Fix "*/" outside comment warning in env_hdfs.h

Summary:
Red diff to remove existing implementation of garbage collection. The current approach is reference counting kind of approach and require a lot of effort to get the size counter right on compaction and deletion. I'm going to go with a simple mark-sweep kind of approach and will send another PR for that.

CompactionEventListener was added solely for blob db and it adds complexity and overhead to compaction iterator. Removing it as well.
Closes https://github.com/facebook/rocksdb/pull/3551

Differential Revision: D7130190

Pulled By: yiwu-arbug

fbshipit-source-id: c3a375ad2639a3f6ed179df6eda602372cc5b8df

HISTORY.md

@@ -1,9 +1,53 @@
 # Rocksdb Change Log
 ## Unreleased
 ### Public API Change
+* RocksDBOptionsParser::Parse()'s `ignore_unknown_options` argument will only be effective if the option file shows it is generated using a higher version of RocksDB than the current version.
+* Remove CompactionEventListener.
+
+### New Features
+* Avoid unnecessarily flushing in `CompactRange()` when the range specified by the user does not overlap unflushed memtables.
+* Add "rocksdb.live-sst-files-size" DB property to return total bytes of all SST files belong to the latest LSM tree.
+
+### Bug Fixes
+* Fix a leak in prepared_section_completed_ where the zeroed entries would not removed from the map.
+
+## 5.12.0 (2/14/2018)
+### Public API Change
 * Iterator::SeekForPrev is now a pure virtual method. This is to prevent user who implement the Iterator interface fail to implement SeekForPrev by mistake.
+* Add `include_end` option to make the range end exclusive when `include_end == false` in `DeleteFilesInRange()`.
+* Add `CompactRangeOptions::allow_write_stall`, which makes `CompactRange` start working immediately, even if it causes user writes to stall. The default value is false, meaning we add delay to `CompactRange` calls until stalling can be avoided when possible. Note this delay is not present in previous RocksDB versions.
+* Creating checkpoint with empty directory now returns `Status::InvalidArgument`; previously, it returned `Status::IOError`.
+* Adds a BlockBasedTableOption to turn off index block compression.
+* Close() method now returns a status when closing a db.
+
+### New Features
+* Improve the performance of iterators doing long range scans by using readahead.
+* Add new function `DeleteFilesInRanges()` to delete files in multiple ranges at once for better performance.
+* FreeBSD build support for RocksDB and RocksJava.
+* Improved performance of long range scans with readahead.
+* Updated to and now continuously tested in Visual Studio 2017.
+
 ### Bug Fixes
 * Fix `DisableFileDeletions()` followed by `GetSortedWalFiles()` to not return obsolete WAL files that `PurgeObsoleteFiles()` is going to delete.
+* Fix Handle error return from WriteBuffer() during WAL file close and DB close.
+* Fix advance reservation of arena block addresses.
+* Fix handling of empty string as checkpoint directory.
+
+## 5.11.0 (01/08/2018)
+### Public API Change
+* Add `autoTune` and `getBytesPerSecond()` to RocksJava RateLimiter
+
+### New Features
+* Add a new histogram stat called rocksdb.db.flush.micros for memtable flush.
+* Add "--use_txn" option to use transactional API in db_stress.
+* Disable onboard cache for compaction output in Windows platform.
+* Improve the performance of iterators doing long range scans by using readahead.
+
+### Bug Fixes
+* Fix a stack-use-after-scope bug in ForwardIterator.
+* Fix builds on platforms including Linux, Windows, and PowerPC.
+* Fix buffer overrun in backup engine for DBs with huge number of files.
+* Fix a mislabel bug for bottom-pri compaction threads.
 * Fix DB::Flush() keep waiting after flush finish under certain condition.
 
 ## 5.10.0 (12/11/2017)

LANGUAGE-BINDINGS.md

@@ -15,3 +15,4 @@ This is the list of all known third-party language bindings for RocksDB. If some
 * D programming language - https://github.com/b1naryth1ef/rocksdb
 * Erlang - https://gitlab.com/barrel-db/erlang-rocksdb
 * Elixir - https://github.com/urbint/rox
+* Nim - https://github.com/status-im/nim-rocksdb

Makefile

@@ -232,9 +232,13 @@ endif
 # USAN doesn't work well with jemalloc. If we're compiling with USAN, we should use regular malloc.
 ifdef COMPILE_WITH_UBSAN
 	DISABLE_JEMALLOC=1
-	EXEC_LDFLAGS += -fsanitize=undefined
-	PLATFORM_CCFLAGS += -fsanitize=undefined -DROCKSDB_UBSAN_RUN
-	PLATFORM_CXXFLAGS += -fsanitize=undefined -DROCKSDB_UBSAN_RUN
+	# Suppress alignment warning because murmurhash relies on casting unaligned
+	# memory to integer. Fixing it may cause performance regression. 3-way crc32
+	# relies on it too, although it can be rewritten to eliminate with minimal
+	# performance regression.
+	EXEC_LDFLAGS += -fsanitize=undefined -fno-sanitize-recover=all
+	PLATFORM_CCFLAGS += -fsanitize=undefined -fno-sanitize-recover=all -DROCKSDB_UBSAN_RUN
+	PLATFORM_CXXFLAGS += -fsanitize=undefined -fno-sanitize-recover=all -DROCKSDB_UBSAN_RUN
 endif
 
 ifdef ROCKSDB_VALGRIND_RUN
@@ -523,6 +527,10 @@ PARALLEL_TEST = \
 	transaction_test \
 	write_prepared_transaction_test \
 
+# options_settable_test doesn't pass with UBSAN as we use hack in the test
+ifdef COMPILE_WITH_UBSAN
+        TESTS := $(shell echo $(TESTS) | sed 's/\boptions_settable_test\b//g')
+endif
 SUBSET := $(TESTS)
 ifdef ROCKSDBTESTS_START
         SUBSET := $(shell echo $(SUBSET) | sed 's/^.*$(ROCKSDBTESTS_START)/$(ROCKSDBTESTS_START)/')

Vagrantfile

@@ -14,6 +14,11 @@ Vagrant.configure("2") do |config|
     box.vm.box = "chef/centos-6.5"
   end
 
+  config.vm.define "centos7" do |box|
+    box.vm.box = "centos/7"
+    box.vm.provision "shell", path: "build_tools/setup_centos7.sh"
+  end
+
   config.vm.define "FreeBSD10" do |box|
     box.vm.guest = :freebsd
     box.vm.box = "robin/freebsd-10"

appveyor.yml

@@ -1,9 +1,9 @@
 version: 1.0.{build}
-image: Visual Studio 2015
+image: Visual Studio 2017
 before_build:
 - md %APPVEYOR_BUILD_FOLDER%\build
 - cd %APPVEYOR_BUILD_FOLDER%\build
-- cmake -G "Visual Studio 14 Win64" -DOPTDBG=1 -DWITH_XPRESS=1 -DPORTABLE=1 ..
+- cmake -G "Visual Studio 15 Win64" -DOPTDBG=1 -DWITH_XPRESS=1 -DPORTABLE=1 ..
 - cd ..
 build:
   project: build\rocksdb.sln

buckifier/buckify_rocksdb.py

@@ -36,7 +36,7 @@ def parse_src_mk(repo_path):
 # get all .cc / .c files
 def get_cc_files(repo_path):
     cc_files = []
-    for root, dirnames, filenames in os.walk(repo_path):
+    for root, dirnames, filenames in os.walk(repo_path):  # noqa: B007 T25377293 Grandfathered in
         root = root[(len(repo_path) + 1):]
         if "java" in root:
             # Skip java

buckifier/rocks_test_runner.sh

@@ -1,4 +1,5 @@
 #!/usr/bin/env bash
 # Create a tmp directory for the test to use
 TEST_DIR=$(mktemp -d /dev/shm/fbcode_rocksdb_XXXXXXX)
+# shellcheck disable=SC2068
 TEST_TMPDIR="$TEST_DIR" $@ && rm -rf "$TEST_DIR"

build_tools/cont_integration.sh

@@ -13,10 +13,12 @@ error=0
 
 function log {
   DATE=`date +%Y-%m-%d:%H:%M:%S`
+  # shellcheck disable=SC2068
   echo $DATE $@
 }
 
 function log_err {
+  # shellcheck disable=SC2145
   log "ERROR: $@ Error code: $error."
 }
 

build_tools/dependencies.sh

@@ -1,3 +1,4 @@
+# shellcheck disable=SC2148
 GCC_BASE=/mnt/gvfs/third-party2/gcc/8219ec1bcedf8ad9da05e121e193364de2cc4f61/5.x/centos6-native/c447969
 CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/64d8d58e3d84f8bde7a029763d4f5baf39d0d5b9/stable/centos6-native/6aaf4de
 LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/ba9be983c81de7299b59fe71950c664a84dcb5f8/5.x/gcc-5-glibc-2.23/339d858

build_tools/dependencies_4.8.1.sh

@@ -1,3 +1,4 @@
+# shellcheck disable=SC2148
 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

build_tools/make_package.sh

@@ -1,3 +1,4 @@
+# shellcheck disable=SC1113
 #/usr/bin/env bash
 
 set -e
@@ -28,12 +29,14 @@ function package() {
     if dpkg --get-selections | grep --quiet $1; then
       log "$1 is already installed. skipping."
     else
+      # shellcheck disable=SC2068
       apt-get install $@ -y
     fi
   elif [[ $OS = "centos" ]]; then
     if rpm -qa | grep --quiet $1; then
       log "$1 is already installed. skipping."
     else
+      # shellcheck disable=SC2068
       yum install $@ -y
     fi
   fi
@@ -52,6 +55,7 @@ function gem_install() {
   if gem list | grep --quiet $1; then
     log "$1 is already installed. skipping."
   else
+    # shellcheck disable=SC2068
     gem install $@
   fi
 }
@@ -125,4 +129,5 @@ function main() {
     include $LIB_DIR
 }
 
+# shellcheck disable=SC2068
 main $@

build_tools/rocksdb-lego-determinator

@@ -352,6 +352,22 @@ LITE_BUILD_COMMANDS="[
     }
 ]"
 
+#
+# Report RocksDB lite binary size to scuba
+REPORT_LITE_BINARY_SIZE_COMMANDS="[
+    {
+        'name':'Rocksdb Lite Binary Size',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Report RocksDB Lite binary size to scuba',
+                'shell':'tools/report_lite_binary_size.sh',
+                'user':'root',
+            },
+        ],
+]"
+
 #
 # RocksDB stress/crash test
 #
@@ -728,6 +744,9 @@ case $1 in
   lite)
     echo $LITE_BUILD_COMMANDS
     ;;
+  report_lite_binary_size)
+    echo $REPORT_LITE_BINARY_SIZE_COMMANDS
+    ;;
   stress_crash)
     echo $STRESS_CRASH_TEST_COMMANDS
     ;;

build_tools/setup_centos7.sh

@@ -0,0 +1,43 @@
+#!/bin/bash
+set -e
+
+ROCKSDB_VERSION="5.10.3"
+ZSTD_VERSION="1.1.3"
+
+echo "This script configures CentOS with everything needed to build and run RocksDB"
+
+yum update -y && yum install epel-release -y
+
+yum install -y \
+  wget \
+  gcc-c++ \
+  snappy snappy-devel \
+  zlib zlib-devel \
+  bzip2 bzip2-devel \
+  lz4-devel \
+  libasan \
+  gflags
+
+mkdir -pv /usr/local/rocksdb-${ROCKSDB_VERSION}
+ln -sfT /usr/local/rocksdb-${ROCKSDB_VERSION} /usr/local/rocksdb
+
+wget -qO /tmp/zstd-${ZSTD_VERSION}.tar.gz https://github.com/facebook/zstd/archive/v${ZSTD_VERSION}.tar.gz
+wget -qO /tmp/rocksdb-${ROCKSDB_VERSION}.tar.gz https://github.com/facebook/rocksdb/archive/v${ROCKSDB_VERSION}.tar.gz
+
+cd /tmp
+
+tar xzvf zstd-${ZSTD_VERSION}.tar.gz
+tar xzvf rocksdb-${ROCKSDB_VERSION}.tar.gz -C /usr/local/
+
+echo "Installing ZSTD..."
+pushd zstd-${ZSTD_VERSION}
+make && make install
+popd
+
+echo "Compiling RocksDB..."
+cd /usr/local/rocksdb
+chown -R vagrant:vagrant /usr/local/rocksdb/
+sudo -u vagrant make static_lib
+cd examples/
+sudo -u vagrant make all
+sudo -u vagrant ./c_simple_example

cache/cache_test.cc

@@ -40,9 +40,9 @@ static int DecodeValue(void* v) {
 const std::string kLRU = "lru";
 const std::string kClock = "clock";
 
-void dumbDeleter(const Slice& key, void* value) {}
+void dumbDeleter(const Slice& /*key*/, void* /*value*/) {}
 
-void eraseDeleter(const Slice& key, void* value) {
+void eraseDeleter(const Slice& /*key*/, void* value) {
   Cache* cache = reinterpret_cast<Cache*>(value);
   cache->Erase("foo");
 }
@@ -470,7 +470,7 @@ class Value {
 };
 
 namespace {
-void deleter(const Slice& key, void* value) {
+void deleter(const Slice& /*key*/, void* value) {
   delete static_cast<Value *>(value);
 }
 }  // namespace

cache/clock_cache.cc

@@ -586,7 +586,7 @@ Status ClockCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
                                size_t charge,
                                void (*deleter)(const Slice& key, void* value),
                                Cache::Handle** out_handle,
-                               Cache::Priority priority) {
+                               Cache::Priority /*priority*/) {
   CleanupContext context;
   HashTable::accessor accessor;
   char* key_data = new char[key.size()];

cache/lru_cache.h

@@ -57,7 +57,7 @@ struct LRUHandle {
   // Include the following flags:
   //   in_cache:    whether this entry is referenced by the hash table.
   //   is_high_pri: whether this entry is high priority entry.
-  //   in_high_pro_pool: whether this entry is in high-pri pool.
+  //   in_high_pri_pool: whether this entry is in high-pri pool.
   char flags;
 
   uint32_t hash;     // Hash of key(); used for fast sharding and comparisons

cache/sharded_cache.cc

@@ -53,7 +53,7 @@ Status ShardedCache::Insert(const Slice& key, void* value, size_t charge,
       ->Insert(key, hash, value, charge, deleter, handle, priority);
 }
 
-Cache::Handle* ShardedCache::Lookup(const Slice& key, Statistics* stats) {
+Cache::Handle* ShardedCache::Lookup(const Slice& key, Statistics* /*stats*/) {
   uint32_t hash = HashSlice(key);
   return GetShard(Shard(hash))->Lookup(key, hash);
 }

coverage/parse_gcov_output.py

@@ -72,7 +72,7 @@ def display_file_coverage(per_file_coverage, total_coverage):
     header_template = \
         "%" + str(max_file_name_length) + "s\t%s\t%s"
     separator = "-" * (max_file_name_length + 10 + 20)
-    print header_template % ("Filename", "Coverage", "Lines")
+    print header_template % ("Filename", "Coverage", "Lines")  # noqa: E999 T25377293 Grandfathered in
     print separator
 
     # -- Print body

db/builder.cc

@@ -61,10 +61,10 @@ TableBuilder* NewTableBuilder(
 
 Status BuildTable(
     const std::string& dbname, Env* env, const ImmutableCFOptions& ioptions,
-    const MutableCFOptions& mutable_cf_options, const EnvOptions& env_options,
-    TableCache* table_cache, InternalIterator* iter,
-    std::unique_ptr<InternalIterator> range_del_iter, FileMetaData* meta,
-    const InternalKeyComparator& internal_comparator,
+    const MutableCFOptions& /*mutable_cf_options*/,
+    const EnvOptions& env_options, TableCache* table_cache,
+    InternalIterator* iter, std::unique_ptr<InternalIterator> range_del_iter,
+    FileMetaData* meta, const InternalKeyComparator& internal_comparator,
     const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
         int_tbl_prop_collector_factories,
     uint32_t column_family_id, const std::string& column_family_name,
@@ -131,7 +131,8 @@ Status BuildTable(
     MergeHelper merge(env, internal_comparator.user_comparator(),
                       ioptions.merge_operator, nullptr, ioptions.info_log,
                       true /* internal key corruption is not ok */,
-                      snapshots.empty() ? 0 : snapshots.back());
+                      snapshots.empty() ? 0 : snapshots.back(),
+                      snapshot_checker);
 
     CompactionIterator c_iter(
         iter, internal_comparator.user_comparator(), &merge, kMaxSequenceNumber,

db/c.cc

@@ -252,7 +252,7 @@ struct rocksdb_comparator_t : public Comparator {
   // No-ops since the C binding does not support key shortening methods.
   virtual void FindShortestSeparator(std::string*,
                                      const Slice&) const override {}
-  virtual void FindShortSuccessor(std::string* key) const override {}
+  virtual void FindShortSuccessor(std::string* /*key*/) const override {}
 };
 
 struct rocksdb_filterpolicy_t : public FilterPolicy {
@@ -367,7 +367,7 @@ struct rocksdb_mergeoperator_t : public MergeOperator {
   virtual bool PartialMergeMulti(const Slice& key,
                                  const std::deque<Slice>& operand_list,
                                  std::string* new_value,
-                                 Logger* logger) const override {
+                                 Logger* /*logger*/) const override {
     size_t operand_count = operand_list.size();
     std::vector<const char*> operand_pointers(operand_count);
     std::vector<size_t> operand_sizes(operand_count);
@@ -2166,8 +2166,8 @@ void rocksdb_options_set_level0_stop_writes_trigger(
   opt->rep.level0_stop_writes_trigger = n;
 }
 
-void rocksdb_options_set_max_mem_compaction_level(rocksdb_options_t* opt,
-                                                  int n) {}
+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);
@@ -2231,8 +2231,8 @@ void rocksdb_options_set_manifest_preallocation_size(
 }
 
 // noop
-void rocksdb_options_set_purge_redundant_kvs_while_flush(rocksdb_options_t* opt,
-                                                         unsigned char v) {}
+void rocksdb_options_set_purge_redundant_kvs_while_flush(
+    rocksdb_options_t* /*opt*/, unsigned char /*v*/) {}
 
 void rocksdb_options_set_use_direct_reads(rocksdb_options_t* opt,
                                           unsigned char v) {
@@ -2402,7 +2402,7 @@ void rocksdb_options_set_table_cache_numshardbits(
 }
 
 void rocksdb_options_set_table_cache_remove_scan_count_limit(
-    rocksdb_options_t* opt, int v) {
+    rocksdb_options_t* /*opt*/, int /*v*/) {
   // this option is deprecated
 }
 
@@ -2973,7 +2973,7 @@ rocksdb_sstfilewriter_t* rocksdb_sstfilewriter_create(
 
 rocksdb_sstfilewriter_t* rocksdb_sstfilewriter_create_with_comparator(
     const rocksdb_envoptions_t* env, const rocksdb_options_t* io_options,
-    const rocksdb_comparator_t* comparator) {
+    const rocksdb_comparator_t* /*comparator*/) {
   rocksdb_sstfilewriter_t* writer = new rocksdb_sstfilewriter_t;
   writer->rep = new SstFileWriter(env->rep, io_options->rep);
   return writer;

db/column_family.cc

@@ -31,6 +31,7 @@
 #include "monitoring/thread_status_util.h"
 #include "options/options_helper.h"
 #include "table/block_based_table_factory.h"
+#include "table/merging_iterator.h"
 #include "util/autovector.h"
 #include "util/compression.h"
 
@@ -344,12 +345,13 @@ void SuperVersionUnrefHandle(void* ptr) {
   // When latter happens, we are in ~ColumnFamilyData(), no get should happen as
   // well.
   SuperVersion* sv = static_cast<SuperVersion*>(ptr);
-  if (sv->Unref()) {
-    sv->db_mutex->Lock();
-    sv->Cleanup();
-    sv->db_mutex->Unlock();
-    delete sv;
-  }
+  bool was_last_ref __attribute__((__unused__));
+  was_last_ref = sv->Unref();
+  // Thread-local SuperVersions can't outlive ColumnFamilyData::super_version_.
+  // This is important because we can't do SuperVersion cleanup here.
+  // That would require locking DB mutex, which would deadlock because
+  // SuperVersionUnrefHandle is called with locked ThreadLocalPtr mutex.
+  assert(!was_last_ref);
 }
 }  // anonymous namespace
 
@@ -381,6 +383,7 @@ ColumnFamilyData::ColumnFamilyData(
       next_(nullptr),
       prev_(nullptr),
       log_number_(0),
+      flush_reason_(FlushReason::kUnknown),
       column_family_set_(column_family_set),
       pending_flush_(false),
       pending_compaction_(false),
@@ -469,7 +472,7 @@ ColumnFamilyData::~ColumnFamilyData() {
     local_sv_.reset();
     super_version_->db_mutex->Lock();
 
-    bool is_last_reference __attribute__((unused));
+    bool is_last_reference __attribute__((__unused__));
     is_last_reference = super_version_->Unref();
     assert(is_last_reference);
     super_version_->Cleanup();
@@ -480,7 +483,7 @@ ColumnFamilyData::~ColumnFamilyData() {
   if (dummy_versions_ != nullptr) {
     // List must be empty
     assert(dummy_versions_->TEST_Next() == dummy_versions_);
-    bool deleted __attribute__((unused));
+    bool deleted __attribute__((__unused__));
     deleted = dummy_versions_->Unref();
     assert(deleted);
   }
@@ -631,6 +634,41 @@ int GetL0ThresholdSpeedupCompaction(int level0_file_num_compaction_trigger,
 }
 }  // namespace
 
+std::pair<WriteStallCondition, ColumnFamilyData::WriteStallCause>
+ColumnFamilyData::GetWriteStallConditionAndCause(
+    int num_unflushed_memtables, int num_l0_files,
+    uint64_t num_compaction_needed_bytes,
+    const MutableCFOptions& mutable_cf_options) {
+  if (num_unflushed_memtables >= mutable_cf_options.max_write_buffer_number) {
+    return {WriteStallCondition::kStopped, WriteStallCause::kMemtableLimit};
+  } else if (!mutable_cf_options.disable_auto_compactions &&
+             num_l0_files >= mutable_cf_options.level0_stop_writes_trigger) {
+    return {WriteStallCondition::kStopped, WriteStallCause::kL0FileCountLimit};
+  } else if (!mutable_cf_options.disable_auto_compactions &&
+             mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
+             num_compaction_needed_bytes >=
+                 mutable_cf_options.hard_pending_compaction_bytes_limit) {
+    return {WriteStallCondition::kStopped,
+            WriteStallCause::kPendingCompactionBytes};
+  } else if (mutable_cf_options.max_write_buffer_number > 3 &&
+             num_unflushed_memtables >=
+                 mutable_cf_options.max_write_buffer_number - 1) {
+    return {WriteStallCondition::kDelayed, WriteStallCause::kMemtableLimit};
+  } else if (!mutable_cf_options.disable_auto_compactions &&
+             mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
+             num_l0_files >=
+                 mutable_cf_options.level0_slowdown_writes_trigger) {
+    return {WriteStallCondition::kDelayed, WriteStallCause::kL0FileCountLimit};
+  } else if (!mutable_cf_options.disable_auto_compactions &&
+             mutable_cf_options.soft_pending_compaction_bytes_limit > 0 &&
+             num_compaction_needed_bytes >=
+                 mutable_cf_options.soft_pending_compaction_bytes_limit) {
+    return {WriteStallCondition::kDelayed,
+            WriteStallCause::kPendingCompactionBytes};
+  }
+  return {WriteStallCondition::kNormal, WriteStallCause::kNone};
+}
+
 WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
       const MutableCFOptions& mutable_cf_options) {
   auto write_stall_condition = WriteStallCondition::kNormal;
@@ -640,25 +678,29 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
     uint64_t compaction_needed_bytes =
         vstorage->estimated_compaction_needed_bytes();
 
+    auto write_stall_condition_and_cause = GetWriteStallConditionAndCause(
+        imm()->NumNotFlushed(), vstorage->l0_delay_trigger_count(),
+        vstorage->estimated_compaction_needed_bytes(), mutable_cf_options);
+    write_stall_condition = write_stall_condition_and_cause.first;
+    auto write_stall_cause = write_stall_condition_and_cause.second;
+
     bool was_stopped = write_controller->IsStopped();
     bool needed_delay = write_controller->NeedsDelay();
 
-    if (imm()->NumNotFlushed() >= mutable_cf_options.max_write_buffer_number) {
+    if (write_stall_condition == WriteStallCondition::kStopped &&
+        write_stall_cause == WriteStallCause::kMemtableLimit) {
       write_controller_token_ = write_controller->GetStopToken();
       internal_stats_->AddCFStats(InternalStats::MEMTABLE_LIMIT_STOPS, 1);
-      write_stall_condition = WriteStallCondition::kStopped;
       ROCKS_LOG_WARN(
           ioptions_.info_log,
           "[%s] Stopping writes because we have %d immutable memtables "
           "(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 (!mutable_cf_options.disable_auto_compactions &&
-               vstorage->l0_delay_trigger_count() >=
-                   mutable_cf_options.level0_stop_writes_trigger) {
+    } else if (write_stall_condition == WriteStallCondition::kStopped &&
+               write_stall_cause == WriteStallCause::kL0FileCountLimit) {
       write_controller_token_ = write_controller->GetStopToken();
       internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_STOPS, 1);
-      write_stall_condition = WriteStallCondition::kStopped;
       if (compaction_picker_->IsLevel0CompactionInProgress()) {
         internal_stats_->AddCFStats(
             InternalStats::LOCKED_L0_FILE_COUNT_LIMIT_STOPS, 1);
@@ -666,28 +708,23 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
       ROCKS_LOG_WARN(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.disable_auto_compactions &&
-               mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
-               compaction_needed_bytes >=
-                   mutable_cf_options.hard_pending_compaction_bytes_limit) {
+    } else if (write_stall_condition == WriteStallCondition::kStopped &&
+               write_stall_cause == WriteStallCause::kPendingCompactionBytes) {
       write_controller_token_ = write_controller->GetStopToken();
       internal_stats_->AddCFStats(
           InternalStats::PENDING_COMPACTION_BYTES_LIMIT_STOPS, 1);
-      write_stall_condition = WriteStallCondition::kStopped;
       ROCKS_LOG_WARN(
           ioptions_.info_log,
           "[%s] Stopping writes because of estimated pending compaction "
           "bytes %" PRIu64,
           name_.c_str(), compaction_needed_bytes);
-    } else if (mutable_cf_options.max_write_buffer_number > 3 &&
-               imm()->NumNotFlushed() >=
-                   mutable_cf_options.max_write_buffer_number - 1) {
+    } else if (write_stall_condition == WriteStallCondition::kDelayed &&
+               write_stall_cause == WriteStallCause::kMemtableLimit) {
       write_controller_token_ =
           SetupDelay(write_controller, compaction_needed_bytes,
                      prev_compaction_needed_bytes_, was_stopped,
                      mutable_cf_options.disable_auto_compactions);
       internal_stats_->AddCFStats(InternalStats::MEMTABLE_LIMIT_SLOWDOWNS, 1);
-      write_stall_condition = WriteStallCondition::kDelayed;
       ROCKS_LOG_WARN(
           ioptions_.info_log,
           "[%s] Stalling writes because we have %d immutable memtables "
@@ -696,10 +733,8 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
           name_.c_str(), imm()->NumNotFlushed(),
           mutable_cf_options.max_write_buffer_number,
           write_controller->delayed_write_rate());
-    } 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) {
+    } else if (write_stall_condition == WriteStallCondition::kDelayed &&
+               write_stall_cause == WriteStallCause::kL0FileCountLimit) {
       // L0 is the last two files from stopping.
       bool near_stop = vstorage->l0_delay_trigger_count() >=
                        mutable_cf_options.level0_stop_writes_trigger - 2;
@@ -709,7 +744,6 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
                      mutable_cf_options.disable_auto_compactions);
       internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_SLOWDOWNS,
                                   1);
-      write_stall_condition = WriteStallCondition::kDelayed;
       if (compaction_picker_->IsLevel0CompactionInProgress()) {
         internal_stats_->AddCFStats(
             InternalStats::LOCKED_L0_FILE_COUNT_LIMIT_SLOWDOWNS, 1);
@@ -719,10 +753,8 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
                      "rate %" PRIu64,
                      name_.c_str(), vstorage->l0_delay_trigger_count(),
                      write_controller->delayed_write_rate());
-    } 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) {
+    } else if (write_stall_condition == WriteStallCondition::kDelayed &&
+               write_stall_cause == WriteStallCause::kPendingCompactionBytes) {
       // If the distance to hard limit is less than 1/4 of the gap between soft
       // and
       // hard bytes limit, we think it is near stop and speed up the slowdown.
@@ -740,7 +772,6 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
                      mutable_cf_options.disable_auto_compactions);
       internal_stats_->AddCFStats(
           InternalStats::PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS, 1);
-      write_stall_condition = WriteStallCondition::kDelayed;
       ROCKS_LOG_WARN(
           ioptions_.info_log,
           "[%s] Stalling writes because of estimated pending compaction "
@@ -748,6 +779,7 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
           name_.c_str(), vstorage->estimated_compaction_needed_bytes(),
           write_controller->delayed_write_rate());
     } else {
+      assert(write_stall_condition == WriteStallCondition::kNormal);
       if (vstorage->l0_delay_trigger_count() >=
           GetL0ThresholdSpeedupCompaction(
               mutable_cf_options.level0_file_num_compaction_trigger,
@@ -814,6 +846,10 @@ uint64_t ColumnFamilyData::GetTotalSstFilesSize() const {
   return VersionSet::GetTotalSstFilesSize(dummy_versions_);
 }
 
+uint64_t ColumnFamilyData::GetLiveSstFilesSize() const {
+  return current_->GetSstFilesSize();
+}
+
 MemTable* ColumnFamilyData::ConstructNewMemtable(
     const MutableCFOptions& mutable_cf_options, SequenceNumber earliest_seq) {
   return new MemTable(internal_comparator_, ioptions_, mutable_cf_options,
@@ -850,6 +886,68 @@ bool ColumnFamilyData::RangeOverlapWithCompaction(
       smallest_user_key, largest_user_key, level);
 }
 
+Status ColumnFamilyData::RangesOverlapWithMemtables(
+    const autovector<Range>& ranges, SuperVersion* super_version,
+    bool* overlap) {
+  assert(overlap != nullptr);
+  *overlap = false;
+  // Create an InternalIterator over all unflushed memtables
+  Arena arena;
+  ReadOptions read_opts;
+  read_opts.total_order_seek = true;
+  MergeIteratorBuilder merge_iter_builder(&internal_comparator_, &arena);
+  merge_iter_builder.AddIterator(
+      super_version->mem->NewIterator(read_opts, &arena));
+  super_version->imm->AddIterators(read_opts, &merge_iter_builder);
+  ScopedArenaIterator memtable_iter(merge_iter_builder.Finish());
+
+  std::vector<InternalIterator*> memtable_range_del_iters;
+  auto* active_range_del_iter =
+      super_version->mem->NewRangeTombstoneIterator(read_opts);
+  if (active_range_del_iter != nullptr) {
+    memtable_range_del_iters.push_back(active_range_del_iter);
+  }
+  super_version->imm->AddRangeTombstoneIterators(read_opts,
+                                                 &memtable_range_del_iters);
+  RangeDelAggregator range_del_agg(internal_comparator_, {} /* snapshots */,
+                                   false /* collapse_deletions */);
+  Status status;
+  {
+    std::unique_ptr<InternalIterator> memtable_range_del_iter(
+        NewMergingIterator(&internal_comparator_,
+                           memtable_range_del_iters.empty()
+                               ? nullptr
+                               : &memtable_range_del_iters[0],
+                           static_cast<int>(memtable_range_del_iters.size())));
+    status = range_del_agg.AddTombstones(std::move(memtable_range_del_iter));
+  }
+  for (size_t i = 0; i < ranges.size() && status.ok() && !*overlap; ++i) {
+    auto* vstorage = super_version->current->storage_info();
+    auto* ucmp = vstorage->InternalComparator()->user_comparator();
+    InternalKey range_start(ranges[i].start, kMaxSequenceNumber,
+                            kValueTypeForSeek);
+    memtable_iter->Seek(range_start.Encode());
+    status = memtable_iter->status();
+    ParsedInternalKey seek_result;
+    if (status.ok()) {
+      if (memtable_iter->Valid() &&
+          !ParseInternalKey(memtable_iter->key(), &seek_result)) {
+        status = Status::Corruption("DB have corrupted keys");
+      }
+    }
+    if (status.ok()) {
+      if (memtable_iter->Valid() &&
+          ucmp->Compare(seek_result.user_key, ranges[i].limit) <= 0) {
+        *overlap = true;
+      } else if (range_del_agg.IsRangeOverlapped(ranges[i].start,
+                                                 ranges[i].limit)) {
+        *overlap = true;
+      }
+    }
+  }
+  return status;
+}
+
 const int ColumnFamilyData::kCompactAllLevels = -1;
 const int ColumnFamilyData::kCompactToBaseLevel = -2;
 
@@ -868,8 +966,7 @@ Compaction* ColumnFamilyData::CompactRange(
 
 SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(
     InstrumentedMutex* db_mutex) {
-  SuperVersion* sv = nullptr;
-  sv = GetThreadLocalSuperVersion(db_mutex);
+  SuperVersion* sv = GetThreadLocalSuperVersion(db_mutex);
   sv->Ref();
   if (!ReturnThreadLocalSuperVersion(sv)) {
     // This Unref() corresponds to the Ref() in GetThreadLocalSuperVersion()
@@ -883,7 +980,6 @@ SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(
 
 SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
     InstrumentedMutex* db_mutex) {
-  SuperVersion* sv = nullptr;
   // The SuperVersion is cached in thread local storage to avoid acquiring
   // mutex when SuperVersion does not change since the last use. When a new
   // SuperVersion is installed, the compaction or flush thread cleans up
@@ -902,7 +998,7 @@ SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
   // should only keep kSVInUse before ReturnThreadLocalSuperVersion call
   // (if no Scrape happens).
   assert(ptr != SuperVersion::kSVInUse);
-  sv = static_cast<SuperVersion*>(ptr);
+  SuperVersion* sv = static_cast<SuperVersion*>(ptr);
   if (sv == SuperVersion::kSVObsolete ||
       sv->version_number != super_version_number_.load()) {
     RecordTick(ioptions_.statistics, NUMBER_SUPERVERSION_ACQUIRES);
@@ -966,6 +1062,12 @@ void ColumnFamilyData::InstallSuperVersion(
       RecalculateWriteStallConditions(mutable_cf_options);
 
   if (old_superversion != nullptr) {
+    // Reset SuperVersions cached in thread local storage.
+    // This should be done before old_superversion->Unref(). That's to ensure
+    // that local_sv_ never holds the last reference to SuperVersion, since
+    // it has no means to safely do SuperVersion cleanup.
+    ResetThreadLocalSuperVersions();
+
     if (old_superversion->mutable_cf_options.write_buffer_size !=
         mutable_cf_options.write_buffer_size) {
       mem_->UpdateWriteBufferSize(mutable_cf_options.write_buffer_size);
@@ -981,9 +1083,6 @@ void ColumnFamilyData::InstallSuperVersion(
       sv_context->superversions_to_free.push_back(old_superversion);
     }
   }
-
-  // Reset SuperVersions cached in thread local storage
-  ResetThreadLocalSuperVersions();
 }
 
 void ColumnFamilyData::ResetThreadLocalSuperVersions() {
@@ -995,10 +1094,12 @@ void ColumnFamilyData::ResetThreadLocalSuperVersions() {
       continue;
     }
     auto sv = static_cast<SuperVersion*>(ptr);
-    if (sv->Unref()) {
-      sv->Cleanup();
-      delete sv;
-    }
+    bool was_last_ref __attribute__((__unused__));
+    was_last_ref = sv->Unref();
+    // sv couldn't have been the last reference because
+    // ResetThreadLocalSuperVersions() is called before
+    // unref'ing super_version_.
+    assert(!was_last_ref);
   }
 }
 
@@ -1060,11 +1161,13 @@ ColumnFamilySet::~ColumnFamilySet() {
   while (column_family_data_.size() > 0) {
     // cfd destructor will delete itself from column_family_data_
     auto cfd = column_family_data_.begin()->second;
-    bool last_ref __attribute__((__unused__)) = cfd->Unref();
+    bool last_ref __attribute__((__unused__));
+    last_ref = cfd->Unref();
     assert(last_ref);
     delete cfd;
   }
-  bool dummy_last_ref __attribute__((__unused__)) = dummy_cfd_->Unref();
+  bool dummy_last_ref __attribute__((__unused__));
+  dummy_last_ref = dummy_cfd_->Unref();
   assert(dummy_last_ref);
   delete dummy_cfd_;
 }

db/column_family.h

@@ -30,6 +30,7 @@ namespace rocksdb {
 
 class Version;
 class VersionSet;
+class VersionStorageInfo;
 class MemTable;
 class MemTableListVersion;
 class CompactionPicker;
@@ -202,6 +203,10 @@ class ColumnFamilyData {
   void SetLogNumber(uint64_t log_number) { log_number_ = log_number; }
   uint64_t GetLogNumber() const { return log_number_; }
 
+  void SetFlushReason(FlushReason flush_reason) {
+    flush_reason_ = flush_reason;
+  }
+  FlushReason GetFlushReason() const { return flush_reason_; }
   // thread-safe
   const EnvOptions* soptions() const;
   const ImmutableCFOptions* ioptions() const { return &ioptions_; }
@@ -239,6 +244,7 @@ class ColumnFamilyData {
   void SetCurrent(Version* _current);
   uint64_t GetNumLiveVersions() const;  // REQUIRE: DB mutex held
   uint64_t GetTotalSstFilesSize() const;  // REQUIRE: DB mutex held
+  uint64_t GetLiveSstFilesSize() const;   // REQUIRE: DB mutex held
   void SetMemtable(MemTable* new_mem) {
     uint64_t memtable_id = last_memtable_id_.fetch_add(1) + 1;
     new_mem->SetID(memtable_id);
@@ -269,6 +275,16 @@ class ColumnFamilyData {
                                   const Slice& largest_user_key,
                                   int level) const;
 
+  // Check if the passed ranges overlap with any unflushed memtables
+  // (immutable or mutable).
+  //
+  // @param super_version A referenced SuperVersion that will be held for the
+  //    duration of this function.
+  //
+  // Thread-safe
+  Status RangesOverlapWithMemtables(const autovector<Range>& ranges,
+                                    SuperVersion* super_version, bool* overlap);
+
   // A flag to tell a manual compaction is to compact all levels together
   // instead of a specific level.
   static const int kCompactAllLevels;
@@ -331,6 +347,17 @@ class ColumnFamilyData {
   bool pending_flush() { return pending_flush_; }
   bool pending_compaction() { return pending_compaction_; }
 
+  enum class WriteStallCause {
+    kNone,
+    kMemtableLimit,
+    kL0FileCountLimit,
+    kPendingCompactionBytes,
+  };
+  static std::pair<WriteStallCondition, WriteStallCause>
+  GetWriteStallConditionAndCause(int num_unflushed_memtables, int num_l0_files,
+                                 uint64_t num_compaction_needed_bytes,
+                                 const MutableCFOptions& mutable_cf_options);
+
   // Recalculate some small conditions, which are changed only during
   // compaction, adding new memtable and/or
   // recalculation of compaction score. These values are used in
@@ -404,6 +431,8 @@ class ColumnFamilyData {
   // recovered from
   uint64_t log_number_;
 
+  std::atomic<FlushReason> flush_reason_;
+
   // An object that keeps all the compaction stats
   // and picks the next compaction
   std::unique_ptr<CompactionPicker> compaction_picker_;

db/column_family_test.cc

@@ -1168,13 +1168,14 @@ TEST_F(ColumnFamilyTest, MemtableNotSupportSnapshot) {
 #endif  // !ROCKSDB_LITE
 
 class TestComparator : public Comparator {
-  int Compare(const rocksdb::Slice& a, const rocksdb::Slice& b) const override {
+  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 {}
+  void FindShortestSeparator(std::string* /*start*/,
+                             const rocksdb::Slice& /*limit*/) const override {}
+  void FindShortSuccessor(std::string* /*key*/) const override {}
 };
 
 static TestComparator third_comparator;

db/compact_files_test.cc

@@ -37,8 +37,7 @@ class FlushedFileCollector : public EventListener {
   FlushedFileCollector() {}
   ~FlushedFileCollector() {}
 
-  virtual void OnFlushCompleted(
-      DB* db, const FlushJobInfo& info) override {
+  virtual void OnFlushCompleted(DB* /*db*/, const FlushJobInfo& info) override {
     std::lock_guard<std::mutex> lock(mutex_);
     flushed_files_.push_back(info.file_path);
   }
@@ -257,9 +256,9 @@ TEST_F(CompactFilesTest, CapturingPendingFiles) {
 TEST_F(CompactFilesTest, CompactionFilterWithGetSv) {
   class FilterWithGet : public CompactionFilter {
    public:
-    virtual bool Filter(int level, const Slice& key, const Slice& value,
-                        std::string* new_value,
-                        bool* value_changed) const override {
+    virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                        const Slice& /*value*/, std::string* /*new_value*/,
+                        bool* /*value_changed*/) const override {
       if (db_ == nullptr) {
         return true;
       }

db/compacted_db_impl.h

@@ -32,55 +32,56 @@ class CompactedDBImpl : public DBImpl {
     override;
 
   using DBImpl::Put;
-  virtual Status Put(const WriteOptions& options,
-                     ColumnFamilyHandle* column_family, const Slice& key,
-                     const Slice& value) override {
+  virtual Status Put(const WriteOptions& /*options*/,
+                     ColumnFamilyHandle* /*column_family*/,
+                     const Slice& /*key*/, const Slice& /*value*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
   using DBImpl::Merge;
-  virtual Status Merge(const WriteOptions& options,
-                       ColumnFamilyHandle* column_family, const Slice& key,
-                       const Slice& value) override {
+  virtual Status Merge(const WriteOptions& /*options*/,
+                       ColumnFamilyHandle* /*column_family*/,
+                       const Slice& /*key*/, const Slice& /*value*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
   using DBImpl::Delete;
-  virtual Status Delete(const WriteOptions& options,
-                        ColumnFamilyHandle* column_family,
-                        const Slice& key) override {
+  virtual Status Delete(const WriteOptions& /*options*/,
+                        ColumnFamilyHandle* /*column_family*/,
+                        const Slice& /*key*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
-  virtual Status Write(const WriteOptions& options,
-                       WriteBatch* updates) override {
+  virtual Status Write(const WriteOptions& /*options*/,
+                       WriteBatch* /*updates*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
   using DBImpl::CompactRange;
-  virtual Status CompactRange(const CompactRangeOptions& options,
-                              ColumnFamilyHandle* column_family,
-                              const Slice* begin, const Slice* end) override {
+  virtual Status CompactRange(const CompactRangeOptions& /*options*/,
+                              ColumnFamilyHandle* /*column_family*/,
+                              const Slice* /*begin*/,
+                              const Slice* /*end*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
 
   virtual Status DisableFileDeletions() override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
-  virtual Status EnableFileDeletions(bool force) override {
+  virtual Status EnableFileDeletions(bool /*force*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
   virtual Status GetLiveFiles(std::vector<std::string>&,
-                              uint64_t* manifest_file_size,
-                              bool flush_memtable = true) override {
+                              uint64_t* /*manifest_file_size*/,
+                              bool /*flush_memtable*/ = true) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
   using DBImpl::Flush;
-  virtual Status Flush(const FlushOptions& options,
-                       ColumnFamilyHandle* column_family) override {
+  virtual Status Flush(const FlushOptions& /*options*/,
+                       ColumnFamilyHandle* /*column_family*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
   using DB::IngestExternalFile;
   virtual Status IngestExternalFile(
-      ColumnFamilyHandle* column_family,
-      const std::vector<std::string>& external_files,
-      const IngestExternalFileOptions& ingestion_options) override {
+      ColumnFamilyHandle* /*column_family*/,
+      const std::vector<std::string>& /*external_files*/,
+      const IngestExternalFileOptions& /*ingestion_options*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
 

db/compaction_iterator.cc

@@ -12,31 +12,6 @@
 
 namespace rocksdb {
 
-#ifndef ROCKSDB_LITE
-CompactionEventListener::CompactionListenerValueType fromInternalValueType(
-    ValueType vt) {
-  switch (vt) {
-    case kTypeDeletion:
-      return CompactionEventListener::CompactionListenerValueType::kDelete;
-    case kTypeValue:
-      return CompactionEventListener::CompactionListenerValueType::kValue;
-    case kTypeMerge:
-      return CompactionEventListener::CompactionListenerValueType::
-          kMergeOperand;
-    case kTypeSingleDeletion:
-      return CompactionEventListener::CompactionListenerValueType::
-          kSingleDelete;
-    case kTypeRangeDeletion:
-      return CompactionEventListener::CompactionListenerValueType::kRangeDelete;
-    case kTypeBlobIndex:
-      return CompactionEventListener::CompactionListenerValueType::kBlobIndex;
-    default:
-      assert(false);
-      return CompactionEventListener::CompactionListenerValueType::kInvalid;
-  }
-}
-#endif  // ROCKSDB_LITE
-
 CompactionIterator::CompactionIterator(
     InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
     SequenceNumber last_sequence, std::vector<SequenceNumber>* snapshots,
@@ -44,7 +19,6 @@ CompactionIterator::CompactionIterator(
     const SnapshotChecker* snapshot_checker, Env* env,
     bool expect_valid_internal_key, RangeDelAggregator* range_del_agg,
     const Compaction* compaction, const CompactionFilter* compaction_filter,
-    CompactionEventListener* compaction_listener,
     const std::atomic<bool>* shutting_down,
     const SequenceNumber preserve_deletes_seqnum)
     : CompactionIterator(
@@ -53,21 +27,18 @@ CompactionIterator::CompactionIterator(
           expect_valid_internal_key, range_del_agg,
           std::unique_ptr<CompactionProxy>(
               compaction ? new CompactionProxy(compaction) : nullptr),
-          compaction_filter, compaction_listener, shutting_down,
-          preserve_deletes_seqnum) {}
+          compaction_filter, shutting_down, preserve_deletes_seqnum) {}
 
 CompactionIterator::CompactionIterator(
     InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
-    SequenceNumber last_sequence, std::vector<SequenceNumber>* snapshots,
+    SequenceNumber /*last_sequence*/, std::vector<SequenceNumber>* snapshots,
     SequenceNumber earliest_write_conflict_snapshot,
     const SnapshotChecker* snapshot_checker, Env* env,
     bool expect_valid_internal_key, RangeDelAggregator* range_del_agg,
     std::unique_ptr<CompactionProxy> compaction,
     const CompactionFilter* compaction_filter,
-    CompactionEventListener* compaction_listener,
     const std::atomic<bool>* shutting_down,
-    const SequenceNumber preserve_deletes_seqnum
-  )
+    const SequenceNumber preserve_deletes_seqnum)
     : input_(input),
       cmp_(cmp),
       merge_helper_(merge_helper),
@@ -79,9 +50,6 @@ CompactionIterator::CompactionIterator(
       range_del_agg_(range_del_agg),
       compaction_(std::move(compaction)),
       compaction_filter_(compaction_filter),
-#ifndef ROCKSDB_LITE
-      compaction_listener_(compaction_listener),
-#endif  // ROCKSDB_LITE
       shutting_down_(shutting_down),
       preserve_deletes_seqnum_(preserve_deletes_seqnum),
       ignore_snapshots_(false),
@@ -293,28 +261,12 @@ void CompactionIterator::NextFromInput() {
           (snapshot_checker_ == nullptr ||
            snapshot_checker_->IsInSnapshot(ikey_.sequence, kMaxSequenceNumber));
 
-#ifndef ROCKSDB_LITE
-      if (compaction_listener_) {
-        compaction_listener_->OnCompaction(compaction_->level(), ikey_.user_key,
-                                           fromInternalValueType(ikey_.type),
-                                           value_, ikey_.sequence, true);
-      }
-#endif  // !ROCKSDB_LITE
-
       // Apply the compaction filter to the first committed version of the user
       // key.
       if (current_key_committed_) {
         InvokeFilterIfNeeded(&need_skip, &skip_until);
       }
     } else {
-#ifndef ROCKSDB_LITE
-      if (compaction_listener_) {
-        compaction_listener_->OnCompaction(compaction_->level(), ikey_.user_key,
-                                           fromInternalValueType(ikey_.type),
-                                           value_, ikey_.sequence, false);
-      }
-#endif  // ROCKSDB_LITE
-
       // Update the current key to reflect the new sequence number/type without
       // copying the user key.
       // TODO(rven): Compaction filter does not process keys in this path
@@ -537,6 +489,11 @@ void CompactionIterator::NextFromInput() {
       //
       // Note:  Dropping this Delete will not affect TransactionDB
       // write-conflict checking since it is earlier than any snapshot.
+      //
+      // It seems that we can also drop deletion later than earliest snapshot
+      // given that:
+      // (1) The deletion is earlier than earliest_write_conflict_snapshot, and
+      // (2) No value exist earlier than the deletion.
       ++iter_stats_.num_record_drop_obsolete;
       if (!bottommost_level_) {
         ++iter_stats_.num_optimized_del_drop_obsolete;
@@ -554,10 +511,6 @@ void CompactionIterator::NextFromInput() {
       // have hit (A)
       // We encapsulate the merge related state machine in a different
       // object to minimize change to the existing flow.
-      // In case snapshot_checker is present, we can probably merge further
-      // beyond prev_snapshot, since there could be more keys with sequence
-      // smaller than prev_snapshot, but reported by snapshot_checker as not
-      // visible by prev_snapshot. But it will make the logic more complicated.
       Status s = merge_helper_->MergeUntil(input_, range_del_agg_,
                                            prev_snapshot, bottommost_level_);
       merge_out_iter_.SeekToFirst();
@@ -621,9 +574,12 @@ void CompactionIterator::PrepareOutput() {
   // and the earliest snapshot is larger than this seqno
   // and the userkey differs from the last userkey in compaction
   // then we can squash the seqno to zero.
-
+  //
   // This is safe for TransactionDB write-conflict checking since transactions
   // only care about sequence number larger than any active snapshots.
+  //
+  // Can we do the same for levels above bottom level as long as
+  // KeyNotExistsBeyondOutputLevel() return true?
   if ((compaction_ != nullptr &&
       !compaction_->allow_ingest_behind()) &&
       ikeyNotNeededForIncrementalSnapshot() &&

db/compaction_iterator.h

@@ -20,8 +20,6 @@
 
 namespace rocksdb {
 
-class CompactionEventListener;
-
 class CompactionIterator {
  public:
   // A wrapper around Compaction. Has a much smaller interface, only what
@@ -32,7 +30,7 @@ class CompactionIterator {
         : compaction_(compaction) {}
 
     virtual ~CompactionProxy() = default;
-    virtual int level(size_t compaction_input_level = 0) const {
+    virtual int level(size_t /*compaction_input_level*/ = 0) const {
       return compaction_->level();
     }
     virtual bool KeyNotExistsBeyondOutputLevel(
@@ -69,7 +67,6 @@ class CompactionIterator {
                      RangeDelAggregator* range_del_agg,
                      const Compaction* compaction = nullptr,
                      const CompactionFilter* compaction_filter = nullptr,
-                     CompactionEventListener* compaction_listener = nullptr,
                      const std::atomic<bool>* shutting_down = nullptr,
                      const SequenceNumber preserve_deletes_seqnum = 0);
 
@@ -83,7 +80,6 @@ class CompactionIterator {
                      RangeDelAggregator* range_del_agg,
                      std::unique_ptr<CompactionProxy> compaction,
                      const CompactionFilter* compaction_filter = nullptr,
-                     CompactionEventListener* compaction_listener = nullptr,
                      const std::atomic<bool>* shutting_down = nullptr,
                      const SequenceNumber preserve_deletes_seqnum = 0);
 
@@ -147,9 +143,6 @@ class CompactionIterator {
   RangeDelAggregator* range_del_agg_;
   std::unique_ptr<CompactionProxy> compaction_;
   const CompactionFilter* compaction_filter_;
-#ifndef ROCKSDB_LITE
-  CompactionEventListener* compaction_listener_;
-#endif  // !ROCKSDB_LITE
   const std::atomic<bool>* shutting_down_;
   const SequenceNumber preserve_deletes_seqnum_;
   bool bottommost_level_;

db/compaction_iterator_test.cc

@@ -9,23 +9,25 @@
 #include <vector>
 
 #include "port/port.h"
+#include "util/string_util.h"
 #include "util/testharness.h"
 #include "util/testutil.h"
+#include "utilities/merge_operators.h"
 
 namespace rocksdb {
 
 // Expects no merging attempts.
 class NoMergingMergeOp : public MergeOperator {
  public:
-  bool FullMergeV2(const MergeOperationInput& merge_in,
-                   MergeOperationOutput* merge_out) const override {
+  bool FullMergeV2(const MergeOperationInput& /*merge_in*/,
+                   MergeOperationOutput* /*merge_out*/) const override {
     ADD_FAILURE();
     return false;
   }
-  bool PartialMergeMulti(const Slice& key,
-                         const std::deque<Slice>& operand_list,
-                         std::string* new_value,
-                         Logger* logger) const override {
+  bool PartialMergeMulti(const Slice& /*key*/,
+                         const std::deque<Slice>& /*operand_list*/,
+                         std::string* /*new_value*/,
+                         Logger* /*logger*/) const override {
     ADD_FAILURE();
     return false;
   }
@@ -39,9 +41,9 @@ class NoMergingMergeOp : public MergeOperator {
 // Always returns Decition::kRemove.
 class StallingFilter : public CompactionFilter {
  public:
-  virtual Decision FilterV2(int level, const Slice& key, ValueType t,
-                            const Slice& existing_value, std::string* new_value,
-                            std::string* skip_until) const override {
+  Decision FilterV2(int /*level*/, const Slice& key, ValueType /*type*/,
+                    const Slice& /*existing_value*/, std::string* /*new_value*/,
+                    std::string* /*skip_until*/) const override {
     int k = std::atoi(key.ToString().c_str());
     last_seen.store(k);
     while (k >= stall_at.load()) {
@@ -72,6 +74,18 @@ class StallingFilter : public CompactionFilter {
   mutable std::atomic<int> last_seen{0};
 };
 
+// Compaction filter that filter out all keys.
+class FilterAllKeysCompactionFilter : public CompactionFilter {
+ public:
+  Decision FilterV2(int /*level*/, const Slice& /*key*/, ValueType /*type*/,
+                    const Slice& /*existing_value*/, std::string* /*new_value*/,
+                    std::string* /*skip_until*/) const override {
+    return Decision::kRemove;
+  }
+
+  const char* Name() const override { return "AllKeysCompactionFilter"; }
+};
+
 class LoggingForwardVectorIterator : public InternalIterator {
  public:
   struct Action {
@@ -112,7 +126,7 @@ class LoggingForwardVectorIterator : public InternalIterator {
                keys_.begin();
   }
 
-  virtual void SeekForPrev(const Slice& target) override { assert(false); }
+  virtual void SeekForPrev(const Slice& /*target*/) override { assert(false); }
 
   virtual void Next() override {
     assert(Valid());
@@ -144,76 +158,149 @@ class FakeCompaction : public CompactionIterator::CompactionProxy {
  public:
   FakeCompaction() = default;
 
-  virtual int level(size_t compaction_input_level) const { return 0; }
-  virtual bool KeyNotExistsBeyondOutputLevel(
-      const Slice& user_key, std::vector<size_t>* level_ptrs) const {
-    return key_not_exists_beyond_output_level;
+  virtual int level(size_t /*compaction_input_level*/) const override {
+    return 0;
   }
-  virtual bool bottommost_level() const { return false; }
-  virtual int number_levels() const { return 1; }
-  virtual Slice GetLargestUserKey() const {
+  virtual bool KeyNotExistsBeyondOutputLevel(
+      const Slice& /*user_key*/,
+      std::vector<size_t>* /*level_ptrs*/) const override {
+    return is_bottommost_level || key_not_exists_beyond_output_level;
+  }
+  virtual bool bottommost_level() const override { return is_bottommost_level; }
+  virtual int number_levels() const override { return 1; }
+  virtual Slice GetLargestUserKey() const override {
     return "\xff\xff\xff\xff\xff\xff\xff\xff\xff";
   }
-  virtual bool allow_ingest_behind() const { return false; }
+  virtual bool allow_ingest_behind() const override { return false; }
 
-  virtual bool preserve_deletes() const {return false; }
+  virtual bool preserve_deletes() const override { return false; }
 
   bool key_not_exists_beyond_output_level = false;
+
+  bool is_bottommost_level = false;
 };
 
-class CompactionIteratorTest : public testing::Test {
+// A simplifed snapshot checker which assumes each snapshot has a global
+// last visible sequence.
+class TestSnapshotChecker : public SnapshotChecker {
+ public:
+  explicit TestSnapshotChecker(
+      SequenceNumber last_committed_sequence,
+      const std::unordered_map<SequenceNumber, SequenceNumber>& snapshots = {})
+      : last_committed_sequence_(last_committed_sequence),
+        snapshots_(snapshots) {}
+
+  bool IsInSnapshot(SequenceNumber seq,
+                    SequenceNumber snapshot_seq) const override {
+    if (snapshot_seq == kMaxSequenceNumber) {
+      return seq <= last_committed_sequence_;
+    }
+    assert(snapshots_.count(snapshot_seq) > 0);
+    return seq <= snapshots_.at(snapshot_seq);
+  }
+
+ private:
+  SequenceNumber last_committed_sequence_;
+  // A map of valid snapshot to last visible sequence to the snapshot.
+  std::unordered_map<SequenceNumber, SequenceNumber> snapshots_;
+};
+
+// Test param:
+//   bool: whether to pass snapshot_checker to compaction iterator.
+class CompactionIteratorTest : public testing::TestWithParam<bool> {
  public:
   CompactionIteratorTest()
       : cmp_(BytewiseComparator()), icmp_(cmp_), snapshots_({}) {}
 
-  void InitIterators(const std::vector<std::string>& ks,
-                     const std::vector<std::string>& vs,
-                     const std::vector<std::string>& range_del_ks,
-                     const std::vector<std::string>& range_del_vs,
-                     SequenceNumber last_sequence,
-                     MergeOperator* merge_op = nullptr,
-                     CompactionFilter* filter = nullptr) {
+  void InitIterators(
+      const std::vector<std::string>& ks, const std::vector<std::string>& vs,
+      const std::vector<std::string>& range_del_ks,
+      const std::vector<std::string>& range_del_vs,
+      SequenceNumber last_sequence,
+      SequenceNumber last_committed_sequence = kMaxSequenceNumber,
+      MergeOperator* merge_op = nullptr, CompactionFilter* filter = nullptr,
+      bool bottommost_level = false,
+      SequenceNumber earliest_write_conflict_snapshot = kMaxSequenceNumber) {
     std::unique_ptr<InternalIterator> range_del_iter(
         new test::VectorIterator(range_del_ks, range_del_vs));
     range_del_agg_.reset(new RangeDelAggregator(icmp_, snapshots_));
     ASSERT_OK(range_del_agg_->AddTombstones(std::move(range_del_iter)));
 
     std::unique_ptr<CompactionIterator::CompactionProxy> compaction;
-    if (filter) {
+    if (filter || bottommost_level) {
       compaction_proxy_ = new FakeCompaction();
+      compaction_proxy_->is_bottommost_level = bottommost_level;
       compaction.reset(compaction_proxy_);
     }
-    // TODO(yiwu) add a mock snapshot checker and add test for it.
-    SnapshotChecker* snapshot_checker = nullptr;
+    bool use_snapshot_checker = UseSnapshotChecker() || GetParam();
+    if (use_snapshot_checker || last_committed_sequence < kMaxSequenceNumber) {
+      snapshot_checker_.reset(
+          new TestSnapshotChecker(last_committed_sequence, snapshot_map_));
+    }
+    merge_helper_.reset(
+        new MergeHelper(Env::Default(), cmp_, merge_op, filter, nullptr, false,
+                        0 /*latest_snapshot*/, snapshot_checker_.get(),
+                        0 /*level*/, nullptr /*statistics*/, &shutting_down_));
 
-    merge_helper_.reset(new MergeHelper(Env::Default(), cmp_, merge_op, filter,
-                                        nullptr, false, 0, 0, nullptr,
-                                        &shutting_down_));
     iter_.reset(new LoggingForwardVectorIterator(ks, vs));
     iter_->SeekToFirst();
     c_iter_.reset(new CompactionIterator(
         iter_.get(), cmp_, merge_helper_.get(), last_sequence, &snapshots_,
-        kMaxSequenceNumber, snapshot_checker, Env::Default(), false,
-        range_del_agg_.get(), std::move(compaction), filter, nullptr,
-        &shutting_down_));
+        earliest_write_conflict_snapshot, snapshot_checker_.get(),
+        Env::Default(), false, range_del_agg_.get(), std::move(compaction),
+        filter, &shutting_down_));
   }
 
-  void AddSnapshot(SequenceNumber snapshot) { snapshots_.push_back(snapshot); }
+  void AddSnapshot(SequenceNumber snapshot,
+                   SequenceNumber last_visible_seq = kMaxSequenceNumber) {
+    snapshots_.push_back(snapshot);
+    snapshot_map_[snapshot] = last_visible_seq;
+  }
+
+  virtual bool UseSnapshotChecker() const { return false; }
+
+  void RunTest(
+      const std::vector<std::string>& input_keys,
+      const std::vector<std::string>& input_values,
+      const std::vector<std::string>& expected_keys,
+      const std::vector<std::string>& expected_values,
+      SequenceNumber last_committed_seq = kMaxSequenceNumber,
+      MergeOperator* merge_operator = nullptr,
+      CompactionFilter* compaction_filter = nullptr,
+      bool bottommost_level = false,
+      SequenceNumber earliest_write_conflict_snapshot = kMaxSequenceNumber) {
+    InitIterators(input_keys, input_values, {}, {}, kMaxSequenceNumber,
+                  last_committed_seq, merge_operator, compaction_filter,
+                  bottommost_level, earliest_write_conflict_snapshot);
+    c_iter_->SeekToFirst();
+    for (size_t i = 0; i < expected_keys.size(); i++) {
+      std::string info = "i = " + ToString(i);
+      ASSERT_TRUE(c_iter_->Valid()) << info;
+      ASSERT_OK(c_iter_->status()) << info;
+      ASSERT_EQ(expected_keys[i], c_iter_->key().ToString()) << info;
+      ASSERT_EQ(expected_values[i], c_iter_->value().ToString()) << info;
+      c_iter_->Next();
+    }
+    ASSERT_FALSE(c_iter_->Valid());
+  }
 
   const Comparator* cmp_;
   const InternalKeyComparator icmp_;
   std::vector<SequenceNumber> snapshots_;
+  // A map of valid snapshot to last visible sequence to the snapshot.
+  std::unordered_map<SequenceNumber, SequenceNumber> snapshot_map_;
   std::unique_ptr<MergeHelper> merge_helper_;
   std::unique_ptr<LoggingForwardVectorIterator> iter_;
   std::unique_ptr<CompactionIterator> c_iter_;
   std::unique_ptr<RangeDelAggregator> range_del_agg_;
+  std::unique_ptr<SnapshotChecker> snapshot_checker_;
   std::atomic<bool> shutting_down_{false};
   FakeCompaction* compaction_proxy_;
 };
 
 // It is possible that the output of the compaction iterator is empty even if
 // the input is not.
-TEST_F(CompactionIteratorTest, EmptyResult) {
+TEST_P(CompactionIteratorTest, EmptyResult) {
   InitIterators({test::KeyStr("a", 5, kTypeSingleDeletion),
                  test::KeyStr("a", 3, kTypeValue)},
                 {"", "val"}, {}, {}, 5);
@@ -223,7 +310,7 @@ TEST_F(CompactionIteratorTest, EmptyResult) {
 
 // If there is a corruption after a single deletion, the corrupted key should
 // be preserved.
-TEST_F(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
+TEST_P(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
   InitIterators({test::KeyStr("a", 5, kTypeSingleDeletion),
                  test::KeyStr("a", 3, kTypeValue, true),
                  test::KeyStr("b", 10, kTypeValue)},
@@ -242,7 +329,7 @@ TEST_F(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
   ASSERT_FALSE(c_iter_->Valid());
 }
 
-TEST_F(CompactionIteratorTest, SimpleRangeDeletion) {
+TEST_P(CompactionIteratorTest, SimpleRangeDeletion) {
   InitIterators({test::KeyStr("morning", 5, kTypeValue),
                  test::KeyStr("morning", 2, kTypeValue),
                  test::KeyStr("night", 3, kTypeValue)},
@@ -258,7 +345,7 @@ TEST_F(CompactionIteratorTest, SimpleRangeDeletion) {
   ASSERT_FALSE(c_iter_->Valid());
 }
 
-TEST_F(CompactionIteratorTest, RangeDeletionWithSnapshots) {
+TEST_P(CompactionIteratorTest, RangeDeletionWithSnapshots) {
   AddSnapshot(10);
   std::vector<std::string> ks1;
   ks1.push_back(test::KeyStr("ma", 28, kTypeRangeDeletion));
@@ -279,11 +366,11 @@ TEST_F(CompactionIteratorTest, RangeDeletionWithSnapshots) {
   ASSERT_FALSE(c_iter_->Valid());
 }
 
-TEST_F(CompactionIteratorTest, CompactionFilterSkipUntil) {
+TEST_P(CompactionIteratorTest, CompactionFilterSkipUntil) {
   class Filter : public CompactionFilter {
-    virtual Decision FilterV2(int level, const Slice& key, ValueType t,
+    virtual Decision FilterV2(int /*level*/, const Slice& key, ValueType t,
                               const Slice& existing_value,
-                              std::string* new_value,
+                              std::string* /*new_value*/,
                               std::string* skip_until) const override {
       std::string k = key.ToString();
       std::string v = existing_value.ToString();
@@ -354,7 +441,7 @@ TEST_F(CompactionIteratorTest, CompactionFilterSkipUntil) {
        test::KeyStr("j", 99, kTypeValue)},
       {"av50", "am45", "bv60", "bv40", "cv35", "dm70", "em71", "fm65", "fm30",
        "fv25", "gv90", "hv91", "im95", "jv99"},
-      {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+      {}, {}, kMaxSequenceNumber, kMaxSequenceNumber, &merge_op, &filter);
 
   // Compaction should output just "a", "e" and "h" keys.
   c_iter_->SeekToFirst();
@@ -389,13 +476,14 @@ TEST_F(CompactionIteratorTest, CompactionFilterSkipUntil) {
   ASSERT_EQ(expected_actions, iter_->log);
 }
 
-TEST_F(CompactionIteratorTest, ShuttingDownInFilter) {
+TEST_P(CompactionIteratorTest, ShuttingDownInFilter) {
   NoMergingMergeOp merge_op;
   StallingFilter filter;
   InitIterators(
       {test::KeyStr("1", 1, kTypeValue), test::KeyStr("2", 2, kTypeValue),
        test::KeyStr("3", 3, kTypeValue), test::KeyStr("4", 4, kTypeValue)},
-      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, kMaxSequenceNumber,
+      &merge_op, &filter);
   // Don't leave tombstones (kTypeDeletion) for filtered keys.
   compaction_proxy_->key_not_exists_beyond_output_level = true;
 
