ -92,7 +89,7 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerSize) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerVarintSize) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;

@@ -100,7 +97,6 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerVarintSize) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceSmallerVarintSize;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller varint size
@@ -120,7 +116,7 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerVarintSize) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackLargeNewValue) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;

@@ -128,7 +124,6 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackLargeNewValue) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceLargerSize;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of larger size
@@ -146,7 +141,7 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackLargeNewValue) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackNoAction) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;

@@ -154,7 +149,6 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackNoAction) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceNoAction;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Callback function requests no actions from db
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -13,18 +13,20 @@
 #include <string>
 #include <limits>

-#include "db/filename.h"
 #include "db/dbformat.h"
+#include "db/filename.h"
+#include "db/merge_context.h"
 #include "port/port.h"
 #include "rocksdb/env.h"
-#include "rocksdb/options.h"
 #include "rocksdb/iterator.h"
 #include "rocksdb/merge_operator.h"
+#include "rocksdb/options.h"
 #include "table/internal_iterator.h"
 #include "util/arena.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
 #include "util/perf_context_imp.h"
+#include "util/string_util.h"

 namespace rocksdb {

@@ -59,9 +61,47 @@ class DBIter: public Iterator {
    kReverse
  };

+  // LocalStatistics contain Statistics counters that will be aggregated per
+  // each iterator instance and then will be sent to the global statistics when
+  // the iterator is destroyed.
+  //
+  // The purpose of this approach is to avoid perf regression happening
+  // when multiple threads bump the atomic counters from a DBIter::Next().
+  struct LocalStatistics {
+    explicit LocalStatistics() { ResetCounters(); }
+
+    void ResetCounters() {
+      next_count_ = 0;
+      next_found_count_ = 0;
+      prev_count_ = 0;
+      prev_found_count_ = 0;
+      bytes_read_ = 0;
+    }
+
+    void BumpGlobalStatistics(Statistics* global_statistics) {
+      RecordTick(global_statistics, NUMBER_DB_NEXT, next_count_);
+      RecordTick(global_statistics, NUMBER_DB_NEXT_FOUND, next_found_count_);
+      RecordTick(global_statistics, NUMBER_DB_PREV, prev_count_);
+      RecordTick(global_statistics, NUMBER_DB_PREV_FOUND, prev_found_count_);
+      RecordTick(global_statistics, ITER_BYTES_READ, bytes_read_);
+      ResetCounters();
+    }
+
+    // Map to Tickers::NUMBER_DB_NEXT
+    uint64_t next_count_;
+    // Map to Tickers::NUMBER_DB_NEXT_FOUND
+    uint64_t next_found_count_;
+    // Map to Tickers::NUMBER_DB_PREV
+    uint64_t prev_count_;
+    // Map to Tickers::NUMBER_DB_PREV_FOUND
+    uint64_t prev_found_count_;
+    // Map to Tickers::ITER_BYTES_READ
+    uint64_t bytes_read_;
+  };
+
  DBIter(Env* env, const ImmutableCFOptions& ioptions, const Comparator* cmp,
         InternalIterator* iter, SequenceNumber s, bool arena_mode,
-         uint64_t max_sequential_skip_in_iterations,
+         uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
         const Slice* iterate_upper_bound = nullptr,
         bool prefix_same_as_start = false)
      : arena_mode_(arena_mode),
@@ -75,14 +115,17 @@ class DBIter: public Iterator {
        valid_(false),
        current_entry_is_merged_(false),
        statistics_(ioptions.statistics),
+        version_number_(version_number),
        iterate_upper_bound_(iterate_upper_bound),
-        prefix_same_as_start_(prefix_same_as_start) {
+        prefix_same_as_start_(prefix_same_as_start),
+        iter_pinned_(false) {
    RecordTick(statistics_, NO_ITERATORS);
    prefix_extractor_ = ioptions.prefix_extractor;
    max_skip_ = max_sequential_skip_in_iterations;
  }
  virtual ~DBIter() {
    RecordTick(statistics_, NO_ITERATORS, -1);
+    local_stats_.BumpGlobalStatistics(statistics_);
    if (!arena_mode_) {
      delete iter_;
    } else {
@@ -92,6 +135,9 @@ class DBIter: public Iterator {
  virtual void SetIter(InternalIterator* iter) {
    assert(iter_ == nullptr);
    iter_ = iter;
+    if (iter_ && iter_pinned_) {
+      iter_->PinData();
+    }
  }
  virtual bool Valid() const override { return valid_; }
  virtual Slice key() const override {
@@ -110,6 +156,50 @@ class DBIter: public Iterator {
      return status_;
    }
  }
+  virtual Status PinData() {
+    Status s;
+    if (iter_) {
+      s = iter_->PinData();
+    }
+    if (s.ok()) {
+      // Even if iter_ is nullptr, we set iter_pinned_ to true so that when
+      // iter_ is updated using SetIter, we Pin it.
+      iter_pinned_ = true;
+    }
+    return s;
+  }
+  virtual Status ReleasePinnedData() {
+    Status s;
+    if (iter_) {
+      s = iter_->ReleasePinnedData();
+    }
+    if (s.ok()) {
+      iter_pinned_ = false;
+    }
+    return s;
+  }
+
+  virtual Status GetProperty(std::string prop_name,
+                             std::string* prop) override {
+    if (prop == nullptr) {
+      return Status::InvalidArgument("prop is nullptr");
+    }
+    if (prop_name == "rocksdb.iterator.super-version-number") {
+      // First try to pass the value returned from inner iterator.
+      if (!iter_->GetProperty(prop_name, prop).ok()) {
+        *prop = ToString(version_number_);
+      }
+      return Status::OK();
+    } else if (prop_name == "rocksdb.iterator.is-key-pinned") {
+      if (valid_) {
+        *prop = (iter_pinned_ && saved_key_.IsKeyPinned()) ? "1" : "0";
+      } else {
+        *prop = "Iterator is not valid.";
+      }
+      return Status::OK();
+    }
+    return Status::InvalidArgument("Undentified property.");
+  }

  virtual void Next() override;
  virtual void Prev() override;
@@ -156,9 +246,14 @@ class DBIter: public Iterator {
  bool current_entry_is_merged_;
  Statistics* statistics_;
  uint64_t max_skip_;
+  uint64_t version_number_;
  const Slice* iterate_upper_bound_;
  IterKey prefix_start_;
  bool prefix_same_as_start_;
+  bool iter_pinned_;
+  // List of operands for merge operator.
+  MergeContext merge_context_;
+  LocalStatistics local_stats_;

  // No copying allowed
  DBIter(const DBIter&);
@@ -186,26 +281,35 @@ 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_
+  if (statistics_ != nullptr) {
+    local_stats_.next_count_++;
+  }
+  // Now we point to the next internal position, for both of merge and
+  // not merge cases.
  if (!iter_->Valid()) {
    valid_ = false;
    return;
  }
  FindNextUserEntry(true /* skipping the current user key */);
-  if (statistics_ != nullptr) {
-    RecordTick(statistics_, NUMBER_DB_NEXT);
-    if (valid_) {
-      RecordTick(statistics_, NUMBER_DB_NEXT_FOUND);
-      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;
  }
+  if (statistics_ != nullptr && valid_) {
+    local_stats_.next_found_count_++;
+    local_stats_.bytes_read_ += (key().size() + value().size());
+  }
 }

 // PRE: saved_key_ has the current user key if skipping
@@ -233,7 +337,7 @@ void DBIter::FindNextUserEntryInternal(bool skipping) {

    if (ParseKey(&ikey)) {
      if (iterate_upper_bound_ != nullptr &&
-          ikey.user_key.compare(*iterate_upper_bound_) >= 0) {
+          user_comparator_->Compare(ikey.user_key, *iterate_upper_bound_) >= 0) {
        break;
      }

@@ -248,18 +352,21 @@ void DBIter::FindNextUserEntryInternal(bool skipping) {
            case kTypeSingleDeletion:
              // Arrange to skip all upcoming entries for this key since
              // they are hidden by this deletion.
-              saved_key_.SetKey(ikey.user_key);
+              saved_key_.SetKey(ikey.user_key,
+                                !iter_->IsKeyPinned() /* copy */);
              skipping = true;
              num_skipped = 0;
              PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
              break;
            case kTypeValue:
              valid_ = true;
-              saved_key_.SetKey(ikey.user_key);
+              saved_key_.SetKey(ikey.user_key,
+                                !iter_->IsKeyPinned() /* copy */);
              return;
            case kTypeMerge:
              // By now, we are sure the current ikey is going to yield a value
-              saved_key_.SetKey(ikey.user_key);
+              saved_key_.SetKey(ikey.user_key,
+                                !iter_->IsKeyPinned() /* copy */);
              current_entry_is_merged_ = true;
              valid_ = true;
              MergeValuesNewToOld();  // Go to a different state machine
@@ -305,9 +412,9 @@ void DBIter::MergeValuesNewToOld() {
    return;
  }

+  merge_context_.Clear();
  // Start the merge process by pushing the first operand
-  std::deque<std::string> operands;
-  operands.push_front(iter_->value().ToString());
+  merge_context_.PushOperand(iter_->value());

  ParsedInternalKey ikey;
  for (iter_->Next(); iter_->Valid(); iter_->Next()) {
@@ -332,7 +439,8 @@ void DBIter::MergeValuesNewToOld() {
      {
        StopWatchNano timer(env_, statistics_ != nullptr);
        PERF_TIMER_GUARD(merge_operator_time_nanos);
-        user_merge_operator_->FullMerge(ikey.user_key, &val, operands,
+        user_merge_operator_->FullMerge(ikey.user_key, &val,
+                                        merge_context_.GetOperands(),
                                        &saved_value_, logger_);
        RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME,
                   timer.ElapsedNanos());
@@ -344,7 +452,7 @@ void DBIter::MergeValuesNewToOld() {
      // hit a merge, add the value as an operand and run associative merge.
      // when complete, add result to operands and continue.
      const Slice& val = iter_->value();
-      operands.push_front(val.ToString());
+      merge_context_.PushOperand(val);
    } else {
      assert(false);
    }
@@ -357,8 +465,9 @@ void DBIter::MergeValuesNewToOld() {
    // a deletion marker.
    // feed null as the existing value to the merge operator, such that
    // client can differentiate this scenario and do things accordingly.
-    user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr, operands,
-                                    &saved_value_, logger_);
+    user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr,
+                                    merge_context_.GetOperands(), &saved_value_,
+                                    logger_);
    RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME, timer.ElapsedNanos());
  }
 }
@@ -370,10 +479,10 @@ void DBIter::Prev() {
  }
  PrevInternal();
  if (statistics_ != nullptr) {
-    RecordTick(statistics_, NUMBER_DB_PREV);
+    local_stats_.prev_count_++;
    if (valid_) {
-      RecordTick(statistics_, NUMBER_DB_PREV_FOUND);
-      RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
+      local_stats_.prev_found_count_++;
+      local_stats_.bytes_read_ += (key().size() + value().size());
    }
  }
  if (valid_ && prefix_extractor_ && prefix_same_as_start_ &&
@@ -419,7 +528,8 @@ void DBIter::PrevInternal() {
  ParsedInternalKey ikey;

  while (iter_->Valid()) {
-    saved_key_.SetKey(ExtractUserKey(iter_->key()));
+    saved_key_.SetKey(ExtractUserKey(iter_->key()),
+                      !iter_->IsKeyPinned() /* copy */);
    if (FindValueForCurrentKey()) {
      valid_ = true;
      if (!iter_->Valid()) {
@@ -449,8 +559,7 @@ void DBIter::PrevInternal() {
 // saved_value_
 bool DBIter::FindValueForCurrentKey() {
  assert(iter_->Valid());
-  // Contains operands for merge operator.
-  std::deque<std::string> operands;
+  merge_context_.Clear();
  // last entry before merge (could be kTypeDeletion, kTypeSingleDeletion or
  // kTypeValue)
  ValueType last_not_merge_type = kTypeDeletion;
@@ -470,19 +579,19 @@ bool DBIter::FindValueForCurrentKey() {
    last_key_entry_type = ikey.type;
    switch (last_key_entry_type) {
      case kTypeValue:
-        operands.clear();
+        merge_context_.Clear();
        saved_value_ = iter_->value().ToString();
        last_not_merge_type = kTypeValue;
        break;
      case kTypeDeletion:
      case kTypeSingleDeletion:
-        operands.clear();
+        merge_context_.Clear();
        last_not_merge_type = last_key_entry_type;
        PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
        break;
      case kTypeMerge:
        assert(user_merge_operator_ != nullptr);
-        operands.push_back(iter_->value().ToString());
+        merge_context_.PushOperandBack(iter_->value());
        break;
      default:
        assert(false);
@@ -504,7 +613,8 @@ bool DBIter::FindValueForCurrentKey() {
      if (last_not_merge_type == kTypeDeletion) {
        StopWatchNano timer(env_, statistics_ != nullptr);
        PERF_TIMER_GUARD(merge_operator_time_nanos);
-        user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr, operands,
+        user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr,
+                                        merge_context_.GetOperands(),
                                        &saved_value_, logger_);
        RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME,
                   timer.ElapsedNanos());
@@ -516,7 +626,8 @@ bool DBIter::FindValueForCurrentKey() {
          StopWatchNano timer(env_, statistics_ != nullptr);
          PERF_TIMER_GUARD(merge_operator_time_nanos);
          user_merge_operator_->FullMerge(saved_key_.GetKey(), &temp_slice,
-                                          operands, &saved_value_, logger_);
+                                          merge_context_.GetOperands(),
+                                          &saved_value_, logger_);
          RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME,
                     timer.ElapsedNanos());
        }
@@ -559,11 +670,11 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {

  // kTypeMerge. We need to collect all kTypeMerge values and save them
  // in operands
-  std::deque<std::string> operands;
+  merge_context_.Clear();
  while (iter_->Valid() &&
         user_comparator_->Equal(ikey.user_key, saved_key_.GetKey()) &&
         ikey.type == kTypeMerge) {
-    operands.push_front(iter_->value().ToString());
+    merge_context_.PushOperand(iter_->value());
    iter_->Next();
    FindParseableKey(&ikey, kForward);
  }
@@ -574,7 +685,8 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
    {
      StopWatchNano timer(env_, statistics_ != nullptr);
      PERF_TIMER_GUARD(merge_operator_time_nanos);
-      user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr, operands,
+      user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr,
+                                      merge_context_.GetOperands(),
                                      &saved_value_, logger_);
      RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME, timer.ElapsedNanos());
    }
@@ -592,8 +704,9 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
  {
    StopWatchNano timer(env_, statistics_ != nullptr);
    PERF_TIMER_GUARD(merge_operator_time_nanos);
-    user_merge_operator_->FullMerge(saved_key_.GetKey(), &val, operands,
-                                    &saved_value_, logger_);
+    user_merge_operator_->FullMerge(saved_key_.GetKey(), &val,
+                                    merge_context_.GetOperands(), &saved_value_,
+                                    logger_);
    RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME, timer.ElapsedNanos());
  }
  valid_ = true;
@@ -735,7 +848,7 @@ void DBIter::SeekToLast() {
  // it will seek to the last key before the
  // ReadOptions.iterate_upper_bound
  if (iter_->Valid() && iterate_upper_bound_ != nullptr) {
-    saved_key_.SetKey(*iterate_upper_bound_);
+    saved_key_.SetKey(*iterate_upper_bound_, false /* copy */);
    std::string last_key;
    AppendInternalKey(&last_key,
                      ParsedInternalKey(saved_key_.GetKey(), kMaxSequenceNumber,
@@ -771,11 +884,17 @@ Iterator* NewDBIterator(Env* env, const ImmutableCFOptions& ioptions,
                        InternalIterator* internal_iter,
                        const SequenceNumber& sequence,
                        uint64_t max_sequential_skip_in_iterations,
+                        uint64_t version_number,
                        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, prefix_same_as_start);
+                        bool prefix_same_as_start, bool pin_data) {
+  DBIter* db_iter =
+      new DBIter(env, ioptions, user_key_comparator, internal_iter, sequence,
+                 false, max_sequential_skip_in_iterations, version_number,
+                 iterate_upper_bound, prefix_same_as_start);
+  if (pin_data) {
+    db_iter->PinData();
+  }
+  return db_iter;
 }

 ArenaWrappedDBIter::~ArenaWrappedDBIter() { db_iter_->~DBIter(); }
@@ -797,6 +916,14 @@ inline void ArenaWrappedDBIter::Prev() { db_iter_->Prev(); }
 inline Slice ArenaWrappedDBIter::key() const { return db_iter_->key(); }
 inline Slice ArenaWrappedDBIter::value() const { return db_iter_->value(); }
 inline Status ArenaWrappedDBIter::status() const { return db_iter_->status(); }
+inline Status ArenaWrappedDBIter::PinData() { return db_iter_->PinData(); }
+inline Status ArenaWrappedDBIter::GetProperty(std::string prop_name,
+                                              std::string* prop) {
+  return db_iter_->GetProperty(prop_name, prop);
+}
+inline Status ArenaWrappedDBIter::ReleasePinnedData() {
+  return db_iter_->ReleasePinnedData();
+}
 void ArenaWrappedDBIter::RegisterCleanup(CleanupFunction function, void* arg1,
                                         void* arg2) {
  db_iter_->RegisterCleanup(function, arg1, arg2);
@@ -805,17 +932,21 @@ void ArenaWrappedDBIter::RegisterCleanup(CleanupFunction function, void* arg1,
 ArenaWrappedDBIter* NewArenaWrappedDbIterator(
    Env* env, const ImmutableCFOptions& ioptions,
    const Comparator* user_key_comparator, const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations,
-    const Slice* iterate_upper_bound, bool prefix_same_as_start) {
+    uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
+    const Slice* iterate_upper_bound, bool prefix_same_as_start,
+    bool pin_data) {
  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,
+                       true, max_sequential_skip_in_iterations, version_number,
                       iterate_upper_bound, prefix_same_as_start);

  iter->SetDBIter(db_iter);
+  if (pin_data) {
+    iter->PinData();
+  }

  return iter;
 }
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -9,6 +9,7 @@

 #pragma once
 #include <stdint.h>
+#include <string>
 #include "rocksdb/db.h"
 #include "rocksdb/iterator.h"
 #include "db/dbformat.h"
@@ -24,13 +25,12 @@ 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,
-                               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);
+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,
+    uint64_t version_number, const Slice* iterate_upper_bound = nullptr,
+    bool prefix_same_as_start = false, bool pin_data = 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
@@ -63,6 +63,9 @@ class ArenaWrappedDBIter : public Iterator {
  virtual Status status() const override;

  void RegisterCleanup(CleanupFunction function, void* arg1, void* arg2);
+  virtual Status PinData();
+  virtual Status ReleasePinnedData();
+  virtual Status GetProperty(std::string prop_name, std::string* prop) override;

 private:
  DBIter* db_iter_;
@@ -73,8 +76,8 @@ class ArenaWrappedDBIter : public Iterator {
 extern ArenaWrappedDBIter* NewArenaWrappedDbIterator(
    Env* env, const ImmutableCFOptions& options,
    const Comparator* user_key_comparator, const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations,
+    uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
    const Slice* iterate_upper_bound = nullptr,
-    bool prefix_same_as_start = false);
+    bool prefix_same_as_start = false, bool pin_data = false);

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -181,10 +181,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->AddPut("b", "val_b");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 10,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        10, options.max_sequential_skip_in_iterations, 0));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -215,7 +214,7 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations));
+        10, options.max_sequential_skip_in_iterations, 0));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -241,7 +240,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -273,7 +273,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -308,7 +309,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
@@ -337,7 +339,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        7, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        7, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    SetPerfLevel(kEnableCount);
    ASSERT_TRUE(GetPerfLevel() == kEnableCount);
@@ -374,7 +377,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        4, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        4, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -399,7 +403,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
@@ -421,7 +426,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -456,7 +462,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        7, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        7, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    SetPerfLevel(kEnableCount);
    ASSERT_TRUE(GetPerfLevel() == kEnableCount);
@@ -482,10 +489,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->AddPut("b", "val_b");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 10,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        10, options.max_sequential_skip_in_iterations, 0));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -524,10 +530,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->AddPut("b", "val_b");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 2,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        2, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -555,10 +560,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->AddPut("c", "val_c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 10,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        10, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -583,10 +587,9 @@ TEST_F(DBIteratorTest, DBIteratorEmpty) {
    TestIterator* internal_iter = new TestIterator(BytewiseComparator());
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 0,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
  }
@@ -595,10 +598,9 @@ TEST_F(DBIteratorTest, DBIteratorEmpty) {
    TestIterator* internal_iter = new TestIterator(BytewiseComparator());
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 0,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToFirst();
    ASSERT_TRUE(!db_iter->Valid());
  }
@@ -617,10 +619,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkipCountSkips) {
  }
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(
-      NewDBIterator(env_, ImmutableCFOptions(options),
-                    BytewiseComparator(), internal_iter, 2,
-                    options.max_sequential_skip_in_iterations));
+  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 2,
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -659,9 +660,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {

      options.statistics = rocksdb::CreateDBStatistics();
      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, i + 2,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i + 2, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -695,9 +695,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, i + 2,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i + 2, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -724,9 +723,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, 202,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, 202, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -756,10 +754,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      }
      internal_iter->AddPut("c", "200");
      internal_iter->Finish();
-      std::unique_ptr<Iterator> db_iter(
-          NewDBIterator(env_, ImmutableCFOptions(options),
-                        BytewiseComparator(), internal_iter, i,
-                        options.max_sequential_skip_in_iterations));
+      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(!db_iter->Valid());

@@ -773,10 +770,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
    }
    internal_iter->AddPut("c", "200");
    internal_iter->Finish();
-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 200,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        200, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -809,9 +805,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, i + 2,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i + 2, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -844,9 +839,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, i + 2,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i + 2, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -887,7 +881,7 @@ TEST_F(DBIteratorTest, DBIterator1) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 1,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -913,7 +907,7 @@ TEST_F(DBIteratorTest, DBIterator2) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 0,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -936,7 +930,7 @@ TEST_F(DBIteratorTest, DBIterator3) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 2,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -958,7 +952,7 @@ TEST_F(DBIteratorTest, DBIterator4) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 4,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -987,7 +981,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        0, options.max_sequential_skip_in_iterations));
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1009,7 +1003,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        1, options.max_sequential_skip_in_iterations));
+        1, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1031,7 +1025,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        2, options.max_sequential_skip_in_iterations));
+        2, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1053,7 +1047,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        3, options.max_sequential_skip_in_iterations));
+        3, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1075,7 +1069,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        4, options.max_sequential_skip_in_iterations));
+        4, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1097,7 +1091,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        5, options.max_sequential_skip_in_iterations));
+        5, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1119,7 +1113,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        6, options.max_sequential_skip_in_iterations));
+        6, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1145,7 +1139,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        0, options.max_sequential_skip_in_iterations));
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1167,7 +1161,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        1, options.max_sequential_skip_in_iterations));
+        1, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1189,7 +1183,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        2, options.max_sequential_skip_in_iterations));
+        2, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1211,7 +1205,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        3, options.max_sequential_skip_in_iterations));
+        3, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
  }
@@ -1229,7 +1223,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        4, options.max_sequential_skip_in_iterations));
+        4, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1251,7 +1245,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        5, options.max_sequential_skip_in_iterations));
+        5, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1273,7 +1267,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        6, options.max_sequential_skip_in_iterations));
+        6, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1311,7 +1305,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        0, options.max_sequential_skip_in_iterations));
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1345,7 +1339,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        2, options.max_sequential_skip_in_iterations));
+        2, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1385,7 +1379,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        4, options.max_sequential_skip_in_iterations));
+        4, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1425,7 +1419,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        5, options.max_sequential_skip_in_iterations));
+        5, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1470,7 +1464,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        6, options.max_sequential_skip_in_iterations));
+        6, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1516,7 +1510,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        7, options.max_sequential_skip_in_iterations));
+        7, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1556,7 +1550,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        9, options.max_sequential_skip_in_iterations));
+        9, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1602,7 +1596,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        13, options.max_sequential_skip_in_iterations));
+        13, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1649,7 +1643,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        14, options.max_sequential_skip_in_iterations));
+        14, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1678,7 +1672,7 @@ TEST_F(DBIteratorTest, DBIterator8) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-      10, options.max_sequential_skip_in_iterations));
+      10, options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -1707,7 +1701,7 @@ TEST_F(DBIteratorTest, DBIterator9) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations));
+        10, options.max_sequential_skip_in_iterations, 0));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -1754,7 +1748,7 @@ TEST_F(DBIteratorTest, DBIterator10) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-      10, options.max_sequential_skip_in_iterations));
+      10, options.max_sequential_skip_in_iterations, 0));

  db_iter->Seek("c");
  ASSERT_TRUE(db_iter->Valid());
@@ -1778,9 +1772,9 @@ TEST_F(DBIteratorTest, SeekToLastOccurrenceSeq0) {
  internal_iter->AddPut("b", "2");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
-      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-      10, 0 /* force seek */));
+  std::unique_ptr<Iterator> db_iter(
+      NewDBIterator(env_, ImmutableCFOptions(options), BytewiseComparator(),
+                    internal_iter, 10, 0 /* force seek */, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1807,7 +1801,7 @@ TEST_F(DBIteratorTest, DBIterator11) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 1,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1832,7 +1826,7 @@ TEST_F(DBIteratorTest, DBIterator12) {

  std::unique_ptr<Iterator> db_iter(
      NewDBIterator(env_, ImmutableCFOptions(options), BytewiseComparator(),
-                    internal_iter, 10, 0));
+                    internal_iter, 10, 0, 0));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -1874,7 +1868,7 @@ class DBIterWithMergeIterTest : public testing::Test {
    db_iter_.reset(NewDBIterator(env_, ImmutableCFOptions(options_),
                                 BytewiseComparator(), merge_iter,
                                 8 /* read data earlier than seqId 8 */,
-                                 3 /* max iterators before reseek */));
+                                 3 /* max iterators before reseek */, 0));
  }

  Env* env_;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -27,7 +27,7 @@ class DBTestXactLogIterator : public DBTestBase {
    Status status = dbfull()->GetUpdatesSince(seq, &iter);
    EXPECT_OK(status);
    EXPECT_TRUE(iter->Valid());
-    return std::move(iter);
+    return iter;
  }
 };

