Disable error as warning

The previuos change that use gcc-5 set USE_SSE to wrong flag by mistake. Fix it.

.gitignore

@@ -32,6 +32,7 @@ build/
 ldb
 manifest_dump
 sst_dump
+column_aware_encoding_exp
 util/build_version.cc
 build_tools/VALGRIND_LOGS/
 coverage/COVERAGE_REPORT
@@ -66,3 +67,5 @@ t
 LOG
 
 db_logs/
+tp2/
+fbcode/

.travis.yml

@@ -9,7 +9,7 @@ compiler:
 addons:
    apt:
       sources: ['ubuntu-toolchain-r-test', 'llvm-toolchain-precise-3.6']
-      packages: ['clang-3.6' , 'zlib1g-dev', 'libbz2-dev', 'libsnappy-dev', 'curl']
+      packages: ['clang-3.6' , 'zlib1g-dev', 'libbz2-dev', 'libsnappy-dev', 'curl', 'gflags']
 env:
   # Run all tests before db_block_cache_test (db_test, db_test2)
   - JOB_NAME=unittests ROCKSDBTESTS_END=db_block_cache_test
@@ -26,15 +26,9 @@ install:
   - pushd /tmp/ && curl -L https://github.com/gflags/gflags/archive/v2.1.2.tar.gz -o gflags.tar.gz && tar xfz gflags.tar.gz && cd gflags-2.1.2 && cmake . && make && popd
 
 before_script:
-  # Add gflags to include/library paths
-  # TODO(noetzli): Remove when gflags available through Travis
-  - export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/tmp/gflags-2.1.2/lib"
-  - export LIBRARY_PATH="$LIBRARY_PATH:/tmp/gflags-2.1.2/lib"
-  - export CPLUS_INCLUDE_PATH="$CPLUS_INCLUDE_PATH:/tmp/gflags-2.1.2/include"
   - if [[ "${TRAVIS_OS_NAME}" == 'linux' ]]; then CXX=clang++-3.6; fi
-  - if [[ "${TRAVIS_OS_NAME}" == 'osx' ]]; then brew install gflags snappy; fi
-  # Limit the maximum number of open file descriptors to 2000
-  - ulimit -n 2000 || true
+  # Limit the maximum number of open file descriptors to 8192
+  - ulimit -n 8192 || true
 
 script:
   - if [[ "${JOB_NAME}" == 'unittests' ]]; then OPT=-DTRAVIS V=1 make -j4 check_some; fi

CMakeLists.txt

@@ -32,8 +32,8 @@ project(rocksdb)
 include(${CMAKE_CURRENT_SOURCE_DIR}/thirdparty.inc)
 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 ${TIME})
+string(REGEX REPLACE "(..)_(..)_..(..).*" "\\1/\\2/\\3" DATE "${DATE}")
+string(REGEX REPLACE "(..):(.....).*" " \\1:\\2" TIME "${TIME}")
 string(CONCAT GIT_DATE_TIME ${DATE} ${TIME})
 
 find_package(Git)
@@ -44,7 +44,7 @@ else()
     set(GIT_SHA 0)
 endif()
 
-string(REGEX REPLACE "[^0-9a-f]+" "" GIT_SHA ${GIT_SHA})
+string(REGEX REPLACE "[^0-9a-f]+" "" GIT_SHA "${GIT_SHA}")
 
 set(BUILD_VERSION_CC ${CMAKE_CURRENT_SOURCE_DIR}/util/build_version.cc)
 
@@ -193,6 +193,8 @@ set(SOURCES
         util/arena.cc
         util/bloom.cc
         util/build_version.cc
+        util/cf_options.cc
+        util/clock_cache.cc
         util/coding.cc
         util/compaction_job_stats_impl.cc
         util/comparator.cc
@@ -220,7 +222,6 @@ set(SOURCES
         util/log_buffer.cc
         util/memenv.cc
         util/murmurhash.cc
-        util/mutable_cf_options.cc
         util/options.cc
         util/options_helper.cc
         util/options_parser.cc
@@ -239,7 +240,7 @@ set(SOURCES
         util/testharness.cc
         util/testutil.cc
         util/thread_local.cc
-        util/threadpool.cc
+        util/threadpool_imp.cc
         util/thread_status_impl.cc
         util/thread_status_updater.cc
         util/thread_status_util.cc
@@ -248,8 +249,10 @@ set(SOURCES
         util/xfunc.cc
         util/xxhash.cc
         utilities/backupable/backupable_db.cc
+        utilities/blob_db/blob_db.cc
         utilities/checkpoint/checkpoint.cc
         utilities/compaction_filters/remove_emptyvalue_compactionfilter.cc
+        utilities/date_tiered/date_tiered_db_impl.cc
         utilities/document/document_db.cc
         utilities/document/json_document.cc
         utilities/document/json_document_builder.cc
@@ -264,7 +267,11 @@ set(SOURCES
         utilities/merge_operators/put.cc
         utilities/merge_operators/max.cc
         utilities/merge_operators/uint64add.cc
+        utilities/option_change_migration/option_change_migration.cc
         utilities/options/options_util.cc
+        utilities/persistent_cache/block_cache_tier.cc
+        utilities/persistent_cache/block_cache_tier_file.cc
+        utilities/persistent_cache/block_cache_tier_metadata.cc
         utilities/persistent_cache/persistent_cache_tier.cc
         utilities/persistent_cache/volatile_tier_impl.cc
         utilities/redis/redis_lists.cc
@@ -282,6 +289,9 @@ set(SOURCES
         utilities/ttl/db_ttl_impl.cc
         utilities/write_batch_with_index/write_batch_with_index.cc
         utilities/write_batch_with_index/write_batch_with_index_internal.cc
+        utilities/col_buf_encoder.cc
+        utilities/col_buf_decoder.cc
+        utilities/column_aware_encoding_util.cc
 )
 
 # For test util library that is build only in DEBUG mode
@@ -327,6 +337,7 @@ set(APPS
         tools/dump/rocksdb_undump.cc
         util/cache_bench.cc
         utilities/persistent_cache/hash_table_bench.cc
+        utilities/column_aware_encoding_exp.cc
 )
 
 set(C_TESTS db/c_test.c)
@@ -359,6 +370,7 @@ set(TESTS
         db/db_flush_test.cc
         db/db_iterator_test.cc
         db/db_sst_test.cc
+        db/external_sst_file_test.cc
         db/db_universal_compaction_test.cc
         db/db_wal_test.cc
         db/dbformat_test.cc
@@ -416,6 +428,7 @@ set(TESTS
         util/heap_test.cc
         util/histogram_test.cc
         util/iostats_context_test.cc
+        util/lru_cache_test.cc
         util/mock_env_test.cc
         util/options_settable_test.cc
         util/options_test.cc
@@ -425,13 +438,16 @@ set(TESTS
         util/thread_list_test.cc
         util/thread_local_test.cc
         utilities/backupable/backupable_db_test.cc
+        utilities/blob_db/blob_db_test.cc
         utilities/checkpoint/checkpoint_test.cc
+        utilities/date_tiered/date_tiered_test.cc
         utilities/document/document_db_test.cc
         utilities/document/json_document_test.cc
         utilities/env_registry_test.cc
         utilities/geodb/geodb_test.cc
         utilities/memory/memory_test.cc
         utilities/merge_operators/string_append/stringappend_test.cc
+        utilities/option_change_migration/option_change_migration_test.cc
         utilities/options/options_util_test.cc
         utilities/persistent_cache/hash_table_test.cc
         utilities/persistent_cache/persistent_cache_test.cc
@@ -442,6 +458,7 @@ set(TESTS
         utilities/transactions/transaction_test.cc
         utilities/ttl/ttl_test.cc
         utilities/write_batch_with_index/write_batch_with_index_test.cc
+        utilities/column_aware_encoding_test.cc
 )
 
 set(EXES ${APPS})

HISTORY.md

@@ -1,12 +1,25 @@
-## 4.10.2
-### Bug Fixes
-* backport the bug fix of the regression data corruption after DB recovery when stale WAL file deletion is reordered.
+# Rocksdb Change Log
+## 4.12.0 (9/12/2016)
+### Public API Change
+* CancelAllBackgroundWork() flushes all memtables for databases containing writes that have bypassed the WAL (writes issued with WriteOptions::disableWAL=true) before shutting down background threads.
+* Merge options source_compaction_factor, max_grandparent_overlap_bytes and expanded_compaction_factor into max_compaction_bytes.
+* Remove ImmutableCFOptions.
+* Add a compression type ZSTD, which can work with ZSTD 0.8.0 or up. Still keep ZSTDNotFinal for compatibility reasons.
 
-## 4.10.1
-### Bug Fixes
-* Fix the regression deadlock bug of DB recovery if options.max_successive_merges hits.
+### New Features
+* Introduce NewClockCache, which is based on CLOCK algorithm with better concurrent performance in some cases. It can be used to replace the default LRU-based block cache and table cache. To use it, RocksDB need to be linked with TBB lib.
+* Change ticker/histogram statistics implementations to accumulate data in thread-local storage, which improves CPU performance by reducing cache coherency costs. Callers of CreateDBStatistics do not need to change anything to use this feature.
+* Block cache mid-point insertion, where index and filter block are inserted into LRU block cache with higher priority. The feature can be enabled by setting BlockBasedTableOptions::cache_index_and_filter_blocks_with_high_priority to true and high_pri_pool_ratio > 0 when creating NewLRUCache.
 
-## 4.10.0
+## 4.11.0 (8/1/2016)
+### Public API Change
+* options.memtable_prefix_bloom_huge_page_tlb_size => memtable_huge_page_size. When it is set, RocksDB will try to allocate memory from huge page for memtable too, rather than just memtable bloom filter.
+
+### New Features
+* A tool to migrate DB after options change. See include/rocksdb/utilities/option_change_migration.h.
+* Add ReadOptions.background_purge_on_iterator_cleanup. If true, we avoid file deletion when destorying iterators.
+
+## 4.10.0 (7/5/2016)
 ### Public API Change
 * options.memtable_prefix_bloom_bits changes to options.memtable_prefix_bloom_bits_ratio and deprecate options.memtable_prefix_bloom_probes
 * enum type CompressionType and PerfLevel changes from char to unsigned char. Value of all PerfLevel shift by one.

Makefile

@@ -130,8 +130,12 @@ am__v_AR_ = $(am__v_AR_$(AM_DEFAULT_VERBOSITY))
 am__v_AR_0 = @echo "  AR      " $@;
 am__v_AR_1 =
 
-AM_LINK = $(AM_V_CCLD)$(CXX) $^ $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+ifdef ROCKSDB_USE_LIBRADOS
+LIB_SOURCES += utilities/env_librados.cc
+LDFLAGS += -lrados
+endif
 
+AM_LINK = $(AM_V_CCLD)$(CXX) $^ $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
 # detect what platform we're building on
 dummy := $(shell (export ROCKSDB_ROOT="$(CURDIR)"; "$(CURDIR)/build_tools/build_detect_platform" "$(CURDIR)/make_config.mk"))
 # this file is generated by the previous line to set build flags and sources
@@ -260,6 +264,8 @@ VALGRIND_OPTS = --error-exitcode=$(VALGRIND_ERROR) --leak-check=full
 
 BENCHTOOLOBJECTS = $(BENCH_LIB_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)
 
+EXPOBJECTS = $(EXP_LIB_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)
+
 TESTS = \
 	db_test \
 	db_test2 \
@@ -275,6 +281,7 @@ TESTS = \
 	db_iterator_test \
 	db_options_test \
 	db_sst_test \
+	external_sst_file_test \
 	db_tailing_iter_test \
 	db_universal_compaction_test \
 	db_wal_test \
@@ -326,7 +333,9 @@ TESTS = \
 	skiplist_test \
 	stringappend_test \
 	ttl_test \
+	date_tiered_test \
 	backupable_db_test \
+	blob_db_test \
 	document_db_test \
 	json_document_test \
 	sim_cache_test \
@@ -359,6 +368,7 @@ TESTS = \
 	compaction_job_test \
 	thread_list_test \
 	sst_dump_test \
+	column_aware_encoding_test \
 	compact_files_test \
 	perf_context_test \
 	optimistic_transaction_test \
@@ -366,19 +376,23 @@ TESTS = \
 	heap_test \
 	compact_on_deletion_collector_test \
 	compaction_job_stats_test \
+	option_change_migration_test \
 	transaction_test \
 	ldb_cmd_test \
 	iostats_context_test \
 	persistent_cache_test \
 	statistics_test \
+	lru_cache_test \
 
 PARALLEL_TEST = \
 	backupable_db_test \
-	compact_on_deletion_collector_test \
 	db_compaction_filter_test \
 	db_compaction_test \
+	db_sst_test \
 	db_test \
 	db_universal_compaction_test \
+	db_wal_test \
+	external_sst_file_test \
 	fault_injection_test \
 	inlineskiplist_test \
 	manual_compaction_test \
@@ -407,7 +421,7 @@ TEST_LIBS = \
 	librocksdb_env_basic_test.a
 
 # TODO: add back forward_iterator_bench, after making it build in all environemnts.
-BENCHMARKS = db_bench table_reader_bench cache_bench memtablerep_bench
+BENCHMARKS = db_bench table_reader_bench cache_bench memtablerep_bench column_aware_encoding_exp persistent_cache_bench
 
 # if user didn't config LIBNAME, set the default
 ifeq ($(LIBNAME),)
@@ -555,7 +569,7 @@ $(parallel_tests): $(PARALLEL_TEST)
       '#!/bin/sh' \
       "d=\$(TMPD)$$TEST_SCRIPT" \
       'mkdir -p $$d' \
-			"TEST_TMPDIR=\$$d ./$$TEST_BINARY --gtest_filter=$$TEST_NAME" \
+      "TEST_TMPDIR=\$$d $(DRIVER) ./$$TEST_BINARY --gtest_filter=$$TEST_NAME" \
 		> $$TEST_SCRIPT; \
 		chmod a=rx $$TEST_SCRIPT; \
 	done
@@ -582,7 +596,7 @@ gen_parallel_tests:
 # 107.816 PASS t/DBTest.EncodeDecompressedBlockSizeTest
 #
 slow_test_regexp = \
-	^t/run-table_test-HarnessTest.Randomized$$|^t/run-db_test-.*(?:FileCreationRandomFailure|EncodeDecompressedBlockSizeTest)$$
+	^t/run-table_test-HarnessTest.Randomized$$|^t/run-db_test-.*(?:FileCreationRandomFailure|EncodeDecompressedBlockSizeTest)$$|^.*RecoverFromCorruptedWALWithoutFlush$$
 prioritize_long_running_tests =						\
   perl -pe 's,($(slow_test_regexp)),100 $$1,'				\
     | sort -k1,1gr							\
@@ -612,7 +626,7 @@ check_0:
 	} \
 	  | $(prioritize_long_running_tests)				\
 	  | grep -E '$(tests-regexp)'					\
-	  | parallel -j$(J) --joblog=LOG $$eta --gnu '{} >& t/log-{/}'
+	  | build_tools/gnu_parallel -j$(J) --plain --joblog=LOG $$eta --gnu '{} >& t/log-{/}'
 
 .PHONY: valgrind_check_0
 valgrind_check_0:
@@ -627,9 +641,9 @@ valgrind_check_0:
 	}								\
 	  | $(prioritize_long_running_tests)				\
 	  | grep -E '$(tests-regexp)'					\
-	  | parallel -j$(J) --joblog=LOG $$eta --gnu \
-      'if [[ "{}" == "./"* ]] ; then $(DRIVER) {} >& t/valgrind_log-{/}; ' \
-      'else {} >& t/valgrind_log-{/}; fi'
+	  | build_tools/gnu_parallel -j$(J) --plain --joblog=LOG $$eta --gnu \
+	  '(if [[ "{}" == "./"* ]] ; then $(DRIVER) {}; else {}; fi) ' \
+	  '>& t/valgrind_log-{/}'
 
 CLEAN_FILES += t LOG $(TMPD)
 
@@ -648,7 +662,7 @@ watch-log:
 check: all
 	$(MAKE) gen_parallel_tests
 	$(AM_V_GEN)if test "$(J)" != 1                                  \
-	    && (parallel --gnu --help 2>/dev/null) |                    \
+	    && (build_tools/gnu_parallel --gnu --help 2>/dev/null) |                    \
 	        grep -q 'GNU Parallel';                                 \
 	then                                                            \
 	    $(MAKE) T="$$t" TMPD=$(TMPD) check_0;                       \
@@ -675,9 +689,15 @@ blackbox_crash_test: db_stress
 	python -u tools/db_crashtest.py --simple blackbox
 	python -u tools/db_crashtest.py blackbox
 
+ifeq ($(CRASH_TEST_KILL_ODD),)
+  CRASH_TEST_KILL_ODD=888887
+endif
+
 whitebox_crash_test: db_stress
-	python -u tools/db_crashtest.py --simple whitebox
-	python -u tools/db_crashtest.py whitebox
+	python -u tools/db_crashtest.py --simple whitebox --random_kill_odd \
+      $(CRASH_TEST_KILL_ODD)
+	python -u tools/db_crashtest.py whitebox  --random_kill_odd \
+      $(CRASH_TEST_KILL_ODD)
 
 asan_check:
 	$(MAKE) clean
@@ -700,9 +720,9 @@ ubsan_crash_test:
 	$(MAKE) clean
 
 valgrind_check: $(TESTS)
-	$(MAKE) gen_parallel_tests
+	$(MAKE) DRIVER="$(VALGRIND_VER) $(VALGRIND_OPTS)" gen_parallel_tests
 	$(AM_V_GEN)if test "$(J)" != 1                                  \
-	    && (parallel --gnu --help 2>/dev/null) |                    \
+	    && (build_tools/gnu_parallel --gnu --help 2>/dev/null) |                    \
 	        grep -q 'GNU Parallel';                                 \
 	then                                                            \
       $(MAKE) TMPD=$(TMPD)                                        \
@@ -724,10 +744,10 @@ parloop:
 	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=$$s;' \
+			seq $(J) | v="$$t" build_tools/gnu_parallel --gnu --plain '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-{};' \
+			seq $(J) | v="./$$t" build_tools/gnu_parallel --gnu --plain 's=$(TMPD)/rdb-{};' \
 			     'export TEST_TMPDIR=$$s; timeout 10m $$v >> $$s/log-{} 2>1'; \
 		fi; \
 		ret_code=$$?; \
@@ -748,7 +768,7 @@ test_names = \
 
 parallel_check: $(TESTS)
 	$(AM_V_GEN)if test "$(J)" > 1                                  \
-	    && (parallel --gnu --help 2>/dev/null) |                    \
+	    && (build_tools/gnu_parallel --gnu --help 2>/dev/null) |                    \
 	        grep -q 'GNU Parallel';                                 \
 	then                                                            \
 	    echo Running in parallel $(J);			\
@@ -758,7 +778,7 @@ parallel_check: $(TESTS)
 	ret_bad=0;							\
 	echo $(J);\
 	echo Test Dir: $(TMPD); \
-        seq $(J) | parallel --gnu 's=$(TMPD)/rdb-{}; rm -rf $$s; mkdir $$s'; \
+        seq $(J) | build_tools/gnu_parallel --gnu --plain 's=$(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))" \
@@ -797,6 +817,7 @@ clean:
 	find . -name "*.[oda]" -exec rm -f {} \;
 	find . -type f -regex ".*\.\(\(gcda\)\|\(gcno\)\)" -exec rm {} \;
 	rm -rf bzip2* snappy* zlib* lz4*
+	cd java; $(MAKE) clean
 
 tags:
 	ctags * -R
@@ -829,6 +850,9 @@ db_bench: tools/db_bench.o $(BENCHTOOLOBJECTS)
 cache_bench: util/cache_bench.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)
 
+persistent_cache_bench: utilities/persistent_cache/persistent_cache_bench.o $(LIBOBJECTS) $(TESTUTIL)
+	$(AM_LINK)
+
 memtablerep_bench: db/memtablerep_bench.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)
 
@@ -850,7 +874,7 @@ arena_test: util/arena_test.o $(LIBOBJECTS) $(TESTHARNESS)
 autovector_test: util/autovector_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
-column_family_test: db/column_family_test.o $(LIBOBJECTS) $(TESTHARNESS)
+column_family_test: db/column_family_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
 table_properties_collector_test: db/table_properties_collector_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -871,6 +895,9 @@ cache_test: util/cache_test.o $(LIBOBJECTS) $(TESTHARNESS)
 coding_test: util/coding_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+option_change_migration_test: utilities/option_change_migration/option_change_migration_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 stringappend_test: utilities/merge_operators/string_append/stringappend_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -934,6 +961,9 @@ db_options_test: db/db_options_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARN
 db_sst_test: db/db_sst_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+external_sst_file_test: db/external_sst_file_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_tailing_iter_test: db/db_tailing_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -976,6 +1006,9 @@ prefix_test: db/prefix_test.o $(LIBOBJECTS) $(TESTHARNESS)
 backupable_db_test: utilities/backupable/backupable_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+blob_db_test: utilities/blob_db/blob_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 checkpoint_test: utilities/checkpoint/checkpoint_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -994,12 +1027,20 @@ spatial_db_test: utilities/spatialdb/spatial_db_test.o $(LIBOBJECTS) $(TESTHARNE
 env_mirror_test: utilities/env_mirror_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+ifdef ROCKSDB_USE_LIBRADOS
+env_librados_test: utilities/env_librados_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_V_CCLD)$(CXX) $^ $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+endif
+
 env_registry_test: utilities/env_registry_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
 ttl_test: utilities/ttl/ttl_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+date_tiered_test: utilities/date_tiered/date_tiered_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 write_batch_with_index_test: utilities/write_batch_with_index/write_batch_with_index_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -1156,6 +1197,9 @@ event_logger_test: util/event_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
 sst_dump_test: tools/sst_dump_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+column_aware_encoding_test: utilities/column_aware_encoding_test.o $(TESTHARNESS) $(EXPOBJECTS)
+	$(AM_LINK)
+
 optimistic_transaction_test: utilities/transactions/optimistic_transaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -1186,6 +1230,9 @@ transaction_test: utilities/transactions/transaction_test.o $(LIBOBJECTS) $(TEST
 sst_dump: tools/sst_dump.o $(LIBOBJECTS)
 	$(AM_LINK)
 
+column_aware_encoding_exp: utilities/column_aware_encoding_exp.o $(EXPOBJECTS)
+	$(AM_LINK)
+
 repair_test: db/repair_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -1204,6 +1251,10 @@ persistent_cache_test: utilities/persistent_cache/persistent_cache_test.o  db/db
 statistics_test: util/statistics_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+lru_cache_test: util/lru_cache_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+
 #-------------------------------------------------
 # make install related stuff
 INSTALL_PATH ?= /usr/local
@@ -1274,29 +1325,29 @@ endif
 
 libz.a:
 	-rm -rf zlib-1.2.8
-	curl -O http://zlib.net/zlib-1.2.8.tar.gz
+	curl -O -L http://zlib.net/zlib-1.2.8.tar.gz
 	tar xvzf zlib-1.2.8.tar.gz
 	cd zlib-1.2.8 && CFLAGS='-fPIC' ./configure --static && make
 	cp zlib-1.2.8/libz.a .
 
 libbz2.a:
 	-rm -rf bzip2-1.0.6
-	curl -O  http://www.bzip.org/1.0.6/bzip2-1.0.6.tar.gz
+	curl -O -L http://www.bzip.org/1.0.6/bzip2-1.0.6.tar.gz
 	tar xvzf bzip2-1.0.6.tar.gz
 	cd bzip2-1.0.6 && make CFLAGS='-fPIC -O2 -g -D_FILE_OFFSET_BITS=64'
 	cp bzip2-1.0.6/libbz2.a .
 
 libsnappy.a:
-	-rm -rf snappy-1.1.1
-	curl -O https://snappy.googlecode.com/files/snappy-1.1.1.tar.gz
-	tar xvzf snappy-1.1.1.tar.gz
-	cd snappy-1.1.1 && ./configure --with-pic --enable-static
-	cd snappy-1.1.1 && make
-	cp snappy-1.1.1/.libs/libsnappy.a .
+	-rm -rf snappy-1.1.3
+	curl -O -L https://github.com/google/snappy/releases/download/1.1.3/snappy-1.1.3.tar.gz
+	tar xvzf snappy-1.1.3.tar.gz
+	cd snappy-1.1.3 && ./configure --with-pic --enable-static
+	cd snappy-1.1.3 && make
+	cp snappy-1.1.3/.libs/libsnappy.a .
 
 liblz4.a:
 	   -rm -rf lz4-r127
-	   curl -O https://codeload.github.com/Cyan4973/lz4/tar.gz/r127
+	   curl -O -L https://codeload.github.com/Cyan4973/lz4/tar.gz/r127
 	   mv r127 lz4-r127.tar.gz
 	   tar xvzf lz4-r127.tar.gz
 	   cd lz4-r127/lib && make CFLAGS='-fPIC' all
@@ -1307,7 +1358,7 @@ java_static_libobjects = $(patsubst %,jls/%,$(LIBOBJECTS))
 CLEAN_FILES += jls
 
 JAVA_STATIC_FLAGS = -DZLIB -DBZIP2 -DSNAPPY -DLZ4
-JAVA_STATIC_INCLUDES = -I./zlib-1.2.8 -I./bzip2-1.0.6 -I./snappy-1.1.1 -I./lz4-r127/lib
+JAVA_STATIC_INCLUDES = -I./zlib-1.2.8 -I./bzip2-1.0.6 -I./snappy-1.1.3 -I./lz4-r127/lib
 
 $(java_static_libobjects): jls/%.o: %.cc libz.a libbz2.a libsnappy.a liblz4.a
 	$(AM_V_CC)mkdir -p $(@D) && $(CXX) $(CXXFLAGS) $(JAVA_STATIC_FLAGS) $(JAVA_STATIC_INCLUDES) -fPIC -c $< -o $@ $(COVERAGEFLAGS)
@@ -1358,6 +1409,12 @@ rocksdbjava: $(java_libobjects)
 jclean:
 	cd java;$(MAKE) clean;
 
+jtest_compile: rocksdbjava
+	cd java;$(MAKE) java_test
+
+jtest_run:
+	cd java;$(MAKE) run_test
+
 jtest: rocksdbjava
 	cd java;$(MAKE) sample;$(MAKE) test;
 
@@ -1366,7 +1423,7 @@ jdb_bench:
 
 commit_prereq: build_tools/rocksdb-lego-determinator \
                build_tools/precommit_checker.py
-	J=$(J) build_tools/precommit_checker.py unit unit_481 clang_unit tsan asan ubsan lite
+	J=$(J) build_tools/precommit_checker.py unit unit_481 clang_unit release release_481 clang_release tsan asan ubsan lite unit_non_shm
 	$(MAKE) clean && $(MAKE) jclean && $(MAKE) rocksdbjava;
 
 xfunc:
@@ -1414,7 +1471,7 @@ endif
 #  	Source files dependencies detection
 # ---------------------------------------------------------------------------
 
-all_sources = $(LIB_SOURCES) $(MAIN_SOURCES) $(MOCK_LIB_SOURCES) $(TOOL_LIB_SOURCES) $(BENCH_LIB_SOURCES) $(TEST_LIB_SOURCES)
+all_sources = $(LIB_SOURCES) $(MAIN_SOURCES) $(MOCK_LIB_SOURCES) $(TOOL_LIB_SOURCES) $(BENCH_LIB_SOURCES) $(TEST_LIB_SOURCES) $(EXP_LIB_SOURCES)
 DEPFILES = $(all_sources:.cc=.d)
 
 # Add proper dependency support so changing a .h file forces a .cc file to

README.md

@@ -1,6 +1,8 @@
 ## RocksDB: A Persistent Key-Value Store for Flash and RAM Storage
 
 [![Build Status](https://travis-ci.org/facebook/rocksdb.svg?branch=master)](https://travis-ci.org/facebook/rocksdb)
+[![Build status](https://ci.appveyor.com/api/projects/status/fbgfu0so3afcno78/branch/master?svg=true)](https://ci.appveyor.com/project/Facebook/rocksdb/branch/master)
+
 
 RocksDB is developed and maintained by Facebook Database Engineering Team.
 It is built on earlier work on LevelDB by Sanjay Ghemawat (sanjay@google.com)

appveyor.yml

@@ -1,8 +1,9 @@
 version: 1.0.{build}
+image: Visual Studio 2015
 before_build:
 - md %APPVEYOR_BUILD_FOLDER%\build
 - cd %APPVEYOR_BUILD_FOLDER%\build
-- cmake -G "Visual Studio 12 Win64" -DOPTDBG=1 -DXPRESS=1 ..
+- cmake -G "Visual Studio 14 2015 Win64" -DOPTDBG=1 -DXPRESS=1 ..
 - cd ..
 build:
   project: build\rocksdb.sln

arcanist_util/__phutil_library_map__.php

@@ -46,7 +46,7 @@ if (class_exists('ArcanistWorkflow')) {
       'BaseDirectoryScopedFormatLinter' => 'cpp_linter/BaseDirectoryScopedFormatLinter.php',
       'FacebookArcanistConfiguration' => 'config/FacebookOldArcanistConfiguration.php',
       'FacebookFbcodeLintEngine' => 'lint_engine/FacebookFbcodeLintEngine.php',
-      'FacebookFbcodeUnitTestEngine' => 'unit_engine/FacebookFbcodeUnitTestEngine.php',
+      'FacebookFbcodeUnitTestEngine' => 'unit_engine/FacebookOldFbcodeUnitTestEngine.php',
       'FacebookHowtoevenLintEngine' => 'lint_engine/FacebookHowtoevenLintEngine.php',
       'FacebookHowtoevenLinter' => 'cpp_linter/FacebookHowtoevenLinter.php',
       'FbcodeClangFormatLinter' => 'cpp_linter/FbcodeClangFormatLinter.php',

arcanist_util/config/RocksDBCommonHelper.php

@@ -62,7 +62,9 @@ function getSteps($applyDiff, $diffID, $username, $test) {
     assert(strlen($diffID) > 0);
     assert(is_numeric($diffID));
 
-    $arcrc_content = exec("cat ~/.arcrc | gzip -f | base64 -w0");
+    $arcrc_content = (PHP_OS == "Darwin" ?
+        exec("cat ~/.arcrc | gzip -f | base64") :
+            exec("cat ~/.arcrc | gzip -f | base64 -w0"));
     assert(strlen($arcrc_content) > 0);
 
     // Sandcastle machines don't have arc setup. We copy the user certificate
@@ -75,9 +77,10 @@ function getSteps($applyDiff, $diffID, $username, $test) {
     );
 
     // arc demands certain permission on its config.
+    // also fix the sticky bit issue in sandcastle
     $fix_permission = array(
       "name" => "Fix environment",
-      "shell" => "chmod 600 ~/.arcrc",
+      "shell" => "chmod 600 ~/.arcrc && chmod +t /dev/shm",
       "user" => "root"
     );
 
@@ -115,7 +118,7 @@ function getSteps($applyDiff, $diffID, $username, $test) {
   }
 
   // Run the actual command.
-  $cmd = $cmd . "./build_tools/precommit_checker.py " . $test
+  $cmd = $cmd . "J=$(nproc) ./build_tools/precommit_checker.py " . $test
            . "; exit_code=$?; ";
 
   if ($applyDiff) {
@@ -125,15 +128,31 @@ function getSteps($applyDiff, $diffID, $username, $test) {
                 . "; ";
   }
 
+  // shell command to sort the tests based on exit code and print
+  // the output of the log files.
+  $cat_sorted_logs = "
+    while read code log_file;
+      do echo \"################ cat \$log_file [exit_code : \$code] ################\";
+      cat \$log_file;
+    done < <(tail -n +2 LOG | sort -k7,7n -k4,4gr | awk '{print \$7,\$NF}')";
+
+  // Shell command to cat all log files
+  $cat_all_logs = "for f in `ls t/!(run-*)`; do echo \$f;cat \$f; done";
+
+  // If LOG file exist use it to cat log files sorted by exit code, otherwise
+  // cat everything
+  $logs_cmd = "if [ -f LOG ]; then {$cat_sorted_logs}; else {$cat_all_logs}; fi";
+
   $cmd = $cmd . " cat /tmp/precommit-check.log"
-           . "; for f in `ls t/log-*`; do echo \$f; cat \$f; done;"
-           . "[[ \$exit_code -eq 0 ]]";
+              . "; shopt -s extglob; {$logs_cmd}"
+              . "; shopt -u extglob; [[ \$exit_code -eq 0 ]]";
   assert(strlen($cmd) > 0);
 
   $run_test = array(
     "name" => "Run " . $test,
     "shell" => $cmd,
     "user" => "root",
+    "parser" => "python build_tools/error_filter.py " . $test,
   );
 
   $steps[] = $run_test;
@@ -212,21 +231,11 @@ function getSandcastleConfig() {
     assert(is_numeric($diffID));
   }
 
-  if (strcmp(getenv("ROCKSDB_CHECK_ALL"), 1) == 0) {
-    // Extract all tests from the CI definition.
-    $output = file_get_contents("build_tools/rocksdb-lego-determinator");
-    assert(strlen($output) > 0);
-
-    preg_match_all('/[ ]{2}([a-zA-Z0-9_]+)[\)]{1}/', $output, $matches);
-    $tests = $matches[1];
-    assert(count($tests) > 0);
-  } else {
-    // Manually list of tests we want to run in Sandcastle.
-    $tests = array(
-      "unit", "unit_481", "clang_unit", "tsan", "asan", "lite_test",
-      "valgrind", "release"
-    );
-  }
+  // List of tests we want to run in Sandcastle.
+  $tests = array("unit", "unit_non_shm", "unit_481", "clang_unit", "tsan",
+                 "asan", "lite_test", "valgrind", "release", "release_481",
+                 "clang_release", "punit", "clang_analyze", "code_cov",
+                 "java_build", "no_compression", "unity", "ubsan");
 
   $send_email_template = array(
     'type' => 'email',
@@ -255,7 +264,9 @@ function getSandcastleConfig() {
   // execute. Why compress the job definitions? Otherwise we run over the max
   // string size.
   $cmd = "echo " . base64_encode(json_encode($arg))
-         . " | gzip -f | base64 -w0";
+         . (PHP_OS == "Darwin" ?
+             " | gzip -f | base64" :
+                 " | gzip -f | base64 -w0");
   assert(strlen($cmd) > 0);
 
   $arg_encoded = shell_exec($cmd);

arcanist_util/unit_engine/FacebookFbcodeUnitTestEngine.php

@@ -4,7 +4,7 @@
 // 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.
 
-class FacebookFbcodeUnitTestEngine extends ArcanistBaseUnitTestEngine {
+class FacebookFbcodeUnitTestEngine extends ArcanistUnitTestEngine {
 
   public function run() {
       // For a call to `arc call-conduit differential.updateunitresults` to

arcanist_util/unit_engine/FacebookOldFbcodeUnitTestEngine.php

@@ -0,0 +1,17 @@
+<?php
+// Copyright 2004-present Facebook. 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.
+
+class FacebookFbcodeUnitTestEngine extends ArcanistBaseUnitTestEngine {
+
+  public function run() {
+      // For a call to `arc call-conduit differential.updateunitresults` to
+      // succeed we need at least one entry here.
+      $result = new ArcanistUnitTestResult();
+      $result->setName("dummy_placeholder_entry");
+      $result->setResult(ArcanistUnitTestResult::RESULT_PASS);
+      return array($result);
+  }
+}

build_tools/build_detect_platform

@@ -19,22 +19,21 @@
 #
 # The PLATFORM_CCFLAGS and PLATFORM_CXXFLAGS might include the following:
 #
-#       -DLEVELDB_PLATFORM_POSIX if cstdatomic is present
-#       -DLEVELDB_PLATFORM_NOATOMIC if it is not
+#       -DROCKSDB_PLATFORM_POSIX    if posix-platform based
 #       -DSNAPPY                    if the Snappy library is present
 #       -DLZ4                       if the LZ4 library is present
 #       -DZSTD                      if the ZSTD library is present
 #       -DNUMA                      if the NUMA library is present
+#       -DTBB                       if the TBB library is present
 #
 # Using gflags in rocksdb:
 # Our project depends on gflags, which requires users to take some extra steps
 # before they can compile the whole repository:
 #   1. Install gflags. You may download it from here:
-#      https://code.google.com/p/gflags/
-#   2. Once install, add the include path/lib path for gflags to CPATH and
-#      LIBRARY_PATH respectively. If installed with default mode, the
-#      lib and include path will be /usr/local/lib and /usr/local/include
-# Mac user can do this by having brew installed and running brew install gflags
+#      https://gflags.github.io/gflags/ (Mac users can `brew install gflags`)
+#   2. Once installed, add the include path for gflags to your CPATH env var and
+#      the lib path to LIBRARY_PATH. If installed with default settings, the lib
+#      will be /usr/local/lib and the include path will be /usr/local/include
 
 OUTPUT=$1
 if test -z "$OUTPUT"; then
@@ -165,7 +164,7 @@ case "$TARGET_OS" in
         ;;
     OS_ANDROID_CROSSCOMPILE)
         PLATFORM=OS_ANDROID
-	COMMON_FLAGS="$COMMON_FLAGS -fno-builtin-memcmp -D_REENTRANT -DOS_ANDROID -DLEVELDB_PLATFORM_POSIX"
+	COMMON_FLAGS="$COMMON_FLAGS -fno-builtin-memcmp -D_REENTRANT -DOS_ANDROID -DROCKSDB_PLATFORM_POSIX"
 	PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS "  # All pthread features are in the Android C library
         # PORT_FILES=port/android/android.cc
         CROSS_COMPILE=true
@@ -283,7 +282,7 @@ EOF
     # Test whether numa is available
     $CXX $CFLAGS -x c++ - -o /dev/null -lnuma 2>/dev/null  <<EOF
       #include <numa.h>
-      #inlcude <numaif.h>
+      #include <numaif.h>
       int main() {}
 EOF
     if [ "$?" = 0 ]; then
@@ -292,6 +291,17 @@ EOF
         JAVA_LDFLAGS="$JAVA_LDFLAGS -lnuma"
     fi
 
+    # Test whether tbb is available
+    $CXX $CFLAGS $LDFLAGS -x c++ - -o /dev/null -ltbb 2>/dev/null  <<EOF
+      #include <tbb/tbb.h>
+      int main() {}
+EOF
+    if [ "$?" = 0 ]; then
+        COMMON_FLAGS="$COMMON_FLAGS -DTBB"
+        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -ltbb"
+        JAVA_LDFLAGS="$JAVA_LDFLAGS -ltbb"
+    fi
+
     # Test whether jemalloc is available
     if echo 'int main() {}' | $CXX $CFLAGS -x c++ - -o /dev/null -ljemalloc \
       2>/dev/null; then

build_tools/dependencies_4.8.1.sh

@@ -1,16 +1,17 @@
-GCC_BASE=/mnt/gvfs/third-party2/gcc/ebc96bc2fb751b5a0300b8d91a95bdf24ac1d88b/4.8.1/centos6-native/cc6c9dc
-CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/b91de48a4974ec839946d824402b098d43454cef/stable/centos6-native/7aaccbe
-LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/53e0eac8911888a105aa98b9a35fe61cf1d8b278/4.8.1/gcc-4.8.1-glibc-2.17/8aac7fc
-GLIBC_BASE=/mnt/gvfs/third-party2/glibc/ee36ac9a72dfac4a995f1b215bb4c0fc8a0f6f91/2.17/gcc-4.8.1-glibc-2.17/99df8fc
+GCC_BASE=/mnt/gvfs/third-party2/gcc/cf7d14c625ce30bae1a4661c2319c5a283e4dd22/4.8.1/centos6-native/cc6c9dc
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/8598c375b0e94e1448182eb3df034704144a838d/stable/centos6-native/3f16ddd
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/d6e0a7da6faba45f5e5b1638f9edd7afc2f34e7d/4.8.1/gcc-4.8.1-glibc-2.17/8aac7fc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/d282e6e8f3d20f4e40a516834847bdc038e07973/2.17/gcc-4.8.1-glibc-2.17/99df8fc
 SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/8c38a4c1e52b4c2cc8a9cdc31b9c947ed7dbfcb4/1.1.3/gcc-4.8.1-glibc-2.17/c3f970a
-ZLIB_BASE=/mnt/gvfs/third-party2/zlib/2b24f1e99454f9ca7b0301720f94836dae1bf71b/1.2.8/gcc-4.8.1-glibc-2.17/c3f970a
-BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/af7c14c9b652cdd5ec34eadd25c3f38a9b306c5d/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/0882df3713c7a84f15abe368dc004581f20b39d7/1.2.8/gcc-4.8.1-glibc-2.17/c3f970a
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/740325875f6729f42d28deaa2147b0854f3a347e/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a
 LZ4_BASE=/mnt/gvfs/third-party2/lz4/0e790b441e2d9acd68d51e1d2e028f88c6a79ddf/r131/gcc-4.8.1-glibc-2.17/c3f970a
-ZSTD_BASE=/mnt/gvfs/third-party2/zstd/1408484d03b15492aa54b10356104e9dc22e1cc5/0.6.1/gcc-4.8.1-glibc-2.17/c3f970a
-GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/5a5c7a6608cb32f1e1e7f814023d5bdfbd136370/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a
-JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/a4c2adecffcaa68d5585d06be2252e3efa52555b/master/gcc-4.8.1-glibc-2.17/8d31e51
-NUMA_BASE=/mnt/gvfs/third-party2/numa/1abc0d3c01743b854676423cf2d3629912f34930/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/9455f75ff7f4831dc9fda02a6a0f8c68922fad8f/1.0.0/gcc-4.8.1-glibc-2.17/c3f970a
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/f001a51b2854957676d07306ef3abf67186b5c8b/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/fc8a13ca1fffa4d0765c716c5a0b49f0c107518f/master/gcc-4.8.1-glibc-2.17/8d31e51
+NUMA_BASE=/mnt/gvfs/third-party2/numa/17c514c4d102a25ca15f4558be564eeed76f4b6a/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a
 LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/ad576de2a1ea560c4d3434304f0fc4e079bede42/trunk/gcc-4.8.1-glibc-2.17/675d945
-KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/f24be37d170e04be6e469af487644d4d62e1c6c1/4.0.9-36_fbk5_2933_gd092e3f/gcc-4.8.1-glibc-2.17/da39a3e
-BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/9d65c666b9adf8f2a989fd4b98a9a5e7d3afa233/2.26/centos6-native/da39a3e
-VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/9cee5a3628dc9d4b93897972c58eba865e25b270/3.8.1/gcc-4.8.1-glibc-2.17/c3f970a
+TBB_BASE=/mnt/gvfs/third-party2/tbb/9d9a554877d0c5bef330fe818ab7178806dd316a/4.0_update2/gcc-4.8.1-glibc-2.17/c3f970a
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/7c111ff27e0c466235163f00f280a9d617c3d2ec/4.0.9-36_fbk5_2933_gd092e3f/gcc-4.8.1-glibc-2.17/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/b7fd454c4b10c6a81015d4524ed06cdeab558490/2.26/centos6-native/da39a3e
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d7f4d4d86674a57668e3a96f76f0e17dd0eb8765/3.8.1/gcc-4.8.1-glibc-2.17/c3f970a

build_tools/error_filter.py

@@ -119,7 +119,9 @@ _TEST_NAME_TO_PARSERS = {
     'unit': [CompilerErrorParser, GTestErrorParser],
     'release': [CompilerErrorParser, GTestErrorParser],
     'unit_481': [CompilerErrorParser, GTestErrorParser],
+    'release_481': [CompilerErrorParser, GTestErrorParser],
     'clang_unit': [CompilerErrorParser, GTestErrorParser],
+    'clang_release': [CompilerErrorParser, GTestErrorParser],
     'clang_analyze': [CompilerErrorParser, ScanBuildErrorParser],
     'code_cov': [CompilerErrorParser, GTestErrorParser],
     'unity': [CompilerErrorParser, GTestErrorParser],

build_tools/fbcode_config4.8.1.sh

@@ -48,13 +48,17 @@ NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"
 # location of libunwind
 LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
 
+# location of tbb
+TBB_INCLUDE=" -isystem $TBB_BASE/include/"
+TBB_LIBS="$TBB_BASE/lib/libtbb.a"
+
 # use Intel SSE support for checksum calculations
 export USE_SSE=1
 
 BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"
 
-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP2_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 $TBB_INCLUDE"
 
 STDLIBS="-L $GCC_BASE/lib64"
 
@@ -90,13 +94,15 @@ fi
 
 CFLAGS+=" $DEPS_INCLUDE"
 CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
-CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4 -DZSTD -DNUMA"
+CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4 -DZSTD -DNUMA -DTBB"
 CXXFLAGS+=" $CFLAGS"
 
-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP2_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 $TBB_LIBS"
 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"
+# required by libtbb
+EXEC_LDFLAGS+=" -ldl"
 