@@ -426,13 +514,14 @@ TEST_F(CompactionIteratorTest, ShuttingDownInFilter) {
 
 // Same as ShuttingDownInFilter, but shutdown happens during filter call for
 // a merge operand, not for a value.
-TEST_F(CompactionIteratorTest, ShuttingDownInMerge) {
+TEST_P(CompactionIteratorTest, ShuttingDownInMerge) {
   NoMergingMergeOp merge_op;
   StallingFilter filter;
   InitIterators(
       {test::KeyStr("1", 1, kTypeValue), test::KeyStr("2", 2, kTypeMerge),
        test::KeyStr("3", 3, kTypeMerge), test::KeyStr("4", 4, kTypeValue)},
-      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, kMaxSequenceNumber,
+      &merge_op, &filter);
   compaction_proxy_->key_not_exists_beyond_output_level = true;
 
   std::atomic<bool> seek_done{false};
@@ -460,12 +549,12 @@ TEST_F(CompactionIteratorTest, ShuttingDownInMerge) {
   EXPECT_EQ(2, filter.last_seen.load());
 }
 
-TEST_F(CompactionIteratorTest, SingleMergeOperand) {
+TEST_P(CompactionIteratorTest, SingleMergeOperand) {
   class Filter : public CompactionFilter {
-    virtual Decision FilterV2(int level, const Slice& key, ValueType t,
+    virtual Decision FilterV2(int /*level*/, const Slice& key, ValueType t,
                               const Slice& existing_value,
-                              std::string* new_value,
-                              std::string* skip_until) const override {
+                              std::string* /*new_value*/,
+                              std::string* /*skip_until*/) const override {
       std::string k = key.ToString();
       std::string v = existing_value.ToString();
 
@@ -516,7 +605,7 @@ TEST_F(CompactionIteratorTest, SingleMergeOperand) {
     bool PartialMergeMulti(const Slice& key,
                            const std::deque<Slice>& operand_list,
                            std::string* new_value,
-                           Logger* logger) const override {
+                           Logger* /*logger*/) const override {
       std::string string_key = key.ToString();
       EXPECT_TRUE(string_key == "a" || string_key == "b");
 
@@ -552,7 +641,7 @@ TEST_F(CompactionIteratorTest, SingleMergeOperand) {
        // c should invoke FullMerge due to kTypeValue at the beginning.
        test::KeyStr("c", 90, kTypeMerge), test::KeyStr("c", 80, kTypeValue)},
       {"av1", "bv2", "bv1", "cv2", "cv1"}, {}, {}, kMaxSequenceNumber,
-      &merge_op, &filter);
+      kMaxSequenceNumber, &merge_op, &filter);
 
   c_iter_->SeekToFirst();
   ASSERT_TRUE(c_iter_->Valid());
@@ -565,6 +654,298 @@ TEST_F(CompactionIteratorTest, SingleMergeOperand) {
   ASSERT_EQ("cv1cv2", c_iter_->value().ToString());
 }
 
+// In bottommost level, values earlier than earliest snapshot can be output
+// with sequence = 0.
+TEST_P(CompactionIteratorTest, ZeroOutSequenceAtBottomLevel) {
+  AddSnapshot(1);
+  RunTest({test::KeyStr("a", 1, kTypeValue), test::KeyStr("b", 2, kTypeValue)},
+          {"v1", "v2"},
+          {test::KeyStr("a", 0, kTypeValue), test::KeyStr("b", 2, kTypeValue)},
+          {"v1", "v2"}, kMaxSequenceNumber /*last_commited_seq*/,
+          nullptr /*merge_operator*/, nullptr /*compaction_filter*/,
+          true /*bottommost_level*/);
+}
+
+// In bottommost level, deletions earlier than earliest snapshot can be removed
+// permanently.
+TEST_P(CompactionIteratorTest, RemoveDeletionAtBottomLevel) {
+  AddSnapshot(1);
+  RunTest({test::KeyStr("a", 1, kTypeDeletion),
+           test::KeyStr("b", 2, kTypeDeletion)},
+          {"", ""}, {test::KeyStr("b", 2, kTypeDeletion)}, {""},
+          kMaxSequenceNumber /*last_commited_seq*/, nullptr /*merge_operator*/,
+          nullptr /*compaction_filter*/, true /*bottommost_level*/);
+}
+
+// In bottommost level, single deletions earlier than earliest snapshot can be
+// removed permanently.
+TEST_P(CompactionIteratorTest, RemoveSingleDeletionAtBottomLevel) {
+  AddSnapshot(1);
+  RunTest({test::KeyStr("a", 1, kTypeSingleDeletion),
+           test::KeyStr("b", 2, kTypeSingleDeletion)},
+          {"", ""}, {test::KeyStr("b", 2, kTypeSingleDeletion)}, {""},
+          kMaxSequenceNumber /*last_commited_seq*/, nullptr /*merge_operator*/,
+          nullptr /*compaction_filter*/, true /*bottommost_level*/);
+}
+
+INSTANTIATE_TEST_CASE_P(CompactionIteratorTestInstance, CompactionIteratorTest,
+                        testing::Values(true, false));
+
+// Tests how CompactionIterator work together with SnapshotChecker.
+class CompactionIteratorWithSnapshotCheckerTest
+    : public CompactionIteratorTest {
+ public:
+  bool UseSnapshotChecker() const override { return true; }
+};
+
+// Uncommitted keys (keys with seq > last_committed_seq) should be output as-is
+// while committed version of these keys should get compacted as usual.
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_Value) {
+  RunTest(
+      {test::KeyStr("foo", 3, kTypeValue), test::KeyStr("foo", 2, kTypeValue),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v3", "v2", "v1"},
+      {test::KeyStr("foo", 3, kTypeValue), test::KeyStr("foo", 2, kTypeValue)},
+      {"v3", "v2"}, 2 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_Deletion) {
+  RunTest({test::KeyStr("foo", 2, kTypeDeletion),
+           test::KeyStr("foo", 1, kTypeValue)},
+          {"", "v1"},
+          {test::KeyStr("foo", 2, kTypeDeletion),
+           test::KeyStr("foo", 1, kTypeValue)},
+          {"", "v1"}, 1 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_Merge) {
+  auto merge_op = MergeOperators::CreateStringAppendOperator();
+  RunTest(
+      {test::KeyStr("foo", 3, kTypeMerge), test::KeyStr("foo", 2, kTypeMerge),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v3", "v2", "v1"},
+      {test::KeyStr("foo", 3, kTypeMerge), test::KeyStr("foo", 2, kTypeValue)},
+      {"v3", "v1,v2"}, 2 /*last_committed_seq*/, merge_op.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_SingleDelete) {
+  RunTest({test::KeyStr("foo", 2, kTypeSingleDeletion),
+           test::KeyStr("foo", 1, kTypeValue)},
+          {"", "v1"},
+          {test::KeyStr("foo", 2, kTypeSingleDeletion),
+           test::KeyStr("foo", 1, kTypeValue)},
+          {"", "v1"}, 1 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_BlobIndex) {
+  RunTest({test::KeyStr("foo", 3, kTypeBlobIndex),
+           test::KeyStr("foo", 2, kTypeBlobIndex),
+           test::KeyStr("foo", 1, kTypeBlobIndex)},
+          {"v3", "v2", "v1"},
+          {test::KeyStr("foo", 3, kTypeBlobIndex),
+           test::KeyStr("foo", 2, kTypeBlobIndex)},
+          {"v3", "v2"}, 2 /*last_committed_seq*/);
+}
+
+// Test compaction iterator dedup keys visible to the same snapshot.
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, DedupSameSnapshot_Value) {
+  AddSnapshot(2, 1);
+  RunTest(
+      {test::KeyStr("foo", 4, kTypeValue), test::KeyStr("foo", 3, kTypeValue),
+       test::KeyStr("foo", 2, kTypeValue), test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "v3", "v2", "v1"},
+      {test::KeyStr("foo", 4, kTypeValue), test::KeyStr("foo", 3, kTypeValue),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "v3", "v1"}, 3 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, DedupSameSnapshot_Deletion) {
+  AddSnapshot(2, 1);
+  RunTest(
+      {test::KeyStr("foo", 4, kTypeValue),
+       test::KeyStr("foo", 3, kTypeDeletion),
+       test::KeyStr("foo", 2, kTypeValue), test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "", "v2", "v1"},
+      {test::KeyStr("foo", 4, kTypeValue),
+       test::KeyStr("foo", 3, kTypeDeletion),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "", "v1"}, 3 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, DedupSameSnapshot_Merge) {
+  AddSnapshot(2, 1);
+  AddSnapshot(4, 3);
+  auto merge_op = MergeOperators::CreateStringAppendOperator();
+  RunTest(
+      {test::KeyStr("foo", 5, kTypeMerge), test::KeyStr("foo", 4, kTypeMerge),
+       test::KeyStr("foo", 3, kTypeMerge), test::KeyStr("foo", 2, kTypeMerge),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v5", "v4", "v3", "v2", "v1"},
+      {test::KeyStr("foo", 5, kTypeMerge), test::KeyStr("foo", 4, kTypeMerge),
+       test::KeyStr("foo", 3, kTypeMerge), test::KeyStr("foo", 1, kTypeValue)},
+      {"v5", "v4", "v2,v3", "v1"}, 4 /*last_committed_seq*/, merge_op.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       DedupSameSnapshot_SingleDeletion) {
+  AddSnapshot(2, 1);
+  RunTest(
+      {test::KeyStr("foo", 4, kTypeValue),
+       test::KeyStr("foo", 3, kTypeSingleDeletion),
+       test::KeyStr("foo", 2, kTypeValue), test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "", "v2", "v1"},
+      {test::KeyStr("foo", 4, kTypeValue), test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "v1"}, 3 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, DedupSameSnapshot_BlobIndex) {
+  AddSnapshot(2, 1);
+  RunTest({test::KeyStr("foo", 4, kTypeBlobIndex),
+           test::KeyStr("foo", 3, kTypeBlobIndex),
+           test::KeyStr("foo", 2, kTypeBlobIndex),
+           test::KeyStr("foo", 1, kTypeBlobIndex)},
+          {"v4", "v3", "v2", "v1"},
+          {test::KeyStr("foo", 4, kTypeBlobIndex),
+           test::KeyStr("foo", 3, kTypeBlobIndex),
+           test::KeyStr("foo", 1, kTypeBlobIndex)},
+          {"v4", "v3", "v1"}, 3 /*last_committed_seq*/);
+}
+
+// At bottom level, sequence numbers can be zero out, and deletions can be
+// removed, but only when they are visible to earliest snapshot.
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       NotZeroOutSequenceIfNotVisibleToEarliestSnapshot) {
+  AddSnapshot(2, 1);
+  RunTest({test::KeyStr("a", 1, kTypeValue), test::KeyStr("b", 2, kTypeValue),
+           test::KeyStr("c", 3, kTypeValue)},
+          {"v1", "v2", "v3"},
+          {test::KeyStr("a", 0, kTypeValue), test::KeyStr("b", 2, kTypeValue),
+           test::KeyStr("c", 3, kTypeValue)},
+          {"v1", "v2", "v3"}, kMaxSequenceNumber /*last_commited_seq*/,
+          nullptr /*merge_operator*/, nullptr /*compaction_filter*/,
+          true /*bottommost_level*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       NotRemoveDeletionIfNotVisibleToEarliestSnapshot) {
+  AddSnapshot(2, 1);
+  RunTest(
+      {test::KeyStr("a", 1, kTypeDeletion), test::KeyStr("b", 2, kTypeDeletion),
+       test::KeyStr("c", 3, kTypeDeletion)},
+      {"", "", ""},
+      {test::KeyStr("b", 2, kTypeDeletion),
+       test::KeyStr("c", 3, kTypeDeletion)},
+      {"", ""}, kMaxSequenceNumber /*last_commited_seq*/,
+      nullptr /*merge_operator*/, nullptr /*compaction_filter*/,
+      true /*bottommost_level*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       NotRemoveSingleDeletionIfNotVisibleToEarliestSnapshot) {
+  AddSnapshot(2, 1);
+  RunTest({test::KeyStr("a", 1, kTypeSingleDeletion),
+           test::KeyStr("b", 2, kTypeSingleDeletion),
+           test::KeyStr("c", 3, kTypeSingleDeletion)},
+          {"", "", ""},
+          {test::KeyStr("b", 2, kTypeSingleDeletion),
+           test::KeyStr("c", 3, kTypeSingleDeletion)},
+          {"", ""}, kMaxSequenceNumber /*last_commited_seq*/,
+          nullptr /*merge_operator*/, nullptr /*compaction_filter*/,
+          true /*bottommost_level*/);
+}
+
+// Single delete should not cancel out values that not visible to the
+// same set of snapshots
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       SingleDeleteAcrossSnapshotBoundary) {
+  AddSnapshot(2, 1);
+  RunTest({test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", "v1"},
+          {test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", "v1"}, 2 /*last_committed_seq*/);
+}
+
+// Single delete should be kept in case it is not visible to the
+// earliest write conflict snapshot. If a single delete is kept for this reason,
+// corresponding value can be trimmed to save space.
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       KeepSingleDeletionForWriteConflictChecking) {
+  AddSnapshot(2, 0);
+  RunTest({test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", "v1"},
+          {test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", ""}, 2 /*last_committed_seq*/, nullptr /*merge_operator*/,
+          nullptr /*compaction_filter*/, false /*bottommost_level*/,
+          2 /*earliest_write_conflict_snapshot*/);
+}
+
+// Compaction filter should keep uncommitted key as-is, and
+//   * Convert the latest velue to deletion, and/or
+//   * if latest value is a merge, apply filter to all suequent merges.
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, CompactionFilter_Value) {
+  std::unique_ptr<CompactionFilter> compaction_filter(
+      new FilterAllKeysCompactionFilter());
+  RunTest(
+      {test::KeyStr("a", 2, kTypeValue), test::KeyStr("a", 1, kTypeValue),
+       test::KeyStr("b", 3, kTypeValue), test::KeyStr("c", 1, kTypeValue)},
+      {"v2", "v1", "v3", "v4"},
+      {test::KeyStr("a", 2, kTypeValue), test::KeyStr("a", 1, kTypeDeletion),
+       test::KeyStr("b", 3, kTypeValue), test::KeyStr("c", 1, kTypeDeletion)},
+      {"v2", "", "v3", ""}, 1 /*last_committed_seq*/,
+      nullptr /*merge_operator*/, compaction_filter.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, CompactionFilter_Deletion) {
+  std::unique_ptr<CompactionFilter> compaction_filter(
+      new FilterAllKeysCompactionFilter());
+  RunTest(
+      {test::KeyStr("a", 2, kTypeDeletion), test::KeyStr("a", 1, kTypeValue)},
+      {"", "v1"},
+      {test::KeyStr("a", 2, kTypeDeletion),
+       test::KeyStr("a", 1, kTypeDeletion)},
+      {"", ""}, 1 /*last_committed_seq*/, nullptr /*merge_operator*/,
+      compaction_filter.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       CompactionFilter_PartialMerge) {
+  std::shared_ptr<MergeOperator> merge_op =
+      MergeOperators::CreateStringAppendOperator();
+  std::unique_ptr<CompactionFilter> compaction_filter(
+      new FilterAllKeysCompactionFilter());
+  RunTest({test::KeyStr("a", 3, kTypeMerge), test::KeyStr("a", 2, kTypeMerge),
+           test::KeyStr("a", 1, kTypeMerge)},
+          {"v3", "v2", "v1"}, {test::KeyStr("a", 3, kTypeMerge)}, {"v3"},
+          2 /*last_committed_seq*/, merge_op.get(), compaction_filter.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, CompactionFilter_FullMerge) {
+  std::shared_ptr<MergeOperator> merge_op =
+      MergeOperators::CreateStringAppendOperator();
+  std::unique_ptr<CompactionFilter> compaction_filter(
+      new FilterAllKeysCompactionFilter());
+  RunTest(
+      {test::KeyStr("a", 3, kTypeMerge), test::KeyStr("a", 2, kTypeMerge),
+       test::KeyStr("a", 1, kTypeValue)},
+      {"v3", "v2", "v1"},
+      {test::KeyStr("a", 3, kTypeMerge), test::KeyStr("a", 1, kTypeDeletion)},
+      {"v3", ""}, 2 /*last_committed_seq*/, merge_op.get(),
+      compaction_filter.get());
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/compaction_job.cc

@@ -747,8 +747,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       compaction_filter, db_options_.info_log.get(),
       false /* internal key corruption is expected */,
       existing_snapshots_.empty() ? 0 : existing_snapshots_.back(),
-      compact_->compaction->level(), db_options_.statistics.get(),
-      shutting_down_);
+      snapshot_checker_, compact_->compaction->level(),
+      db_options_.statistics.get(), shutting_down_);
 
   TEST_SYNC_POINT("CompactionJob::Run():Inprogress");
 
@@ -762,24 +762,13 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     input->SeekToFirst();
   }
 
-  // we allow only 1 compaction event listener. Used by blob storage
-  CompactionEventListener* comp_event_listener = nullptr;
-#ifndef ROCKSDB_LITE
-  for (auto& celitr : cfd->ioptions()->listeners) {
-    comp_event_listener = celitr->GetCompactionEventListener();
-    if (comp_event_listener != nullptr) {
-      break;
-    }
-  }
-#endif  // ROCKSDB_LITE
-
   Status status;
   sub_compact->c_iter.reset(new CompactionIterator(
       input.get(), cfd->user_comparator(), &merge, versions_->LastSequence(),
       &existing_snapshots_, earliest_write_conflict_snapshot_,
       snapshot_checker_, env_, false, range_del_agg.get(),
-      sub_compact->compaction, compaction_filter, comp_event_listener,
-      shutting_down_, preserve_deletes_seqnum_));
+      sub_compact->compaction, compaction_filter, shutting_down_,
+      preserve_deletes_seqnum_));
   auto c_iter = sub_compact->c_iter.get();
   c_iter->SeekToFirst();
   if (c_iter->Valid() &&

db/compaction_job_stats_test.cc

@@ -426,7 +426,7 @@ class CompactionJobStatsChecker : public EventListener {
   // Once a compaction completed, this function will verify the returned
   // CompactionJobInfo with the oldest CompactionJobInfo added earlier
   // in "expected_stats_" which has not yet being used for verification.
-  virtual void OnCompactionCompleted(DB *db, const CompactionJobInfo& ci) {
+  virtual void OnCompactionCompleted(DB* /*db*/, const CompactionJobInfo& ci) {
     if (verify_next_comp_io_stats_) {
       ASSERT_GT(ci.stats.file_write_nanos, 0);
       ASSERT_GT(ci.stats.file_range_sync_nanos, 0);

db/compaction_picker.cc

@@ -199,7 +199,7 @@ void CompactionPicker::GetRange(const std::vector<CompactionInputFiles>& inputs,
   assert(initialized);
 }
 
-bool CompactionPicker::ExpandInputsToCleanCut(const std::string& cf_name,
+bool CompactionPicker::ExpandInputsToCleanCut(const std::string& /*cf_name*/,
                                               VersionStorageInfo* vstorage,
                                               CompactionInputFiles* inputs) {
   // This isn't good compaction
@@ -309,7 +309,7 @@ Compaction* CompactionPicker::CompactFiles(
 Status CompactionPicker::GetCompactionInputsFromFileNumbers(
     std::vector<CompactionInputFiles>* input_files,
     std::unordered_set<uint64_t>* input_set, const VersionStorageInfo* vstorage,
-    const CompactionOptions& compact_options) const {
+    const CompactionOptions& /*compact_options*/) const {
   if (input_set->size() == 0U) {
     return Status::InvalidArgument(
         "Compaction must include at least one file.");
@@ -1612,8 +1612,9 @@ Compaction* FIFOCompactionPicker::PickCompaction(
 Compaction* FIFOCompactionPicker::CompactRange(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     VersionStorageInfo* vstorage, int input_level, int output_level,
-    uint32_t output_path_id, const InternalKey* begin, const InternalKey* end,
-    InternalKey** compaction_end, bool* manual_conflict) {
+    uint32_t /*output_path_id*/, const InternalKey* /*begin*/,
+    const InternalKey* /*end*/, InternalKey** compaction_end,
+    bool* /*manual_conflict*/) {
   assert(input_level == 0);
   assert(output_level == 0);
   *compaction_end = nullptr;

db/compaction_picker.h

@@ -267,27 +267,29 @@ class NullCompactionPicker : public CompactionPicker {
   virtual ~NullCompactionPicker() {}
 
   // Always return "nullptr"
-  Compaction* PickCompaction(const std::string& cf_name,
-                             const MutableCFOptions& mutable_cf_options,
-                             VersionStorageInfo* vstorage,
-                             LogBuffer* log_buffer) override {
+  Compaction* PickCompaction(const std::string& /*cf_name*/,
+                             const MutableCFOptions& /*mutable_cf_options*/,
+                             VersionStorageInfo* /*vstorage*/,
+                             LogBuffer* /*log_buffer*/) override {
     return nullptr;
   }
 
   // Always return "nullptr"
-  Compaction* CompactRange(const std::string& cf_name,
-                           const MutableCFOptions& mutable_cf_options,
-                           VersionStorageInfo* vstorage, int input_level,
-                           int output_level, uint32_t output_path_id,
-                           const InternalKey* begin, const InternalKey* end,
-                           InternalKey** compaction_end,
-                           bool* manual_conflict) override {
+  Compaction* CompactRange(const std::string& /*cf_name*/,
+                           const MutableCFOptions& /*mutable_cf_options*/,
+                           VersionStorageInfo* /*vstorage*/,
+                           int /*input_level*/, int /*output_level*/,
+                           uint32_t /*output_path_id*/,
+                           const InternalKey* /*begin*/,
+                           const InternalKey* /*end*/,
+                           InternalKey** /*compaction_end*/,
+                           bool* /*manual_conflict*/) override {
     return nullptr;
   }
 
   // Always returns false.
   virtual bool NeedsCompaction(
-      const VersionStorageInfo* vstorage) const override {
+      const VersionStorageInfo* /*vstorage*/) const override {
     return false;
   }
 };

db/compaction_picker_test.cc

@@ -20,7 +20,9 @@ namespace rocksdb {
 class CountingLogger : public Logger {
  public:
   using Logger::Logv;
-  virtual void Logv(const char* format, va_list ap) override { log_count++; }
+  virtual void Logv(const char* /*format*/, va_list /*ap*/) override {
+    log_count++;
+  }
   size_t log_count;
 };
 

db/compaction_picker_universal.cc

@@ -204,7 +204,7 @@ void UniversalCompactionPicker::SortedRun::DumpSizeInfo(
 
 std::vector<UniversalCompactionPicker::SortedRun>
 UniversalCompactionPicker::CalculateSortedRuns(
-    const VersionStorageInfo& vstorage, const ImmutableCFOptions& ioptions,
+    const VersionStorageInfo& vstorage, const ImmutableCFOptions& /*ioptions*/,
     const MutableCFOptions& mutable_cf_options) {
   std::vector<UniversalCompactionPicker::SortedRun> ret;
   for (FileMetaData* f : vstorage.LevelFiles(0)) {
@@ -603,9 +603,9 @@ Compaction* UniversalCompactionPicker::PickCompactionToReduceSortedRuns(
 
   CompactionReason compaction_reason;
   if (max_number_of_files_to_compact == UINT_MAX) {
-    compaction_reason = CompactionReason::kUniversalSortedRunNum;
-  } else {
     compaction_reason = CompactionReason::kUniversalSizeRatio;
+  } else {
+    compaction_reason = CompactionReason::kUniversalSortedRunNum;
   }
   return new Compaction(
       vstorage, ioptions_, mutable_cf_options, std::move(inputs), output_level,

db/comparator_db_test.cc

@@ -188,10 +188,10 @@ class DoubleComparator : public Comparator {
       return -1;
     }
   }
-  virtual void FindShortestSeparator(std::string* start,
-                                     const Slice& limit) const override {}
+  virtual void FindShortestSeparator(std::string* /*start*/,
+                                     const Slice& /*limit*/) const override {}
 
-  virtual void FindShortSuccessor(std::string* key) const override {}
+  virtual void FindShortSuccessor(std::string* /*key*/) const override {}
 };
 
 class HashComparator : public Comparator {
@@ -211,10 +211,10 @@ class HashComparator : public Comparator {
       return -1;
     }
   }
-  virtual void FindShortestSeparator(std::string* start,
-                                     const Slice& limit) const override {}
+  virtual void FindShortestSeparator(std::string* /*start*/,
+                                     const Slice& /*limit*/) const override {}
 
-  virtual void FindShortSuccessor(std::string* key) const override {}
+  virtual void FindShortSuccessor(std::string* /*key*/) const override {}
 };
 
 class TwoStrComparator : public Comparator {
@@ -243,10 +243,10 @@ class TwoStrComparator : public Comparator {
     }
     return a2.compare(b2);
   }
-  virtual void FindShortestSeparator(std::string* start,
-                                     const Slice& limit) const override {}
+  virtual void FindShortestSeparator(std::string* /*start*/,
+                                     const Slice& /*limit*/) const override {}
 
-  virtual void FindShortSuccessor(std::string* key) const override {}
+  virtual void FindShortSuccessor(std::string* /*key*/) const override {}
 };
 }  // namespace
 

db/convenience.cc

@@ -19,9 +19,17 @@ void CancelAllBackgroundWork(DB* db, bool wait) {
 }
 
 Status DeleteFilesInRange(DB* db, ColumnFamilyHandle* column_family,
-                          const Slice* begin, const Slice* end) {
+                          const Slice* begin, const Slice* end,
+                          bool include_end) {
+  RangePtr range(begin, end);
+  return DeleteFilesInRanges(db, column_family, &range, 1, include_end);
+}
+
+Status DeleteFilesInRanges(DB* db, ColumnFamilyHandle* column_family,
+                           const RangePtr* ranges, size_t n,
+                           bool include_end) {
   return (static_cast_with_check<DBImpl, DB>(db->GetRootDB()))
-      ->DeleteFilesInRange(column_family, begin, end);
+      ->DeleteFilesInRanges(column_family, ranges, n, include_end);
 }
 
 Status VerifySstFileChecksum(const Options& options,

db/db_basic_test.cc

@@ -9,6 +9,7 @@
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/perf_context.h"
+#include "util/fault_injection_test_env.h"
 #if !defined(ROCKSDB_LITE)
 #include "util/sync_point.h"
 #endif
@@ -849,23 +850,44 @@ TEST_F(DBBasicTest, MmapAndBufferOptions) {
 
 class TestEnv : public EnvWrapper {
   public:
-    explicit TestEnv(Env* base) : EnvWrapper(base) { };
+    explicit TestEnv() : EnvWrapper(Env::Default()),
+                close_count(0) { }
 
     class TestLogger : public Logger {
       public:
         using Logger::Logv;
-        virtual void Logv(const char *format, va_list ap) override { };
-      private:
-        virtual Status CloseImpl() override {
-          return Status::NotSupported();
+        TestLogger(TestEnv *env_ptr) : Logger() { env = env_ptr; }
+        ~TestLogger() {
+          if (!closed_) {
+            CloseHelper();
+          }
         }
+        virtual void Logv(const char* /*format*/, va_list /*ap*/) override{};
+
+       protected:
+        virtual Status CloseImpl() override {
+          return CloseHelper();
+        }
+      private:
+        Status CloseHelper() {
+          env->CloseCountInc();;
+          return Status::IOError();
+        }
+        TestEnv *env;
     };
 
-    virtual Status NewLogger(const std::string& fname,
+    void CloseCountInc() { close_count++; }
+
+    int GetCloseCount() { return close_count; }
+
+    virtual Status NewLogger(const std::string& /*fname*/,
                              shared_ptr<Logger>* result) {
-      result->reset(new TestLogger());
+      result->reset(new TestLogger(this));
       return Status::OK();
     }
+
+  private:
+    int close_count;
 };
 
 TEST_F(DBBasicTest, DBClose) {
@@ -874,19 +896,29 @@ TEST_F(DBBasicTest, DBClose) {
   ASSERT_OK(DestroyDB(dbname, options));
 
   DB* db = nullptr;
+  TestEnv *env = new TestEnv();
   options.create_if_missing = true;
-  options.env = new TestEnv(Env::Default());
+  options.env = env;
   Status s = DB::Open(options, dbname, &db);
   ASSERT_OK(s);
   ASSERT_TRUE(db != nullptr);
 
   s = db->Close();
-  ASSERT_EQ(s, Status::NotSupported());
+  ASSERT_EQ(env->GetCloseCount(), 1);
+  ASSERT_EQ(s, Status::IOError());
 
   delete db;
+  ASSERT_EQ(env->GetCloseCount(), 1);
+
+  // Do not call DB::Close() and ensure our logger Close() still gets called
+  s = DB::Open(options, dbname, &db);
+  ASSERT_OK(s);
+  ASSERT_TRUE(db != nullptr);
+  delete db;
+  ASSERT_EQ(env->GetCloseCount(), 2);
 
   // Provide our own logger and ensure DB::Close() does not close it
-  options.info_log.reset(new TestEnv::TestLogger());
+  options.info_log.reset(new TestEnv::TestLogger(env));
   options.create_if_missing = false;
   s = DB::Open(options, dbname, &db);
   ASSERT_OK(s);
@@ -895,9 +927,35 @@ TEST_F(DBBasicTest, DBClose) {
   s = db->Close();
   ASSERT_EQ(s, Status::OK());
   delete db;
+  ASSERT_EQ(env->GetCloseCount(), 2);
+  options.info_log.reset();
+  ASSERT_EQ(env->GetCloseCount(), 3);
+
   delete options.env;
 }
 
+TEST_F(DBBasicTest, DBCloseFlushError) {
+  std::unique_ptr<FaultInjectionTestEnv> fault_injection_env(
+      new FaultInjectionTestEnv(Env::Default()));
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.manual_wal_flush = true;
+  options.write_buffer_size=100;
+  options.env = fault_injection_env.get();
+
+  Reopen(options);
+  ASSERT_OK(Put("key1", "value1"));
+  ASSERT_OK(Put("key2", "value2"));
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable());
+  ASSERT_OK(Put("key3", "value3"));
+  fault_injection_env->SetFilesystemActive(false);
+  Status s = dbfull()->Close();
+  fault_injection_env->SetFilesystemActive(true);
+  ASSERT_NE(s, Status::OK());
+
+  Destroy(options);
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/db_block_cache_test.cc

@@ -47,7 +47,7 @@ class DBBlockCacheTest : public DBTestBase {
     return options;
   }
 
-  void InitTable(const Options& options) {
+  void InitTable(const Options& /*options*/) {
     std::string value(kValueSize, 'a');
     for (size_t i = 0; i < kNumBlocks; i++) {
       ASSERT_OK(Put(ToString(i), value.c_str()));
@@ -111,6 +111,31 @@ class DBBlockCacheTest : public DBTestBase {
   }
 };
 
+TEST_F(DBBlockCacheTest, IteratorBlockCacheUsage) {
+  ReadOptions read_options;
+  read_options.fill_cache = false;
+  auto table_options = GetTableOptions();
+  auto options = GetOptions(table_options);
+  InitTable(options);
+
+  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  table_options.block_cache = cache;
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  Reopen(options);
+  RecordCacheCounters(options);
+
+  std::vector<std::unique_ptr<Iterator>> iterators(kNumBlocks - 1);
+  Iterator* iter = nullptr;
+
+  ASSERT_EQ(0, cache->GetUsage());
+  iter = db_->NewIterator(read_options);
+  iter->Seek(ToString(0));
+  ASSERT_LT(0, cache->GetUsage());
+  delete iter;
+  iter = nullptr;
+  ASSERT_EQ(0, cache->GetUsage());
+}
+
 TEST_F(DBBlockCacheTest, TestWithoutCompressedBlockCache) {
   ReadOptions read_options;
   auto table_options = GetTableOptions();

db/db_compaction_filter_test.cc

@@ -56,9 +56,9 @@ INSTANTIATE_TEST_CASE_P(
 
 class KeepFilter : public CompactionFilter {
  public:
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value, bool* value_changed) const
-      override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& /*value*/, std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     cfilter_count++;
     return false;
   }
@@ -68,9 +68,9 @@ class KeepFilter : public CompactionFilter {
 
 class DeleteFilter : public CompactionFilter {
  public:
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value, bool* value_changed) const
-      override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& /*value*/, std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     cfilter_count++;
     return true;
   }
@@ -80,9 +80,9 @@ class DeleteFilter : public CompactionFilter {
 
 class DeleteISFilter : public CompactionFilter {
  public:
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value,
-                      bool* value_changed) const override {
+  virtual bool Filter(int /*level*/, const Slice& key, const Slice& /*value*/,
+                      std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     cfilter_count++;
     int i = std::stoi(key.ToString());
     if (i > 5 && i <= 105) {
@@ -100,8 +100,10 @@ class DeleteISFilter : public CompactionFilter {
 // zero-padded to length 10.
 class SkipEvenFilter : public CompactionFilter {
  public:
-  virtual Decision FilterV2(int level, const Slice& key, ValueType value_type,
-                            const Slice& existing_value, std::string* new_value,
+  virtual Decision FilterV2(int /*level*/, const Slice& key,
+                            ValueType /*value_type*/,
+                            const Slice& /*existing_value*/,
+                            std::string* /*new_value*/,
                             std::string* skip_until) const override {
     cfilter_count++;
     int i = std::stoi(key.ToString());
@@ -123,9 +125,9 @@ class SkipEvenFilter : public CompactionFilter {
 class DelayFilter : public CompactionFilter {
  public:
   explicit DelayFilter(DBTestBase* d) : db_test(d) {}
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value,
-                      bool* value_changed) const override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& /*value*/, std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     db_test->env_->addon_time_.fetch_add(1000);
     return true;
   }
@@ -140,9 +142,9 @@ class ConditionalFilter : public CompactionFilter {
  public:
   explicit ConditionalFilter(const std::string* filtered_value)
       : filtered_value_(filtered_value) {}
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value,
-                      bool* value_changed) const override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/, const Slice& value,
+                      std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     return value.ToString() == *filtered_value_;
   }
 