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -92,6 +92,10 @@ 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(
@@ -101,10 +105,10 @@ DBTablePropertiesTest::TestGetPropertiesOfTablesInRange(
  // only which fall within requested ranges
  std::vector<LiveFileMetaData> vmd;
  db_->GetLiveFilesMetaData(&vmd);
-  for (auto md : vmd) {
+  for (auto& md : vmd) {
    std::string fn = md.db_path + md.name;
    bool in_range = false;
-    for (auto r : ranges) {
+    for (auto& r : ranges) {
      // smallestkey < limit && largestkey >= start
      if (r.limit.compare(md.smallestkey) >= 0 &&
          r.start.compare(md.largestkey) <= 0) {
@@ -187,11 +191,20 @@ TEST_F(DBTablePropertiesTest, GetPropertiesOfTablesInRange) {
  for (int j = 0; j < 100; j++) {
    // create a bunch of ranges
    std::vector<std::string> random_keys;
-    auto n = 2 * rnd.Uniform(50);
+    // 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()) {
@@ -202,6 +215,34 @@ TEST_F(DBTablePropertiesTest, GetPropertiesOfTablesInRange) {
    TestGetPropertiesOfTablesInRange(std::move(ranges));
  }
 }
+
+TEST_F(DBTablePropertiesTest, GetColumnFamilyNameProperty) {
+  std::string kExtraCfName = "pikachu";
+  CreateAndReopenWithCF({kExtraCfName}, CurrentOptions());
+
+  // Create one table per CF, then verify it was created with the column family
+  // name property.
+  for (int cf = 0; cf < 2; ++cf) {
+    Put(cf, "key", "val");
+    Flush(cf);
+
+    TablePropertiesCollection fname_to_props;
+    ASSERT_OK(db_->GetPropertiesOfAllTables(handles_[cf], &fname_to_props));
+    ASSERT_EQ(1U, fname_to_props.size());
+
+    std::string expected_cf_name;
+    if (cf > 0) {
+      expected_cf_name = kExtraCfName;
+    } else {
+      expected_cf_name = kDefaultColumnFamilyName;
+    }
+    ASSERT_EQ(expected_cf_name,
+              fname_to_props.begin()->second->column_family_name);
+    ASSERT_EQ(cf, static_cast<uint32_t>(
+                      fname_to_props.begin()->second->column_family_id));
+  }
+}
+
 }  // namespace rocksdb

 #endif  // ROCKSDB_LITE
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -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) {
@@ -160,7 +168,7 @@ TEST_F(DBTestTailingIterator, TailingIteratorTrimSeekToNext) {
    char buf3[32];
    char buf4[32];
    snprintf(buf1, sizeof(buf1), "00a0%016d", i * 5);
-    snprintf(buf3, sizeof(buf1), "00b0%016d", i * 5);
+    snprintf(buf3, sizeof(buf3), "00b0%016d", i * 5);

    Slice key(buf1, 20);
    ASSERT_OK(Put(1, key, value));
@@ -173,7 +181,7 @@ TEST_F(DBTestTailingIterator, TailingIteratorTrimSeekToNext) {
      if (i == 299) {
        file_iters_deleted = true;
      }
-      snprintf(buf4, sizeof(buf1), "00a0%016d", i * 5 / 2);
+      snprintf(buf4, sizeof(buf4), "00a0%016d", i * 5 / 2);
      Slice target(buf4, 20);
      iterh->Seek(target);
      ASSERT_TRUE(iter->Valid());
@@ -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;
@@ -644,6 +654,51 @@ TEST_F(DBTestTailingIterator, ManagedTailingIteratorSeekToSame) {
  ASSERT_EQ(found, iter->key().ToString());
 }

+TEST_F(DBTestTailingIterator, ForwardIteratorVersionProperty) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 1000;
+
+  ReadOptions read_options;
+  read_options.tailing = true;
+
+  Put("foo", "bar");
+
+  uint64_t v1, v2, v3, v4;
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    iter->Seek("foo");
+    std::string prop_value;
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                &prop_value));
+    v1 = static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+    Put("foo1", "bar1");
+    Flush();
+
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                &prop_value));
+    v2 = static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+    iter->Seek("f");
+
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                &prop_value));
+    v3 = static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+    ASSERT_EQ(v1, v2);
+    ASSERT_GT(v3, v2);
+  }
+
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    iter->Seek("foo");
+    std::string prop_value;
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                &prop_value));
+    v4 = static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+  }
+  ASSERT_EQ(v3, v4);
+}
 }  // namespace rocksdb