 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
 

build_tools/precommit_checker.py

@@ -94,7 +94,7 @@ class PreCommitChecker(Env):
     #
     def get_commands(self, test):
         status, out = self.GetOutput(
-            "build_tools/rocksdb-lego-determinator %s" % test, ".")
+            "RATIO=1 build_tools/rocksdb-lego-determinator %s" % test, ".")
         return status, out
 
     #

build_tools/rocksdb-lego-determinator

@@ -8,7 +8,7 @@
 # Input         Value
 # -------------------------------------------------------------------------
 # EMAIL         Email address to report on trigger conditions
-# ONCAL         Email address to raise a task on failure
+# ONCALL        Email address to raise a task on failure
 # TRIGGER       Trigger conditions for email. Valid values are fail, warn, all
 # SUBSCRIBER    Email addresss to add as subscriber for task
 #
@@ -42,6 +42,9 @@ if [ ! -z $ONCALL ]; then
   },"
 fi
 
+# For now, create the tasks using only the dedicated task creation tool.
+CREATE_TASK=
+
 REPORT=
 if [[ ! -z $REPORT_EMAIL || ! -z $CREATE_TASK ]]; then
   REPORT="'report': [
@@ -56,10 +59,26 @@ fi
 CLEANUP_ENV="
 {
     'name':'Cleanup environment',
-    'shell':'rm -rf /dev/shm/rocksdb && mkdir /dev/shm/rocksdb && make clean',
+    'shell':'rm -rf /dev/shm/rocksdb && mkdir /dev/shm/rocksdb && (chmod +t /dev/shm || true)  && make clean',
     'user':'root'
 }"
 
+# We will eventually set the RATIO to 1, but we want do this
+# in steps. RATIO=$(nproc) will make it work as J=1
+if [ -z $RATIO ]; then
+  RATIO=$(nproc)
+fi
+
+if [ -z $PARALLEL_J ]; then
+  PARALLEL_J="J=$(expr $(nproc) / ${RATIO})"
+fi
+
+if [ -z $PARALLEL_j ]; then
+  PARALLEL_j="-j$(expr $(nproc) / ${RATIO})"
+fi
+
+PARALLELISM="$PARALLEL_J $PARALLEL_j"
+
 DEBUG="OPT=-g"
 SHM="TEST_TMPDIR=/dev/shm/rocksdb"
 GCC_481="ROCKSDB_FBCODE_BUILD_WITH_481=1"
@@ -73,6 +92,19 @@ HTTP_PROXY="https_proxy=http://fwdproxy.29.prn1:8080 http_proxy=http://fwdproxy.
 SETUP_JAVA_ENV="export $HTTP_PROXY; export JAVA_HOME=/usr/local/jdk-7u10-64/; export PATH=\$PATH:\$JAVA_HOME/bin"
 PARSER="'parser':'python build_tools/error_filter.py $1'"
 
+CONTRUN_NAME="ROCKSDB_CONTRUN_NAME"
+
+# This code is getting called under various scenarios. What we care about is to
+# understand when it's called from nightly contruns because in that case we'll
+# create tasks for any failures. To follow the existing pattern, we'll check
+# the value of $ONCALL. If it's a diff then just call `false` to make sure
+# that errors will be properly propagated to the caller.
+if [ ! -z $ONCALL ]; then
+  TASK_CREATION_TOOL="/usr/local/bin/mysql_mtr_filter --rocksdb"
+else
+  TASK_CREATION_TOOL="false"
+fi
+
 ARTIFACTS=" 'artifacts': [
     {
         'name':'database',
@@ -110,14 +142,14 @@ PARALLEL_UNIT_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build and test RocksDB debug version',
-                'shell':'$DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || cat t/log-*',
+                'shell':'$DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || $CONTRUN_NAME=punit_check $TASK_CREATION_TOOL || cat t/log-*',
                 'user':'root',
                 $PARSER
             },
             $CLEANUP_ENV,
             {
                 'name':'Build and test RocksDB debug version under gcc-4.8.1',
-                'shell':'$GCC_481 $DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || cat t/log-*',
+                'shell':'$GCC_481 $DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || $CONTRUN_NAME=punit_check_gcc481 $TASK_CREATION_TOOL || cat t/log-*',
                 'user':'root',
                 $PARSER
             },
@@ -137,7 +169,29 @@ UNIT_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build and test RocksDB debug version',
-                'shell':'$SHM $DEBUG make J=1 check',
+                'shell':'$SHM $DEBUG make $PARALLELISM check || $CONTRUN_NAME=check $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB unit test not under /dev/shm
+#
+UNIT_TEST_NON_SHM_COMMANDS="[
+    {
+        'name':'Rocksdb Unit Test',
+        'oncall':'$ONCALL',
+        'timeout': 86400,
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build and test RocksDB debug version',
+                'timeout': 86400,
+                'shell':'$DEBUG make $PARALLELISM check || $CONTRUN_NAME=non_shm_check $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -149,15 +203,15 @@ UNIT_TEST_COMMANDS="[
 #
 # RocksDB release build and unit tests
 #
-RELEASE_UNIT_TEST_COMMANDS="[
+RELEASE_BUILD_COMMANDS="[
     {
-        'name':'Rocksdb Release Unit Test',
+        'name':'Rocksdb Release Build',
         'oncall':'$ONCALL',
         'steps': [
             $CLEANUP_ENV,
             {
-                'name':'Build and test RocksDB debug version',
-                'shell':'$SHM $DEBUG make J=1 release',
+                'name':'Build RocksDB release',
+                'shell':'make $PARALLEL_j release || $CONTRUN_NAME=release $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -177,7 +231,27 @@ UNIT_TEST_COMMANDS_481="[
             $CLEANUP_ENV,
             {
                 'name':'Build and test RocksDB debug version',
-                'shell':'$SHM $GCC_481 $DEBUG make J=1 check',
+                'shell':'$SHM $GCC_481 $DEBUG make $PARALLELISM check || $CONTRUN_NAME=unit_gcc_481_check $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB release build and unit tests
+#
+RELEASE_BUILD_COMMANDS_481="[
+    {
+        'name':'Rocksdb Release on GCC 4.8.1',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build RocksDB release on GCC 4.8.1',
+                'shell':'$GCC_481 make $PARALLEL_j release || $CONTRUN_NAME=release_gcc481 $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -197,7 +271,27 @@ CLANG_UNIT_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build and test RocksDB debug',
-                'shell':'$CLANG $SHM $DEBUG make J=1 check',
+                'shell':'$CLANG $SHM $DEBUG make $PARALLELISM check || $CONTRUN_NAME=clang_check $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB release build with CLANG
+#
+CLANG_RELEASE_BUILD_COMMANDS="[
+    {
+        'name':'Rocksdb CLANG Release Build',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build RocksDB release',
+                'shell':'$CLANG make $PARALLEL_j release|| $CONTRUN_NAME=clang_release $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -217,7 +311,7 @@ CLANG_ANALYZE_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'RocksDB build and analyze',
-                'shell':'$CLANG $SHM $DEBUG make J=1 analyze',
+                'shell':'$CLANG $SHM $DEBUG make $PARALLEL_j analyze || $CONTRUN_NAME=clang_analyze $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -237,7 +331,7 @@ CODE_COV_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build, test and collect code coverage info',
-                'shell':'$SHM $DEBUG make J=1 coverage',
+                'shell':'$SHM $DEBUG make $PARALLELISM coverage || $CONTRUN_NAME=coverage $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -257,7 +351,7 @@ UNITY_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build, test unity test',
-                'shell':'$SHM $DEBUG V=1 make J=1 unity_test',
+                'shell':'$SHM $DEBUG V=1 make J=1 unity_test || $CONTRUN_NAME=unity_test $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -277,7 +371,7 @@ LITE_BUILD_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build RocksDB debug version',
-                'shell':'$LITE make J=1 static_lib',
+                'shell':'$LITE make J=1 static_lib || $CONTRUN_NAME=lite_static_lib $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -297,7 +391,7 @@ LITE_UNIT_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build RocksDB debug version',
-                'shell':'$SHM $LITE make J=1 check',
+                'shell':'$SHM $LITE make J=1 check || $CONTRUN_NAME=lite_check $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -318,14 +412,14 @@ STRESS_CRASH_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build and run RocksDB debug stress tests',
-                'shell':'$SHM $DEBUG make J=1 db_stress',
+                'shell':'$SHM $DEBUG make J=1 db_stress || $CONTRUN_NAME=db_stress $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
             {
                 'name':'Build and run RocksDB debug crash tests',
                 'timeout': 86400,
-                'shell':'$SHM $DEBUG make J=1 crash_test',
+                'shell':'$SHM $DEBUG make J=1 crash_test || $CONTRUN_NAME=crash_test $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             }
@@ -346,7 +440,7 @@ WRITE_STRESS_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build and run RocksDB write stress tests',
-                'shell':'make write_stress && python tools/write_stress_runner.py --runtime_sec=3600 --db=/tmp/rocksdb_write_stress',
+                'shell':'make write_stress && python tools/write_stress_runner.py --runtime_sec=3600 --db=/tmp/rocksdb_write_stress || $CONTRUN_NAME=write_stress $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             }
@@ -368,7 +462,7 @@ ASAN_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Test RocksDB debug under ASAN',
-                'shell':'set -o pipefail && $SHM $ASAN $DEBUG make J=1 asan_check |& /usr/facebook/ops/scripts/asan_symbolize.py -d',
+'shell':'set -o pipefail && ($SHM $ASAN $DEBUG make $PARALLELISM asan_check || $CONTRUN_NAME=asan_check $TASK_CREATION_TOOL) |& /usr/facebook/ops/scripts/asan_symbolize.py -d',
                 'user':'root',
                 $PARSER
             }
@@ -390,7 +484,7 @@ ASAN_CRASH_TEST_COMMANDS="[
             {
                 'name':'Build and run RocksDB debug asan_crash_test',
                 'timeout': 86400,
-                'shell':'$SHM $DEBUG make J=1 asan_crash_test',
+                'shell':'$SHM $DEBUG make J=1 asan_crash_test || $CONTRUN_NAME=asan_crash_test $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -410,7 +504,7 @@ UBSAN_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Test RocksDB debug under UBSAN',
-                'shell':'set -o pipefail && $SHM $UBSAN $DEBUG make J=1 ubsan_check',
+                'shell':'set -o pipefail && $SHM $UBSAN $DEBUG make $PARALLELISM ubsan_check || $CONTRUN_NAME=ubsan_check $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             }
@@ -432,7 +526,7 @@ UBSAN_CRASH_TEST_COMMANDS="[
             {
                 'name':'Build and run RocksDB debug ubsan_crash_test',
                 'timeout': 86400,
-                'shell':'$SHM $DEBUG make J=1 ubsan_crash_test',
+                'shell':'$SHM $DEBUG make J=1 ubsan_crash_test || $CONTRUN_NAME=ubsan_crash_test $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -448,11 +542,13 @@ VALGRIND_TEST_COMMANDS="[
     {
         'name':'Rocksdb Unit Test under valgrind',
         'oncall':'$ONCALL',
+        'timeout': 86400,
         'steps': [
             $CLEANUP_ENV,
             {
                 'name':'Run RocksDB debug unit tests',
-                'shell':'$DISABLE_JEMALLOC $SHM $DEBUG make valgrind_check',
+                'timeout': 86400,
+                'shell':'$DISABLE_JEMALLOC $SHM $DEBUG make $PARALLELISM valgrind_check || $CONTRUN_NAME=valgrind_check $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -468,11 +564,13 @@ TSAN_UNIT_TEST_COMMANDS="[
     {
         'name':'Rocksdb Unit Test under TSAN',
         'oncall':'$ONCALL',
+        'timeout': 86400,
         'steps': [
             $CLEANUP_ENV,
             {
                 'name':'Run RocksDB debug unit test',
-                'shell':'set -o pipefail && $SHM $DEBUG $TSAN make J=1 check',
+                'timeout': 86400,
+                'shell':'set -o pipefail && $SHM $DEBUG $TSAN make $PARALLELISM check || $CONTRUN_NAME=tsan_check $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -494,7 +592,7 @@ TSAN_CRASH_TEST_COMMANDS="[
             {
                 'name':'Compile and run',
                 'timeout': 86400,
-                'shell':'set -o pipefail && $SHM $DEBUG $TSAN make J=1 crash_test',
+                'shell':'set -o pipefail && $SHM $DEBUG $TSAN CRASH_TEST_KILL_ODD=1887 make J=1 crash_test || $CONTRUN_NAME=tsan_crash_test $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -503,8 +601,6 @@ TSAN_CRASH_TEST_COMMANDS="[
     }
 ]"
 
-TSAN_CRASH_TEST_COMMANDS=$DISABLE_COMMANDS
-
 #
 # RocksDB format compatible
 #
@@ -558,7 +654,7 @@ FORMAT_COMPATIBLE_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Run RocksDB debug unit test',
-                'shell':'build_tools/rocksdb-lego-determinator run_format_compatible',
+                'shell':'build_tools/rocksdb-lego-determinator run_format_compatible || $CONTRUN_NAME=run_format_compatible $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -591,7 +687,7 @@ NO_COMPRESSION_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Run RocksDB debug unit test',
-                'shell':'build_tools/rocksdb-lego-determinator run_no_compression',
+                'shell':'build_tools/rocksdb-lego-determinator run_no_compression || $CONTRUN_NAME=run_no_compression $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -648,7 +744,7 @@ REGRESSION_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Make and run script',
-                'shell':'build_tools/rocksdb-lego-determinator run_regression',
+                'shell':'build_tools/rocksdb-lego-determinator run_regression || $CONTRUN_NAME=run_regression $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -668,7 +764,7 @@ JAVA_BUILD_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Build RocksDB for Java',
-                'shell':'$SETUP_JAVA_ENV; $SHM make rocksdbjava',
+                'shell':'$SETUP_JAVA_ENV; $SHM make rocksdbjava || $CONTRUN_NAME=rocksdbjava $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -685,15 +781,24 @@ case $1 in
   unit)
     echo $UNIT_TEST_COMMANDS
     ;;
+  unit_non_shm)
+    echo $UNIT_TEST_NON_SHM_COMMANDS
+    ;;
   release)
-    echo $RELEASE_UNIT_TEST_COMMANDS
+    echo $RELEASE_BUILD_COMMANDS
     ;;
   unit_481)
     echo $UNIT_TEST_COMMANDS_481
     ;;
+  release_481)
+    echo $RELEASE_BUILD_COMMANDS_481
+    ;;
   clang_unit)
     echo $CLANG_UNIT_TEST_COMMANDS
     ;;
+  clang_release)
+    echo $CLANG_RELEASE_BUILD_COMMANDS
+    ;;
   clang_analyze)
     echo $CLANG_ANALYZE_COMMANDS
     ;;

build_tools/update_dependencies.sh

@@ -82,6 +82,7 @@ get_lib_base gflags     LATEST
 get_lib_base jemalloc   LATEST
 get_lib_base numa       LATEST
 get_lib_base libunwind  LATEST
+get_lib_base tbb        4.0_update2 gcc-4.9-glibc-2.20
 
 get_lib_base kernel-headers LATEST 
 get_lib_base binutils   LATEST centos6-native 
@@ -119,6 +120,7 @@ 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 tbb        4.0_update2 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 

db/builder.h

@@ -12,14 +12,13 @@
 #include "db/table_properties_collector.h"
 #include "rocksdb/comparator.h"
 #include "rocksdb/env.h"
-#include "rocksdb/immutable_options.h"
 #include "rocksdb/listener.h"
 #include "rocksdb/options.h"
 #include "rocksdb/status.h"
 #include "rocksdb/table_properties.h"
 #include "rocksdb/types.h"
+#include "util/cf_options.h"
 #include "util/event_logger.h"
-#include "util/mutable_cf_options.h"
 
 namespace rocksdb {
 

db/c.cc

@@ -43,6 +43,7 @@ using rocksdb::CompactionFilterContext;
 using rocksdb::CompactionOptionsFIFO;
 using rocksdb::Comparator;
 using rocksdb::CompressionType;
+using rocksdb::WALRecoveryMode;
 using rocksdb::DB;
 using rocksdb::DBOptions;
 using rocksdb::Env;
@@ -123,6 +124,7 @@ struct rocksdb_compactionfilter_t : public CompactionFilter {
       char** new_value, size_t *new_value_length,
       unsigned char* value_changed);
   const char* (*name_)(void*);
+  unsigned char ignore_snapshots_;
 
   virtual ~rocksdb_compactionfilter_t() {
     (*destructor_)(state_);
@@ -148,6 +150,8 @@ struct rocksdb_compactionfilter_t : public CompactionFilter {
   }
 
   virtual const char* Name() const override { return (*name_)(state_); }
+
+  virtual bool IgnoreSnapshots() const override { return ignore_snapshots_; }
 };
 
 struct rocksdb_compactionfilterfactory_t : public CompactionFilterFactory {
@@ -269,33 +273,31 @@ struct rocksdb_mergeoperator_t : public MergeOperator {
 
   virtual const char* Name() const override { return (*name_)(state_); }
 
-  virtual bool FullMerge(const Slice& key, const Slice* existing_value,
-                         const std::deque<std::string>& operand_list,
-                         std::string* new_value,
-                         Logger* logger) const override {
-    size_t n = operand_list.size();
+  virtual bool FullMergeV2(const MergeOperationInput& merge_in,
+                           MergeOperationOutput* merge_out) const override {
+    size_t n = merge_in.operand_list.size();
     std::vector<const char*> operand_pointers(n);
     std::vector<size_t> operand_sizes(n);
     for (size_t i = 0; i < n; i++) {
-      Slice operand(operand_list[i]);
+      Slice operand(merge_in.operand_list[i]);
       operand_pointers[i] = operand.data();
       operand_sizes[i] = operand.size();
     }
 
     const char* existing_value_data = nullptr;
     size_t existing_value_len = 0;
-    if (existing_value != nullptr) {
-      existing_value_data = existing_value->data();
-      existing_value_len = existing_value->size();
+    if (merge_in.existing_value != nullptr) {
+      existing_value_data = merge_in.existing_value->data();
+      existing_value_len = merge_in.existing_value->size();
     }
 
     unsigned char success;
     size_t new_value_len;
     char* tmp_new_value = (*full_merge_)(
-        state_, key.data(), key.size(), existing_value_data, existing_value_len,
-        &operand_pointers[0], &operand_sizes[0], static_cast<int>(n), &success,
-        &new_value_len);
-    new_value->assign(tmp_new_value, new_value_len);
+        state_, merge_in.key.data(), merge_in.key.size(), existing_value_data,
+        existing_value_len, &operand_pointers[0], &operand_sizes[0],
+        static_cast<int>(n), &success, &new_value_len);
+    merge_out->new_value.assign(tmp_new_value, new_value_len);
 
     if (delete_value_ != nullptr) {
       (*delete_value_)(state_, tmp_new_value, new_value_len);
@@ -1514,14 +1516,9 @@ void rocksdb_options_set_max_bytes_for_level_multiplier(
   opt->rep.max_bytes_for_level_multiplier = n;
 }
 
-void rocksdb_options_set_expanded_compaction_factor(
-    rocksdb_options_t* opt, int n) {
-  opt->rep.expanded_compaction_factor = n;
-}
-
-void rocksdb_options_set_max_grandparent_overlap_factor(
-    rocksdb_options_t* opt, int n) {
-  opt->rep.max_grandparent_overlap_factor = n;
+void rocksdb_options_set_max_compaction_bytes(rocksdb_options_t* opt,
+                                              uint64_t n) {
+  opt->rep.max_compaction_bytes = n;
 }
 
 void rocksdb_options_set_max_bytes_for_level_multiplier_additional(
@@ -1558,6 +1555,10 @@ void rocksdb_options_set_level0_stop_writes_trigger(
 void rocksdb_options_set_max_mem_compaction_level(rocksdb_options_t* opt,
                                                   int n) {}
 
+void rocksdb_options_set_wal_recovery_mode(rocksdb_options_t* opt,int mode) {
+  opt->rep.wal_recovery_mode = static_cast<WALRecoveryMode>(mode);
+}
+
 void rocksdb_options_set_compression(rocksdb_options_t* opt, int t) {
   opt->rep.compression = static_cast<CompressionType>(t);
 }
@@ -1777,11 +1778,6 @@ void rocksdb_options_set_delete_obsolete_files_period_micros(
   opt->rep.delete_obsolete_files_period_micros = v;
 }
 
-void rocksdb_options_set_source_compaction_factor(
-    rocksdb_options_t* opt, int n) {
-  opt->rep.expanded_compaction_factor = n;
-}
-
 void rocksdb_options_prepare_for_bulk_load(rocksdb_options_t* opt) {
   opt->rep.PrepareForBulkLoad();
 }
@@ -1799,9 +1795,9 @@ void rocksdb_options_set_memtable_prefix_bloom_size_ratio(
   opt->rep.memtable_prefix_bloom_size_ratio = 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_memtable_huge_page_size(rocksdb_options_t* opt,
+                                                 size_t v) {
+  opt->rep.memtable_huge_page_size = v;
 }
 
 void rocksdb_options_set_hash_skip_list_rep(
@@ -1921,10 +1917,17 @@ rocksdb_compactionfilter_t* rocksdb_compactionfilter_create(
   result->state_ = state;
   result->destructor_ = destructor;
   result->filter_ = filter;
+  result->ignore_snapshots_ = false;
   result->name_ = name;
   return result;
 }
 
+void rocksdb_compactionfilter_set_ignore_snapshots(
+  rocksdb_compactionfilter_t* filter,
+  unsigned char whether_ignore) {
+  filter->ignore_snapshots_ = whether_ignore;
+}
+
 void rocksdb_compactionfilter_destroy(rocksdb_compactionfilter_t* filter) {
   delete filter;
 }

db/c_test.c

@@ -575,6 +575,7 @@ int main(int argc, char** argv) {
     rocksdb_close(db);
     rocksdb_options_set_create_if_missing(options, 0);
     rocksdb_options_set_error_if_exists(options, 0);
+    rocksdb_options_set_wal_recovery_mode(options, 2);
     rocksdb_repair_db(options, dbname, &err);
     CheckNoError(err);
     db = rocksdb_open(options, dbname, &err);

db/column_family.cc

@@ -81,7 +81,7 @@ Status ColumnFamilyHandleImpl::GetDescriptor(ColumnFamilyDescriptor* desc) {
 #endif  // !ROCKSDB_LITE
 }
 
-const Comparator* ColumnFamilyHandleImpl::user_comparator() const {
+const Comparator* ColumnFamilyHandleImpl::GetComparator() const {
   return cfd()->user_comparator();
 }
 
@@ -204,6 +204,10 @@ ColumnFamilyOptions SanitizeOptions(const DBOptions& db_options,
     result.level0_stop_writes_trigger = std::numeric_limits<int>::max();
   }
 
+  if (result.max_bytes_for_level_multiplier <= 0) {
+    result.max_bytes_for_level_multiplier = 1;
+  }
+
   if (result.level0_file_num_compaction_trigger == 0) {
     Warn(db_options.info_log.get(),
          "level0_file_num_compaction_trigger cannot be 0");
@@ -262,6 +266,10 @@ ColumnFamilyOptions SanitizeOptions(const DBOptions& db_options,
     }
   }
 
+  if (result.max_compaction_bytes == 0) {
+    result.max_compaction_bytes = result.target_file_size_base * 25;
+  }
+
   return result;
 }
 
@@ -558,8 +566,9 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
           "(waiting for flush), max_write_buffer_number is set to %d",
           name_.c_str(), imm()->NumNotFlushed(),
           mutable_cf_options.max_write_buffer_number);
-    } else if (vstorage->l0_delay_trigger_count() >=
-               mutable_cf_options.level0_stop_writes_trigger) {
+    } else if (!mutable_cf_options.disable_auto_compactions &&
+               vstorage->l0_delay_trigger_count() >=
+                   mutable_cf_options.level0_stop_writes_trigger) {
       write_controller_token_ = write_controller->GetStopToken();
       internal_stats_->AddCFStats(InternalStats::LEVEL0_NUM_FILES_TOTAL, 1);
       if (compaction_picker_->IsLevel0CompactionInProgress()) {
@@ -569,7 +578,8 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
       Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
           "[%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 &&
+    } else if (!mutable_cf_options.disable_auto_compactions &&
+               mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
                compaction_needed_bytes >=
                    mutable_cf_options.hard_pending_compaction_bytes_limit) {
       write_controller_token_ = write_controller->GetStopToken();
@@ -594,7 +604,8 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
           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 &&
+    } else if (!mutable_cf_options.disable_auto_compactions &&
+               mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
                vstorage->l0_delay_trigger_count() >=
                    mutable_cf_options.level0_slowdown_writes_trigger) {
       write_controller_token_ =
@@ -611,7 +622,8 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
           "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 &&
+    } else if (!mutable_cf_options.disable_auto_compactions &&
+               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_ =
@@ -1007,8 +1019,7 @@ uint32_t GetColumnFamilyID(ColumnFamilyHandle* column_family) {
 const Comparator* GetColumnFamilyUserComparator(
     ColumnFamilyHandle* column_family) {
   if (column_family != nullptr) {
-    auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
-    return cfh->user_comparator();
+    return column_family->GetComparator();
   }
   return nullptr;
 }

db/column_family.h

@@ -15,15 +15,15 @@
 #include <atomic>
 
 #include "db/memtable_list.h"
-#include "db/write_batch_internal.h"
-#include "db/write_controller.h"
 #include "db/table_cache.h"
 #include "db/table_properties_collector.h"
+#include "db/write_batch_internal.h"
+#include "db/write_controller.h"
 #include "rocksdb/compaction_job_stats.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
-#include "util/mutable_cf_options.h"
+#include "util/cf_options.h"
 #include "util/thread_local.h"
 
 namespace rocksdb {
@@ -55,11 +55,11 @@ class ColumnFamilyHandleImpl : public ColumnFamilyHandle {
   // destroy without mutex
   virtual ~ColumnFamilyHandleImpl();
   virtual ColumnFamilyData* cfd() const { return cfd_; }
-  virtual const Comparator* user_comparator() const;
 
   virtual uint32_t GetID() const override;
   virtual const std::string& GetName() const override;
   virtual Status GetDescriptor(ColumnFamilyDescriptor* desc) override;
+  virtual const Comparator* GetComparator() const override;
 
  private:
   ColumnFamilyData* cfd_;

db/column_family_test.cc

@@ -1,7 +1,8 @@
 //  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.
+//  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
@@ -912,6 +913,38 @@ TEST_F(ColumnFamilyTest, LogDeletionTest) {
   Close();
 }
 
+TEST_F(ColumnFamilyTest, CrashAfterFlush) {
+  std::unique_ptr<FaultInjectionTestEnv> fault_env(
+      new FaultInjectionTestEnv(env_));
+  db_options_.env = fault_env.get();
+  Open();
+  CreateColumnFamilies({"one"});
+
+  WriteBatch batch;
+  batch.Put(handles_[0], Slice("foo"), Slice("bar"));
+  batch.Put(handles_[1], Slice("foo"), Slice("bar"));
+  ASSERT_OK(db_->Write(WriteOptions(), &batch));
+  Flush(0);
+  fault_env->SetFilesystemActive(false);
+
+  std::vector<std::string> names;
+  for (auto name : names_) {
+    if (name != "") {
+      names.push_back(name);
+    }
+  }
+  Close();
+  fault_env->DropUnsyncedFileData();
+  fault_env->ResetState();
+  Open(names, {});
+
+  // Write batch should be atomic.
+  ASSERT_EQ(Get(0, "foo"), Get(1, "foo"));
+
+  Close();
+  db_options_.env = env_;
+}
+
 // Makes sure that obsolete log files get deleted
 TEST_F(ColumnFamilyTest, DifferentWriteBufferSizes) {
   // disable flushing stale column families
@@ -1040,6 +1073,39 @@ TEST_F(ColumnFamilyTest, MemtableNotSupportSnapshot) {
 }
 #endif  // !ROCKSDB_LITE
 
+class TestComparator : public Comparator {
+  int Compare(const rocksdb::Slice& a, const rocksdb::Slice& b) const override {
+    return 0;
+  }
+  const char* Name() const override { return "Test"; }
+  void FindShortestSeparator(std::string* start,
+                             const rocksdb::Slice& limit) const override {}
+  void FindShortSuccessor(std::string* key) const override {}
+};
+
+static TestComparator third_comparator;
+static TestComparator fourth_comparator;
+
+// Test that we can retrieve the comparator from a created CF
+TEST_F(ColumnFamilyTest, GetComparator) {
+  Open();
+  // Add a column family with no comparator specified
+  CreateColumnFamilies({"first"});
+  const Comparator* comp = handles_[0]->GetComparator();
+  ASSERT_EQ(comp, BytewiseComparator());
+
+  // Add three column families - one with no comparator and two
+  // with comparators specified
+  ColumnFamilyOptions second, third, fourth;
+  second.comparator = &third_comparator;
+  third.comparator = &fourth_comparator;
+  CreateColumnFamilies({"second", "third", "fourth"}, {second, third, fourth});
+  ASSERT_EQ(handles_[1]->GetComparator(), BytewiseComparator());
+  ASSERT_EQ(handles_[2]->GetComparator(), &third_comparator);
+  ASSERT_EQ(handles_[3]->GetComparator(), &fourth_comparator);
+  Close();
+}
+
 TEST_F(ColumnFamilyTest, DifferentMergeOperators) {
   Open();
   CreateColumnFamilies({"first", "second"});
@@ -1081,7 +1147,8 @@ TEST_F(ColumnFamilyTest, DifferentCompactionStyles) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = static_cast<uint64_t>(1) << 60;
+
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1153,7 +1220,7 @@ TEST_F(ColumnFamilyTest, MultipleManualCompactions) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1252,7 +1319,7 @@ TEST_F(ColumnFamilyTest, AutomaticAndManualCompactions) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1345,7 +1412,7 @@ TEST_F(ColumnFamilyTest, ManualAndAutomaticCompactions) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1441,7 +1508,7 @@ TEST_F(ColumnFamilyTest, SameCFManualManualCompactions) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1540,7 +1607,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactions) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1630,7 +1697,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactionsLevel) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1727,7 +1794,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticConflict) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1847,7 +1914,7 @@ TEST_F(ColumnFamilyTest, SameCFAutomaticManualCompactions) {
   default_cf.num_levels = 3;
   default_cf.write_buffer_size = 64 << 10;  // 64KB
   default_cf.target_file_size_base = 30 << 10;
-  default_cf.source_compaction_factor = 100;
+  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
   BlockBasedTableOptions table_options;
   table_options.no_block_cache = true;
   default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -2381,6 +2448,7 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
   mutable_cf_options.level0_stop_writes_trigger = 10000;
   mutable_cf_options.soft_pending_compaction_bytes_limit = 200;
   mutable_cf_options.hard_pending_compaction_bytes_limit = 2000;
+  mutable_cf_options.disable_auto_compactions = false;
 
   vstorage->TEST_set_estimated_compaction_needed_bytes(50);
   cfd->RecalculateWriteStallConditions(mutable_cf_options);
@@ -2527,16 +2595,17 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
   vstorage->set_l0_delay_trigger_count(50);
   cfd->RecalculateWriteStallConditions(mutable_cf_options);
   ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
-  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
+  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
   ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
 
   vstorage->set_l0_delay_trigger_count(60);
   vstorage->TEST_set_estimated_compaction_needed_bytes(300);
   cfd->RecalculateWriteStallConditions(mutable_cf_options);
   ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
-  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
+  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
   ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
 
+  mutable_cf_options.disable_auto_compactions = false;
   vstorage->set_l0_delay_trigger_count(70);
   vstorage->TEST_set_estimated_compaction_needed_bytes(500);
   cfd->RecalculateWriteStallConditions(mutable_cf_options);
@@ -2544,7 +2613,6 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
   ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
   ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
 
-  mutable_cf_options.disable_auto_compactions = false;
   vstorage->set_l0_delay_trigger_count(71);
   vstorage->TEST_set_estimated_compaction_needed_bytes(501);
   cfd->RecalculateWriteStallConditions(mutable_cf_options);
@@ -2771,6 +2839,223 @@ TEST_F(ColumnFamilyTest, CompactionSpeedupTwoColumnFamilies) {
   ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
 }
 
+#ifndef ROCKSDB_LITE
+TEST_F(ColumnFamilyTest, FlushCloseWALFiles) {
+  SpecialEnv env(Env::Default());
+  db_options_.env = &env;
+  db_options_.max_background_flushes = 1;
+  column_family_options_.memtable_factory.reset(new SpecialSkipListFactory(2));
+  Open();
+  CreateColumnFamilies({"one"});
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  ASSERT_OK(Put(0, "fodor", "mirko"));
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      {"DBImpl::BGWorkFlush:done", "FlushCloseWALFiles:0"},
+  });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // Block flush jobs from running
+  test::SleepingBackgroundTask sleeping_task;
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
+                 Env::Priority::HIGH);
+
+  WriteOptions wo;
+  wo.sync = true;
+  ASSERT_OK(db_->Put(wo, handles_[1], "fodor", "mirko"));
+
+  ASSERT_EQ(2, env.num_open_wal_file_.load());
+
+  sleeping_task.WakeUp();
+  sleeping_task.WaitUntilDone();
+  TEST_SYNC_POINT("FlushCloseWALFiles:0");
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  ASSERT_EQ(1, env.num_open_wal_file_.load());
+
+  Reopen();
+  ASSERT_EQ("mirko", Get(0, "fodor"));
+  ASSERT_EQ("mirko", Get(1, "fodor"));
+  db_options_.env = env_;
+  Close();
+}
+#endif  // !ROCKSDB_LITE
+
+TEST_F(ColumnFamilyTest, IteratorCloseWALFile1) {
+  SpecialEnv env(Env::Default());
+  db_options_.env = &env;
+  db_options_.max_background_flushes = 1;
+  column_family_options_.memtable_factory.reset(new SpecialSkipListFactory(2));
+  Open();
+  CreateColumnFamilies({"one"});
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  // Create an iterator holding the current super version.
+  Iterator* it = db_->NewIterator(ReadOptions(), handles_[1]);
+  // A flush will make `it` hold the last reference of its super version.
+  Flush(1);
+
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  ASSERT_OK(Put(0, "fodor", "mirko"));
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+
+  // Flush jobs will close previous WAL files after finishing. By
+  // block flush jobs from running, we trigger a condition where
+  // the iterator destructor should close the WAL files.
+  test::SleepingBackgroundTask sleeping_task;
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
+                 Env::Priority::HIGH);
+
+  WriteOptions wo;
+  wo.sync = true;
+  ASSERT_OK(db_->Put(wo, handles_[1], "fodor", "mirko"));
+
+  ASSERT_EQ(2, env.num_open_wal_file_.load());
+  // Deleting the iterator will clear its super version, triggering
+  // closing all files
+  delete it;
+  ASSERT_EQ(1, env.num_open_wal_file_.load());
+
+  sleeping_task.WakeUp();
+  sleeping_task.WaitUntilDone();
+  WaitForFlush(1);
+
+  Reopen();
+  ASSERT_EQ("mirko", Get(0, "fodor"));
+  ASSERT_EQ("mirko", Get(1, "fodor"));
+  db_options_.env = env_;
+  Close();
+}
+
+TEST_F(ColumnFamilyTest, IteratorCloseWALFile2) {
+  SpecialEnv env(Env::Default());
+  // Allow both of flush and purge job to schedule.
+  env.SetBackgroundThreads(2, Env::HIGH);
+  db_options_.env = &env;
+  db_options_.max_background_flushes = 1;
+  column_family_options_.memtable_factory.reset(new SpecialSkipListFactory(2));
+  Open();
+  CreateColumnFamilies({"one"});
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  // Create an iterator holding the current super version.
+  ReadOptions ro;
+  ro.background_purge_on_iterator_cleanup = true;
+  Iterator* it = db_->NewIterator(ro, handles_[1]);
+  // A flush will make `it` hold the last reference of its super version.
+  Flush(1);
+
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  ASSERT_OK(Put(0, "fodor", "mirko"));
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      {"ColumnFamilyTest::IteratorCloseWALFile2:0",
+       "DBImpl::BGWorkPurge:start"},
+      {"ColumnFamilyTest::IteratorCloseWALFile2:2",
+       "DBImpl::BackgroundCallFlush:start"},
+      {"DBImpl::BGWorkPurge:end", "ColumnFamilyTest::IteratorCloseWALFile2:1"},
+  });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  WriteOptions wo;
+  wo.sync = true;
+  ASSERT_OK(db_->Put(wo, handles_[1], "fodor", "mirko"));
+
+  ASSERT_EQ(2, env.num_open_wal_file_.load());
+  // Deleting the iterator will clear its super version, triggering
+  // closing all files
+  delete it;
+  ASSERT_EQ(2, env.num_open_wal_file_.load());
+
+  TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:0");
+  TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:1");
+  ASSERT_EQ(1, env.num_open_wal_file_.load());
+  TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:2");
+  WaitForFlush(1);
+  ASSERT_EQ(1, env.num_open_wal_file_.load());
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  Reopen();
+  ASSERT_EQ("mirko", Get(0, "fodor"));
+  ASSERT_EQ("mirko", Get(1, "fodor"));
+  db_options_.env = env_;
+  Close();
+}
+
+#ifndef ROCKSDB_LITE  // TEST functions are not supported in lite
+TEST_F(ColumnFamilyTest, ForwardIteratorCloseWALFile) {
+  SpecialEnv env(Env::Default());
+  // Allow both of flush and purge job to schedule.
+  env.SetBackgroundThreads(2, Env::HIGH);
+  db_options_.env = &env;
+  db_options_.max_background_flushes = 1;
+  column_family_options_.memtable_factory.reset(new SpecialSkipListFactory(3));
+  column_family_options_.level0_file_num_compaction_trigger = 2;
+  Open();
+  CreateColumnFamilies({"one"});
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  ASSERT_OK(Put(1, "fodar2", "mirko"));
+  Flush(1);
+
+  // Create an iterator holding the current super version, as well as
+  // the SST file just flushed.
+  ReadOptions ro;
+  ro.tailing = true;
+  ro.background_purge_on_iterator_cleanup = true;
+  Iterator* it = db_->NewIterator(ro, handles_[1]);
+  // A flush will make `it` hold the last reference of its super version.
+
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  ASSERT_OK(Put(1, "fodar2", "mirko"));
+  Flush(1);
+
+  WaitForCompaction();
+
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+  ASSERT_OK(Put(0, "fodor", "mirko"));
+  ASSERT_OK(Put(1, "fodor", "mirko"));
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      {"ColumnFamilyTest::IteratorCloseWALFile2:0",
+       "DBImpl::BGWorkPurge:start"},
+      {"ColumnFamilyTest::IteratorCloseWALFile2:2",
+       "DBImpl::BackgroundCallFlush:start"},
+      {"DBImpl::BGWorkPurge:end", "ColumnFamilyTest::IteratorCloseWALFile2:1"},
+  });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  WriteOptions wo;
+  wo.sync = true;
+  ASSERT_OK(db_->Put(wo, handles_[1], "fodor", "mirko"));
+
+  env.delete_count_.store(0);
+  ASSERT_EQ(2, env.num_open_wal_file_.load());
+  // Deleting the iterator will clear its super version, triggering
+  // closing all files
+  it->Seek("");
+  ASSERT_EQ(2, env.num_open_wal_file_.load());
+  ASSERT_EQ(0, env.delete_count_.load());
+
+  TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:0");
+  TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:1");
+  ASSERT_EQ(1, env.num_open_wal_file_.load());
+  ASSERT_EQ(1, env.delete_count_.load());
+  TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:2");
+  WaitForFlush(1);
+  ASSERT_EQ(1, env.num_open_wal_file_.load());
+  ASSERT_EQ(1, env.delete_count_.load());
+
+  delete it;
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  Reopen();
+  ASSERT_EQ("mirko", Get(0, "fodor"));
+  ASSERT_EQ("mirko", Get(1, "fodor"));
+  db_options_.env = env_;
+  Close();
+}
+#endif  // !ROCKSDB_LITE
+
 // Disable on windows because SyncWAL requires env->IsSyncThreadSafe()
 // to return true which is not so in unbuffered mode.
 #ifndef OS_WIN

db/compact_files_test.cc

@@ -10,6 +10,7 @@
 #include <thread>
 #include <vector>
 
+#include "db/db_impl.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "util/string_util.h"
@@ -49,6 +50,7 @@ class FlushedFileCollector : public EventListener {
     }
     return result;
   }
+  void ClearFlushedFiles() { flushed_files_.clear(); }
 
  private:
   std::vector<std::string> flushed_files_;
@@ -115,13 +117,12 @@ TEST_F(CompactFilesTest, L0ConflictsFiles) {
 TEST_F(CompactFilesTest, ObsoleteFiles) {
   Options options;
   // to trigger compaction more easily
-  const int kWriteBufferSize = 10000;
+  const int kWriteBufferSize = 65536;
   options.create_if_missing = true;
   // Disable RocksDB background compaction.
   options.compaction_style = kCompactionStyleNone;
-  // Small slowdown and stop trigger for experimental purpose.
-  options.level0_slowdown_writes_trigger = 20;
-  options.level0_stop_writes_trigger = 20;
+  options.level0_slowdown_writes_trigger = (1 << 30);
+  options.level0_stop_writes_trigger = (1 << 30);
   options.write_buffer_size = kWriteBufferSize;
   options.max_write_buffer_number = 2;
   options.compression = kNoCompression;
@@ -144,6 +145,7 @@ TEST_F(CompactFilesTest, ObsoleteFiles) {
 
   auto l0_files = collector->GetFlushedFiles();
   ASSERT_OK(db->CompactFiles(CompactionOptions(), l0_files, 1));
+  reinterpret_cast<DBImpl*>(db)->TEST_WaitForCompact();
 
   // verify all compaction input files are deleted
   for (auto fname : l0_files) {
@@ -152,6 +154,46 @@ TEST_F(CompactFilesTest, ObsoleteFiles) {
   delete db;
 }
 
+TEST_F(CompactFilesTest, NotCutOutputOnLevel0) {
+  Options options;
+  options.create_if_missing = true;
+  // Disable RocksDB background compaction.
+  options.compaction_style = kCompactionStyleNone;
+  options.level0_slowdown_writes_trigger = 1000;
+  options.level0_stop_writes_trigger = 1000;
+  options.write_buffer_size = 65536;
+  options.max_write_buffer_number = 2;
+  options.compression = kNoCompression;
+  options.max_compaction_bytes = 5000;
+
+  // 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 couple files
+  for (int i = 0; i < 500; ++i) {
+    db->Put(WriteOptions(), ToString(i), std::string(1000, 'a' + (i % 26)));
+  }
+  reinterpret_cast<DBImpl*>(db)->TEST_WaitForFlushMemTable();
+  auto l0_files_1 = collector->GetFlushedFiles();
+  collector->ClearFlushedFiles();
+  for (int i = 0; i < 500; ++i) {
+    db->Put(WriteOptions(), ToString(i), std::string(1000, 'a' + (i % 26)));
+  }
+  reinterpret_cast<DBImpl*>(db)->TEST_WaitForFlushMemTable();
+  auto l0_files_2 = collector->GetFlushedFiles();
+  ASSERT_OK(db->CompactFiles(CompactionOptions(), l0_files_1, 0));
+  ASSERT_OK(db->CompactFiles(CompactionOptions(), l0_files_2, 0));
+  // no assertion failure
+  delete db;
+}
+
 TEST_F(CompactFilesTest, CapturingPendingFiles) {
   Options options;
   options.create_if_missing = true;

db/compaction.cc

@@ -143,8 +143,8 @@ Compaction::Compaction(VersionStorageInfo* vstorage,
                        const MutableCFOptions& _mutable_cf_options,
                        std::vector<CompactionInputFiles> _inputs,
                        int _output_level, uint64_t _target_file_size,
-                       uint64_t _max_grandparent_overlap_bytes,
-                       uint32_t _output_path_id, CompressionType _compression,
+                       uint64_t _max_compaction_bytes, uint32_t _output_path_id,
+                       CompressionType _compression,
                        std::vector<FileMetaData*> _grandparents,
                        bool _manual_compaction, double _score,
                        bool _deletion_compaction,
@@ -152,7 +152,7 @@ Compaction::Compaction(VersionStorageInfo* vstorage,
     : start_level_(_inputs[0].level),
       output_level_(_output_level),
       max_output_file_size_(_target_file_size),
-      max_grandparent_overlap_bytes_(_max_grandparent_overlap_bytes),
+      max_compaction_bytes_(_max_compaction_bytes),
       mutable_cf_options_(_mutable_cf_options),
       input_version_(nullptr),
       number_levels_(vstorage->num_levels()),
@@ -247,7 +247,7 @@ bool Compaction::IsTrivialMove() const {
   return (start_level_ != output_level_ && num_input_levels() == 1 &&
           input(0, 0)->fd.GetPathId() == output_path_id() &&
           InputCompressionMatchesOutput() &&
-          TotalFileSize(grandparents_) <= max_grandparent_overlap_bytes_);
+          TotalFileSize(grandparents_) <= max_compaction_bytes_);
 }
 
 void Compaction::AddInputDeletions(VersionEdit* out_edit) {

db/compaction.h

@@ -8,10 +8,10 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #pragma once
+#include "db/version_set.h"
 #include "util/arena.h"
 #include "util/autovector.h"
-#include "util/mutable_cf_options.h"
-#include "db/version_set.h"
+#include "util/cf_options.h"
 
 namespace rocksdb {
 
@@ -37,7 +37,7 @@ class Compaction {
   Compaction(VersionStorageInfo* input_version,
              const MutableCFOptions& mutable_cf_options,
              std::vector<CompactionInputFiles> inputs, int output_level,
-             uint64_t target_file_size, uint64_t max_grandparent_overlap_bytes,
+             uint64_t target_file_size, uint64_t max_compaction_bytes,
              uint32_t output_path_id, CompressionType compression,
              std::vector<FileMetaData*> grandparents,
              bool manual_compaction = false, double score = -1,
@@ -229,9 +229,7 @@ class Compaction {
     return grandparents_;
   }
 
-  uint64_t max_grandparent_overlap_bytes() const {
-    return max_grandparent_overlap_bytes_;
-  }
+  uint64_t max_compaction_bytes() const { return max_compaction_bytes_; }
 
  private:
   // mark (or clear) all files that are being compacted
@@ -254,7 +252,7 @@ class Compaction {
   const int start_level_;    // the lowest level to be compacted
   const int output_level_;  // levels to which output files are stored
   uint64_t max_output_file_size_;
-  uint64_t max_grandparent_overlap_bytes_;
+  uint64_t max_compaction_bytes_;
   MutableCFOptions mutable_cf_options_;
   Version* input_version_;
   VersionEdit edit_;

db/compaction_iterator.cc

@@ -49,6 +49,12 @@ CompactionIterator::CompactionIterator(
   } else {
     ignore_snapshots_ = false;
   }
+  input_->SetPinnedItersMgr(&pinned_iters_mgr_);
+}
+
+CompactionIterator::~CompactionIterator() {
+  // input_ Iteartor lifetime is longer than pinned_iters_mgr_ lifetime
+  input_->SetPinnedItersMgr(nullptr);
 }
 
 void CompactionIterator::ResetRecordCounts() {
@@ -83,6 +89,8 @@ void CompactionIterator::Next() {
       ikey_.user_key = current_key_.GetUserKey();
       valid_ = true;
     } else {
+      // We consumed all pinned merge operands, release pinned iterators
+      pinned_iters_mgr_.ReleasePinnedData();
       // MergeHelper moves the iterator to the first record after the merged
       // records, so even though we reached the end of the merge output, we do
       // not want to advance the iterator.
@@ -246,6 +254,7 @@ void CompactionIterator::NextFromInput() {
       // 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.
+      // We will report counts on these anomalous cases.
 