@@ -156,9 +158,9 @@ class ChangeFilter : public CompactionFilter {
  public:
   explicit ChangeFilter() {}
 
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value, bool* value_changed) const
-      override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& /*value*/, std::string* new_value,
+                      bool* value_changed) const override {
     assert(new_value != nullptr);
     *new_value = NEW_VALUE;
     *value_changed = true;
@@ -247,7 +249,7 @@ class DelayFilterFactory : public CompactionFilterFactory {
  public:
   explicit DelayFilterFactory(DBTestBase* d) : db_test(d) {}
   virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
-      const CompactionFilter::Context& context) override {
+      const CompactionFilter::Context& /*context*/) override {
     return std::unique_ptr<CompactionFilter>(new DelayFilter(db_test));
   }
 
@@ -263,7 +265,7 @@ class ConditionalFilterFactory : public CompactionFilterFactory {
       : filtered_value_(filtered_value.ToString()) {}
 
   virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
-      const CompactionFilter::Context& context) override {
+      const CompactionFilter::Context& /*context*/) override {
     return std::unique_ptr<CompactionFilter>(
         new ConditionalFilter(&filtered_value_));
   }
@@ -281,7 +283,7 @@ class ChangeFilterFactory : public CompactionFilterFactory {
   explicit ChangeFilterFactory() {}
 
   virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
-      const CompactionFilter::Context& context) override {
+      const CompactionFilter::Context& /*context*/) override {
     return std::unique_ptr<CompactionFilter>(new ChangeFilter());
   }
 

db/db_compaction_test.cc

@@ -53,7 +53,7 @@ class FlushedFileCollector : public EventListener {
   FlushedFileCollector() {}
   ~FlushedFileCollector() {}
 
-  virtual void OnFlushCompleted(DB* db, const FlushJobInfo& info) override {
+  virtual void OnFlushCompleted(DB* /*db*/, const FlushJobInfo& info) override {
     std::lock_guard<std::mutex> lock(mutex_);
     flushed_files_.push_back(info.file_path);
   }
@@ -1517,6 +1517,122 @@ TEST_F(DBCompactionTest, DeleteFileRange) {
   ASSERT_GT(old_num_files, new_num_files);
 }
 
+TEST_F(DBCompactionTest, DeleteFilesInRanges) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 10 * 1024 * 1024;
+  options.max_bytes_for_level_multiplier = 2;
+  options.num_levels = 4;
+  options.max_background_compactions = 3;
+  options.disable_auto_compactions = true;
+
+  DestroyAndReopen(options);
+  int32_t value_size = 10 * 1024;  // 10 KB
+
+  Random rnd(301);
+  std::map<int32_t, std::string> values;
+
+  // file [0 => 100), [100 => 200), ... [900, 1000)
+  for (auto i = 0; i < 10; i++) {
+    for (auto j = 0; j < 100; j++) {
+      auto k = i * 100 + j;
+      values[k] = RandomString(&rnd, value_size);
+      ASSERT_OK(Put(Key(k), values[k]));
+    }
+    ASSERT_OK(Flush());
+  }
+  ASSERT_EQ("10", FilesPerLevel(0));
+  CompactRangeOptions compact_options;
+  compact_options.change_level = true;
+  compact_options.target_level = 2;
+  ASSERT_OK(db_->CompactRange(compact_options, nullptr, nullptr));
+  ASSERT_EQ("0,0,10", FilesPerLevel(0));
+
+  // file [0 => 100), [200 => 300), ... [800, 900)
+  for (auto i = 0; i < 10; i+=2) {
+    for (auto j = 0; j < 100; j++) {
+      auto k = i * 100 + j;
+      ASSERT_OK(Put(Key(k), values[k]));
+    }
+    ASSERT_OK(Flush());
+  }
+  ASSERT_EQ("5,0,10", FilesPerLevel(0));
+  ASSERT_OK(dbfull()->TEST_CompactRange(0, nullptr, nullptr));
+  ASSERT_EQ("0,5,10", FilesPerLevel(0));
+
+  // Delete files in range [0, 299] (inclusive)
+  {
+    auto begin_str1 = Key(0), end_str1 = Key(100);
+    auto begin_str2 = Key(100), end_str2 = Key(200);
+    auto begin_str3 = Key(200), end_str3 = Key(299);
+    Slice begin1(begin_str1), end1(end_str1);
+    Slice begin2(begin_str2), end2(end_str2);
+    Slice begin3(begin_str3), end3(end_str3);
+    std::vector<RangePtr> ranges;
+    ranges.push_back(RangePtr(&begin1, &end1));
+    ranges.push_back(RangePtr(&begin2, &end2));
+    ranges.push_back(RangePtr(&begin3, &end3));
+    ASSERT_OK(DeleteFilesInRanges(db_, db_->DefaultColumnFamily(),
+                                  ranges.data(), ranges.size()));
+    ASSERT_EQ("0,3,7", FilesPerLevel(0));
+
+    // Keys [0, 300) should not exist.
+    for (auto i = 0; i < 300; i++) {
+      ReadOptions ropts;
+      std::string result;
+      auto s = db_->Get(ropts, Key(i), &result);
+      ASSERT_TRUE(s.IsNotFound());
+    }
+    for (auto i = 300; i < 1000; i++) {
+      ASSERT_EQ(Get(Key(i)), values[i]);
+    }
+  }
+
+  // Delete files in range [600, 999) (exclusive)
+  {
+    auto begin_str1 = Key(600), end_str1 = Key(800);
+    auto begin_str2 = Key(700), end_str2 = Key(900);
+    auto begin_str3 = Key(800), end_str3 = Key(999);
+    Slice begin1(begin_str1), end1(end_str1);
+    Slice begin2(begin_str2), end2(end_str2);
+    Slice begin3(begin_str3), end3(end_str3);
+    std::vector<RangePtr> ranges;
+    ranges.push_back(RangePtr(&begin1, &end1));
+    ranges.push_back(RangePtr(&begin2, &end2));
+    ranges.push_back(RangePtr(&begin3, &end3));
+    ASSERT_OK(DeleteFilesInRanges(db_, db_->DefaultColumnFamily(),
+                                  ranges.data(), ranges.size(), false));
+    ASSERT_EQ("0,1,4", FilesPerLevel(0));
+
+    // Keys [600, 900) should not exist.
+    for (auto i = 600; i < 900; i++) {
+      ReadOptions ropts;
+      std::string result;
+      auto s = db_->Get(ropts, Key(i), &result);
+      ASSERT_TRUE(s.IsNotFound());
+    }
+    for (auto i = 300; i < 600; i++) {
+      ASSERT_EQ(Get(Key(i)), values[i]);
+    }
+    for (auto i = 900; i < 1000; i++) {
+      ASSERT_EQ(Get(Key(i)), values[i]);
+    }
+  }
+
+  // Delete all files.
+  {
+    RangePtr range;
+    ASSERT_OK(DeleteFilesInRanges(db_, db_->DefaultColumnFamily(), &range, 1));
+    ASSERT_EQ("", FilesPerLevel(0));
+
+    for (auto i = 0; i < 1000; i++) {
+      ReadOptions ropts;
+      std::string result;
+      auto s = db_->Get(ropts, Key(i), &result);
+      ASSERT_TRUE(s.IsNotFound());
+    }
+  }
+}
+
 TEST_F(DBCompactionTest, DeleteFileRangeFileEndpointsOverlapBug) {
   // regression test for #2833: groups of files whose user-keys overlap at the
   // endpoints could be split by `DeleteFilesInRange`. This caused old data to
@@ -2989,6 +3105,276 @@ TEST_F(DBCompactionTest, CompactBottomLevelFilesWithDeletions) {
   }
 }
 
+TEST_F(DBCompactionTest, CompactRangeDelayedByL0FileCount) {
+  // Verify that, when `CompactRangeOptions::allow_write_stall == false`, manual
+  // compaction only triggers flush after it's sure stall won't be triggered for
+  // L0 file count going too high.
+  const int kNumL0FilesTrigger = 4;
+  const int kNumL0FilesLimit = 8;
+  // i == 0: verifies normal case where stall is avoided by delay
+  // i == 1: verifies no delay in edge case where stall trigger is same as
+  //         compaction trigger, so stall can't be avoided
+  for (int i = 0; i < 2; ++i) {
+    Options options = CurrentOptions();
+    options.level0_slowdown_writes_trigger = kNumL0FilesLimit;
+    if (i == 0) {
+      options.level0_file_num_compaction_trigger = kNumL0FilesTrigger;
+    } else {
+      options.level0_file_num_compaction_trigger = kNumL0FilesLimit;
+    }
+    Reopen(options);
+
+    if (i == 0) {
+      // ensure the auto compaction doesn't finish until manual compaction has
+      // had a chance to be delayed.
+      rocksdb::SyncPoint::GetInstance()->LoadDependency(
+          {{"DBImpl::CompactRange:StallWait", "CompactionJob::Run():End"}});
+    } else {
+      // ensure the auto-compaction doesn't finish until manual compaction has
+      // continued without delay.
+      rocksdb::SyncPoint::GetInstance()->LoadDependency(
+          {{"DBImpl::CompactRange:StallWaitDone", "CompactionJob::Run():End"}});
+    }
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+    Random rnd(301);
+    for (int j = 0; j < kNumL0FilesLimit - 1; ++j) {
+      for (int k = 0; k < 2; ++k) {
+        ASSERT_OK(Put(Key(k), RandomString(&rnd, 1024)));
+      }
+      Flush();
+    }
+    auto manual_compaction_thread = port::Thread([this]() {
+      CompactRangeOptions cro;
+      cro.allow_write_stall = false;
+      db_->CompactRange(cro, nullptr, nullptr);
+    });
+
+    manual_compaction_thread.join();
+    dbfull()->TEST_WaitForCompact();
+    ASSERT_EQ(0, NumTableFilesAtLevel(0));
+    ASSERT_GT(NumTableFilesAtLevel(1), 0);
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
+
+TEST_F(DBCompactionTest, CompactRangeDelayedByImmMemTableCount) {
+  // Verify that, when `CompactRangeOptions::allow_write_stall == false`, manual
+  // compaction only triggers flush after it's sure stall won't be triggered for
+  // immutable memtable count going too high.
+  const int kNumImmMemTableLimit = 8;
+  // i == 0: verifies normal case where stall is avoided by delay
+  // i == 1: verifies no delay in edge case where stall trigger is same as flush
+  //         trigger, so stall can't be avoided
+  for (int i = 0; i < 2; ++i) {
+    Options options = CurrentOptions();
+    options.disable_auto_compactions = true;
+    // the delay limit is one less than the stop limit. This test focuses on
+    // avoiding delay limit, but this option sets stop limit, so add one.
+    options.max_write_buffer_number = kNumImmMemTableLimit + 1;
+    if (i == 1) {
+      options.min_write_buffer_number_to_merge = kNumImmMemTableLimit;
+    }
+    Reopen(options);
+
+    if (i == 0) {
+      // ensure the flush doesn't finish until manual compaction has had a
+      // chance to be delayed.
+      rocksdb::SyncPoint::GetInstance()->LoadDependency(
+          {{"DBImpl::CompactRange:StallWait", "FlushJob::WriteLevel0Table"}});
+    } else {
+      // ensure the flush doesn't finish until manual compaction has continued
+      // without delay.
+      rocksdb::SyncPoint::GetInstance()->LoadDependency(
+          {{"DBImpl::CompactRange:StallWaitDone",
+            "FlushJob::WriteLevel0Table"}});
+    }
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+    Random rnd(301);
+    for (int j = 0; j < kNumImmMemTableLimit - 1; ++j) {
+      ASSERT_OK(Put(Key(0), RandomString(&rnd, 1024)));
+      FlushOptions flush_opts;
+      flush_opts.wait = false;
+      dbfull()->Flush(flush_opts);
+    }
+
+    auto manual_compaction_thread = port::Thread([this]() {
+      CompactRangeOptions cro;
+      cro.allow_write_stall = false;
+      db_->CompactRange(cro, nullptr, nullptr);
+    });
+
+    manual_compaction_thread.join();
+    dbfull()->TEST_WaitForFlushMemTable();
+    ASSERT_EQ(0, NumTableFilesAtLevel(0));
+    ASSERT_GT(NumTableFilesAtLevel(1), 0);
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
+
+TEST_F(DBCompactionTest, CompactRangeShutdownWhileDelayed) {
+  // Verify that, when `CompactRangeOptions::allow_write_stall == false`, delay
+  // does not hang if CF is dropped or DB is closed
+  const int kNumL0FilesTrigger = 4;
+  const int kNumL0FilesLimit = 8;
+  Options options = CurrentOptions();
+  options.level0_file_num_compaction_trigger = kNumL0FilesTrigger;
+  options.level0_slowdown_writes_trigger = kNumL0FilesLimit;
+  // i == 0: DB::DropColumnFamily() on CompactRange's target CF unblocks it
+  // i == 1: DB::CancelAllBackgroundWork() unblocks CompactRange. This is to
+  //         simulate what happens during Close as we can't call Close (it
+  //         blocks on the auto-compaction, making a cycle).
+  for (int i = 0; i < 2; ++i) {
+    CreateAndReopenWithCF({"one"}, options);
+    // The calls to close CF/DB wait until the manual compaction stalls.
+    // The auto-compaction waits until the manual compaction finishes to ensure
+    // the signal comes from closing CF/DB, not from compaction making progress.
+    rocksdb::SyncPoint::GetInstance()->LoadDependency(
+        {{"DBImpl::CompactRange:StallWait",
+          "DBCompactionTest::CompactRangeShutdownWhileDelayed:PreShutdown"},
+         {"DBCompactionTest::CompactRangeShutdownWhileDelayed:PostManual",
+          "CompactionJob::Run():End"}});
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+    Random rnd(301);
+    for (int j = 0; j < kNumL0FilesLimit - 1; ++j) {
+      for (int k = 0; k < 2; ++k) {
+        ASSERT_OK(Put(1, Key(k), RandomString(&rnd, 1024)));
+      }
+      Flush(1);
+    }
+    auto manual_compaction_thread = port::Thread([this]() {
+      CompactRangeOptions cro;
+      cro.allow_write_stall = false;
+      ASSERT_TRUE(db_->CompactRange(cro, handles_[1], nullptr, nullptr)
+                      .IsShutdownInProgress());
+    });
+
+    TEST_SYNC_POINT(
+        "DBCompactionTest::CompactRangeShutdownWhileDelayed:PreShutdown");
+    if (i == 0) {
+      ASSERT_OK(db_->DropColumnFamily(handles_[1]));
+    } else {
+      dbfull()->CancelAllBackgroundWork(false /* wait */);
+    }
+    manual_compaction_thread.join();
+    TEST_SYNC_POINT(
+        "DBCompactionTest::CompactRangeShutdownWhileDelayed:PostManual");
+    dbfull()->TEST_WaitForCompact();
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
+
+TEST_F(DBCompactionTest, CompactRangeSkipFlushAfterDelay) {
+  // Verify that, when `CompactRangeOptions::allow_write_stall == false`,
+  // CompactRange skips its flush if the delay is long enough that the memtables
+  // existing at the beginning of the call have already been flushed.
+  const int kNumL0FilesTrigger = 4;
+  const int kNumL0FilesLimit = 8;
+  Options options = CurrentOptions();
+  options.level0_slowdown_writes_trigger = kNumL0FilesLimit;
+  options.level0_file_num_compaction_trigger = kNumL0FilesTrigger;
+  Reopen(options);
+
+  Random rnd(301);
+  // The manual flush includes the memtable that was active when CompactRange
+  // began. So it unblocks CompactRange and precludes its flush. Throughout the
+  // test, stall conditions are upheld via high L0 file count.
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::CompactRange:StallWait",
+        "DBCompactionTest::CompactRangeSkipFlushAfterDelay:PreFlush"},
+       {"DBCompactionTest::CompactRangeSkipFlushAfterDelay:PostFlush",
+        "DBImpl::CompactRange:StallWaitDone"},
+       {"DBImpl::CompactRange:StallWaitDone", "CompactionJob::Run():End"}});
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  for (int i = 0; i < kNumL0FilesLimit - 1; ++i) {
+    for (int j = 0; j < 2; ++j) {
+      ASSERT_OK(Put(Key(j), RandomString(&rnd, 1024)));
+    }
+    Flush();
+  }
+  auto manual_compaction_thread = port::Thread([this]() {
+    CompactRangeOptions cro;
+    cro.allow_write_stall = false;
+    db_->CompactRange(cro, nullptr, nullptr);
+  });
+
+  TEST_SYNC_POINT("DBCompactionTest::CompactRangeSkipFlushAfterDelay:PreFlush");
+  Put(ToString(0), RandomString(&rnd, 1024));
+  Flush();
+  Put(ToString(0), RandomString(&rnd, 1024));
+  TEST_SYNC_POINT("DBCompactionTest::CompactRangeSkipFlushAfterDelay:PostFlush");
+  manual_compaction_thread.join();
+
+  // If CompactRange's flush was skipped, the final Put above will still be
+  // in the active memtable.
+  std::string num_keys_in_memtable;
+  db_->GetProperty(DB::Properties::kNumEntriesActiveMemTable, &num_keys_in_memtable);
+  ASSERT_EQ(ToString(1), num_keys_in_memtable);
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+TEST_F(DBCompactionTest, CompactRangeFlushOverlappingMemtable) {
+  // Verify memtable only gets flushed if it contains data overlapping the range
+  // provided to `CompactRange`. Tests all kinds of overlap/non-overlap.
+  const int kNumEndpointKeys = 5;
+  std::string keys[kNumEndpointKeys] = {"a", "b", "c", "d", "e"};
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  Reopen(options);
+
+  // One extra iteration for nullptr, which means left side of interval is
+  // unbounded.
+  for (int i = 0; i <= kNumEndpointKeys; ++i) {
+    Slice begin;
+    Slice* begin_ptr;
+    if (i == 0) {
+      begin_ptr = nullptr;
+    } else {
+      begin = keys[i - 1];
+      begin_ptr = &begin;
+    }
+    // Start at `i` so right endpoint comes after left endpoint. One extra
+    // iteration for nullptr, which means right side of interval is unbounded.
+    for (int j = std::max(0, i - 1); j <= kNumEndpointKeys; ++j) {
+      Slice end;
+      Slice* end_ptr;
+      if (j == kNumEndpointKeys) {
+        end_ptr = nullptr;
+      } else {
+        end = keys[j];
+        end_ptr = &end;
+      }
+      ASSERT_OK(Put("b", "val"));
+      ASSERT_OK(Put("d", "val"));
+      CompactRangeOptions compact_range_opts;
+      ASSERT_OK(db_->CompactRange(compact_range_opts, begin_ptr, end_ptr));
+
+      uint64_t get_prop_tmp, num_memtable_entries = 0;
+      ASSERT_TRUE(db_->GetIntProperty(DB::Properties::kNumEntriesImmMemTables,
+                                      &get_prop_tmp));
+      num_memtable_entries += get_prop_tmp;
+      ASSERT_TRUE(db_->GetIntProperty(DB::Properties::kNumEntriesActiveMemTable,
+                                      &get_prop_tmp));
+      num_memtable_entries += get_prop_tmp;
+      if (begin_ptr == nullptr || end_ptr == nullptr ||
+          (i <= 4 && j >= 1 && (begin != "c" || end != "c"))) {
+        // In this case `CompactRange`'s range overlapped in some way with the
+        // memtable's range, so flush should've happened. Then "b" and "d" won't
+        // be in the memtable.
+        ASSERT_EQ(0, num_memtable_entries);
+      } else {
+        ASSERT_EQ(2, num_memtable_entries);
+        // flush anyways to prepare for next iteration
+        db_->Flush(FlushOptions());
+      }
+    }
+  }
+}
+
 INSTANTIATE_TEST_CASE_P(DBCompactionTestWithParam, DBCompactionTestWithParam,
                         ::testing::Values(std::make_tuple(1, true),
                                           std::make_tuple(1, false),

db/db_filesnapshot.cc

@@ -93,7 +93,7 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
       }
       cfd->Ref();
       mutex_.Unlock();
-      status = FlushMemTable(cfd, FlushOptions());
+      status = FlushMemTable(cfd, FlushOptions(), FlushReason::kGetLiveFiles);
       TEST_SYNC_POINT("DBImpl::GetLiveFiles:1");
       TEST_SYNC_POINT("DBImpl::GetLiveFiles:2");
       mutex_.Lock();

db/db_flush_test.cc

@@ -137,7 +137,7 @@ TEST_F(DBFlushTest, ManualFlushWithMinWriteBufferNumberToMerge) {
       {{"DBImpl::BGWorkFlush",
         "DBFlushTest::ManualFlushWithMinWriteBufferNumberToMerge:1"},
        {"DBFlushTest::ManualFlushWithMinWriteBufferNumberToMerge:2",
-        "DBImpl::FlushMemTableToOutputFile:BeforeInstallSV"}});
+        "FlushJob::WriteLevel0Table"}});
   SyncPoint::GetInstance()->EnableProcessing();
 
   ASSERT_OK(Put("key1", "value1"));
@@ -185,6 +185,26 @@ TEST_P(DBFlushDirectIOTest, DirectIO) {
   delete options.env;
 }
 
+TEST_F(DBFlushTest, FlushError) {
+  Options options;
+  std::unique_ptr<FaultInjectionTestEnv> fault_injection_env(
+      new FaultInjectionTestEnv(env_));
+  options.write_buffer_size = 100;
+  options.max_write_buffer_number = 4;
+  options.min_write_buffer_number_to_merge = 3;
+  options.disable_auto_compactions = true;
+  options.env = fault_injection_env.get();
+  Reopen(options);
+
+  ASSERT_OK(Put("key1", "value1"));
+  ASSERT_OK(Put("key2", "value2"));
+  fault_injection_env->SetFilesystemActive(false);
+  Status s = dbfull()->TEST_SwitchMemtable();
+  fault_injection_env->SetFilesystemActive(true);
+  Destroy(options);
+  ASSERT_NE(s, Status::OK());
+}
+
 INSTANTIATE_TEST_CASE_P(DBFlushDirectIOTest, DBFlushDirectIOTest,
                         testing::Bool());
 

db/db_impl.cc

@@ -258,7 +258,7 @@ void DBImpl::CancelAllBackgroundWork(bool wait) {
       if (!cfd->IsDropped() && cfd->initialized() && !cfd->mem()->IsEmpty()) {
         cfd->Ref();
         mutex_.Unlock();
-        FlushMemTable(cfd, FlushOptions());
+        FlushMemTable(cfd, FlushOptions(), FlushReason::kShutDown);
         mutex_.Lock();
         cfd->Unref();
       }
@@ -278,7 +278,7 @@ void DBImpl::CancelAllBackgroundWork(bool wait) {
   }
 }
 
-Status DBImpl::CloseImpl() {
+Status DBImpl::CloseHelper() {
   // 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))
@@ -287,6 +287,7 @@ Status DBImpl::CloseImpl() {
       env_->UnSchedule(this, Env::Priority::BOTTOM);
   int compactions_unscheduled = env_->UnSchedule(this, Env::Priority::LOW);
   int flushes_unscheduled = env_->UnSchedule(this, Env::Priority::HIGH);
+  Status ret;
   mutex_.Lock();
   bg_bottom_compaction_scheduled_ -= bottom_compactions_unscheduled;
   bg_compaction_scheduled_ -= compactions_unscheduled;
@@ -349,7 +350,18 @@ Status DBImpl::CloseImpl() {
     delete l;
   }
   for (auto& log : logs_) {
-    log.ClearWriter();
+    uint64_t log_number = log.writer->get_log_number();
+    Status s = log.ClearWriter();
+    if (!s.ok()) {
+      ROCKS_LOG_WARN(immutable_db_options_.info_log,
+                     "Unable to Sync WAL file %s with error -- %s",
+                     LogFileName(immutable_db_options_.wal_dir, log_number).c_str(),
+                     s.ToString().c_str());
+      // Retain the first error
+      if (ret.ok()) {
+        ret = s;
+      }
+    }
   }
   logs_.clear();
 
@@ -383,14 +395,25 @@ Status DBImpl::CloseImpl() {
   ROCKS_LOG_INFO(immutable_db_options_.info_log, "Shutdown complete");
   LogFlush(immutable_db_options_.info_log);
 
-  Status s = Status::OK();
   if (immutable_db_options_.info_log && own_info_log_) {
-    s = immutable_db_options_.info_log->Close();
+    Status s = immutable_db_options_.info_log->Close();
+    if (ret.ok()) {
+      ret = s;
+    }
   }
-  return s;
+  return ret;
 }
 
-DBImpl::~DBImpl() { Close(); }
+Status DBImpl::CloseImpl() {
+  return CloseHelper();
+}
+
+DBImpl::~DBImpl() {
+  if (!closed_) {
+    closed_ = true;
+    CloseHelper();
+  }
+}
 
 void DBImpl::MaybeIgnoreError(Status* s) const {
   if (s->ok() || immutable_db_options_.paranoid_checks) {
@@ -535,6 +558,7 @@ Status DBImpl::SetOptions(ColumnFamilyHandle* column_family,
 
       persist_options_status = WriteOptionsFile(
           false /*need_mutex_lock*/, true /*need_enter_write_thread*/);
+      bg_cv_.SignalAll();
     }
   }
   sv_context.Clean();
@@ -651,8 +675,9 @@ Status DBImpl::SetDBOptions(
 }
 
 // return the same level if it cannot be moved
-int DBImpl::FindMinimumEmptyLevelFitting(ColumnFamilyData* cfd,
-    const MutableCFOptions& mutable_cf_options, int level) {
+int DBImpl::FindMinimumEmptyLevelFitting(
+    ColumnFamilyData* cfd, const MutableCFOptions& /*mutable_cf_options*/,
+    int level) {
   mutex_.AssertHeld();
   const auto* vstorage = cfd->current()->storage_info();
   int minimum_level = level;
@@ -873,7 +898,7 @@ struct IterState {
   bool background_purge;
 };
 
-static void CleanupIteratorState(void* arg1, void* arg2) {
+static void CleanupIteratorState(void* arg1, void* /*arg2*/) {
   IterState* state = reinterpret_cast<IterState*>(arg1);
 
   if (state->super_version->Unref()) {
@@ -1419,6 +1444,7 @@ Status DBImpl::DropColumnFamilyImpl(ColumnFamilyHandle* column_family) {
       }
       is_snapshot_supported_ = new_is_snapshot_supported;
     }
+    bg_cv_.SignalAll();
   }
 
   if (s.ok()) {
@@ -2138,52 +2164,63 @@ Status DBImpl::DeleteFile(std::string name) {
   return status;
 }
 
-Status DBImpl::DeleteFilesInRange(ColumnFamilyHandle* column_family,
-                                  const Slice* begin, const Slice* end) {
+Status DBImpl::DeleteFilesInRanges(ColumnFamilyHandle* column_family,
+                                   const RangePtr* ranges, size_t n,
+                                   bool include_end) {
   Status status;
   auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
   ColumnFamilyData* cfd = cfh->cfd();
   VersionEdit edit;
-  std::vector<FileMetaData*> deleted_files;
+  std::set<FileMetaData*> deleted_files;
   JobContext job_context(next_job_id_.fetch_add(1), true);
   {
     InstrumentedMutexLock l(&mutex_);
     Version* input_version = cfd->current();
 
     auto* vstorage = input_version->storage_info();
-    for (int i = 1; i < cfd->NumberLevels(); i++) {
-      if (vstorage->LevelFiles(i).empty() ||
-          !vstorage->OverlapInLevel(i, begin, end)) {
-        continue;
-      }
-      std::vector<FileMetaData*> level_files;
-      InternalKey begin_storage, end_storage, *begin_key, *end_key;
-      if (begin == nullptr) {
-        begin_key = nullptr;
-      } else {
-        begin_storage.SetMinPossibleForUserKey(*begin);
-        begin_key = &begin_storage;
-      }
-      if (end == nullptr) {
-        end_key = nullptr;
-      } else {
-        end_storage.SetMaxPossibleForUserKey(*end);
-        end_key = &end_storage;
-      }
-
-      vstorage->GetCleanInputsWithinInterval(i, begin_key, end_key,
-                                             &level_files, -1 /* hint_index */,
-                                             nullptr /* file_index */);
-      FileMetaData* level_file;
-      for (uint32_t j = 0; j < level_files.size(); j++) {
-        level_file = level_files[j];
-        if (level_file->being_compacted) {
+    for (size_t r = 0; r < n; r++) {
+      auto begin = ranges[r].start, end = ranges[r].limit;
+      for (int i = 1; i < cfd->NumberLevels(); i++) {
+        if (vstorage->LevelFiles(i).empty() ||
+            !vstorage->OverlapInLevel(i, begin, end)) {
           continue;
         }
-        edit.SetColumnFamily(cfd->GetID());
-        edit.DeleteFile(i, level_file->fd.GetNumber());
-        deleted_files.push_back(level_file);
-        level_file->being_compacted = true;
+        std::vector<FileMetaData*> level_files;
+        InternalKey begin_storage, end_storage, *begin_key, *end_key;
+        if (begin == nullptr) {
+          begin_key = nullptr;
+        } else {
+          begin_storage.SetMinPossibleForUserKey(*begin);
+          begin_key = &begin_storage;
+        }
+        if (end == nullptr) {
+          end_key = nullptr;
+        } else {
+          end_storage.SetMaxPossibleForUserKey(*end);
+          end_key = &end_storage;
+        }
+
+        vstorage->GetCleanInputsWithinInterval(i, begin_key, end_key,
+                                               &level_files, -1 /* hint_index */,
+                                               nullptr /* file_index */);
+        FileMetaData* level_file;
+        for (uint32_t j = 0; j < level_files.size(); j++) {
+          level_file = level_files[j];
+          if (level_file->being_compacted) {
+            continue;
+          }
+          if (deleted_files.find(level_file) != deleted_files.end()) {
+            continue;
+          }
+          if (!include_end && end != nullptr &&
+              cfd->user_comparator()->Compare(level_file->largest.user_key(), *end) == 0) {
+            continue;
+          }
+          edit.SetColumnFamily(cfd->GetID());
+          edit.DeleteFile(i, level_file->fd.GetNumber());
+          deleted_files.insert(level_file);
+          level_file->being_compacted = true;
+        }
       }
     }
     if (edit.GetDeletedFiles().empty()) {
@@ -2299,31 +2336,31 @@ Status DBImpl::GetDbIdentity(std::string& identity) const {
 }
 
 // Default implementation -- returns not supported status
-Status DB::CreateColumnFamily(const ColumnFamilyOptions& cf_options,
-                              const std::string& column_family_name,
-                              ColumnFamilyHandle** handle) {
+Status DB::CreateColumnFamily(const ColumnFamilyOptions& /*cf_options*/,
+                              const std::string& /*column_family_name*/,
+                              ColumnFamilyHandle** /*handle*/) {
   return Status::NotSupported("");
 }
 
 Status DB::CreateColumnFamilies(
-    const ColumnFamilyOptions& cf_options,
-    const std::vector<std::string>& column_family_names,
-    std::vector<ColumnFamilyHandle*>* handles) {
+    const ColumnFamilyOptions& /*cf_options*/,
+    const std::vector<std::string>& /*column_family_names*/,
+    std::vector<ColumnFamilyHandle*>* /*handles*/) {
   return Status::NotSupported("");
 }
 
 Status DB::CreateColumnFamilies(
-    const std::vector<ColumnFamilyDescriptor>& column_families,
-    std::vector<ColumnFamilyHandle*>* handles) {
+    const std::vector<ColumnFamilyDescriptor>& /*column_families*/,
+    std::vector<ColumnFamilyHandle*>* /*handles*/) {
   return Status::NotSupported("");
 }
 
-Status DB::DropColumnFamily(ColumnFamilyHandle* column_family) {
+Status DB::DropColumnFamily(ColumnFamilyHandle* /*column_family*/) {
   return Status::NotSupported("");
 }
 
 Status DB::DropColumnFamilies(
-    const std::vector<ColumnFamilyHandle*>& column_families) {
+    const std::vector<ColumnFamilyHandle*>& /*column_families*/) {
   return Status::NotSupported("");
 }
 