 #endif  // !defined(ROCKSDB_LITE)
@@ -0,0 +1,864 @@
+//  Copyright (c) 2011-present, 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 <cstdlib>
+#include "db/db_test_util.h"
+#include "port/stack_trace.h"
+#include "rocksdb/wal_filter.h"
+
+namespace rocksdb {
+
+static uint64_t TestGetTickerCount(const Options& options,
+                                   Tickers ticker_type) {
+  return options.statistics->getTickerCount(ticker_type);
+}
+
+class DBTest2 : public DBTestBase {
+ public:
+  DBTest2() : DBTestBase("/db_test2") {}
+};
+
+class PrefixFullBloomWithReverseComparator
+    : public DBTestBase,
+      public ::testing::WithParamInterface<bool> {
+ public:
+  PrefixFullBloomWithReverseComparator()
+      : DBTestBase("/prefix_bloom_reverse") {}
+  virtual void SetUp() override { if_cache_filter_ = GetParam(); }
+  bool if_cache_filter_;
+};
+
+TEST_P(PrefixFullBloomWithReverseComparator,
+       PrefixFullBloomWithReverseComparator) {
+  Options options = last_options_;
+  options.comparator = ReverseBytewiseComparator();
+  options.prefix_extractor.reset(NewCappedPrefixTransform(3));
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions bbto;
+  if (if_cache_filter_) {
+    bbto.no_block_cache = false;
+    bbto.cache_index_and_filter_blocks = true;
+    bbto.block_cache = NewLRUCache(1);
+  }
+  bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  bbto.whole_key_filtering = false;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  DestroyAndReopen(options);
+
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "bar123", "foo"));
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "bar234", "foo2"));
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "foo123", "foo3"));
+
+  dbfull()->Flush(FlushOptions());
+
+  if (bbto.block_cache) {
+    bbto.block_cache->EraseUnRefEntries();
+  }
+
+  unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions()));
+  iter->Seek("bar345");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ("bar234", iter->key().ToString());
+  ASSERT_EQ("foo2", iter->value().ToString());
+  iter->Next();
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ("bar123", iter->key().ToString());
+  ASSERT_EQ("foo", iter->value().ToString());
+
+  iter->Seek("foo234");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ("foo123", iter->key().ToString());
+  ASSERT_EQ("foo3", iter->value().ToString());
+
+  iter->Seek("bar");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(!iter->Valid());
+}
+
+INSTANTIATE_TEST_CASE_P(PrefixFullBloomWithReverseComparator,
+                        PrefixFullBloomWithReverseComparator, testing::Bool());
+
+TEST_F(DBTest2, IteratorPropertyVersionNumber) {
+  Put("", "");
+  Iterator* iter1 = db_->NewIterator(ReadOptions());
+  std::string prop_value;
+  ASSERT_OK(
+      iter1->GetProperty("rocksdb.iterator.super-version-number", &prop_value));
+  uint64_t version_number1 =
+      static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+  Put("", "");
+  Flush();
+
+  Iterator* iter2 = db_->NewIterator(ReadOptions());
+  ASSERT_OK(
+      iter2->GetProperty("rocksdb.iterator.super-version-number", &prop_value));
+  uint64_t version_number2 =
+      static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+  ASSERT_GT(version_number2, version_number1);
+
+  Put("", "");
+
+  Iterator* iter3 = db_->NewIterator(ReadOptions());
+  ASSERT_OK(
+      iter3->GetProperty("rocksdb.iterator.super-version-number", &prop_value));
+  uint64_t version_number3 =
+      static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+  ASSERT_EQ(version_number2, version_number3);
+
+  iter1->SeekToFirst();
+  ASSERT_OK(
+      iter1->GetProperty("rocksdb.iterator.super-version-number", &prop_value));
+  uint64_t version_number1_new =
+      static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+  ASSERT_EQ(version_number1, version_number1_new);
+
+  delete iter1;
+  delete iter2;
+  delete iter3;
+}
+
+TEST_F(DBTest2, CacheIndexAndFilterWithDBRestart) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions table_options;
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  Put(1, "a", "begin");
+  Put(1, "z", "end");
+  ASSERT_OK(Flush(1));
+  TryReopenWithColumnFamilies({"default", "pikachu"}, options);
+
+  std::string value;
+  value = Get(1, "a");
+}
+
+#ifndef ROCKSDB_LITE
+namespace {
+  void ValidateKeyExistence(DB* db, const std::vector<Slice>& keys_must_exist,
+    const std::vector<Slice>& keys_must_not_exist) {
+    // Ensure that expected keys exist
+    std::vector<std::string> values;
+    if (keys_must_exist.size() > 0) {
+      std::vector<Status> status_list =
+        db->MultiGet(ReadOptions(), keys_must_exist, &values);
+      for (size_t i = 0; i < keys_must_exist.size(); i++) {
+        ASSERT_OK(status_list[i]);
+      }
+    }
+
+    // Ensure that given keys don't exist
+    if (keys_must_not_exist.size() > 0) {
+      std::vector<Status> status_list =
+        db->MultiGet(ReadOptions(), keys_must_not_exist, &values);
+      for (size_t i = 0; i < keys_must_not_exist.size(); i++) {
+        ASSERT_TRUE(status_list[i].IsNotFound());
+      }
+    }
+  }
+
+}  // namespace
+
+TEST_F(DBTest2, WalFilterTest) {
+  class TestWalFilter : public WalFilter {
+  private:
+    // Processing option that is requested to be applied at the given index
+    WalFilter::WalProcessingOption wal_processing_option_;
+    // Index at which to apply wal_processing_option_
+    // At other indexes default wal_processing_option::kContinueProcessing is
+    // returned.
+    size_t apply_option_at_record_index_;
+    // Current record index, incremented with each record encountered.
+    size_t current_record_index_;
+
+  public:
+    TestWalFilter(WalFilter::WalProcessingOption wal_processing_option,
+      size_t apply_option_for_record_index)
+      : wal_processing_option_(wal_processing_option),
+      apply_option_at_record_index_(apply_option_for_record_index),
+      current_record_index_(0) {}
+
+    virtual WalProcessingOption LogRecord(const WriteBatch& batch,
+      WriteBatch* new_batch,
+      bool* batch_changed) const override {
+      WalFilter::WalProcessingOption option_to_return;
+
+      if (current_record_index_ == apply_option_at_record_index_) {
+        option_to_return = wal_processing_option_;
+      }
+      else {
+        option_to_return = WalProcessingOption::kContinueProcessing;
+      }
+
+      // Filter is passed as a const object for RocksDB to not modify the
+      // object, however we modify it for our own purpose here and hence
+      // cast the constness away.
+      (const_cast<TestWalFilter*>(this)->current_record_index_)++;
+
+      return option_to_return;
+    }
+
+    virtual const char* Name() const override { return "TestWalFilter"; }
+  };
+
+  // Create 3 batches with two keys each
+  std::vector<std::vector<std::string>> batch_keys(3);
+
+  batch_keys[0].push_back("key1");
+  batch_keys[0].push_back("key2");
+  batch_keys[1].push_back("key3");
+  batch_keys[1].push_back("key4");
+  batch_keys[2].push_back("key5");
+  batch_keys[2].push_back("key6");
+
+  // Test with all WAL processing options
+  for (int option = 0;
+    option < static_cast<int>(
+    WalFilter::WalProcessingOption::kWalProcessingOptionMax);
+  option++) {
+    Options options = OptionsForLogIterTest();
+    DestroyAndReopen(options);
+    CreateAndReopenWithCF({ "pikachu" }, options);
+
+    // Write given keys in given batches
+    for (size_t i = 0; i < batch_keys.size(); i++) {
+      WriteBatch batch;
+      for (size_t j = 0; j < batch_keys[i].size(); j++) {
+        batch.Put(handles_[0], batch_keys[i][j], DummyString(1024));
+      }
+      dbfull()->Write(WriteOptions(), &batch);
+    }
+
+    WalFilter::WalProcessingOption wal_processing_option =
+      static_cast<WalFilter::WalProcessingOption>(option);
+
+    // Create a test filter that would apply wal_processing_option at the first
+    // record
+    size_t apply_option_for_record_index = 1;
+    TestWalFilter test_wal_filter(wal_processing_option,
+      apply_option_for_record_index);
+
+    // Reopen database with option to use WAL filter
+    options = OptionsForLogIterTest();
+    options.wal_filter = &test_wal_filter;
+    Status status =
+      TryReopenWithColumnFamilies({ "default", "pikachu" }, options);
+    if (wal_processing_option ==
+      WalFilter::WalProcessingOption::kCorruptedRecord) {
+      assert(!status.ok());
+      // In case of corruption we can turn off paranoid_checks to reopen
+      // databse
+      options.paranoid_checks = false;
+      ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+    }
+    else {
+      assert(status.ok());
+    }
+
+    // Compute which keys we expect to be found
+    // and which we expect not to be found after recovery.
+    std::vector<Slice> keys_must_exist;
+    std::vector<Slice> keys_must_not_exist;
+    switch (wal_processing_option) {
+    case WalFilter::WalProcessingOption::kCorruptedRecord:
+    case WalFilter::WalProcessingOption::kContinueProcessing: {
+      fprintf(stderr, "Testing with complete WAL processing\n");
+      // we expect all records to be processed
+      for (size_t i = 0; i < batch_keys.size(); i++) {
+        for (size_t j = 0; j < batch_keys[i].size(); j++) {
+          keys_must_exist.push_back(Slice(batch_keys[i][j]));
+        }
+      }
+      break;
+    }
+    case WalFilter::WalProcessingOption::kIgnoreCurrentRecord: {
+      fprintf(stderr,
+        "Testing with ignoring record %" ROCKSDB_PRIszt " only\n",
+        apply_option_for_record_index);
+      // We expect the record with apply_option_for_record_index to be not
+      // found.
+      for (size_t i = 0; i < batch_keys.size(); i++) {
+        for (size_t j = 0; j < batch_keys[i].size(); j++) {
+          if (i == apply_option_for_record_index) {
+            keys_must_not_exist.push_back(Slice(batch_keys[i][j]));
+          }
+          else {
+            keys_must_exist.push_back(Slice(batch_keys[i][j]));
+          }
+        }
+      }
+      break;
+    }
+    case WalFilter::WalProcessingOption::kStopReplay: {
+      fprintf(stderr,
+        "Testing with stopping replay from record %" ROCKSDB_PRIszt
+        "\n",
+        apply_option_for_record_index);
+      // We expect records beyond apply_option_for_record_index to be not
+      // found.
+      for (size_t i = 0; i < batch_keys.size(); i++) {
+        for (size_t j = 0; j < batch_keys[i].size(); j++) {
+          if (i >= apply_option_for_record_index) {
+            keys_must_not_exist.push_back(Slice(batch_keys[i][j]));
+          }
+          else {
+            keys_must_exist.push_back(Slice(batch_keys[i][j]));
+          }
+        }
+      }
+      break;
+    }
+    default:
+      assert(false);  // unhandled case
+    }
+
+    bool checked_after_reopen = false;
+
+    while (true) {
+      // Ensure that expected keys exists
+      // and not expected keys don't exist after recovery
+      ValidateKeyExistence(db_, keys_must_exist, keys_must_not_exist);
+
+      if (checked_after_reopen) {
+        break;
+      }
+
+      // reopen database again to make sure previous log(s) are not used
+      //(even if they were skipped)
+      // reopn database with option to use WAL filter
+      options = OptionsForLogIterTest();
+      ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+
+      checked_after_reopen = true;
+    }
+  }
+}
+
+TEST_F(DBTest2, WalFilterTestWithChangeBatch) {
+  class ChangeBatchHandler : public WriteBatch::Handler {
+  private:
+    // Batch to insert keys in
+    WriteBatch* new_write_batch_;
+    // Number of keys to add in the new batch
+    size_t num_keys_to_add_in_new_batch_;
+    // Number of keys added to new batch
+    size_t num_keys_added_;
+
+  public:
+    ChangeBatchHandler(WriteBatch* new_write_batch,
+      size_t num_keys_to_add_in_new_batch)
+      : new_write_batch_(new_write_batch),
+      num_keys_to_add_in_new_batch_(num_keys_to_add_in_new_batch),
+      num_keys_added_(0) {}
+    virtual void Put(const Slice& key, const Slice& value) override {
+      if (num_keys_added_ < num_keys_to_add_in_new_batch_) {
+        new_write_batch_->Put(key, value);
+        ++num_keys_added_;
+      }
+    }
+  };
+
+  class TestWalFilterWithChangeBatch : public WalFilter {
+  private:
+    // Index at which to start changing records
+    size_t change_records_from_index_;
+    // Number of keys to add in the new batch
+    size_t num_keys_to_add_in_new_batch_;
+    // Current record index, incremented with each record encountered.
+    size_t current_record_index_;
+
+  public:
+    TestWalFilterWithChangeBatch(size_t change_records_from_index,
+      size_t num_keys_to_add_in_new_batch)
+      : change_records_from_index_(change_records_from_index),
+      num_keys_to_add_in_new_batch_(num_keys_to_add_in_new_batch),
+      current_record_index_(0) {}
+
+    virtual WalProcessingOption LogRecord(const WriteBatch& batch,
+      WriteBatch* new_batch,
+      bool* batch_changed) const override {
+      if (current_record_index_ >= change_records_from_index_) {
+        ChangeBatchHandler handler(new_batch, num_keys_to_add_in_new_batch_);
+        batch.Iterate(&handler);
+        *batch_changed = true;
+      }
+
+      // Filter is passed as a const object for RocksDB to not modify the
+      // object, however we modify it for our own purpose here and hence
+      // cast the constness away.
+      (const_cast<TestWalFilterWithChangeBatch*>(this)
+        ->current_record_index_)++;
+
+      return WalProcessingOption::kContinueProcessing;
+    }
+
+    virtual const char* Name() const override {
+      return "TestWalFilterWithChangeBatch";
+    }
+  };
+
+  std::vector<std::vector<std::string>> batch_keys(3);
+
+  batch_keys[0].push_back("key1");
+  batch_keys[0].push_back("key2");
+  batch_keys[1].push_back("key3");
+  batch_keys[1].push_back("key4");
+  batch_keys[2].push_back("key5");
+  batch_keys[2].push_back("key6");
+
+  Options options = OptionsForLogIterTest();
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({ "pikachu" }, options);
+
+  // Write given keys in given batches
+  for (size_t i = 0; i < batch_keys.size(); i++) {
+    WriteBatch batch;
+    for (size_t j = 0; j < batch_keys[i].size(); j++) {
+      batch.Put(handles_[0], batch_keys[i][j], DummyString(1024));
+    }
+    dbfull()->Write(WriteOptions(), &batch);
+  }
+
+  // Create a test filter that would apply wal_processing_option at the first
+  // record
+  size_t change_records_from_index = 1;
+  size_t num_keys_to_add_in_new_batch = 1;
+  TestWalFilterWithChangeBatch test_wal_filter_with_change_batch(
+    change_records_from_index, num_keys_to_add_in_new_batch);
+
+  // Reopen database with option to use WAL filter
+  options = OptionsForLogIterTest();
+  options.wal_filter = &test_wal_filter_with_change_batch;
+  ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+
+  // Ensure that all keys exist before change_records_from_index_
+  // And after that index only single key exists
+  // as our filter adds only single key for each batch
+  std::vector<Slice> keys_must_exist;
+  std::vector<Slice> keys_must_not_exist;
+
+  for (size_t i = 0; i < batch_keys.size(); i++) {
+    for (size_t j = 0; j < batch_keys[i].size(); j++) {
+      if (i >= change_records_from_index && j >= 
+        num_keys_to_add_in_new_batch) {
+        keys_must_not_exist.push_back(Slice(batch_keys[i][j]));
+      }
+      else {
+        keys_must_exist.push_back(Slice(batch_keys[i][j]));
+      }
+    }
+  }
+
+  bool checked_after_reopen = false;
+
+  while (true) {
+    // Ensure that expected keys exists
+    // and not expected keys don't exist after recovery
+    ValidateKeyExistence(db_, keys_must_exist, keys_must_not_exist);
+
+    if (checked_after_reopen) {
+      break;
+    }
+
+    // reopen database again to make sure previous log(s) are not used
+    //(even if they were skipped)
+    // reopn database with option to use WAL filter
+    options = OptionsForLogIterTest();
+    ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+
+    checked_after_reopen = true;
+  }
+}
+
+TEST_F(DBTest2, WalFilterTestWithChangeBatchExtraKeys) {
+  class TestWalFilterWithChangeBatchAddExtraKeys : public WalFilter {
+  public:
+    virtual WalProcessingOption LogRecord(const WriteBatch& batch,
+      WriteBatch* new_batch,
+      bool* batch_changed) const override {
+      *new_batch = batch;
+      new_batch->Put("key_extra", "value_extra");
+      *batch_changed = true;
+      return WalProcessingOption::kContinueProcessing;
+    }
+
+    virtual const char* Name() const override {
+      return "WalFilterTestWithChangeBatchExtraKeys";
+    }
+  };
+
+  std::vector<std::vector<std::string>> batch_keys(3);
+
+  batch_keys[0].push_back("key1");
+  batch_keys[0].push_back("key2");
+  batch_keys[1].push_back("key3");
+  batch_keys[1].push_back("key4");
+  batch_keys[2].push_back("key5");
+  batch_keys[2].push_back("key6");
+
+  Options options = OptionsForLogIterTest();
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({ "pikachu" }, options);
+
+  // Write given keys in given batches
+  for (size_t i = 0; i < batch_keys.size(); i++) {
+    WriteBatch batch;
+    for (size_t j = 0; j < batch_keys[i].size(); j++) {
+      batch.Put(handles_[0], batch_keys[i][j], DummyString(1024));
+    }
+    dbfull()->Write(WriteOptions(), &batch);
+  }
+
+  // Create a test filter that would add extra keys
+  TestWalFilterWithChangeBatchAddExtraKeys test_wal_filter_extra_keys;
+
+  // Reopen database with option to use WAL filter
+  options = OptionsForLogIterTest();
+  options.wal_filter = &test_wal_filter_extra_keys;
+  Status status = TryReopenWithColumnFamilies({ "default", "pikachu" }, 
+    options);
+  ASSERT_TRUE(status.IsNotSupported());
+
+  // Reopen without filter, now reopen should succeed - previous
+  // attempt to open must not have altered the db.
+  options = OptionsForLogIterTest();
+  ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+
+  std::vector<Slice> keys_must_exist;
+  std::vector<Slice> keys_must_not_exist;  // empty vector
+
+  for (size_t i = 0; i < batch_keys.size(); i++) {
+    for (size_t j = 0; j < batch_keys[i].size(); j++) {
+      keys_must_exist.push_back(Slice(batch_keys[i][j]));
+    }
+  }
+
+  ValidateKeyExistence(db_, keys_must_exist, keys_must_not_exist);
+}
+
+TEST_F(DBTest2, WalFilterTestWithColumnFamilies) {
+  class TestWalFilterWithColumnFamilies : public WalFilter {
+  private:
+    // column_family_id -> log_number map (provided to WALFilter)
+    std::map<uint32_t, uint64_t> cf_log_number_map_;
+    // column_family_name -> column_family_id map (provided to WALFilter)
+    std::map<std::string, uint32_t> cf_name_id_map_;
+    // column_family_name -> keys_found_in_wal map
+    // We store keys that are applicable to the column_family
+    // during recovery (i.e. aren't already flushed to SST file(s))
+    // for verification against the keys we expect.
+    std::map<uint32_t, std::vector<std::string>> cf_wal_keys_;
+  public:
+    virtual void ColumnFamilyLogNumberMap(
+      const std::map<uint32_t, uint64_t>& cf_lognumber_map,
+      const std::map<std::string, uint32_t>& cf_name_id_map) override {
+      cf_log_number_map_ = cf_lognumber_map;
+      cf_name_id_map_ = cf_name_id_map;
+    }
+
+    virtual WalProcessingOption LogRecordFound(unsigned long long log_number,
+      const std::string& log_file_name,
+      const WriteBatch& batch,
+      WriteBatch* new_batch,
+      bool* batch_changed) override {
+      class LogRecordBatchHandler : public WriteBatch::Handler {
+      private:
+        const std::map<uint32_t, uint64_t> & cf_log_number_map_;
+        std::map<uint32_t, std::vector<std::string>> & cf_wal_keys_;
+        unsigned long long log_number_;
+      public:
+        LogRecordBatchHandler(unsigned long long current_log_number,
+          const std::map<uint32_t, uint64_t> & cf_log_number_map,
+          std::map<uint32_t, std::vector<std::string>> & cf_wal_keys) :
+          cf_log_number_map_(cf_log_number_map),
+          cf_wal_keys_(cf_wal_keys),
+          log_number_(current_log_number){}
+
+        virtual Status PutCF(uint32_t column_family_id, const Slice& key,
+          const Slice& /*value*/) override {
+          auto it = cf_log_number_map_.find(column_family_id);
+          assert(it != cf_log_number_map_.end());
+          unsigned long long log_number_for_cf = it->second;
+          // If the current record is applicable for column_family_id
+          // (i.e. isn't flushed to SST file(s) for column_family_id)
+          // add it to the cf_wal_keys_ map for verification.
+          if (log_number_ >= log_number_for_cf) {
+            cf_wal_keys_[column_family_id].push_back(std::string(key.data(),
+              key.size()));
+          }
+          return Status::OK();
+        }
+      } handler(log_number, cf_log_number_map_, cf_wal_keys_);
+
+      batch.Iterate(&handler);
+
+      return WalProcessingOption::kContinueProcessing;
+    }
+
+    virtual const char* Name() const override {
+      return "WalFilterTestWithColumnFamilies";
+    }
+
+    const std::map<uint32_t, std::vector<std::string>> & 
+    GetColumnFamilyKeys() {
+      return cf_wal_keys_;
+    }
+
+    const std::map<std::string, uint32_t> & GetColumnFamilyNameIdMap() {
+      return cf_name_id_map_;
+    }
+  };
+
+  std::vector<std::vector<std::string>> batch_keys_pre_flush(3);
+
+  batch_keys_pre_flush[0].push_back("key1");
+  batch_keys_pre_flush[0].push_back("key2");
+  batch_keys_pre_flush[1].push_back("key3");
+  batch_keys_pre_flush[1].push_back("key4");
+  batch_keys_pre_flush[2].push_back("key5");
+  batch_keys_pre_flush[2].push_back("key6");
+
+  Options options = OptionsForLogIterTest();
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({ "pikachu" }, options);
+
+  // Write given keys in given batches
+  for (size_t i = 0; i < batch_keys_pre_flush.size(); i++) {
+    WriteBatch batch;
+    for (size_t j = 0; j < batch_keys_pre_flush[i].size(); j++) {
+      batch.Put(handles_[0], batch_keys_pre_flush[i][j], DummyString(1024));
+      batch.Put(handles_[1], batch_keys_pre_flush[i][j], DummyString(1024));
+    }
+    dbfull()->Write(WriteOptions(), &batch);
+  }
+
+  //Flush default column-family
+  db_->Flush(FlushOptions(), handles_[0]);
+
+  // Do some more writes
+  std::vector<std::vector<std::string>> batch_keys_post_flush(3);
+
+  batch_keys_post_flush[0].push_back("key7");
+  batch_keys_post_flush[0].push_back("key8");
+  batch_keys_post_flush[1].push_back("key9");
+  batch_keys_post_flush[1].push_back("key10");
+  batch_keys_post_flush[2].push_back("key11");
+  batch_keys_post_flush[2].push_back("key12");
+
+  // Write given keys in given batches
+  for (size_t i = 0; i < batch_keys_post_flush.size(); i++) {
+    WriteBatch batch;
+    for (size_t j = 0; j < batch_keys_post_flush[i].size(); j++) {
+      batch.Put(handles_[0], batch_keys_post_flush[i][j], DummyString(1024));
+      batch.Put(handles_[1], batch_keys_post_flush[i][j], DummyString(1024));
+    }
+    dbfull()->Write(WriteOptions(), &batch);
+  }
+
+  // On Recovery we should only find the second batch applicable to default CF
+  // But both batches applicable to pikachu CF
+
+  // Create a test filter that would add extra keys
+  TestWalFilterWithColumnFamilies test_wal_filter_column_families;
+
+  // Reopen database with option to use WAL filter
+  options = OptionsForLogIterTest();
+  options.wal_filter = &test_wal_filter_column_families;
+  Status status =
+    TryReopenWithColumnFamilies({ "default", "pikachu" }, options);
+  ASSERT_TRUE(status.ok());
+
+  // verify that handles_[0] only has post_flush keys
+  // while handles_[1] has pre and post flush keys
+  auto cf_wal_keys = test_wal_filter_column_families.GetColumnFamilyKeys();
+  auto name_id_map = 
+    test_wal_filter_column_families.GetColumnFamilyNameIdMap();
+  size_t index = 0;
+  auto keys_cf = cf_wal_keys[name_id_map[kDefaultColumnFamilyName]];
+  //default column-family, only post_flush keys are expected
+  for (size_t i = 0; i < batch_keys_post_flush.size(); i++) {
+    for (size_t j = 0; j < batch_keys_post_flush[i].size(); j++) {
+      Slice key_from_the_log(keys_cf[index++]);
+      Slice batch_key(batch_keys_post_flush[i][j]);
+      ASSERT_TRUE(key_from_the_log.compare(batch_key) == 0);
+    }
+  }
+  ASSERT_TRUE(index == keys_cf.size());
+
+  index = 0;
+  keys_cf = cf_wal_keys[name_id_map["pikachu"]];
+  //pikachu column-family, all keys are expected
+  for (size_t i = 0; i < batch_keys_pre_flush.size(); i++) {
+    for (size_t j = 0; j < batch_keys_pre_flush[i].size(); j++) {
+      Slice key_from_the_log(keys_cf[index++]);
+      Slice batch_key(batch_keys_pre_flush[i][j]);
+      ASSERT_TRUE(key_from_the_log.compare(batch_key) == 0);
+    }
+  }
+
+  for (size_t i = 0; i < batch_keys_post_flush.size(); i++) {
+    for (size_t j = 0; j < batch_keys_post_flush[i].size(); j++) {
+      Slice key_from_the_log(keys_cf[index++]);
+      Slice batch_key(batch_keys_post_flush[i][j]);
+      ASSERT_TRUE(key_from_the_log.compare(batch_key) == 0);
+    }
+  }
+  ASSERT_TRUE(index == keys_cf.size());
+}
+#endif  // ROCKSDB_LITE
+
+TEST_F(DBTest2, DISABLED_FirstSnapshotTest) {
+  Options options;
+  options.write_buffer_size = 100000;  // Small write buffer
+  options = CurrentOptions(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  // This snapshot will have sequence number 0.  When compaction encounters
+  // this snapshot, CompactionIterator::findEarliestVisibleSnapshot() will
+  // assert as it expects non-zero snapshots.
+  //
+  // One fix would be to simply remove this assert.  However, a better fix
+  // might
+  // be to always have db sequence numbers start from 1 so that no code is
+  // ever
+  // confused by 0.
+  const Snapshot* s1 = db_->GetSnapshot();
+
+  Put(1, "k1", std::string(100000, 'x'));  // Fill memtable
+  Put(1, "k2", std::string(100000, 'y'));  // Trigger flush
+
+  db_->ReleaseSnapshot(s1);
+}
+
+class PinL0IndexAndFilterBlocksTest : public DBTestBase,
+                                      public testing::WithParamInterface<bool> {
+ public:
+  PinL0IndexAndFilterBlocksTest() : DBTestBase("/db_pin_l0_index_bloom_test") {}
+  virtual void SetUp() override { infinite_max_files_ = GetParam(); }
+
+  bool infinite_max_files_;
+};
+
+TEST_P(PinL0IndexAndFilterBlocksTest,
+       IndexAndFilterBlocksOfNewTableAddedToCacheWithPinning) {
+  Options options = CurrentOptions();
+  if (infinite_max_files_) {
+    options.max_open_files = -1;
+  }
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions table_options;
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.pin_l0_filter_and_index_blocks_in_cache = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  ASSERT_OK(Put(1, "key", "val"));
+  // Create a new table.
+  ASSERT_OK(Flush(1));
+
+  // index/filter blocks added to block cache right after table creation.
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+
+  // only index/filter were added
+  ASSERT_EQ(2, TestGetTickerCount(options, BLOCK_CACHE_ADD));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_DATA_MISS));
+
+  std::string value;
+  // Miss and hit count should remain the same, they're all pinned.
+  db_->KeyMayExist(ReadOptions(), handles_[1], "key", &value);
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+
+  // Miss and hit count should remain the same, they're all pinned.
+  value = Get(1, "key");
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+}
+
+TEST_P(PinL0IndexAndFilterBlocksTest,
+       MultiLevelIndexAndFilterBlocksCachedWithPinning) {
+  Options options = CurrentOptions();
+  if (infinite_max_files_) {
+    options.max_open_files = -1;
+  }
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions table_options;
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.pin_l0_filter_and_index_blocks_in_cache = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  Put(1, "a", "begin");
+  Put(1, "z", "end");
+  ASSERT_OK(Flush(1));
+  // move this table to L1
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr, handles_[1]);
+
+  // reset block cache
+  table_options.block_cache = NewLRUCache(64 * 1024);
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  TryReopenWithColumnFamilies({"default", "pikachu"}, options);
+  // create new table at L0
+  Put(1, "a2", "begin2");
+  Put(1, "z2", "end2");
+  ASSERT_OK(Flush(1));
+
+  table_options.block_cache->EraseUnRefEntries();
+
+  // get base cache values
+  uint64_t fm = TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS);
+  uint64_t fh = TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT);
+  uint64_t im = TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS);
+  uint64_t ih = TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT);
+
+  std::string value;
+  // this should be read from L0
+  // so cache values don't change
+  value = Get(1, "a2");
+  ASSERT_EQ(fm, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(fh, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(im, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(ih, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+
+  // this should be read from L1
+  // the file is opened, prefetching results in a cache filter miss
+  // the block is loaded and added to the cache,
+  // then the get results in a cache hit for L1
+  value = Get(1, "a");
+  ASSERT_EQ(fm + 1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(im + 1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+}
+
+INSTANTIATE_TEST_CASE_P(PinL0IndexAndFilterBlocksTest,
+                        PinL0IndexAndFilterBlocksTest, ::testing::Bool());
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -1,4 +1,4 @@
-// Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+// Copyright (c) 2011-present, 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.
@@ -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);
@@ -55,6 +56,7 @@ DBTestBase::DBTestBase(const std::string path)
  EXPECT_OK(DestroyDB(dbname_, options));
  db_ = nullptr;
  Reopen(options);
+  Random::GetTLSInstance()->Reset(0xdeadbeef);
 }