       // The easiest way to process a SingleDelete during iteration is to peek
       // ahead at the next key.
@@ -268,6 +277,7 @@ void CompactionIterator::NextFromInput() {
             // First SingleDelete has been skipped since we already called
             // input_->Next().
             ++iter_stats_.num_record_drop_obsolete;
+            ++iter_stats_.num_single_del_mismatch;
           } else if ((ikey_.sequence <= earliest_write_conflict_snapshot_) ||
                      has_outputted_key_) {
             // Found a matching value, we can drop the single delete and the
@@ -277,7 +287,12 @@ void CompactionIterator::NextFromInput() {
 
             // 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.
+            // either way. We will maintain counts of how many mismatches
+            // happened
+            if (next_ikey.type != kTypeValue) {
+              ++iter_stats_.num_single_del_mismatch;
+            }
+
             ++iter_stats_.num_record_drop_hidden;
             ++iter_stats_.num_record_drop_obsolete;
             // Already called input_->Next() once.  Call it a second time to
@@ -318,6 +333,7 @@ void CompactionIterator::NextFromInput() {
           // Key doesn't exist outside of this range.
           // Can compact out this SingleDelete.
           ++iter_stats_.num_record_drop_obsolete;
+          ++iter_stats_.num_single_del_fallthru;
         } else {
           // Output SingleDelete
           valid_ = true;
@@ -368,6 +384,7 @@ void CompactionIterator::NextFromInput() {
         return;
       }
 
+      pinned_iters_mgr_.StartPinning();
       // We know the merge type entry is not hidden, otherwise we would
       // have hit (A)
       // We encapsulate the merge related state machine in a different
@@ -395,6 +412,7 @@ void CompactionIterator::NextFromInput() {
         // batch consumed by the merge operator should not shadow any keys
         // coming after the merges
         has_current_user_key_ = false;
+        pinned_iters_mgr_.ReleasePinnedData();
       }
     } else {
       valid_ = true;

db/compaction_iterator.h

@@ -13,6 +13,7 @@
 
 #include "db/compaction.h"
 #include "db/merge_helper.h"
+#include "db/pinned_iterators_manager.h"
 #include "rocksdb/compaction_filter.h"
 #include "util/log_buffer.h"
 
@@ -33,6 +34,10 @@ struct CompactionIteratorStats {
   uint64_t num_input_corrupt_records = 0;
   uint64_t total_input_raw_key_bytes = 0;
   uint64_t total_input_raw_value_bytes = 0;
+
+  // Single-Delete diagnostics for exceptional situations
+  uint64_t num_single_del_fallthru = 0;
+  uint64_t num_single_del_mismatch = 0;
 };
 
 class CompactionIterator {
@@ -46,6 +51,8 @@ class CompactionIterator {
                      const CompactionFilter* compaction_filter = nullptr,
                      LogBuffer* log_buffer = nullptr);
 
+  ~CompactionIterator();
+
   void ResetRecordCounts();
 
   // Seek to the beginning of the compaction iterator output.
@@ -136,6 +143,9 @@ class CompactionIterator {
   bool clear_and_output_next_key_ = false;
 
   MergeOutputIterator merge_out_iter_;
+  // PinnedIteratorsManager used to pin input_ Iterator blocks while reading
+  // merge operands and then releasing them after consuming them.
+  PinnedIteratorsManager pinned_iters_mgr_;
   std::string compaction_filter_value_;
   // "level_ptrs" holds indices that remember which file of an associated
   // level we were last checking during the last call to compaction->

db/compaction_job.cc

@@ -99,6 +99,8 @@ struct CompactionJob::SubcompactionState {
     }
   }
 
+  uint64_t current_output_file_size;
+
   // State during the subcompaction
   uint64_t total_bytes;
   uint64_t num_input_records;
@@ -121,6 +123,7 @@ struct CompactionJob::SubcompactionState {
         end(_end),
         outfile(nullptr),
         builder(nullptr),
+        current_output_file_size(0),
         total_bytes(0),
         num_input_records(0),
         num_output_records(0),
@@ -142,6 +145,7 @@ struct CompactionJob::SubcompactionState {
     outputs = std::move(o.outputs);
     outfile = std::move(o.outfile);
     builder = std::move(o.builder);
+    current_output_file_size = std::move(o.current_output_file_size);
     total_bytes = std::move(o.total_bytes);
     num_input_records = std::move(o.num_input_records);
     num_output_records = std::move(o.num_output_records);
@@ -161,7 +165,7 @@ struct CompactionJob::SubcompactionState {
 
   // Returns true iff we should stop building the current output
   // before processing "internal_key".
-  bool ShouldStopBefore(const Slice& internal_key) {
+  bool ShouldStopBefore(const Slice& internal_key, uint64_t curr_file_size) {
     const InternalKeyComparator* icmp =
         &compaction->column_family_data()->internal_comparator();
     const std::vector<FileMetaData*>& grandparents = compaction->grandparents();
@@ -182,7 +186,8 @@ struct CompactionJob::SubcompactionState {
     }
     seen_key = true;
 
-    if (overlapped_bytes > compaction->max_grandparent_overlap_bytes()) {
+    if (overlapped_bytes + curr_file_size >
+        compaction->max_compaction_bytes()) {
       // Too much overlap for current output; start new output
       overlapped_bytes = 0;
       return true;
@@ -624,6 +629,13 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options) {
          << "num_output_records" << compact_->num_output_records
          << "num_subcompactions" << compact_->sub_compact_states.size();
 
+  if (compaction_job_stats_ != nullptr) {
+    stream << "num_single_delete_mismatches"
+           << compaction_job_stats_->num_single_del_mismatch;
+    stream << "num_single_delete_fallthrough"
+           << compaction_job_stats_->num_single_del_fallthru;
+  }
+
   if (measure_io_stats_ && compaction_job_stats_ != nullptr) {
     stream << "file_write_nanos" << compaction_job_stats_->file_write_nanos;
     stream << "file_range_sync_nanos"
@@ -750,7 +762,9 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     if (end != nullptr &&
         cfd->user_comparator()->Compare(c_iter->user_key(), *end) >= 0) {
       break;
-    } else if (sub_compact->ShouldStopBefore(key) &&
+    } else if (sub_compact->compaction->output_level() != 0 &&
+               sub_compact->ShouldStopBefore(
+                   key, sub_compact->current_output_file_size) &&
                sub_compact->builder != nullptr) {
       status = FinishCompactionOutputFile(input->status(), sub_compact);
       if (!status.ok()) {
@@ -775,6 +789,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     assert(sub_compact->builder != nullptr);
     assert(sub_compact->current_output() != nullptr);
     sub_compact->builder->Add(key, value);
+    sub_compact->current_output_file_size = sub_compact->builder->FileSize();
     sub_compact->current_output()->meta.UpdateBoundaries(
         key, c_iter->ikey().sequence);
     sub_compact->num_output_records++;
@@ -833,7 +848,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     // during subcompactions (i.e. if output size, estimated by input size, is
     // going to be 1.2MB and max_output_file_size = 1MB, prefer to have 0.6MB
     // and 0.6MB instead of 1MB and 0.2MB)
-    if (sub_compact->builder->FileSize() >=
+    if (sub_compact->current_output_file_size >=
         sub_compact->compaction->max_output_file_size()) {
       status = FinishCompactionOutputFile(input->status(), sub_compact);
       if (sub_compact->outputs.size() == 1) {
@@ -850,6 +865,10 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       c_iter_stats.num_input_deletion_records;
   sub_compact->compaction_job_stats.num_corrupt_keys =
       c_iter_stats.num_input_corrupt_records;
+  sub_compact->compaction_job_stats.num_single_del_fallthru =
+      c_iter_stats.num_single_del_fallthru;
+  sub_compact->compaction_job_stats.num_single_del_mismatch =
+      c_iter_stats.num_single_del_mismatch;
   sub_compact->compaction_job_stats.total_input_raw_key_bytes +=
       c_iter_stats.total_input_raw_key_bytes;
   sub_compact->compaction_job_stats.total_input_raw_value_bytes +=
@@ -1010,6 +1029,7 @@ Status CompactionJob::FinishCompactionOutputFile(
   }
 
   sub_compact->builder.reset();
+  sub_compact->current_output_file_size = 0;
   return s;
 }
 

db/compaction_job_stats_test.cc

@@ -957,7 +957,7 @@ TEST_P(CompactionJobStatsTest, UniversalCompactionTest) {
   uint64_t key_base = 100000000l;
   // Note: key_base must be multiple of num_keys_per_L0_file
   int num_keys_per_table = 100;
-  const uint32_t kTestScale = 8;
+  const uint32_t kTestScale = 6;
   const int kKeySize = 10;
   const int kValueSize = 900;
   double compression_ratio = 1.0;
@@ -1008,8 +1008,9 @@ TEST_P(CompactionJobStatsTest, UniversalCompactionTest) {
             num_input_units,
             num_keys_per_table * num_input_units,
             1.0, 0, false));
+    dbfull()->TEST_WaitForCompact();
   }
-  ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 4U);
+  ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 3U);
 
   for (uint64_t start_key = key_base;
                 start_key <= key_base * kTestScale;

db/compaction_job_test.cc

@@ -241,8 +241,8 @@ class CompactionJobTest : public testing::Test {
 
     Compaction compaction(cfd->current()->storage_info(),
                           *cfd->GetLatestMutableCFOptions(),
-                          compaction_input_files, 1, 1024 * 1024, 10, 0,
-                          kNoCompression, {}, true);
+                          compaction_input_files, 1, 1024 * 1024,
+                          10 * 1024 * 1024, 0, kNoCompression, {}, true);
     compaction.SetInputVersion(cfd->current());
 
     LogBuffer log_buffer(InfoLogLevel::INFO_LEVEL, db_options_.info_log.get());

db/compaction_picker.cc

@@ -266,10 +266,6 @@ Compaction* CompactionPicker::FormCompaction(
     const std::vector<CompactionInputFiles>& input_files, int output_level,
     VersionStorageInfo* vstorage, const MutableCFOptions& mutable_cf_options,
     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());
 
   // TODO(rven ): we might be able to run concurrent level 0 compaction
@@ -278,10 +274,11 @@ Compaction* CompactionPicker::FormCompaction(
   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);
+  auto c =
+      new Compaction(vstorage, mutable_cf_options, input_files, output_level,
+                     compact_options.output_file_size_limit,
+                     mutable_cf_options.max_compaction_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
@@ -390,6 +387,9 @@ bool CompactionPicker::SetupOtherInputs(
   vstorage->GetOverlappingInputs(output_level, &smallest, &largest,
                                  &output_level_inputs->files, *parent_index,
                                  parent_index);
+  if (!output_level_inputs->empty()) {
+    ExpandWhileOverlapping(cf_name, vstorage, output_level_inputs);
+  }
 
   if (FilesInCompaction(output_level_inputs->files)) {
     return false;
@@ -413,8 +413,7 @@ bool CompactionPicker::SetupOtherInputs(
     const uint64_t inputs1_size =
         TotalCompensatedFileSize(output_level_inputs->files);
     const uint64_t expanded0_size = TotalCompensatedFileSize(expanded0.files);
-    uint64_t limit =
-        mutable_cf_options.ExpandedCompactionByteSizeLimit(input_level);
+    uint64_t limit = mutable_cf_options.max_compaction_bytes;
     if (expanded0.size() > inputs->size() &&
         inputs1_size + expanded0_size < limit &&
         !FilesInCompaction(expanded0.files) &&
@@ -510,7 +509,7 @@ Compaction* CompactionPicker::CompactRange(
     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,
+        /* max_compaction_bytes */ LLONG_MAX, output_path_id,
         GetCompressionType(ioptions_, vstorage, mutable_cf_options,
                            output_level, 1),
         /* grandparents */ {}, /* is manual */ true);
@@ -548,8 +547,7 @@ Compaction* CompactionPicker::CompactRange(
   // and we must not pick one file and drop another older file if the
   // two files overlap.
   if (input_level > 0) {
-    const uint64_t limit = mutable_cf_options.MaxFileSizeForLevel(input_level) *
-                           mutable_cf_options.source_compaction_factor;
+    const uint64_t limit = mutable_cf_options.max_compaction_bytes;
     uint64_t total = 0;
     for (size_t i = 0; i + 1 < inputs.size(); ++i) {
       uint64_t s = inputs[i]->compensated_file_size;
@@ -611,8 +609,7 @@ Compaction* CompactionPicker::CompactRange(
   Compaction* compaction = new Compaction(
       vstorage, mutable_cf_options, std::move(compaction_inputs), output_level,
       mutable_cf_options.MaxFileSizeForLevel(output_level),
-      mutable_cf_options.MaxGrandParentOverlapBytes(input_level),
-      output_path_id,
+      mutable_cf_options.max_compaction_bytes, output_path_id,
       GetCompressionType(ioptions_, vstorage, mutable_cf_options, output_level,
                          vstorage->base_level()),
       std::move(grandparents), /* is manual compaction */ true);
@@ -1017,7 +1014,7 @@ Compaction* LevelCompactionPicker::PickCompaction(
   auto c = new Compaction(
       vstorage, mutable_cf_options, std::move(compaction_inputs), output_level,
       mutable_cf_options.MaxFileSizeForLevel(output_level),
-      mutable_cf_options.MaxGrandParentOverlapBytes(level),
+      mutable_cf_options.max_compaction_bytes,
       GetPathId(ioptions_, mutable_cf_options, output_level),
       GetCompressionType(ioptions_, vstorage, mutable_cf_options, output_level,
                          vstorage->base_level()),
@@ -1310,6 +1307,8 @@ Compaction* UniversalCompactionPicker::PickCompaction(
       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());
+    TEST_SYNC_POINT_CALLBACK("UniversalCompactionPicker::PickCompaction:Return",
+                             nullptr);
     return nullptr;
   }
   VersionStorageInfo::LevelSummaryStorage tmp;
@@ -1344,19 +1343,34 @@ Compaction* UniversalCompactionPicker::PickCompaction(
       assert(sorted_runs.size() >=
              static_cast<size_t>(
                  mutable_cf_options.level0_file_num_compaction_trigger));
-      unsigned int num_files =
-          static_cast<unsigned int>(sorted_runs.size()) -
-          mutable_cf_options.level0_file_num_compaction_trigger;
-      if ((c = PickCompactionUniversalReadAmp(
-               cf_name, mutable_cf_options, vstorage, score, UINT_MAX,
-               num_files, sorted_runs, log_buffer)) != nullptr) {
-        LogToBuffer(log_buffer,
-                    "[%s] Universal: compacting for file num -- %u\n",
-                    cf_name.c_str(), num_files);
+      // Get the total number of sorted runs that are not being compacted
+      int num_sr_not_compacted = 0;
+      for (size_t i = 0; i < sorted_runs.size(); i++) {
+        if (sorted_runs[i].being_compacted == false) {
+          num_sr_not_compacted++;
+        }
+      }
+
+      // The number of sorted runs that are not being compacted is greater than
+      // the maximum allowed number of sorted runs
+      if (num_sr_not_compacted >
+          mutable_cf_options.level0_file_num_compaction_trigger) {
+        unsigned int num_files =
+            num_sr_not_compacted -
+            mutable_cf_options.level0_file_num_compaction_trigger + 1;
+        if ((c = PickCompactionUniversalReadAmp(
+                 cf_name, mutable_cf_options, vstorage, score, UINT_MAX,
+                 num_files, sorted_runs, log_buffer)) != nullptr) {
+          LogToBuffer(log_buffer,
+                      "[%s] Universal: compacting for file num -- %u\n",
+                      cf_name.c_str(), num_files);
+        }
       }
     }
   }
   if (c == nullptr) {
+    TEST_SYNC_POINT_CALLBACK("UniversalCompactionPicker::PickCompaction:Return",
+                             nullptr);
     return nullptr;
   }
 
@@ -1409,6 +1423,8 @@ Compaction* UniversalCompactionPicker::PickCompaction(
 
   level0_compactions_in_progress_.insert(c);
 
+  TEST_SYNC_POINT_CALLBACK("UniversalCompactionPicker::PickCompaction:Return",
+                           c);
   return c;
 }
 

db/compaction_picker.h

@@ -20,8 +20,7 @@
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
 #include "rocksdb/status.h"
-#include "util/mutable_cf_options.h"
-
+#include "util/cf_options.h"
 
 namespace rocksdb {
 

db/compaction_picker_test.cc

@@ -621,9 +621,9 @@ TEST_F(CompactionPickerTest, OverlappingUserKeys2) {
   // Overlapping user keys on same level and output level
   Add(1, 1U, "200", "400", 1000000000U);
   Add(1, 2U, "400", "500", 1U, 0, 0);
-  Add(2, 3U, "400", "600", 1U);
-  // The following file is not in the compaction despite overlapping user keys
-  Add(2, 4U, "600", "700", 1U, 0, 0);
+  Add(2, 3U, "000", "100", 1U);
+  Add(2, 4U, "100", "600", 1U, 0, 0);
+  Add(2, 5U, "600", "700", 1U, 0, 0);
   UpdateVersionStorageInfo();
 
   std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
@@ -631,10 +631,12 @@ TEST_F(CompactionPickerTest, OverlappingUserKeys2) {
   ASSERT_TRUE(compaction.get() != nullptr);
   ASSERT_EQ(2U, compaction->num_input_levels());
   ASSERT_EQ(2U, compaction->num_input_files(0));
-  ASSERT_EQ(1U, compaction->num_input_files(1));
+  ASSERT_EQ(3U, compaction->num_input_files(1));
   ASSERT_EQ(1U, compaction->input(0, 0)->fd.GetNumber());
   ASSERT_EQ(2U, compaction->input(0, 1)->fd.GetNumber());
   ASSERT_EQ(3U, compaction->input(1, 0)->fd.GetNumber());
+  ASSERT_EQ(4U, compaction->input(1, 1)->fd.GetNumber());
+  ASSERT_EQ(5U, compaction->input(1, 2)->fd.GetNumber());
 }
 
 TEST_F(CompactionPickerTest, OverlappingUserKeys3) {
@@ -666,6 +668,31 @@ TEST_F(CompactionPickerTest, OverlappingUserKeys3) {
   ASSERT_EQ(7U, compaction->input(1, 1)->fd.GetNumber());
 }
 
+TEST_F(CompactionPickerTest, OverlappingUserKeys4) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.max_bytes_for_level_base = 1000000;
+
+  Add(1, 1U, "100", "150", 1U);
+  Add(1, 2U, "150", "199", 1U, 0, 0);
+  Add(1, 3U, "200", "250", 1100000U, 0, 0);
+  Add(1, 4U, "251", "300", 1U, 0, 0);
+  Add(1, 5U, "300", "350", 1U, 0, 0);
+
+  Add(2, 6U, "100", "115", 1U);
+  Add(2, 7U, "125", "325", 1U);
+  Add(2, 8U, "350", "400", 1U);
+  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(2U, compaction->num_input_levels());
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  ASSERT_EQ(1U, compaction->num_input_files(1));
+  ASSERT_EQ(3U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(7U, compaction->input(1, 0)->fd.GetNumber());
+}
+
 TEST_F(CompactionPickerTest, NotScheduleL1IfL0WithHigherPri1) {
   NewVersionStorage(6, kCompactionStyleLevel);
   mutable_cf_options_.level0_file_num_compaction_trigger = 2;
@@ -1034,6 +1061,57 @@ TEST_F(CompactionPickerTest, IsBottommostLevelTest) {
   DeleteVersionStorage();
 }
 
+TEST_F(CompactionPickerTest, MaxCompactionBytesHit) {
+  mutable_cf_options_.max_bytes_for_level_base = 1000000u;
+  mutable_cf_options_.max_compaction_bytes = 800000u;
+  ioptions_.level_compaction_dynamic_level_bytes = false;
+  NewVersionStorage(6, kCompactionStyleLevel);
+  // A compaction should be triggered and pick file 2 and 5.
+  // It cannot expand because adding file 1 and 3, the compaction size will
+  // exceed mutable_cf_options_.max_bytes_for_level_base.
+  Add(1, 1U, "100", "150", 300000U);
+  Add(1, 2U, "151", "200", 300001U, 0, 0);
+  Add(1, 3U, "201", "250", 300000U, 0, 0);
+  Add(1, 4U, "251", "300", 300000U, 0, 0);
+  Add(2, 5U, "160", "256", 1U);
+  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(2U, compaction->num_input_levels());
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  ASSERT_EQ(1U, compaction->num_input_files(1));
+  ASSERT_EQ(2U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(5U, compaction->input(1, 0)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, MaxCompactionBytesNotHit) {
+  mutable_cf_options_.max_bytes_for_level_base = 1000000u;
+  mutable_cf_options_.max_compaction_bytes = 1000000u;
+  ioptions_.level_compaction_dynamic_level_bytes = false;
+  NewVersionStorage(6, kCompactionStyleLevel);
+  // A compaction should be triggered and pick file 2 and 5.
+  // and it expands to file 1 and 3 too.
+  Add(1, 1U, "100", "150", 300000U);
+  Add(1, 2U, "151", "200", 300001U, 0, 0);
+  Add(1, 3U, "201", "250", 300000U, 0, 0);
+  Add(1, 4U, "251", "300", 300000U, 0, 0);
+  Add(2, 5U, "000", "233", 1U);
+  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(2U, compaction->num_input_levels());
+  ASSERT_EQ(3U, compaction->num_input_files(0));
+  ASSERT_EQ(1U, compaction->num_input_files(1));
+  ASSERT_EQ(1U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(2U, compaction->input(0, 1)->fd.GetNumber());
+  ASSERT_EQ(3U, compaction->input(0, 2)->fd.GetNumber());
+  ASSERT_EQ(5U, compaction->input(1, 0)->fd.GetNumber());
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/db_block_cache_test.cc

@@ -9,6 +9,7 @@
 #include <cstdlib>
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
+#include "util/lru_cache.h"
 
 namespace rocksdb {
 
@@ -171,7 +172,7 @@ TEST_F(DBBlockCacheTest, TestWithCompressedBlockCache) {
   InitTable(options);
 
   std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
-  std::shared_ptr<Cache> compressed_cache = NewLRUCache(0, 0, false);
+  std::shared_ptr<Cache> compressed_cache = NewLRUCache(1 << 25, 0, false);
   table_options.block_cache = cache;
   table_options.block_cache_compressed = compressed_cache;
   options.table_factory.reset(new BlockBasedTableFactory(table_options));
@@ -203,9 +204,6 @@ TEST_F(DBBlockCacheTest, TestWithCompressedBlockCache) {
   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());
@@ -270,16 +268,16 @@ TEST_F(DBBlockCacheTest, IndexAndFilterBlocksOfNewTableAddedToCache) {
   ASSERT_EQ(2, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
 
   // Make sure index block is in cache.
-  auto index_block_hit = TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT);
+  auto index_block_hit = TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT);
   value = Get(1, "key");
-  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
   ASSERT_EQ(index_block_hit + 1,
-            TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+            TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
 
   value = Get(1, "key");
-  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
   ASSERT_EQ(index_block_hit + 2,
-            TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+            TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
 }
 
 TEST_F(DBBlockCacheTest, IndexAndFilterBlocksStats) {
@@ -323,6 +321,91 @@ TEST_F(DBBlockCacheTest, IndexAndFilterBlocksStats) {
             filter_bytes_insert);
 }
 
+namespace {
+
+// A mock cache wraps LRUCache, and record how many entries have been
+// inserted for each priority.
+class MockCache : public LRUCache {
+ public:
+  static uint32_t high_pri_insert_count;
+  static uint32_t low_pri_insert_count;
+
+  MockCache() : LRUCache(1 << 25, 0, false, 0.0) {}
+
+  virtual Status Insert(const Slice& key, void* value, size_t charge,
+                        void (*deleter)(const Slice& key, void* value),
+                        Handle** handle, Priority priority) override {
+    if (priority == Priority::LOW) {
+      low_pri_insert_count++;
+    } else {
+      high_pri_insert_count++;
+    }
+    return LRUCache::Insert(key, value, charge, deleter, handle, priority);
+  }
+};
+
+uint32_t MockCache::high_pri_insert_count = 0;
+uint32_t MockCache::low_pri_insert_count = 0;
+
+}  // anonymous namespace
+
+TEST_F(DBBlockCacheTest, IndexAndFilterBlocksCachePriority) {
+  for (auto priority : {Cache::Priority::LOW, Cache::Priority::HIGH}) {
+    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.block_cache.reset(new MockCache());
+    table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+    table_options.cache_index_and_filter_blocks_with_high_priority =
+        priority == Cache::Priority::HIGH ? true : false;
+    options.table_factory.reset(new BlockBasedTableFactory(table_options));
+    DestroyAndReopen(options);
+
+    MockCache::high_pri_insert_count = 0;
+    MockCache::low_pri_insert_count = 0;
+
+    // Create a new table.
+    ASSERT_OK(Put("foo", "value"));
+    ASSERT_OK(Put("bar", "value"));
+    ASSERT_OK(Flush());
+    ASSERT_EQ(1, NumTableFilesAtLevel(0));
+
+    // index/filter blocks added to block cache right after table creation.
+    ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+    ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+    ASSERT_EQ(2, /* only index/filter were added */
+              TestGetTickerCount(options, BLOCK_CACHE_ADD));
+    ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_DATA_MISS));
+    if (priority == Cache::Priority::LOW) {
+      ASSERT_EQ(0, MockCache::high_pri_insert_count);
+      ASSERT_EQ(2, MockCache::low_pri_insert_count);
+    } else {
+      ASSERT_EQ(2, MockCache::high_pri_insert_count);
+      ASSERT_EQ(0, MockCache::low_pri_insert_count);
+    }
+
+    // Access data block.
+    ASSERT_EQ("value", Get("foo"));
+
+    ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+    ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+    ASSERT_EQ(3, /*adding data block*/
+              TestGetTickerCount(options, BLOCK_CACHE_ADD));
+    ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_DATA_MISS));
+
+    // Data block should be inserted with low priority.
+    if (priority == Cache::Priority::LOW) {
+      ASSERT_EQ(0, MockCache::high_pri_insert_count);
+      ASSERT_EQ(3, MockCache::low_pri_insert_count);
+    } else {
+      ASSERT_EQ(2, MockCache::high_pri_insert_count);
+      ASSERT_EQ(1, MockCache::low_pri_insert_count);
+    }
+  }
+}
+
 TEST_F(DBBlockCacheTest, ParanoidFileChecks) {
   Options options = CurrentOptions();
   options.create_if_missing = true;

db/db_compaction_test.cc

@@ -649,6 +649,72 @@ TEST_F(DBCompactionTest, MinorCompactionsHappen) {
   } while (ChangeCompactOptions());
 }
 
+TEST_F(DBCompactionTest, UserKeyCrossFile1) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleLevel;
+  options.level0_file_num_compaction_trigger = 3;
+
+  DestroyAndReopen(options);
+
+  // create first file and flush to l0
+  Put("4", "A");
+  Put("3", "A");
+  Flush();
+  dbfull()->TEST_WaitForFlushMemTable();
+
+  Put("2", "A");
+  Delete("3");
+  Flush();
+  dbfull()->TEST_WaitForFlushMemTable();
+  ASSERT_EQ("NOT_FOUND", Get("3"));
+
+  // move both files down to l1
+  dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  ASSERT_EQ("NOT_FOUND", Get("3"));
+
+  for (int i = 0; i < 3; i++) {
+    Put("2", "B");
+    Flush();
+    dbfull()->TEST_WaitForFlushMemTable();
+  }
+  dbfull()->TEST_WaitForCompact();
+
+  ASSERT_EQ("NOT_FOUND", Get("3"));
+}
+
+TEST_F(DBCompactionTest, UserKeyCrossFile2) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleLevel;
+  options.level0_file_num_compaction_trigger = 3;
+
+  DestroyAndReopen(options);
+
+  // create first file and flush to l0
+  Put("4", "A");
+  Put("3", "A");
+  Flush();
+  dbfull()->TEST_WaitForFlushMemTable();
+
+  Put("2", "A");
+  SingleDelete("3");
+  Flush();
+  dbfull()->TEST_WaitForFlushMemTable();
+  ASSERT_EQ("NOT_FOUND", Get("3"));
+
+  // move both files down to l1
+  dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  ASSERT_EQ("NOT_FOUND", Get("3"));
+
+  for (int i = 0; i < 3; i++) {
+    Put("2", "B");
+    Flush();
+    dbfull()->TEST_WaitForFlushMemTable();
+  }
+  dbfull()->TEST_WaitForCompact();
+
+  ASSERT_EQ("NOT_FOUND", Get("3"));
+}
+
 TEST_F(DBCompactionTest, ZeroSeqIdCompaction) {
   Options options = CurrentOptions();
   options.compaction_style = kCompactionStyleLevel;

db/db_dynamic_level_test.cc

@@ -137,7 +137,7 @@ TEST_F(DBTestDynamicLevel, DynamicLevelMaxBytesBase2) {
   options.max_bytes_for_level_multiplier = 4;
   options.max_background_compactions = 2;
   options.num_levels = 5;
-  options.expanded_compaction_factor = 0;  // Force not expanding in compactions
+  options.max_compaction_bytes = 0;  // Force not expanding in compactions
   BlockBasedTableOptions table_options;
   table_options.block_size = 1024;
   options.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -223,8 +223,9 @@ TEST_F(DBTestDynamicLevel, DynamicLevelMaxBytesBase2) {
   ASSERT_OK(dbfull()->SetOptions({
       {"disable_auto_compactions", "true"},
   }));
-  // Write about 600K more
-  for (int i = 0; i < 1500; i++) {
+  // Write about 650K more.
+  // Each file is about 11KB, with 9KB of data.
+  for (int i = 0; i < 1300; i++) {
     ASSERT_OK(Put(Key(static_cast<int>(rnd.Uniform(kMaxKey))),
                   RandomString(&rnd, 380)));
   }
@@ -294,7 +295,7 @@ TEST_F(DBTestDynamicLevel, DynamicLevelMaxBytesCompactRange) {
   options.max_background_compactions = 1;
   const int kNumLevels = 5;
   options.num_levels = kNumLevels;
-  options.expanded_compaction_factor = 0;  // Force not expanding in compactions
+  options.max_compaction_bytes = 1;  // Force not expanding in compactions
   BlockBasedTableOptions table_options;
   table_options.block_size = 1024;
   options.table_factory.reset(NewBlockBasedTableFactory(table_options));

db/db_impl.cc

@@ -79,6 +79,7 @@
 #include "table/two_level_iterator.h"
 #include "util/autovector.h"
 #include "util/build_version.h"
+#include "util/cf_options.h"
 #include "util/coding.h"
 #include "util/compression.h"
 #include "util/crc32c.h"
@@ -340,6 +341,7 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname)
       next_job_id_(1),
       has_unpersisted_data_(false),
       env_options_(db_options_),
+      num_running_addfile_(0),
 #ifndef ROCKSDB_LITE
       wal_manager_(db_options_, env_options_),
 #endif  // ROCKSDB_LITE
@@ -372,19 +374,6 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname)
 // Will lock the mutex_,  will wait for completion if wait is true
 void DBImpl::CancelAllBackgroundWork(bool wait) {
   InstrumentedMutexLock l(&mutex_);
-  shutting_down_.store(true, std::memory_order_release);
-  bg_cv_.SignalAll();
-  if (!wait) {
-    return;
-  }
-  // Wait for background work to finish
-  while (bg_compaction_scheduled_ || bg_flush_scheduled_) {
-    bg_cv_.Wait();
-  }
-}
-
-DBImpl::~DBImpl() {
-  mutex_.Lock();
 
   if (!shutting_down_.load(std::memory_order_acquire) &&
       has_unpersisted_data_) {
@@ -399,7 +388,19 @@ DBImpl::~DBImpl() {
     }
     versions_->GetColumnFamilySet()->FreeDeadColumnFamilies();
   }
-  mutex_.Unlock();
+
+  shutting_down_.store(true, std::memory_order_release);
+  bg_cv_.SignalAll();
+  if (!wait) {
+    return;
+  }
+  // Wait for background work to finish
+  while (bg_compaction_scheduled_ || bg_flush_scheduled_) {
+    bg_cv_.Wait();
+  }
+}
+
+DBImpl::~DBImpl() {
   // CancelAllBackgroundWork called with false means we just set the shutdown
   // marker. After this we do a variant of the waiting and unschedule work
   // (to consider: moving all the waiting into CancelAllBackgroundWork(true))
@@ -716,6 +717,16 @@ uint64_t DBImpl::FindMinLogContainingOutstandingPrep() {
   return min_log;
 }
 
+void DBImpl::ScheduleBgLogWriterClose(JobContext* job_context) {
+  if (!job_context->logs_to_free.empty()) {
+    for (auto l : job_context->logs_to_free) {
+      AddToLogsToFreeQueue(l);
+    }
+    job_context->logs_to_free.clear();
+    SchedulePurge();
+  }
+}
+
 // * Returns the list of live files in 'sst_live'
 // If it's doing full scan:
 // * Returns the list of all files in the filesystem in
@@ -1383,7 +1394,6 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
   bool stop_replay_by_wal_filter = false;
   bool stop_replay_for_corruption = false;
   bool flushed = false;
-  SequenceNumber recovered_sequence = 0;
   for (auto log_number : log_numbers) {
     // The previous incarnation may not have written any MANIFEST
     // records after allocating this log number.  So we manually
@@ -1462,13 +1472,13 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
       WriteBatchInternal::SetContents(&batch, record);
       SequenceNumber sequence = WriteBatchInternal::Sequence(&batch);
 
-      // In point-in-time recovery mode, if sequence id of log files are
-      // consecutive, we continue recovery despite corruption. This could happen
-      // when we open and write to a corrupted DB, where sequence id will start
-      // from the last sequence id we recovered.
       if (db_options_.wal_recovery_mode ==
           WALRecoveryMode::kPointInTimeRecovery) {
-        if (sequence == recovered_sequence + 1) {
+        // In point-in-time recovery mode, if sequence id of log files are
+        // consecutive, we continue recovery despite corruption. This could
+        // happen when we open and write to a corrupted DB, where sequence id
+        // will start from the last sequence id we recovered.
+        if (sequence == *next_sequence) {
           stop_replay_for_corruption = false;
         }
         if (stop_replay_for_corruption) {
@@ -1477,13 +1487,16 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
         }
       }
 
-      recovered_sequence = sequence;
       bool no_prev_seq = true;
-      if (*next_sequence == kMaxSequenceNumber) {
+      if (!db_options_.allow_2pc) {
         *next_sequence = sequence;
       } else {
-        no_prev_seq = false;
-        WriteBatchInternal::SetSequence(&batch, *next_sequence);
+        if (*next_sequence == kMaxSequenceNumber) {
+          *next_sequence = sequence;
+        } else {
+          no_prev_seq = false;
+          WriteBatchInternal::SetSequence(&batch, *next_sequence);
+        }
       }
 
 #ifndef ROCKSDB_LITE
@@ -1565,23 +1578,25 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
       // insert. We don't want to fail the whole write batch in that case --
       // we just ignore the update.
       // That's why we set ignore missing column families to true
-      //
+      bool has_valid_writes = false;
       // If we pass DB through and options.max_successive_merges is hit
       // during recovery, Get() will be issued which will try to acquire
       // DB mutex and cause deadlock, as DB mutex is already held.
       // The DB pointer is not needed unless 2PC is used.
       // TODO(sdong) fix the allow_2pc case too.
-      bool has_valid_writes = false;
       status = WriteBatchInternal::InsertInto(
           &batch, column_family_memtables_.get(), &flush_scheduler_, true,
-          log_number, db_options_.allow_2pc ? this : nullptr, false /* concurrent_memtable_writes */,
-          next_sequence, &has_valid_writes);
+          log_number, db_options_.allow_2pc ? this : nullptr,
+          false /* concurrent_memtable_writes */, next_sequence,
+          &has_valid_writes);
       // If it is the first log file and there is no column family updated
       // after replaying the file, this file may be a stale file. We ignore
       // sequence IDs from the file. Otherwise, if a newer stale log file that
       // has been deleted, the sequenceID may be wrong.
-      if (no_prev_seq && !has_valid_writes) {
-        *next_sequence = kMaxSequenceNumber;
+      if (db_options_.allow_2pc) {
+        if (no_prev_seq && !has_valid_writes) {
+          *next_sequence = kMaxSequenceNumber;
+        }
       }
       MaybeIgnoreError(&status);
       if (!status.ok()) {
@@ -1691,7 +1706,7 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
       // recovered and should be ignored on next reincarnation.
       // Since we already recovered max_log_number, we want all logs
       // with numbers `<= max_log_number` (includes this one) to be ignored
-      if (flushed) {
+      if (flushed || cfd->mem()->GetFirstSequenceNumber() == 0) {
         edit->SetLogNumber(max_log_number + 1);
       }
       // we must mark the next log number as used, even though it's
@@ -1787,6 +1802,49 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
   return s;
 }
 
+Status DBImpl::SyncClosedLogs(JobContext* job_context) {
+  mutex_.AssertHeld();
+  autovector<log::Writer*, 1> logs_to_sync;
+  uint64_t current_log_number = logfile_number_;
+  while (logs_.front().number < current_log_number &&
+         logs_.front().getting_synced) {
+    log_sync_cv_.Wait();
+  }
+  for (auto it = logs_.begin();
+       it != logs_.end() && it->number < current_log_number; ++it) {
+    auto& log = *it;
+    assert(!log.getting_synced);
+    log.getting_synced = true;
+    logs_to_sync.push_back(log.writer);
+  }
+
+  Status s;
+  if (!logs_to_sync.empty()) {
+    mutex_.Unlock();
+
+    for (log::Writer* log : logs_to_sync) {
+      Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
+          "[JOB %d] Syncing log #%" PRIu64, job_context->job_id,
+          log->get_log_number());
+      s = log->file()->Sync(db_options_.use_fsync);
+    }
+    if (s.ok()) {
+      s = directories_.GetWalDir()->Fsync();
+    }
+
+    mutex_.Lock();
+
+    // "number <= current_log_number - 1" is equivalent to
+    // "number < current_log_number".
+    MarkLogsSynced(current_log_number - 1, true, s);
+    if (!s.ok()) {
+      bg_error_ = s;
+      return s;
+    }
+  }
+  return s;
+}
+
 Status DBImpl::FlushMemTableToOutputFile(
     ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
     bool* made_progress, JobContext* job_context, LogBuffer* log_buffer) {
@@ -1808,13 +1866,30 @@ Status DBImpl::FlushMemTableToOutputFile(
 
   FileMetaData file_meta;
 
+  flush_job.PickMemTable();
+
+  Status s;
+  if (logfile_number_ > 0 &&
+      versions_->GetColumnFamilySet()->NumberOfColumnFamilies() > 0 &&
+      !db_options_.disableDataSync) {
+    // If there are more than one column families, we need to make sure that
+    // all the log files except the most recent one are synced. Otherwise if
+    // the host crashes after flushing and before WAL is persistent, the
+    // flushed SST may contain data from write batches whose updates to
+    // other column families are missing.
+    // SyncClosedLogs() may unlock and re-lock the db_mutex.
+    s = SyncClosedLogs(job_context);
+  }
+
   // Within flush_job.Run, rocksdb may call event listener to notify
   // file creation and deletion.
   //
   // Note that flush_job.Run will unlock and lock the db_mutex,
   // and EventListener callback will be called when the db_mutex
   // is unlocked by the current thread.
-  Status s = flush_job.Run(&file_meta);
+  if (s.ok()) {
+    s = flush_job.Run(&file_meta);
+  }
 
   if (s.ok()) {
     InstallSuperVersionAndScheduleWorkWrapper(cfd, job_context,
@@ -2035,6 +2110,10 @@ Status DBImpl::CompactFiles(
   {
     InstrumentedMutexLock l(&mutex_);
 
+    // This call will unlock/lock the mutex to wait for current running
+    // AddFile() calls to finish.
+    WaitForAddFile();
+
     s = CompactFilesImpl(compact_options, cfd, sv->current,
                          input_file_names, output_level,
                          output_path_id, &job_context, &log_buffer);
@@ -2304,20 +2383,6 @@ void DBImpl::NotifyOnCompactionCompleted(
 #endif  // ROCKSDB_LITE
 }
 
-bool DBImpl::NeedFlushOrCompaction(const MutableCFOptions& base_options,
-                                   const MutableCFOptions& new_options) {
-  return (base_options.disable_auto_compactions &&
-          !new_options.disable_auto_compactions) ||
-         base_options.level0_slowdown_writes_trigger <
-             new_options.level0_slowdown_writes_trigger ||
-         base_options.level0_stop_writes_trigger <
-             new_options.level0_stop_writes_trigger ||
-         base_options.soft_pending_compaction_bytes_limit <
-             new_options.soft_pending_compaction_bytes_limit ||
-         base_options.hard_pending_compaction_bytes_limit <
-             new_options.hard_pending_compaction_bytes_limit;
-}
-
 Status DBImpl::SetOptions(ColumnFamilyHandle* column_family,
     const std::unordered_map<std::string, std::string>& options_map) {
 #ifdef ROCKSDB_LITE
@@ -2331,7 +2396,6 @@ Status DBImpl::SetOptions(ColumnFamilyHandle* column_family,
     return Status::InvalidArgument("empty input");
   }
 
-  MutableCFOptions prev_options = *cfd->GetLatestMutableCFOptions();
   MutableCFOptions new_options;
   Status s;
   Status persist_options_status;
@@ -2340,14 +2404,12 @@ Status DBImpl::SetOptions(ColumnFamilyHandle* column_family,
     s = cfd->SetOptions(options_map);
     if (s.ok()) {
       new_options = *cfd->GetLatestMutableCFOptions();
-      if (NeedFlushOrCompaction(prev_options, new_options)) {
-        // Trigger possible flush/compactions. This has to be before we persist
-        // options to file, otherwise there will be a deadlock with writer
-        // thread.
-        auto* old_sv =
-            InstallSuperVersionAndScheduleWork(cfd, nullptr, new_options);
-        delete old_sv;
-      }
+      // Trigger possible flush/compactions. This has to be before we persist
+      // options to file, otherwise there will be a deadlock with writer
+      // thread.
+      auto* old_sv =
+          InstallSuperVersionAndScheduleWork(cfd, nullptr, new_options);
+      delete old_sv;
 