@@ -2796,12 +2833,14 @@ Status DBImpl::IngestExternalFile(
     // Figure out if we need to flush the memtable first
     if (status.ok()) {
       bool need_flush = false;
-      status = ingestion_job.NeedsFlush(&need_flush);
+      status = ingestion_job.NeedsFlush(&need_flush, cfd->GetSuperVersion());
       TEST_SYNC_POINT_CALLBACK("DBImpl::IngestExternalFile:NeedFlush",
                                &need_flush);
       if (status.ok() && need_flush) {
         mutex_.Unlock();
-        status = FlushMemTable(cfd, FlushOptions(), true /* writes_stopped */);
+        status = FlushMemTable(cfd, FlushOptions(),
+                               FlushReason::kExternalFileIngestion,
+                               true /* writes_stopped */);
         mutex_.Lock();
       }
     }

db/db_impl.h

@@ -14,7 +14,6 @@
 #include <limits>
 #include <list>
 #include <map>
-#include <queue>
 #include <set>
 #include <string>
 #include <utility>
@@ -247,8 +246,9 @@ class DBImpl : public DB {
       const TransactionLogIterator::ReadOptions&
           read_options = TransactionLogIterator::ReadOptions()) override;
   virtual Status DeleteFile(std::string name) override;
-  Status DeleteFilesInRange(ColumnFamilyHandle* column_family,
-                            const Slice* begin, const Slice* end);
+  Status DeleteFilesInRanges(ColumnFamilyHandle* column_family,
+                             const RangePtr* ranges, size_t n,
+                             bool include_end = true);
 
   virtual void GetLiveFilesMetaData(
       std::vector<LiveFileMetaData>* metadata) override;
@@ -432,6 +432,8 @@ class DBImpl : public DB {
 
   uint64_t TEST_FindMinLogContainingOutstandingPrep();
   uint64_t TEST_FindMinPrepLogReferencedByMemTable();
+  size_t TEST_PreparedSectionCompletedSize();
+  size_t TEST_LogsWithPrepSize();
 
   int TEST_BGCompactionsAllowed() const;
   int TEST_BGFlushesAllowed() const;
@@ -672,10 +674,14 @@ class DBImpl : public DB {
   // batch which will be written to memtable later during the commit, and in
   // WritePrepared it is guaranteed since it will be used only for WAL markers
   // which will never be written to memtable.
+  // batch_cnt is expected to be non-zero in seq_per_batch mode and indicates
+  // the number of sub-patches. A sub-patch is a subset of the write batch that
+  // does not have duplicate keys.
   Status WriteImpl(const WriteOptions& options, WriteBatch* updates,
                    WriteCallback* callback = nullptr,
                    uint64_t* log_used = nullptr, uint64_t log_ref = 0,
                    bool disable_memtable = false, uint64_t* seq_used = nullptr,
+                   size_t batch_cnt = 0,
                    PreReleaseCallback* pre_release_callback = nullptr);
 
   Status PipelinedWriteImpl(const WriteOptions& options, WriteBatch* updates,
@@ -684,10 +690,13 @@ class DBImpl : public DB {
                             bool disable_memtable = false,
                             uint64_t* seq_used = nullptr);
 
+  // batch_cnt is expected to be non-zero in seq_per_batch mode and indicates
+  // the number of sub-patches. A sub-patch is a subset of the write batch that
+  // does not have duplicate keys.
   Status WriteImplWALOnly(const WriteOptions& options, WriteBatch* updates,
                           WriteCallback* callback = nullptr,
                           uint64_t* log_used = nullptr, uint64_t log_ref = 0,
-                          uint64_t* seq_used = nullptr,
+                          uint64_t* seq_used = nullptr, size_t batch_cnt = 0,
                           PreReleaseCallback* pre_release_callback = nullptr);
 
   uint64_t FindMinLogContainingOutstandingPrep();
@@ -696,6 +705,9 @@ class DBImpl : public DB {
   // The writer must be the leader in write_thread_ and holding mutex_
   Status WriteRecoverableState();
 
+  // Actual implementation of Close()
+  Status CloseImpl();
+
  private:
   friend class DB;
   friend class InternalStats;
@@ -811,7 +823,7 @@ class DBImpl : public DB {
 
   // Force current memtable contents to be flushed.
   Status FlushMemTable(ColumnFamilyData* cfd, const FlushOptions& options,
-                       bool writes_stopped = false);
+                       FlushReason flush_reason, bool writes_stopped = false);
 
   // Wait for memtable flushed.
   // If flush_memtable_id is non-null, wait until the memtable with the ID
@@ -873,7 +885,7 @@ class DBImpl : public DB {
   ColumnFamilyData* GetColumnFamilyDataByName(const std::string& cf_name);
 
   void MaybeScheduleFlushOrCompaction();
-  void SchedulePendingFlush(ColumnFamilyData* cfd);
+  void SchedulePendingFlush(ColumnFamilyData* cfd, FlushReason flush_reason);
   void SchedulePendingCompaction(ColumnFamilyData* cfd);
   void SchedulePendingPurge(std::string fname, FileType type, uint64_t number,
                             uint32_t path_id, int job_id);
@@ -912,7 +924,7 @@ class DBImpl : public DB {
   // helper functions for adding and removing from flush & compaction queues
   void AddToCompactionQueue(ColumnFamilyData* cfd);
   ColumnFamilyData* PopFirstFromCompactionQueue();
-  void AddToFlushQueue(ColumnFamilyData* cfd);
+  void AddToFlushQueue(ColumnFamilyData* cfd, FlushReason flush_reason);
   ColumnFamilyData* PopFirstFromFlushQueue();
 
   // helper function to call after some of the logs_ were synced
@@ -922,8 +934,7 @@ class DBImpl : public DB {
 
   uint64_t GetMaxTotalWalSize() const;
 
-  // Actual implementation of Close()
-  virtual Status CloseImpl();
+  Status CloseHelper();
 
   // table_cache_ provides its own synchronization
   std::shared_ptr<Cache> table_cache_;
@@ -955,6 +966,8 @@ class DBImpl : public DB {
   // * whenever num_running_ingest_file_ goes to 0.
   // * whenever pending_purge_obsolete_files_ goes to 0.
   // * whenever disable_delete_obsolete_files_ goes to 0.
+  // * whenever SetOptions successfully updates options.
+  // * whenever a column family is dropped.
   InstrumentedCondVar bg_cv_;
   // Writes are protected by locking both mutex_ and log_write_mutex_, and reads
   // must be under either mutex_ or log_write_mutex_. Since after ::Open,
@@ -992,9 +1005,11 @@ class DBImpl : public DB {
       writer = nullptr;
       return w;
     }
-    void ClearWriter() {
+    Status ClearWriter() {
+      Status s = writer->WriteBuffer();
       delete writer;
       writer = nullptr;
+      return s;
     }
 
     uint64_t number;
@@ -1284,27 +1299,33 @@ class DBImpl : public DB {
   // Indicate DB was opened successfully
   bool opened_successfully_;
 
-  // minimum log number still containing prepared data.
+  // REQUIRES: logs_with_prep_mutex_ held
+  //
+  // sorted list of log numbers still containing prepared data.
   // this is used by FindObsoleteFiles to determine which
   // flushed logs we must keep around because they still
-  // contain prepared data which has not been flushed or rolled back
-  std::priority_queue<uint64_t, std::vector<uint64_t>, std::greater<uint64_t>>
-      min_log_with_prep_;
+  // contain prepared data which has not been committed or rolled back
+  struct LogCnt {
+    uint64_t log;  // the log number
+    uint64_t cnt;  // number of prepared sections in the log
+  };
+  std::vector<LogCnt> logs_with_prep_;
+  std::mutex logs_with_prep_mutex_;
 
-  // to be used in conjunction with min_log_with_prep_.
+  // REQUIRES: prepared_section_completed_mutex_ held
+  //
+  // to be used in conjunction with logs_with_prep_.
   // once a transaction with data in log L is committed or rolled back
-  // rather than removing the value from the heap we add that value
-  // to prepared_section_completed_ which maps LOG -> instance_count
-  // since a log could contain multiple prepared sections
+  // rather than updating logs_with_prep_ directly we keep track of that
+  // in prepared_section_completed_ which maps LOG -> instance_count. This helps
+  // avoiding contention between a commit thread and the prepare threads.
   //
   // when trying to determine the minimum log still active we first
-  // consult min_log_with_prep_. while that root value maps to
-  // a value > 0 in prepared_section_completed_ we decrement the
-  // instance_count for that log and pop the root value in
-  // min_log_with_prep_. This will work the same as a min_heap
-  // where we are deleteing arbitrary elements and the up heaping.
+  // consult logs_with_prep_. while that root value maps to
+  // an equal value in prepared_section_completed_ we erase the log from
+  // both logs_with_prep_ and prepared_section_completed_.
   std::unordered_map<uint64_t, uint64_t> prepared_section_completed_;
-  std::mutex prep_heap_mutex_;
+  std::mutex prepared_section_completed_mutex_;
 
   // Callback for compaction to check if a key is visible to a snapshot.
   // REQUIRES: mutex held
@@ -1353,8 +1374,8 @@ class DBImpl : public DB {
     return Env::WLTH_SHORT;
   }
 
-  // When set, we use a seprate queue for writes that dont write to memtable. In
-  // 2PC these are the writes at Prepare phase.
+  // When set, we use a separate queue for writes that dont write to memtable.
+  // In 2PC these are the writes at Prepare phase.
   const bool two_write_queues_;
   const bool manual_wal_flush_;
   // Increase the sequence number after writing each batch, whether memtable is

db/db_impl_compaction_flush.cc

@@ -23,6 +23,7 @@
 #include "util/sync_point.h"
 
 namespace rocksdb {
+
 Status DBImpl::SyncClosedLogs(JobContext* job_context) {
   TEST_SYNC_POINT("DBImpl::SyncClosedLogs:Start");
   mutex_.AssertHeld();
@@ -134,7 +135,6 @@ Status DBImpl::FlushMemTableToOutputFile(
   }
 
   if (s.ok()) {
-    TEST_SYNC_POINT("DBImpl::FlushMemTableToOutputFile:BeforeInstallSV");
     InstallSuperVersionAndScheduleWork(cfd, &job_context->superversion_context,
                                        mutable_cf_options);
     if (made_progress) {
@@ -223,6 +223,7 @@ void DBImpl::NotifyOnFlushBegin(ColumnFamilyData* cfd, FileMetaData* file_meta,
     info.smallest_seqno = file_meta->smallest_seqno;
     info.largest_seqno = file_meta->largest_seqno;
     info.table_properties = prop;
+    info.flush_reason = cfd->GetFlushReason();
     for (auto listener : immutable_db_options_.listeners) {
       listener->OnFlushBegin(this, info);
     }
@@ -267,6 +268,7 @@ void DBImpl::NotifyOnFlushCompleted(ColumnFamilyData* cfd,
     info.smallest_seqno = file_meta->smallest_seqno;
     info.largest_seqno = file_meta->largest_seqno;
     info.table_properties = prop;
+    info.flush_reason = cfd->GetFlushReason();
     for (auto listener : immutable_db_options_.listeners) {
       listener->OnFlushCompleted(this, info);
     }
@@ -288,10 +290,75 @@ Status DBImpl::CompactRange(const CompactRangeOptions& options,
   auto cfd = cfh->cfd();
   bool exclusive = options.exclusive_manual_compaction;
 
-  Status s = FlushMemTable(cfd, FlushOptions());
-  if (!s.ok()) {
-    LogFlush(immutable_db_options_.info_log);
-    return s;
+  bool flush_needed = true;
+  if (begin != nullptr && end != nullptr) {
+    // TODO(ajkr): We could also optimize away the flush in certain cases where
+    // one/both sides of the interval are unbounded. But it requires more
+    // changes to RangesOverlapWithMemtables.
+    Range range(*begin, *end);
+    SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+    cfd->RangesOverlapWithMemtables({range}, super_version, &flush_needed);
+    CleanupSuperVersion(super_version);
+  }
+
+  if (!options.allow_write_stall && flush_needed) {
+    InstrumentedMutexLock l(&mutex_);
+    uint64_t orig_active_memtable_id = cfd->mem()->GetID();
+    WriteStallCondition write_stall_condition = WriteStallCondition::kNormal;
+    do {
+      if (write_stall_condition != WriteStallCondition::kNormal) {
+        TEST_SYNC_POINT("DBImpl::CompactRange:StallWait");
+        ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                       "[%s] CompactRange waiting on stall conditions to clear",
+                       cfd->GetName().c_str());
+        bg_cv_.Wait();
+      }
+      if (cfd->IsDropped() || shutting_down_.load(std::memory_order_acquire)) {
+        return Status::ShutdownInProgress();
+      }
+
+      uint64_t earliest_memtable_id =
+          std::min(cfd->mem()->GetID(), cfd->imm()->GetEarliestMemTableID());
+      if (earliest_memtable_id > orig_active_memtable_id) {
+        // We waited so long that the memtable we were originally waiting on was
+        // flushed.
+        flush_needed = false;
+        break;
+      }
+
+      const auto& mutable_cf_options = *cfd->GetLatestMutableCFOptions();
+      const auto* vstorage = cfd->current()->storage_info();
+
+      // Skip stalling check if we're below auto-flush and auto-compaction
+      // triggers. If it stalled in these conditions, that'd mean the stall
+      // triggers are so low that stalling is needed for any background work. In
+      // that case we shouldn't wait since background work won't be scheduled.
+      if (cfd->imm()->NumNotFlushed() <
+              cfd->ioptions()->min_write_buffer_number_to_merge &&
+          vstorage->l0_delay_trigger_count() <
+              mutable_cf_options.level0_file_num_compaction_trigger) {
+        break;
+      }
+
+      // check whether one extra immutable memtable or an extra L0 file would
+      // cause write stalling mode to be entered. It could still enter stall
+      // mode due to pending compaction bytes, but that's less common
+      write_stall_condition =
+          ColumnFamilyData::GetWriteStallConditionAndCause(
+              cfd->imm()->NumNotFlushed() + 1,
+              vstorage->l0_delay_trigger_count() + 1,
+              vstorage->estimated_compaction_needed_bytes(), mutable_cf_options)
+              .first;
+    } while (write_stall_condition != WriteStallCondition::kNormal);
+  }
+  TEST_SYNC_POINT("DBImpl::CompactRange:StallWaitDone");
+  Status s;
+  if (flush_needed) {
+    s = FlushMemTable(cfd, FlushOptions(), FlushReason::kManualCompaction);
+    if (!s.ok()) {
+      LogFlush(immutable_db_options_.info_log);
+      return s;
+    }
   }
 
   int max_level_with_files = 0;
@@ -797,7 +864,7 @@ int DBImpl::NumberLevels(ColumnFamilyHandle* column_family) {
   return cfh->cfd()->NumberLevels();
 }
 
-int DBImpl::MaxMemCompactionLevel(ColumnFamilyHandle* column_family) {
+int DBImpl::MaxMemCompactionLevel(ColumnFamilyHandle* /*column_family*/) {
   return 0;
 }
 
@@ -813,7 +880,8 @@ Status DBImpl::Flush(const FlushOptions& flush_options,
   auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
   ROCKS_LOG_INFO(immutable_db_options_.info_log, "[%s] Manual flush start.",
                  cfh->GetName().c_str());
-  Status s = FlushMemTable(cfh->cfd(), flush_options);
+  Status s =
+      FlushMemTable(cfh->cfd(), flush_options, FlushReason::kManualCompaction);
   ROCKS_LOG_INFO(immutable_db_options_.info_log,
                  "[%s] Manual flush finished, status: %s\n",
                  cfh->GetName().c_str(), s.ToString().c_str());
@@ -950,7 +1018,7 @@ Status DBImpl::RunManualCompaction(ColumnFamilyData* cfd, int input_level,
 
 Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
                              const FlushOptions& flush_options,
-                             bool writes_stopped) {
+                             FlushReason flush_reason, bool writes_stopped) {
   Status s;
   uint64_t flush_memtable_id = 0;
   {
@@ -979,7 +1047,7 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
     cfd->imm()->FlushRequested();
 
     // schedule flush
-    SchedulePendingFlush(cfd);
+    SchedulePendingFlush(cfd, flush_reason);
     MaybeScheduleFlushOrCompaction();
   }
 
@@ -1135,11 +1203,12 @@ ColumnFamilyData* DBImpl::PopFirstFromCompactionQueue() {
   return cfd;
 }
 
-void DBImpl::AddToFlushQueue(ColumnFamilyData* cfd) {
+void DBImpl::AddToFlushQueue(ColumnFamilyData* cfd, FlushReason flush_reason) {
   assert(!cfd->pending_flush());
   cfd->Ref();
   flush_queue_.push_back(cfd);
   cfd->set_pending_flush(true);
+  cfd->SetFlushReason(flush_reason);
 }
 
 ColumnFamilyData* DBImpl::PopFirstFromFlushQueue() {
@@ -1148,12 +1217,14 @@ ColumnFamilyData* DBImpl::PopFirstFromFlushQueue() {
   flush_queue_.pop_front();
   assert(cfd->pending_flush());
   cfd->set_pending_flush(false);
+  // TODO: need to unset flush reason?
   return cfd;
 }
 
-void DBImpl::SchedulePendingFlush(ColumnFamilyData* cfd) {
+void DBImpl::SchedulePendingFlush(ColumnFamilyData* cfd,
+                                  FlushReason flush_reason) {
   if (!cfd->pending_flush() && cfd->imm()->IsFlushPending()) {
-    AddToFlushQueue(cfd);
+    AddToFlushQueue(cfd, flush_reason);
     ++unscheduled_flushes_;
   }
 }
@@ -1696,7 +1767,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
     env_->Schedule(&DBImpl::BGWorkBottomCompaction, ca, Env::Priority::BOTTOM,
                    this, &DBImpl::UnscheduleCallback);
   } else {
-    int output_level  __attribute__((unused));
+    int output_level  __attribute__((__unused__));
     output_level = c->output_level();
     TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:NonTrivial",
                              &output_level);
@@ -1930,7 +2001,7 @@ void DBImpl::InstallSuperVersionAndScheduleWork(
 
   // Whenever we install new SuperVersion, we might need to issue new flushes or
   // compactions.
-  SchedulePendingFlush(cfd);
+  SchedulePendingFlush(cfd, FlushReason::kSuperVersionChange);
   SchedulePendingCompaction(cfd);
   MaybeScheduleFlushOrCompaction();
 

db/db_impl_debug.cc

@@ -103,7 +103,7 @@ Status DBImpl::TEST_FlushMemTable(bool wait, ColumnFamilyHandle* cfh) {
     auto cfhi = reinterpret_cast<ColumnFamilyHandleImpl*>(cfh);
     cfd = cfhi->cfd();
   }
-  return FlushMemTable(cfd, fo);
+  return FlushMemTable(cfd, fo, FlushReason::kTest);
 }
 
 Status DBImpl::TEST_WaitForFlushMemTable(ColumnFamilyHandle* column_family) {
@@ -186,6 +186,12 @@ uint64_t DBImpl::TEST_FindMinLogContainingOutstandingPrep() {
   return FindMinLogContainingOutstandingPrep();
 }
 
+size_t DBImpl::TEST_PreparedSectionCompletedSize() {
+  return prepared_section_completed_.size();
+}
+
+size_t DBImpl::TEST_LogsWithPrepSize() { return logs_with_prep_.size(); }
+
 uint64_t DBImpl::TEST_FindMinPrepLogReferencedByMemTable() {
   return FindMinPrepLogReferencedByMemTable();
 }

db/db_impl_files.cc

@@ -48,58 +48,61 @@ uint64_t DBImpl::FindMinPrepLogReferencedByMemTable() {
   return min_log;
 }
 
-// TODO(myabandeh): Avoid using locks
 void DBImpl::MarkLogAsHavingPrepSectionFlushed(uint64_t log) {
   assert(log != 0);
-  std::lock_guard<std::mutex> lock(prep_heap_mutex_);
+  std::lock_guard<std::mutex> lock(prepared_section_completed_mutex_);
   auto it = prepared_section_completed_.find(log);
-  assert(it != prepared_section_completed_.end());
-  it->second += 1;
+  if (UNLIKELY(it == prepared_section_completed_.end())) {
+    prepared_section_completed_[log] = 1;
+  } else {
+    it->second += 1;
+  }
 }
 
-// TODO(myabandeh): Avoid using locks
 void DBImpl::MarkLogAsContainingPrepSection(uint64_t log) {
   assert(log != 0);
-  std::lock_guard<std::mutex> lock(prep_heap_mutex_);
-  min_log_with_prep_.push(log);
-  auto it = prepared_section_completed_.find(log);
-  if (it == prepared_section_completed_.end()) {
-    prepared_section_completed_[log] = 0;
+  std::lock_guard<std::mutex> lock(logs_with_prep_mutex_);
+
+  auto rit = logs_with_prep_.rbegin();
+  bool updated = false;
+  // Most probabely the last log is the one that is being marked for
+  // having a prepare section; so search from the end.
+  for (; rit != logs_with_prep_.rend() && rit->log >= log; ++rit) {
+    if (rit->log == log) {
+      rit->cnt++;
+      updated = true;
+      break;
+    }
+  }
+  if (!updated) {
+    // We are either at the start, or at a position with rit->log < log
+    logs_with_prep_.insert(rit.base(), {log, 1});
   }
 }
 
 uint64_t DBImpl::FindMinLogContainingOutstandingPrep() {
-
-  if (!allow_2pc()) {
-    return 0;
-  }
-
-  std::lock_guard<std::mutex> lock(prep_heap_mutex_);
-  uint64_t min_log = 0;
-
-  // first we look in the prepared heap where we keep
-  // track of transactions that have been prepared (written to WAL)
-  // but not yet committed.
-  while (!min_log_with_prep_.empty()) {
-    min_log = min_log_with_prep_.top();
-
-    auto it = prepared_section_completed_.find(min_log);
-
-    // value was marked as 'deleted' from heap
-    if (it != prepared_section_completed_.end() && it->second > 0) {
-      it->second -= 1;
-      min_log_with_prep_.pop();
-
-      // back to squere one...
-      min_log = 0;
-      continue;
-    } else {
-      // found a valid value
-      break;
+  std::lock_guard<std::mutex> lock(logs_with_prep_mutex_);
+  auto it = logs_with_prep_.begin();
+  // start with the smallest log
+  for (; it != logs_with_prep_.end();) {
+    auto min_log = it->log;
+    {
+      std::lock_guard<std::mutex> lock2(prepared_section_completed_mutex_);
+      auto completed_it = prepared_section_completed_.find(min_log);
+      if (completed_it == prepared_section_completed_.end() ||
+          completed_it->second < it->cnt) {
+        return min_log;
+      }
+      assert(completed_it != prepared_section_completed_.end() &&
+             completed_it->second == it->cnt);
+      prepared_section_completed_.erase(completed_it);
     }
+    // erase from beigning in vector is not efficient but this function is not
+    // on the fast path.
+    it = logs_with_prep_.erase(it);
   }
-
-  return min_log;
+  // no such log found
+  return 0;
 }
 
 uint64_t DBImpl::MinLogNumberToKeep() {

db/db_impl_open.cc

@@ -356,6 +356,29 @@ Status DBImpl::Recover(
       assert(s.IsIOError());
       return s;
     }
+    // Verify compatibility of env_options_ and filesystem
+    {
+      unique_ptr<RandomAccessFile> idfile;
+      EnvOptions customized_env(env_options_);
+      customized_env.use_direct_reads |=
+          immutable_db_options_.use_direct_io_for_flush_and_compaction;
+      s = env_->NewRandomAccessFile(IdentityFileName(dbname_), &idfile,
+                                    customized_env);
+      if (!s.ok()) {
+        const char* error_msg = s.ToString().c_str();
+        // Check if unsupported Direct I/O is the root cause
+        customized_env.use_direct_reads = false;
+        s = env_->NewRandomAccessFile(IdentityFileName(dbname_), &idfile,
+                                      customized_env);
+        if (s.ok()) {
+          return Status::InvalidArgument(
+              "Direct I/O is not supported by the specified DB.");
+        } else {
+          return Status::InvalidArgument(
+              "Found options incompatible with filesystem", error_msg);
+        }
+      }
+    }
   }
 
   Status s = versions_->Recover(column_families, read_only);

db/db_impl_readonly.cc

@@ -107,7 +107,7 @@ Status DBImplReadOnly::NewIterators(
 }
 
 Status DB::OpenForReadOnly(const Options& options, const std::string& dbname,
-                           DB** dbptr, bool error_if_log_file_exist) {
+                           DB** dbptr, bool /*error_if_log_file_exist*/) {
   *dbptr = nullptr;
 
   // Try to first open DB as fully compacted DB

db/db_impl_readonly.h

@@ -36,46 +36,47 @@ class DBImplReadOnly : public DBImpl {
       std::vector<Iterator*>* iterators) override;
 
   using DBImpl::Put;
-  virtual Status Put(const WriteOptions& options,
-                     ColumnFamilyHandle* column_family, const Slice& key,
-                     const Slice& value) override {
+  virtual Status Put(const WriteOptions& /*options*/,
+                     ColumnFamilyHandle* /*column_family*/,
+                     const Slice& /*key*/, const Slice& /*value*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
   using DBImpl::Merge;
-  virtual Status Merge(const WriteOptions& options,
-                       ColumnFamilyHandle* column_family, const Slice& key,
-                       const Slice& value) override {
+  virtual Status Merge(const WriteOptions& /*options*/,
+                       ColumnFamilyHandle* /*column_family*/,
+                       const Slice& /*key*/, const Slice& /*value*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
   using DBImpl::Delete;
-  virtual Status Delete(const WriteOptions& options,
-                        ColumnFamilyHandle* column_family,
-                        const Slice& key) override {
+  virtual Status Delete(const WriteOptions& /*options*/,
+                        ColumnFamilyHandle* /*column_family*/,
+                        const Slice& /*key*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
   using DBImpl::SingleDelete;
-  virtual Status SingleDelete(const WriteOptions& options,
-                              ColumnFamilyHandle* column_family,
-                              const Slice& key) override {
+  virtual Status SingleDelete(const WriteOptions& /*options*/,
+                              ColumnFamilyHandle* /*column_family*/,
+                              const Slice& /*key*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
-  virtual Status Write(const WriteOptions& options,
-                       WriteBatch* updates) override {
+  virtual Status Write(const WriteOptions& /*options*/,
+                       WriteBatch* /*updates*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
   using DBImpl::CompactRange;
-  virtual Status CompactRange(const CompactRangeOptions& options,
-                              ColumnFamilyHandle* column_family,
-                              const Slice* begin, const Slice* end) override {
+  virtual Status CompactRange(const CompactRangeOptions& /*options*/,
+                              ColumnFamilyHandle* /*column_family*/,
+                              const Slice* /*begin*/,
+                              const Slice* /*end*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
 
   using DBImpl::CompactFiles;
   virtual Status CompactFiles(
-      const CompactionOptions& compact_options,
-      ColumnFamilyHandle* column_family,
-      const std::vector<std::string>& input_file_names,
-      const int output_level, const int output_path_id = -1) override {
+      const CompactionOptions& /*compact_options*/,
+      ColumnFamilyHandle* /*column_family*/,
+      const std::vector<std::string>& /*input_file_names*/,
+      const int /*output_level*/, const int /*output_path_id*/ = -1) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
 
@@ -83,18 +84,18 @@ class DBImplReadOnly : public DBImpl {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
 
-  virtual Status EnableFileDeletions(bool force) override {
+  virtual Status EnableFileDeletions(bool /*force*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
   virtual Status GetLiveFiles(std::vector<std::string>&,
-                              uint64_t* manifest_file_size,
-                              bool flush_memtable = true) override {
+                              uint64_t* /*manifest_file_size*/,
+                              bool /*flush_memtable*/ = true) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
 
   using DBImpl::Flush;
-  virtual Status Flush(const FlushOptions& options,
-                       ColumnFamilyHandle* column_family) override {
+  virtual Status Flush(const FlushOptions& /*options*/,
+                       ColumnFamilyHandle* /*column_family*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
 
@@ -105,9 +106,9 @@ class DBImplReadOnly : public DBImpl {
 
   using DB::IngestExternalFile;
   virtual Status IngestExternalFile(
-      ColumnFamilyHandle* column_family,
-      const std::vector<std::string>& external_files,
-      const IngestExternalFileOptions& ingestion_options) override {
+      ColumnFamilyHandle* /*column_family*/,
+      const std::vector<std::string>& /*external_files*/,
+      const IngestExternalFileOptions& /*ingestion_options*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
 

db/db_impl_write.cc

@@ -64,7 +64,9 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
                          WriteBatch* my_batch, WriteCallback* callback,
                          uint64_t* log_used, uint64_t log_ref,
                          bool disable_memtable, uint64_t* seq_used,
+                         size_t batch_cnt,
                          PreReleaseCallback* pre_release_callback) {
+  assert(!seq_per_batch_ || batch_cnt != 0);
   if (my_batch == nullptr) {
     return Status::Corruption("Batch is nullptr!");
   }
@@ -76,6 +78,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
         "pipelined_writes is not compatible with concurrent prepares");
   }
   if (seq_per_batch_ && immutable_db_options_.enable_pipelined_write) {
+    // TODO(yiwu): update pipeline write with seq_per_batch and batch_cnt
     return Status::NotSupported(
         "pipelined_writes is not compatible with seq_per_batch");
   }
@@ -93,7 +96,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
 
   if (two_write_queues_ && disable_memtable) {
     return WriteImplWALOnly(write_options, my_batch, callback, log_used,
-                            log_ref, seq_used, pre_release_callback);
+                            log_ref, seq_used, batch_cnt, pre_release_callback);
   }
 
   if (immutable_db_options_.enable_pipelined_write) {
@@ -103,7 +106,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
 
   PERF_TIMER_GUARD(write_pre_and_post_process_time);
   WriteThread::Writer w(write_options, my_batch, callback, log_ref,
-                        disable_memtable, pre_release_callback);
+                        disable_memtable, batch_cnt, pre_release_callback);
 
   if (!write_options.disableWAL) {
     RecordTick(stats_, WRITE_WITH_WAL);
@@ -122,7 +125,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
       w.status = WriteBatchInternal::InsertInto(
           &w, w.sequence, &column_family_memtables, &flush_scheduler_,
           write_options.ignore_missing_column_families, 0 /*log_number*/, this,
-          true /*concurrent_memtable_writes*/, seq_per_batch_);
+          true /*concurrent_memtable_writes*/, seq_per_batch_, w.batch_cnt);
     }
 
     if (write_thread_.CompleteParallelMemTableWriter(&w)) {
@@ -214,7 +217,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
     uint64_t total_byte_size = 0;
     for (auto* writer : write_group) {
       if (writer->CheckCallback(this)) {
-        valid_batches++;
+        valid_batches += writer->batch_cnt;
         if (writer->ShouldWriteToMemtable()) {
           total_count += WriteBatchInternal::Count(writer->batch);
           parallel = parallel && !writer->batch->HasMerge();
@@ -303,7 +306,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
           }
           writer->sequence = next_sequence;
           if (seq_per_batch_) {
-            next_sequence++;
+            assert(writer->batch_cnt);
+            next_sequence += writer->batch_cnt;
           } else if (writer->ShouldWriteToMemtable()) {
             next_sequence += WriteBatchInternal::Count(writer->batch);
           }
@@ -323,7 +327,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
           w.status = WriteBatchInternal::InsertInto(
               &w, w.sequence, &column_family_memtables, &flush_scheduler_,
               write_options.ignore_missing_column_families, 0 /*log_number*/,
-              this, true /*concurrent_memtable_writes*/, seq_per_batch_);
+              this, true /*concurrent_memtable_writes*/, seq_per_batch_,
+              w.batch_cnt);
         }
       }
       if (seq_used != nullptr) {
@@ -515,12 +520,13 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
 Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
                                 WriteBatch* my_batch, WriteCallback* callback,
                                 uint64_t* log_used, uint64_t log_ref,
-                                uint64_t* seq_used,
+                                uint64_t* seq_used, size_t batch_cnt,
                                 PreReleaseCallback* pre_release_callback) {
   Status status;
   PERF_TIMER_GUARD(write_pre_and_post_process_time);
   WriteThread::Writer w(write_options, my_batch, callback, log_ref,
-                        true /* disable_memtable */, pre_release_callback);
+                        true /* disable_memtable */, batch_cnt,
+                        pre_release_callback);
   RecordTick(stats_, WRITE_WITH_WAL);
   StopWatch write_sw(env_, immutable_db_options_.statistics.get(), DB_WRITE);
 
@@ -576,7 +582,15 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
   PERF_TIMER_GUARD(write_wal_time);
   // LastAllocatedSequence is increased inside WriteToWAL under
   // wal_write_mutex_ to ensure ordered events in WAL
-  size_t seq_inc = seq_per_batch_ ? write_group.size : 0 /*total_count*/;
+  size_t seq_inc = 0 /* total_count */;
+  if (seq_per_batch_) {
+    size_t total_batch_cnt = 0;
+    for (auto* writer : write_group) {
+      assert(writer->batch_cnt);
+      total_batch_cnt += writer->batch_cnt;
+    }
+    seq_inc = total_batch_cnt;
+  }
   if (!write_options.disableWAL) {
     status =
         ConcurrentWriteToWAL(write_group, log_used, &last_sequence, seq_inc);
@@ -591,7 +605,8 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
     }
     writer->sequence = curr_seq;
     if (seq_per_batch_) {
-      curr_seq++;
+      assert(writer->batch_cnt);
+      curr_seq += writer->batch_cnt;
     }
     // else seq advances only by memtable writes
   }
@@ -1004,7 +1019,7 @@ Status DBImpl::SwitchWAL(WriteContext* write_context) {
         break;
       }
       cfd->imm()->FlushRequested();
-      SchedulePendingFlush(cfd);
+      SchedulePendingFlush(cfd, FlushReason::kWriteBufferManager);
     }
   }
   MaybeScheduleFlushOrCompaction();
@@ -1050,7 +1065,7 @@ Status DBImpl::HandleWriteBufferFull(WriteContext* write_context) {
     status = SwitchMemtable(cfd_picked, write_context);
     if (status.ok()) {
       cfd_picked->imm()->FlushRequested();
-      SchedulePendingFlush(cfd_picked);
+      SchedulePendingFlush(cfd_picked, FlushReason::kWriteBufferFull);
       MaybeScheduleFlushOrCompaction();
     }
   }
@@ -1160,7 +1175,7 @@ Status DBImpl::ScheduleFlushes(WriteContext* context) {
 }
 