 DBTestBase::~DBTestBase() {
@@ -71,63 +73,69 @@ DBTestBase::~DBTestBase() {
  delete env_;
 }

-// Switch to a fresh database with the next option configuration to
-// 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_++) {
+bool DBTestBase::ShouldSkipOptions(int option_config, int skip_mask) {
 #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;
+  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 ||
+      option_config == kConcurrentSkipList) {
+    return true;
    }
 #endif

    if ((skip_mask & kSkipDeletesFilterFirst) &&
-        option_config_ == kDeletesFilterFirst) {
-      continue;
+        option_config == kDeletesFilterFirst) {
+      return true;
    }
    if ((skip_mask & kSkipUniversalCompaction) &&
-        (option_config_ == kUniversalCompaction ||
-         option_config_ == kUniversalCompactionMultiLevel)) {
-      continue;
+        (option_config == kUniversalCompaction ||
+         option_config == kUniversalCompactionMultiLevel)) {
+      return true;
    }
-    if ((skip_mask & kSkipMergePut) && option_config_ == kMergePut) {
-      continue;
+    if ((skip_mask & kSkipMergePut) && option_config == kMergePut) {
+      return true;
    }
    if ((skip_mask & kSkipNoSeekToLast) &&
-        (option_config_ == kHashLinkList || option_config_ == kHashSkipList)) {
-      continue;
+        (option_config == kHashLinkList || option_config == kHashSkipList)) {
+      return true;
    }
    if ((skip_mask & kSkipPlainTable) &&
-        (option_config_ == kPlainTableAllBytesPrefix ||
-         option_config_ == kPlainTableFirstBytePrefix ||
-         option_config_ == kPlainTableCappedPrefix ||
-         option_config_ == kPlainTableCappedPrefixNonMmap)) {
-      continue;
+        (option_config == kPlainTableAllBytesPrefix ||
+         option_config == kPlainTableFirstBytePrefix ||
+         option_config == kPlainTableCappedPrefix ||
+         option_config == kPlainTableCappedPrefixNonMmap)) {
+      return true;
    }
    if ((skip_mask & kSkipHashIndex) &&
-        (option_config_ == kBlockBasedTableWithPrefixHashIndex ||
-         option_config_ == kBlockBasedTableWithWholeKeyHashIndex)) {
-      continue;
+        (option_config == kBlockBasedTableWithPrefixHashIndex ||
+         option_config == kBlockBasedTableWithWholeKeyHashIndex)) {
+      return true;
    }
-    if ((skip_mask & kSkipHashCuckoo) && (option_config_ == kHashCuckoo)) {
-      continue;
+    if ((skip_mask & kSkipHashCuckoo) && (option_config == kHashCuckoo)) {
+      return true;
    }
-    if ((skip_mask & kSkipFIFOCompaction) &&
-        option_config_ == kFIFOCompaction) {
-      continue;
+    if ((skip_mask & kSkipFIFOCompaction) && option_config == kFIFOCompaction) {
+      return true;
    }
-    if ((skip_mask & kSkipMmapReads) && option_config_ == kWalDirAndMmapReads) {
+    if ((skip_mask & kSkipMmapReads) && option_config == kWalDirAndMmapReads) {
+      return true;
+    }
+    return false;
+}
+
+// Switch to a fresh database with the next option configuration to
+// 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_++) {
+    if (ShouldSkipOptions(option_config_, skip_mask)) {
      continue;
    }
    break;
@@ -202,6 +210,8 @@ Options DBTestBase::CurrentOptions(
    const anon::OptionsOverride& options_override) {
  Options options;
  options.write_buffer_size = 4090 * 4096;
+  options.target_file_size_base = 2 * 1024 * 1024;
+  options.max_bytes_for_level_base = 10 * 1024 * 1024;
  return CurrentOptions(options, options_override);
 }

@@ -293,6 +303,7 @@ Options DBTestBase::CurrentOptions(
    case kPerfOptions:
      options.soft_rate_limit = 2.0;
      options.delayed_write_rate = 8 * 1024 * 1024;
+      options.report_bg_io_stats = true;
      // TODO(3.13) -- test more options
      break;
    case kDeletesFilterFirst:
@@ -331,6 +342,10 @@ Options DBTestBase::CurrentOptions(
      options.prefix_extractor.reset(NewNoopTransform());
      break;
    }
+    case kBlockBasedTableWithIndexRestartInterval: {
+      table_options.index_block_restart_interval = 8;
+      break;
+    }
    case kOptimizeFiltersForHits: {
      options.optimize_filters_for_hits = true;
      set_block_based_table_factory = true;
@@ -354,6 +369,11 @@ Options DBTestBase::CurrentOptions(
      options.max_subcompactions = 4;
      break;
    }
+    case kConcurrentSkipList: {
+      options.allow_concurrent_memtable_write = true;
+      options.enable_write_thread_adaptive_yield = true;
+      break;
+    }

    default:
      break;
@@ -367,6 +387,7 @@ Options DBTestBase::CurrentOptions(
  }
  options.env = env_;
  options.create_if_missing = true;
+  options.fail_if_options_file_error = true;
  return options;
 }

@@ -658,14 +679,14 @@ uint64_t DBTestBase::SizeAtLevel(int level) {
  return sum;
 }

-int DBTestBase::TotalLiveFiles(int cf) {
+size_t DBTestBase::TotalLiveFiles(int cf) {
  ColumnFamilyMetaData cf_meta;
  if (cf == 0) {
    db_->GetColumnFamilyMetaData(&cf_meta);
  } else {
    db_->GetColumnFamilyMetaData(handles_[cf], &cf_meta);
  }
-  int num_files = 0;
+  size_t num_files = 0;
  for (auto& level : cf_meta.levels) {
    num_files += level.files.size();
  }
@@ -830,7 +851,7 @@ 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)++;
  }
@@ -842,7 +863,7 @@ 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)++;
  }
@@ -852,8 +873,10 @@ 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)));
@@ -995,4 +1018,29 @@ void DBTestBase::CopyFile(const std::string& source,
  ASSERT_OK(destfile->Close());
 }

+std::unordered_map<std::string, uint64_t> DBTestBase::GetAllSSTFiles(
+    uint64_t* total_size) {
+  std::unordered_map<std::string, uint64_t> res;
+
+  if (total_size) {
+    *total_size = 0;
+  }
+  std::vector<std::string> files;
+  env_->GetChildren(dbname_, &files);
+  for (auto& file_name : files) {
+    uint64_t number;
+    FileType type;
+    std::string file_path = dbname_ + "/" + file_name;
+    if (ParseFileName(file_name, &number, &type) && type == kTableFile) {
+      uint64_t file_size = 0;
+      env_->GetFileSize(file_path, &file_size);
+      res[file_path] = file_size;
+      if (total_size) {
+        *total_size += file_size;
+      }
+    }
+  }
+  return res;
+}
+
 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-// Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+// Copyright (c) 2011-present, 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.
@@ -19,6 +19,7 @@
 #endif

 #include <algorithm>
+#include <map>
 #include <set>
 #include <string>
 #include <thread>
@@ -27,7 +28,6 @@
 #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"
@@ -118,6 +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:
@@ -327,15 +405,19 @@ class SpecialEnv : public EnvWrapper {