       // Persist RocksDB options under the single write thread
       WriteThread::Writer w;
@@ -2565,12 +2627,11 @@ Status DBImpl::SyncWAL() {
   }
 
   TEST_SYNC_POINT("DBImpl::SyncWAL:BeforeMarkLogsSynced:1");
-  TEST_SYNC_POINT("DBImpl::SyncWAL:BeforeMarkLogsSynced:2");
-
   {
     InstrumentedMutexLock l(&mutex_);
     MarkLogsSynced(current_log_number, need_log_dir_sync, status);
   }
+  TEST_SYNC_POINT("DBImpl::SyncWAL:BeforeMarkLogsSynced:2");
 
   return status;
 }
@@ -2644,6 +2705,8 @@ Status DBImpl::RunManualCompaction(ColumnFamilyData* cfd, int input_level,
     manual.end = &end_storage;
   }
 
+  TEST_SYNC_POINT("DBImpl::RunManualCompaction:0");
+  TEST_SYNC_POINT("DBImpl::RunManualCompaction:1");
   InstrumentedMutexLock l(&mutex_);
 
   // When a manual compaction arrives, temporarily disable scheduling of
@@ -2712,6 +2775,10 @@ Status DBImpl::RunManualCompaction(ColumnFamilyData* cfd, int input_level,
       ca->m = &manual;
       manual.incomplete = false;
       bg_compaction_scheduled_++;
+      // manual.compaction will be added to running_compactions_ and erased
+      // inside BackgroundCompaction() but we need to put it now since we
+      // will unlock the mutex.
+      running_compactions_.insert(manual.compaction);
       env_->Schedule(&DBImpl::BGWorkCompaction, ca, Env::Priority::LOW, this,
                      &DBImpl::UnscheduleCallback);
       scheduled = true;
@@ -2952,8 +3019,9 @@ void DBImpl::BGWorkCompaction(void* arg) {
 
 void DBImpl::BGWorkPurge(void* db) {
   IOSTATS_SET_THREAD_POOL_ID(Env::Priority::HIGH);
-  TEST_SYNC_POINT("DBImpl::BGWorkPurge");
+  TEST_SYNC_POINT("DBImpl::BGWorkPurge:start");
   reinterpret_cast<DBImpl*>(db)->BackgroundCallPurge();
+  TEST_SYNC_POINT("DBImpl::BGWorkPurge:end");
 }
 
 void DBImpl::UnscheduleCallback(void* arg) {
@@ -2968,20 +3036,32 @@ void DBImpl::UnscheduleCallback(void* arg) {
 void DBImpl::BackgroundCallPurge() {
   mutex_.Lock();
 
-  while (!purge_queue_.empty()) {
-    auto purge_file = purge_queue_.begin();
-    auto fname = purge_file->fname;
-    auto type = purge_file->type;
-    auto number = purge_file->number;
-    auto path_id = purge_file->path_id;
-    auto job_id = purge_file->job_id;
-    purge_queue_.pop_front();
+  // We use one single loop to clear both queues so that after existing the loop
+  // both queues are empty. This is stricter than what is needed, but can make
+  // it easier for us to reason the correctness.
+  while (!purge_queue_.empty() || !logs_to_free_queue_.empty()) {
+    if (!purge_queue_.empty()) {
+      auto purge_file = purge_queue_.begin();
+      auto fname = purge_file->fname;
+      auto type = purge_file->type;
+      auto number = purge_file->number;
+      auto path_id = purge_file->path_id;
+      auto job_id = purge_file->job_id;
+      purge_queue_.pop_front();
 
-    mutex_.Unlock();
-    Status file_deletion_status;
-    DeleteObsoleteFileImpl(file_deletion_status, job_id, fname, type, number,
-                           path_id);
-    mutex_.Lock();
+      mutex_.Unlock();
+      Status file_deletion_status;
+      DeleteObsoleteFileImpl(file_deletion_status, job_id, fname, type, number,
+                             path_id);
+      mutex_.Lock();
+    } else {
+      assert(!logs_to_free_queue_.empty());
+      log::Writer* log_writer = *(logs_to_free_queue_.begin());
+      logs_to_free_queue_.pop_front();
+      mutex_.Unlock();
+      delete log_writer;
+      mutex_.Lock();
+    }
   }
   bg_purge_scheduled_--;
 
@@ -3048,6 +3128,8 @@ void DBImpl::BackgroundCallFlush() {
   JobContext job_context(next_job_id_.fetch_add(1), true);
   assert(bg_flush_scheduled_);
 
+  TEST_SYNC_POINT("DBImpl::BackgroundCallFlush:start");
+
   LogBuffer log_buffer(InfoLogLevel::INFO_LEVEL, db_options_.info_log.get());
   {
     InstrumentedMutexLock l(&mutex_);
@@ -3119,6 +3201,11 @@ void DBImpl::BackgroundCallCompaction(void* arg) {
   LogBuffer log_buffer(InfoLogLevel::INFO_LEVEL, db_options_.info_log.get());
   {
     InstrumentedMutexLock l(&mutex_);
+
+    // This call will unlock/lock the mutex to wait for current running
+    // AddFile() calls to finish.
+    WaitForAddFile();
+
     num_running_compactions_++;
 
     auto pending_outputs_inserted_elem =
@@ -3290,7 +3377,9 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       // NOTE: try to avoid unnecessary copy of MutableCFOptions if
       // compaction is not necessary. Need to make sure mutex is held
       // until we make a copy in the following code
+      TEST_SYNC_POINT("DBImpl::BackgroundCompaction():BeforePickCompaction");
       c.reset(cfd->PickCompaction(*mutable_cf_options, log_buffer));
+      TEST_SYNC_POINT("DBImpl::BackgroundCompaction():AfterPickCompaction");
       if (c != nullptr) {
         // update statistics
         MeasureTime(stats_, NUM_FILES_IN_SINGLE_COMPACTION,
@@ -3528,6 +3617,11 @@ void DBImpl::RemoveManualCompaction(DBImpl::ManualCompaction* m) {
 }
 
 bool DBImpl::ShouldntRunManualCompaction(ManualCompaction* m) {
+  if (num_running_addfile_ > 0) {
+    // We need to wait for other AddFile() calls to finish
+    // before running a manual compaction.
+    return true;
+  }
   if (m->exclusive) {
     return (bg_compaction_scheduled_ > 0);
   }
@@ -3617,6 +3711,9 @@ static void CleanupIteratorState(void* arg1, void* arg2) {
     state->mu->Lock();
     state->super_version->Cleanup();
     state->db->FindObsoleteFiles(&job_context, false, true);
+    if (state->background_purge) {
+      state->db->ScheduleBgLogWriterClose(&job_context);
+    }
     state->mu->Unlock();
 
     delete state->super_version;

db/db_impl.h

@@ -252,6 +252,7 @@ class DBImpl : public DB {
   //
   // Returns OK or NotFound on success,
   // other status on unexpected error.
+  // TODO(andrewkr): this API need to be aware of range deletion operations
   Status GetLatestSequenceForKey(SuperVersion* sv, const Slice& key,
                                  bool cache_only, SequenceNumber* seq,
                                  bool* found_record_for_key);
@@ -259,10 +260,10 @@ class DBImpl : public DB {
   using DB::AddFile;
   virtual Status AddFile(ColumnFamilyHandle* column_family,
                          const std::vector<ExternalSstFileInfo>& file_info_list,
-                         bool move_file) override;
+                         bool move_file, bool skip_snapshot_check) override;
   virtual Status AddFile(ColumnFamilyHandle* column_family,
                          const std::vector<std::string>& file_path_list,
-                         bool move_file) override;
+                         bool move_file, bool skip_snapshot_check) override;
 
 #endif  // ROCKSDB_LITE
 
@@ -365,6 +366,10 @@ class DBImpl : public DB {
   // compaction status.
   int BGCompactionsAllowed() const;
 
+  // move logs pending closing from job_context to the DB queue and
+  // schedule a purge
+  void ScheduleBgLogWriterClose(JobContext* job_context);
+
   // 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
@@ -493,6 +498,9 @@ class DBImpl : public DB {
 
   void MarkLogAsHavingPrepSectionFlushed(uint64_t log);
   void MarkLogAsContainingPrepSection(uint64_t log);
+  void AddToLogsToFreeQueue(log::Writer* log_writer) {
+    logs_to_free_queue_.push_back(log_writer);
+  }
 
   Status NewDB();
 
@@ -601,6 +609,8 @@ class DBImpl : public DB {
   // and blocked by any other pending_outputs_ calls)
   void ReleaseFileNumberFromPendingOutputs(std::list<uint64_t>::iterator v);
 
+  Status SyncClosedLogs(JobContext* job_context);
+
   // Flush the in-memory write buffer to storage.  Switches to a new
   // log-file/memtable and writes a new descriptor iff successful.
   Status FlushMemTableToOutputFile(ColumnFamilyData* cfd,
@@ -610,7 +620,7 @@ class DBImpl : public DB {
 
   // REQUIRES: log_numbers are sorted in ascending order
   Status RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
-                         SequenceNumber* max_sequence, bool read_only);
+                         SequenceNumber* next_sequence, bool read_only);
 
   // The following two methods are used to flush a memtable to
   // storage. The first one is used at database RecoveryTime (when the
@@ -640,15 +650,24 @@ class DBImpl : public DB {
   int PickLevelForIngestedFile(ColumnFamilyData* cfd,
                                const ExternalSstFileInfo& file_info);
 
-  Status CompactFilesImpl(
-      const CompactionOptions& compact_options, ColumnFamilyData* cfd,
-      Version* version, const std::vector<std::string>& input_file_names,
-      const int output_level, int output_path_id, JobContext* job_context,
-      LogBuffer* log_buffer);
+  // Wait for current AddFile() calls to finish.
+  // REQUIRES: mutex_ held
+  void WaitForAddFile();
+
+  Status CompactFilesImpl(const CompactionOptions& compact_options,
+                          ColumnFamilyData* cfd, Version* version,
+                          const std::vector<std::string>& input_file_names,
+                          const int output_level, int output_path_id,
+                          JobContext* job_context, LogBuffer* log_buffer);
+
   Status ReadExternalSstFileInfo(ColumnFamilyHandle* column_family,
                                  const std::string& file_path,
                                  ExternalSstFileInfo* file_info);
 
+#else
+  // AddFile is not supported in ROCKSDB_LITE so this function
+  // will be no-op
+  void WaitForAddFile() {}
 #endif  // ROCKSDB_LITE
 
   ColumnFamilyData* GetColumnFamilyDataByName(const std::string& cf_name);
@@ -670,11 +689,6 @@ class DBImpl : public DB {
   Status BackgroundFlush(bool* madeProgress, JobContext* job_context,
                          LogBuffer* log_buffer);
 
-  // Compare options before and after to see whether flush or compaction is
-  // needed immediately after dynamic option change.
-  bool NeedFlushOrCompaction(const MutableCFOptions& base_options,
-                             const MutableCFOptions& new_options);
-
   void PrintStatistics();
 
   // dump rocksdb.stats to LOG
@@ -723,6 +737,7 @@ class DBImpl : public DB {
   // * whenever bg_flush_scheduled_ or bg_purge_scheduled_ value decreases
   // (i.e. whenever a flush is done, even if it didn't make any progress)
   // * whenever there is an error in background purge, flush or compaction
+  // * whenever num_running_addfile_ goes to 0.
   InstrumentedCondVar bg_cv_;
   uint64_t logfile_number_;
   std::deque<uint64_t>
@@ -882,6 +897,9 @@ class DBImpl : public DB {
 
   // A queue to store filenames of the files to be purged
   std::deque<PurgeFileInfo> purge_queue_;
+
+  // A queue to store log writers to close
+  std::deque<log::Writer*> logs_to_free_queue_;
   int unscheduled_flushes_;
   int unscheduled_compactions_;
 
@@ -965,6 +983,10 @@ class DBImpl : public DB {
   // REQUIRES: mutex held
   std::unordered_set<Compaction*> running_compactions_;
 
+  // Number of running AddFile() calls.
+  // REQUIRES: mutex held
+  int num_running_addfile_;
+
 #ifndef ROCKSDB_LITE
   WalManager wal_manager_;
 #endif  // ROCKSDB_LITE

db/db_impl_add_file.cc

@@ -22,21 +22,6 @@
 
 namespace rocksdb {
 
-namespace {
-// RAII Timer
-class StatsTimer {
- public:
-  explicit StatsTimer(Env* env, uint64_t* counter)
-      : env_(env), counter_(counter), start_micros_(env_->NowMicros()) {}
-  ~StatsTimer() { *counter_ += env_->NowMicros() - start_micros_; }
-
- private:
-  Env* env_;
-  uint64_t* counter_;
-  uint64_t start_micros_;
-};
-}  // anonymous namespace
-
 #ifndef ROCKSDB_LITE
 
 Status DBImpl::ReadExternalSstFileInfo(ColumnFamilyHandle* column_family,
@@ -120,14 +105,13 @@ Status DBImpl::ReadExternalSstFileInfo(ColumnFamilyHandle* column_family,
 
 Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
                        const std::vector<std::string>& file_path_list,
-                       bool move_file) {
+                       bool move_file, bool skip_snapshot_check) {
   Status status;
   auto num_files = file_path_list.size();
   if (num_files == 0) {
     return Status::InvalidArgument("The list of files is empty");
   }
 
-  std::vector<ExternalSstFileInfo> file_infos(num_files);
   std::vector<ExternalSstFileInfo> file_info_list(num_files);
   for (size_t i = 0; i < num_files; i++) {
     status = ReadExternalSstFileInfo(column_family, file_path_list[i],
@@ -136,15 +120,16 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
       return status;
     }
   }
-  return AddFile(column_family, file_info_list, move_file);
+  return AddFile(column_family, file_info_list, move_file, skip_snapshot_check);
 }
 
 Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
                        const std::vector<ExternalSstFileInfo>& file_info_list,
-                       bool move_file) {
+                       bool move_file, bool skip_snapshot_check) {
   Status status;
   auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
   ColumnFamilyData* cfd = cfh->cfd();
+  const Comparator* user_cmp = cfd->internal_comparator().user_comparator();
 
   auto num_files = file_info_list.size();
   if (num_files == 0) {
@@ -158,19 +143,17 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
       sorted_file_info_list[i] = &file_info_list[i];
     }
 
-    auto* vstorage = cfd->current()->storage_info();
-    std::sort(
-        sorted_file_info_list.begin(), sorted_file_info_list.end(),
-        [&vstorage, &file_info_list](const ExternalSstFileInfo* info1,
-                                     const ExternalSstFileInfo* info2) {
-          return vstorage->InternalComparator()->user_comparator()->Compare(
-                     info1->smallest_key, info2->smallest_key) < 0;
-        });
+    std::sort(sorted_file_info_list.begin(), sorted_file_info_list.end(),
+              [&user_cmp, &file_info_list](const ExternalSstFileInfo* info1,
+                                           const ExternalSstFileInfo* info2) {
+                return user_cmp->Compare(info1->smallest_key,
+                                         info2->smallest_key) < 0;
+              });
 
     for (size_t i = 0; i < num_files - 1; i++) {
-      if (sorted_file_info_list[i]->largest_key >=
-          sorted_file_info_list[i + 1]->smallest_key) {
-        return Status::NotSupported("Cannot add overlapping range among files");
+      if (user_cmp->Compare(sorted_file_info_list[i]->largest_key,
+                            sorted_file_info_list[i + 1]->smallest_key) >= 0) {
+        return Status::NotSupported("Files have overlapping ranges");
       }
     }
   }
@@ -178,7 +161,7 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
   std::vector<uint64_t> micro_list(num_files, 0);
   std::vector<FileMetaData> meta_list(num_files);
   for (size_t i = 0; i < num_files; i++) {
-    StatsTimer t(env_, &micro_list[i]);
+    StopWatch sw(env_, nullptr, 0, &micro_list[i], false);
     if (file_info_list[i].num_entries == 0) {
       return Status::InvalidArgument("File contain no entries");
     }
@@ -212,7 +195,7 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
   {
     InstrumentedMutexLock l(&mutex_);
     for (size_t i = 0; i < num_files; i++) {
-      StatsTimer t(env_, &micro_list[i]);
+      StopWatch sw(env_, nullptr, 0, &micro_list[i], false);
       pending_outputs_inserted_elem_list[i] =
           CaptureCurrentFileNumberInPendingOutputs();
       meta_list[i].fd = FileDescriptor(versions_->NewFileNumber(), 0,
@@ -224,7 +207,7 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
   std::vector<std::string> db_fname_list(num_files);
   size_t j = 0;
   for (; j < num_files; j++) {
-    StatsTimer t(env_, &micro_list[j]);
+    StopWatch sw(env_, nullptr, 0, &micro_list[j], false);
     db_fname_list[j] =
         TableFileName(db_options_.db_paths, meta_list[j].fd.GetNumber(),
                       meta_list[j].fd.GetPathId());
@@ -254,13 +237,17 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
 
   {
     InstrumentedMutexLock l(&mutex_);
+    TEST_SYNC_POINT("DBImpl::AddFile:MutexLock");
+
     const MutableCFOptions mutable_cf_options =
         *cfd->GetLatestMutableCFOptions();
 
     WriteThread::Writer w;
     write_thread_.EnterUnbatched(&w, &mutex_);
 
-    if (!snapshots_.empty()) {
+    num_running_addfile_++;
+
+    if (!skip_snapshot_check && !snapshots_.empty()) {
       // Check that no snapshots are being held
       status =
           Status::NotSupported("Cannot add a file while holding snapshots");
@@ -275,7 +262,7 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
       ScopedArenaIterator iter(NewInternalIterator(ro, cfd, sv, &arena));
 
       for (size_t i = 0; i < num_files; i++) {
-        StatsTimer t(env_, &micro_list[i]);
+        StopWatch sw(env_, nullptr, 0, &micro_list[i], false);
         InternalKey range_start(file_info_list[i].smallest_key,
                                 kMaxSequenceNumber, kTypeValue);
         iter->Seek(range_start.Encode());
@@ -284,9 +271,8 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
         if (status.ok() && iter->Valid()) {
           ParsedInternalKey seek_result;
           if (ParseInternalKey(iter->key(), &seek_result)) {
-            auto* vstorage = cfd->current()->storage_info();
-            if (vstorage->InternalComparator()->user_comparator()->Compare(
-                    seek_result.user_key, file_info_list[i].largest_key) <= 0) {
+            if (user_cmp->Compare(seek_result.user_key,
+                                  file_info_list[i].largest_key) <= 0) {
               status = Status::NotSupported("Cannot add overlapping range");
               break;
             }
@@ -298,14 +284,13 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
       }
     }
 
-    // The level the file will be ingested into
+    // The levels that the files will be ingested into
     std::vector<int> target_level_list(num_files, 0);
     if (status.ok()) {
-      // Add files to L0
       VersionEdit edit;
       edit.SetColumnFamily(cfd->GetID());
       for (size_t i = 0; i < num_files; i++) {
-        StatsTimer t(env_, &micro_list[i]);
+        StopWatch sw(env_, nullptr, 0, &micro_list[i], false);
         // Add file to the lowest possible level
         target_level_list[i] = PickLevelForIngestedFile(cfd, file_info_list[i]);
         edit.AddFile(target_level_list[i], meta_list[i].fd.GetNumber(),
@@ -335,7 +320,13 @@ Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
       ReleaseFileNumberFromPendingOutputs(
           pending_outputs_inserted_elem_list[i]);
     }
-  }
+
+    num_running_addfile_--;
+    if (num_running_addfile_ == 0) {
+      bg_cv_.SignalAll();
+    }
+    TEST_SYNC_POINT("DBImpl::AddFile:MutexUnlock");
+  }  // mutex_ is unlocked here;
 
   if (!status.ok()) {
     // We failed to add the files to the database
@@ -414,6 +405,13 @@ int DBImpl::PickLevelForIngestedFile(ColumnFamilyData* cfd,
 
   return target_level;
 }
+
+void DBImpl::WaitForAddFile() {
+  mutex_.AssertHeld();
+  while (num_running_addfile_ > 0) {
+    bg_cv_.Wait();
+  }
+}
 #endif  // ROCKSDB_LITE
 
 }  // namespace rocksdb

db/db_io_failure_test.cc

@@ -111,6 +111,7 @@ TEST_F(DBIOFailureTest, NoSpaceCompactRange) {
     Status s = dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr,
                                            true /* disallow trivial move */);
     ASSERT_TRUE(s.IsIOError());
+    ASSERT_TRUE(s.IsNoSpace());
 
     env_->no_space_.store(false, std::memory_order_release);
   } while (ChangeCompactOptions());

db/db_iter.cc

@@ -133,7 +133,9 @@ class DBIter: public Iterator {
   }
   virtual ~DBIter() {
     // Release pinned data if any
-    pinned_iters_mgr_.ReleasePinnedIterators();
+    if (pinned_iters_mgr_.PinningEnabled()) {
+      pinned_iters_mgr_.ReleasePinnedData();
+    }
     RecordTick(statistics_, NO_ITERATORS, -1);
     local_stats_.BumpGlobalStatistics(statistics_);
     if (!arena_mode_) {
@@ -155,7 +157,9 @@ class DBIter: public Iterator {
   virtual Slice value() const override {
     assert(valid_);
     if (current_entry_is_merged_) {
-      return saved_value_;
+      // If pinned_value_ is set then the result of merge operator is one of
+      // the merge operands and we should return it.
+      return pinned_value_.data() ? pinned_value_ : saved_value_;
     } else if (direction_ == kReverse) {
       return pinned_value_;
     } else {
@@ -221,8 +225,8 @@ class DBIter: public Iterator {
 
   // Release blocks pinned by TempPinData()
   void ReleaseTempPinnedData() {
-    if (!pin_thru_lifetime_) {
-      pinned_iters_mgr_.ReleasePinnedIterators();
+    if (!pin_thru_lifetime_ && pinned_iters_mgr_.PinningEnabled()) {
+      pinned_iters_mgr_.ReleasePinnedData();
     }
   }
 
@@ -286,9 +290,9 @@ inline bool DBIter::ParseKey(ParsedInternalKey* ikey) {
 void DBIter::Next() {
   assert(valid_);
 
+  // Release temporarily pinned blocks from last operation
+  ReleaseTempPinnedData();
   if (direction_ == kReverse) {
-    // We only pin blocks when doing kReverse
-    ReleaseTempPinnedData();
     FindNextUserKey();
     direction_ = kForward;
     if (!iter_->Valid()) {
@@ -433,9 +437,12 @@ void DBIter::MergeValuesNewToOld() {
     return;
   }
 
+  // Temporarily pin the blocks that hold merge operands
+  TempPinData();
   merge_context_.Clear();
   // Start the merge process by pushing the first operand
-  merge_context_.PushOperand(iter_->value());
+  merge_context_.PushOperand(iter_->value(),
+                             iter_->IsValuePinned() /* operand_pinned */);
 
   ParsedInternalKey ikey;
   for (iter_->Next(); iter_->Valid(); iter_->Next()) {
@@ -459,15 +466,15 @@ void DBIter::MergeValuesNewToOld() {
       const Slice val = iter_->value();
       MergeHelper::TimedFullMerge(merge_operator_, ikey.user_key, &val,
                                   merge_context_.GetOperands(), &saved_value_,
-                                  logger_, statistics_, env_);
+                                  logger_, statistics_, env_, &pinned_value_);
       // iter_ is positioned after put
       iter_->Next();
       return;
     } else if (kTypeMerge == ikey.type) {
       // 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();
-      merge_context_.PushOperand(val);
+      merge_context_.PushOperand(iter_->value(),
+                                 iter_->IsValuePinned() /* operand_pinned */);
     } else {
       assert(false);
     }
@@ -479,15 +486,15 @@ void DBIter::MergeValuesNewToOld() {
   // client can differentiate this scenario and do things accordingly.
   MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(), nullptr,
                               merge_context_.GetOperands(), &saved_value_,
-                              logger_, statistics_, env_);
+                              logger_, statistics_, env_, &pinned_value_);
 }
 
 void DBIter::Prev() {
   assert(valid_);
+  ReleaseTempPinnedData();
   if (direction_ == kForward) {
     ReverseToBackward();
   }
-  ReleaseTempPinnedData();
   PrevInternal();
   if (statistics_ != nullptr) {
     local_stats_.prev_count_++;
@@ -580,6 +587,9 @@ bool DBIter::FindValueForCurrentKey() {
   ParsedInternalKey ikey;
   FindParseableKey(&ikey, kReverse);
 
+  // Temporarily pin blocks that hold (merge operands / the value)
+  ReleaseTempPinnedData();
+  TempPinData();
   size_t num_skipped = 0;
   while (iter_->Valid() && ikey.sequence <= sequence_ &&
          user_comparator_->Equal(ikey.user_key, saved_key_.GetKey())) {
@@ -592,8 +602,7 @@ bool DBIter::FindValueForCurrentKey() {
     switch (last_key_entry_type) {
       case kTypeValue:
         merge_context_.Clear();
-        ReleaseTempPinnedData();
-        TempPinData();
+        assert(iter_->IsValuePinned());
         pinned_value_ = iter_->value();
         last_not_merge_type = kTypeValue;
         break;
@@ -605,7 +614,8 @@ bool DBIter::FindValueForCurrentKey() {
         break;
       case kTypeMerge:
         assert(merge_operator_ != nullptr);
-        merge_context_.PushOperandBack(iter_->value());
+        merge_context_.PushOperandBack(
+            iter_->value(), iter_->IsValuePinned() /* operand_pinned */);
         break;
       default:
         assert(false);
@@ -628,13 +638,14 @@ bool DBIter::FindValueForCurrentKey() {
       if (last_not_merge_type == kTypeDeletion) {
         MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(),
                                     nullptr, merge_context_.GetOperands(),
-                                    &saved_value_, logger_, statistics_, env_);
+                                    &saved_value_, logger_, statistics_, env_,
+                                    &pinned_value_);
       } else {
         assert(last_not_merge_type == kTypeValue);
         MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(),
                                     &pinned_value_,
                                     merge_context_.GetOperands(), &saved_value_,
-                                    logger_, statistics_, env_);
+                                    logger_, statistics_, env_, &pinned_value_);
       }
       break;
     case kTypeValue:
@@ -651,6 +662,9 @@ bool DBIter::FindValueForCurrentKey() {
 // This function is used in FindValueForCurrentKey.
 // We use Seek() function instead of Prev() to find necessary value
 bool DBIter::FindValueForCurrentKeyUsingSeek() {
+  // FindValueForCurrentKey will enable pinning before calling
+  // FindValueForCurrentKeyUsingSeek()
+  assert(pinned_iters_mgr_.PinningEnabled());
   std::string last_key;
   AppendInternalKey(&last_key, ParsedInternalKey(saved_key_.GetKey(), sequence_,
                                                  kValueTypeForSeek));
@@ -664,8 +678,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
   if (ikey.type == kTypeValue || ikey.type == kTypeDeletion ||
       ikey.type == kTypeSingleDeletion) {
     if (ikey.type == kTypeValue) {
-      ReleaseTempPinnedData();
-      TempPinData();
+      assert(iter_->IsValuePinned());
       pinned_value_ = iter_->value();
       valid_ = true;
       return true;
@@ -681,7 +694,8 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
   while (iter_->Valid() &&
          user_comparator_->Equal(ikey.user_key, saved_key_.GetKey()) &&
          ikey.type == kTypeMerge) {
-    merge_context_.PushOperand(iter_->value());
+    merge_context_.PushOperand(iter_->value(),
+                               iter_->IsValuePinned() /* operand_pinned */);
     iter_->Next();
     FindParseableKey(&ikey, kForward);
   }
@@ -691,7 +705,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
       ikey.type == kTypeDeletion || ikey.type == kTypeSingleDeletion) {
     MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(), nullptr,
                                 merge_context_.GetOperands(), &saved_value_,
-                                logger_, statistics_, env_);
+                                logger_, statistics_, env_, &pinned_value_);
     // Make iter_ valid and point to saved_key_
     if (!iter_->Valid() ||
         !user_comparator_->Equal(ikey.user_key, saved_key_.GetKey())) {
@@ -705,7 +719,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
   const Slice& val = iter_->value();
   MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(), &val,
                               merge_context_.GetOperands(), &saved_value_,
-                              logger_, statistics_, env_);
+                              logger_, statistics_, env_, &pinned_value_);
   valid_ = true;
   return true;
 }

db/db_iter.h

@@ -10,11 +10,12 @@
 #pragma once
 #include <stdint.h>
 #include <string>
+#include "db/dbformat.h"
 #include "rocksdb/db.h"
 #include "rocksdb/iterator.h"
-#include "db/dbformat.h"
 #include "util/arena.h"
 #include "util/autovector.h"
+#include "util/cf_options.h"
 
 namespace rocksdb {
 

db/db_iter_test.cc

@@ -150,6 +150,9 @@ class TestIterator : public InternalIterator {
     return Status::OK();
   }
 
+  virtual bool IsKeyPinned() const override { return true; }
+  virtual bool IsValuePinned() const override { return true; }
+
  private:
   bool initialized_;
   bool valid_;

db/db_options_test.cc

@@ -6,6 +6,9 @@
 // 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 <limits>
+#include <string>
+
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "util/sync_point.h"
@@ -17,80 +20,113 @@ class DBOptionsTest : public DBTestBase {
   DBOptionsTest() : DBTestBase("/db_options_test") {}
 };
 
-// When write stalls, user can enable auto compaction to unblock writes.
-// However, we had an issue where the stalled write thread blocks the attempt
-// to persist auto compaction option, thus creating a deadlock. The test
-// verifies the issue is fixed.
-TEST_F(DBOptionsTest, EnableAutoCompactionToUnblockWrites) {
-  Options options;
-  options.disable_auto_compactions = true;
-  options.write_buffer_size = 1000 * 1000;  // 1M
-  options.level0_file_num_compaction_trigger = 1;
-  options.level0_slowdown_writes_trigger = 1;
-  options.level0_stop_writes_trigger = 1;
-  options.compression = kNoCompression;
+// RocksDB lite don't support dynamic options.
+#ifndef ROCKSDB_LITE
 
-  SyncPoint::GetInstance()->LoadDependency(
-      {{"DBImpl::DelayWrite:Wait",
-        "DBOptionsTest::EnableAutoCompactionToUnblockWrites:1"},
-       {"DBImpl::BackgroundCompaction:Finish",
-        "DBOptionsTest::EnableAutoCompactionToUnblockWrites:1"}});
-  SyncPoint::GetInstance()->EnableProcessing();
+TEST_F(DBOptionsTest, EnableAutoCompactionAndTriggerStall) {
+  const std::string kValue(1024, 'v');
+  for (int method_type = 0; method_type < 2; method_type++) {
+    for (int option_type = 0; option_type < 4; option_type++) {
+      Options options;
+      options.create_if_missing = true;
+      options.disable_auto_compactions = true;
+      options.write_buffer_size = 1024 * 1024 * 10;
+      options.compression = CompressionType::kNoCompression;
+      options.level0_file_num_compaction_trigger = 1;
+      options.level0_stop_writes_trigger = std::numeric_limits<int>::max();
+      options.level0_slowdown_writes_trigger = std::numeric_limits<int>::max();
+      options.hard_pending_compaction_bytes_limit =
+          std::numeric_limits<uint64_t>::max();
+      options.soft_pending_compaction_bytes_limit =
+          std::numeric_limits<uint64_t>::max();
 
-  // Stall writes.
-  Reopen(options);
-  env_->StartThread(
-      [](void* arg) {
-        std::string value(1000, 'v');
-        auto* t = static_cast<DBOptionsTest*>(arg);
-        for (int i = 0; i < 2000; i++) {
-          ASSERT_OK(t->Put(t->Key(i), value));
-        }
-      },
-      this);
-  TEST_SYNC_POINT("DBOptionsTest::EnableAutoCompactionToUnblockWrites:1");
-  ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
-  ColumnFamilyHandle* handle = dbfull()->DefaultColumnFamily();
-  // We will get a deadlock here if we hit the issue.
-  dbfull()->EnableAutoCompaction({handle});
-  env_->WaitForJoin();
+      DestroyAndReopen(options);
+      int i = 0;
+      for (; i < 1024; i++) {
+        Put(Key(i), kValue);
+      }
+      Flush();
+      for (; i < 1024 * 2; i++) {
+        Put(Key(i), kValue);
+      }
+      Flush();
+      dbfull()->TEST_WaitForFlushMemTable();
+      ASSERT_EQ(2, NumTableFilesAtLevel(0));
+      uint64_t l0_size = SizeAtLevel(0);
+
+      switch (option_type) {
+        case 0:
+          // test with level0_stop_writes_trigger
+          options.level0_stop_writes_trigger = 2;
+          options.level0_slowdown_writes_trigger = 2;
+          break;
+        case 1:
+          options.level0_slowdown_writes_trigger = 2;
+          break;
+        case 2:
+          options.hard_pending_compaction_bytes_limit = l0_size;
+          options.soft_pending_compaction_bytes_limit = l0_size;
+          break;
+        case 3:
+          options.soft_pending_compaction_bytes_limit = l0_size;
+          break;
+      }
+      Reopen(options);
+      dbfull()->TEST_WaitForCompact();
+      ASSERT_FALSE(dbfull()->TEST_write_controler().IsStopped());
+      ASSERT_FALSE(dbfull()->TEST_write_controler().NeedsDelay());
+
+      SyncPoint::GetInstance()->LoadDependency(
+          {{"DBOptionsTest::EnableAutoCompactionAndTriggerStall:1",
+            "BackgroundCallCompaction:0"},
+           {"DBImpl::BackgroundCompaction():BeforePickCompaction",
+            "DBOptionsTest::EnableAutoCompactionAndTriggerStall:2"},
+           {"DBOptionsTest::EnableAutoCompactionAndTriggerStall:3",
+            "DBImpl::BackgroundCompaction():AfterPickCompaction"}});
+      // Block background compaction.
+      SyncPoint::GetInstance()->EnableProcessing();
+
+      switch (method_type) {
+        case 0:
+          ASSERT_OK(
+              dbfull()->SetOptions({{"disable_auto_compactions", "false"}}));
+          break;
+        case 1:
+          ASSERT_OK(dbfull()->EnableAutoCompaction(
+              {dbfull()->DefaultColumnFamily()}));
+          break;
+      }
+      TEST_SYNC_POINT("DBOptionsTest::EnableAutoCompactionAndTriggerStall:1");
+      // Wait for stall condition recalculate.
+      TEST_SYNC_POINT("DBOptionsTest::EnableAutoCompactionAndTriggerStall:2");
+
+      switch (option_type) {
+        case 0:
+          ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
+          break;
+        case 1:
+          ASSERT_FALSE(dbfull()->TEST_write_controler().IsStopped());
+          ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
+          break;
+        case 2:
+          ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
+          break;
+        case 3:
+          ASSERT_FALSE(dbfull()->TEST_write_controler().IsStopped());
+          ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
+          break;
+      }
+      TEST_SYNC_POINT("DBOptionsTest::EnableAutoCompactionAndTriggerStall:3");
+
+      // Background compaction executed.
+      dbfull()->TEST_WaitForCompact();
+      ASSERT_FALSE(dbfull()->TEST_write_controler().IsStopped());
+      ASSERT_FALSE(dbfull()->TEST_write_controler().NeedsDelay());
+    }
+  }
 }
 
-// Similar to EnableAutoCompactionAfterStallDeadlock. See comments there.
-TEST_F(DBOptionsTest, ToggleStopTriggerToUnblockWrites) {
-  Options options;
-  options.disable_auto_compactions = true;
-  options.write_buffer_size = 1000 * 1000;  // 1M
-  options.level0_file_num_compaction_trigger = 1;
-  options.level0_slowdown_writes_trigger = 1;
-  options.level0_stop_writes_trigger = 1;
-  options.compression = kNoCompression;
-
-  SyncPoint::GetInstance()->LoadDependency(
-      {{"DBImpl::DelayWrite:Wait",
-        "DBOptionsTest::ToggleStopTriggerToUnblockWrites:1"},
-       {"DBImpl::BackgroundCompaction:Finish",
-        "DBOptionsTest::ToggleStopTriggerToUnblockWrites:1"}});
-  SyncPoint::GetInstance()->EnableProcessing();
-
-  // Stall writes.
-  Reopen(options);
-  env_->StartThread(
-      [](void* arg) {
-        std::string value(1000, 'v');
-        auto* t = static_cast<DBOptionsTest*>(arg);
-        for (int i = 0; i < 2000; i++) {
-          ASSERT_OK(t->Put(t->Key(i), value));
-        }
-      },
-      this);
-  TEST_SYNC_POINT("DBOptionsTest::ToggleStopTriggerToUnblockWrites:1");
-  ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
-  // We will get a deadlock here if we hit the issue.
-  dbfull()->SetOptions({{"level0_stop_writes_trigger", "1000000"},
-                        {"level0_slowdown_writes_trigger", "1000000"}});
-  env_->WaitForJoin();
-}
+#endif  // ROCKSDB_LITE
 
 }  // namespace rocksdb
 

db/db_properties_test.cc

@@ -594,7 +594,8 @@ TEST_F(DBPropertiesTest, NumImmutableMemTable) {
   } while (ChangeCompactOptions());
 }
 
-TEST_F(DBPropertiesTest, GetProperty) {
+// TODO(techdept) : Disabled flaky test #12863555
+TEST_F(DBPropertiesTest, DISABLED_GetProperty) {
   // Set sizes to both background thread pool to be 1 and block them.
   env_->SetBackgroundThreads(1, Env::HIGH);
   env_->SetBackgroundThreads(1, Env::LOW);

db/db_table_properties_test.cc

@@ -138,14 +138,14 @@ TEST_F(DBTablePropertiesTest, GetPropertiesOfTablesInRange) {
   Options options;
   options.create_if_missing = true;
   options.write_buffer_size = 4096;
-  options.max_write_buffer_number = 8;
+  options.max_write_buffer_number = 3;
   options.level0_file_num_compaction_trigger = 2;
   options.level0_slowdown_writes_trigger = 2;
   options.level0_stop_writes_trigger = 4;
   options.target_file_size_base = 2048;
   options.max_bytes_for_level_base = 10240;
   options.max_bytes_for_level_multiplier = 4;
-  options.soft_rate_limit = 1.1;
+  options.hard_pending_compaction_bytes_limit = 16 * 1024;
   options.num_levels = 8;
 
   DestroyAndReopen(options);
@@ -155,6 +155,12 @@ TEST_F(DBTablePropertiesTest, GetPropertiesOfTablesInRange) {
     ASSERT_OK(Put(test::RandomKey(&rnd, 5), RandomString(&rnd, 102)));
   }
   Flush();
+  dbfull()->TEST_WaitForCompact();
+  if (NumTableFilesAtLevel(0) == 0) {
+    ASSERT_OK(Put(test::RandomKey(&rnd, 5), RandomString(&rnd, 102)));
+    Flush();
+  }
+
   db_->PauseBackgroundWork();
 
   // Ensure that we have at least L0, L1 and L2

db/db_test.cc

@@ -1193,7 +1193,7 @@ bool MinLevelToCompress(CompressionType& type, Options& options, int wbits,
     type = kXpressCompression;
     fprintf(stderr, "using xpress\n");
   } else if (ZSTD_Supported()) {
-    type = kZSTDNotFinalCompression;
+    type = kZSTD;
     fprintf(stderr, "using ZSTD\n");
   } else {
     fprintf(stderr, "skipping test, compression disabled\n");
@@ -2649,12 +2649,12 @@ class ModelDB : public DB {
   using DB::AddFile;
   virtual Status AddFile(ColumnFamilyHandle* column_family,
                          const std::vector<ExternalSstFileInfo>& file_info_list,
-                         bool move_file) override {
+                         bool move_file, bool skip_snapshot_check) override {
     return Status::NotSupported("Not implemented.");
   }
   virtual Status AddFile(ColumnFamilyHandle* column_family,
                          const std::vector<std::string>& file_path_list,
-                         bool move_file) override {
+                         bool move_file, bool skip_snapshot_check) override {
     return Status::NotSupported("Not implemented.");
   }
 
@@ -4271,10 +4271,8 @@ TEST_F(DBTest, DynamicCompactionOptions) {
   options.level0_file_num_compaction_trigger = 3;
   options.level0_slowdown_writes_trigger = 4;
   options.level0_stop_writes_trigger = 8;
-  options.max_grandparent_overlap_factor = 10;
-  options.expanded_compaction_factor = 25;
-  options.source_compaction_factor = 1;
   options.target_file_size_base = k64KB;
+  options.max_compaction_bytes = options.target_file_size_base * 10;
   options.target_file_size_multiplier = 1;
   options.max_bytes_for_level_base = k128KB;
   options.max_bytes_for_level_multiplier = 4;
@@ -4710,7 +4708,7 @@ TEST_F(DBTest, CompressionStatsTest) {
     type = kXpressCompression;
     fprintf(stderr, "using xpress\n");
   } else if (ZSTD_Supported()) {
-    type = kZSTDNotFinalCompression;
+    type = kZSTD;
     fprintf(stderr, "using ZSTD\n");
   } else {
     fprintf(stderr, "skipping test, compression disabled\n");
@@ -4720,7 +4718,7 @@ TEST_F(DBTest, CompressionStatsTest) {
   Options options = CurrentOptions();
   options.compression = type;
   options.statistics = rocksdb::CreateDBStatistics();
-  options.statistics->stats_level_ = StatsLevel::kAll;
+  options.statistics->stats_level_ = StatsLevel::kExceptTimeForMutex;
   DestroyAndReopen(options);
 
   int kNumKeysWritten = 100000;
@@ -4731,6 +4729,7 @@ TEST_F(DBTest, CompressionStatsTest) {
     // compressible string
     ASSERT_OK(Put(Key(i), RandomString(&rnd, 128) + std::string(128, 'a')));
   }
+  ASSERT_OK(Flush());
   ASSERT_GT(options.statistics->getTickerCount(NUMBER_BLOCK_COMPRESSED), 0);
 
   for (int i = 0; i < kNumKeysWritten; ++i) {
@@ -4750,6 +4749,7 @@ TEST_F(DBTest, CompressionStatsTest) {
     // compressible string
     ASSERT_OK(Put(Key(i), RandomString(&rnd, 128) + std::string(128, 'a')));
   }
+  ASSERT_OK(Flush());
   ASSERT_EQ(options.statistics->getTickerCount(NUMBER_BLOCK_COMPRESSED)
             - currentCompressions, 0);
 
@@ -4848,12 +4848,11 @@ class DelayedMergeOperator : public MergeOperator {
 
  public:
   explicit DelayedMergeOperator(DBTest* d) : db_test_(d) {}
-  virtual bool FullMerge(const Slice& key, const Slice* existing_value,
-                         const std::deque<std::string>& operand_list,
-                         std::string* new_value,
-                         Logger* logger) const override {
+
+  virtual bool FullMergeV2(const MergeOperationInput& merge_in,
+                           MergeOperationOutput* merge_out) const override {
     db_test_->env_->addon_time_.fetch_add(1000);
-    *new_value = "";
+    merge_out->new_value = "";
     return true;
   }
 
@@ -5022,7 +5021,7 @@ TEST_F(DBTest, SuggestCompactRangeTest) {
   options.compression = kNoCompression;
   options.max_bytes_for_level_base = 450 << 10;
   options.target_file_size_base = 98 << 10;
-  options.max_grandparent_overlap_factor = 1 << 20;  // inf
+  options.max_compaction_bytes = static_cast<uint64_t>(1) << 60;  // inf
 
   Reopen(options);
 

db/db_test2.cc

@@ -961,9 +961,9 @@ TEST_F(DBTest2, PresetCompressionDict) {
   compression_types.push_back(kLZ4Compression);
   compression_types.push_back(kLZ4HCCompression);
 #endif                          // LZ4_VERSION_NUMBER >= 10400
-#if ZSTD_VERSION_NUMBER >= 500  // v0.5.0+
-  compression_types.push_back(kZSTDNotFinalCompression);
-#endif  // ZSTD_VERSION_NUMBER >= 500
+  if (ZSTD_Supported()) {
+    compression_types.push_back(kZSTD);
+  }
 
   for (auto compression_type : compression_types) {
     options.compression = compression_type;
@@ -1210,6 +1210,37 @@ class PinL0IndexAndFilterBlocksTest : public DBTestBase,
   PinL0IndexAndFilterBlocksTest() : DBTestBase("/db_pin_l0_index_bloom_test") {}
   virtual void SetUp() override { infinite_max_files_ = GetParam(); }
 
+  void CreateTwoLevels(Options* options) {
+    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();
+  }
+
   bool infinite_max_files_;
 };
 