 #ifndef ROCKSDB_LITE
-void DBImpl::NotifyOnMemTableSealed(ColumnFamilyData* cfd,
+void DBImpl::NotifyOnMemTableSealed(ColumnFamilyData* /*cfd*/,
                                     const MemTableInfo& mem_table_info) {
   if (immutable_db_options_.listeners.size() == 0U) {
     return;
@@ -1292,6 +1307,29 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
                  ". Immutable memtables: %d.\n",
                  cfd->GetName().c_str(), new_log_number, num_imm_unflushed);
   mutex_.Lock();
+  if (s.ok() && creating_new_log) {
+    log_write_mutex_.Lock();
+    logfile_number_ = new_log_number;
+    assert(new_log != nullptr);
+    log_empty_ = true;
+    log_dir_synced_ = false;
+    if (!logs_.empty()) {
+      // Alway flush the buffer of the last log before switching to a new one
+      log::Writer* cur_log_writer = logs_.back().writer;
+      s = cur_log_writer->WriteBuffer();
+      if (!s.ok()) {
+        ROCKS_LOG_WARN(immutable_db_options_.info_log,
+                       "[%s] Failed to switch from #%" PRIu64 " to #%" PRIu64
+                       "  WAL file -- %s\n",
+                       cfd->GetName().c_str(), cur_log_writer->get_log_number(),
+                       new_log_number);
+      }
+    }
+    logs_.emplace_back(logfile_number_, new_log);
+    alive_log_files_.push_back(LogFileNumberSize(logfile_number_));
+    log_write_mutex_.Unlock();
+  }
+
   if (!s.ok()) {
     // how do we fail if we're not creating new log?
     assert(creating_new_log);
@@ -1302,21 +1340,7 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
     }
     return s;
   }
-  if (creating_new_log) {
-    log_write_mutex_.Lock();
-    logfile_number_ = new_log_number;
-    assert(new_log != nullptr);
-    log_empty_ = true;
-    log_dir_synced_ = false;
-    if (!logs_.empty()) {
-      // Alway flush the buffer of the last log before switching to a new one
-      log::Writer* cur_log_writer = logs_.back().writer;
-      cur_log_writer->WriteBuffer();
-    }
-    logs_.emplace_back(logfile_number_, new_log);
-    alive_log_files_.push_back(LogFileNumberSize(logfile_number_));
-    log_write_mutex_.Unlock();
-  }
+
   for (auto loop_cfd : *versions_->GetColumnFamilySet()) {
     // all this is just optimization to delete logs that
     // are no longer needed -- if CF is empty, that means it
@@ -1371,7 +1395,10 @@ Status DB::Put(const WriteOptions& opt, ColumnFamilyHandle* column_family,
   // 8 bytes are taken by header, 4 bytes for count, 1 byte for type,
   // and we allocate 11 extra bytes for key length, as well as value length.
   WriteBatch batch(key.size() + value.size() + 24);
-  batch.Put(column_family, key, value);
+  Status s = batch.Put(column_family, key, value);
+  if (!s.ok()) {
+    return s;
+  }
   return Write(opt, &batch);
 }
 
@@ -1400,7 +1427,10 @@ Status DB::DeleteRange(const WriteOptions& opt,
 Status DB::Merge(const WriteOptions& opt, ColumnFamilyHandle* column_family,
                  const Slice& key, const Slice& value) {
   WriteBatch batch;
-  batch.Merge(column_family, key, value);
+  Status s = batch.Merge(column_family, key, value);
+  if (!s.ok()) {
+    return s;
+  }
   return Write(opt, &batch);
 }
 }  // namespace rocksdb

db/db_iter.cc

@@ -217,6 +217,9 @@ class DBIter final: public Iterator {
         *prop = "Iterator is not valid.";
       }
       return Status::OK();
+    } else if (prop_name == "rocksdb.iterator.internal-key") {
+      *prop = saved_key_.GetUserKey().ToString();
+      return Status::OK();
     }
     return Status::InvalidArgument("Undentified property.");
   }
@@ -611,10 +614,12 @@ void DBIter::MergeValuesNewToOld() {
   // Start the merge process by pushing the first operand
   merge_context_.PushOperand(iter_->value(),
                              iter_->IsValuePinned() /* operand_pinned */);
+  TEST_SYNC_POINT("DBIter::MergeValuesNewToOld:PushedFirstOperand");
 
   ParsedInternalKey ikey;
   Status s;
   for (iter_->Next(); iter_->Valid(); iter_->Next()) {
+    TEST_SYNC_POINT("DBIter::MergeValuesNewToOld:SteppedToNextOperand");
     if (!ParseKey(&ikey)) {
       // skip corrupted key
       continue;

db/db_iterator_test.cc

@@ -18,14 +18,39 @@
 
 namespace rocksdb {
 
-class DBIteratorTest : public DBTestBase {
+// A dumb ReadCallback which saying every key is committed.
+class DummyReadCallback : public ReadCallback {
+  bool IsCommitted(SequenceNumber /*seq*/) { return true; }
+};
+
+// Test param:
+//   bool: whether to pass read_callback to NewIterator().
+class DBIteratorTest : public DBTestBase,
+                       public testing::WithParamInterface<bool> {
  public:
   DBIteratorTest() : DBTestBase("/db_iterator_test") {}
+
+  Iterator* NewIterator(const ReadOptions& read_options,
+                        ColumnFamilyHandle* column_family = nullptr) {
+    if (column_family == nullptr) {
+      column_family = db_->DefaultColumnFamily();
+    }
+    auto* cfd = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family)->cfd();
+    SequenceNumber seq = read_options.snapshot != nullptr
+                             ? read_options.snapshot->GetSequenceNumber()
+                             : db_->GetLatestSequenceNumber();
+    bool use_read_callback = GetParam();
+    ReadCallback* read_callback = use_read_callback ? &read_callback_ : nullptr;
+    return dbfull()->NewIteratorImpl(read_options, cfd, seq, read_callback);
+  }
+
+ private:
+  DummyReadCallback read_callback_;
 };
 
 class FlushBlockEveryKeyPolicy : public FlushBlockPolicy {
  public:
-  virtual bool Update(const Slice& key, const Slice& value) override {
+  virtual bool Update(const Slice& /*key*/, const Slice& /*value*/) override {
     if (!start_) {
       start_ = true;
       return false;
@@ -45,34 +70,41 @@ class FlushBlockEveryKeyPolicyFactory : public FlushBlockPolicyFactory {
   }
 
   FlushBlockPolicy* NewFlushBlockPolicy(
-    const BlockBasedTableOptions& table_options,
-    const BlockBuilder& data_block_builder) const override {
+      const BlockBasedTableOptions& /*table_options*/,
+      const BlockBuilder& /*data_block_builder*/) const override {
     return new FlushBlockEveryKeyPolicy;
   }
 };
 
-TEST_F(DBIteratorTest, IteratorProperty) {
+TEST_P(DBIteratorTest, IteratorProperty) {
   // The test needs to be changed if kPersistedTier is supported in iterator.
   Options options = CurrentOptions();
   CreateAndReopenWithCF({"pikachu"}, options);
   Put(1, "1", "2");
+  Delete(1, "2");
   ReadOptions ropt;
   ropt.pin_data = false;
   {
-    unique_ptr<Iterator> iter(db_->NewIterator(ropt, handles_[1]));
+    unique_ptr<Iterator> iter(NewIterator(ropt, handles_[1]));
     iter->SeekToFirst();
     std::string prop_value;
     ASSERT_NOK(iter->GetProperty("non_existing.value", &prop_value));
     ASSERT_OK(iter->GetProperty("rocksdb.iterator.is-key-pinned", &prop_value));
     ASSERT_EQ("0", prop_value);
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.internal-key", &prop_value));
+    ASSERT_EQ("1", prop_value);
     iter->Next();
     ASSERT_OK(iter->GetProperty("rocksdb.iterator.is-key-pinned", &prop_value));
     ASSERT_EQ("Iterator is not valid.", prop_value);
+
+    // Get internal key at which the iteration stopped (tombstone in this case).
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.internal-key", &prop_value));
+    ASSERT_EQ("2", prop_value);
   }
   Close();
 }
 
-TEST_F(DBIteratorTest, PersistedTierOnIterator) {
+TEST_P(DBIteratorTest, PersistedTierOnIterator) {
   // The test needs to be changed if kPersistedTier is supported in iterator.
   Options options = CurrentOptions();
   CreateAndReopenWithCF({"pikachu"}, options);
@@ -88,7 +120,7 @@ TEST_F(DBIteratorTest, PersistedTierOnIterator) {
   Close();
 }
 
-TEST_F(DBIteratorTest, NonBlockingIteration) {
+TEST_P(DBIteratorTest, NonBlockingIteration) {
   do {
     ReadOptions non_blocking_opts, regular_opts;
     Options options = CurrentOptions();
@@ -100,7 +132,7 @@ TEST_F(DBIteratorTest, NonBlockingIteration) {
 
     // scan using non-blocking iterator. We should find it because
     // it is in memtable.
-    Iterator* iter = db_->NewIterator(non_blocking_opts, handles_[1]);
+    Iterator* iter = NewIterator(non_blocking_opts, handles_[1]);
     int count = 0;
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
       ASSERT_OK(iter->status());
@@ -117,7 +149,7 @@ TEST_F(DBIteratorTest, NonBlockingIteration) {
     // kvs. Neither does it do any IOs to storage.
     uint64_t numopen = TestGetTickerCount(options, NO_FILE_OPENS);
     uint64_t cache_added = TestGetTickerCount(options, BLOCK_CACHE_ADD);
-    iter = db_->NewIterator(non_blocking_opts, handles_[1]);
+    iter = NewIterator(non_blocking_opts, handles_[1]);
     count = 0;
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
       count++;
@@ -134,7 +166,7 @@ TEST_F(DBIteratorTest, NonBlockingIteration) {
     // verify that we can find it via a non-blocking scan
     numopen = TestGetTickerCount(options, NO_FILE_OPENS);
     cache_added = TestGetTickerCount(options, BLOCK_CACHE_ADD);
-    iter = db_->NewIterator(non_blocking_opts, handles_[1]);
+    iter = NewIterator(non_blocking_opts, handles_[1]);
     count = 0;
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
       ASSERT_OK(iter->status());
@@ -153,7 +185,7 @@ TEST_F(DBIteratorTest, NonBlockingIteration) {
 }
 
 #ifndef ROCKSDB_LITE
-TEST_F(DBIteratorTest, ManagedNonBlockingIteration) {
+TEST_P(DBIteratorTest, ManagedNonBlockingIteration) {
   do {
     ReadOptions non_blocking_opts, regular_opts;
     Options options = CurrentOptions();
@@ -166,7 +198,7 @@ TEST_F(DBIteratorTest, ManagedNonBlockingIteration) {
 
     // scan using non-blocking iterator. We should find it because
     // it is in memtable.
-    Iterator* iter = db_->NewIterator(non_blocking_opts, handles_[1]);
+    Iterator* iter = NewIterator(non_blocking_opts, handles_[1]);
     int count = 0;
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
       ASSERT_OK(iter->status());
@@ -183,7 +215,7 @@ TEST_F(DBIteratorTest, ManagedNonBlockingIteration) {
     // kvs. Neither does it do any IOs to storage.
     int64_t numopen = TestGetTickerCount(options, NO_FILE_OPENS);
     int64_t cache_added = TestGetTickerCount(options, BLOCK_CACHE_ADD);
-    iter = db_->NewIterator(non_blocking_opts, handles_[1]);
+    iter = NewIterator(non_blocking_opts, handles_[1]);
     count = 0;
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
       count++;
@@ -200,7 +232,7 @@ TEST_F(DBIteratorTest, ManagedNonBlockingIteration) {
     // verify that we can find it via a non-blocking scan
     numopen = TestGetTickerCount(options, NO_FILE_OPENS);
     cache_added = TestGetTickerCount(options, BLOCK_CACHE_ADD);
-    iter = db_->NewIterator(non_blocking_opts, handles_[1]);
+    iter = NewIterator(non_blocking_opts, handles_[1]);
     count = 0;
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
       ASSERT_OK(iter->status());
@@ -219,7 +251,7 @@ TEST_F(DBIteratorTest, ManagedNonBlockingIteration) {
 }
 #endif  // ROCKSDB_LITE
 
-TEST_F(DBIteratorTest, IterSeekBeforePrev) {
+TEST_P(DBIteratorTest, IterSeekBeforePrev) {
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
   dbfull()->Flush(FlushOptions());
@@ -227,7 +259,7 @@ TEST_F(DBIteratorTest, IterSeekBeforePrev) {
   ASSERT_OK(Put("1", "h"));
   dbfull()->Flush(FlushOptions());
   ASSERT_OK(Put("2", "j"));
-  auto iter = db_->NewIterator(ReadOptions());
+  auto iter = NewIterator(ReadOptions());
   iter->Seek(Slice("c"));
   iter->Prev();
   iter->Seek(Slice("a"));
@@ -235,7 +267,7 @@ TEST_F(DBIteratorTest, IterSeekBeforePrev) {
   delete iter;
 }
 
-TEST_F(DBIteratorTest, IterSeekForPrevBeforeNext) {
+TEST_P(DBIteratorTest, IterSeekForPrevBeforeNext) {
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
   dbfull()->Flush(FlushOptions());
@@ -243,7 +275,7 @@ TEST_F(DBIteratorTest, IterSeekForPrevBeforeNext) {
   ASSERT_OK(Put("1", "h"));
   dbfull()->Flush(FlushOptions());
   ASSERT_OK(Put("2", "j"));
-  auto iter = db_->NewIterator(ReadOptions());
+  auto iter = NewIterator(ReadOptions());
   iter->SeekForPrev(Slice("0"));
   iter->Next();
   iter->SeekForPrev(Slice("1"));
@@ -257,7 +289,7 @@ std::string MakeLongKey(size_t length, char c) {
 }
 }  // namespace
 
-TEST_F(DBIteratorTest, IterLongKeys) {
+TEST_P(DBIteratorTest, IterLongKeys) {
   ASSERT_OK(Put(MakeLongKey(20, 0), "0"));
   ASSERT_OK(Put(MakeLongKey(32, 2), "2"));
   ASSERT_OK(Put("a", "b"));
@@ -265,7 +297,7 @@ TEST_F(DBIteratorTest, IterLongKeys) {
   ASSERT_OK(Put(MakeLongKey(50, 1), "1"));
   ASSERT_OK(Put(MakeLongKey(127, 3), "3"));
   ASSERT_OK(Put(MakeLongKey(64, 4), "4"));
-  auto iter = db_->NewIterator(ReadOptions());
+  auto iter = NewIterator(ReadOptions());
 
   // Create a key that needs to be skipped for Seq too new
   iter->Seek(MakeLongKey(20, 0));
@@ -287,7 +319,7 @@ TEST_F(DBIteratorTest, IterLongKeys) {
   ASSERT_EQ(IterStatus(iter), MakeLongKey(50, 1) + "->1");
   delete iter;
 
-  iter = db_->NewIterator(ReadOptions());
+  iter = NewIterator(ReadOptions());
   iter->Seek(MakeLongKey(50, 1));
   ASSERT_EQ(IterStatus(iter), MakeLongKey(50, 1) + "->1");
   iter->Next();
@@ -297,13 +329,13 @@ TEST_F(DBIteratorTest, IterLongKeys) {
   delete iter;
 }
 
-TEST_F(DBIteratorTest, IterNextWithNewerSeq) {
+TEST_P(DBIteratorTest, IterNextWithNewerSeq) {
   ASSERT_OK(Put("0", "0"));
   dbfull()->Flush(FlushOptions());
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
   ASSERT_OK(Put("d", "e"));
-  auto iter = db_->NewIterator(ReadOptions());
+  auto iter = NewIterator(ReadOptions());
 
   // Create a key that needs to be skipped for Seq too new
   for (uint64_t i = 0; i < last_options_.max_sequential_skip_in_iterations + 1;
@@ -323,13 +355,13 @@ TEST_F(DBIteratorTest, IterNextWithNewerSeq) {
   delete iter;
 }
 
-TEST_F(DBIteratorTest, IterPrevWithNewerSeq) {
+TEST_P(DBIteratorTest, IterPrevWithNewerSeq) {
   ASSERT_OK(Put("0", "0"));
   dbfull()->Flush(FlushOptions());
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
   ASSERT_OK(Put("d", "e"));
-  auto iter = db_->NewIterator(ReadOptions());
+  auto iter = NewIterator(ReadOptions());
 
   // Create a key that needs to be skipped for Seq too new
   for (uint64_t i = 0; i < last_options_.max_sequential_skip_in_iterations + 1;
@@ -354,14 +386,14 @@ TEST_F(DBIteratorTest, IterPrevWithNewerSeq) {
   delete iter;
 }
 
-TEST_F(DBIteratorTest, IterPrevWithNewerSeq2) {
+TEST_P(DBIteratorTest, IterPrevWithNewerSeq2) {
   ASSERT_OK(Put("0", "0"));
   dbfull()->Flush(FlushOptions());
   ASSERT_OK(Put("a", "b"));
   ASSERT_OK(Put("c", "d"));
   ASSERT_OK(Put("e", "f"));
-  auto iter = db_->NewIterator(ReadOptions());
-  auto iter2 = db_->NewIterator(ReadOptions());
+  auto iter = NewIterator(ReadOptions());
+  auto iter2 = NewIterator(ReadOptions());
   iter->Seek(Slice("c"));
   iter2->SeekForPrev(Slice("d"));
   ASSERT_EQ(IterStatus(iter), "c->d");
@@ -383,10 +415,10 @@ TEST_F(DBIteratorTest, IterPrevWithNewerSeq2) {
   delete iter2;
 }
 
-TEST_F(DBIteratorTest, IterEmpty) {
+TEST_P(DBIteratorTest, IterEmpty) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
-    Iterator* iter = db_->NewIterator(ReadOptions(), handles_[1]);
+    Iterator* iter = NewIterator(ReadOptions(), handles_[1]);
 
     iter->SeekToFirst();
     ASSERT_EQ(IterStatus(iter), "(invalid)");
@@ -404,11 +436,11 @@ TEST_F(DBIteratorTest, IterEmpty) {
   } while (ChangeCompactOptions());
 }
 
-TEST_F(DBIteratorTest, IterSingle) {
+TEST_P(DBIteratorTest, IterSingle) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
     ASSERT_OK(Put(1, "a", "va"));
-    Iterator* iter = db_->NewIterator(ReadOptions(), handles_[1]);
+    Iterator* iter = NewIterator(ReadOptions(), handles_[1]);
 
     iter->SeekToFirst();
     ASSERT_EQ(IterStatus(iter), "a->va");
@@ -455,13 +487,13 @@ TEST_F(DBIteratorTest, IterSingle) {
   } while (ChangeCompactOptions());
 }
 
-TEST_F(DBIteratorTest, IterMulti) {
+TEST_P(DBIteratorTest, IterMulti) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
     ASSERT_OK(Put(1, "a", "va"));
     ASSERT_OK(Put(1, "b", "vb"));
     ASSERT_OK(Put(1, "c", "vc"));
-    Iterator* iter = db_->NewIterator(ReadOptions(), handles_[1]);
+    Iterator* iter = NewIterator(ReadOptions(), handles_[1]);
 
     iter->SeekToFirst();
     ASSERT_EQ(IterStatus(iter), "a->va");
@@ -554,7 +586,7 @@ TEST_F(DBIteratorTest, IterMulti) {
 
 // Check that we can skip over a run of user keys
 // by using reseek rather than sequential scan
-TEST_F(DBIteratorTest, IterReseek) {
+TEST_P(DBIteratorTest, IterReseek) {
   anon::OptionsOverride options_override;
   options_override.skip_policy = kSkipNoSnapshot;
   Options options = CurrentOptions(options_override);
@@ -571,7 +603,7 @@ TEST_F(DBIteratorTest, IterReseek) {
   ASSERT_OK(Put(1, "a", "one"));
   ASSERT_OK(Put(1, "a", "two"));
   ASSERT_OK(Put(1, "b", "bone"));
-  Iterator* iter = db_->NewIterator(ReadOptions(), handles_[1]);
+  Iterator* iter = NewIterator(ReadOptions(), handles_[1]);
   iter->SeekToFirst();
   ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 0);
   ASSERT_EQ(IterStatus(iter), "a->two");
@@ -583,7 +615,7 @@ TEST_F(DBIteratorTest, IterReseek) {
   // insert a total of three keys with same userkey and verify
   // that reseek is still not invoked.
   ASSERT_OK(Put(1, "a", "three"));
-  iter = db_->NewIterator(ReadOptions(), handles_[1]);
+  iter = NewIterator(ReadOptions(), handles_[1]);
   iter->SeekToFirst();
   ASSERT_EQ(IterStatus(iter), "a->three");
   iter->Next();
@@ -594,7 +626,7 @@ TEST_F(DBIteratorTest, IterReseek) {
   // insert a total of four keys with same userkey and verify
   // that reseek is invoked.
   ASSERT_OK(Put(1, "a", "four"));
-  iter = db_->NewIterator(ReadOptions(), handles_[1]);
+  iter = NewIterator(ReadOptions(), handles_[1]);
   iter->SeekToFirst();
   ASSERT_EQ(IterStatus(iter), "a->four");
   ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION), 0);
@@ -611,7 +643,7 @@ TEST_F(DBIteratorTest, IterReseek) {
 
   // Insert another version of b and assert that reseek is not invoked
   ASSERT_OK(Put(1, "b", "btwo"));
-  iter = db_->NewIterator(ReadOptions(), handles_[1]);
+  iter = NewIterator(ReadOptions(), handles_[1]);
   iter->SeekToLast();
   ASSERT_EQ(IterStatus(iter), "b->btwo");
   ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION),
@@ -626,7 +658,7 @@ TEST_F(DBIteratorTest, IterReseek) {
   // of b and 4 versions of a.
   ASSERT_OK(Put(1, "b", "bthree"));
   ASSERT_OK(Put(1, "b", "bfour"));
-  iter = db_->NewIterator(ReadOptions(), handles_[1]);
+  iter = NewIterator(ReadOptions(), handles_[1]);
   iter->SeekToLast();
   ASSERT_EQ(IterStatus(iter), "b->bfour");
   ASSERT_EQ(TestGetTickerCount(options, NUMBER_OF_RESEEKS_IN_ITERATION),
@@ -640,7 +672,7 @@ TEST_F(DBIteratorTest, IterReseek) {
   delete iter;
 }
 
-TEST_F(DBIteratorTest, IterSmallAndLargeMix) {
+TEST_P(DBIteratorTest, IterSmallAndLargeMix) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
     ASSERT_OK(Put(1, "a", "va"));
@@ -649,7 +681,7 @@ TEST_F(DBIteratorTest, IterSmallAndLargeMix) {
     ASSERT_OK(Put(1, "d", std::string(100000, 'd')));
     ASSERT_OK(Put(1, "e", std::string(100000, 'e')));
 
-    Iterator* iter = db_->NewIterator(ReadOptions(), handles_[1]);
+    Iterator* iter = NewIterator(ReadOptions(), handles_[1]);
 
     iter->SeekToFirst();
     ASSERT_EQ(IterStatus(iter), "a->va");
@@ -681,7 +713,7 @@ TEST_F(DBIteratorTest, IterSmallAndLargeMix) {
   } while (ChangeCompactOptions());
 }
 
-TEST_F(DBIteratorTest, IterMultiWithDelete) {
+TEST_P(DBIteratorTest, IterMultiWithDelete) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
     ASSERT_OK(Put(1, "ka", "va"));
@@ -690,7 +722,7 @@ TEST_F(DBIteratorTest, IterMultiWithDelete) {
     ASSERT_OK(Delete(1, "kb"));
     ASSERT_EQ("NOT_FOUND", Get(1, "kb"));
 
-    Iterator* iter = db_->NewIterator(ReadOptions(), handles_[1]);
+    Iterator* iter = NewIterator(ReadOptions(), handles_[1]);
     iter->Seek("kc");
     ASSERT_EQ(IterStatus(iter), "kc->vc");
     if (!CurrentOptions().merge_operator) {
@@ -707,7 +739,7 @@ TEST_F(DBIteratorTest, IterMultiWithDelete) {
   } while (ChangeOptions());
 }
 
-TEST_F(DBIteratorTest, IterPrevMaxSkip) {
+TEST_P(DBIteratorTest, IterPrevMaxSkip) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
     for (int i = 0; i < 2; i++) {
@@ -737,7 +769,7 @@ TEST_F(DBIteratorTest, IterPrevMaxSkip) {
   } while (ChangeOptions(kSkipMergePut | kSkipNoSeekToLast));
 }
 
-TEST_F(DBIteratorTest, IterWithSnapshot) {
+TEST_P(DBIteratorTest, IterWithSnapshot) {
   anon::OptionsOverride options_override;
   options_override.skip_policy = kSkipNoSnapshot;
   do {
@@ -751,7 +783,7 @@ TEST_F(DBIteratorTest, IterWithSnapshot) {
     const Snapshot* snapshot = db_->GetSnapshot();
     ReadOptions options;
     options.snapshot = snapshot;
-    Iterator* iter = db_->NewIterator(options, handles_[1]);
+    Iterator* iter = NewIterator(options, handles_[1]);
 
     ASSERT_OK(Put(1, "key0", "val0"));
     // Put more values after the snapshot
@@ -804,13 +836,13 @@ TEST_F(DBIteratorTest, IterWithSnapshot) {
   } while (ChangeOptions(kSkipHashCuckoo));
 }
 
-TEST_F(DBIteratorTest, IteratorPinsRef) {
+TEST_P(DBIteratorTest, IteratorPinsRef) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
     Put(1, "foo", "hello");
 
     // Get iterator that will yield the current contents of the DB.
-    Iterator* iter = db_->NewIterator(ReadOptions(), handles_[1]);
+    Iterator* iter = NewIterator(ReadOptions(), handles_[1]);
 
     // Write to force compactions
     Put(1, "foo", "newvalue1");
@@ -830,7 +862,7 @@ TEST_F(DBIteratorTest, IteratorPinsRef) {
   } while (ChangeCompactOptions());
 }
 
-TEST_F(DBIteratorTest, DBIteratorBoundTest) {
+TEST_P(DBIteratorTest, DBIteratorBoundTest) {
   Options options = CurrentOptions();
   options.env = env_;
   options.create_if_missing = true;
@@ -847,7 +879,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundTest) {
     ReadOptions ro;
     ro.iterate_upper_bound = nullptr;
 
-    std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+    std::unique_ptr<Iterator> iter(NewIterator(ro));
 
     iter->Seek("foo");
 
@@ -884,7 +916,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundTest) {
     Slice prefix("foo2");
     ro.iterate_upper_bound = &prefix;
 
-    std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+    std::unique_ptr<Iterator> iter(NewIterator(ro));
 
     iter->Seek("foo");
 
@@ -906,7 +938,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundTest) {
     Slice prefix("foo");
     ro.iterate_upper_bound = &prefix;
 
-    std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+    std::unique_ptr<Iterator> iter(NewIterator(ro));
 
     iter->SeekToLast();
     ASSERT_TRUE(iter->Valid());
@@ -930,7 +962,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundTest) {
     Slice upper_bound("g");
     ro.iterate_upper_bound = &upper_bound;
 
-    std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+    std::unique_ptr<Iterator> iter(NewIterator(ro));
 
     iter->Seek("foo");
 
@@ -963,7 +995,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundTest) {
     ReadOptions ro;
     ro.iterate_upper_bound = nullptr;
 
-    std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+    std::unique_ptr<Iterator> iter(NewIterator(ro));
 
     iter->Seek("b");
     ASSERT_TRUE(iter->Valid());
@@ -983,7 +1015,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundTest) {
     Slice prefix("c");
     ro.iterate_upper_bound = &prefix;
 
-    iter.reset(db_->NewIterator(ro));
+    iter.reset(NewIterator(ro));
 
     get_perf_context()->Reset();
 
@@ -1004,7 +1036,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundTest) {
   }
 }
 
-TEST_F(DBIteratorTest, DBIteratorBoundOptimizationTest) {
+TEST_P(DBIteratorTest, DBIteratorBoundOptimizationTest) {
   int upper_bound_hits = 0;
   Options options = CurrentOptions();
   rocksdb::SyncPoint::GetInstance()->SetCallBack(
@@ -1032,7 +1064,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundOptimizationTest) {
   ReadOptions ro;
   ro.iterate_upper_bound = &ub;
 
-  std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+  std::unique_ptr<Iterator> iter(NewIterator(ro));
 
   iter->Seek("foo");
   ASSERT_TRUE(iter->Valid());
@@ -1050,7 +1082,7 @@ TEST_F(DBIteratorTest, DBIteratorBoundOptimizationTest) {
 }
 // TODO(3.13): fix the issue of Seek() + Prev() which might not necessary
 //             return the biggest key which is smaller than the seek key.
-TEST_F(DBIteratorTest, PrevAfterAndNextAfterMerge) {
+TEST_P(DBIteratorTest, PrevAfterAndNextAfterMerge) {
   Options options;
   options.create_if_missing = true;
   options.merge_operator = MergeOperators::CreatePutOperator();
@@ -1063,7 +1095,7 @@ TEST_F(DBIteratorTest, PrevAfterAndNextAfterMerge) {
   db_->Merge(wopts, "2", "data2");
   db_->Merge(wopts, "3", "data3");
 
-  std::unique_ptr<Iterator> it(db_->NewIterator(ReadOptions()));
+  std::unique_ptr<Iterator> it(NewIterator(ReadOptions()));
 
   it->Seek("2");
   ASSERT_TRUE(it->Valid());
@@ -1159,7 +1191,7 @@ class DBIteratorTestForPinnedData : public DBIteratorTest {
 
     ReadOptions ro;
     ro.pin_data = true;
-    auto iter = db_->NewIterator(ro);
+    auto iter = NewIterator(ro);
 
     {
       // Test Seek to random keys
@@ -1251,25 +1283,25 @@ class DBIteratorTestForPinnedData : public DBIteratorTest {
 }
 };
 
-TEST_F(DBIteratorTestForPinnedData, PinnedDataIteratorRandomizedNormal) {
+TEST_P(DBIteratorTestForPinnedData, PinnedDataIteratorRandomizedNormal) {
   PinnedDataIteratorRandomized(TestConfig::NORMAL);
 }
 
-TEST_F(DBIteratorTestForPinnedData, PinnedDataIteratorRandomizedCLoseAndOpen) {
+TEST_P(DBIteratorTestForPinnedData, PinnedDataIteratorRandomizedCLoseAndOpen) {
   PinnedDataIteratorRandomized(TestConfig::CLOSE_AND_OPEN);
 }
 
-TEST_F(DBIteratorTestForPinnedData,
+TEST_P(DBIteratorTestForPinnedData,
        PinnedDataIteratorRandomizedCompactBeforeRead) {
   PinnedDataIteratorRandomized(TestConfig::COMPACT_BEFORE_READ);
 }
 
-TEST_F(DBIteratorTestForPinnedData, PinnedDataIteratorRandomizedFlush) {
+TEST_P(DBIteratorTestForPinnedData, PinnedDataIteratorRandomizedFlush) {
   PinnedDataIteratorRandomized(TestConfig::FLUSH_EVERY_1000);
 }
 
 #ifndef ROCKSDB_LITE
-TEST_F(DBIteratorTest, PinnedDataIteratorMultipleFiles) {
+TEST_P(DBIteratorTest, PinnedDataIteratorMultipleFiles) {
   Options options = CurrentOptions();
   BlockBasedTableOptions table_options;
   table_options.use_delta_encoding = false;
@@ -1318,7 +1350,7 @@ TEST_F(DBIteratorTest, PinnedDataIteratorMultipleFiles) {
 
   ReadOptions ro;
   ro.pin_data = true;
-  auto iter = db_->NewIterator(ro);
+  auto iter = NewIterator(ro);
 
   std::vector<std::pair<Slice, std::string>> results;
   for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
@@ -1340,7 +1372,7 @@ TEST_F(DBIteratorTest, PinnedDataIteratorMultipleFiles) {
 }
 #endif
 
-TEST_F(DBIteratorTest, PinnedDataIteratorMergeOperator) {
+TEST_P(DBIteratorTest, PinnedDataIteratorMergeOperator) {
   Options options = CurrentOptions();
   BlockBasedTableOptions table_options;
   table_options.use_delta_encoding = false;
@@ -1373,7 +1405,7 @@ TEST_F(DBIteratorTest, PinnedDataIteratorMergeOperator) {
 
   ReadOptions ro;
   ro.pin_data = true;
-  auto iter = db_->NewIterator(ro);
+  auto iter = NewIterator(ro);
 
   std::vector<std::pair<Slice, std::string>> results;
   for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
@@ -1400,7 +1432,7 @@ TEST_F(DBIteratorTest, PinnedDataIteratorMergeOperator) {
   delete iter;
 }
 
-TEST_F(DBIteratorTest, PinnedDataIteratorReadAfterUpdate) {
+TEST_P(DBIteratorTest, PinnedDataIteratorReadAfterUpdate) {
   Options options = CurrentOptions();
   BlockBasedTableOptions table_options;
   table_options.use_delta_encoding = false;
@@ -1420,7 +1452,7 @@ TEST_F(DBIteratorTest, PinnedDataIteratorReadAfterUpdate) {
 
   ReadOptions ro;
   ro.pin_data = true;
-  auto iter = db_->NewIterator(ro);
+  auto iter = NewIterator(ro);
 
   // Delete 50% of the keys and update the other 50%
   for (auto& kv : true_data) {
@@ -1450,7 +1482,7 @@ TEST_F(DBIteratorTest, PinnedDataIteratorReadAfterUpdate) {
   delete iter;
 }
 