  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();
    }
@@ -343,11 +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_;
@@ -399,6 +483,9 @@ 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};
@@ -438,9 +525,11 @@ class DBTestBase : public testing::Test {
    kOptimizeFiltersForHits = 27,
    kRowCache = 28,
    kRecycleLogFiles = 29,
-    kLevelSubcompactions = 30,
-    kUniversalSubcompactions = 31,
-    kEnd = 30
+    kConcurrentSkipList = 30,
+    kEnd = 31,
+    kLevelSubcompactions = 31,
+    kUniversalSubcompactions = 32,
+    kBlockBasedTableWithIndexRestartInterval = 33,
  };
  int option_config_;

@@ -486,6 +575,8 @@ class DBTestBase : public testing::Test {
    return std::string(buf);
  }

+  static bool ShouldSkipOptions(int option_config, int skip_mask = kNoSkip);
+
  // Switch to a fresh database with the next option configuration to
  // test.  Return false if there are no more configurations to test.
  bool ChangeOptions(int skip_mask = kNoSkip);
@@ -574,7 +665,7 @@ class DBTestBase : public testing::Test {

  uint64_t SizeAtLevel(int level);

-  int TotalLiveFiles(int cf = 0);
+  size_t TotalLiveFiles(int cf = 0);

  size_t CountLiveFiles();
 #endif  // ROCKSDB_LITE
@@ -622,6 +713,9 @@ 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);
@@ -659,6 +753,9 @@ class DBTestBase : public testing::Test {

  void CopyFile(const std::string& source, const std::string& destination,
                uint64_t size = 0);
+
+  std::unordered_map<std::string, uint64_t> GetAllSSTFiles(
+      uint64_t* total_size = nullptr);
 };

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -14,6 +14,11 @@

 namespace rocksdb {

+static uint64_t TestGetTickerCount(const Options& options,
+                                   Tickers ticker_type) {
+  return options.statistics->getTickerCount(ticker_type);
+}
+
 static std::string CompressibleString(Random* rnd, int len) {
  std::string r;
  test::CompressibleString(rnd, 0.8, len, &r);
@@ -22,12 +27,16 @@ static std::string CompressibleString(Random* rnd, int len) {

 class DBTestUniversalCompactionBase
    : public DBTestBase,
-      public ::testing::WithParamInterface<int> {
+      public ::testing::WithParamInterface<std::tuple<int, bool>> {
 public:
  explicit DBTestUniversalCompactionBase(
      const std::string& path) : DBTestBase(path) {}
-  virtual void SetUp() override { num_levels_ = GetParam(); }
+  virtual void SetUp() override {
+    num_levels_ = std::get<0>(GetParam());
+    exclusive_manual_compaction_ = std::get<1>(GetParam());
+  }
  int num_levels_;
+  bool exclusive_manual_compaction_;
 };

 class DBTestUniversalCompaction : public DBTestUniversalCompactionBase {
@@ -112,6 +121,108 @@ 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 = CurrentOptions();
+
+  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);
+  }
+}
+
+TEST_P(DBTestUniversalCompaction, OptimizeFiltersForHits) {
+  Options options = CurrentOptions();
+  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;
+  options.target_file_size_base = 32 << 10;  // 32KB
+  // trigger compaction if there are >= 4 files
+  options.level0_file_num_compaction_trigger = 4;
+  BlockBasedTableOptions bbto;
+  bbto.cache_index_and_filter_blocks = true;
+  bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  bbto.whole_key_filtering = true;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  options.optimize_filters_for_hits = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  options.memtable_factory.reset(new SpecialSkipListFactory(3));
+
+  DestroyAndReopen(options);
+
+  // block compaction from happening
+  env_->SetBackgroundThreads(1, Env::LOW);
+  test::SleepingBackgroundTask sleeping_task_low;
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
+                 Env::Priority::LOW);
+
+  for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+    Put(Key(num * 10), "val");
+    if (num) {
+      dbfull()->TEST_WaitForFlushMemTable();
+    }
+    Put(Key(30 + num * 10), "val");
+    Put(Key(60 + num * 10), "val");
+  }
+  Put("", "");
+  dbfull()->TEST_WaitForFlushMemTable();
+
+  // Query set of non existing keys
+  for (int i = 5; i < 90; i += 10) {
+    ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
+  }
+
+  // Make sure bloom filter is used at least once.
+  ASSERT_GT(TestGetTickerCount(options, BLOOM_FILTER_USEFUL), 0);
+  auto prev_counter = TestGetTickerCount(options, BLOOM_FILTER_USEFUL);
+
+  // Make sure bloom filter is used for all but the last L0 file when looking
+  // up a non-existent key that's in the range of all L0 files.
+  ASSERT_EQ(Get(Key(35)), "NOT_FOUND");
+  ASSERT_EQ(prev_counter + NumTableFilesAtLevel(0) - 1,
+            TestGetTickerCount(options, BLOOM_FILTER_USEFUL));
+  prev_counter = TestGetTickerCount(options, BLOOM_FILTER_USEFUL);
+
+  // Unblock compaction and wait it for happening.
+  sleeping_task_low.WakeUp();
+  dbfull()->TEST_WaitForCompact();
+
+  // The same queries will not trigger bloom filter
+  for (int i = 5; i < 90; i += 10) {
+    ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
+  }
+  ASSERT_EQ(prev_counter, TestGetTickerCount(options, BLOOM_FILTER_USEFUL));
+}
+
 // 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 +230,7 @@ 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;
@@ -223,13 +335,12 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrigger) {
 }

 TEST_P(DBTestUniversalCompaction, UniversalCompactionSizeAmplification) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = num_levels_;
  options.write_buffer_size = 100 << 10;     // 100KB
  options.target_file_size_base = 32 << 10;  // 32KB
  options.level0_file_num_compaction_trigger = 3;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -274,12 +385,12 @@ TEST_P(DBTestUniversalCompaction, CompactFilesOnUniversalCompaction) {
  ChangeCompactOptions();
  Options options;
  options.create_if_missing = true;
-  options.write_buffer_size = kEntrySize * kEntriesPerBuffer;
  options.compaction_style = kCompactionStyleLevel;
  options.num_levels = 1;
  options.target_file_size_base = options.write_buffer_size;
  options.compression = kNoCompression;
  options = CurrentOptions(options);
+  options.write_buffer_size = kEntrySize * kEntriesPerBuffer;
  CreateAndReopenWithCF({"pikachu"}, options);
  ASSERT_EQ(options.compaction_style, kCompactionStyleUniversal);
  Random rnd(301);
@@ -337,12 +448,11 @@ TEST_P(DBTestUniversalCompaction, CompactFilesOnUniversalCompaction) {
 }

 TEST_P(DBTestUniversalCompaction, UniversalCompactionTargetLevel) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.write_buffer_size = 100 << 10;     // 100KB
  options.num_levels = 7;
  options.disable_auto_compactions = true;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  // Generate 3 overlapping files
@@ -367,6 +477,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTargetLevel) {
  CompactRangeOptions compact_options;
  compact_options.change_level = true;
  compact_options.target_level = 4;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
  db_->CompactRange(compact_options, nullptr, nullptr);
  ASSERT_EQ("0,0,0,0,1", FilesPerLevel(0));
 }
@@ -381,14 +492,13 @@ class DBTestUniversalCompactionMultiLevels
 };

 TEST_P(DBTestUniversalCompactionMultiLevels, UniversalCompactionMultiLevels) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = num_levels_;
  options.write_buffer_size = 100 << 10;  // 100KB
  options.level0_file_num_compaction_trigger = 8;
  options.max_background_compactions = 3;
  options.target_file_size_base = 32 * 1024;
-  options = CurrentOptions(options);
  CreateAndReopenWithCF({"pikachu"}, options);

  // Trigger compaction if size amplification exceeds 110%
@@ -424,7 +534,7 @@ TEST_P(DBTestUniversalCompactionMultiLevels, UniversalCompactionTrivialMove) {
      });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.compaction_options_universal.allow_trivial_move = true;
  options.num_levels = 3;
@@ -432,7 +542,6 @@ TEST_P(DBTestUniversalCompactionMultiLevels, UniversalCompactionTrivialMove) {
  options.level0_file_num_compaction_trigger = 3;
  options.max_background_compactions = 2;
  options.target_file_size_base = 32 * 1024;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -459,7 +568,8 @@ TEST_P(DBTestUniversalCompactionMultiLevels, UniversalCompactionTrivialMove) {

 INSTANTIATE_TEST_CASE_P(DBTestUniversalCompactionMultiLevels,
                        DBTestUniversalCompactionMultiLevels,
-                        ::testing::Values(3, 20));
+                        ::testing::Combine(::testing::Values(3, 20),
+                                           ::testing::Bool()));

 class DBTestUniversalCompactionParallel :
    public DBTestUniversalCompactionBase {
@@ -470,7 +580,7 @@ class DBTestUniversalCompactionParallel :
 };

 TEST_P(DBTestUniversalCompactionParallel, UniversalCompactionParallel) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = num_levels_;
  options.write_buffer_size = 1 << 10;  // 1KB
@@ -479,7 +589,6 @@ TEST_P(DBTestUniversalCompactionParallel, UniversalCompactionParallel) {
  options.max_background_flushes = 3;
  options.target_file_size_base = 1 * 1024;
  options.compaction_options_universal.max_size_amplification_percent = 110;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -532,10 +641,11 @@ TEST_P(DBTestUniversalCompactionParallel, UniversalCompactionParallel) {

 INSTANTIATE_TEST_CASE_P(DBTestUniversalCompactionParallel,
                        DBTestUniversalCompactionParallel,
-                        ::testing::Values(1, 10));
+                        ::testing::Combine(::testing::Values(1, 10),
+                                           ::testing::Bool()));

 TEST_P(DBTestUniversalCompaction, UniversalCompactionOptions) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.write_buffer_size = 105 << 10;    // 105KB
  options.arena_block_size = 4 << 10;       // 4KB
@@ -543,7 +653,6 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionOptions) {
  options.level0_file_num_compaction_trigger = 4;
  options.num_levels = num_levels_;
  options.compaction_options_universal.compression_size_percent = -1;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -656,14 +765,13 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionCompressRatio1) {
    return;
  }

-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.write_buffer_size = 100 << 10;     // 100KB
  options.target_file_size_base = 32 << 10;  // 32KB
  options.level0_file_num_compaction_trigger = 2;
  options.num_levels = num_levels_;
  options.compaction_options_universal.compression_size_percent = 70;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  Random rnd(301);
@@ -724,14 +832,13 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionCompressRatio2) {
  if (!Snappy_Supported()) {
    return;
  }
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.write_buffer_size = 100 << 10;     // 100KB
  options.target_file_size_base = 32 << 10;  // 32KB
  options.level0_file_num_compaction_trigger = 2;
  options.num_levels = num_levels_;
  options.compaction_options_universal.compression_size_percent = 95;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  Random rnd(301);
@@ -767,7 +874,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest1) {
      });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.compaction_options_universal.allow_trivial_move = true;
  options.num_levels = 2;
@@ -775,7 +882,6 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest1) {
  options.level0_file_num_compaction_trigger = 3;
  options.max_background_compactions = 1;
  options.target_file_size_base = 32 * 1024;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -812,7 +918,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.compaction_options_universal.allow_trivial_move = true;
  options.num_levels = 15;
@@ -820,7 +926,6 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {
  options.level0_file_num_compaction_trigger = 8;
  options.max_background_compactions = 4;
  options.target_file_size_base = 64 * 1024;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -846,17 +951,19 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {
 }

 TEST_P(DBTestUniversalCompaction, UniversalCompactionFourPaths) {
-  Options options;
+  Options options = CurrentOptions();
  options.db_paths.emplace_back(dbname_, 300 * 1024);
  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;
  options.num_levels = 1;
-  options = CurrentOptions(options);

  std::vector<std::string> filenames;
  env_->GetChildren(options.db_paths[1].path, &filenames);
@@ -968,29 +1075,26 @@ 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);

  for (int i = 0; i <= max_key1; i++) {
    // each value is 10K
    ASSERT_OK(Put(1, Key(i), RandomString(&rnd, 10000)));
+    dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
  }
  ASSERT_OK(Flush(1));
  dbfull()->TEST_WaitForCompact();

-  int non_level0_num_files = 0;
-  for (int i = 1; i < options.num_levels; i++) {
-    non_level0_num_files += NumTableFilesAtLevel(i, 1);
-  }
-  ASSERT_EQ(non_level0_num_files, 0);
-
  // Stage 2: reopen with universal compaction, num_levels=4
  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = 4;
@@ -1003,6 +1107,7 @@ TEST_P(DBTestUniversalCompaction, IncreaseUniversalCompactionNumLevels) {
  for (int i = max_key1 + 1; i <= max_key2; i++) {
    // each value is 10K
    ASSERT_OK(Put(1, Key(i), RandomString(&rnd, 10000)));
+    dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
  }
  ASSERT_OK(Flush(1));
  dbfull()->TEST_WaitForCompact();
@@ -1019,6 +1124,7 @@ TEST_P(DBTestUniversalCompaction, IncreaseUniversalCompactionNumLevels) {
  CompactRangeOptions compact_options;
  compact_options.change_level = true;
  compact_options.target_level = 0;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
  dbfull()->CompactRange(compact_options, handles_[1], nullptr, nullptr);
  // Need to restart it once to remove higher level records in manifest.
  ReopenWithColumnFamilies({"default", "pikachu"}, options);
@@ -1032,6 +1138,7 @@ TEST_P(DBTestUniversalCompaction, IncreaseUniversalCompactionNumLevels) {
  for (int i = max_key2 + 1; i <= max_key3; i++) {
    // each value is 10K
    ASSERT_OK(Put(1, Key(i), RandomString(&rnd, 10000)));
+    dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
  }
  ASSERT_OK(Flush(1));
  dbfull()->TEST_WaitForCompact();
@@ -1043,16 +1150,18 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionSecondPathRatio) {
  if (!Snappy_Supported()) {
    return;
  }
-  Options options;
+  Options options = CurrentOptions();
  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 = CurrentOptions(options);
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));

  std::vector<std::string> filenames;
  env_->GetChildren(options.db_paths[1].path, &filenames);
@@ -1139,7 +1248,8 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionSecondPathRatio) {
 }