@@ -1261,35 +1292,7 @@ TEST_P(PinL0IndexAndFilterBlocksTest,
 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();
-
+  PinL0IndexAndFilterBlocksTest::CreateTwoLevels(&options);
   // get base cache values
   uint64_t fm = TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS);
   uint64_t fh = TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT);
@@ -1309,14 +1312,125 @@ TEST_P(PinL0IndexAndFilterBlocksTest,
   // 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
+  // When we have inifinite max_files, there is still cache miss because we have
+  // reset the block cache
   value = Get(1, "a");
   ASSERT_EQ(fm + 1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
   ASSERT_EQ(im + 1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
 }
 
+TEST_P(PinL0IndexAndFilterBlocksTest, DisablePrefetchingNonL0IndexAndFilter) {
+  Options options = CurrentOptions();
+  PinL0IndexAndFilterBlocksTest::CreateTwoLevels(&options);
+
+  // 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);
+
+  // Reopen database. If max_open_files is set as -1, table readers will be
+  // preloaded. This will trigger a BlockBasedTable::Open() and prefetch
+  // L0 index and filter. Level 1's prefetching is disabled in DB::Open()
+  TryReopenWithColumnFamilies({"default", "pikachu"}, options);
+
+  if (infinite_max_files_) {
+    // After reopen, cache miss are increased by one because we read (and only
+    // read) filter and index on L0
+    ASSERT_EQ(fm + 1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+    ASSERT_EQ(fh, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+    ASSERT_EQ(im + 1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+    ASSERT_EQ(ih, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+  } else {
+    // If max_open_files is not -1, we do not preload table readers, so there is
+    // no change.
+    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));
+  }
+  std::string value;
+  // this should be read from L0
+  value = Get(1, "a2");
+  // If max_open_files is -1, we have pinned index and filter in Rep, so there
+  // will not be changes in index and filter misses or hits. If max_open_files
+  // is not -1, Get() will open a TableReader and prefetch index and filter.
+  ASSERT_EQ(fm + 1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(fh, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(im + 1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(ih, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+
+  // this should be read from L1
+  value = Get(1, "a");
+  if (infinite_max_files_) {
+    // In inifinite max files case, there's a cache miss in executing Get()
+    // because index and filter are not prefetched before.
+    ASSERT_EQ(fm + 2, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+    ASSERT_EQ(fh, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+    ASSERT_EQ(im + 2, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+    ASSERT_EQ(ih, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+  } else {
+    // In this case, cache miss will be increased by one in
+    // BlockBasedTable::Open() because this is not in DB::Open() code path so we
+    // will prefetch L1's index and filter. Cache hit will also be increased by
+    // one because Get() will read index and filter from the block cache
+    // prefetched in previous Open() call.
+    ASSERT_EQ(fm + 2, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+    ASSERT_EQ(fh + 1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+    ASSERT_EQ(im + 2, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+    ASSERT_EQ(ih + 1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+  }
+}
+
 INSTANTIATE_TEST_CASE_P(PinL0IndexAndFilterBlocksTest,
                         PinL0IndexAndFilterBlocksTest, ::testing::Bool());
+
 #ifndef ROCKSDB_LITE
+TEST_F(DBTest2, MaxCompactionBytesTest) {
+  Options options = CurrentOptions();
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(DBTestBase::kNumKeysByGenerateNewRandomFile));
+  options.compaction_style = kCompactionStyleLevel;
+  options.write_buffer_size = 200 << 10;
+  options.arena_block_size = 4 << 10;
+  options.level0_file_num_compaction_trigger = 4;
+  options.num_levels = 4;
+  options.compression = kNoCompression;
+  options.max_bytes_for_level_base = 450 << 10;
+  options.target_file_size_base = 100 << 10;
+  // Infinite for full compaction.
+  options.max_compaction_bytes = options.target_file_size_base * 100;
+
+  Reopen(options);
+
+  Random rnd(301);
+
+  for (int num = 0; num < 8; num++) {
+    GenerateNewRandomFile(&rnd);
+  }
+  CompactRangeOptions cro;
+  cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
+  ASSERT_EQ("0,0,8", FilesPerLevel(0));
+
+  // When compact from Ln -> Ln+1, cut a file if the file overlaps with
+  // more than three files in Ln+1.
+  options.max_compaction_bytes = options.target_file_size_base * 3;
+  Reopen(options);
+
+  GenerateNewRandomFile(&rnd);
+  // Add three more small files that overlap with the previous file
+  for (int i = 0; i < 3; i++) {
+    Put("a", "z");
+    ASSERT_OK(Flush());
+  }
+  dbfull()->TEST_WaitForCompact();
+
+  // Output files to L1 are cut to three pieces, according to
+  // options.max_compaction_bytes
+  ASSERT_EQ("0,3,8", FilesPerLevel(0));
+}
+
 static void UniqueIdCallback(void* arg) {
   int* result = reinterpret_cast<int*>(arg);
   if (*result == -1) {
@@ -1488,6 +1602,293 @@ TEST_F(DBTest2, SyncPointMarker) {
 }
 #endif
 
+class MergeOperatorPinningTest : public DBTest2,
+                                 public testing::WithParamInterface<bool> {
+ public:
+  MergeOperatorPinningTest() { disable_block_cache_ = GetParam(); }
+
+  bool disable_block_cache_;
+};
+
+INSTANTIATE_TEST_CASE_P(MergeOperatorPinningTest, MergeOperatorPinningTest,
+                        ::testing::Bool());
+
+#ifndef ROCKSDB_LITE
+TEST_P(MergeOperatorPinningTest, OperandsMultiBlocks) {
+  Options options = CurrentOptions();
+  BlockBasedTableOptions table_options;
+  table_options.block_size = 1;  // every block will contain one entry
+  table_options.no_block_cache = disable_block_cache_;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  options.merge_operator = MergeOperators::CreateStringAppendTESTOperator();
+  options.level0_slowdown_writes_trigger = (1 << 30);
+  options.level0_stop_writes_trigger = (1 << 30);
+  options.disable_auto_compactions = true;
+  DestroyAndReopen(options);
+
+  const int kKeysPerFile = 10;
+  const int kOperandsPerKeyPerFile = 7;
+  const int kOperandSize = 100;
+  // Filse to write in L0 before compacting to lower level
+  const int kFilesPerLevel = 3;
+
+  Random rnd(301);
+  std::map<std::string, std::string> true_data;
+  int batch_num = 1;
+  int lvl_to_fill = 4;
+  int key_id = 0;
+  while (true) {
+    for (int j = 0; j < kKeysPerFile; j++) {
+      std::string key = Key(key_id % 35);
+      key_id++;
+      for (int k = 0; k < kOperandsPerKeyPerFile; k++) {
+        std::string val = RandomString(&rnd, kOperandSize);
+        ASSERT_OK(db_->Merge(WriteOptions(), key, val));
+        if (true_data[key].size() == 0) {
+          true_data[key] = val;
+        } else {
+          true_data[key] += "," + val;
+        }
+      }
+    }
+
+    if (lvl_to_fill == -1) {
+      // Keep last batch in memtable and stop
+      break;
+    }
+
+    ASSERT_OK(Flush());
+    if (batch_num % kFilesPerLevel == 0) {
+      if (lvl_to_fill != 0) {
+        MoveFilesToLevel(lvl_to_fill);
+      }
+      lvl_to_fill--;
+    }
+    batch_num++;
+  }
+
+  // 3 L0 files
+  // 1 L1 file
+  // 3 L2 files
+  // 1 L3 file
+  // 3 L4 Files
+  ASSERT_EQ(FilesPerLevel(), "3,1,3,1,3");
+
+  VerifyDBFromMap(true_data);
+}
+
+TEST_P(MergeOperatorPinningTest, Randomized) {
+  do {
+    Options options = CurrentOptions();
+    options.merge_operator = MergeOperators::CreateMaxOperator();
+    BlockBasedTableOptions table_options;
+    table_options.no_block_cache = disable_block_cache_;
+    options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+    DestroyAndReopen(options);
+
+    Random rnd(301);
+    std::map<std::string, std::string> true_data;
+
+    const int kTotalMerges = 10000;
+    // Every key gets ~10 operands
+    const int kKeyRange = kTotalMerges / 10;
+    const int kOperandSize = 20;
+    const int kNumPutBefore = kKeyRange / 10;  // 10% value
+    const int kNumPutAfter = kKeyRange / 10;   // 10% overwrite
+    const int kNumDelete = kKeyRange / 10;     // 10% delete
+
+    // kNumPutBefore keys will have base values
+    for (int i = 0; i < kNumPutBefore; i++) {
+      std::string key = Key(rnd.Next() % kKeyRange);
+      std::string value = RandomString(&rnd, kOperandSize);
+      ASSERT_OK(db_->Put(WriteOptions(), key, value));
+
+      true_data[key] = value;
+    }
+
+    // Do kTotalMerges merges
+    for (int i = 0; i < kTotalMerges; i++) {
+      std::string key = Key(rnd.Next() % kKeyRange);
+      std::string value = RandomString(&rnd, kOperandSize);
+      ASSERT_OK(db_->Merge(WriteOptions(), key, value));
+
+      if (true_data[key] < value) {
+        true_data[key] = value;
+      }
+    }
+
+    // Overwrite random kNumPutAfter keys
+    for (int i = 0; i < kNumPutAfter; i++) {
+      std::string key = Key(rnd.Next() % kKeyRange);
+      std::string value = RandomString(&rnd, kOperandSize);
+      ASSERT_OK(db_->Put(WriteOptions(), key, value));
+
+      true_data[key] = value;
+    }
+
+    // Delete random kNumDelete keys
+    for (int i = 0; i < kNumDelete; i++) {
+      std::string key = Key(rnd.Next() % kKeyRange);
+      ASSERT_OK(db_->Delete(WriteOptions(), key));
+
+      true_data.erase(key);
+    }
+
+    VerifyDBFromMap(true_data);
+
+    // Skip HashCuckoo since it does not support merge operators
+  } while (ChangeOptions(kSkipMergePut | kSkipHashCuckoo));
+}
+
+class MergeOperatorHook : public MergeOperator {
+ public:
+  explicit MergeOperatorHook(std::shared_ptr<MergeOperator> _merge_op)
+      : merge_op_(_merge_op) {}
+
+  virtual bool FullMergeV2(const MergeOperationInput& merge_in,
+                           MergeOperationOutput* merge_out) const override {
+    before_merge_();
+    bool res = merge_op_->FullMergeV2(merge_in, merge_out);
+    after_merge_();
+    return res;
+  }
+
+  virtual const char* Name() const override { return merge_op_->Name(); }
+
+  std::shared_ptr<MergeOperator> merge_op_;
+  std::function<void()> before_merge_ = []() {};
+  std::function<void()> after_merge_ = []() {};
+};
+
+TEST_P(MergeOperatorPinningTest, EvictCacheBeforeMerge) {
+  Options options = CurrentOptions();
+
+  auto merge_hook =
+      std::make_shared<MergeOperatorHook>(MergeOperators::CreateMaxOperator());
+  options.merge_operator = merge_hook;
+  options.disable_auto_compactions = true;
+  options.level0_slowdown_writes_trigger = (1 << 30);
+  options.level0_stop_writes_trigger = (1 << 30);
+  options.max_open_files = 20;
+  BlockBasedTableOptions bbto;
+  bbto.no_block_cache = disable_block_cache_;
+  if (bbto.no_block_cache == false) {
+    bbto.block_cache = NewLRUCache(64 * 1024 * 1024);
+  } else {
+    bbto.block_cache = nullptr;
+  }
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  DestroyAndReopen(options);
+
+  const int kNumOperands = 30;
+  const int kNumKeys = 1000;
+  const int kOperandSize = 100;
+  Random rnd(301);
+
+  // 1000 keys every key have 30 operands, every operand is in a different file
+  std::map<std::string, std::string> true_data;
+  for (int i = 0; i < kNumOperands; i++) {
+    for (int j = 0; j < kNumKeys; j++) {
+      std::string k = Key(j);
+      std::string v = RandomString(&rnd, kOperandSize);
+      ASSERT_OK(db_->Merge(WriteOptions(), k, v));
+
+      true_data[k] = std::max(true_data[k], v);
+    }
+    ASSERT_OK(Flush());
+  }
+
+  std::vector<uint64_t> file_numbers = ListTableFiles(env_, dbname_);
+  ASSERT_EQ(file_numbers.size(), kNumOperands);
+  int merge_cnt = 0;
+
+  // Code executed before merge operation
+  merge_hook->before_merge_ = [&]() {
+    // Evict all tables from cache before every merge operation
+    for (uint64_t num : file_numbers) {
+      TableCache::Evict(dbfull()->TEST_table_cache(), num);
+    }
+    // Decrease cache capacity to force all unrefed blocks to be evicted
+    if (bbto.block_cache) {
+      bbto.block_cache->SetCapacity(1);
+    }
+    merge_cnt++;
+  };
+
+  // Code executed after merge operation
+  merge_hook->after_merge_ = [&]() {
+    // Increase capacity again after doing the merge
+    if (bbto.block_cache) {
+      bbto.block_cache->SetCapacity(64 * 1024 * 1024);
+    }
+  };
+
+  size_t total_reads;
+  VerifyDBFromMap(true_data, &total_reads);
+  ASSERT_EQ(merge_cnt, total_reads);
+
+  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+
+  VerifyDBFromMap(true_data, &total_reads);
+}
+
+TEST_P(MergeOperatorPinningTest, TailingIterator) {
+  Options options = CurrentOptions();
+  options.merge_operator = MergeOperators::CreateMaxOperator();
+  BlockBasedTableOptions bbto;
+  bbto.no_block_cache = disable_block_cache_;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  DestroyAndReopen(options);
+
+  const int kNumOperands = 100;
+  const int kNumWrites = 100000;
+
+  std::function<void()> writer_func = [&]() {
+    int k = 0;
+    for (int i = 0; i < kNumWrites; i++) {
+      db_->Merge(WriteOptions(), Key(k), Key(k));
+
+      if (i && i % kNumOperands == 0) {
+        k++;
+      }
+      if (i && i % 127 == 0) {
+        ASSERT_OK(Flush());
+      }
+      if (i && i % 317 == 0) {
+        ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+      }
+    }
+  };
+
+  std::function<void()> reader_func = [&]() {
+    ReadOptions ro;
+    ro.tailing = true;
+    Iterator* iter = db_->NewIterator(ro);
+
+    iter->SeekToFirst();
+    for (int i = 0; i < (kNumWrites / kNumOperands); i++) {
+      while (!iter->Valid()) {
+        // wait for the key to be written
+        env_->SleepForMicroseconds(100);
+        iter->Seek(Key(i));
+      }
+      ASSERT_EQ(iter->key(), Key(i));
+      ASSERT_EQ(iter->value(), Key(i));
+
+      iter->Next();
+    }
+
+    delete iter;
+  };
+
+  std::thread writer_thread(writer_func);
+  std::thread reader_thread(reader_func);
+
+  writer_thread.join();
+  reader_thread.join();
+}
+#endif  // ROCKSDB_LITE
+
 TEST_F(DBTest2, MaxSuccessiveMergesInRecovery) {
   Options options;
   options = CurrentOptions(options);
@@ -1504,6 +1905,167 @@ TEST_F(DBTest2, MaxSuccessiveMergesInRecovery) {
   options.max_successive_merges = 3;
   Reopen(options);
 }
+
+size_t GetEncodedEntrySize(size_t key_size, size_t value_size) {
+  std::string buffer;
+
+  PutVarint32(&buffer, static_cast<uint32_t>(0));
+  PutVarint32(&buffer, static_cast<uint32_t>(key_size));
+  PutVarint32(&buffer, static_cast<uint32_t>(value_size));
+
+  return buffer.size() + key_size + value_size;
+}
+
+TEST_F(DBTest2, ReadAmpBitmap) {
+  Options options = CurrentOptions();
+  BlockBasedTableOptions bbto;
+  // Disable delta encoding to make it easier to calculate read amplification
+  bbto.use_delta_encoding = false;
+  // Huge block cache to make it easier to calculate read amplification
+  bbto.block_cache = NewLRUCache(1024 * 1024 * 1024);
+  bbto.read_amp_bytes_per_bit = 16;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  options.statistics = rocksdb::CreateDBStatistics();
+  DestroyAndReopen(options);
+
+  const size_t kNumEntries = 10000;
+
+  Random rnd(301);
+  for (size_t i = 0; i < kNumEntries; i++) {
+    ASSERT_OK(Put(Key(static_cast<int>(i)), RandomString(&rnd, 100)));
+  }
+  ASSERT_OK(Flush());
+
+  Close();
+  Reopen(options);
+
+  // Read keys/values randomly and verify that reported read amp error
+  // is less than 2%
+  uint64_t total_useful_bytes = 0;
+  std::set<int> read_keys;
+  std::string value;
+  for (size_t i = 0; i < kNumEntries * 5; i++) {
+    int key_idx = rnd.Next() % kNumEntries;
+    std::string k = Key(key_idx);
+    ASSERT_OK(db_->Get(ReadOptions(), k, &value));
+
+    if (read_keys.find(key_idx) == read_keys.end()) {
+      auto ik = InternalKey(k, 0, ValueType::kTypeValue);
+      total_useful_bytes += GetEncodedEntrySize(ik.size(), value.size());
+      read_keys.insert(key_idx);
+    }
+
+    double expected_read_amp =
+        static_cast<double>(total_useful_bytes) /
+        options.statistics->getTickerCount(READ_AMP_TOTAL_READ_BYTES);
+
+    double read_amp =
+        static_cast<double>(options.statistics->getTickerCount(
+            READ_AMP_ESTIMATE_USEFUL_BYTES)) /
+        options.statistics->getTickerCount(READ_AMP_TOTAL_READ_BYTES);
+
+    double error_pct = fabs(expected_read_amp - read_amp) * 100;
+    // Error between reported read amp and real read amp should be less than 2%
+    EXPECT_LE(error_pct, 2);
+  }
+
+  // Make sure we read every thing in the DB (which is smaller than our cache)
+  Iterator* iter = db_->NewIterator(ReadOptions());
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+    ASSERT_EQ(iter->value().ToString(), Get(iter->key().ToString()));
+  }
+  delete iter;
+
+  // Read amp is 100% since we read all what we loaded in memory
+  ASSERT_EQ(options.statistics->getTickerCount(READ_AMP_ESTIMATE_USEFUL_BYTES),
+            options.statistics->getTickerCount(READ_AMP_TOTAL_READ_BYTES));
+}
+
+TEST_F(DBTest2, ReadAmpBitmapLiveInCacheAfterDBClose) {
+  if (dbname_.find("dev/shm") != std::string::npos) {
+    // /dev/shm dont support getting a unique file id, this mean that
+    // running this test on /dev/shm will fail because lru_cache will load
+    // the blocks again regardless of them being already in the cache
+    return;
+  }
+
+  std::shared_ptr<Cache> lru_cache = NewLRUCache(1024 * 1024 * 1024);
+  std::shared_ptr<Statistics> stats = rocksdb::CreateDBStatistics();
+
+  Options options = CurrentOptions();
+  BlockBasedTableOptions bbto;
+  // Disable delta encoding to make it easier to calculate read amplification
+  bbto.use_delta_encoding = false;
+  // Huge block cache to make it easier to calculate read amplification
+  bbto.block_cache = lru_cache;
+  bbto.read_amp_bytes_per_bit = 16;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  options.statistics = stats;
+  DestroyAndReopen(options);
+
+  const int kNumEntries = 10000;
+
+  Random rnd(301);
+  for (int i = 0; i < kNumEntries; i++) {
+    ASSERT_OK(Put(Key(i), RandomString(&rnd, 100)));
+  }
+  ASSERT_OK(Flush());
+
+  Close();
+  Reopen(options);
+
+  uint64_t total_useful_bytes = 0;
+  std::set<int> read_keys;
+  std::string value;
+  // Iter1: Read half the DB, Read even keys
+  // Key(0), Key(2), Key(4), Key(6), Key(8), ...
+  for (int i = 0; i < kNumEntries; i += 2) {
+    std::string k = Key(i);
+    ASSERT_OK(db_->Get(ReadOptions(), k, &value));
+
+    if (read_keys.find(i) == read_keys.end()) {
+      auto ik = InternalKey(k, 0, ValueType::kTypeValue);
+      total_useful_bytes += GetEncodedEntrySize(ik.size(), value.size());
+      read_keys.insert(i);
+    }
+  }
+
+  size_t total_useful_bytes_iter1 =
+      options.statistics->getTickerCount(READ_AMP_ESTIMATE_USEFUL_BYTES);
+  size_t total_loaded_bytes_iter1 =
+      options.statistics->getTickerCount(READ_AMP_TOTAL_READ_BYTES);
+
+  Close();
+  std::shared_ptr<Statistics> new_statistics = rocksdb::CreateDBStatistics();
+  // Destroy old statistics obj that the blocks in lru_cache are pointing to
+  options.statistics.reset();
+  // Use the statistics object that we just created
+  options.statistics = new_statistics;
+  Reopen(options);
+
+  // Iter2: Read half the DB, Read odd keys
+  // Key(1), Key(3), Key(5), Key(7), Key(9), ...
+  for (int i = 1; i < kNumEntries; i += 2) {
+    std::string k = Key(i);
+    ASSERT_OK(db_->Get(ReadOptions(), k, &value));
+
+    if (read_keys.find(i) == read_keys.end()) {
+      auto ik = InternalKey(k, 0, ValueType::kTypeValue);
+      total_useful_bytes += GetEncodedEntrySize(ik.size(), value.size());
+      read_keys.insert(i);
+    }
+  }
+
+  size_t total_useful_bytes_iter2 =
+      options.statistics->getTickerCount(READ_AMP_ESTIMATE_USEFUL_BYTES);
+  size_t total_loaded_bytes_iter2 =
+      options.statistics->getTickerCount(READ_AMP_TOTAL_READ_BYTES);
+
+  // We reached read_amp of 100% because we read all the keys in the DB
+  ASSERT_EQ(total_useful_bytes_iter1 + total_useful_bytes_iter2,
+            total_loaded_bytes_iter1 + total_loaded_bytes_iter2);
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/db_test_util.cc

@@ -30,6 +30,8 @@ SpecialEnv::SpecialEnv(Env* base)
   manifest_write_error_.store(false, std::memory_order_release);
   log_write_error_.store(false, std::memory_order_release);
   random_file_open_counter_.store(0, std::memory_order_relaxed);
+  delete_count_.store(0, std::memory_order_relaxed);
+  num_open_wal_file_.store(0);
   log_write_slowdown_ = 0;
   bytes_written_ = 0;
   sync_counter_ = 0;
@@ -730,7 +732,7 @@ double DBTestBase::CompressionRatioAtLevel(int level, int cf) {
 
 int DBTestBase::TotalTableFiles(int cf, int levels) {
   if (levels == -1) {
-    levels = CurrentOptions().num_levels;
+    levels = (cf == 0) ? db_->NumberLevels() : db_->NumberLevels(handles_[1]);
   }
   int result = 0;
   for (int level = 0; level < levels; level++) {
@@ -1074,6 +1076,92 @@ std::vector<std::uint64_t> DBTestBase::ListTableFiles(Env* env,
   return file_numbers;
 }
 
+void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
+                                 size_t* total_reads_res) {
+  size_t total_reads = 0;
+
+  for (auto& kv : true_data) {
+    ASSERT_EQ(Get(kv.first), kv.second);
+    total_reads++;
+  }
+
+  // Normal Iterator
+  {
+    int iter_cnt = 0;
+    ReadOptions ro;
+    ro.total_order_seek = true;
+    Iterator* iter = db_->NewIterator(ro);
+    // Verify Iterator::Next()
+    iter_cnt = 0;
+    auto data_iter = true_data.begin();
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next(), data_iter++) {
+      ASSERT_EQ(iter->key().ToString(), data_iter->first);
+      ASSERT_EQ(iter->value().ToString(), data_iter->second);
+      iter_cnt++;
+      total_reads++;
+    }
+    ASSERT_EQ(data_iter, true_data.end()) << iter_cnt << " / "
+                                          << true_data.size();
+
+    // Verify Iterator::Prev()
+    iter_cnt = 0;
+    auto data_rev = true_data.rbegin();
+    for (iter->SeekToLast(); iter->Valid(); iter->Prev(), data_rev++) {
+      ASSERT_EQ(iter->key().ToString(), data_rev->first);
+      ASSERT_EQ(iter->value().ToString(), data_rev->second);
+      iter_cnt++;
+      total_reads++;
+    }
+    ASSERT_EQ(data_rev, true_data.rend()) << iter_cnt << " / "
+                                          << true_data.size();
+
+    // Verify Iterator::Seek()
+    for (auto kv : true_data) {
+      iter->Seek(kv.first);
+      ASSERT_EQ(kv.first, iter->key().ToString());
+      ASSERT_EQ(kv.second, iter->value().ToString());
+      total_reads++;
+    }
+
+    delete iter;
+  }
+
+#ifndef ROCKSDB_LITE
+  // Tailing iterator
+  int iter_cnt = 0;
+  ReadOptions ro;
+  ro.tailing = true;
+  ro.total_order_seek = true;
+  Iterator* iter = db_->NewIterator(ro);
+
+  // Verify ForwardIterator::Next()
+  iter_cnt = 0;
+  auto data_iter = true_data.begin();
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next(), data_iter++) {
+    ASSERT_EQ(iter->key().ToString(), data_iter->first);
+    ASSERT_EQ(iter->value().ToString(), data_iter->second);
+    iter_cnt++;
+    total_reads++;
+  }
+  ASSERT_EQ(data_iter, true_data.end()) << iter_cnt << " / "
+                                        << true_data.size();
+
+  // Verify ForwardIterator::Seek()
+  for (auto kv : true_data) {
+    iter->Seek(kv.first);
+    ASSERT_EQ(kv.first, iter->key().ToString());
+    ASSERT_EQ(kv.second, iter->value().ToString());
+    total_reads++;
+  }
+
+  delete iter;
+#endif  // ROCKSDB_LITE
+
+  if (total_reads_res) {
+    *total_reads_res = total_reads;
+  }
+}
+
 #ifndef ROCKSDB_LITE
 
 Status DBTestBase::GenerateAndAddExternalFile(const Options options,

db/db_test_util.h

@@ -221,7 +221,7 @@ class SpecialEnv : public EnvWrapper {
           // Drop writes on the floor
           return Status::OK();
         } else if (env_->no_space_.load(std::memory_order_acquire)) {
-          return Status::IOError("No space left on device");
+          return Status::NoSpace("No space left on device");
         } else {
           env_->bytes_written_ += data.size();
           return base_->Append(data);
@@ -285,7 +285,10 @@ class SpecialEnv : public EnvWrapper {
     class WalFile : public WritableFile {
      public:
       WalFile(SpecialEnv* env, unique_ptr<WritableFile>&& b)
-          : env_(env), base_(std::move(b)) {}
+          : env_(env), base_(std::move(b)) {
+        env_->num_open_wal_file_.fetch_add(1);
+      }
+      virtual ~WalFile() { env_->num_open_wal_file_.fetch_add(-1); }
       Status Append(const Slice& data) override {
 #if !(defined NDEBUG) || !defined(OS_WIN)
         TEST_SYNC_POINT("SpecialEnv::WalFile::Append:1");
@@ -443,6 +446,11 @@ class SpecialEnv : public EnvWrapper {
            addon_time_.load();
   }
 
+  virtual Status DeleteFile(const std::string& fname) override {
+    delete_count_.fetch_add(1);
+    return target()->DeleteFile(fname);
+  }
+
   Random rnd_;
   port::Mutex rnd_mutex_;  // Lock to pretect rnd_
 
@@ -470,6 +478,9 @@ class SpecialEnv : public EnvWrapper {
   // Slow down every log write, in micro-seconds.
   std::atomic<int> log_write_slowdown_;
 
+  // Number of WAL files that are still open for write.
+  std::atomic<int> num_open_wal_file_;
+
   bool count_random_reads_;
   anon::AtomicCounter random_read_counter_;
   std::atomic<size_t> random_read_bytes_counter_;
@@ -494,6 +505,8 @@ class SpecialEnv : public EnvWrapper {
 
   std::atomic<int64_t> addon_time_;
 
+  std::atomic<int> delete_count_;
+
   bool time_elapse_only_sleep_;
 
   bool no_sleep_;
@@ -797,6 +810,9 @@ class DBTestBase : public testing::Test {
 
   std::vector<std::uint64_t> ListTableFiles(Env* env, const std::string& path);
 
+  void VerifyDBFromMap(std::map<std::string, std::string> true_data,
+                       size_t* total_reads_res = nullptr);
+
 #ifndef ROCKSDB_LITE
   Status GenerateAndAddExternalFile(const Options options,
                                     std::vector<int> keys, size_t file_id);

db/db_universal_compaction_test.cc

@@ -637,10 +637,132 @@ TEST_P(DBTestUniversalCompactionParallel, UniversalCompactionParallel) {
   }
 }
 
+TEST_P(DBTestUniversalCompactionParallel, PickByFileNumberBug) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.num_levels = num_levels_;
+  options.write_buffer_size = 1 * 1024;  // 1KB
+  options.level0_file_num_compaction_trigger = 7;
+  options.max_background_compactions = 2;
+  options.target_file_size_base = 1024 * 1024;  // 1MB
+
+  // Disable size amplifiction compaction
+  options.compaction_options_universal.max_size_amplification_percent =
+      UINT_MAX;
+  DestroyAndReopen(options);
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBTestUniversalCompactionParallel::PickByFileNumberBug:0",
+        "BackgroundCallCompaction:0"},
+       {"UniversalCompactionPicker::PickCompaction:Return",
+        "DBTestUniversalCompactionParallel::PickByFileNumberBug:1"},
+       {"DBTestUniversalCompactionParallel::PickByFileNumberBug:2",
+        "CompactionJob::Run():Start"}});
+
+  int total_picked_compactions = 0;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "UniversalCompactionPicker::PickCompaction:Return", [&](void* arg) {
+        if (arg) {
+          total_picked_compactions++;
+        }
+      });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // Write 7 files to trigger compaction
+  int key_idx = 1;
+  for (int i = 1; i <= 70; i++) {
+    std::string k = Key(key_idx++);
+    ASSERT_OK(Put(k, k));
+    if (i % 10 == 0) {
+      ASSERT_OK(Flush());
+    }
+  }
+
+  // Wait for the 1st background compaction process to start
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:0");
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:1");
+  rocksdb::SyncPoint::GetInstance()->ClearTrace();
+
+  // Write 3 files while 1st compaction is held
+  // These 3 files have different sizes to avoid compacting based on size_ratio
+  int num_keys = 1000;
+  for (int i = 0; i < 3; i++) {
+    for (int j = 1; j <= num_keys; j++) {
+      std::string k = Key(key_idx++);
+      ASSERT_OK(Put(k, k));
+    }
+    ASSERT_OK(Flush());
+    num_keys -= 100;
+  }
+
+  // Wait for the 2nd background compaction process to start
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:0");
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:1");
+
+  // Hold the 1st and 2nd compaction from finishing
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:2");
+  dbfull()->TEST_WaitForCompact();
+
+  // Although 2 compaction threads started, the second one did not compact
+  // anything because the number of files not being compacted is less than
+  // level0_file_num_compaction_trigger
+  EXPECT_EQ(total_picked_compactions, 1);
+  EXPECT_EQ(TotalTableFiles(), 4);
+
+  // Stop SyncPoint and destroy the DB and reopen it again
+  rocksdb::SyncPoint::GetInstance()->ClearTrace();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  key_idx = 1;
+  total_picked_compactions = 0;
+  DestroyAndReopen(options);
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // Write 7 files to trigger compaction
+  for (int i = 1; i <= 70; i++) {
+    std::string k = Key(key_idx++);
+    ASSERT_OK(Put(k, k));
+    if (i % 10 == 0) {
+      ASSERT_OK(Flush());
+    }
+  }
+
+  // Wait for the 1st background compaction process to start
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:0");
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:1");
+  rocksdb::SyncPoint::GetInstance()->ClearTrace();
+
+  // Write 8 files while 1st compaction is held
+  // These 8 files have different sizes to avoid compacting based on size_ratio
+  num_keys = 1000;
+  for (int i = 0; i < 8; i++) {
+    for (int j = 1; j <= num_keys; j++) {
+      std::string k = Key(key_idx++);
+      ASSERT_OK(Put(k, k));
+    }
+    ASSERT_OK(Flush());
+    num_keys -= 100;
+  }
+
+  // Wait for the 2nd background compaction process to start
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:0");
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:1");
+
+  // Hold the 1st and 2nd compaction from finishing
+  TEST_SYNC_POINT("DBTestUniversalCompactionParallel::PickByFileNumberBug:2");
+  dbfull()->TEST_WaitForCompact();
+
+  // This time we will trigger a compaction because of size ratio and
+  // another compaction because of number of files that are not compacted
+  // greater than 7
+  EXPECT_GE(total_picked_compactions, 2);
+}
+
 INSTANTIATE_TEST_CASE_P(DBTestUniversalCompactionParallel,
                         DBTestUniversalCompactionParallel,
                         ::testing::Combine(::testing::Values(1, 10),
-                                           ::testing::Bool()));
+                                           ::testing::Values(false)));
 
 TEST_P(DBTestUniversalCompaction, UniversalCompactionOptions) {
   Options options = CurrentOptions();
@@ -996,13 +1118,13 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionFourPaths) {
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
   ASSERT_EQ(0, GetSstFileCount(dbname_));
 
-  // (1, 2, 4)
+  // (1, 2, 4) -> (3, 4)
   GenerateNewFile(&rnd, &key_idx);
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[2].path));
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
-  ASSERT_EQ(1, GetSstFileCount(dbname_));
+  ASSERT_EQ(0, GetSstFileCount(dbname_));
 
-  // (1, 1, 2, 4) -> (8)
+  // (1, 3, 4) -> (8)
   GenerateNewFile(&rnd, &key_idx);
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[3].path));
 
@@ -1016,22 +1138,22 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionFourPaths) {
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[3].path));
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
 
-  // (1, 2, 8)
+  // (1, 2, 8) -> (3, 8)
   GenerateNewFile(&rnd, &key_idx);
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[3].path));
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
-  ASSERT_EQ(1, GetSstFileCount(dbname_));
+  ASSERT_EQ(0, GetSstFileCount(dbname_));
 
-  // (1, 1, 2, 8) -> (4, 8)
+  // (1, 3, 8) -> (4, 8)
   GenerateNewFile(&rnd, &key_idx);
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[2].path));
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[3].path));
 
-  // (1, 4, 8)
+  // (1, 4, 8) -> (5, 8)
   GenerateNewFile(&rnd, &key_idx);
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[3].path));
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[2].path));
-  ASSERT_EQ(1, GetSstFileCount(dbname_));
+  ASSERT_EQ(0, GetSstFileCount(dbname_));
 
   for (int i = 0; i < key_idx; i++) {
     auto v = Get(Key(i));
@@ -1195,12 +1317,12 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionSecondPathRatio) {
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
   ASSERT_EQ(1, GetSstFileCount(dbname_));
 
-  // (1, 2, 4)
+  // (1, 2, 4) -> (3, 4)
   GenerateNewFile(&rnd, &key_idx);
-  ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
-  ASSERT_EQ(2, GetSstFileCount(dbname_));
+  ASSERT_EQ(2, GetSstFileCount(options.db_paths[1].path));
+  ASSERT_EQ(0, GetSstFileCount(dbname_));
 
-  // (1, 1, 2, 4) -> (8)
+  // (1, 3, 4) -> (8)
   GenerateNewFile(&rnd, &key_idx);
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
   ASSERT_EQ(0, GetSstFileCount(dbname_));
@@ -1215,20 +1337,20 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionSecondPathRatio) {
   ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
   ASSERT_EQ(1, GetSstFileCount(dbname_));
 
-  // (1, 2, 8)
-  GenerateNewFile(&rnd, &key_idx);
-  ASSERT_EQ(1, GetSstFileCount(options.db_paths[1].path));
-  ASSERT_EQ(2, GetSstFileCount(dbname_));
-
-  // (1, 1, 2, 8) -> (4, 8)
+  // (1, 2, 8) -> (3, 8)
   GenerateNewFile(&rnd, &key_idx);
   ASSERT_EQ(2, GetSstFileCount(options.db_paths[1].path));
   ASSERT_EQ(0, GetSstFileCount(dbname_));
 
-  // (1, 4, 8)
+  // (1, 3, 8) -> (4, 8)
   GenerateNewFile(&rnd, &key_idx);
   ASSERT_EQ(2, GetSstFileCount(options.db_paths[1].path));
-  ASSERT_EQ(1, GetSstFileCount(dbname_));
+  ASSERT_EQ(0, GetSstFileCount(dbname_));
+
+  // (1, 4, 8) -> (5, 8)
+  GenerateNewFile(&rnd, &key_idx);
+  ASSERT_EQ(2, GetSstFileCount(options.db_paths[1].path));
+  ASSERT_EQ(0, GetSstFileCount(dbname_));
 
   for (int i = 0; i < key_idx; i++) {
     auto v = Get(Key(i));

db/db_wal_test.cc

@@ -9,6 +9,7 @@
 
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
+#include "util/fault_injection_test_env.h"
 #include "util/options_helper.h"
 #include "util/sync_point.h"
 
@@ -292,6 +293,44 @@ TEST_F(DBWALTest, RecoveryWithEmptyLog) {
 }
 
 #ifndef ROCKSDB_LITE
+TEST_F(DBWALTest, GetSortedWalFiles) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    VectorLogPtr log_files;
+    ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
+    ASSERT_EQ(0, log_files.size());
+
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
+    ASSERT_EQ(1, log_files.size());
+  } while (ChangeOptions());
+}
+
+TEST_F(DBWALTest, RecoveryWithLogDataForSomeCFs) {
+  // Test for regression of WAL cleanup missing files that don't contain data
+  // for every column family.
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_OK(Put(1, "foo", "v2"));
+    uint64_t earliest_log_nums[2];
+    for (int i = 0; i < 2; ++i) {
+      if (i > 0) {
+        ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+      }
+      VectorLogPtr log_files;
+      ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
+      if (log_files.size() > 0) {
+        earliest_log_nums[i] = log_files[0]->LogNumber();
+      } else {
+        earliest_log_nums[i] = port::kMaxUint64;
+      }
+    }
+    // Check at least the first WAL was cleaned up during the recovery.
+    ASSERT_LT(earliest_log_nums[0], earliest_log_nums[1]);
+  } while (ChangeOptions());
+}
+
 TEST_F(DBWALTest, RecoverWithLargeLog) {
   do {
     {
@@ -471,6 +510,41 @@ TEST_F(DBWALTest, SyncMultipleLogs) {
   ASSERT_OK(dbfull()->SyncWAL());
 }
 
+// Github issue 1339. Prior the fix we read sequence id from the first log to
+// a local variable, then keep increase the variable as we replay logs,
+// ignoring actual sequence id of the records. This is incorrect if some writes
+// come with WAL disabled.
+TEST_F(DBWALTest, PartOfWritesWithWALDisabled) {
+  std::unique_ptr<FaultInjectionTestEnv> fault_env(
+      new FaultInjectionTestEnv(env_));
+  Options options = CurrentOptions();
+  options.env = fault_env.get();
+  options.disable_auto_compactions = true;
+  // TODO(yiwu): fix for 2PC.
+  options.allow_2pc = false;
+  WriteOptions wal_on, wal_off;
+  wal_on.sync = true;
+  wal_on.disableWAL = false;
+  wal_off.disableWAL = true;
+  CreateAndReopenWithCF({"dummy"}, options);
+  ASSERT_OK(Put(1, "dummy", "d1", wal_on));  // seq id 1
+  ASSERT_OK(Put(1, "dummy", "d2", wal_off));
+  ASSERT_OK(Put(1, "dummy", "d3", wal_off));
+  ASSERT_OK(Put(0, "key", "v4", wal_on));  // seq id 4
+  ASSERT_OK(Flush(0));
+  ASSERT_OK(Put(0, "key", "v5", wal_on));  // seq id 5
+  ASSERT_EQ("v5", Get(0, "key"));
+  // Simulate a crash.
+  fault_env->SetFilesystemActive(false);
+  Close();
+  fault_env->ResetState();
+  ReopenWithColumnFamilies({"default", "dummy"}, options);
+  // Prior to the fix, we may incorrectly recover "v5" with sequence id = 3.
+  ASSERT_EQ("v5", Get(0, "key"));
+  // Destroy DB before destruct fault_env.
+  Destroy(options);
+}
+
 //
 // Test WAL recovery for the various modes available
 //

db/dbformat.cc

@@ -22,7 +22,7 @@ namespace rocksdb {
 
 uint64_t PackSequenceAndType(uint64_t seq, ValueType t) {
   assert(seq <= kMaxSequenceNumber);
-  assert(IsValueType(t));
+  assert(IsExtendedValueType(t));
   return (seq << 8) | t;
 }
 
@@ -31,7 +31,7 @@ void UnPackSequenceAndType(uint64_t packed, uint64_t* seq, ValueType* t) {
   *t = static_cast<ValueType>(packed & 0xff);
 
   assert(*seq <= kMaxSequenceNumber);
-  assert(IsValueType(*t));
+  assert(IsExtendedValueType(*t));
 }
 
 void AppendInternalKey(std::string* result, const ParsedInternalKey& key) {

db/dbformat.h

@@ -10,6 +10,7 @@
 #pragma once
 #include <stdio.h>
 #include <string>
+#include <utility>
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/filter_policy.h"
@@ -44,6 +45,8 @@ enum ValueType : unsigned char {
   kTypeCommitXID = 0xB,                   // WAL only.
   kTypeRollbackXID = 0xC,                 // WAL only.
   kTypeNoop = 0xD,                        // WAL only.
+  kTypeColumnFamilyRangeDeletion = 0xE,   // WAL only.
+  kTypeRangeDeletion = 0xF,               // meta block
   kMaxValue = 0x7F                        // Not used for storing records.
 };
 
@@ -55,12 +58,18 @@ enum ValueType : unsigned char {
 // ValueType, not the lowest).
 static const ValueType kValueTypeForSeek = kTypeSingleDeletion;
 
-// Checks whether a type is a value type (i.e. a type used in memtables and sst
-// files).
+// Checks whether a type is an inline value type
+// (i.e. a type used in memtable skiplist and sst file datablock).
 inline bool IsValueType(ValueType t) {
   return t <= kTypeMerge || t == kTypeSingleDeletion;
 }
 
+// Checks whether a type is from user operation
+// kTypeRangeDeletion is in meta block so this API is separated from above
+inline bool IsExtendedValueType(ValueType t) {
+  return IsValueType(t) || t == kTypeRangeDeletion;
+}
+
 // We leave eight bits empty at the bottom so a type and sequence#
 // can be packed together into 64-bits.
 static const SequenceNumber kMaxSequenceNumber =
@@ -208,7 +217,7 @@ inline bool ParseInternalKey(const Slice& internal_key,
   result->type = static_cast<ValueType>(c);
   assert(result->type <= ValueType::kMaxValue);
   result->user_key = Slice(internal_key.data(), n - 8);
-  return IsValueType(result->type);
+  return IsExtendedValueType(result->type);
 }
 
 // Update the sequence number in the internal key.
@@ -488,4 +497,37 @@ extern bool ReadKeyFromWriteBatchEntry(Slice* input, Slice* key,
 extern Status ReadRecordFromWriteBatch(Slice* input, char* tag,
                                        uint32_t* column_family, Slice* key,
                                        Slice* value, Slice* blob, Slice* xid);
+
+// When user call DeleteRange() to delete a range of keys,
+// we will store a serialized RangeTombstone in MemTable and SST.
+// the struct here is a easy-understood form
+// start/end_key_ is the start/end user key of the range to be deleted
+struct RangeTombstone {
+  Slice start_key_;
+  Slice end_key_;
+  SequenceNumber seq_;
+  explicit RangeTombstone(Slice sk, Slice ek, SequenceNumber sn)
+      : start_key_(sk), end_key_(ek), seq_(sn) {}
+
+  explicit RangeTombstone(Slice internal_key, Slice value) {
+    ParsedInternalKey parsed_key;
+    if (ParseInternalKey(internal_key, &parsed_key)) {
+      start_key_ = parsed_key.user_key;
+      seq_ = parsed_key.sequence;
+      end_key_ = value;
+    }
+  }
+
+  // be careful to use Serialize(); InternalKey() allocates new memory
+  std::pair<InternalKey, Slice> Serialize() {
+    auto key = InternalKey(start_key_, seq_, kTypeRangeDeletion);
+    Slice value = end_key_;
+    return std::make_pair(std::move(key), std::move(value));
+  }
+
+  InternalKey SerializeKey() {
+    return InternalKey(start_key_, seq_, kTypeRangeDeletion);
+  }
+};
+
 }  // namespace rocksdb

db/fault_injection_test.cc

@@ -405,6 +405,7 @@ TEST_P(FaultInjectionTest, WriteOptionSyncTest) {
   env_->SetFilesystemActive(false);
   NoWriteTestReopenWithFault(kResetDropAndDeleteUnsynced);
   sleeping_task_low.WakeUp();
+  sleeping_task_low.WaitUntilDone();
 