-TEST_F(DBIteratorTest, IterSeekForPrevCrossingFiles) {
+TEST_P(DBIteratorTest, IterSeekForPrevCrossingFiles) {
   Options options = CurrentOptions();
   options.prefix_extractor.reset(NewFixedPrefixTransform(1));
   options.disable_auto_compactions = true;
@@ -1479,7 +1511,7 @@ TEST_F(DBIteratorTest, IterSeekForPrevCrossingFiles) {
   MoveFilesToLevel(1);
   {
     ReadOptions ro;
-    Iterator* iter = db_->NewIterator(ro);
+    Iterator* iter = NewIterator(ro);
 
     iter->SeekForPrev("a4");
     ASSERT_EQ(iter->key().ToString(), "a3");
@@ -1497,14 +1529,14 @@ TEST_F(DBIteratorTest, IterSeekForPrevCrossingFiles) {
   {
     ReadOptions ro;
     ro.prefix_same_as_start = true;
-    Iterator* iter = db_->NewIterator(ro);
+    Iterator* iter = NewIterator(ro);
     iter->SeekForPrev("c2");
     ASSERT_TRUE(!iter->Valid());
     delete iter;
   }
 }
 
-TEST_F(DBIteratorTest, IterPrevKeyCrossingBlocks) {
+TEST_P(DBIteratorTest, IterPrevKeyCrossingBlocks) {
   Options options = CurrentOptions();
   BlockBasedTableOptions table_options;
   table_options.block_size = 1;  // every block will contain one entry
@@ -1546,7 +1578,7 @@ TEST_F(DBIteratorTest, IterPrevKeyCrossingBlocks) {
   {
     ReadOptions ro;
     ro.fill_cache = false;
-    Iterator* iter = db_->NewIterator(ro);
+    Iterator* iter = NewIterator(ro);
 
     iter->SeekToLast();
     ASSERT_EQ(iter->key().ToString(), "key5");
@@ -1572,7 +1604,7 @@ TEST_F(DBIteratorTest, IterPrevKeyCrossingBlocks) {
   }
 }
 
-TEST_F(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
+TEST_P(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
   Options options = CurrentOptions();
   options.merge_operator = MergeOperators::CreateStringAppendTESTOperator();
   options.disable_auto_compactions = true;
@@ -1648,7 +1680,7 @@ TEST_F(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
   {
     ReadOptions ro;
     ro.fill_cache = false;
-    Iterator* iter = db_->NewIterator(ro);
+    Iterator* iter = NewIterator(ro);
     auto data_iter = true_data.rbegin();
 
     for (iter->SeekToLast(); iter->Valid(); iter->Prev()) {
@@ -1664,7 +1696,7 @@ TEST_F(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
   {
     ReadOptions ro;
     ro.fill_cache = false;
-    Iterator* iter = db_->NewIterator(ro);
+    Iterator* iter = NewIterator(ro);
     auto data_iter = true_data.rbegin();
 
     int entries_right = 0;
@@ -1719,7 +1751,7 @@ TEST_F(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
   }
 }
 
-TEST_F(DBIteratorTest, IteratorWithLocalStatistics) {
+TEST_P(DBIteratorTest, IteratorWithLocalStatistics) {
   Options options = CurrentOptions();
   options.statistics = rocksdb::CreateDBStatistics();
   DestroyAndReopen(options);
@@ -1740,7 +1772,7 @@ TEST_F(DBIteratorTest, IteratorWithLocalStatistics) {
   std::function<void()> reader_func_next = [&]() {
     SetPerfLevel(kEnableCount);
     get_perf_context()->Reset();
-    Iterator* iter = db_->NewIterator(ReadOptions());
+    Iterator* iter = NewIterator(ReadOptions());
 
     iter->SeekToFirst();
     // Seek will bump ITER_BYTES_READ
@@ -1767,7 +1799,7 @@ TEST_F(DBIteratorTest, IteratorWithLocalStatistics) {
 
   std::function<void()> reader_func_prev = [&]() {
     SetPerfLevel(kEnableCount);
-    Iterator* iter = db_->NewIterator(ReadOptions());
+    Iterator* iter = NewIterator(ReadOptions());
 
     iter->SeekToLast();
     // Seek will bump ITER_BYTES_READ
@@ -1813,7 +1845,7 @@ TEST_F(DBIteratorTest, IteratorWithLocalStatistics) {
 
 }
 
-TEST_F(DBIteratorTest, ReadAhead) {
+TEST_P(DBIteratorTest, ReadAhead) {
   Options options;
   env_->count_random_reads_ = true;
   options.env = env_;
@@ -1850,7 +1882,7 @@ TEST_F(DBIteratorTest, ReadAhead) {
   env_->random_read_bytes_counter_ = 0;
   options.statistics->setTickerCount(NO_FILE_OPENS, 0);
   ReadOptions read_options;
-  auto* iter = db_->NewIterator(read_options);
+  auto* iter = NewIterator(read_options);
   iter->SeekToFirst();
   int64_t num_file_opens = TestGetTickerCount(options, NO_FILE_OPENS);
   size_t bytes_read = env_->random_read_bytes_counter_;
@@ -1859,7 +1891,7 @@ TEST_F(DBIteratorTest, ReadAhead) {
   env_->random_read_bytes_counter_ = 0;
   options.statistics->setTickerCount(NO_FILE_OPENS, 0);
   read_options.readahead_size = 1024 * 10;
-  iter = db_->NewIterator(read_options);
+  iter = NewIterator(read_options);
   iter->SeekToFirst();
   int64_t num_file_opens_readahead = TestGetTickerCount(options, NO_FILE_OPENS);
   size_t bytes_read_readahead = env_->random_read_bytes_counter_;
@@ -1869,7 +1901,7 @@ TEST_F(DBIteratorTest, ReadAhead) {
   ASSERT_GT(bytes_read_readahead, read_options.readahead_size * 3);
 
   // Verify correctness.
-  iter = db_->NewIterator(read_options);
+  iter = NewIterator(read_options);
   int count = 0;
   for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
     ASSERT_EQ(value, iter->value());
@@ -1886,7 +1918,7 @@ TEST_F(DBIteratorTest, ReadAhead) {
 // Insert a key, create a snapshot iterator, overwrite key lots of times,
 // seek to a smaller key. Expect DBIter to fall back to a seek instead of
 // going through all the overwrites linearly.
-TEST_F(DBIteratorTest, DBIteratorSkipRecentDuplicatesTest) {
+TEST_P(DBIteratorTest, DBIteratorSkipRecentDuplicatesTest) {
   Options options = CurrentOptions();
   options.env = env_;
   options.create_if_missing = true;
@@ -1901,7 +1933,7 @@ TEST_F(DBIteratorTest, DBIteratorSkipRecentDuplicatesTest) {
 
   // Create iterator.
   ReadOptions ro;
-  std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+  std::unique_ptr<Iterator> iter(NewIterator(ro));
 
   // Insert a lot.
   for (int i = 0; i < 100; ++i) {
@@ -1939,10 +1971,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipRecentDuplicatesTest) {
                  NUMBER_OF_RESEEKS_IN_ITERATION));
 }
 
-TEST_F(DBIteratorTest, Refresh) {
+TEST_P(DBIteratorTest, Refresh) {
   ASSERT_OK(Put("x", "y"));
 
-  std::unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions()));
+  std::unique_ptr<Iterator> iter(NewIterator(ReadOptions()));
   iter->Seek(Slice("a"));
   ASSERT_TRUE(iter->Valid());
   ASSERT_EQ(iter->key().compare(Slice("x")), 0);
@@ -1998,20 +2030,20 @@ TEST_F(DBIteratorTest, Refresh) {
   iter.reset();
 }
 
-TEST_F(DBIteratorTest, CreationFailure) {
+TEST_P(DBIteratorTest, CreationFailure) {
   SyncPoint::GetInstance()->SetCallBack(
       "DBImpl::NewInternalIterator:StatusCallback", [](void* arg) {
         *(reinterpret_cast<Status*>(arg)) = Status::Corruption("test status");
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
-  Iterator* iter = db_->NewIterator(ReadOptions());
+  Iterator* iter = NewIterator(ReadOptions());
   ASSERT_FALSE(iter->Valid());
   ASSERT_TRUE(iter->status().IsCorruption());
   delete iter;
 }
 
-TEST_F(DBIteratorTest, TableFilter) {
+TEST_P(DBIteratorTest, TableFilter) {
   ASSERT_OK(Put("a", "1"));
   dbfull()->Flush(FlushOptions());
   ASSERT_OK(Put("b", "2"));
@@ -2036,7 +2068,7 @@ TEST_F(DBIteratorTest, TableFilter) {
       }
       return true;
     };
-    auto iter = db_->NewIterator(opts);
+    auto iter = NewIterator(opts);
     iter->SeekToFirst();
     ASSERT_EQ(IterStatus(iter), "a->1");
     iter->Next();
@@ -2062,7 +2094,7 @@ TEST_F(DBIteratorTest, TableFilter) {
     opts.table_filter = [](const TableProperties& props) {
       return props.num_entries != 2;
     };
-    auto iter = db_->NewIterator(opts);
+    auto iter = NewIterator(opts);
     iter->SeekToFirst();
     ASSERT_EQ(IterStatus(iter), "a->1");
     iter->Next();
@@ -2077,7 +2109,7 @@ TEST_F(DBIteratorTest, TableFilter) {
   }
 }
 
-TEST_F(DBIteratorTest, SkipStatistics) {
+TEST_P(DBIteratorTest, SkipStatistics) {
   Options options = CurrentOptions();
   options.statistics = rocksdb::CreateDBStatistics();
   DestroyAndReopen(options);
@@ -2099,7 +2131,7 @@ TEST_F(DBIteratorTest, SkipStatistics) {
   ASSERT_OK(Delete("e"));
   ASSERT_OK(Delete("f"));
 
-  Iterator* iter = db_->NewIterator(ReadOptions());
+  Iterator* iter = NewIterator(ReadOptions());
   int count = 0;
   for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
     ASSERT_OK(iter->status());
@@ -2110,7 +2142,7 @@ TEST_F(DBIteratorTest, SkipStatistics) {
   skip_count += 8; // 3 deletes + 3 original keys + 2 lower in sequence
   ASSERT_EQ(skip_count, TestGetTickerCount(options, NUMBER_ITER_SKIP));
 
-  iter = db_->NewIterator(ReadOptions());
+  iter = NewIterator(ReadOptions());
   count = 0;
   for (iter->SeekToLast(); iter->Valid(); iter->Prev()) {
     ASSERT_OK(iter->status());
@@ -2134,7 +2166,7 @@ TEST_F(DBIteratorTest, SkipStatistics) {
   Slice prefix("b");
   ro.iterate_upper_bound = &prefix;
 
-  iter = db_->NewIterator(ro);
+  iter = NewIterator(ro);
   count = 0;
   for(iter->Seek("aa"); iter->Valid(); iter->Next()) {
     ASSERT_OK(iter->status());
@@ -2145,7 +2177,7 @@ TEST_F(DBIteratorTest, SkipStatistics) {
   skip_count += 6; // 3 deletes + 3 original keys
   ASSERT_EQ(skip_count, TestGetTickerCount(options, NUMBER_ITER_SKIP));
 
-  iter = db_->NewIterator(ro);
+  iter = NewIterator(ro);
   count = 0;
   for(iter->SeekToLast(); iter->Valid(); iter->Prev()) {
     ASSERT_OK(iter->status());
@@ -2158,7 +2190,55 @@ TEST_F(DBIteratorTest, SkipStatistics) {
   ASSERT_EQ(skip_count, TestGetTickerCount(options, NUMBER_ITER_SKIP));
 }
 
-TEST_F(DBIteratorTest, ReadCallback) {
+TEST_P(DBIteratorTest, SeekAfterHittingManyInternalKeys) {
+  Options options = CurrentOptions();
+  DestroyAndReopen(options);
+  ReadOptions ropts;
+  ropts.max_skippable_internal_keys = 2;
+
+  Put("1", "val_1");
+  // Add more tombstones than max_skippable_internal_keys so that Next() fails.
+  Delete("2");
+  Delete("3");
+  Delete("4");
+  Delete("5");
+  Put("6", "val_6");
+
+  unique_ptr<Iterator> iter(NewIterator(ropts));
+  iter->SeekToFirst();
+
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(iter->key().ToString(), "1");
+  ASSERT_EQ(iter->value().ToString(), "val_1");
+
+  // This should fail as incomplete due to too many non-visible internal keys on
+  // the way to the next valid user key.
+  iter->Next();
+  ASSERT_TRUE(!iter->Valid());
+  ASSERT_TRUE(iter->status().IsIncomplete());
+
+  // Get the internal key at which Next() failed.
+  std::string prop_value;
+  ASSERT_OK(iter->GetProperty("rocksdb.iterator.internal-key", &prop_value));
+  ASSERT_EQ("4", prop_value);
+
+  // Create a new iterator to seek to the internal key.
+  unique_ptr<Iterator> iter2(NewIterator(ropts));
+  iter2->Seek(prop_value);
+  ASSERT_TRUE(iter2->Valid());
+  ASSERT_OK(iter2->status());
+
+  ASSERT_EQ(iter2->key().ToString(), "6");
+  ASSERT_EQ(iter2->value().ToString(), "val_6");
+}
+
+INSTANTIATE_TEST_CASE_P(DBIteratorTestInstance, DBIteratorTest,
+                        testing::Values(true, false));
+
+// Tests how DBIter work with ReadCallback
+class DBIteratorWithReadCallbackTest : public DBIteratorTest {};
+
+TEST_F(DBIteratorWithReadCallbackTest, ReadCallback) {
   class TestReadCallback : public ReadCallback {
    public:
     explicit TestReadCallback(SequenceNumber last_visible_seq)

db/db_memtable_test.cc

@@ -28,13 +28,11 @@ class MockMemTableRep : public MemTableRep {
     return rep_->Allocate(len, buf);
   }
 
-  virtual void Insert(KeyHandle handle) override {
-    return rep_->Insert(handle);
-  }
+  virtual void Insert(KeyHandle handle) override { rep_->Insert(handle); }
 
   virtual void InsertWithHint(KeyHandle handle, void** hint) override {
     num_insert_with_hint_++;
-    ASSERT_NE(nullptr, hint);
+    EXPECT_NE(nullptr, hint);
     last_hint_in_ = *hint;
     rep_->InsertWithHint(handle, hint);
     last_hint_out_ = *hint;
@@ -121,7 +119,7 @@ class TestPrefixExtractor : public SliceTransform {
     return separator(key) != nullptr;
   }
 
-  virtual bool InRange(const Slice& key) const override { return false; }
+  virtual bool InRange(const Slice& /*key*/) const override { return false; }
 
  private:
   const char* separator(const Slice& key) const {
@@ -129,6 +127,84 @@ class TestPrefixExtractor : public SliceTransform {
   }
 };
 
+// Test that ::Add properly returns false when inserting duplicate keys
+TEST_F(DBMemTableTest, DuplicateSeq) {
+  SequenceNumber seq = 123;
+  std::string value;
+  Status s;
+  MergeContext merge_context;
+  Options options;
+  InternalKeyComparator ikey_cmp(options.comparator);
+  RangeDelAggregator range_del_agg(ikey_cmp, {} /* snapshots */);
+
+  // Create a MemTable
+  InternalKeyComparator cmp(BytewiseComparator());
+  auto factory = std::make_shared<SkipListFactory>();
+  options.memtable_factory = factory;
+  ImmutableCFOptions ioptions(options);
+  WriteBufferManager wb(options.db_write_buffer_size);
+  MemTable* mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
+                               kMaxSequenceNumber, 0 /* column_family_id */);
+
+  // Write some keys and make sure it returns false on duplicates
+  bool res;
+  res = mem->Add(seq, kTypeValue, "key", "value2");
+  ASSERT_TRUE(res);
+  res = mem->Add(seq, kTypeValue, "key", "value2");
+  ASSERT_FALSE(res);
+  // Changing the type should still cause the duplicatae key
+  res = mem->Add(seq, kTypeMerge, "key", "value2");
+  ASSERT_FALSE(res);
+  // Changing the seq number will make the key fresh
+  res = mem->Add(seq + 1, kTypeMerge, "key", "value2");
+  ASSERT_TRUE(res);
+  // Test with different types for duplicate keys
+  res = mem->Add(seq, kTypeDeletion, "key", "");
+  ASSERT_FALSE(res);
+  res = mem->Add(seq, kTypeSingleDeletion, "key", "");
+  ASSERT_FALSE(res);
+
+  // Test the duplicate keys under stress
+  for (int i = 0; i < 10000; i++) {
+    bool insert_dup = i % 10 == 1;
+    if (!insert_dup) {
+      seq++;
+    }
+    res = mem->Add(seq, kTypeValue, "foo", "value" + ToString(seq));
+    if (insert_dup) {
+      ASSERT_FALSE(res);
+    } else {
+      ASSERT_TRUE(res);
+    }
+  }
+  delete mem;
+
+  // Test with InsertWithHint
+  options.memtable_insert_with_hint_prefix_extractor.reset(
+      new TestPrefixExtractor());  // which uses _ to extract the prefix
+  ioptions = ImmutableCFOptions(options);
+  mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
+                     kMaxSequenceNumber, 0 /* column_family_id */);
+  // Insert a duplicate key with _ in it
+  res = mem->Add(seq, kTypeValue, "key_1", "value");
+  ASSERT_TRUE(res);
+  res = mem->Add(seq, kTypeValue, "key_1", "value");
+  ASSERT_FALSE(res);
+  delete mem;
+
+  // Test when InsertConcurrently will be invoked
+  options.allow_concurrent_memtable_write = true;
+  ioptions = ImmutableCFOptions(options);
+  mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
+                     kMaxSequenceNumber, 0 /* column_family_id */);
+  MemTablePostProcessInfo post_process_info;
+  res = mem->Add(seq, kTypeValue, "key", "value", true, &post_process_info);
+  ASSERT_TRUE(res);
+  res = mem->Add(seq, kTypeValue, "key", "value", true, &post_process_info);
+  ASSERT_FALSE(res);
+  delete mem;
+}
+
 TEST_F(DBMemTableTest, InsertWithHint) {
   Options options;
   options.allow_concurrent_memtable_write = false;

db/db_merge_operator_test.cc

@@ -14,10 +14,43 @@
 
 namespace rocksdb {
 
+class TestReadCallback : public ReadCallback {
+ public:
+  TestReadCallback(SnapshotChecker* snapshot_checker,
+                   SequenceNumber snapshot_seq)
+      : snapshot_checker_(snapshot_checker), snapshot_seq_(snapshot_seq) {}
+
+  bool IsCommitted(SequenceNumber seq) override {
+    return snapshot_checker_->IsInSnapshot(seq, snapshot_seq_);
+  }
+
+ private:
+  SnapshotChecker* snapshot_checker_;
+  SequenceNumber snapshot_seq_;
+};
+
 // Test merge operator functionality.
 class DBMergeOperatorTest : public DBTestBase {
  public:
   DBMergeOperatorTest() : DBTestBase("/db_merge_operator_test") {}
+
+  std::string GetWithReadCallback(SnapshotChecker* snapshot_checker,
+                                  const Slice& key,
+                                  const Snapshot* snapshot = nullptr) {
+    SequenceNumber seq = snapshot == nullptr ? db_->GetLatestSequenceNumber()
+                                             : snapshot->GetSequenceNumber();
+    TestReadCallback read_callback(snapshot_checker, seq);
+    ReadOptions read_opt;
+    read_opt.snapshot = snapshot;
+    PinnableSlice value;
+    Status s =
+        dbfull()->GetImpl(read_opt, db_->DefaultColumnFamily(), key, &value,
+                          nullptr /*value_found*/, &read_callback);
+    if (!s.ok()) {
+      return s.ToString();
+    }
+    return value.ToString();
+  }
 };
 
 TEST_F(DBMergeOperatorTest, LimitMergeOperands) {
@@ -449,8 +482,152 @@ TEST_P(MergeOperatorPinningTest, TailingIterator) {
   writer_thread.join();
   reader_thread.join();
 }
+
+TEST_F(DBMergeOperatorTest, TailingIteratorMemtableUnrefedBySomeoneElse) {
+  Options options = CurrentOptions();
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  DestroyAndReopen(options);
+
+  // Overview of the test:
+  //  * There are two merge operands for the same key: one in an sst file,
+  //    another in a memtable.
+  //  * Seek a tailing iterator to this key.
+  //  * As part of the seek, the iterator will:
+  //      (a) first visit the operand in the memtable and tell ForwardIterator
+  //          to pin this operand, then
+  //      (b) move on to the operand in the sst file, then pass both operands
+  //          to merge operator.
+  //  * The memtable may get flushed and unreferenced by another thread between
+  //    (a) and (b). The test simulates it by flushing the memtable inside a
+  //    SyncPoint callback located between (a) and (b).
+  //  * In this case it's ForwardIterator's responsibility to keep the memtable
+  //    pinned until (b) is complete. There used to be a bug causing
+  //    ForwardIterator to not pin it in some circumstances. This test
+  //    reproduces it.
+
+  db_->Merge(WriteOptions(), "key", "sst");
+  db_->Flush(FlushOptions()); // Switch to SuperVersion A
+  db_->Merge(WriteOptions(), "key", "memtable");
+
+  // Pin SuperVersion A
+  std::unique_ptr<Iterator> someone_else(db_->NewIterator(ReadOptions()));
+
+  bool pushed_first_operand = false;
+  bool stepped_to_next_operand = false;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBIter::MergeValuesNewToOld:PushedFirstOperand", [&](void*) {
+        EXPECT_FALSE(pushed_first_operand);
+        pushed_first_operand = true;
+        db_->Flush(FlushOptions()); // Switch to SuperVersion B
+      });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBIter::MergeValuesNewToOld:SteppedToNextOperand", [&](void*) {
+        EXPECT_FALSE(stepped_to_next_operand);
+        stepped_to_next_operand = true;
+        someone_else.reset(); // Unpin SuperVersion A
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  ReadOptions ro;
+  ro.tailing = true;
+  std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+  iter->Seek("key");
+
+  ASSERT_TRUE(iter->status().ok());
+  ASSERT_TRUE(iter->Valid());
+  EXPECT_EQ(std::string("sst,memtable"), iter->value().ToString());
+  EXPECT_TRUE(pushed_first_operand);
+  EXPECT_TRUE(stepped_to_next_operand);
+}
 #endif  // ROCKSDB_LITE
 
+TEST_F(DBMergeOperatorTest, SnapshotCheckerAndReadCallback) {
+  Options options = CurrentOptions();
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  DestroyAndReopen(options);
+
+  class TestSnapshotChecker : public SnapshotChecker {
+    bool IsInSnapshot(SequenceNumber seq,
+                      SequenceNumber snapshot_seq) const override {
+      switch (snapshot_seq) {
+        case 0:
+          return seq == 0;
+        case 1:
+          return seq <= 1;
+        case 2:
+          // seq = 2 not visible to snapshot with seq = 2
+          return seq <= 1;
+        case 3:
+          return seq <= 3;
+        case 4:
+          // seq = 4 not visible to snpahost with seq = 4
+          return seq <= 3;
+        default:
+          // seq >=4 is uncommitted
+          return seq <= 4;
+      };
+    }
+  };
+  TestSnapshotChecker* snapshot_checker = new TestSnapshotChecker();
+  dbfull()->SetSnapshotChecker(snapshot_checker);
+
+  std::string value;
+  ASSERT_OK(Merge("foo", "v1"));
+  ASSERT_EQ(1, db_->GetLatestSequenceNumber());
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo"));
+  ASSERT_OK(Merge("foo", "v2"));
+  ASSERT_EQ(2, db_->GetLatestSequenceNumber());
+  // v2 is not visible to latest snapshot, which has seq = 2.
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo"));
+  // Take a snapshot with seq = 2.
+  const Snapshot* snapshot1 = db_->GetSnapshot();
+  ASSERT_EQ(2, snapshot1->GetSequenceNumber());
+  // v2 is not visible to snapshot1, which has seq = 2
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo", snapshot1));
+
+  // Verify flush doesn't alter the result.
+  ASSERT_OK(Flush());
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo", snapshot1));
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo"));
+
+  ASSERT_OK(Merge("foo", "v3"));
+  ASSERT_EQ(3, db_->GetLatestSequenceNumber());
+  ASSERT_EQ("v1,v2,v3", GetWithReadCallback(snapshot_checker, "foo"));
+  ASSERT_OK(Merge("foo", "v4"));
+  ASSERT_EQ(4, db_->GetLatestSequenceNumber());
+  // v4 is not visible to latest snapshot, which has seq = 4.
+  ASSERT_EQ("v1,v2,v3", GetWithReadCallback(snapshot_checker, "foo"));
+  const Snapshot* snapshot2 = db_->GetSnapshot();
+  ASSERT_EQ(4, snapshot2->GetSequenceNumber());
+  // v4 is not visible to snapshot2, which has seq = 4.
+  ASSERT_EQ("v1,v2,v3",
+            GetWithReadCallback(snapshot_checker, "foo", snapshot2));
+
+  // Verify flush doesn't alter the result.
+  ASSERT_OK(Flush());
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo", snapshot1));
+  ASSERT_EQ("v1,v2,v3",
+            GetWithReadCallback(snapshot_checker, "foo", snapshot2));
+  ASSERT_EQ("v1,v2,v3", GetWithReadCallback(snapshot_checker, "foo"));
+
+  ASSERT_OK(Merge("foo", "v5"));
+  ASSERT_EQ(5, db_->GetLatestSequenceNumber());
+  // v5 is uncommitted
+  ASSERT_EQ("v1,v2,v3,v4", GetWithReadCallback(snapshot_checker, "foo"));
+
+  // full manual compaction.
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+
+  // Verify compaction doesn't alter the result.
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo", snapshot1));
+  ASSERT_EQ("v1,v2,v3",
+            GetWithReadCallback(snapshot_checker, "foo", snapshot2));
+  ASSERT_EQ("v1,v2,v3,v4", GetWithReadCallback(snapshot_checker, "foo"));
+
+  db_->ReleaseSnapshot(snapshot1);
+  db_->ReleaseSnapshot(snapshot2);
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/db_properties_test.cc

@@ -14,6 +14,7 @@
 
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
+#include "rocksdb/listener.h"
 #include "rocksdb/options.h"
 #include "rocksdb/perf_context.h"
 #include "rocksdb/perf_level.h"
@@ -985,8 +986,9 @@ class CountingUserTblPropCollector : public TablePropertiesCollector {
     return Status::OK();
   }
 
-  Status AddUserKey(const Slice& user_key, const Slice& value, EntryType type,
-                    SequenceNumber seq, uint64_t file_size) override {
+  Status AddUserKey(const Slice& /*user_key*/, const Slice& /*value*/,
+                    EntryType /*type*/, SequenceNumber /*seq*/,
+                    uint64_t /*file_size*/) override {
     ++count_;
     return Status::OK();
   }
@@ -1027,8 +1029,9 @@ class CountingDeleteTabPropCollector : public TablePropertiesCollector {
  public:
   const char* Name() const override { return "CountingDeleteTabPropCollector"; }
 
-  Status AddUserKey(const Slice& user_key, const Slice& value, EntryType type,
-                    SequenceNumber seq, uint64_t file_size) override {
+  Status AddUserKey(const Slice& /*user_key*/, const Slice& /*value*/,
+                    EntryType type, SequenceNumber /*seq*/,
+                    uint64_t /*file_size*/) override {
     if (type == kEntryDelete) {
       num_deletes_++;
     }
@@ -1055,7 +1058,7 @@ class CountingDeleteTabPropCollectorFactory
     : public TablePropertiesCollectorFactory {
  public:
   virtual TablePropertiesCollector* CreateTablePropertiesCollector(
-      TablePropertiesCollectorFactory::Context context) override {
+      TablePropertiesCollectorFactory::Context /*context*/) override {
     return new CountingDeleteTabPropCollector();
   }
   const char* Name() const override {
@@ -1383,6 +1386,56 @@ TEST_F(DBPropertiesTest, EstimateOldestKeyTime) {
   Close();
 }
 
+TEST_F(DBPropertiesTest, SstFilesSize) {
+  struct TestListener : public EventListener {
+    void OnCompactionCompleted(DB* db,
+                               const CompactionJobInfo& /*info*/) override {
+      assert(callback_triggered == false);
+      assert(size_before_compaction > 0);
+      callback_triggered = true;
+      uint64_t total_sst_size = 0;
+      uint64_t live_sst_size = 0;
+      bool ok = db->GetIntProperty(DB::Properties::kTotalSstFilesSize,
+                                   &total_sst_size);
+      ASSERT_TRUE(ok);
+      // total_sst_size include files before and after compaction.
+      ASSERT_GT(total_sst_size, size_before_compaction);
+      ok =
+          db->GetIntProperty(DB::Properties::kLiveSstFilesSize, &live_sst_size);
+      ASSERT_TRUE(ok);
+      // live_sst_size only include files after compaction.
+      ASSERT_GT(live_sst_size, 0);
+      ASSERT_LT(live_sst_size, size_before_compaction);
+    }
+
+    uint64_t size_before_compaction = 0;
+    bool callback_triggered = false;
+  };
+  std::shared_ptr<TestListener> listener = std::make_shared<TestListener>();
+
+  Options options;
+  options.disable_auto_compactions = true;
+  options.listeners.push_back(listener);
+  Reopen(options);
+
+  for (int i = 0; i < 10; i++) {
+    ASSERT_OK(Put("key" + ToString(i), std::string(1000, 'v')));
+  }
+  ASSERT_OK(Flush());
+  for (int i = 0; i < 5; i++) {
+    ASSERT_OK(Delete("key" + ToString(i)));
+  }
+  ASSERT_OK(Flush());
+  uint64_t sst_size;
+  bool ok = db_->GetIntProperty(DB::Properties::kTotalSstFilesSize, &sst_size);
+  ASSERT_TRUE(ok);
+  ASSERT_GT(sst_size, 0);
+  listener->size_before_compaction = sst_size;
+  // Compact to clean all keys and trigger listener.
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_TRUE(listener->callback_triggered);
+}
+
 #endif  // ROCKSDB_LITE
 }  // namespace rocksdb
 

db/db_test.cc

@@ -480,6 +480,36 @@ TEST_F(DBTest, SingleDeletePutFlush) {
                          kSkipUniversalCompaction | kSkipMergePut));
 }
 
+// Disable because not all platform can run it.
+// It requires more than 9GB memory to run it, With single allocation
+// of more than 3GB.
+TEST_F(DBTest, DISABLED_SanitizeVeryVeryLargeValue) {
+  const size_t kValueSize = 4 * size_t{1024 * 1024 * 1024};  // 4GB value
+  std::string raw(kValueSize, 'v');
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.merge_operator = MergeOperators::CreatePutOperator();
+  options.write_buffer_size = 100000;  // Small write buffer
+  options.paranoid_checks = true;
+  DestroyAndReopen(options);
+
+  ASSERT_OK(Put("boo", "v1"));
+  ASSERT_TRUE(Put("foo", raw).IsInvalidArgument());
+  ASSERT_TRUE(Merge("foo", raw).IsInvalidArgument());
+
+  WriteBatch wb;
+  ASSERT_TRUE(wb.Put("foo", raw).IsInvalidArgument());
+  ASSERT_TRUE(wb.Merge("foo", raw).IsInvalidArgument());
+
+  Slice value_slice = raw;
+  Slice key_slice = "foo";
+  SliceParts sp_key(&key_slice, 1);
+  SliceParts sp_value(&value_slice, 1);
+
+  ASSERT_TRUE(wb.Put(sp_key, sp_value).IsInvalidArgument());
+  ASSERT_TRUE(wb.Merge(sp_key, sp_value).IsInvalidArgument());
+}
+
 // Disable because not all platform can run it.
 // It requires more than 9GB memory to run it, With single allocation
 // of more than 3GB.
@@ -720,9 +750,9 @@ TEST_F(DBTest, FlushSchedule) {
 namespace {
 class KeepFilter : public CompactionFilter {
  public:
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value,
-                      bool* value_changed) const override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& /*value*/, std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     return false;
   }
 
@@ -752,9 +782,9 @@ class KeepFilterFactory : public CompactionFilterFactory {
 class DelayFilter : public CompactionFilter {
  public:
   explicit DelayFilter(DBTestBase* d) : db_test(d) {}
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value,
-                      bool* value_changed) const override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& /*value*/, std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     db_test->env_->addon_time_.fetch_add(1000);
     return true;
   }
@@ -769,7 +799,7 @@ class DelayFilterFactory : public CompactionFilterFactory {
  public:
   explicit DelayFilterFactory(DBTestBase* d) : db_test(d) {}
   virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
-      const CompactionFilter::Context& context) override {
+      const CompactionFilter::Context& /*context*/) override {
     return std::unique_ptr<CompactionFilter>(new DelayFilter(db_test));
   }
 
@@ -2217,17 +2247,17 @@ class ModelDB : public DB {
     return Write(o, &batch);
   }
   using DB::Get;
-  virtual Status Get(const ReadOptions& options, ColumnFamilyHandle* cf,
-                     const Slice& key, PinnableSlice* value) override {
+  virtual Status Get(const ReadOptions& /*options*/, ColumnFamilyHandle* /*cf*/,
+                     const Slice& key, PinnableSlice* /*value*/) override {
     return Status::NotSupported(key);
   }
 
   using DB::MultiGet;
   virtual std::vector<Status> MultiGet(
-      const ReadOptions& options,
-      const std::vector<ColumnFamilyHandle*>& column_family,
+      const ReadOptions& /*options*/,
+      const std::vector<ColumnFamilyHandle*>& /*column_family*/,
       const std::vector<Slice>& keys,
-      std::vector<std::string>* values) override {
+      std::vector<std::string>* /*values*/) override {
     std::vector<Status> s(keys.size(),
                           Status::NotSupported("Not implemented."));
     return s;
@@ -2236,9 +2266,9 @@ class ModelDB : public DB {
 #ifndef ROCKSDB_LITE
   using DB::IngestExternalFile;
   virtual Status IngestExternalFile(
-      ColumnFamilyHandle* column_family,
-      const std::vector<std::string>& external_files,
-      const IngestExternalFileOptions& options) override {
+      ColumnFamilyHandle* /*column_family*/,
+      const std::vector<std::string>& /*external_files*/,
+      const IngestExternalFileOptions& /*options*/) override {
     return Status::NotSupported("Not implemented.");
   }
 