 INSTANTIATE_TEST_CASE_P(UniversalCompactionNumLevels, DBTestUniversalCompaction,
-                        ::testing::Values(1, 3, 5));
+                        ::testing::Combine(::testing::Values(1, 3, 5),
+                                           ::testing::Bool()));

 class DBTestUniversalManualCompactionOutputPathId
    : public DBTestUniversalCompactionBase {
@@ -1171,6 +1281,7 @@ TEST_P(DBTestUniversalManualCompactionOutputPathId,
  // Full compaction to DB path 0
  CompactRangeOptions compact_options;
  compact_options.target_path_id = 1;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
  db_->CompactRange(compact_options, handles_[1], nullptr, nullptr);
  ASSERT_EQ(1, TotalLiveFiles(1));
  ASSERT_EQ(0, GetSstFileCount(options.db_paths[0].path));
@@ -1193,6 +1304,7 @@ TEST_P(DBTestUniversalManualCompactionOutputPathId,

  // Full compaction to DB path 0
  compact_options.target_path_id = 0;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
  db_->CompactRange(compact_options, handles_[1], nullptr, nullptr);
  ASSERT_EQ(1, TotalLiveFiles(1));
  ASSERT_EQ(1, GetSstFileCount(options.db_paths[0].path));
@@ -1200,13 +1312,15 @@ TEST_P(DBTestUniversalManualCompactionOutputPathId,

  // Fail when compacting to an invalid path ID
  compact_options.target_path_id = 2;
+  compact_options.exclusive_manual_compaction = exclusive_manual_compaction_;
  ASSERT_TRUE(db_->CompactRange(compact_options, handles_[1], nullptr, nullptr)
                  .IsInvalidArgument());
 }

 INSTANTIATE_TEST_CASE_P(DBTestUniversalManualCompactionOutputPathId,
                        DBTestUniversalManualCompactionOutputPathId,
-                        ::testing::Values(1, 8));
+                        ::testing::Combine(::testing::Values(1, 8),
+                                           ::testing::Bool()));

 }  // namespace rocksdb

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -271,7 +271,8 @@ inline LookupKey::~LookupKey() {

 class IterKey {
 public:
-  IterKey() : key_(space_), buf_size_(sizeof(space_)), key_size_(0) {}
+  IterKey()
+      : buf_(space_), buf_size_(sizeof(space_)), key_(buf_), key_size_(0) {}

  ~IterKey() { ResetBuffer(); }

@@ -293,31 +294,41 @@ class IterKey {
  void TrimAppend(const size_t shared_len, const char* non_shared_data,
                  const size_t non_shared_len) {
    assert(shared_len <= key_size_);
-
    size_t total_size = shared_len + non_shared_len;
-    if (total_size <= buf_size_) {
-      key_size_ = total_size;
-    } else {
+
+    if (IsKeyPinned() /* key is not in buf_ */) {
+      // Copy the key from external memory to buf_ (copy shared_len bytes)
+      EnlargeBufferIfNeeded(total_size);
+      memcpy(buf_, key_, shared_len);
+    } else if (total_size > buf_size_) {
      // Need to allocate space, delete previous space
      char* p = new char[total_size];
      memcpy(p, key_, shared_len);

-      if (key_ != space_) {
-        delete[] key_;
+      if (buf_ != space_) {
+        delete[] buf_;
      }

-      key_ = p;
-      key_size_ = total_size;
+      buf_ = p;
      buf_size_ = total_size;
    }

-    memcpy(key_ + shared_len, non_shared_data, non_shared_len);
+    memcpy(buf_ + shared_len, non_shared_data, non_shared_len);
+    key_ = buf_;
+    key_size_ = total_size;
  }

-  Slice SetKey(const Slice& key) {
+  Slice SetKey(const Slice& key, bool copy = true) {
    size_t size = key.size();
-    EnlargeBufferIfNeeded(size);
-    memcpy(key_, key.data(), size);
+    if (copy) {
+      // Copy key to buf_
+      EnlargeBufferIfNeeded(size);
+      memcpy(buf_, key.data(), size);
+      key_ = buf_;
+    } else {
+      // Update key_ to point to external memory
+      key_ = key.data();
+    }
    key_size_ = size;
    return Slice(key_, key_size_);
  }
@@ -335,11 +346,14 @@ class IterKey {
  // Update the sequence number in the internal key.  Guarantees not to
  // invalidate slices to the key (and the user key).
  void UpdateInternalKey(uint64_t seq, ValueType t) {
+    assert(!IsKeyPinned());
    assert(key_size_ >= 8);
    uint64_t newval = (seq << 8) | t;
-    EncodeFixed64(&key_[key_size_ - 8], newval);
+    EncodeFixed64(&buf_[key_size_ - 8], newval);
  }

+  bool IsKeyPinned() const { return (key_ != buf_); }
+
  void SetInternalKey(const Slice& key_prefix, const Slice& user_key,
                      SequenceNumber s,
                      ValueType value_type = kValueTypeForSeek) {
@@ -347,10 +361,12 @@ class IterKey {
    size_t usize = user_key.size();
    EnlargeBufferIfNeeded(psize + usize + sizeof(uint64_t));
    if (psize > 0) {
-      memcpy(key_, key_prefix.data(), psize);
+      memcpy(buf_, key_prefix.data(), psize);
    }
-    memcpy(key_ + psize, user_key.data(), usize);
-    EncodeFixed64(key_ + usize + psize, PackSequenceAndType(s, value_type));
+    memcpy(buf_ + psize, user_key.data(), usize);
+    EncodeFixed64(buf_ + usize + psize, PackSequenceAndType(s, value_type));
+
+    key_ = buf_;
    key_size_ = psize + usize + sizeof(uint64_t);
  }

@@ -377,20 +393,22 @@ class IterKey {
  void EncodeLengthPrefixedKey(const Slice& key) {
    auto size = key.size();
    EnlargeBufferIfNeeded(size + static_cast<size_t>(VarintLength(size)));
-    char* ptr = EncodeVarint32(key_, static_cast<uint32_t>(size));
+    char* ptr = EncodeVarint32(buf_, static_cast<uint32_t>(size));
    memcpy(ptr, key.data(), size);
+    key_ = buf_;
  }

 private:
-  char* key_;
+  char* buf_;
  size_t buf_size_;
+  const char* key_;
  size_t key_size_;
  char space_[32];  // Avoid allocation for short keys

  void ResetBuffer() {
-    if (key_ != space_) {
-      delete[] key_;
-      key_ = space_;
+    if (buf_ != space_) {
+      delete[] buf_;
+      buf_ = space_;
    }
    buf_size_ = sizeof(space_);
    key_size_ = 0;
@@ -407,7 +425,7 @@ class IterKey {
    if (key_size > buf_size_) {
      // Need to enlarge the buffer.
      ResetBuffer();
-      key_ = new char[key_size];
+      buf_ = new char[key_size];
      buf_size_ = key_size;
    }
  }
@@ -447,6 +465,12 @@ class InternalKeySliceTransform : public SliceTransform {
  const SliceTransform* const transform_;
 };

+// Read the key of a record from a write batch.
+// if this record represent the default column family then cf_record
+// must be passed as false, otherwise it must be passed as true.
+extern bool ReadKeyFromWriteBatchEntry(Slice* input, Slice* key,
+                                       bool cf_record);
+
 // Read record from a write batch piece from input.
 // tag, column_family, key, value and blob are return values. Callers own the
 // Slice they point to.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -37,6 +37,7 @@ class DeleteFileTest : public testing::Test {
  DeleteFileTest() {
    db_ = nullptr;
    env_ = Env::Default();
+    options_.delete_obsolete_files_period_micros = 0;  // always do full purge
    options_.enable_thread_tracking = true;
    options_.write_buffer_size = 1024*1024*1000;
    options_.target_file_size_base = 1024*1024*1000;
@@ -74,6 +75,7 @@ class DeleteFileTest : public testing::Test {

  void CloseDB() {
    delete db_;
+    db_ = nullptr;
  }

  void AddKeys(int numkeys, int startkey = 0) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -8,7 +8,10 @@
 namespace rocksdb {

 namespace {
-inline double SafeDivide(double a, double b) { return b == 0.0 ? 0 : a / b; }
+template<class T>
+inline T SafeDivide(T a, T b) {
+  return b == 0 ? 0 : a / b;
+}
 }  // namespace

 void EventHelpers::AppendCurrentTime(JSONWriter* jwriter) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2015, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -645,18 +645,17 @@ class FaultInjectionTest : public testing::Test,
    return test::RandomString(&r, kValueSize, storage);
  }

-  Status OpenDB() {
-    delete db_;
-    db_ = NULL;
-    env_->ResetState();
-    return DB::Open(options_, dbname_, &db_);
-  }
-
  void CloseDB() {
    delete db_;
    db_ = NULL;
  }

+  Status OpenDB() {
+    CloseDB();
+    env_->ResetState();
+    return DB::Open(options_, dbname_, &db_);
+  }
+
  void DeleteAllData() {
    Iterator* iter = db_->NewIterator(ReadOptions());
    WriteOptions options;
@@ -784,6 +783,7 @@ TEST_P(FaultInjectionTest, WriteOptionSyncTest) {
  // Block the job queue to prevent flush job from running.
  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
                 Env::Priority::HIGH);
+  sleeping_task_low.WaitUntilSleeping();

  WriteOptions write_options;
  write_options.sync = false;
@@ -867,6 +867,7 @@ TEST_P(FaultInjectionTest, ManualLogSyncTest) {
  // Block the job queue to prevent flush job from running.
  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
                 Env::Priority::HIGH);
+  sleeping_task_low.WaitUntilSleeping();

  WriteOptions write_options;
  write_options.sync = false;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -26,9 +26,9 @@ struct FileLevel;
 // The file tree structure in Version is prebuilt and the range of each file
 // is known. On Version::Get(), it uses binary search to find a potential file
 // and then check if a target key can be found in the file by comparing the key
-// to each file's smallest and largest key. The results of these comparisions
+// to each file's smallest and largest key. The results of these comparisons
 // can be reused beyond checking if a key falls into a file's range.
-// With some pre-calculated knowledge, each key comparision that has been done
+// With some pre-calculated knowledge, each key comparison that has been done
 // can serve as a hint to narrow down further searches: if a key compared to
 // be smaller than a file's smallest or largest, that comparison can be used
 // to find out the right bound of next binary search. Similarly, if a key
@@ -48,7 +48,7 @@ class FileIndexer {
  size_t LevelIndexSize(size_t level) const;

  // Return a file index range in the next level to search for a key based on
-  // smallest and largest key comparision for the current file specified by
+  // smallest and largest key comparison for the current file specified by
  // level and file_index. When *left_index < *right_index, both index should
  // be valid and fit in the vector size.
  void GetNextLevelIndex(const size_t level, const size_t file_index,
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -21,6 +21,7 @@
 #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 {
@@ -47,8 +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++;
  }
@@ -146,7 +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,
@@ -186,6 +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",
@@ -206,6 +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,
@@ -268,6 +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
@@ -302,7 +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;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -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,
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -62,10 +62,11 @@ FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
                   InstrumentedMutex* db_mutex,
                   std::atomic<bool>* shutting_down,
                   std::vector<SequenceNumber> existing_snapshots,
+                   SequenceNumber earliest_write_conflict_snapshot,
                   JobContext* job_context, LogBuffer* log_buffer,
                   Directory* db_directory, Directory* output_file_directory,
                   CompressionType output_compression, Statistics* stats,
-                   EventLogger* event_logger)
+                   EventLogger* event_logger, bool measure_io_stats)
    : dbname_(dbname),
      cfd_(cfd),
      db_options_(db_options),
@@ -75,13 +76,15 @@ FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
      db_mutex_(db_mutex),
      shutting_down_(shutting_down),
      existing_snapshots_(std::move(existing_snapshots)),
+      earliest_write_conflict_snapshot_(earliest_write_conflict_snapshot),
      job_context_(job_context),
      log_buffer_(log_buffer),
      db_directory_(db_directory),
      output_file_directory_(output_file_directory),
      output_compression_(output_compression),
      stats_(stats),
-      event_logger_(event_logger) {
+      event_logger_(event_logger),
+      measure_io_stats_(measure_io_stats) {
  // Update the thread status to indicate flush.
  ReportStartedFlush();
  TEST_SYNC_POINT("FlushJob::FlushJob()");
@@ -92,7 +95,8 @@ FlushJob::~FlushJob() {
 }

 void FlushJob::ReportStartedFlush() {
-  ThreadStatusUtil::SetColumnFamily(cfd_);
+  ThreadStatusUtil::SetColumnFamily(cfd_, cfd_->ioptions()->env,
+                                    cfd_->options()->enable_thread_tracking);
  ThreadStatusUtil::SetThreadOperation(ThreadStatus::OP_FLUSH);
  ThreadStatusUtil::SetThreadOperationProperty(
      ThreadStatus::COMPACTION_JOB_ID,
@@ -118,6 +122,21 @@ void FlushJob::RecordFlushIOStats() {
 Status FlushJob::Run(FileMetaData* file_meta) {
  AutoThreadOperationStageUpdater stage_run(
      ThreadStatus::STAGE_FLUSH_RUN);
+  // I/O measurement variables
+  PerfLevel prev_perf_level = PerfLevel::kEnableTime;
+  uint64_t prev_write_nanos = 0;
+  uint64_t prev_fsync_nanos = 0;
+  uint64_t prev_range_sync_nanos = 0;
+  uint64_t prev_prepare_write_nanos = 0;
+  if (measure_io_stats_) {
+    prev_perf_level = GetPerfLevel();
+    SetPerfLevel(PerfLevel::kEnableTime);
+    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);
+  }
+
  // Save the contents of the earliest memtable as a new Table
  FileMetaData meta;
  autovector<MemTable*> mems;
@@ -177,6 +196,18 @@ Status FlushJob::Run(FileMetaData* file_meta) {
  }
  stream.EndArray();