   ASSERT_OK(OpenDB());
   std::string val;
@@ -489,6 +490,7 @@ TEST_P(FaultInjectionTest, ManualLogSyncTest) {
   env_->SetFilesystemActive(false);
   NoWriteTestReopenWithFault(kResetDropAndDeleteUnsynced);
   sleeping_task_low.WakeUp();
+  sleeping_task_low.WaitUntilDone();
 
   ASSERT_OK(OpenDB());
   std::string val;

db/flush_job.cc

@@ -84,7 +84,8 @@ FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
       output_compression_(output_compression),
       stats_(stats),
       event_logger_(event_logger),
-      measure_io_stats_(measure_io_stats) {
+      measure_io_stats_(measure_io_stats),
+      pick_memtable_called(false) {
   // Update the thread status to indicate flush.
   ReportStartedFlush();
   TEST_SYNC_POINT("FlushJob::FlushJob()");
@@ -121,9 +122,47 @@ void FlushJob::RecordFlushIOStats() {
   IOSTATS_RESET(bytes_written);
 }
 
+void FlushJob::PickMemTable() {
+  db_mutex_->AssertHeld();
+  assert(!pick_memtable_called);
+  pick_memtable_called = true;
+  // Save the contents of the earliest memtable as a new Table
+  cfd_->imm()->PickMemtablesToFlush(&mems_);
+  if (mems_.empty()) {
+    return;
+  }
+
+  ReportFlushInputSize(mems_);
+
+  // entries mems are (implicitly) sorted in ascending order by their created
+  // time. We will use the first memtable's `edit` to keep the meta info for
+  // this flush.
+  MemTable* m = mems_[0];
+  edit_ = m->GetEdits();
+  edit_->SetPrevLogNumber(0);
+  // SetLogNumber(log_num) indicates logs with number smaller than log_num
+  // will no longer be picked up for recovery.
+  edit_->SetLogNumber(mems_.back()->GetNextLogNumber());
+  edit_->SetColumnFamily(cfd_->GetID());
+
+  // path 0 for level 0 file.
+  meta_.fd = FileDescriptor(versions_->NewFileNumber(), 0, 0);
+
+  base_ = cfd_->current();
+  base_->Ref();  // it is likely that we do not need this reference
+}
+
 Status FlushJob::Run(FileMetaData* file_meta) {
+  db_mutex_->AssertHeld();
+  assert(pick_memtable_called);
   AutoThreadOperationStageUpdater stage_run(
       ThreadStatus::STAGE_FLUSH_RUN);
+  if (mems_.empty()) {
+    LogToBuffer(log_buffer_, "[%s] Nothing in memtable to flush",
+                cfd_->GetName().c_str());
+    return Status::OK();
+  }
+
   // I/O measurement variables
   PerfLevel prev_perf_level = PerfLevel::kEnableTime;
   uint64_t prev_write_nanos = 0;
@@ -139,31 +178,8 @@ Status FlushJob::Run(FileMetaData* file_meta) {
     prev_prepare_write_nanos = IOSTATS(prepare_write_nanos);
   }
 
-  // Save the contents of the earliest memtable as a new Table
-  FileMetaData meta;
-  autovector<MemTable*> mems;
-  cfd_->imm()->PickMemtablesToFlush(&mems);
-  if (mems.empty()) {
-    LogToBuffer(log_buffer_, "[%s] Nothing in memtable to flush",
-                cfd_->GetName().c_str());
-    return Status::OK();
-  }
-
-  ReportFlushInputSize(mems);
-
-  // entries mems are (implicitly) sorted in ascending order by their created
-  // time. We will use the first memtable's `edit` to keep the meta info for
-  // this flush.
-  MemTable* m = mems[0];
-  VersionEdit* edit = m->GetEdits();
-  edit->SetPrevLogNumber(0);
-  // SetLogNumber(log_num) indicates logs with number smaller than log_num
-  // will no longer be picked up for recovery.
-  edit->SetLogNumber(mems.back()->GetNextLogNumber());
-  edit->SetColumnFamily(cfd_->GetID());
-
   // This will release and re-acquire the mutex.
-  Status s = WriteLevel0Table(mems, edit, &meta);
+  Status s = WriteLevel0Table();
 
   if (s.ok() &&
       (shutting_down_->load(std::memory_order_acquire) || cfd_->IsDropped())) {
@@ -172,18 +188,18 @@ Status FlushJob::Run(FileMetaData* file_meta) {
   }
 
   if (!s.ok()) {
-    cfd_->imm()->RollbackMemtableFlush(mems, meta.fd.GetNumber());
+    cfd_->imm()->RollbackMemtableFlush(mems_, meta_.fd.GetNumber());
   } else {
     TEST_SYNC_POINT("FlushJob::InstallResults");
     // Replace immutable memtable with the generated Table
     s = cfd_->imm()->InstallMemtableFlushResults(
-        cfd_, mutable_cf_options_, mems, versions_, db_mutex_,
-        meta.fd.GetNumber(), &job_context_->memtables_to_free, db_directory_,
+        cfd_, mutable_cf_options_, mems_, versions_, db_mutex_,
+        meta_.fd.GetNumber(), &job_context_->memtables_to_free, db_directory_,
         log_buffer_);
   }
 
   if (s.ok() && file_meta != nullptr) {
-    *file_meta = meta;
+    *file_meta = meta_;
   }
   RecordFlushIOStats();
 
@@ -214,17 +230,11 @@ Status FlushJob::Run(FileMetaData* file_meta) {
   return s;
 }
 
-Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
-                                  VersionEdit* edit, FileMetaData* meta) {
+Status FlushJob::WriteLevel0Table() {
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_FLUSH_WRITE_L0);
   db_mutex_->AssertHeld();
   const uint64_t start_micros = db_options_.env->NowMicros();
-  // path 0 for level 0 file.
-  meta->fd = FileDescriptor(versions_->NewFileNumber(), 0, 0);
-
-  Version* base = cfd_->current();
-  base->Ref();  // it is likely that we do not need this reference
   Status s;
   {
     db_mutex_->Unlock();
@@ -237,7 +247,7 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
     Arena arena;
     uint64_t total_num_entries = 0, total_num_deletes = 0;
     size_t total_memory_usage = 0;
-    for (MemTable* m : mems) {
+    for (MemTable* m : mems_) {
       Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
           "[%s] [JOB %d] Flushing memtable with next log file: %" PRIu64 "\n",
           cfd_->GetName().c_str(), job_context_->job_id, m->GetNextLogNumber());
@@ -249,7 +259,7 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
 
     event_logger_->Log() << "job" << job_context_->job_id << "event"
                          << "flush_started"
-                         << "num_memtables" << mems.size() << "num_entries"
+                         << "num_memtables" << mems_.size() << "num_entries"
                          << total_num_entries << "num_deletes"
                          << total_num_deletes << "memory_usage"
                          << total_memory_usage;
@@ -260,13 +270,13 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
                              static_cast<int>(memtables.size()), &arena));
       Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
           "[%s] [JOB %d] Level-0 flush table #%" PRIu64 ": started",
-          cfd_->GetName().c_str(), job_context_->job_id, meta->fd.GetNumber());
+          cfd_->GetName().c_str(), job_context_->job_id, meta_.fd.GetNumber());
 
       TEST_SYNC_POINT_CALLBACK("FlushJob::WriteLevel0Table:output_compression",
                                &output_compression_);
       s = BuildTable(
           dbname_, db_options_.env, *cfd_->ioptions(), mutable_cf_options_,
-          env_options_, cfd_->table_cache(), iter.get(), meta,
+          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_,
@@ -280,9 +290,9 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
         "[%s] [JOB %d] Level-0 flush table #%" PRIu64 ": %" PRIu64
         " bytes %s"
         "%s",
-        cfd_->GetName().c_str(), job_context_->job_id, meta->fd.GetNumber(),
-        meta->fd.GetFileSize(), s.ToString().c_str(),
-        meta->marked_for_compaction ? " (needs compaction)" : "");
+        cfd_->GetName().c_str(), job_context_->job_id, meta_.fd.GetNumber(),
+        meta_.fd.GetFileSize(), s.ToString().c_str(),
+        meta_.marked_for_compaction ? " (needs compaction)" : "");
 
     if (!db_options_.disableDataSync && output_file_directory_ != nullptr) {
       output_file_directory_->Fsync();
@@ -290,29 +300,29 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
     TEST_SYNC_POINT("FlushJob::WriteLevel0Table");
     db_mutex_->Lock();
   }
-  base->Unref();
+  base_->Unref();
 
   // Note that if file_size is zero, the file has been deleted and
   // should not be added to the manifest.
-  if (s.ok() && meta->fd.GetFileSize() > 0) {
+  if (s.ok() && meta_.fd.GetFileSize() > 0) {
     // if we have more than 1 background thread, then we cannot
     // insert files directly into higher levels because some other
     // threads could be concurrently producing compacted files for
     // that key range.
     // Add file to L0
-    edit->AddFile(0 /* level */, meta->fd.GetNumber(), meta->fd.GetPathId(),
-                  meta->fd.GetFileSize(), meta->smallest, meta->largest,
-                  meta->smallest_seqno, meta->largest_seqno,
-                  meta->marked_for_compaction);
+    edit_->AddFile(0 /* level */, meta_.fd.GetNumber(), meta_.fd.GetPathId(),
+                   meta_.fd.GetFileSize(), meta_.smallest, meta_.largest,
+                   meta_.smallest_seqno, meta_.largest_seqno,
+                   meta_.marked_for_compaction);
   }
 
   // Note that here we treat flush as level 0 compaction in internal stats
   InternalStats::CompactionStats stats(1);
   stats.micros = db_options_.env->NowMicros() - start_micros;
-  stats.bytes_written = meta->fd.GetFileSize();
+  stats.bytes_written = meta_.fd.GetFileSize();
   cfd_->internal_stats()->AddCompactionStats(0 /* level */, stats);
   cfd_->internal_stats()->AddCFStats(InternalStats::BYTES_FLUSHED,
-                                     meta->fd.GetFileSize());
+                                     meta_.fd.GetFileSize());
   RecordFlushIOStats();
   return s;
 }

db/flush_job.h

@@ -16,12 +16,17 @@
 #include <vector>
 #include <string>
 
-#include "db/dbformat.h"
 #include "db/column_family.h"
+#include "db/dbformat.h"
+#include "db/flush_scheduler.h"
+#include "db/internal_stats.h"
+#include "db/job_context.h"
 #include "db/log_writer.h"
 #include "db/memtable_list.h"
 #include "db/snapshot_impl.h"
 #include "db/version_edit.h"
+#include "db/write_controller.h"
+#include "db/write_thread.h"
 #include "port/port.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
@@ -33,11 +38,6 @@
 #include "util/instrumented_mutex.h"
 #include "util/stop_watch.h"
 #include "util/thread_local.h"
-#include "db/internal_stats.h"
-#include "db/write_controller.h"
-#include "db/flush_scheduler.h"
-#include "db/write_thread.h"
-#include "db/job_context.h"
 
 namespace rocksdb {
 
@@ -66,6 +66,8 @@ class FlushJob {
 
   ~FlushJob();
 
+  // Require db_mutex held
+  void PickMemTable();
   Status Run(FileMetaData* file_meta = nullptr);
   TableProperties GetTableProperties() const { return table_properties_; }
 
@@ -73,8 +75,7 @@ class FlushJob {
   void ReportStartedFlush();
   void ReportFlushInputSize(const autovector<MemTable*>& mems);
   void RecordFlushIOStats();
-  Status WriteLevel0Table(const autovector<MemTable*>& mems, VersionEdit* edit,
-                          FileMetaData* meta);
+  Status WriteLevel0Table();
   const std::string& dbname_;
   ColumnFamilyData* cfd_;
   const DBOptions& db_options_;
@@ -94,6 +95,13 @@ class FlushJob {
   EventLogger* event_logger_;
   TableProperties table_properties_;
   bool measure_io_stats_;
+
+  // Variables below are set by PickMemTable():
+  FileMetaData meta_;
+  autovector<MemTable*> mems_;
+  VersionEdit* edit_;
+  Version* base_;
+  bool pick_memtable_called;
 };
 
 }  // namespace rocksdb

db/flush_job_test.cc

@@ -94,7 +94,11 @@ TEST_F(FlushJobTest, Empty) {
                      env_options_, versions_.get(), &mutex_, &shutting_down_,
                      {}, kMaxSequenceNumber, &job_context, nullptr, nullptr,
                      nullptr, kNoCompression, nullptr, &event_logger, false);
-  ASSERT_OK(flush_job.Run());
+  {
+    InstrumentedMutexLock l(&mutex_);
+    flush_job.PickMemTable();
+    ASSERT_OK(flush_job.Run());
+  }
   job_context.Clean();
 }
 
@@ -135,6 +139,7 @@ TEST_F(FlushJobTest, NonEmpty) {
                      nullptr, kNoCompression, nullptr, &event_logger, true);
   FileMetaData fd;
   mutex_.Lock();
+  flush_job.PickMemTable();
   ASSERT_OK(flush_job.Run(&fd));
   mutex_.Unlock();
   ASSERT_EQ(ToString(0), fd.smallest.user_key().ToString());
@@ -198,6 +203,7 @@ TEST_F(FlushJobTest, Snapshots) {
       &shutting_down_, snapshots, kMaxSequenceNumber, &job_context, nullptr,
       nullptr, nullptr, kNoCompression, nullptr, &event_logger, true);
   mutex_.Lock();
+  flush_job.PickMemTable();
   ASSERT_OK(flush_job.Run());
   mutex_.Unlock();
   mock_table_factory_->AssertSingleFile(inserted_keys);

db/forward_iterator.cc

@@ -32,10 +32,24 @@ namespace rocksdb {
 class LevelIterator : public InternalIterator {
  public:
   LevelIterator(const ColumnFamilyData* const cfd,
-      const ReadOptions& read_options,
-      const std::vector<FileMetaData*>& files)
-    : cfd_(cfd), read_options_(read_options), files_(files), valid_(false),
-      file_index_(std::numeric_limits<uint32_t>::max()) {}
+                const ReadOptions& read_options,
+                const std::vector<FileMetaData*>& files)
+      : cfd_(cfd),
+        read_options_(read_options),
+        files_(files),
+        valid_(false),
+        file_index_(std::numeric_limits<uint32_t>::max()),
+        file_iter_(nullptr),
+        pinned_iters_mgr_(nullptr) {}
+
+  ~LevelIterator() {
+    // Reset current pointer
+    if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
+      pinned_iters_mgr_->PinIterator(file_iter_);
+    } else {
+      delete file_iter_;
+    }
+  }
 
   void SetFileIndex(uint32_t file_index) {
     assert(file_index < files_.size());
@@ -47,10 +61,20 @@ class LevelIterator : public InternalIterator {
   }
   void Reset() {
     assert(file_index_ < files_.size());
-    file_iter_.reset(cfd_->table_cache()->NewIterator(
+
+    // Reset current pointer
+    if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
+      pinned_iters_mgr_->PinIterator(file_iter_);
+    } else {
+      delete file_iter_;
+    }
+
+    file_iter_ = cfd_->table_cache()->NewIterator(
         read_options_, *(cfd_->soptions()), cfd_->internal_comparator(),
         files_[file_index_]->fd, nullptr /* table_reader_ptr */, nullptr,
-        false));
+        false);
+
+    file_iter_->SetPinnedItersMgr(pinned_iters_mgr_);
   }
   void SeekToLast() override {
     status_ = Status::NotSupported("LevelIterator::SeekToLast()");
@@ -105,6 +129,20 @@ class LevelIterator : public InternalIterator {
     }
     return Status::OK();
   }
+  bool IsKeyPinned() const override {
+    return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
+           file_iter_->IsKeyPinned();
+  }
+  bool IsValuePinned() const override {
+    return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
+           file_iter_->IsValuePinned();
+  }
+  void SetPinnedItersMgr(PinnedIteratorsManager* pinned_iters_mgr) override {
+    pinned_iters_mgr_ = pinned_iters_mgr;
+    if (file_iter_) {
+      file_iter_->SetPinnedItersMgr(pinned_iters_mgr_);
+    }
+  }
 
  private:
   const ColumnFamilyData* const cfd_;
@@ -114,7 +152,8 @@ class LevelIterator : public InternalIterator {
   bool valid_;
   uint32_t file_index_;
   Status status_;
-  std::unique_ptr<InternalIterator> file_iter_;
+  InternalIterator* file_iter_;
+  PinnedIteratorsManager* pinned_iters_mgr_;
 };
 
 ForwardIterator::ForwardIterator(DBImpl* db, const ReadOptions& read_options,
@@ -135,7 +174,8 @@ ForwardIterator::ForwardIterator(DBImpl* db, const ReadOptions& read_options,
       has_iter_trimmed_for_upper_bound_(false),
       current_over_upper_bound_(false),
       is_prev_set_(false),
-      is_prev_inclusive_(false) {
+      is_prev_inclusive_(false),
+      pinned_iters_mgr_(nullptr) {
   if (sv_) {
     RebuildIterators(false);
   }
@@ -145,6 +185,13 @@ ForwardIterator::~ForwardIterator() {
   Cleanup(true);
 }
 
+namespace {
+// Used in PinnedIteratorsManager to release pinned SuperVersion
+static void ReleaseSuperVersionFunc(void* sv) {
+  delete reinterpret_cast<SuperVersion*>(sv);
+}
+}  // namespace
+
 void ForwardIterator::SVCleanup() {
   if (sv_ != nullptr && sv_->Unref()) {
     // Job id == 0 means that this is not our background process, but rather
@@ -153,10 +200,18 @@ void ForwardIterator::SVCleanup() {
     db_->mutex_.Lock();
     sv_->Cleanup();
     db_->FindObsoleteFiles(&job_context, false, true);
+    if (read_options_.background_purge_on_iterator_cleanup) {
+      db_->ScheduleBgLogWriterClose(&job_context);
+    }
     db_->mutex_.Unlock();
-    delete sv_;
+    if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
+      pinned_iters_mgr_->PinPtr(sv_, &ReleaseSuperVersionFunc);
+    } else {
+      delete sv_;
+    }
     if (job_context.HaveSomethingToDelete()) {
-      db_->PurgeObsoleteFiles(job_context);
+      db_->PurgeObsoleteFiles(
+          job_context, read_options_.background_purge_on_iterator_cleanup);
     }
     job_context.Clean();
   }
@@ -164,18 +219,21 @@ void ForwardIterator::SVCleanup() {
 
 void ForwardIterator::Cleanup(bool release_sv) {
   if (mutable_iter_ != nullptr) {
-    mutable_iter_->~InternalIterator();
+    DeleteIterator(mutable_iter_, true /* is_arena */);
   }
+
   for (auto* m : imm_iters_) {
-    m->~InternalIterator();
+    DeleteIterator(m, true /* is_arena */);
   }
   imm_iters_.clear();
+
   for (auto* f : l0_iters_) {
-    delete f;
+    DeleteIterator(f);
   }
   l0_iters_.clear();
+
   for (auto* l : level_iters_) {
-    delete l;
+    DeleteIterator(l);
   }
   level_iters_.clear();
 
@@ -276,7 +334,7 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,
               l0[i]->largest.user_key()) > 0) {
           if (read_options_.iterate_upper_bound != nullptr) {
             has_iter_trimmed_for_upper_bound_ = true;
-            delete l0_iters_[i];
+            DeleteIterator(l0_iters_[i]);
             l0_iters_[i] = nullptr;
           }
           continue;
@@ -291,7 +349,7 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,
           immutable_min_heap_.push(l0_iters_[i]);
         } else {
           has_iter_trimmed_for_upper_bound_ = true;
-          delete l0_iters_[i];
+          DeleteIterator(l0_iters_[i]);
           l0_iters_[i] = nullptr;
         }
       }
@@ -368,7 +426,7 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,
           } else {
             // Nothing in this level is interesting. Remove.
             has_iter_trimmed_for_upper_bound_ = true;
-            delete level_iters_[level - 1];
+            DeleteIterator(level_iters_[level - 1]);
             level_iters_[level - 1] = nullptr;
           }
         }
@@ -481,6 +539,50 @@ Status ForwardIterator::GetProperty(std::string prop_name, std::string* prop) {
   return Status::InvalidArgument();
 }
 
+void ForwardIterator::SetPinnedItersMgr(
+    PinnedIteratorsManager* pinned_iters_mgr) {
+  pinned_iters_mgr_ = pinned_iters_mgr;
+  UpdateChildrenPinnedItersMgr();
+}
+
+void ForwardIterator::UpdateChildrenPinnedItersMgr() {
+  // Set PinnedIteratorsManager for mutable memtable iterator.
+  if (mutable_iter_) {
+    mutable_iter_->SetPinnedItersMgr(pinned_iters_mgr_);
+  }
+
+  // Set PinnedIteratorsManager for immutable memtable iterators.
+  for (InternalIterator* child_iter : imm_iters_) {
+    if (child_iter) {
+      child_iter->SetPinnedItersMgr(pinned_iters_mgr_);
+    }
+  }
+
+  // Set PinnedIteratorsManager for L0 files iterators.
+  for (InternalIterator* child_iter : l0_iters_) {
+    if (child_iter) {
+      child_iter->SetPinnedItersMgr(pinned_iters_mgr_);
+    }
+  }
+
+  // Set PinnedIteratorsManager for L1+ levels iterators.
+  for (LevelIterator* child_iter : level_iters_) {
+    if (child_iter) {
+      child_iter->SetPinnedItersMgr(pinned_iters_mgr_);
+    }
+  }
+}
+
+bool ForwardIterator::IsKeyPinned() const {
+  return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
+         current_->IsKeyPinned();
+}
+
+bool ForwardIterator::IsValuePinned() const {
+  return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
+         current_->IsValuePinned();
+}
+
 void ForwardIterator::RebuildIterators(bool refresh_sv) {
   // Clean up
   Cleanup(refresh_sv);
@@ -509,6 +611,8 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
   BuildLevelIterators(vstorage);
   current_ = nullptr;
   is_prev_set_ = false;
+
+  UpdateChildrenPinnedItersMgr();
 }
 
 void ForwardIterator::RenewIterators() {
@@ -517,10 +621,10 @@ void ForwardIterator::RenewIterators() {
   svnew = cfd_->GetReferencedSuperVersion(&(db_->mutex_));
 
   if (mutable_iter_ != nullptr) {
-    mutable_iter_->~InternalIterator();
+    DeleteIterator(mutable_iter_, true /* is_arena */);
   }
   for (auto* m : imm_iters_) {
-    m->~InternalIterator();
+    DeleteIterator(m, true /* is_arena */);
   }
   imm_iters_.clear();
 
@@ -561,13 +665,13 @@ void ForwardIterator::RenewIterators() {
   }
 
   for (auto* f : l0_iters_) {
-    delete f;
+    DeleteIterator(f);
   }
   l0_iters_.clear();
   l0_iters_ = l0_iters_new;
 
   for (auto* l : level_iters_) {
-    delete l;
+    DeleteIterator(l);
   }
   level_iters_.clear();
   BuildLevelIterators(vstorage_new);
@@ -575,6 +679,8 @@ void ForwardIterator::RenewIterators() {
   is_prev_set_ = false;
   SVCleanup();
   sv_ = svnew;
+
+  UpdateChildrenPinnedItersMgr();
 }
 
 void ForwardIterator::BuildLevelIterators(const VersionStorageInfo* vstorage) {
@@ -604,10 +710,11 @@ void ForwardIterator::ResetIncompleteIterators() {
     if (!l0_iters_[i] || !l0_iters_[i]->status().IsIncomplete()) {
       continue;
     }
-    delete l0_iters_[i];
+    DeleteIterator(l0_iters_[i]);
     l0_iters_[i] = cfd_->table_cache()->NewIterator(
         read_options_, *cfd_->soptions(), cfd_->internal_comparator(),
         l0_files[i]->fd);
+    l0_iters_[i]->SetPinnedItersMgr(pinned_iters_mgr_);
   }
 
   for (auto* level_iter : level_iters_) {
@@ -696,7 +803,7 @@ void ForwardIterator::DeleteCurrentIter() {
     }
     if (l0_iters_[i] == current_) {
       has_iter_trimmed_for_upper_bound_ = true;
-      delete l0_iters_[i];
+      DeleteIterator(l0_iters_[i]);
       l0_iters_[i] = nullptr;
       return;
     }
@@ -708,7 +815,7 @@ void ForwardIterator::DeleteCurrentIter() {
     }
     if (level_iters_[level - 1] == current_) {
       has_iter_trimmed_for_upper_bound_ = true;
-      delete level_iters_[level - 1];
+      DeleteIterator(level_iters_[level - 1]);
       level_iters_[level - 1] = nullptr;
     }
   }
@@ -772,6 +879,22 @@ uint32_t ForwardIterator::FindFileInRange(
   return right;
 }
 
+void ForwardIterator::DeleteIterator(InternalIterator* iter, bool is_arena) {
+  if (iter == nullptr) {
+    return;
+  }
+
+  if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
+    pinned_iters_mgr_->PinIterator(iter, is_arena);
+  } else {
+    if (is_arena) {
+      iter->~InternalIterator();
+    } else {
+      delete iter;
+    }
+  }
+}
+
 }  // namespace rocksdb
 
 #endif  // ROCKSDB_LITE

db/forward_iterator.h

@@ -72,6 +72,10 @@ class ForwardIterator : public InternalIterator {
   virtual Slice value() const override;
   virtual Status status() const override;
   virtual Status GetProperty(std::string prop_name, std::string* prop) override;
+  virtual void SetPinnedItersMgr(
+      PinnedIteratorsManager* pinned_iters_mgr) override;
+  virtual bool IsKeyPinned() const override;
+  virtual bool IsValuePinned() const override;
 
   bool TEST_CheckDeletedIters(int* deleted_iters, int* num_iters);
 
@@ -92,6 +96,14 @@ class ForwardIterator : public InternalIterator {
 
   bool IsOverUpperBound(const Slice& internal_key) const;
 
+  // Set PinnedIteratorsManager for all children Iterators, this function should
+  // be called whenever we update children Iterators or pinned_iters_mgr_.
+  void UpdateChildrenPinnedItersMgr();
+
+  // A helper function that will release iter in the proper manner, or pass it
+  // to pinned_iters_mgr_ to release it later if pinning is enabled.
+  void DeleteIterator(InternalIterator* iter, bool is_arena = false);
+
   DBImpl* const db_;
   const ReadOptions read_options_;
   ColumnFamilyData* const cfd_;
@@ -129,6 +141,7 @@ class ForwardIterator : public InternalIterator {
   bool is_prev_set_;
   bool is_prev_inclusive_;
 
+  PinnedIteratorsManager* pinned_iters_mgr_;
   Arena arena_;
 };
 

db/internal_stats.cc

@@ -615,15 +615,15 @@ void InternalStats::DumpDBStats(std::string* value) {
   // Data
   // writes: total number of write requests.
   // keys: total number of key updates issued by all the write requests
-  // batches: number of group commits issued to the DB. Each group can contain
-  //          one or more writes.
+  // commit groups: number of group commits issued to the DB. Each group can
+  //                contain one or more writes.
   // so writes/keys is the average number of put in multi-put or put
-  // writes/batches is the average group commit size.
+  // writes/groups is the average group commit size.
   //
   // The format is the same for interval stats.
   snprintf(buf, sizeof(buf),
-           "Cumulative writes: %s writes, %s keys, %s batches, "
-           "%.1f writes per batch, ingest: %.2f GB, %.2f MB/s\n",
+           "Cumulative writes: %s writes, %s keys, %s commit groups, "
+           "%.1f writes per commit group, ingest: %.2f GB, %.2f MB/s\n",
            NumberToHumanString(write_other + write_self).c_str(),
            NumberToHumanString(num_keys_written).c_str(),
            NumberToHumanString(write_self).c_str(),
@@ -654,8 +654,8 @@ void InternalStats::DumpDBStats(std::string* value) {
   uint64_t interval_num_keys_written =
       num_keys_written - db_stats_snapshot_.num_keys_written;
   snprintf(buf, sizeof(buf),
-           "Interval writes: %s writes, %s keys, %s batches, "
-           "%.1f writes per batch, ingest: %.2f MB, %.2f MB/s\n",
+           "Interval writes: %s writes, %s keys, %s commit groups, "
+           "%.1f writes per commit group, ingest: %.2f MB, %.2f MB/s\n",
            NumberToHumanString(
                interval_write_other + interval_write_self).c_str(),
            NumberToHumanString(interval_num_keys_written).c_str(),

db/job_context.h

@@ -12,7 +12,6 @@
 #include <string>
 #include <vector>
 
-#include "db/column_family.h"
 #include "db/log_writer.h"
 
 namespace rocksdb {

db/log_writer.h

@@ -81,6 +81,8 @@ class Writer {
   WritableFileWriter* file() { return dest_.get(); }
   const WritableFileWriter* file() const { return dest_.get(); }
 
+  uint64_t get_log_number() const { return log_number_; }
+
  private:
   unique_ptr<WritableFileWriter> dest_;
   size_t block_offset_;       // Current offset in block

db/memtable.cc

@@ -44,8 +44,7 @@ MemTableOptions::MemTableOptions(const ImmutableCFOptions& ioptions,
               static_cast<double>(mutable_cf_options.write_buffer_size) *
               mutable_cf_options.memtable_prefix_bloom_size_ratio) *
           8u),
-      memtable_prefix_bloom_huge_page_tlb_size(
-          mutable_cf_options.memtable_prefix_bloom_huge_page_tlb_size),
+      memtable_huge_page_size(mutable_cf_options.memtable_huge_page_size),
       inplace_update_support(ioptions.inplace_update_support),
       inplace_update_num_locks(mutable_cf_options.inplace_update_num_locks),
       inplace_callback(ioptions.inplace_callback),
@@ -63,7 +62,8 @@ MemTable::MemTable(const InternalKeyComparator& cmp,
       moptions_(ioptions, mutable_cf_options),
       refs_(0),
       kArenaBlockSize(OptimizeBlockSize(moptions_.arena_block_size)),
-      arena_(moptions_.arena_block_size, 0),
+      arena_(moptions_.arena_block_size,
+             mutable_cf_options.memtable_huge_page_size),
       allocator_(&arena_, write_buffer_manager),
       table_(ioptions.memtable_factory->CreateMemTableRep(
           comparator_, &allocator_, ioptions.prefix_extractor,
@@ -92,7 +92,7 @@ MemTable::MemTable(const InternalKeyComparator& cmp,
     prefix_bloom_.reset(new DynamicBloom(
         &allocator_, moptions_.memtable_prefix_bloom_bits,
         ioptions.bloom_locality, 6 /* hard coded 6 probes */, nullptr,
-        moptions_.memtable_prefix_bloom_huge_page_tlb_size, ioptions.info_log));
+        moptions_.memtable_huge_page_size, ioptions.info_log));
   }
 }
 
@@ -218,12 +218,13 @@ const char* EncodeKey(std::string* scratch, const Slice& target) {
 
 class MemTableIterator : public InternalIterator {
  public:
-  MemTableIterator(
-      const MemTable& mem, const ReadOptions& read_options, Arena* arena)
+  MemTableIterator(const MemTable& mem, const ReadOptions& read_options,
+                   Arena* arena)
       : bloom_(nullptr),
         prefix_extractor_(mem.prefix_extractor_),
         valid_(false),
-        arena_mode_(arena != nullptr) {
+        arena_mode_(arena != nullptr),
+        value_pinned_(!mem.GetMemTableOptions()->inplace_update_support) {
     if (prefix_extractor_ != nullptr && !read_options.total_order_seek) {
       bloom_ = mem.prefix_bloom_.get();
       iter_ = mem.table_->GetDynamicPrefixIterator(arena);
@@ -306,12 +307,18 @@ class MemTableIterator : public InternalIterator {
     return true;
   }
 
+  virtual bool IsValuePinned() const override {
+    // memtable value is always pinned, except if we allow inplace update.
+    return value_pinned_;
+  }
+
  private:
   DynamicBloom* bloom_;
   const SliceTransform* const prefix_extractor_;
   MemTableRep::Iterator* iter_;
   bool valid_;
   bool arena_mode_;
+  bool value_pinned_;
 
   // No copying allowed
   MemTableIterator(const MemTableIterator&);
@@ -508,7 +515,6 @@ static bool SaveValue(void* arg, const char* entry) {
       case kTypeDeletion:
       case kTypeSingleDeletion: {
         if (*(s->merge_in_progress)) {
-          *(s->status) = Status::OK();
           *(s->status) = MergeHelper::TimedFullMerge(
               merge_operator, s->key->user_key(), nullptr,
               merge_context->GetOperands(), s->value, s->logger, s->statistics,
@@ -532,7 +538,8 @@ static bool SaveValue(void* arg, const char* entry) {
         }
         Slice v = GetLengthPrefixedSlice(key_ptr + key_length);
         *(s->merge_in_progress) = true;
-        merge_context->PushOperand(v);
+        merge_context->PushOperand(
+            v, s->inplace_update_support == false /* operand_pinned */);
         return true;
       }
       default:

db/memtable.h

@@ -15,17 +15,16 @@
 #include <string>
 #include <vector>
 #include "db/dbformat.h"
+#include "db/memtable_allocator.h"
 #include "db/skiplist.h"
 #include "db/version_edit.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
-#include "rocksdb/immutable_options.h"
-#include "db/memtable_allocator.h"
+#include "util/cf_options.h"
 #include "util/concurrent_arena.h"
 #include "util/dynamic_bloom.h"
 #include "util/instrumented_mutex.h"
-#include "util/mutable_cf_options.h"
 
 namespace rocksdb {
 
@@ -41,7 +40,7 @@ struct MemTableOptions {
   size_t write_buffer_size;
   size_t arena_block_size;
   uint32_t memtable_prefix_bloom_bits;
-  size_t memtable_prefix_bloom_huge_page_tlb_size;
+  size_t memtable_huge_page_size;
   bool inplace_update_support;
   size_t inplace_update_num_locks;
   UpdateStatus (*inplace_callback)(char* existing_value,
@@ -232,8 +231,10 @@ class MemTable {
     num_entries_.fetch_add(update_counters.num_entries,
                            std::memory_order_relaxed);
     data_size_.fetch_add(update_counters.data_size, std::memory_order_relaxed);
-    num_deletes_.fetch_add(update_counters.num_deletes,
-                           std::memory_order_relaxed);
+    if (update_counters.num_deletes != 0) {
+      num_deletes_.fetch_add(update_counters.num_deletes,
+                             std::memory_order_relaxed);
+    }
     UpdateFlushState();
   }
 

db/merge_context.h

@@ -4,70 +4,113 @@
 //  of patent rights can be found in the PATENTS file in the same directory.
 //
 #pragma once
+#include <string>
+#include <vector>
 #include "db/dbformat.h"
 #include "rocksdb/slice.h"
-#include <string>
-#include <deque>
 
 namespace rocksdb {
 
-const std::deque<std::string> empty_operand_list;
+const std::vector<Slice> empty_operand_list;
 
 // The merge context for merging a user key.
 // When doing a Get(), DB will create such a class and pass it when
 // issuing Get() operation to memtables and version_set. The operands
 // will be fetched from the context when issuing partial of full merge.
 class MergeContext {
-public:
+ public:
   // Clear all the operands
   void Clear() {
-    if (operand_list) {
-      operand_list->clear();
+    if (operand_list_) {
+      operand_list_->clear();
+      copied_operands_->clear();
     }
   }
-  // Replace all operands with merge_result, which are expected to be the
-  // merge result of them.
-  void PushPartialMergeResult(std::string& merge_result) {
-    assert (operand_list);
-    operand_list->clear();
-    operand_list->push_front(std::move(merge_result));
-  }
+
   // Push a merge operand
-  void PushOperand(const Slice& operand_slice) {
+  void PushOperand(const Slice& operand_slice, bool operand_pinned = false) {
     Initialize();
-    operand_list->push_front(operand_slice.ToString());
+    SetDirectionBackward();
+
+    if (operand_pinned) {
+      operand_list_->push_back(operand_slice);
+    } else {
+      // We need to have our own copy of the operand since it's not pinned
+      copied_operands_->emplace_back(
+          new std::string(operand_slice.data(), operand_slice.size()));
+      operand_list_->push_back(*copied_operands_->back());
+    }
   }
+
   // Push back a merge operand
-  void PushOperandBack(const Slice& operand_slice) {
+  void PushOperandBack(const Slice& operand_slice,
+                       bool operand_pinned = false) {
     Initialize();
-    operand_list->push_back(operand_slice.ToString());
+    SetDirectionForward();
+
+    if (operand_pinned) {
+      operand_list_->push_back(operand_slice);
+    } else {
+      // We need to have our own copy of the operand since it's not pinned
+      copied_operands_->emplace_back(
+          new std::string(operand_slice.data(), operand_slice.size()));
+      operand_list_->push_back(*copied_operands_->back());
+    }
   }
+
   // return total number of operands in the list
   size_t GetNumOperands() const {
-    if (!operand_list) {
+    if (!operand_list_) {
       return 0;
     }
-    return operand_list->size();
+    return operand_list_->size();
   }
+
   // Get the operand at the index.
-  Slice GetOperand(int index) const {
-    assert (operand_list);
-    return (*operand_list)[index];
+  Slice GetOperand(int index) {
+    assert(operand_list_);
+
+    SetDirectionForward();
+    return (*operand_list_)[index];
   }
+
   // Return all the operands.
-  const std::deque<std::string>& GetOperands() const {
-    if (!operand_list) {
+  const std::vector<Slice>& GetOperands() {
+    if (!operand_list_) {
       return empty_operand_list;
     }
-    return *operand_list;
+
+    SetDirectionForward();
+    return *operand_list_;
   }
-private:
+
+ private:
   void Initialize() {
-    if (!operand_list) {
-      operand_list.reset(new std::deque<std::string>());
+    if (!operand_list_) {
+      operand_list_.reset(new std::vector<Slice>());
+      copied_operands_.reset(new std::vector<std::unique_ptr<std::string>>());
     }
   }
-  std::unique_ptr<std::deque<std::string>> operand_list;
+
+  void SetDirectionForward() {
+    if (operands_reversed_ == true) {
+      std::reverse(operand_list_->begin(), operand_list_->end());
+      operands_reversed_ = false;
+    }
+  }
+
+  void SetDirectionBackward() {
+    if (operands_reversed_ == false) {
+      std::reverse(operand_list_->begin(), operand_list_->end());
+      operands_reversed_ = true;
+    }
+  }
+
+  // List of operands
+  std::unique_ptr<std::vector<Slice>> operand_list_;
+  // Copy of operands that are not pinned.
+  std::unique_ptr<std::vector<std::unique_ptr<std::string>>> copied_operands_;
+  bool operands_reversed_ = true;
 };
 
 } // namespace rocksdb

db/merge_helper.cc

@@ -20,9 +20,10 @@ namespace rocksdb {
 
 Status MergeHelper::TimedFullMerge(const MergeOperator* merge_operator,
                                    const Slice& key, const Slice* value,
-                                   const std::deque<std::string>& operands,
+                                   const std::vector<Slice>& operands,
                                    std::string* result, Logger* logger,
-                                   Statistics* statistics, Env* env) {
+                                   Statistics* statistics, Env* env,
+                                   Slice* result_operand) {
   assert(merge_operator != nullptr);
 
   if (operands.size() == 0) {
@@ -32,13 +33,28 @@ Status MergeHelper::TimedFullMerge(const MergeOperator* merge_operator,
   }
 
   bool success;
+  Slice tmp_result_operand(nullptr, 0);
+  const MergeOperator::MergeOperationInput merge_in(key, value, operands,
+                                                    logger);
+  MergeOperator::MergeOperationOutput merge_out(*result, tmp_result_operand);
   {
     // Setup to time the merge
     StopWatchNano timer(env, statistics != nullptr);
     PERF_TIMER_GUARD(merge_operator_time_nanos);
 
     // Do the merge
-    success = merge_operator->FullMerge(key, value, operands, result, logger);
+    success = merge_operator->FullMergeV2(merge_in, &merge_out);
+
+    if (tmp_result_operand.data()) {
+      // FullMergeV2 result is an existing operand
+      if (result_operand != nullptr) {
+        *result_operand = tmp_result_operand;
+      } else {
+        result->assign(tmp_result_operand.data(), tmp_result_operand.size());
+      }
+    } else if (result_operand) {
+      *result_operand = Slice(nullptr, 0);
+    }
 
     RecordTick(statistics, MERGE_OPERATION_TOTAL_TIME,
                statistics ? timer.ElapsedNanos() : 0);
@@ -65,7 +81,7 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
   // Also maintain the list of merge operands seen.
   assert(HasOperator());
   keys_.clear();
-  operands_.clear();
+  merge_context_.Clear();
   assert(user_merge_operator_);
   bool first_key = true;
 
@@ -87,7 +103,7 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
   bool hit_the_next_user_key = false;
   for (; iter->Valid(); iter->Next(), original_key_is_iter = false) {
     ParsedInternalKey ikey;
-    assert(keys_.size() == operands_.size());
+    assert(keys_.size() == merge_context_.GetNumOperands());
 
     if (!ParseInternalKey(iter->key(), &ikey)) {
       // stop at corrupted key
@@ -142,7 +158,8 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
       const Slice* val_ptr = (kTypeValue == ikey.type) ? &val : nullptr;
       std::string merge_result;
       s = TimedFullMerge(user_merge_operator_, ikey.user_key, val_ptr,
-                         operands_, &merge_result, logger_, stats_, env_);
+                         merge_context_.GetOperands(), &merge_result, logger_,
+                         stats_, env_);
 
       // We store the result in keys_.back() and operands_.back()
       // if nothing went wrong (i.e.: no operand corruption on disk)
@@ -152,9 +169,9 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
         orig_ikey.type = kTypeValue;
         UpdateInternalKey(&original_key, orig_ikey.sequence, orig_ikey.type);
         keys_.clear();
-        operands_.clear();
+        merge_context_.Clear();
         keys_.emplace_front(std::move(original_key));
-        operands_.emplace_front(std::move(merge_result));
+        merge_context_.PushOperand(merge_result);
       }
 
       // move iter to the next entry
@@ -188,12 +205,13 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
           // original_key before
           ParseInternalKey(keys_.back(), &orig_ikey);
         }
-        operands_.push_front(value_slice.ToString());
+        merge_context_.PushOperand(value_slice,
+                                   iter->IsValuePinned() /* operand_pinned */);
       }
     }
   }
 
-  if (operands_.size() == 0) {
+  if (merge_context_.GetNumOperands() == 0) {
     // we filtered out all the merge operands
     return Status::OK();
   }
@@ -218,11 +236,12 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
     // do a final merge with nullptr as the existing value and say
     // bye to the merge type (it's now converted to a Put)
     assert(kTypeMerge == orig_ikey.type);
-    assert(operands_.size() >= 1);
-    assert(operands_.size() == keys_.size());
+    assert(merge_context_.GetNumOperands() >= 1);
+    assert(merge_context_.GetNumOperands() == keys_.size());
     std::string merge_result;
     s = TimedFullMerge(user_merge_operator_, orig_ikey.user_key, nullptr,
-                       operands_, &merge_result, logger_, stats_, env_);
+                       merge_context_.GetOperands(), &merge_result, logger_,
+                       stats_, env_);
     if (s.ok()) {
       // The original key encountered
       // We are certain that keys_ is not empty here (see assertions couple of
@@ -231,9 +250,9 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
       orig_ikey.type = kTypeValue;
       UpdateInternalKey(&original_key, orig_ikey.sequence, orig_ikey.type);
       keys_.clear();
-      operands_.clear();
+      merge_context_.Clear();
       keys_.emplace_front(std::move(original_key));
-      operands_.emplace_front(std::move(merge_result));
+      merge_context_.PushOperand(merge_result);
     }
   } else {
     // We haven't seen the beginning of the key nor a Put/Delete.
@@ -244,8 +263,8 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
     // partial merge returns Status::OK(). Should we change the status code
     // after a successful partial merge?
     s = Status::MergeInProgress();
-    if (operands_.size() >= 2 &&
-        operands_.size() >= min_partial_merge_operands_) {
+    if (merge_context_.GetNumOperands() >= 2 &&
+        merge_context_.GetNumOperands() >= min_partial_merge_operands_) {
       bool merge_success = false;
       std::string merge_result;
       {
@@ -253,7 +272,8 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
         PERF_TIMER_GUARD(merge_operator_time_nanos);
         merge_success = user_merge_operator_->PartialMergeMulti(
             orig_ikey.user_key,
-            std::deque<Slice>(operands_.begin(), operands_.end()),
+            std::deque<Slice>(merge_context_.GetOperands().begin(),
+                              merge_context_.GetOperands().end()),
             &merge_result, logger_);
         RecordTick(stats_, MERGE_OPERATION_TOTAL_TIME,
                    stats_ ? timer.ElapsedNanosSafe() : 0);
@@ -261,8 +281,8 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
       if (merge_success) {
         // Merging of operands (associative merge) was successful.
         // Replace operands with the merge result
-        operands_.clear();
-        operands_.emplace_front(std::move(merge_result));
+        merge_context_.Clear();
+        merge_context_.PushOperand(merge_result);
         keys_.erase(keys_.begin(), keys_.end() - 1);
       }
     }

db/merge_helper.h

@@ -8,8 +8,10 @@
 
 #include <deque>
 #include <string>
+#include <vector>
 
 #include "db/dbformat.h"
+#include "db/merge_context.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/env.h"
 #include "rocksdb/slice.h"
@@ -42,14 +44,13 @@ class MergeHelper {
         latest_snapshot_(latest_snapshot),
         level_(level),
         keys_(),
-        operands_(),
         filter_timer_(env_),
         total_filter_time_(0U),
         stats_(stats) {
     assert(user_comparator_ != nullptr);
   }
 