+  if (measure_io_stats_) {
+    if (prev_perf_level != PerfLevel::kEnableTime) {
+      SetPerfLevel(prev_perf_level);
+    }
+    stream << "file_write_nanos" << (IOSTATS(write_nanos) - prev_write_nanos);
+    stream << "file_range_sync_nanos"
+           << (IOSTATS(range_sync_nanos) - prev_range_sync_nanos);
+    stream << "file_fsync_nanos" << (IOSTATS(fsync_nanos) - prev_fsync_nanos);
+    stream << "file_prepare_write_nanos"
+           << (IOSTATS(prepare_write_nanos) - prev_prepare_write_nanos);
+  }
+
  return s;
 }

@@ -231,14 +262,15 @@ 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(), cfd_->GetID(),
-                     existing_snapshots_, output_compression_,
-                     cfd_->ioptions()->compression_opts,
-                     mutable_cf_options_.paranoid_file_checks,
-                     cfd_->internal_stats(), Env::IO_HIGH, &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(),
+          cfd_->GetName(), existing_snapshots_,
+          earliest_write_conflict_snapshot_, output_compression_,
+          cfd_->ioptions()->compression_opts,
+          mutable_cf_options_.paranoid_file_checks, cfd_->internal_stats(),
+          Env::IO_HIGH, &table_properties_, 0 /* level */);
      info.table_properties = table_properties_;
      LogFlush(db_options_.info_log);
    }
@@ -268,6 +300,7 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
    if (!db_options_.disableDataSync && output_file_directory_ != nullptr) {
      output_file_directory_->Fsync();
    }
+    TEST_SYNC_POINT("FlushJob::WriteLevel0Table");
    db_mutex_->Lock();
  }
  base->Unref();
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -58,10 +58,11 @@ class FlushJob {
           const EnvOptions& env_options, VersionSet* versions,
           InstrumentedMutex* db_mutex, std::atomic<bool>* shutting_down,
           std::vector<SequenceNumber> existing_snapshots,
+           SequenceNumber earliest_write_conflict_snapshot,
           JobContext* job_context, LogBuffer* log_buffer,
           Directory* db_directory, Directory* output_file_directory,
           CompressionType output_compression, Statistics* stats,
-           EventLogger* event_logger);
+           EventLogger* event_logger, bool measure_io_stats);

  ~FlushJob();

@@ -83,6 +84,7 @@ class FlushJob {
  InstrumentedMutex* db_mutex_;
  std::atomic<bool>* shutting_down_;
  std::vector<SequenceNumber> existing_snapshots_;
+  SequenceNumber earliest_write_conflict_snapshot_;
  JobContext* job_context_;
  LogBuffer* log_buffer_;
  Directory* db_directory_;
@@ -91,6 +93,7 @@ class FlushJob {
  Statistics* stats_;
  EventLogger* event_logger_;
  TableProperties table_properties_;
+  bool measure_io_stats_;
 };

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -92,8 +92,8 @@ TEST_F(FlushJobTest, Empty) {
  FlushJob flush_job(dbname_, versions_->GetColumnFamilySet()->GetDefault(),
                     db_options_, *cfd->GetLatestMutableCFOptions(),
                     env_options_, versions_.get(), &mutex_, &shutting_down_,
-                     {}, &job_context, nullptr, nullptr, nullptr,
-                     kNoCompression, nullptr, &event_logger);
+                     {}, kMaxSequenceNumber, &job_context, nullptr, nullptr,
+                     nullptr, kNoCompression, nullptr, &event_logger, false);
  ASSERT_OK(flush_job.Run());
  job_context.Clean();
 }
@@ -131,8 +131,8 @@ TEST_F(FlushJobTest, NonEmpty) {
  FlushJob flush_job(dbname_, versions_->GetColumnFamilySet()->GetDefault(),
                     db_options_, *cfd->GetLatestMutableCFOptions(),
                     env_options_, versions_.get(), &mutex_, &shutting_down_,
-                     {}, &job_context, nullptr, nullptr, nullptr,
-                     kNoCompression, nullptr, &event_logger);
+                     {}, kMaxSequenceNumber, &job_context, nullptr, nullptr,
+                     nullptr, kNoCompression, nullptr, &event_logger, true);
  FileMetaData fd;
  mutex_.Lock();
  ASSERT_OK(flush_job.Run(&fd));
@@ -192,11 +192,11 @@ TEST_F(FlushJobTest, Snapshots) {
  }

  EventLogger event_logger(db_options_.info_log.get());
-  FlushJob flush_job(dbname_, versions_->GetColumnFamilySet()->GetDefault(),
-                     db_options_, *cfd->GetLatestMutableCFOptions(),
-                     env_options_, versions_.get(), &mutex_, &shutting_down_,
-                     snapshots, &job_context, nullptr, nullptr, nullptr,
-                     kNoCompression, nullptr, &event_logger);
+  FlushJob flush_job(
+      dbname_, versions_->GetColumnFamilySet()->GetDefault(), db_options_,
+      *cfd->GetLatestMutableCFOptions(), env_options_, versions_.get(), &mutex_,
+      &shutting_down_, snapshots, kMaxSequenceNumber, &job_context, nullptr,
+      nullptr, nullptr, kNoCompression, nullptr, &event_logger, true);
  mutex_.Lock();
  ASSERT_OK(flush_job.Run());
  mutex_.Unlock();
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -13,51 +13,69 @@ namespace rocksdb {

 void FlushScheduler::ScheduleFlush(ColumnFamilyData* cfd) {
 #ifndef NDEBUG
-  assert(column_families_set_.find(cfd) == column_families_set_.end());
-  column_families_set_.insert(cfd);
+  {
+    std::lock_guard<std::mutex> lock(checking_mutex_);
+    assert(checking_set_.count(cfd) == 0);
+    checking_set_.insert(cfd);
+  }
 #endif  // NDEBUG
  cfd->Ref();
-  column_families_.push_back(cfd);
+  Node* node = new Node{cfd, head_.load(std::memory_order_relaxed)};
+  while (!head_.compare_exchange_strong(
+      node->next, node, std::memory_order_relaxed, std::memory_order_relaxed)) {
+    // failing CAS updates the first param, so we are already set for
+    // retry.  TakeNextColumnFamily won't happen until after another
+    // inter-thread synchronization, so we don't even need release
+    // semantics for this CAS
+  }
 }

-ColumnFamilyData* FlushScheduler::GetNextColumnFamily() {
-  ColumnFamilyData* cfd = nullptr;
-  while (column_families_.size() > 0) {
-    cfd = column_families_.front();
-    column_families_.pop_front();
-    if (cfd->IsDropped()) {
-      if (cfd->Unref()) {
-        delete cfd;
-        cfd = nullptr;
-      }
-    } else {
-      break;
+ColumnFamilyData* FlushScheduler::TakeNextColumnFamily() {
+  while (true) {
+    if (Empty()) {
+      return nullptr;
    }
-  }
-#ifndef NDEBUG
-  if (cfd != nullptr) {
-    auto itr = column_families_set_.find(cfd);
-    assert(itr != column_families_set_.end());
-    column_families_set_.erase(itr);
-  }
-#endif  // NDEBUG
-  return cfd;
-}

-bool FlushScheduler::Empty() { return column_families_.empty(); }
+    // dequeue the head
+    Node* node = head_.load(std::memory_order_relaxed);
+    head_.store(node->next, std::memory_order_relaxed);
+    ColumnFamilyData* cfd = node->column_family;
+    delete node;

-void FlushScheduler::Clear() {
-  for (auto cfd : column_families_) {
 #ifndef NDEBUG
-    auto itr = column_families_set_.find(cfd);
-    assert(itr != column_families_set_.end());
-    column_families_set_.erase(itr);
+    {
+      auto iter = checking_set_.find(cfd);
+      assert(iter != checking_set_.end());
+      checking_set_.erase(iter);
+    }
 #endif  // NDEBUG
+
+    if (!cfd->IsDropped()) {
+      // success
+      return cfd;
+    }
+
+    // no longer relevant, retry
    if (cfd->Unref()) {
      delete cfd;
    }
  }
-  column_families_.clear();
+}
+
+bool FlushScheduler::Empty() {
+  auto rv = head_.load(std::memory_order_relaxed) == nullptr;
+  assert(rv == checking_set_.empty());
+  return rv;
+}
+
+void FlushScheduler::Clear() {
+  ColumnFamilyData* cfd;
+  while ((cfd = TakeNextColumnFamily()) != nullptr) {
+    if (cfd->Unref()) {
+      delete cfd;
+    }
+  }
+  assert(Empty());
 }

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -6,34 +6,42 @@
 #pragma once

 #include <stdint.h>
-#include <deque>
+#include <atomic>
+#include <mutex>
 #include <set>
-#include <vector>

 namespace rocksdb {

 class ColumnFamilyData;

-// This class is thread-compatible. It's should only be accessed from single
-// write thread (between BeginWrite() and EndWrite())
+// Unless otherwise noted, all methods on FlushScheduler should be called
+// only with the DB mutex held or from a single-threaded recovery context.
 class FlushScheduler {
 public:
-  FlushScheduler() = default;
-  ~FlushScheduler() = default;
+  FlushScheduler() : head_(nullptr) {}

+  // May be called from multiple threads at once, but not concurrent with
+  // any other method calls on this instance
  void ScheduleFlush(ColumnFamilyData* cfd);
-  // Returns Ref()-ed column family. Client needs to Unref()
-  // REQUIRES: db mutex is held (exception is single-threaded recovery)
-  ColumnFamilyData* GetNextColumnFamily();
+
+  // Removes and returns Ref()-ed column family. Client needs to Unref().
+  // Filters column families that have been dropped.
+  ColumnFamilyData* TakeNextColumnFamily();

  bool Empty();

  void Clear();

 private:
-  std::deque<ColumnFamilyData*> column_families_;
+  struct Node {
+    ColumnFamilyData* column_family;
+    Node* next;
+  };
+
+  std::atomic<Node*> head_;
 #ifndef NDEBUG
-  std::set<ColumnFamilyData*> column_families_set_;
+  std::mutex checking_mutex_;
+  std::set<ColumnFamilyData*> checking_set_;
 #endif  // NDEBUG
 };

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, 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.
@@ -10,15 +10,16 @@
 #include <string>
 #include <utility>

-#include "db/job_context.h"
+#include "db/column_family.h"
 #include "db/db_impl.h"
 #include "db/db_iter.h"
-#include "db/column_family.h"
+#include "db/dbformat.h"
+#include "db/job_context.h"
 #include "rocksdb/env.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/slice_transform.h"
 #include "table/merger.h"
-#include "db/dbformat.h"
+#include "util/string_util.h"
 #include "util/sync_point.h"

 namespace rocksdb {
@@ -144,6 +145,23 @@ 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_->~InternalIterator();
@@ -162,20 +180,7 @@ 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 +190,10 @@ 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 +211,10 @@ 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 +234,9 @@ 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
@@ -254,7 +263,7 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,
    }
    const VersionStorageInfo* vstorage = sv_->current->storage_info();
    const std::vector<FileMetaData*>& l0 = vstorage->LevelFiles(0);
-    for (uint32_t i = 0; i < l0.size(); ++i) {
+    for (size_t i = 0; i < l0.size(); ++i) {
      if (!l0_iters_[i]) {
        continue;
      }
@@ -393,7 +402,11 @@ 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;
@@ -459,6 +472,15 @@ Status ForwardIterator::status() const {
  return immutable_status_;
 }

+Status ForwardIterator::GetProperty(std::string prop_name, std::string* prop) {
+  assert(prop != nullptr);
+  if (prop_name == "rocksdb.iterator.super-version-number") {
+    *prop = ToString(sv_->version_number);
+    return Status::OK();
+  }
+  return Status::InvalidArgument();
+}
+
 void ForwardIterator::RebuildIterators(bool refresh_sv) {
  // Clean up
  Cleanup(refresh_sv);
@@ -484,10 +506,81 @@ 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);
+  size_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,
@@ -502,14 +595,11 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
          new LevelIterator(cfd_, read_options_, level_files));
    }
  }
-
-  current_ = nullptr;
-  is_prev_set_ = false;
 }

 void ForwardIterator::ResetIncompleteIterators() {
  const auto& l0_files = sv_->current->storage_info()->LevelFiles(0);
-  for (uint32_t i = 0; i < l0_iters_.size(); ++i) {
+  for (size_t i = 0; i < l0_iters_.size(); ++i) {
    assert(i < l0_files.size());
    if (!l0_iters_[i] || !l0_iters_[i]->status().IsIncomplete()) {
      continue;
@@ -600,7 +690,7 @@ bool ForwardIterator::NeedToSeekImmutable(const Slice& target) {
 void ForwardIterator::DeleteCurrentIter() {
  const VersionStorageInfo* vstorage = sv_->current->storage_info();
  const std::vector<FileMetaData*>& l0 = vstorage->LevelFiles(0);
-  for (uint32_t i = 0; i < l0.size(); ++i) {
+  for (size_t i = 0; i < l0.size(); ++i) {
    if (!l0_iters_[i]) {
      continue;
    }
@@ -632,7 +722,7 @@ bool ForwardIterator::TEST_CheckDeletedIters(int* pdeleted_iters,

  const VersionStorageInfo* vstorage = sv_->current->storage_info();
  const std::vector<FileMetaData*>& l0 = vstorage->LevelFiles(0);
-  for (uint32_t i = 0; i < l0.size(); ++i) {
+  for (size_t i = 0; i < l0.size(); ++i) {
    if (!l0_iters_[i]) {
      retval = true;
      deleted_iters++;