-  // Wrapper around MergeOperator::FullMerge() that records perf statistics.
+  // Wrapper around MergeOperator::FullMergeV2() that records perf statistics.
   // Result of merge will be written to result if status returned is OK.
   // If operands is empty, the value will simply be copied to result.
   // Returns one of the following statuses:
@@ -57,9 +58,10 @@ class MergeHelper {
   // - Corruption: Merge operator reported unsuccessful merge.
   static Status TimedFullMerge(const MergeOperator* merge_operator,
                                const Slice& key, const Slice* value,
-                               const std::deque<std::string>& operands,
+                               const std::vector<Slice>& operands,
                                std::string* result, Logger* logger,
-                               Statistics* statistics, Env* env);
+                               Statistics* statistics, Env* env,
+                               Slice* result_operand = nullptr);
 
   // Merge entries until we hit
   //     - a corrupted key
@@ -116,7 +118,9 @@ class MergeHelper {
   //                So keys().back() was the first key seen by iterator.
   // TODO: Re-style this comment to be like the first one
   const std::deque<std::string>& keys() const { return keys_; }
-  const std::deque<std::string>& values() const { return operands_; }
+  const std::vector<Slice>& values() const {
+    return merge_context_.GetOperands();
+  }
   uint64_t TotalFilterTime() const { return total_filter_time_; }
   bool HasOperator() const { return user_merge_operator_ != nullptr; }
 
@@ -133,8 +137,11 @@ class MergeHelper {
 
   // the scratch area that holds the result of MergeUntil
   // valid up to the next MergeUntil call
-  std::deque<std::string> keys_;    // Keeps track of the sequence of keys seen
-  std::deque<std::string> operands_;  // Parallel with keys_; stores the values
+
+  // Keeps track of the sequence of keys seen
+  std::deque<std::string> keys_;
+  // Parallel with keys_; stores the operands
+  mutable MergeContext merge_context_;
 
   StopWatchNano filter_timer_;
   uint64_t total_filter_time_;
@@ -159,7 +166,7 @@ class MergeOutputIterator {
  private:
   const MergeHelper* merge_helper_;
   std::deque<std::string>::const_reverse_iterator it_keys_;
-  std::deque<std::string>::const_reverse_iterator it_values_;
+  std::vector<Slice>::const_reverse_iterator it_values_;
 };
 
 } // namespace rocksdb

db/merge_operator.cc

@@ -11,6 +11,18 @@
 
 namespace rocksdb {
 
+bool MergeOperator::FullMergeV2(const MergeOperationInput& merge_in,
+                                MergeOperationOutput* merge_out) const {
+  // If FullMergeV2 is not implemented, we convert the operand_list to
+  // std::deque<std::string> and pass it to FullMerge
+  std::deque<std::string> operand_list_str;
+  for (auto& op : merge_in.operand_list) {
+    operand_list_str.emplace_back(op.data(), op.size());
+  }
+  return FullMerge(merge_in.key, merge_in.existing_value, operand_list_str,
+                   &merge_out->new_value, merge_in.logger);
+}
+
 // The default implementation of PartialMergeMulti, which invokes
 // PartialMerge multiple times internally and merges two operands at
 // a time.
@@ -39,23 +51,20 @@ bool MergeOperator::PartialMergeMulti(const Slice& key,
 // Given a "real" merge from the library, call the user's
 // associative merge function one-by-one on each of the operands.
 // NOTE: It is assumed that the client's merge-operator will handle any errors.
-bool AssociativeMergeOperator::FullMerge(
-    const Slice& key,
-    const Slice* existing_value,
-    const std::deque<std::string>& operand_list,
-    std::string* new_value,
-    Logger* logger) const {
-
+bool AssociativeMergeOperator::FullMergeV2(
+    const MergeOperationInput& merge_in,
+    MergeOperationOutput* merge_out) const {
   // Simply loop through the operands
   Slice temp_existing;
-  for (const auto& operand : operand_list) {
-    Slice value(operand);
+  const Slice* existing_value = merge_in.existing_value;
+  for (const auto& operand : merge_in.operand_list) {
     std::string temp_value;
-    if (!Merge(key, existing_value, value, &temp_value, logger)) {
+    if (!Merge(merge_in.key, existing_value, operand, &temp_value,
+               merge_in.logger)) {
       return false;
     }
-    swap(temp_value, *new_value);
-    temp_existing = Slice(*new_value);
+    swap(temp_value, merge_out->new_value);
+    temp_existing = Slice(merge_out->new_value);
     existing_value = &temp_existing;
   }
 

db/pinned_iterators_manager.h

@@ -6,6 +6,7 @@
 #pragma once
 #include <algorithm>
 #include <memory>
+#include <utility>
 #include <vector>
 
 #include "table/internal_iterator.h"
@@ -17,50 +18,70 @@ namespace rocksdb {
 // not needed anymore.
 class PinnedIteratorsManager {
  public:
-  PinnedIteratorsManager() : pinning_enabled(false), pinned_iters_(nullptr) {}
-  ~PinnedIteratorsManager() { assert(!pinning_enabled); }
+  PinnedIteratorsManager() : pinning_enabled(false) {}
+  ~PinnedIteratorsManager() {
+    if (pinning_enabled) {
+      ReleasePinnedData();
+    }
+  }
 
   // Enable Iterators pinning
   void StartPinning() {
-    if (!pinning_enabled) {
-      pinning_enabled = true;
-      if (!pinned_iters_) {
-        pinned_iters_.reset(new std::vector<InternalIterator*>());
-      }
-    }
+    assert(pinning_enabled == false);
+    pinning_enabled = true;
   }
 
   // Is pinning enabled ?
   bool PinningEnabled() { return pinning_enabled; }
 
-  // Take ownership of iter if pinning is enabled and delete it when
-  // ReleasePinnedIterators() is called
-  void PinIteratorIfNeeded(InternalIterator* iter) {
-    if (!pinning_enabled || !iter) {
+  // Take ownership of iter and delete it when ReleasePinnedData() is called
+  void PinIterator(InternalIterator* iter, bool arena = false) {
+    if (arena) {
+      PinPtr(iter, &PinnedIteratorsManager::ReleaseArenaInternalIterator);
+    } else {
+      PinPtr(iter, &PinnedIteratorsManager::ReleaseInternalIterator);
+    }
+  }
+
+  typedef void (*ReleaseFunction)(void* arg1);
+  void PinPtr(void* ptr, ReleaseFunction release_func) {
+    assert(pinning_enabled);
+    if (ptr == nullptr) {
       return;
     }
-    pinned_iters_->push_back(iter);
+    pinned_ptrs_.emplace_back(ptr, release_func);
   }
 
   // Release pinned Iterators
-  void ReleasePinnedIterators() {
-    if (pinning_enabled) {
-      pinning_enabled = false;
+  inline void ReleasePinnedData() {
+    assert(pinning_enabled == true);
+    pinning_enabled = false;
 
-      // Remove duplicate pointers
-      std::sort(pinned_iters_->begin(), pinned_iters_->end());
-      std::unique(pinned_iters_->begin(), pinned_iters_->end());
+    // Remove duplicate pointers
+    std::sort(pinned_ptrs_.begin(), pinned_ptrs_.end());
+    std::unique(pinned_ptrs_.begin(), pinned_ptrs_.end());
 
-      for (auto& iter : *pinned_iters_) {
-        delete iter;
-      }
-      pinned_iters_->clear();
+    for (size_t i = 0; i < pinned_ptrs_.size(); i++) {
+      assert(i == 0 || pinned_ptrs_[i].first != pinned_ptrs_[i - 1].first);
+
+      void* ptr = pinned_ptrs_[i].first;
+      ReleaseFunction release_func = pinned_ptrs_[i].second;
+      release_func(ptr);
     }
+    pinned_ptrs_.clear();
   }
 
  private:
+  static void ReleaseInternalIterator(void* ptr) {
+    delete reinterpret_cast<InternalIterator*>(ptr);
+  }
+
+  static void ReleaseArenaInternalIterator(void* ptr) {
+    reinterpret_cast<InternalIterator*>(ptr)->~InternalIterator();
+  }
+
   bool pinning_enabled;
-  std::unique_ptr<std::vector<InternalIterator*>> pinned_iters_;
+  std::vector<std::pair<void*, ReleaseFunction>> pinned_ptrs_;
 };
 
 }  // namespace rocksdb

db/plain_table_db_test.cc

@@ -324,10 +324,11 @@ class TestPlainTableFactory : public PlainTableFactory {
         column_family_id_(column_family_id),
         column_family_name_(std::move(column_family_name)) {}
 
-  Status NewTableReader(const TableReaderOptions& table_reader_options,
-                        unique_ptr<RandomAccessFileReader>&& file,
-                        uint64_t file_size,
-                        unique_ptr<TableReader>* table) const override {
+  Status NewTableReader(
+      const TableReaderOptions& table_reader_options,
+      unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
+      unique_ptr<TableReader>* table,
+      bool prefetch_index_and_filter_in_cache) const override {
     TableProperties* props = nullptr;
     auto s =
         ReadTableProperties(file.get(), file_size, kPlainTableMagicNumber,

db/prefix_test.cc

@@ -21,9 +21,9 @@ int main() {
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/filter_policy.h"
+#include "rocksdb/memtablerep.h"
 #include "rocksdb/perf_context.h"
 #include "rocksdb/slice_transform.h"
-#include "rocksdb/memtablerep.h"
 #include "rocksdb/table.h"
 #include "util/histogram.h"
 #include "util/stop_watch.h"
@@ -44,7 +44,7 @@ DEFINE_int32(max_write_buffer_number, 2, "");
 DEFINE_int32(min_write_buffer_number_to_merge, 1, "");
 DEFINE_int32(skiplist_height, 4, "");
 DEFINE_double(memtable_prefix_bloom_size_ratio, 0.1, "");
-DEFINE_int32(memtable_prefix_bloom_huge_page_tlb_size, 2 * 1024 * 1024, "");
+DEFINE_int32(memtable_huge_page_size, 2 * 1024 * 1024, "");
 DEFINE_int32(value_size, 40, "");
 
 // Path to the database on file system
@@ -146,6 +146,35 @@ std::string Get(DB* db, const ReadOptions& read_options, uint64_t prefix,
   }
   return result;
 }
+
+class SamePrefixTransform : public SliceTransform {
+ private:
+  const Slice prefix_;
+  std::string name_;
+
+ public:
+  explicit SamePrefixTransform(const Slice& prefix)
+      : prefix_(prefix), name_("rocksdb.SamePrefix." + prefix.ToString()) {}
+
+  virtual const char* Name() const override { return name_.c_str(); }
+
+  virtual Slice Transform(const Slice& src) const override {
+    assert(InDomain(src));
+    return prefix_;
+  }
+
+  virtual bool InDomain(const Slice& src) const override {
+    if (src.size() >= prefix_.size()) {
+      return Slice(src.data(), prefix_.size()) == prefix_;
+    }
+    return false;
+  }
+
+  virtual bool InRange(const Slice& dst) const override {
+    return dst == prefix_;
+  }
+};
+
 }  // namespace
 
 class PrefixTest : public testing::Test {
@@ -161,8 +190,7 @@ class PrefixTest : public testing::Test {
 
     options.memtable_prefix_bloom_size_ratio =
         FLAGS_memtable_prefix_bloom_size_ratio;
-    options.memtable_prefix_bloom_huge_page_tlb_size =
-        FLAGS_memtable_prefix_bloom_huge_page_tlb_size;
+    options.memtable_huge_page_size = FLAGS_memtable_huge_page_size;
 
     options.prefix_extractor.reset(NewFixedPrefixTransform(8));
     BlockBasedTableOptions bbto;
@@ -227,6 +255,55 @@ class PrefixTest : public testing::Test {
   Options options;
 };
 
+TEST(SamePrefixTest, InDomainTest) {
+  DB* db;
+  Options options;
+  options.create_if_missing = true;
+  options.prefix_extractor.reset(new SamePrefixTransform("HHKB"));
+  BlockBasedTableOptions bbto;
+  bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  bbto.whole_key_filtering = false;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  WriteOptions write_options;
+  ReadOptions read_options;
+  {
+    ASSERT_OK(DB::Open(options, kDbName, &db));
+    ASSERT_OK(db->Put(write_options, "HHKB pro2", "Mar 24, 2006"));
+    ASSERT_OK(db->Put(write_options, "HHKB pro2 Type-S", "June 29, 2011"));
+    ASSERT_OK(db->Put(write_options, "Realforce 87u", "idk"));
+    db->Flush(FlushOptions());
+    std::string result;
+    auto db_iter = db->NewIterator(ReadOptions());
+
+    db_iter->Seek("Realforce 87u");
+    ASSERT_TRUE(db_iter->Valid());
+    ASSERT_OK(db_iter->status());
+    ASSERT_EQ(db_iter->key(), "Realforce 87u");
+    ASSERT_EQ(db_iter->value(), "idk");
+
+    delete db_iter;
+    delete db;
+    ASSERT_OK(DestroyDB(kDbName, Options()));
+  }
+
+  {
+    ASSERT_OK(DB::Open(options, kDbName, &db));
+    ASSERT_OK(db->Put(write_options, "pikachu", "1"));
+    ASSERT_OK(db->Put(write_options, "Meowth", "1"));
+    ASSERT_OK(db->Put(write_options, "Mewtwo", "idk"));
+    db->Flush(FlushOptions());
+    std::string result;
+    auto db_iter = db->NewIterator(ReadOptions());
+
+    db_iter->Seek("Mewtwo");
+    ASSERT_TRUE(db_iter->Valid());
+    ASSERT_OK(db_iter->status());
+    delete db_iter;
+    delete db;
+    ASSERT_OK(DestroyDB(kDbName, Options()));
+  }
+}
+
 TEST_F(PrefixTest, TestResult) {
   for (int num_buckets = 1; num_buckets <= 2; num_buckets++) {
     FirstOption();

db/repair.cc

@@ -78,10 +78,10 @@
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
-#include "rocksdb/immutable_options.h"
 #include "rocksdb/options.h"
 #include "rocksdb/write_buffer_manager.h"
 #include "table/scoped_arena_iterator.h"
+#include "util/cf_options.h"
 #include "util/file_reader_writer.h"
 #include "util/string_util.h"
 

db/table_cache.cc

@@ -89,17 +89,17 @@ Status TableCache::GetTableReader(
     const InternalKeyComparator& internal_comparator, const FileDescriptor& fd,
     bool sequential_mode, size_t readahead, bool record_read_stats,
     HistogramImpl* file_read_hist, unique_ptr<TableReader>* table_reader,
-    bool skip_filters, int level) {
+    bool skip_filters, int level, bool prefetch_index_and_filter_in_cache) {
   std::string fname =
       TableFileName(ioptions_.db_paths, fd.GetNumber(), fd.GetPathId());
   unique_ptr<RandomAccessFile> file;
   Status s = ioptions_.env->NewRandomAccessFile(fname, &file, env_options);
 
-  if (readahead > 0) {
-    file = NewReadaheadRandomAccessFile(std::move(file), readahead);
-  }
   RecordTick(ioptions_.statistics, NO_FILE_OPENS);
   if (s.ok()) {
+    if (readahead > 0) {
+      file = NewReadaheadRandomAccessFile(std::move(file), readahead);
+    }
     if (!sequential_mode && ioptions_.advise_random_on_open) {
       file->Hint(RandomAccessFile::RANDOM);
     }
@@ -111,7 +111,8 @@ Status TableCache::GetTableReader(
     s = ioptions_.table_factory->NewTableReader(
         TableReaderOptions(ioptions_, env_options, internal_comparator,
                            skip_filters, level),
-        std::move(file_reader), fd.GetFileSize(), table_reader);
+        std::move(file_reader), fd.GetFileSize(), table_reader,
+        prefetch_index_and_filter_in_cache);
     TEST_SYNC_POINT("TableCache::GetTableReader:0");
   }
   return s;
@@ -129,7 +130,8 @@ Status TableCache::FindTable(const EnvOptions& env_options,
                              const FileDescriptor& fd, Cache::Handle** handle,
                              const bool no_io, bool record_read_stats,
                              HistogramImpl* file_read_hist, bool skip_filters,
-                             int level) {
+                             int level,
+                             bool prefetch_index_and_filter_in_cache) {
   PERF_TIMER_GUARD(find_table_nanos);
   Status s;
   uint64_t number = fd.GetNumber();
@@ -146,7 +148,7 @@ Status TableCache::FindTable(const EnvOptions& env_options,
     s = GetTableReader(env_options, internal_comparator, fd,
                        false /* sequential mode */, 0 /* readahead */,
                        record_read_stats, file_read_hist, &table_reader,
-                       skip_filters, level);
+                       skip_filters, level, prefetch_index_and_filter_in_cache);
     if (!s.ok()) {
       assert(table_reader == nullptr);
       RecordTick(ioptions_.statistics, NO_FILE_ERRORS);

db/table_cache.h

@@ -18,9 +18,10 @@
 #include "port/port.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/env.h"
-#include "rocksdb/table.h"
 #include "rocksdb/options.h"
+#include "rocksdb/table.h"
 #include "table/table_reader.h"
+#include "util/cf_options.h"
 
 namespace rocksdb {
 
@@ -79,7 +80,8 @@ class TableCache {
                    const FileDescriptor& file_fd, Cache::Handle**,
                    const bool no_io = false, bool record_read_stats = true,
                    HistogramImpl* file_read_hist = nullptr,
-                   bool skip_filters = false, int level = -1);
+                   bool skip_filters = false, int level = -1,
+                   bool prefetch_index_and_filter_in_cache = true);
 
   // Get TableReader from a cache handle.
   TableReader* GetTableReaderFromHandle(Cache::Handle* handle);
@@ -114,7 +116,8 @@ class TableCache {
                         size_t readahead, bool record_read_stats,
                         HistogramImpl* file_read_hist,
                         unique_ptr<TableReader>* table_reader,
-                        bool skip_filters = false, int level = -1);
+                        bool skip_filters = false, int level = -1,
+                        bool prefetch_index_and_filter_in_cache = true);
 
   const ImmutableCFOptions& ioptions_;
   const EnvOptions& env_options_;

db/table_properties_collector_test.cc

@@ -12,12 +12,12 @@
 #include "db/db_impl.h"
 #include "db/dbformat.h"
 #include "db/table_properties_collector.h"
-#include "rocksdb/immutable_options.h"
 #include "rocksdb/table.h"
 #include "table/block_based_table_factory.h"
 #include "table/meta_blocks.h"
 #include "table/plain_table_factory.h"
 #include "table/table_builder.h"
+#include "util/cf_options.h"
 #include "util/coding.h"
 #include "util/file_reader_writer.h"
 #include "util/testharness.h"

db/version_builder.cc

@@ -287,7 +287,8 @@ class VersionBuilder::Rep {
     CheckConsistency(vstorage);
   }
 
-  void LoadTableHandlers(InternalStats* internal_stats, int max_threads) {
+  void LoadTableHandlers(InternalStats* internal_stats, int max_threads,
+                         bool prefetch_index_and_filter_in_cache) {
     assert(table_cache_ != nullptr);
     // <file metadata, level>
     std::vector<std::pair<FileMetaData*, int>> files_meta;
@@ -309,11 +310,12 @@ class VersionBuilder::Rep {
 
         auto* file_meta = files_meta[file_idx].first;
         int level = files_meta[file_idx].second;
-        table_cache_->FindTable(
-            env_options_, *(base_vstorage_->InternalComparator()),
-            file_meta->fd, &file_meta->table_reader_handle, false /*no_io */,
-            true /* record_read_stats */,
-            internal_stats->GetFileReadHist(level), false, level);
+        table_cache_->FindTable(env_options_,
+                                *(base_vstorage_->InternalComparator()),
+                                file_meta->fd, &file_meta->table_reader_handle,
+                                false /*no_io */, true /* record_read_stats */,
+                                internal_stats->GetFileReadHist(level), false,
+                                level, prefetch_index_and_filter_in_cache);
         if (file_meta->table_reader_handle != nullptr) {
           // Load table_reader
           file_meta->fd.table_reader = table_cache_->GetTableReaderFromHandle(
@@ -363,9 +365,11 @@ void VersionBuilder::Apply(VersionEdit* edit) { rep_->Apply(edit); }
 void VersionBuilder::SaveTo(VersionStorageInfo* vstorage) {
   rep_->SaveTo(vstorage);
 }
-void VersionBuilder::LoadTableHandlers(InternalStats* internal_stats,
-                                       int max_threads) {
-  rep_->LoadTableHandlers(internal_stats, max_threads);
+void VersionBuilder::LoadTableHandlers(
+    InternalStats* internal_stats, int max_threads,
+    bool prefetch_index_and_filter_in_cache) {
+  rep_->LoadTableHandlers(internal_stats, max_threads,
+                          prefetch_index_and_filter_in_cache);
 }
 void VersionBuilder::MaybeAddFile(VersionStorageInfo* vstorage, int level,
                                   FileMetaData* f) {

db/version_builder.h

@@ -31,7 +31,8 @@ class VersionBuilder {
                                   int level);
   void Apply(VersionEdit* edit);
   void SaveTo(VersionStorageInfo* vstorage);
-  void LoadTableHandlers(InternalStats* internal_stats, int max_threads = 1);
+  void LoadTableHandlers(InternalStats* internal_stats, int max_threads,
+                         bool prefetch_index_and_filter_in_cache);
   void MaybeAddFile(VersionStorageInfo* vstorage, int level, FileMetaData* f);
 
  private:

db/version_set.cc

@@ -31,6 +31,7 @@
 #include "db/memtable.h"
 #include "db/merge_context.h"
 #include "db/merge_helper.h"
+#include "db/pinned_iterators_manager.h"
 #include "db/table_cache.h"
 #include "db/version_builder.h"
 #include "rocksdb/env.h"
@@ -917,10 +918,17 @@ void Version::Get(const ReadOptions& read_options, const LookupKey& k,
     *key_exists = true;
   }
 
+  PinnedIteratorsManager pinned_iters_mgr;
   GetContext get_context(
       user_comparator(), merge_operator_, info_log_, db_statistics_,
       status->ok() ? GetContext::kNotFound : GetContext::kMerge, user_key,
-      value, value_found, merge_context, this->env_, seq);
+      value, value_found, merge_context, this->env_, seq,
+      merge_operator_ ? &pinned_iters_mgr : nullptr);
+
+  // Pin blocks that we read to hold merge operands
+  if (merge_operator_) {
+    pinned_iters_mgr.StartPinning();
+  }
 
   FilePicker fp(
       storage_info_.files_, user_key, ikey, &storage_info_.level_files_brief_,
@@ -2264,7 +2272,8 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
       // Need to do it out of the mutex.
       builder_guard->version_builder()->LoadTableHandlers(
           column_family_data->internal_stats(),
-          column_family_data->ioptions()->optimize_filters_for_hits);
+          column_family_data->ioptions()->optimize_filters_for_hits,
+          true /* prefetch_index_and_filter_in_cache */);
     }
 
     // This is fine because everything inside of this block is serialized --
@@ -2706,8 +2715,9 @@ Status VersionSet::Recover(
       if (db_options_->max_open_files == -1) {
         // unlimited table cache. Pre-load table handle now.
         // Need to do it out of the mutex.
-        builder->LoadTableHandlers(cfd->internal_stats(),
-                                   db_options_->max_file_opening_threads);
+        builder->LoadTableHandlers(
+            cfd->internal_stats(), db_options_->max_file_opening_threads,
+            false /* prefetch_index_and_filter_in_cache */);
       }
 
       Version* v = new Version(cfd, this, current_version_number_++);

db/wal_manager.cc

@@ -383,7 +383,7 @@ Status WalManager::ReadFirstRecord(const WalFileType type,
   Status s;
   if (type == kAliveLogFile) {
     std::string fname = LogFileName(db_options_.wal_dir, number);
-    s = ReadFirstLine(fname, sequence);
+    s = ReadFirstLine(fname, number, sequence);
     if (env_->FileExists(fname).ok() && !s.ok()) {
       // return any error that is not caused by non-existing file
       return s;
@@ -394,7 +394,7 @@ Status WalManager::ReadFirstRecord(const WalFileType type,
     //  check if the file got moved to archive.
     std::string archived_file =
         ArchivedLogFileName(db_options_.wal_dir, number);
-    s = ReadFirstLine(archived_file, sequence);
+    s = ReadFirstLine(archived_file, number, sequence);
     // maybe the file was deleted from archive dir. If that's the case, return
     // Status::OK(). The caller with identify this as empty file because
     // *sequence == 0
@@ -413,6 +413,7 @@ Status WalManager::ReadFirstRecord(const WalFileType type,
 // the function returns status.ok() and sequence == 0 if the file exists, but is
 // empty
 Status WalManager::ReadFirstLine(const std::string& fname,
+                                 const uint64_t number,
                                  SequenceNumber* sequence) {
   struct LogReporter : public log::Reader::Reporter {
     Env* env;
@@ -449,7 +450,7 @@ Status WalManager::ReadFirstLine(const std::string& fname,
   reporter.status = &status;
   reporter.ignore_error = !db_options_.paranoid_checks;
   log::Reader reader(db_options_.info_log, std::move(file_reader), &reporter,
-                     true /*checksum*/, 0 /*initial_offset*/, *sequence);
+                     true /*checksum*/, 0 /*initial_offset*/, number);
   std::string scratch;
   Slice record;
 

db/wal_manager.h

@@ -54,9 +54,9 @@ class WalManager {
     return ReadFirstRecord(type, number, sequence);
   }
 
-  Status TEST_ReadFirstLine(const std::string& fname,
+  Status TEST_ReadFirstLine(const std::string& fname, const uint64_t number,
                             SequenceNumber* sequence) {
-    return ReadFirstLine(fname, sequence);
+    return ReadFirstLine(fname, number, sequence);
   }
 
  private:
@@ -71,7 +71,8 @@ class WalManager {
   Status ReadFirstRecord(const WalFileType type, const uint64_t number,
                          SequenceNumber* sequence);
 
-  Status ReadFirstLine(const std::string& fname, SequenceNumber* sequence);
+  Status ReadFirstLine(const std::string& fname, const uint64_t number,
+                       SequenceNumber* sequence);
 
   // ------- state from DBImpl ------
   const DBOptions& db_options_;

db/wal_manager_test.cc

@@ -119,10 +119,11 @@ TEST_F(WalManagerTest, ReadFirstRecordCache) {
   ASSERT_OK(env_->NewWritableFile(path, &file, EnvOptions()));
 
   SequenceNumber s;
-  ASSERT_OK(wal_manager_->TEST_ReadFirstLine(path, &s));
+  ASSERT_OK(wal_manager_->TEST_ReadFirstLine(path, 1 /* number */, &s));
   ASSERT_EQ(s, 0U);
 
-  ASSERT_OK(wal_manager_->TEST_ReadFirstRecord(kAliveLogFile, 1, &s));
+  ASSERT_OK(
+      wal_manager_->TEST_ReadFirstRecord(kAliveLogFile, 1 /* number */, &s));
   ASSERT_EQ(s, 0U);
 
   unique_ptr<WritableFileWriter> file_writer(

db/write_batch.cc

@@ -64,6 +64,7 @@ enum ContentFlags : uint32_t {
   HAS_END_PREPARE = 1 << 6,
   HAS_COMMIT = 1 << 7,
   HAS_ROLLBACK = 1 << 8,
+  HAS_DELETE_RANGE = 1 << 9,
 };
 
 struct BatchContentClassifier : public WriteBatch::Handler {
@@ -84,6 +85,11 @@ struct BatchContentClassifier : public WriteBatch::Handler {
     return Status::OK();
   }
 
+  Status DeleteRangeCF(uint32_t, const Slice&, const Slice&) override {
+    content_flags |= ContentFlags::HAS_DELETE_RANGE;
+    return Status::OK();
+  }
+
   Status MergeCF(uint32_t, const Slice&, const Slice&) override {
     content_flags |= ContentFlags::HAS_MERGE;
     return Status::OK();
@@ -219,6 +225,10 @@ bool WriteBatch::HasSingleDelete() const {
   return (ComputeContentFlags() & ContentFlags::HAS_SINGLE_DELETE) != 0;
 }
 
+bool WriteBatch::HasDeleteRange() const {
+  return (ComputeContentFlags() & ContentFlags::HAS_DELETE_RANGE) != 0;
+}
+
 bool WriteBatch::HasMerge() const {
   return (ComputeContentFlags() & ContentFlags::HAS_MERGE) != 0;
 }
@@ -287,6 +297,18 @@ Status ReadRecordFromWriteBatch(Slice* input, char* tag,
         return Status::Corruption("bad WriteBatch Delete");
       }
       break;
+    case kTypeColumnFamilyRangeDeletion:
+      if (!GetVarint32(input, column_family)) {
+        return Status::Corruption("bad WriteBatch DeleteRange");
+      }
+    // intentional fallthrough
+    case kTypeRangeDeletion:
+      // for range delete, "key" is begin_key, "value" is end_key
+      if (!GetLengthPrefixedSlice(input, key) ||
+          !GetLengthPrefixedSlice(input, value)) {
+        return Status::Corruption("bad WriteBatch DeleteRange");
+      }
+      break;
     case kTypeColumnFamilyMerge:
       if (!GetVarint32(input, column_family)) {
         return Status::Corruption("bad WriteBatch Merge");
@@ -370,6 +392,13 @@ Status WriteBatch::Iterate(Handler* handler) const {
         s = handler->SingleDeleteCF(column_family, key);
         found++;
         break;
+      case kTypeColumnFamilyRangeDeletion:
+      case kTypeRangeDeletion:
+        assert(content_flags_.load(std::memory_order_relaxed) &
+               (ContentFlags::DEFERRED | ContentFlags::HAS_DELETE_RANGE));
+        s = handler->DeleteRangeCF(column_family, key, value);
+        found++;
+        break;
       case kTypeColumnFamilyMerge:
       case kTypeMerge:
         assert(content_flags_.load(std::memory_order_relaxed) &
@@ -598,6 +627,53 @@ void WriteBatch::SingleDelete(ColumnFamilyHandle* column_family,
   WriteBatchInternal::SingleDelete(this, GetColumnFamilyID(column_family), key);
 }
 
+void WriteBatchInternal::DeleteRange(WriteBatch* b, uint32_t column_family_id,
+                                     const Slice& begin_key,
+                                     const Slice& end_key) {
+  WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
+  if (column_family_id == 0) {
+    b->rep_.push_back(static_cast<char>(kTypeRangeDeletion));
+  } else {
+    b->rep_.push_back(static_cast<char>(kTypeColumnFamilyRangeDeletion));
+    PutVarint32(&b->rep_, column_family_id);
+  }
+  PutLengthPrefixedSlice(&b->rep_, begin_key);
+  PutLengthPrefixedSlice(&b->rep_, end_key);
+  b->content_flags_.store(b->content_flags_.load(std::memory_order_relaxed) |
+                              ContentFlags::HAS_DELETE_RANGE,
+                          std::memory_order_relaxed);
+}
+
+void WriteBatch::DeleteRange(ColumnFamilyHandle* column_family,
+                             const Slice& begin_key, const Slice& end_key) {
+  WriteBatchInternal::DeleteRange(this, GetColumnFamilyID(column_family),
+                                  begin_key, end_key);
+}
+
+void WriteBatchInternal::DeleteRange(WriteBatch* b, uint32_t column_family_id,
+                                     const SliceParts& begin_key,
+                                     const SliceParts& end_key) {
+  WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
+  if (column_family_id == 0) {
+    b->rep_.push_back(static_cast<char>(kTypeRangeDeletion));
+  } else {
+    b->rep_.push_back(static_cast<char>(kTypeColumnFamilyRangeDeletion));
+    PutVarint32(&b->rep_, column_family_id);
+  }
+  PutLengthPrefixedSliceParts(&b->rep_, begin_key);
+  PutLengthPrefixedSliceParts(&b->rep_, end_key);
+  b->content_flags_.store(b->content_flags_.load(std::memory_order_relaxed) |
+                              ContentFlags::HAS_DELETE_RANGE,
+                          std::memory_order_relaxed);
+}
+
+void WriteBatch::DeleteRange(ColumnFamilyHandle* column_family,
+                             const SliceParts& begin_key,
+                             const SliceParts& end_key) {
+  WriteBatchInternal::DeleteRange(this, GetColumnFamilyID(column_family),
+                                  begin_key, end_key);
+}
+
 void WriteBatchInternal::Merge(WriteBatch* b, uint32_t column_family_id,
                                const Slice& key, const Slice& value) {
   WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
@@ -836,9 +912,9 @@ class MemTableInserter : public WriteBatch::Handler {
   }
 
   Status DeleteImpl(uint32_t column_family_id, const Slice& key,
-                    ValueType delete_type) {
+                    const Slice& value, ValueType delete_type) {
     MemTable* mem = cf_mems_->GetMemTable();
-    mem->Add(sequence_, delete_type, key, Slice(), concurrent_memtable_writes_,
+    mem->Add(sequence_, delete_type, key, value, concurrent_memtable_writes_,
              get_post_process_info(mem));
     sequence_++;
     CheckMemtableFull();
@@ -858,7 +934,7 @@ class MemTableInserter : public WriteBatch::Handler {
       return seek_status;
     }
 
-    return DeleteImpl(column_family_id, key, kTypeDeletion);
+    return DeleteImpl(column_family_id, key, Slice(), kTypeDeletion);
   }
 
   virtual Status SingleDeleteCF(uint32_t column_family_id,
@@ -874,7 +950,25 @@ class MemTableInserter : public WriteBatch::Handler {
       return seek_status;
     }
 
-    return DeleteImpl(column_family_id, key, kTypeSingleDeletion);
+    return DeleteImpl(column_family_id, key, Slice(), kTypeSingleDeletion);
+  }
+
+  virtual Status DeleteRangeCF(uint32_t column_family_id,
+                               const Slice& begin_key,
+                               const Slice& end_key) override {
+    if (rebuilding_trx_ != nullptr) {
+      WriteBatchInternal::DeleteRange(rebuilding_trx_, column_family_id,
+                                      begin_key, end_key);
+      return Status::OK();
+    }
+
+    Status seek_status;
+    if (!SeekToColumnFamily(column_family_id, &seek_status)) {
+      ++sequence_;
+      return seek_status;
+    }
+
+    return DeleteImpl(column_family_id, begin_key, end_key, kTypeRangeDeletion);
   }
 
   virtual Status MergeCF(uint32_t column_family_id, const Slice& key,
@@ -929,12 +1023,10 @@ class MemTableInserter : public WriteBatch::Handler {
       auto merge_operator = moptions->merge_operator;
       assert(merge_operator);
 
-      std::deque<std::string> operands;
-      operands.push_front(value.ToString());
       std::string new_value;
 
       Status merge_status = MergeHelper::TimedFullMerge(
-          merge_operator, key, &get_value_slice, operands, &new_value,
+          merge_operator, key, &get_value_slice, {value}, &new_value,
           moptions->info_log, moptions->statistics, Env::Default());
 
       if (!merge_status.ok()) {

db/write_batch_base.cc

@@ -56,8 +56,25 @@ void WriteBatchBase::SingleDelete(const SliceParts& key) {
   SingleDelete(key_slice);
 }
 
+void WriteBatchBase::DeleteRange(ColumnFamilyHandle* column_family,
+                                 const SliceParts& begin_key,
+                                 const SliceParts& end_key) {
+  std::string begin_key_buf, end_key_buf;
+  Slice begin_key_slice(begin_key, &begin_key_buf);
+  Slice end_key_slice(end_key, &end_key_buf);
+  DeleteRange(column_family, begin_key_slice, end_key_slice);
+}
+
+void WriteBatchBase::DeleteRange(const SliceParts& begin_key,
+                                 const SliceParts& end_key) {
+  std::string begin_key_buf, end_key_buf;
+  Slice begin_key_slice(begin_key, &begin_key_buf);
+  Slice end_key_slice(end_key, &end_key_buf);
+  DeleteRange(begin_key_slice, end_key_slice);
+}
+
 void WriteBatchBase::Merge(ColumnFamilyHandle* column_family,
-                         const SliceParts& key, const SliceParts& value) {
+                           const SliceParts& key, const SliceParts& value) {
   std::string key_buf, value_buf;
   Slice key_slice(key, &key_buf);
   Slice value_slice(value, &value_buf);

db/write_batch_internal.h

@@ -86,6 +86,13 @@ class WriteBatchInternal {
   static void SingleDelete(WriteBatch* batch, uint32_t column_family_id,
                            const Slice& key);
 
+  static void DeleteRange(WriteBatch* b, uint32_t column_family_id,
+                          const Slice& begin_key, const Slice& end_key);
+
+  static void DeleteRange(WriteBatch* b, uint32_t column_family_id,
+                          const SliceParts& begin_key,
+                          const SliceParts& end_key);
+
   static void Merge(WriteBatch* batch, uint32_t column_family_id,
                     const Slice& key, const Slice& value);
 

db/write_batch_test.cc

@@ -42,6 +42,7 @@ static std::string PrintContents(WriteBatch* b) {
   int put_count = 0;
   int delete_count = 0;
   int single_delete_count = 0;
+  int delete_range_count = 0;
   int merge_count = 0;
   Arena arena;
   ScopedArenaIterator iter(mem->NewIterator(ReadOptions(), &arena));
@@ -73,6 +74,15 @@ static std::string PrintContents(WriteBatch* b) {
         count++;
         single_delete_count++;
         break;
+      case kTypeRangeDeletion:
+        state.append("DeleteRange(");
+        state.append(ikey.user_key.ToString());
+        state.append(", ");
+        state.append(iter->value().ToString());
+        state.append(")");
+        count++;
+        delete_range_count++;
+        break;
       case kTypeMerge:
         state.append("Merge(");
         state.append(ikey.user_key.ToString());
@@ -92,6 +102,7 @@ static std::string PrintContents(WriteBatch* b) {
   EXPECT_EQ(b->HasPut(), put_count > 0);
   EXPECT_EQ(b->HasDelete(), delete_count > 0);
   EXPECT_EQ(b->HasSingleDelete(), single_delete_count > 0);
+  EXPECT_EQ(b->HasDeleteRange(), delete_range_count > 0);
   EXPECT_EQ(b->HasMerge(), merge_count > 0);
   if (!s.ok()) {
     state.append(s.ToString());
@@ -115,15 +126,18 @@ TEST_F(WriteBatchTest, Multiple) {
   WriteBatch batch;
   batch.Put(Slice("foo"), Slice("bar"));
   batch.Delete(Slice("box"));
+  batch.DeleteRange(Slice("bar"), Slice("foo"));
   batch.Put(Slice("baz"), Slice("boo"));
   WriteBatchInternal::SetSequence(&batch, 100);
   ASSERT_EQ(100U, WriteBatchInternal::Sequence(&batch));
-  ASSERT_EQ(3, WriteBatchInternal::Count(&batch));
-  ASSERT_EQ("Put(baz, boo)@102"
-            "Delete(box)@101"
-            "Put(foo, bar)@100",
-            PrintContents(&batch));
-  ASSERT_EQ(3, batch.Count());
+  ASSERT_EQ(4, WriteBatchInternal::Count(&batch));
+  ASSERT_EQ(
+      "DeleteRange(bar, foo)@102"
+      "Put(baz, boo)@103"
+      "Delete(box)@101"
+      "Put(foo, bar)@100",
+      PrintContents(&batch));
+  ASSERT_EQ(4, batch.Count());
 }
 
 TEST_F(WriteBatchTest, Corruption) {
@@ -218,6 +232,18 @@ namespace {
       }
       return Status::OK();
     }
+    virtual Status DeleteRangeCF(uint32_t column_family_id,
+                                 const Slice& begin_key,
+                                 const Slice& end_key) override {
+      if (column_family_id == 0) {
+        seen += "DeleteRange(" + begin_key.ToString() + ", " +
+                end_key.ToString() + ")";
+      } else {
+        seen += "DeleteRangeCF(" + ToString(column_family_id) + ", " +
+                begin_key.ToString() + ", " + end_key.ToString() + ")";
+      }
+      return Status::OK();
+    }
     virtual Status MergeCF(uint32_t column_family_id, const Slice& key,
                            const Slice& value) override {
       if (column_family_id == 0) {
@@ -546,7 +572,7 @@ class ColumnFamilyHandleImplDummy : public ColumnFamilyHandleImpl {
   explicit ColumnFamilyHandleImplDummy(int id)
       : ColumnFamilyHandleImpl(nullptr, nullptr, nullptr), id_(id) {}
   uint32_t GetID() const override { return id_; }
-  const Comparator* user_comparator() const override {
+  const Comparator* GetComparator() const override {
     return BytewiseComparator();
   }
 
@@ -563,6 +589,7 @@ TEST_F(WriteBatchTest, ColumnFamiliesBatchTest) {
   batch.Put(&eight, Slice("eightfoo"), Slice("bar8"));
   batch.Delete(&eight, Slice("eightfoo"));
   batch.SingleDelete(&two, Slice("twofoo"));
+  batch.DeleteRange(&two, Slice("3foo"), Slice("4foo"));
   batch.Merge(&three, Slice("threethree"), Slice("3three"));
   batch.Put(&zero, Slice("foo"), Slice("bar"));
   batch.Merge(Slice("omom"), Slice("nom"));
@@ -575,6 +602,7 @@ TEST_F(WriteBatchTest, ColumnFamiliesBatchTest) {
       "PutCF(8, eightfoo, bar8)"
       "DeleteCF(8, eightfoo)"
       "SingleDeleteCF(2, twofoo)"
+      "DeleteRangeCF(2, 3foo, 4foo)"
       "MergeCF(3, threethree, 3three)"
       "Put(foo, bar)"
       "Merge(omom, nom)",
@@ -590,6 +618,7 @@ TEST_F(WriteBatchTest, ColumnFamiliesBatchWithIndexTest) {
   batch.Put(&eight, Slice("eightfoo"), Slice("bar8"));
   batch.Delete(&eight, Slice("eightfoo"));
   batch.SingleDelete(&two, Slice("twofoo"));
+  batch.DeleteRange(&two, Slice("twofoo"), Slice("threefoo"));
   batch.Merge(&three, Slice("threethree"), Slice("3three"));
   batch.Put(&zero, Slice("foo"), Slice("bar"));
   batch.Merge(Slice("omom"), Slice("nom"));
@@ -628,6 +657,13 @@ TEST_F(WriteBatchTest, ColumnFamiliesBatchWithIndexTest) {
   ASSERT_EQ(WriteType::kSingleDeleteRecord, iter->Entry().type);
   ASSERT_EQ("twofoo", iter->Entry().key.ToString());
 
+  iter->Next();
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(WriteType::kDeleteRangeRecord, iter->Entry().type);
+  ASSERT_EQ("twofoo", iter->Entry().key.ToString());
+  ASSERT_EQ("threefoo", iter->Entry().value.ToString());
+
   iter->Next();
   ASSERT_OK(iter->status());
   ASSERT_TRUE(!iter->Valid());
@@ -678,6 +714,7 @@ TEST_F(WriteBatchTest, ColumnFamiliesBatchWithIndexTest) {
       "PutCF(8, eightfoo, bar8)"
       "DeleteCF(8, eightfoo)"
       "SingleDeleteCF(2, twofoo)"
+      "DeleteRangeCF(2, twofoo, threefoo)"
       "MergeCF(3, threethree, 3three)"
       "Put(foo, bar)"
       "Merge(omom, nom)",

db/write_thread.cc

@@ -185,8 +185,8 @@ void WriteThread::SetState(Writer* w, uint8_t new_state) {
 void WriteThread::LinkOne(Writer* w, bool* linked_as_leader) {
   assert(w->state == STATE_INIT);
 
-  Writer* writers = newest_writer_.load(std::memory_order_relaxed);
   while (true) {
+    Writer* writers = newest_writer_.load(std::memory_order_relaxed);
     w->link_older = writers;
     if (newest_writer_.compare_exchange_strong(writers, w)) {
       if (writers == nullptr) {

doc/doc.css

@@ -1,89 +0,0 @@
-body {
-  margin-left: 0.5in;
-  margin-right: 0.5in;
-  background: white;
-  color: black;
-}
-
-h1 {
-  margin-left: -0.2in;
-  font-size: 14pt;
-}
-h2 {
-  margin-left: -0in;
-  font-size: 12pt;
-}
-h3 {
-  margin-left: -0in;
-}
-h4 {
-  margin-left: -0in;
-}
-hr {
-  margin-left: -0in;
-}
-
-/* Definition lists: definition term bold */
-dt {
-  font-weight: bold;
-}
-
-address {
-  text-align: center;
-}
-code,samp,var {
-  color: blue;
-}
-kbd {
-  color: #600000;
-}
-div.note p {
-  float: right;
-  width: 3in;
-  margin-right: 0%;
-  padding: 1px;
-  border: 2px solid #6060a0;
-  background-color: #fffff0;
-}
-
-ul {
-  margin-top: -0em;
-  margin-bottom: -0em;
-}
-
-ol {
-  margin-top: -0em;
-  margin-bottom: -0em;
-}
-
-UL.nobullets {
-  list-style-type: none;
-  list-style-image: none;
-  margin-left: -1em;
-}
-
-p {
-  margin: 1em 0 1em 0;
-  padding: 0 0 0 0;
-}
-
-pre {
-  line-height: 1.3em;
-  padding: 0.4em 0 0.8em 0;
-  margin:  0 0 0 0;
-  border:  0 0 0 0;
-  color: blue;
-}
-
-.datatable {
-  margin-left: auto;
-  margin-right: auto;
-  margin-top: 2em;
-  margin-bottom: 2em;
-  border: 1px solid;
-}
-
-.datatable td,th {
-  padding: 0 0.5em 0 0.5em;
-  text-align: right;
-}

doc/index.html

@@ -1,827 +0,0 @@
-<!DOCTYPE html>
-<html>
-<head>
-<link rel="stylesheet" type="text/css" href="doc.css" />
-<title>RocksDB</title>
-</head>
-
-<body>
-<h1>RocksDB</h1>
-<address>The Facebook Database Engineering Team</address>
-<address>Build on earlier work on leveldb by Sanjay Ghemawat
-               (sanjay@google.com) and Jeff Dean (jeff@google.com)</address>
-<p>
-The <code>rocksdb</code> library provides a persistent key value store.  Keys and
-values are arbitrary byte arrays.  The keys are ordered within the key
-value store according to a user-specified comparator function.
-
-<p>
-<h1>Opening A Database</h1>
-<p>
-A <code>rocksdb</code> database has a name which corresponds to a file system
-directory.  All of the contents of database are stored in this
-directory.  The following example shows how to open a database,
-creating it if necessary:
-<p>
-<pre>
-  #include &lt;assert&gt;
-  #include "rocksdb/db.h"
-
-  rocksdb::DB* db;
-  rocksdb::Options options;
-  options.create_if_missing = true;
-  rocksdb::Status status = rocksdb::DB::Open(options, "/tmp/testdb", &amp;db);
-  assert(status.ok());
-  ...
-</pre>
-If you want to raise an error if the database already exists, add
-the following line before the <code>rocksdb::DB::Open</code> call:
-<pre>
-  options.error_if_exists = true;
-</pre>
-<h1>Status</h1>
-<p>
-You may have noticed the <code>rocksdb::Status</code> type above.  Values of this
-type are returned by most functions in <code>rocksdb</code> that may encounter an
-error.  You can check if such a result is ok, and also print an
-associated error message:
-<p>
-<pre>
-   rocksdb::Status s = ...;
-   if (!s.ok()) cerr &lt;&lt; s.ToString() &lt;&lt; endl;
-</pre>
-<h1>Closing A Database</h1>
-<p>
-When you are done with a database, just delete the database object.
-Example:
-<p>
-<pre>
-  ... open the db as described above ...
-  ... do something with db ...
-  delete db;
-</pre>
-<h1>Reads And Writes</h1>
-<p>
-The database provides <code>Put</code>, <code>Delete</code>, and <code>Get</code> methods to
-modify/query the database.  For example, the following code
-moves the value stored under key1 to key2.
-<pre>
-  std::string value;
-  rocksdb::Status s = db-&gt;Get(rocksdb::ReadOptions(), key1, &amp;value);
-  if (s.ok()) s = db-&gt;Put(rocksdb::WriteOptions(), key2, value);
-  if (s.ok()) s = db-&gt;Delete(rocksdb::WriteOptions(), key1);
-</pre>
-
-<h1>Atomic Updates</h1>
-<p>
-Note that if the process dies after the Put of key2 but before the
-delete of key1, the same value may be left stored under multiple keys.
-Such problems can be avoided by using the <code>WriteBatch</code> class to
-atomically apply a set of updates:
-<p>
-<pre>
-  #include "rocksdb/write_batch.h"
-  ...
-  std::string value;
-  rocksdb::Status s = db-&gt;Get(rocksdb::ReadOptions(), key1, &amp;value);
-  if (s.ok()) {
-    rocksdb::WriteBatch batch;
-    batch.Delete(key1);
-    batch.Put(key2, value);
-    s = db-&gt;Write(rocksdb::WriteOptions(), &amp;batch);
-  }
-</pre>
-The <code>WriteBatch</code> holds a sequence of edits to be made to the database,
-and these edits within the batch are applied in order.  Note that we
-called <code>Delete</code> before <code>Put</code> so that if <code>key1</code> is identical to <code>key2</code>,
-we do not end up erroneously dropping the value entirely.
-<p>
-Apart from its atomicity benefits, <code>WriteBatch</code> may also be used to
-speed up bulk updates by placing lots of individual mutations into the
-same batch.
-
-<h1>Synchronous Writes</h1>
-By default, each write to <code>leveldb</code> is asynchronous: it
-returns after pushing the write from the process into the operating
-system.  The transfer from operating system memory to the underlying
-persistent storage happens asynchronously.  The <code>sync</code> flag
-can be turned on for a particular write to make the write operation
-not return until the data being written has been pushed all the way to
-persistent storage.  (On Posix systems, this is implemented by calling
-either <code>fsync(...)</code> or <code>fdatasync(...)</code> or
-<code>msync(..., MS_SYNC)</code> before the write operation returns.)
-<pre>
-  rocksdb::WriteOptions write_options;
-  write_options.sync = true;
-  db-&gt;Put(write_options, ...);
-</pre>
-Asynchronous writes are often more than a thousand times as fast as
-synchronous writes.  The downside of asynchronous writes is that a
-crash of the machine may cause the last few updates to be lost.  Note
-that a crash of just the writing process (i.e., not a reboot) will not
-cause any loss since even when <code>sync</code> is false, an update
-is pushed from the process memory into the operating system before it
-is considered done.
-
-<p>
-Asynchronous writes can often be used safely.  For example, when
-loading a large amount of data into the database you can handle lost
-updates by restarting the bulk load after a crash.  A hybrid scheme is
-also possible where every Nth write is synchronous, and in the event
-of a crash, the bulk load is restarted just after the last synchronous
-write finished by the previous run.  (The synchronous write can update
-a marker that describes where to restart on a crash.)
-
-<p>
-<code>WriteBatch</code> provides an alternative to asynchronous writes.
-Multiple updates may be placed in the same <code>WriteBatch</code> and
-applied together using a synchronous write (i.e.,
-<code>write_options.sync</code> is set to true).  The extra cost of
-the synchronous write will be amortized across all of the writes in
-the batch.
-
-<p>
-We also provide a way to completely disable Write Ahead Log for a
-particular write. If you set write_option.disableWAL to true, the
-write will not go to the log at all and may be lost in an event of
-process crash.
-
-<p>
-When opening a DB, you can disable syncing of data files by setting
-Options::disableDataSync to true. This can be useful when doing
-bulk-loading or big idempotent operations. Once the operation is
-finished, you can manually call sync() to flush all dirty buffers
-to stable storage.
-
-<p>
-RocksDB by default uses faster fdatasync() to sync files. If you want
-to use fsync(), you can set Options::use_fsync to true. You should set
-this to true on filesystems like ext3 that can lose files after a
-reboot.
-
-<p>
-<h1>Concurrency</h1>
-<p>
-A database may only be opened by one process at a time.
-The <code>rocksdb</code> implementation acquires a lock from the
-operating system to prevent misuse.  Within a single process, the
-same <code>rocksdb::DB</code> object may be safely shared by multiple
-concurrent threads.  I.e., different threads may write into or fetch
-iterators or call <code>Get</code> on the same database without any
-external synchronization (the leveldb implementation will
-automatically do the required synchronization).  However other objects
-(like Iterator and WriteBatch) may require external synchronization.
-If two threads share such an object, they must protect access to it
-using their own locking protocol.  More details are available in
-the public header files.
-
-<p>
-<h1>Merge operators</h1>
-<p>
-Merge operators provide efficient support for read-modify-write operation.
-More on the interface and implementation can be found on:
-<p>
-<a href="https://github.com/facebook/rocksdb/wiki/Merge-Operator">
-    Merge Operator</a>
-<p>
-<a href="https://github.com/facebook/rocksdb/wiki/Merge-Operator-Implementation">
-    Merge Operator Implementation</a>
-
-<p>
-<h1>Iteration</h1>
-<p>
-The following example demonstrates how to print all key,value pairs
-in a database.
-<p>
-<pre>
-  rocksdb::Iterator* it = db-&gt;NewIterator(rocksdb::ReadOptions());
-  for (it-&gt;SeekToFirst(); it-&gt;Valid(); it-&gt;Next()) {
-    cout &lt;&lt; it-&gt;key().ToString() &lt;&lt; ": "  &lt;&lt; it-&gt;value().ToString() &lt;&lt; endl;
-  }
-  assert(it-&gt;status().ok());  // Check for any errors found during the scan
-  delete it;
-</pre>
-The following variation shows how to process just the keys in the
-range <code>[start,limit)</code>:
-<p>
-<pre>
-  for (it-&gt;Seek(start);
-       it-&gt;Valid() &amp;&amp; it-&gt;key().ToString() &lt; limit;
-       it-&gt;Next()) {
-    ...
-  }
-</pre>
-You can also process entries in reverse order.  (Caveat: reverse
-iteration may be somewhat slower than forward iteration.)
-<p>
-<pre>
-  for (it-&gt;SeekToLast(); it-&gt;Valid(); it-&gt;Prev()) {
-    ...
-  }
-</pre>
-<h1>Snapshots</h1>
-<p>
-Snapshots provide consistent read-only views over the entire state of
-the key-value store.  <code>ReadOptions::snapshot</code> may be non-NULL to indicate
-that a read should operate on a particular version of the DB state.
-If <code>ReadOptions::snapshot</code> is NULL, the read will operate on an
-implicit snapshot of the current state.
-<p>
-Snapshots are created by the DB::GetSnapshot() method:
-<p>
-<pre>
-  rocksdb::ReadOptions options;
-  options.snapshot = db-&gt;GetSnapshot();
-  ... apply some updates to db ...
-  rocksdb::Iterator* iter = db-&gt;NewIterator(options);
-  ... read using iter to view the state when the snapshot was created ...
-  delete iter;
-  db-&gt;ReleaseSnapshot(options.snapshot);
-</pre>
-Note that when a snapshot is no longer needed, it should be released
-using the DB::ReleaseSnapshot interface.  This allows the
-implementation to get rid of state that was being maintained just to
-support reading as of that snapshot.
-<h1>Slice</h1>
-<p>
-The return value of the <code>it->key()</code> and <code>it->value()</code> calls above
-are instances of the <code>rocksdb::Slice</code> type.  <code>Slice</code> is a simple
-structure that contains a length and a pointer to an external byte
-array.  Returning a <code>Slice</code> is a cheaper alternative to returning a
-<code>std::string</code> since we do not need to copy potentially large keys and
-values.  In addition, <code>rocksdb</code> methods do not return null-terminated
-C-style strings since <code>rocksdb</code> keys and values are allowed to
-contain '\0' bytes.
-<p>
-C++ strings and null-terminated C-style strings can be easily converted
-to a Slice:
-<p>
-<pre>
-   rocksdb::Slice s1 = "hello";
-
-   std::string str("world");
-   rocksdb::Slice s2 = str;
-</pre>
-A Slice can be easily converted back to a C++ string:
-<pre>
-   std::string str = s1.ToString();
-   assert(str == std::string("hello"));
-</pre>
-Be careful when using Slices since it is up to the caller to ensure that
-the external byte array into which the Slice points remains live while
-the Slice is in use.  For example, the following is buggy:
-<p>
-<pre>
-   rocksdb::Slice slice;
-   if (...) {
-     std::string str = ...;
-     slice = str;
-   }
-   Use(slice);
-</pre>
-When the <code>if</code> statement goes out of scope, <code>str</code> will be destroyed and the
-backing storage for <code>slice</code> will disappear.
-<p>
-<h1>Comparators</h1>
-<p>
-The preceding examples used the default ordering function for key,
-which orders bytes lexicographically.  You can however supply a custom
-comparator when opening a database.  For example, suppose each
-database key consists of two numbers and we should sort by the first
-number, breaking ties by the second number.  First, define a proper
-subclass of <code>rocksdb::Comparator</code> that expresses these rules:
-<p>
-<pre>
-  class TwoPartComparator : public rocksdb::Comparator {
-   public:
-    // Three-way comparison function:
-    //   if a &lt; b: negative result
-    //   if a &gt; b: positive result
-    //   else: zero result
-    int Compare(const rocksdb::Slice&amp; a, const rocksdb::Slice&amp; b) const {
-      int a1, a2, b1, b2;
-      ParseKey(a, &amp;a1, &amp;a2);
-      ParseKey(b, &amp;b1, &amp;b2);
-      if (a1 &lt; b1) return -1;
-      if (a1 &gt; b1) return +1;
-      if (a2 &lt; b2) return -1;
-      if (a2 &gt; b2) return +1;
-      return 0;
-    }
-
-    // Ignore the following methods for now:
-    const char* Name() const { return "TwoPartComparator"; }
-    void FindShortestSeparator(std::string*, const rocksdb::Slice&amp;) const { }
-    void FindShortSuccessor(std::string*) const { }
-  };
-</pre>
-Now create a database using this custom comparator:
-<p>
-<pre>
-  TwoPartComparator cmp;
-  rocksdb::DB* db;
-  rocksdb::Options options;
-  options.create_if_missing = true;
-  options.comparator = &amp;cmp;
-  rocksdb::Status status = rocksdb::DB::Open(options, "/tmp/testdb", &amp;db);
-  ...
-</pre>
-<h2>Backwards compatibility</h2>
-<p>
-The result of the comparator's <code>Name</code> method is attached to the
-database when it is created, and is checked on every subsequent
-database open.  If the name changes, the <code>rocksdb::DB::Open</code> call will
-fail.  Therefore, change the name if and only if the new key format
-and comparison function are incompatible with existing databases, and
-it is ok to discard the contents of all existing databases.
-<p>
-You can however still gradually evolve your key format over time with
-a little bit of pre-planning.  For example, you could store a version
-number at the end of each key (one byte should suffice for most uses).
-When you wish to switch to a new key format (e.g., adding an optional
-third part to the keys processed by <code>TwoPartComparator</code>),
-(a) keep the same comparator name (b) increment the version number
-for new keys (c) change the comparator function so it uses the
-version numbers found in the keys to decide how to interpret them.
-
-
-<p>
-<h1>MemTable and Table factories</h1>
-<p>
-By default, we keep the data in memory in skiplist memtable and the data
-on disk in a table format described here:
-<a href="https://github.com/facebook/rocksdb/wiki/Rocksdb-Table-Format">
-    RocksDB Table Format</a>.
-<p>
-Since one of the goals of RocksDB is to have
-different parts of the system easily pluggable, we support different
-implementations of both memtable and table format. You can supply
-your own memtable factory by setting <code>Options::memtable_factory</code>
-and your own table factory by setting <code>Options::table_factory</code>.
-For available memtable factories, please refer to
-<code>rocksdb/memtablerep.h</code> and for table factores to
-<code>rocksdb/table.h</code>. These features are both in active development
-and please be wary of any API changes that might break your application
-going forward.
-<p>
-You can also read more about memtables here:
-<a href="https://github.com/facebook/rocksdb/wiki/Rocksdb-Architecture-Guide#memtables">
-Memtables wiki
-</a>
-
-<p>
-<h1>Performance</h1>
-<p>
-Performance can be tuned by changing the default values of the
-types defined in <code>include/rocksdb/options.h</code>.
-
-<p>
-<h2>Block size</h2>
-<p>
-<code>rocksdb</code> groups adjacent keys together into the same block and such a
-block is the unit of transfer to and from persistent storage.  The
-default block size is approximately 4096 uncompressed bytes.
-Applications that mostly do bulk scans over the contents of the
-database may wish to increase this size.  Applications that do a lot
-of point reads of small values may wish to switch to a smaller block
-size if performance measurements indicate an improvement.  There isn't
-much benefit in using blocks smaller than one kilobyte, or larger than
-a few megabytes.  Also note that compression will be more effective
-with larger block sizes. To change block size parameter, use
-<code>Options::block_size</code>.
-<p>
-<h2>Write buffer</h2>
-<p>
-<code>Options::write_buffer_size</code> specifies the amount of data
-to build up in memory before converting to a sorted on-disk file.
-Larger values increase performance, especially during bulk loads.
-Up to max_write_buffer_number write buffers may be held in memory
-at the same time,
-so you may wish to adjust this parameter to control memory usage.
-Also, a larger write buffer will result in a longer recovery time
-the next time the database is opened.
-Related option is
-<code>Options::max_write_buffer_number</code>, which is maximum number
-of write buffers that are built up in memory. The default is 2, so that
-when 1 write buffer is being flushed to storage, new writes can continue
-to the other write buffer.
-<code>Options::min_write_buffer_number_to_merge</code> is the minimum number
-of write buffers that will be merged together before writing to storage.
-If set to 1, then all write buffers are flushed to L0 as individual files and
-this increases read amplification because a get request has to check in all
-of these files. Also, an in-memory merge may result in writing lesser
-data to storage if there are duplicate records in each of these
-individual write buffers.  Default: 1
-<p>
-<h2>Compression</h2>
-<p>
-Each block is individually compressed before being written to
-persistent storage.  Compression is on by default since the default
-compression method is very fast, and is automatically disabled for
-uncompressible data.  In rare cases, applications may want to disable
-compression entirely, but should only do so if benchmarks show a
-performance improvement:
-<p>
-<pre>
-  rocksdb::Options options;
-  options.compression = rocksdb::kNoCompression;
-  ... rocksdb::DB::Open(options, name, ...) ....
-</pre>
-<h2>Cache</h2>
-<p>
-The contents of the database are stored in a set of files in the
-filesystem and each file stores a sequence of compressed blocks.  If
-<code>options.block_cache</code> is non-NULL, it is used to cache frequently
-used uncompressed block contents. If <code>options.block_cache_compressed</code>
-is non-NULL, it is used to cache frequently used compressed blocks. Compressed
-cache is an alternative to OS cache, which also caches compressed blocks. If
-compressed cache is used, the OS cache will be disabled automatically by setting
-<code>options.allow_os_buffer</code> to false.
-<p>
-<pre>
-  #include "rocksdb/cache.h"
-
-  rocksdb::Options options;
-  options.block_cache = rocksdb::NewLRUCache(100 * 1048576);  // 100MB uncompressed cache
-  options.block_cache_compressed = rocksdb::NewLRUCache(100 * 1048576);  // 100MB compressed cache
-  rocksdb::DB* db;
-  rocksdb::DB::Open(options, name, &db);
-  ... use the db ...
-  delete db
-  delete options.block_cache;
-  delete options.block_cache_compressed;
-</pre>
-<p>
-When performing a bulk read, the application may wish to disable
-caching so that the data processed by the bulk read does not end up
-displacing most of the cached contents.  A per-iterator option can be
-used to achieve this:
-<p>
-<pre>
-  rocksdb::ReadOptions options;
-  options.fill_cache = false;
-  rocksdb::Iterator* it = db-&gt;NewIterator(options);
-  for (it-&gt;SeekToFirst(); it-&gt;Valid(); it-&gt;Next()) {
-    ...
-  }
-</pre>
-<p>
-You can also disable block cache by setting <code>options.no_block_cache</code>
-to true.
-<h2>Key Layout</h2>
-<p>
-Note that the unit of disk transfer and caching is a block.  Adjacent
-keys (according to the database sort order) will usually be placed in
-the same block.  Therefore the application can improve its performance
-by placing keys that are accessed together near each other and placing
-infrequently used keys in a separate region of the key space.
-<p>
-For example, suppose we are implementing a simple file system on top
-of <code>rocksdb</code>.  The types of entries we might wish to store are:
-<p>
-<pre>
-   filename -&gt; permission-bits, length, list of file_block_ids
-   file_block_id -&gt; data
-</pre>
-We might want to prefix <code>filename</code> keys with one letter (say '/') and the
-<code>file_block_id</code> keys with a different letter (say '0') so that scans
-over just the metadata do not force us to fetch and cache bulky file
-contents.
-<p>
-<h2>Filters</h2>
-<p>
-Because of the way <code>rocksdb</code> data is organized on disk,
-a single <code>Get()</code> call may involve multiple reads from disk.
-The optional <code>FilterPolicy</code> mechanism can be used to reduce
-the number of disk reads substantially.
-<pre>
-   rocksdb::Options options;
-   options.filter_policy = NewBloomFilter(10);
-   rocksdb::DB* db;
-   rocksdb::DB::Open(options, "/tmp/testdb", &amp;db);
-   ... use the database ...
-   delete db;
-   delete options.filter_policy;
-</pre>
-The preceding code associates a
-<a href="http://en.wikipedia.org/wiki/Bloom_filter">Bloom filter</a>
-based filtering policy with the database.  Bloom filter based
-filtering relies on keeping some number of bits of data in memory per
-key (in this case 10 bits per key since that is the argument we passed
-to NewBloomFilter).  This filter will reduce the number of unnecessary
-disk reads needed for <code>Get()</code> calls by a factor of
-approximately a 100.  Increasing the bits per key will lead to a
-larger reduction at the cost of more memory usage.  We recommend that
-applications whose working set does not fit in memory and that do a
-lot of random reads set a filter policy.
-<p>
-If you are using a custom comparator, you should ensure that the filter
-policy you are using is compatible with your comparator.  For example,
-consider a comparator that ignores trailing spaces when comparing keys.
-<code>NewBloomFilter</code> must not be used with such a comparator.
-Instead, the application should provide a custom filter policy that
-also ignores trailing spaces.  For example:
-<pre>
-  class CustomFilterPolicy : public rocksdb::FilterPolicy {
-   private:
-    FilterPolicy* builtin_policy_;
-   public:
-    CustomFilterPolicy() : builtin_policy_(NewBloomFilter(10)) { }
-    ~CustomFilterPolicy() { delete builtin_policy_; }
-
-    const char* Name() const { return "IgnoreTrailingSpacesFilter"; }
-
-    void CreateFilter(const Slice* keys, int n, std::string* dst) const {
-      // Use builtin bloom filter code after removing trailing spaces
-      std::vector&lt;Slice&gt; trimmed(n);
-      for (int i = 0; i &lt; n; i++) {
-        trimmed[i] = RemoveTrailingSpaces(keys[i]);
-      }
-      return builtin_policy_-&gt;CreateFilter(&amp;trimmed[i], n, dst);
-    }
-
-    bool KeyMayMatch(const Slice& key, const Slice& filter) const {
-      // Use builtin bloom filter code after removing trailing spaces
-      return builtin_policy_-&gt;KeyMayMatch(RemoveTrailingSpaces(key), filter);
-    }
-  };
-</pre>
-<p>
-Advanced applications may provide a filter policy that does not use
-a bloom filter but uses some other mechanism for summarizing a set
-of keys.  See <code>rocksdb/filter_policy.h</code> for detail.
-<p>
-<h1>Checksums</h1>
-<p>
-<code>rocksdb</code> associates checksums with all data it stores in the file system.
-There are two separate controls provided over how aggressively these
-checksums are verified:
-<p>
-<ul>
-<li> <code>ReadOptions::verify_checksums</code> may be set to true to force
-  checksum verification of all data that is read from the file system on
-  behalf of a particular read.  By default, no such verification is
-  done.
-<p>
-<li> <code>Options::paranoid_checks</code> may be set to true before opening a
-  database to make the database implementation raise an error as soon as
-  it detects an internal corruption.  Depending on which portion of the
-  database has been corrupted, the error may be raised when the database
-  is opened, or later by another database operation.  By default,
-  paranoid checking is off so that the database can be used even if
-  parts of its persistent storage have been corrupted.
-<p>
-  If a database is corrupted (perhaps it cannot be opened when
-  paranoid checking is turned on), the <code>rocksdb::RepairDB</code> function
-  may be used to recover as much of the data as possible.
-<p>
-</ul>
-
-<p>
-<h1>Compaction</h1>
-<p>
-You can read more on Compactions here:
-<a href="https://github.com/facebook/rocksdb/wiki/Rocksdb-Architecture-Guide#multi-threaded-compactions">
-    Multi-threaded compactions
-</a>
-<p>
-Here we give overview of the options that impact behavior of Compactions:
-<ul>
-<p>
-<li><code>Options::compaction_style</code> - RocksDB currently supports two
-compaction algorithms - Universal  style and Level style. This option switches
-between the two.  Can be kCompactionStyleUniversal or kCompactionStyleLevel.
-If this is kCompactionStyleUniversal, then you can configure universal style
-parameters with <code>Options::compaction_options_universal</code>.
-<p>
-<li><code>Options::disable_auto_compactions</code> - Disable automatic compactions.
-Manual compactions can still be issued on this database.
-<p>
-<li><code>Options::compaction_filter</code> - Allows an application to modify/delete
-a key-value during background compaction. The client must provide
-compaction_filter_factory if it requires a new compaction filter to be used
-for different compaction processes. Client should specify only one of filter
-or factory.
-<p>
-<li><code>Options::compaction_filter_factory</code> - a factory that provides
-compaction filter objects which allow an application to modify/delete a
-key-value during background compaction.
-</ul>
-<p>
-Other options impacting performance of compactions and when they get triggered
-are:
-<ul>
-<p>
-<li> <code>Options::access_hint_on_compaction_start</code> - Specify the file access
-pattern once a compaction is started. It will be applied to all input files of a compaction. Default: NORMAL
-<p>
-<li> <code>Options::level0_file_num_compaction_trigger</code> -  Number of files to trigger level-0 compaction.
-A negative value means that level-0 compaction will not be triggered by number of files at all.
-<p>
-<li> <code>Options::max_mem_compaction_level</code> -  Maximum level to which a new compacted memtable is pushed if it
-does not create overlap.  We try to push to level 2 to avoid the relatively expensive level 0=>1 compactions and to avoid some
-expensive manifest file operations.  We do not push all the way to the largest level since that can generate a lot of wasted disk
-space if the same key space is being repeatedly overwritten.
-<p>
-<li> <code>Options::target_file_size_base</code> and <code>Options::target_file_size_multiplier</code> -
-Target file size for compaction.  target_file_size_base is per-file size for level-1.
-Target file size for level L can be calculated by target_file_size_base * (target_file_size_multiplier ^ (L-1))
-For example, if target_file_size_base is 2MB and target_file_size_multiplier is 10, then each file on level-1 will
-be 2MB, and each file on level 2 will be 20MB, and each file on level-3 will be 200MB. Default target_file_size_base is 2MB
-and default target_file_size_multiplier is 1.
-<p>
-<li> <code>Options::expanded_compaction_factor</code> -  Maximum number of bytes in all compacted files.  We avoid expanding
-the lower level file set of a compaction if it would make the total compaction cover more than
-(expanded_compaction_factor * targetFileSizeLevel()) many bytes.
-<p>
-<li> <code>Options::source_compaction_factor</code> -    Maximum number of bytes in all source files to be compacted in a
-single compaction run. We avoid picking too many files in the source level so that we do not exceed the total source bytes
-for compaction to exceed (source_compaction_factor * targetFileSizeLevel()) many bytes.
-Default:1, i.e. pick maxfilesize amount of data as the source of a compaction.
-<p>
-<li> <code>Options::max_grandparent_overlap_factor</code> -   Control maximum bytes of overlaps in grandparent (i.e., level+2) before we
-stop building a single file in a level->level+1 compaction.
-<p>
-<li> <code>Options::max_background_compactions</code> - Maximum number of concurrent background jobs, submitted to
-the default LOW priority thread pool
-</ul>
-
-<p>
-You can learn more about all of those options in <code>rocksdb/options.h</code>
-
-<h2> Universal style compaction specific settings</h2>
-<p>
-If you're using Universal style compaction, there is an object <code>CompactionOptionsUniversal</code>
-that hold all the different options for that compaction. The exact definition is in
-<code>rocksdb/universal_compaction.h</code> and you can set it in <code>Options::compaction_options_universal</code>.
-Here we give short overview of options in <code>CompactionOptionsUniversal</code>:
-<ul>
-<p>
-<li> <code>CompactionOptionsUniversal::size_ratio</code> - Percentage flexibility while comparing file size. If the candidate file(s)
-   size is 1% smaller than the next file's size, then include next file into
-   this candidate set.  Default: 1
-<p>
-<li> <code>CompactionOptionsUniversal::min_merge_width</code> - The minimum number of files in a single compaction run. Default: 2
-<p>
-<li> <code>CompactionOptionsUniversal::max_merge_width</code> - The maximum number of files in a single compaction run. Default: UINT_MAX
-<p>
-<li> <code>CompactionOptionsUniversal::max_size_amplification_percent</code> - The size amplification is defined as the amount (in percentage) of
-additional storage needed to store a single byte of data in the database.  For example, a size amplification of 2% means that a database that
-contains 100 bytes of user-data may occupy upto 102 bytes of physical storage. By this definition, a fully compacted database has
-a size amplification of 0%. Rocksdb uses the following heuristic to calculate size amplification: it assumes that all files excluding
-the earliest file contribute to the size amplification.  Default: 200, which means that a 100 byte database could require upto
-300 bytes of storage.
-<p>
-<li> <code>CompactionOptionsUniversal::compression_size_percent</code> - If this option is set to be -1 (the default value), all the output files
-will follow compression type specified.  If this option is not negative, we will try to make sure compressed
-size is just above this value. In normal cases, at least this percentage
-of data will be compressed.
-When we are compacting to a new file, here is the criteria whether
-it needs to be compressed: assuming here are the list of files sorted
-by generation time: [ A1...An B1...Bm C1...Ct ],
-where A1 is the newest and Ct is the oldest, and we are going to compact
-B1...Bm, we calculate the total size of all the files as total_size, as
-well as  the total size of C1...Ct as total_C, the compaction output file
-will be compressed iff total_C / total_size < this percentage
-<p>
-<li> <code>CompactionOptionsUniversal::stop_style</code> - The algorithm used to stop picking files into a single compaction run.
-Can be kCompactionStopStyleSimilarSize (pick files of similar size) or kCompactionStopStyleTotalSize (total size of picked files > next file).
-Default: kCompactionStopStyleTotalSize
-</ul>
-
-<h1>Thread pools</h1>
-<p>
-A thread pool is associated with Env environment object. The client has to create a thread pool by setting the number of background
-threads using method <code>Env::SetBackgroundThreads()</code> defined in <code>rocksdb/env.h</code>.
-We use the thread pool for compactions and memtable flushes.
-Since memtable flushes are in critical code path (stalling memtable flush can stall writes, increasing p99), we suggest
-having two thread pools - with priorities HIGH and LOW. Memtable flushes can be set up to be scheduled on HIGH thread pool.
-There are two options available for configuration of background compactions and flushes:
-<ul>
-<p>
-<li> <code>Options::max_background_compactions</code> - Maximum number of concurrent background jobs,
-submitted to the default LOW priority thread pool
-<p>
-<li> <code>Options::max_background_flushes</code> - Maximum number of concurrent background memtable flush jobs, submitted to
-the HIGH priority thread pool.  By default, all background jobs (major compaction and memtable flush) go
-to the LOW priority pool. If this option is set to a positive number, memtable flush jobs will be submitted to the HIGH priority pool.
-It is important when the same Env is shared by multiple db instances.  Without a separate pool, long running major compaction jobs could
-potentially block memtable flush jobs of other db instances, leading to unnecessary Put stalls.
-</ul>
-<p>
-<pre>
-  #include "rocksdb/env.h"
-  #include "rocksdb/db.h"
-
-  auto env = rocksdb::Env::Default();
-  env->SetBackgroundThreads(2, rocksdb::Env::LOW);
-  env->SetBackgroundThreads(1, rocksdb::Env::HIGH);
-  rocksdb::DB* db;
-  rocksdb::Options options;
-  options.env = env;
-  options.max_background_compactions = 2;
-  options.max_background_flushes = 1;
-  rocksdb::Status status = rocksdb::DB::Open(options, "/tmp/testdb", &amp;db);
-  assert(status.ok());
-  ...
-</pre>
-<h1>Approximate Sizes</h1>
-<p>
-The <code>GetApproximateSizes</code> method can used to get the approximate
-number of bytes of file system space used by one or more key ranges.
-<p>
-<pre>
-   rocksdb::Range ranges[2];
-   ranges[0] = rocksdb::Range("a", "c");
-   ranges[1] = rocksdb::Range("x", "z");
-   uint64_t sizes[2];
-   rocksdb::Status s = db-&gt;GetApproximateSizes(ranges, 2, sizes);
-</pre>
-The preceding call will set <code>sizes[0]</code> to the approximate number of
-bytes of file system space used by the key range <code>[a..c)</code> and
-<code>sizes[1]</code> to the approximate number of bytes used by the key range
-<code>[x..z)</code>.
-<p>
-<h1>Environment</h1>
-<p>
-All file operations (and other operating system calls) issued by the
-<code>rocksdb</code> implementation are routed through a <code>rocksdb::Env</code> object.
-Sophisticated clients may wish to provide their own <code>Env</code>
-implementation to get better control.  For example, an application may
-introduce artificial delays in the file IO paths to limit the impact
-of <code>rocksdb</code> on other activities in the system.
-<p>
-<pre>
-  class SlowEnv : public rocksdb::Env {
-    .. implementation of the Env interface ...
-  };
-
-  SlowEnv env;
-  rocksdb::Options options;
-  options.env = &amp;env;
-  Status s = rocksdb::DB::Open(options, ...);
-</pre>
-<h1>Porting</h1>
-<p>
-<code>rocksdb</code> may be ported to a new platform by providing platform
-specific implementations of the types/methods/functions exported by
-<code>rocksdb/port/port.h</code>.  See <code>rocksdb/port/port_example.h</code> for more
-details.
-<p>
-In addition, the new platform may need a new default <code>rocksdb::Env</code>
-implementation.  See <code>rocksdb/util/env_posix.h</code> for an example.
-
-<h1>Statistics</h1>
-<p>
-To be able to efficiently tune your application, it is always helpful if you
-have access to usage statistics. You can collect those statistics by setting
-<code>Options::table_properties_collectors</code> or
-<code>Options::statistics</code>. For more information, refer to
-<code>rocksdb/table_properties.h</code> and <code>rocksdb/statistics.h</code>.
-These should not add significant overhead to your application and we
-recommend exporting them to other monitoring tools.
-
-<h1>Purging WAL files</h1>
-<p>
-By default, old write-ahead logs are deleted automatically when they fall out
-of scope and application doesn't need them anymore. There are options that
-enable the user to archive the logs and then delete them lazily, either in
-TTL fashion or based on size limit.
-
-The options are <code>Options::WAL_ttl_seconds</code> and
-<code>Options::WAL_size_limit_MB</code>. Here is how they can be used:
-<ul>
-<li>
-<p>
-If both set to 0, logs will be deleted asap and will never get into the archive.
-<li>
-<p>
-If <code>WAL_ttl_seconds</code> is 0 and WAL_size_limit_MB is not 0, WAL
-files will be checked every 10 min and if total size is greater then
-<code>WAL_size_limit_MB</code>, they will be deleted starting with the
-earliest until size_limit is met. All empty files will be deleted.
-<li>
-<p>
-If <code>WAL_ttl_seconds</code> is not 0 and WAL_size_limit_MB is 0, then
-WAL files will be checked every <code>WAL_ttl_seconds / 2</code> and those
-that are older than WAL_ttl_seconds will be deleted.
-<li>
-<p>
-If both are not 0, WAL files will be checked every 10 min and both
-checks will be performed with ttl being first.
-</ul>
-
-<h1>Other Information</h1>
-<p>
-Details about the <code>rocksdb</code> implementation may be found in
-the following documents:
-<ul>
-<li> <a href="https://github.com/facebook/rocksdb/wiki/Rocksdb-Architecture-Guide">
-  RocksDB Architecture Guide</a>
-<li> <a href="https://github.com/facebook/rocksdb/wiki/Rocksdb-Table-Format">
-    Format of an immutable Table file</a>
-<li> <a href="log_format.txt">Format of a log file</a>
-</ul>
-
-</body>
-</html>

doc/log_format.txt

@@ -1,75 +0,0 @@
-The log file contents are a sequence of 32KB blocks.  The only
-exception is that the tail of the file may contain a partial block.
-
-Each block consists of a sequence of records:
-   block := record* trailer?
-   record :=
-	checksum: uint32	// crc32c of type and data[]
-	length: uint16
-	type: uint8		// One of FULL, FIRST, MIDDLE, LAST
-	data: uint8[length]
-
-A record never starts within the last six bytes of a block (since it
-won't fit).  Any leftover bytes here form the trailer, which must
-consist entirely of zero bytes and must be skipped by readers.
-
-Aside: if exactly seven bytes are left in the current block, and a new
-non-zero length record is added, the writer must emit a FIRST record
-(which contains zero bytes of user data) to fill up the trailing seven
-bytes of the block and then emit all of the user data in subsequent
-blocks.
-
-More types may be added in the future.  Some Readers may skip record
-types they do not understand, others may report that some data was
-skipped.
-
-FULL == 1
-FIRST == 2
-MIDDLE == 3
-LAST == 4
-
-The FULL record contains the contents of an entire user record.
-
-FIRST, MIDDLE, LAST are types used for user records that have been
-split into multiple fragments (typically because of block boundaries).
-FIRST is the type of the first fragment of a user record, LAST is the
-type of the last fragment of a user record, and MID is the type of all
-interior fragments of a user record.
-
-Example: consider a sequence of user records:
-   A: length 1000
-   B: length 97270
-   C: length 8000
-A will be stored as a FULL record in the first block.
-
-B will be split into three fragments: first fragment occupies the rest
-of the first block, second fragment occupies the entirety of the
-second block, and the third fragment occupies a prefix of the third
-block.  This will leave six bytes free in the third block, which will
-be left empty as the trailer.
-
-C will be stored as a FULL record in the fourth block.
-
-===================
-
-Some benefits over the recordio format:
-
-(1) We do not need any heuristics for resyncing - just go to next
-block boundary and scan.  If there is a corruption, skip to the next
-block.  As a side-benefit, we do not get confused when part of the
-contents of one log file are embedded as a record inside another log
-file.
-
-(2) Splitting at approximate boundaries (e.g., for mapreduce) is
-simple: find the next block boundary and skip records until we
-hit a FULL or FIRST record.
-
-(3) We do not need extra buffering for large records.
-
-Some downsides compared to recordio format:
-
-(1) No packing of tiny records.  This could be fixed by adding a new
-record type, so it is a shortcoming of the current implementation,
-not necessarily the format.
-
-(2) No compression.  Again, this could be fixed by adding new record types.

docs/.gitignore

@@ -0,0 +1,9 @@
+.DS_STORE
+_site/
+*.swo
+*.swp
+_site
+.sass-cache
+*.psd
+*~
+

docs/Gemfile

@@ -0,0 +1,3 @@
+source 'https://rubygems.org'
+gem 'github-pages', group: :jekyll_plugins
+gem 'jekyll-redirect-from'

docs/Gemfile.lock

@@ -0,0 +1,140 @@
+GEM
+  remote: https://rubygems.org/
+  specs:
+    activesupport (4.2.7)
+      i18n (~> 0.7)
+      json (~> 1.7, >= 1.7.7)
+      minitest (~> 5.1)
+      thread_safe (~> 0.3, >= 0.3.4)
+      tzinfo (~> 1.1)
+    addressable (2.4.0)
+    coffee-script (2.4.1)
+      coffee-script-source
+      execjs
+    coffee-script-source (1.10.0)
+    colorator (1.1.0)
+    ethon (0.9.0)
+      ffi (>= 1.3.0)
+    execjs (2.7.0)
+    faraday (0.9.2)
+      multipart-post (>= 1.2, < 3)
+    ffi (1.9.14)
+    forwardable-extended (2.6.0)
+    gemoji (2.1.0)
+    github-pages (94)
+      activesupport (= 4.2.7)
+      github-pages-health-check (= 1.2.0)
+      jekyll (= 3.2.1)
+      jekyll-coffeescript (= 1.0.1)
+      jekyll-feed (= 0.5.1)
+      jekyll-gist (= 1.4.0)
+      jekyll-github-metadata (= 2.0.2)
+      jekyll-mentions (= 1.2.0)
+      jekyll-paginate (= 1.1.0)
+      jekyll-redirect-from (= 0.11.0)
+      jekyll-sass-converter (= 1.3.0)
+      jekyll-seo-tag (= 2.0.0)
+      jekyll-sitemap (= 0.10.0)
+      jemoji (= 0.7.0)
+      kramdown (= 1.11.1)
+      liquid (= 3.0.6)
+      listen (= 3.0.6)
+      mercenary (~> 0.3)
+      minima (= 1.0.1)
+      rouge (= 1.11.1)
+      terminal-table (~> 1.4)
+    github-pages-health-check (1.2.0)
+      addressable (~> 2.3)
+      net-dns (~> 0.8)
+      octokit (~> 4.0)
+      public_suffix (~> 1.4)
+      typhoeus (~> 0.7)
+    html-pipeline (2.4.2)
+      activesupport (>= 2)
+      nokogiri (>= 1.4)
+    i18n (0.7.0)
+    jekyll (3.2.1)
+      colorator (~> 1.0)
+      jekyll-sass-converter (~> 1.0)
+      jekyll-watch (~> 1.1)
+      kramdown (~> 1.3)
+      liquid (~> 3.0)
+      mercenary (~> 0.3.3)
+      pathutil (~> 0.9)
+      rouge (~> 1.7)
+      safe_yaml (~> 1.0)
+    jekyll-coffeescript (1.0.1)
+      coffee-script (~> 2.2)
+    jekyll-feed (0.5.1)
+    jekyll-gist (1.4.0)
+      octokit (~> 4.2)
+    jekyll-github-metadata (2.0.2)
+      jekyll (~> 3.1)
+      octokit (~> 4.0)
+    jekyll-mentions (1.2.0)
+      activesupport (~> 4.0)
+      html-pipeline (~> 2.3)
+      jekyll (~> 3.0)
+    jekyll-paginate (1.1.0)
+    jekyll-redirect-from (0.11.0)
+      jekyll (>= 2.0)
+    jekyll-sass-converter (1.3.0)
+      sass (~> 3.2)
+    jekyll-seo-tag (2.0.0)
+      jekyll (~> 3.1)
+    jekyll-sitemap (0.10.0)
+    jekyll-watch (1.5.0)
+      listen (~> 3.0, < 3.1)
+    jemoji (0.7.0)
+      activesupport (~> 4.0)
+      gemoji (~> 2.0)
+      html-pipeline (~> 2.2)
+      jekyll (>= 3.0)
+    json (1.8.3)
+    kramdown (1.11.1)
+    liquid (3.0.6)
+    listen (3.0.6)
+      rb-fsevent (>= 0.9.3)
+      rb-inotify (>= 0.9.7)
+    mercenary (0.3.6)
+    mini_portile2 (2.1.0)
+    minima (1.0.1)
+    minitest (5.9.0)
+    multipart-post (2.0.0)
+    net-dns (0.8.0)
+    nokogiri (1.6.8)
+      mini_portile2 (~> 2.1.0)
+      pkg-config (~> 1.1.7)
+    octokit (4.3.0)
+      sawyer (~> 0.7.0, >= 0.5.3)
+    pathutil (0.14.0)
+      forwardable-extended (~> 2.6)
+    pkg-config (1.1.7)
+    public_suffix (1.5.3)
+    rb-fsevent (0.9.7)
+    rb-inotify (0.9.7)
+      ffi (>= 0.5.0)
+    rouge (1.11.1)
+    safe_yaml (1.0.4)
+    sass (3.4.22)
+    sawyer (0.7.0)
+      addressable (>= 2.3.5, < 2.5)
+      faraday (~> 0.8, < 0.10)
+    terminal-table (1.7.0)
+      unicode-display_width (~> 1.1)
+    thread_safe (0.3.5)
+    typhoeus (0.8.0)
+      ethon (>= 0.8.0)
+    tzinfo (1.2.2)
+      thread_safe (~> 0.1)
+    unicode-display_width (1.1.0)
+
+PLATFORMS
+  ruby
+
+DEPENDENCIES
+  github-pages
+  jekyll-redirect-from
+
+BUNDLED WITH
+   1.12.5