@@ -2248,22 +2278,22 @@ class ModelDB : public DB {
 
   using DB::GetPropertiesOfAllTables;
   virtual Status GetPropertiesOfAllTables(
-      ColumnFamilyHandle* column_family,
-      TablePropertiesCollection* props) override {
+      ColumnFamilyHandle* /*column_family*/,
+      TablePropertiesCollection* /*props*/) override {
     return Status();
   }
 
   virtual Status GetPropertiesOfTablesInRange(
-      ColumnFamilyHandle* column_family, const Range* range, std::size_t n,
-      TablePropertiesCollection* props) override {
+      ColumnFamilyHandle* /*column_family*/, const Range* /*range*/,
+      std::size_t /*n*/, TablePropertiesCollection* /*props*/) override {
     return Status();
   }
 #endif  // ROCKSDB_LITE
 
   using DB::KeyMayExist;
-  virtual bool KeyMayExist(const ReadOptions& options,
-                           ColumnFamilyHandle* column_family, const Slice& key,
-                           std::string* value,
+  virtual bool KeyMayExist(const ReadOptions& /*options*/,
+                           ColumnFamilyHandle* /*column_family*/,
+                           const Slice& /*key*/, std::string* /*value*/,
                            bool* value_found = nullptr) override {
     if (value_found != nullptr) {
       *value_found = false;
@@ -2271,8 +2301,9 @@ class ModelDB : public DB {
     return true;  // Not Supported directly
   }
   using DB::NewIterator;
-  virtual Iterator* NewIterator(const ReadOptions& options,
-                                ColumnFamilyHandle* column_family) override {
+  virtual Iterator* NewIterator(
+      const ReadOptions& options,
+      ColumnFamilyHandle* /*column_family*/) override {
     if (options.snapshot == nullptr) {
       KVMap* saved = new KVMap;
       *saved = map_;
@@ -2284,9 +2315,9 @@ class ModelDB : public DB {
     }
   }
   virtual Status NewIterators(
-      const ReadOptions& options,
-      const std::vector<ColumnFamilyHandle*>& column_family,
-      std::vector<Iterator*>* iterators) override {
+      const ReadOptions& /*options*/,
+      const std::vector<ColumnFamilyHandle*>& /*column_family*/,
+      std::vector<Iterator*>* /*iterators*/) override {
     return Status::NotSupported("Not supported yet");
   }
   virtual const Snapshot* GetSnapshot() override {
@@ -2299,7 +2330,7 @@ class ModelDB : public DB {
     delete reinterpret_cast<const ModelSnapshot*>(snapshot);
   }
 
-  virtual Status Write(const WriteOptions& options,
+  virtual Status Write(const WriteOptions& /*options*/,
                        WriteBatch* batch) override {
     class Handler : public WriteBatch::Handler {
      public:
@@ -2307,7 +2338,8 @@ class ModelDB : public DB {
       virtual void Put(const Slice& key, const Slice& value) override {
         (*map_)[key.ToString()] = value.ToString();
       }
-      virtual void Merge(const Slice& key, const Slice& value) override {
+      virtual void Merge(const Slice& /*key*/,
+                         const Slice& /*value*/) override {
         // ignore merge for now
         // (*map_)[key.ToString()] = value.ToString();
       }
@@ -2321,62 +2353,65 @@ class ModelDB : public DB {
   }
 
   using DB::GetProperty;
-  virtual bool GetProperty(ColumnFamilyHandle* column_family,
-                           const Slice& property, std::string* value) override {
+  virtual bool GetProperty(ColumnFamilyHandle* /*column_family*/,
+                           const Slice& /*property*/,
+                           std::string* /*value*/) override {
     return false;
   }
   using DB::GetIntProperty;
-  virtual bool GetIntProperty(ColumnFamilyHandle* column_family,
-                              const Slice& property, uint64_t* value) override {
+  virtual bool GetIntProperty(ColumnFamilyHandle* /*column_family*/,
+                              const Slice& /*property*/,
+                              uint64_t* /*value*/) override {
     return false;
   }
   using DB::GetMapProperty;
   virtual bool GetMapProperty(
-      ColumnFamilyHandle* column_family, const Slice& property,
-      std::map<std::string, std::string>* value) override {
+      ColumnFamilyHandle* /*column_family*/, const Slice& /*property*/,
+      std::map<std::string, std::string>* /*value*/) override {
     return false;
   }
   using DB::GetAggregatedIntProperty;
-  virtual bool GetAggregatedIntProperty(const Slice& property,
-                                        uint64_t* value) override {
+  virtual bool GetAggregatedIntProperty(const Slice& /*property*/,
+                                        uint64_t* /*value*/) override {
     return false;
   }
   using DB::GetApproximateSizes;
-  virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
-                                   const Range* range, int n, uint64_t* sizes,
-                                   uint8_t include_flags
+  virtual void GetApproximateSizes(ColumnFamilyHandle* /*column_family*/,
+                                   const Range* /*range*/, int n,
+                                   uint64_t* sizes,
+                                   uint8_t /*include_flags*/
                                    = INCLUDE_FILES) override {
     for (int i = 0; i < n; i++) {
       sizes[i] = 0;
     }
   }
   using DB::GetApproximateMemTableStats;
-  virtual void GetApproximateMemTableStats(ColumnFamilyHandle* column_family,
-                                           const Range& range,
-                                           uint64_t* const count,
-                                           uint64_t* const size) override {
+  virtual void GetApproximateMemTableStats(
+      ColumnFamilyHandle* /*column_family*/, const Range& /*range*/,
+      uint64_t* const count, uint64_t* const size) override {
     *count = 0;
     *size = 0;
   }
   using DB::CompactRange;
-  virtual Status CompactRange(const CompactRangeOptions& options,
-                              ColumnFamilyHandle* column_family,
-                              const Slice* start, const Slice* end) override {
+  virtual Status CompactRange(const CompactRangeOptions& /*options*/,
+                              ColumnFamilyHandle* /*column_family*/,
+                              const Slice* /*start*/,
+                              const Slice* /*end*/) override {
     return Status::NotSupported("Not supported operation.");
   }
 
   virtual Status SetDBOptions(
-      const std::unordered_map<std::string, std::string>& new_options)
+      const std::unordered_map<std::string, std::string>& /*new_options*/)
       override {
     return Status::NotSupported("Not supported operation.");
   }
 
   using DB::CompactFiles;
-  virtual Status CompactFiles(const CompactionOptions& compact_options,
-                              ColumnFamilyHandle* column_family,
-                              const std::vector<std::string>& input_file_names,
-                              const int output_level,
-                              const int output_path_id = -1) override {
+  virtual Status CompactFiles(
+      const CompactionOptions& /*compact_options*/,
+      ColumnFamilyHandle* /*column_family*/,
+      const std::vector<std::string>& /*input_file_names*/,
+      const int /*output_level*/, const int /*output_path_id*/ = -1) override {
     return Status::NotSupported("Not supported operation.");
   }
 
@@ -2389,24 +2424,25 @@ class ModelDB : public DB {
   }
 
   Status EnableAutoCompaction(
-      const std::vector<ColumnFamilyHandle*>& column_family_handles) override {
+      const std::vector<ColumnFamilyHandle*>& /*column_family_handles*/)
+      override {
     return Status::NotSupported("Not supported operation.");
   }
 
   using DB::NumberLevels;
-  virtual int NumberLevels(ColumnFamilyHandle* column_family) override {
+  virtual int NumberLevels(ColumnFamilyHandle* /*column_family*/) override {
     return 1;
   }
 
   using DB::MaxMemCompactionLevel;
   virtual int MaxMemCompactionLevel(
-      ColumnFamilyHandle* column_family) override {
+      ColumnFamilyHandle* /*column_family*/) override {
     return 1;
   }
 
   using DB::Level0StopWriteTrigger;
   virtual int Level0StopWriteTrigger(
-      ColumnFamilyHandle* column_family) override {
+      ColumnFamilyHandle* /*column_family*/) override {
     return -1;
   }
 
@@ -2415,7 +2451,8 @@ class ModelDB : public DB {
   virtual Env* GetEnv() const override { return nullptr; }
 
   using DB::GetOptions;
-  virtual Options GetOptions(ColumnFamilyHandle* column_family) const override {
+  virtual Options GetOptions(
+      ColumnFamilyHandle* /*column_family*/) const override {
     return options_;
   }
 
@@ -2423,8 +2460,8 @@ class ModelDB : public DB {
   virtual DBOptions GetDBOptions() const override { return options_; }
 
   using DB::Flush;
-  virtual Status Flush(const rocksdb::FlushOptions& options,
-                       ColumnFamilyHandle* column_family) override {
+  virtual Status Flush(const rocksdb::FlushOptions& /*options*/,
+                       ColumnFamilyHandle* /*column_family*/) override {
     Status ret;
     return ret;
   }
@@ -2434,39 +2471,42 @@ class ModelDB : public DB {
 #ifndef ROCKSDB_LITE
   virtual Status DisableFileDeletions() override { return Status::OK(); }
 
-  virtual Status EnableFileDeletions(bool force) override {
+  virtual Status EnableFileDeletions(bool /*force*/) override {
     return Status::OK();
   }
-  virtual Status GetLiveFiles(std::vector<std::string>&, uint64_t* size,
-                              bool flush_memtable = true) override {
+  virtual Status GetLiveFiles(std::vector<std::string>&, uint64_t* /*size*/,
+                              bool /*flush_memtable*/ = true) override {
     return Status::OK();
   }
 
-  virtual Status GetSortedWalFiles(VectorLogPtr& files) override {
+  virtual Status GetSortedWalFiles(VectorLogPtr& /*files*/) override {
     return Status::OK();
   }
 
-  virtual Status DeleteFile(std::string name) override { return Status::OK(); }
+  virtual Status DeleteFile(std::string /*name*/) override {
+    return Status::OK();
+  }
 
   virtual Status GetUpdatesSince(
       rocksdb::SequenceNumber, unique_ptr<rocksdb::TransactionLogIterator>*,
-      const TransactionLogIterator::ReadOptions& read_options =
+      const TransactionLogIterator::ReadOptions& /*read_options*/ =
           TransactionLogIterator::ReadOptions()) override {
     return Status::NotSupported("Not supported in Model DB");
   }
 
   virtual void GetColumnFamilyMetaData(
-      ColumnFamilyHandle* column_family,
-      ColumnFamilyMetaData* metadata) override {}
+      ColumnFamilyHandle* /*column_family*/,
+      ColumnFamilyMetaData* /*metadata*/) override {}
 #endif  // ROCKSDB_LITE
 
-  virtual Status GetDbIdentity(std::string& identity) const override {
+  virtual Status GetDbIdentity(std::string& /*identity*/) const override {
     return Status::OK();
   }
 
   virtual SequenceNumber GetLatestSequenceNumber() const override { return 0; }
 
-  virtual bool SetPreserveDeletesSequenceNumber(SequenceNumber seqnum) override {
+  virtual bool SetPreserveDeletesSequenceNumber(
+      SequenceNumber /*seqnum*/) override {
     return true;
   }
 
@@ -4746,7 +4786,7 @@ class DelayedMergeOperator : public MergeOperator {
  public:
   explicit DelayedMergeOperator(DBTest* d) : db_test_(d) {}
 
-  virtual bool FullMergeV2(const MergeOperationInput& merge_in,
+  virtual bool FullMergeV2(const MergeOperationInput& /*merge_in*/,
                            MergeOperationOutput* merge_out) const override {
     db_test_->env_->addon_time_.fetch_add(1000);
     merge_out->new_value = "";
@@ -5114,10 +5154,10 @@ TEST_F(DBTest, AutomaticConflictsWithManualCompaction) {
     ASSERT_OK(Flush());
   }
   std::thread manual_compaction_thread([this]() {
-      CompactRangeOptions croptions;
-      croptions.exclusive_manual_compaction = true;
-      ASSERT_OK(db_->CompactRange(croptions, nullptr, nullptr));
-    });
+    CompactRangeOptions croptions;
+    croptions.exclusive_manual_compaction = true;
+    ASSERT_OK(db_->CompactRange(croptions, nullptr, nullptr));
+  });
 
   TEST_SYNC_POINT("DBTest::AutomaticConflictsWithManualCompaction:PrePuts");
   for (int i = 0; i < kNumL0Files; ++i) {
@@ -5692,6 +5732,50 @@ TEST_F(DBTest, PauseBackgroundWorkTest) {
   // now it's done
   ASSERT_TRUE(done.load());
 }
+
+// Keep spawning short-living threads that create an iterator and quit.
+// Meanwhile in another thread keep flushing memtables.
+// This used to cause a deadlock.
+TEST_F(DBTest, ThreadLocalPtrDeadlock) {
+  std::atomic<int> flushes_done{0};
+  std::atomic<int> threads_destroyed{0};
+  auto done = [&] {
+    return flushes_done.load() > 10;
+  };
+
+  std::thread flushing_thread([&] {
+    for (int i = 0; !done(); ++i) {
+      ASSERT_OK(db_->Put(WriteOptions(), Slice("hi"),
+                         Slice(std::to_string(i).c_str())));
+      ASSERT_OK(db_->Flush(FlushOptions()));
+      int cnt = ++flushes_done;
+      fprintf(stderr, "Flushed %d times\n", cnt);
+    }
+  });
+
+  std::vector<std::thread> thread_spawning_threads(10);
+  for (auto& t: thread_spawning_threads) {
+    t = std::thread([&] {
+      while (!done()) {
+        {
+          std::thread tmp_thread([&] {
+            auto it = db_->NewIterator(ReadOptions());
+            delete it;
+          });
+          tmp_thread.join();
+        }
+        ++threads_destroyed;
+      }
+    });
+  }
+
+  for (auto& t: thread_spawning_threads) {
+    t.join();
+  }
+  flushing_thread.join();
+  fprintf(stderr, "Done. Flushed %d times, destroyed %d threads\n",
+          flushes_done.load(), threads_destroyed.load());
+}
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/db_test2.cc

@@ -498,9 +498,9 @@ TEST_F(DBTest2, WalFilterTest) {
       apply_option_at_record_index_(apply_option_for_record_index),
       current_record_index_(0) {}
 
-    virtual WalProcessingOption LogRecord(const WriteBatch& batch,
-      WriteBatch* new_batch,
-      bool* batch_changed) const override {
+    virtual WalProcessingOption LogRecord(
+        const WriteBatch& /*batch*/, WriteBatch* /*new_batch*/,
+        bool* /*batch_changed*/) const override {
       WalFilter::WalProcessingOption option_to_return;
 
       if (current_record_index_ == apply_option_at_record_index_) {
@@ -874,11 +874,10 @@ TEST_F(DBTest2, WalFilterTestWithColumnFamilies) {
       cf_name_id_map_ = cf_name_id_map;
     }
 
-    virtual WalProcessingOption LogRecordFound(unsigned long long log_number,
-      const std::string& log_file_name,
-      const WriteBatch& batch,
-      WriteBatch* new_batch,
-      bool* batch_changed) override {
+    virtual WalProcessingOption LogRecordFound(
+        unsigned long long log_number, const std::string& /*log_file_name*/,
+        const WriteBatch& batch, WriteBatch* /*new_batch*/,
+        bool* /*batch_changed*/) override {
       class LogRecordBatchHandler : public WriteBatch::Handler {
       private:
         const std::map<uint32_t, uint64_t> & cf_log_number_map_;
@@ -1231,7 +1230,7 @@ class CompactionStallTestListener : public EventListener {
  public:
   CompactionStallTestListener() : compacted_files_cnt_(0) {}
 
-  void OnCompactionCompleted(DB* db, const CompactionJobInfo& ci) override {
+  void OnCompactionCompleted(DB* /*db*/, const CompactionJobInfo& ci) override {
     ASSERT_EQ(ci.cf_name, "default");
     ASSERT_EQ(ci.base_input_level, 0);
     ASSERT_EQ(ci.compaction_reason, CompactionReason::kLevelL0FilesNum);
@@ -1820,11 +1819,17 @@ TEST_F(DBTest2, ReadAmpBitmap) {
 
 #ifndef OS_SOLARIS // GetUniqueIdFromFile is not implemented
 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;
+  {
+    const int kIdBufLen = 100;
+    char id_buf[kIdBufLen];
+    std::unique_ptr<RandomAccessFile> file;
+    env_->NewRandomAccessFile(dbname_, &file, EnvOptions());
+    if (file->GetUniqueId(id_buf, kIdBufLen) == 0) {
+      // fs holding db directory doesn't support getting a unique file id,
+      // this means that running this test will fail because lru_cache will load
+      // the blocks again regardless of them being already in the cache
+      return;
+    }
   }
   uint32_t bytes_per_bit[2] = {1, 16};
   for (size_t k = 0; k < 2; k++) {

db/db_test_util.cc

@@ -1143,17 +1143,18 @@ UpdateStatus DBTestBase::updateInPlaceSmallerVarintSize(char* prevValue,
   }
 }
 
-UpdateStatus DBTestBase::updateInPlaceLargerSize(char* prevValue,
-                                                 uint32_t* prevSize,
+UpdateStatus DBTestBase::updateInPlaceLargerSize(char* /*prevValue*/,
+                                                 uint32_t* /*prevSize*/,
                                                  Slice delta,
                                                  std::string* newValue) {
   *newValue = std::string(delta.size(), 'c');
   return UpdateStatus::UPDATED;
 }
 
-UpdateStatus DBTestBase::updateInPlaceNoAction(char* prevValue,
-                                               uint32_t* prevSize, Slice delta,
-                                               std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceNoAction(char* /*prevValue*/,
+                                               uint32_t* /*prevSize*/,
+                                               Slice /*delta*/,
+                                               std::string* /*newValue*/) {
   return UpdateStatus::UPDATE_FAILED;
 }
 

db/db_test_util.h

@@ -137,8 +137,8 @@ class SpecialMemTableRep : public MemTableRep {
   // Insert key into the list.
   // REQUIRES: nothing that compares equal to key is currently in the list.
   virtual void Insert(KeyHandle handle) override {
-    memtable_->Insert(handle);
     num_entries_++;
+    memtable_->Insert(handle);
   }
 
   // Returns true iff an entry that compares equal to key is in the list.
@@ -187,7 +187,7 @@ class SpecialSkipListFactory : public MemTableRepFactory {
   using MemTableRepFactory::CreateMemTableRep;
   virtual MemTableRep* CreateMemTableRep(
       const MemTableRep::KeyComparator& compare, Allocator* allocator,
-      const SliceTransform* transform, Logger* logger) override {
+      const SliceTransform* transform, Logger* /*logger*/) override {
     return new SpecialMemTableRep(
         allocator, factory_.CreateMemTableRep(compare, allocator, transform, 0),
         num_entries_flush_);

db/db_universal_compaction_test.cc

@@ -56,9 +56,9 @@ void VerifyCompactionResult(
 
 class KeepFilter : public CompactionFilter {
  public:
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value, bool* value_changed) const
-      override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& /*value*/, std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     return false;
   }
 
@@ -88,9 +88,9 @@ class KeepFilterFactory : public CompactionFilterFactory {
 class DelayFilter : public CompactionFilter {
  public:
   explicit DelayFilter(DBTestBase* d) : db_test(d) {}
-  virtual bool Filter(int level, const Slice& key, const Slice& value,
-                      std::string* new_value,
-                      bool* value_changed) const override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& /*value*/, std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     db_test->env_->addon_time_.fetch_add(1000);
     return true;
   }
@@ -105,7 +105,7 @@ class DelayFilterFactory : public CompactionFilterFactory {
  public:
   explicit DelayFilterFactory(DBTestBase* d) : db_test(d) {}
   virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
-      const CompactionFilter::Context& context) override {
+      const CompactionFilter::Context& /*context*/) override {
     return std::unique_ptr<CompactionFilter>(new DelayFilter(db_test));
   }
 
@@ -388,10 +388,16 @@ TEST_P(DBTestUniversalCompaction, DynamicUniversalCompactionSizeAmplification) {
   DestroyAndReopen(options);
 
   int total_picked_compactions = 0;
+  int total_size_amp_compactions = 0;
   rocksdb::SyncPoint::GetInstance()->SetCallBack(
       "UniversalCompactionPicker::PickCompaction:Return", [&](void* arg) {
         if (arg) {
           total_picked_compactions++;
+          Compaction* c = static_cast<Compaction*>(arg);
+          if (c->compaction_reason() ==
+              CompactionReason::kUniversalSizeAmplification) {
+            total_size_amp_compactions++;
+          }
         }
       });
   rocksdb::SyncPoint::GetInstance()->EnableProcessing();
@@ -445,6 +451,102 @@ TEST_P(DBTestUniversalCompaction, DynamicUniversalCompactionSizeAmplification) {
   // Verify that size amplification did happen
   ASSERT_EQ(NumSortedRuns(1), 1);
   ASSERT_EQ(total_picked_compactions, 1);
+  ASSERT_EQ(total_size_amp_compactions, 1);
+}
+
+TEST_P(DBTestUniversalCompaction, DynamicUniversalCompactionReadAmplification) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.num_levels = 1;
+  options.write_buffer_size = 100 << 10;     // 100KB
+  options.target_file_size_base = 32 << 10;  // 32KB
+  options.level0_file_num_compaction_trigger = 3;
+  // Initial setup of compaction_options_universal will prevent universal
+  // compaction from happening
+  options.compaction_options_universal.max_size_amplification_percent = 2000;
+  options.compaction_options_universal.size_ratio = 0;
+  options.compaction_options_universal.min_merge_width = 100;
+  DestroyAndReopen(options);
+
+  int total_picked_compactions = 0;
+  int total_size_ratio_compactions = 0;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "UniversalCompactionPicker::PickCompaction:Return", [&](void* arg) {
+        if (arg) {
+          total_picked_compactions++;
+          Compaction* c = static_cast<Compaction*>(arg);
+          if (c->compaction_reason() == CompactionReason::kUniversalSizeRatio) {
+            total_size_ratio_compactions++;
+          }
+        }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  MutableCFOptions mutable_cf_options;
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  Random rnd(301);
+  int key_idx = 0;
+
+  // Generate three files in Level 0. All files are approx the same size.
+  for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+    // Write 110KB (11 values, each 10K)
+    for (int i = 0; i < 11; i++) {
+      ASSERT_OK(Put(1, Key(key_idx), RandomString(&rnd, 10000)));
+      key_idx++;
+    }
+    dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+    ASSERT_EQ(NumSortedRuns(1), num + 1);
+  }
+  ASSERT_EQ(NumSortedRuns(1), options.level0_file_num_compaction_trigger);
+
+  // Flush whatever is remaining in memtable. This is typically small, about
+  // 30KB.
+  ASSERT_OK(Flush(1));
+  dbfull()->TEST_WaitForCompact();
+  // Verify compaction did not happen
+  ASSERT_EQ(NumSortedRuns(1), options.level0_file_num_compaction_trigger + 1);
+  ASSERT_EQ(total_picked_compactions, 0);
+
+  ASSERT_OK(dbfull()->SetOptions(
+      handles_[1],
+      {{"compaction_options_universal",
+        "{min_merge_width=2;max_merge_width=2;size_ratio=100;}"}}));
+  ASSERT_EQ(dbfull()
+                ->GetOptions(handles_[1])
+                .compaction_options_universal.min_merge_width,
+            2);
+  ASSERT_EQ(dbfull()
+                ->GetOptions(handles_[1])
+                .compaction_options_universal.max_merge_width,
+            2);
+  ASSERT_EQ(
+      dbfull()->GetOptions(handles_[1]).compaction_options_universal.size_ratio,
+      100);
+
+  ASSERT_OK(dbfull()->TEST_GetLatestMutableCFOptions(handles_[1],
+                                                     &mutable_cf_options));
+  ASSERT_EQ(mutable_cf_options.compaction_options_universal.size_ratio, 100);
+  ASSERT_EQ(mutable_cf_options.compaction_options_universal.min_merge_width, 2);
+  ASSERT_EQ(mutable_cf_options.compaction_options_universal.max_merge_width, 2);
+
+  dbfull()->TEST_WaitForCompact();
+
+  // Files in L0 are approx: 0.3 (30KB), 1, 1, 1.
+  // On compaction: the files are below the size amp threshold, so we
+  // fallthrough to checking read amp conditions. The configured size ratio is
+  // not big enough to take 0.3 into consideration. So the next files 1 and 1
+  // are compacted together first as they satisfy size ratio condition and
+  // (min_merge_width, max_merge_width) condition, to give out a file size of 2.
+  // Next, the newly generated 2 and the last file 1 are compacted together. So
+  // at the end: #sortedRuns = 2, #picked_compactions = 2, and all the picked
+  // ones are size ratio based compactions.
+  ASSERT_EQ(NumSortedRuns(1), 2);
+  // If max_merge_width had not been changed dynamically above, and if it
+  // continued to be the default value of UINIT_MAX, total_picked_compactions
+  // would have been 1.
+  ASSERT_EQ(total_picked_compactions, 2);
+  ASSERT_EQ(total_size_ratio_compactions, 2);
 }
 
 TEST_P(DBTestUniversalCompaction, CompactFilesOnUniversalCompaction) {

db/dbformat.cc

@@ -125,6 +125,26 @@ int InternalKeyComparator::Compare(const Slice& akey, const Slice& bkey) const {
   return r;
 }
 
+int InternalKeyComparator::CompareKeySeq(const Slice& akey,
+                                         const Slice& bkey) const {
+  // Order by:
+  //    increasing user key (according to user-supplied comparator)
+  //    decreasing sequence number
+  int r = user_comparator_->Compare(ExtractUserKey(akey), ExtractUserKey(bkey));
+  PERF_COUNTER_ADD(user_key_comparison_count, 1);
+  if (r == 0) {
+    // Shift the number to exclude the last byte which contains the value type
+    const uint64_t anum = DecodeFixed64(akey.data() + akey.size() - 8) >> 8;
+    const uint64_t bnum = DecodeFixed64(bkey.data() + bkey.size() - 8) >> 8;
+    if (anum > bnum) {
+      r = -1;
+    } else if (anum < bnum) {
+      r = +1;
+    }
+  }
+  return r;
+}
+
 int InternalKeyComparator::Compare(const ParsedInternalKey& a,
                                    const ParsedInternalKey& b) const {
   // Order by:

db/dbformat.h

@@ -162,6 +162,8 @@ class InternalKeyComparator
 
   virtual const char* Name() const override;
   virtual int Compare(const Slice& a, const Slice& b) const override;
+  // Same as Compare except that it excludes the value type from comparison
+  virtual int CompareKeySeq(const Slice& a, const Slice& b) const;
   virtual void FindShortestSeparator(std::string* start,
                                      const Slice& limit) const override;
   virtual void FindShortSuccessor(std::string* key) const override;

db/deletefile_test.cc

@@ -159,7 +159,7 @@ class DeleteFileTest : public testing::Test {
   }
 
   // An empty job to guard all jobs are processed
-  static void GuardFinish(void* arg) {
+  static void GuardFinish(void* /*arg*/) {
     TEST_SYNC_POINT("DeleteFileTest::GuardFinish");
   }
 };

db/external_sst_file_ingestion_job.cc

@@ -132,11 +132,15 @@ Status ExternalSstFileIngestionJob::Prepare(
   return status;
 }
 
-Status ExternalSstFileIngestionJob::NeedsFlush(bool* flush_needed) {
-  SuperVersion* super_version = cfd_->GetSuperVersion();
+Status ExternalSstFileIngestionJob::NeedsFlush(bool* flush_needed,
+                                               SuperVersion* super_version) {
+  autovector<Range> ranges;
+  for (const IngestedFileInfo& file_to_ingest : files_to_ingest_) {
+    ranges.emplace_back(file_to_ingest.smallest_user_key,
+                        file_to_ingest.largest_user_key);
+  }
   Status status =
-      IngestedFilesOverlapWithMemtables(super_version, flush_needed);
-
+      cfd_->RangesOverlapWithMemtables(ranges, super_version, flush_needed);
   if (status.ok() && *flush_needed &&
       !ingestion_options_.allow_blocking_flush) {
     status = Status::InvalidArgument("External file requires flush");
@@ -148,11 +152,12 @@ Status ExternalSstFileIngestionJob::NeedsFlush(bool* flush_needed) {
 // nonmem_write_thread_
 Status ExternalSstFileIngestionJob::Run() {
   Status status;
+  SuperVersion* super_version = cfd_->GetSuperVersion();
 #ifndef NDEBUG
   // We should never run the job with a memtable that is overlapping
   // with the files we are ingesting
   bool need_flush = false;
-  status = NeedsFlush(&need_flush);
+  status = NeedsFlush(&need_flush, super_version);
   assert(status.ok() && need_flush == false);
 #endif
 
@@ -167,7 +172,6 @@ Status ExternalSstFileIngestionJob::Run() {
   // It is safe to use this instead of LastAllocatedSequence since we are
   // the only active writer, and hence they are equal
   const SequenceNumber last_seqno = versions_->LastSequence();
-  SuperVersion* super_version = cfd_->GetSuperVersion();
   edit_.SetColumnFamily(cfd_->GetID());
   // The levels that the files will be ingested into
 
@@ -375,54 +379,6 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
   return status;
 }
 
-Status ExternalSstFileIngestionJob::IngestedFilesOverlapWithMemtables(
-    SuperVersion* sv, bool* overlap) {
-  *overlap = false;
-  // Create an InternalIterator over all memtables
-  Arena arena;
-  ReadOptions ro;
-  ro.total_order_seek = true;
-  MergeIteratorBuilder merge_iter_builder(&cfd_->internal_comparator(), &arena);
-  merge_iter_builder.AddIterator(sv->mem->NewIterator(ro, &arena));
-  sv->imm->AddIterators(ro, &merge_iter_builder);
-  ScopedArenaIterator memtable_iter(merge_iter_builder.Finish());
-
-  std::vector<InternalIterator*> memtable_range_del_iters;
-  auto* active_range_del_iter = sv->mem->NewRangeTombstoneIterator(ro);
-  if (active_range_del_iter != nullptr) {
-    memtable_range_del_iters.push_back(active_range_del_iter);
-  }
-  sv->imm->AddRangeTombstoneIterators(ro, &memtable_range_del_iters);
-  RangeDelAggregator range_del_agg(cfd_->internal_comparator(),
-                                   {} /* snapshots */,
-                                   false /* collapse_deletions */);
-  Status status;
-  {
-    std::unique_ptr<InternalIterator> memtable_range_del_iter(
-        NewMergingIterator(&cfd_->internal_comparator(),
-                           memtable_range_del_iters.empty()
-                               ? nullptr
-                               : &memtable_range_del_iters[0],
-                           static_cast<int>(memtable_range_del_iters.size())));
-    status = range_del_agg.AddTombstones(std::move(memtable_range_del_iter));
-  }
-  if (status.ok()) {
-    for (IngestedFileInfo& f : files_to_ingest_) {
-      status = IngestedFileOverlapWithIteratorRange(&f, memtable_iter.get(),
-                                                    overlap);
-      if (!status.ok() || *overlap == true) {
-        break;
-      }
-      if (range_del_agg.IsRangeOverlapped(f.smallest_user_key,
-                                          f.largest_user_key)) {
-        *overlap = true;
-        break;
-      }
-    }
-  }
-  return status;
-}
-
 Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(
     SuperVersion* sv, bool force_global_seqno, CompactionStyle compaction_style,
     IngestedFileInfo* file_to_ingest, SequenceNumber* assigned_seqno) {
@@ -473,7 +429,10 @@ Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(
                             return f1->largest_seqno < f2->largest_seqno;
                           }))
                 ->largest_seqno;
-        if (level_largest_seqno != 0) {
+        // should only assign seqno to current level's largest seqno when
+        // the file fits
+        if (level_largest_seqno != 0 &&
+            IngestedFileFitInLevel(file_to_ingest, lvl)) {
           *assigned_seqno = level_largest_seqno;
         } else {
           continue;

db/external_sst_file_ingestion_job.h

@@ -84,8 +84,12 @@ class ExternalSstFileIngestionJob {
   // Check if we need to flush the memtable before running the ingestion job
   // This will be true if the files we are ingesting are overlapping with any
   // key range in the memtable.
-  // REQUIRES: Mutex held
-  Status NeedsFlush(bool* flush_needed);
+  //
+  // @param super_version A referenced SuperVersion that will be held for the
+  //    duration of this function.
+  //
+  // Thread-safe
+  Status NeedsFlush(bool* flush_needed, SuperVersion* super_version);
 
   // Will execute the ingestion job and prepare edit() to be applied.
   // REQUIRES: Mutex held
@@ -110,10 +114,6 @@ class ExternalSstFileIngestionJob {
   Status GetIngestedFileInfo(const std::string& external_file,
                              IngestedFileInfo* file_to_ingest);
 
-  // Check if the files we are ingesting overlap with any memtable.
-  // REQUIRES: Mutex held
-  Status IngestedFilesOverlapWithMemtables(SuperVersion* sv, bool* overlap);
-
   // Assign `file_to_ingest` the appropriate sequence number and  the lowest
   // possible level that it can be ingested to according to compaction_style.
   // REQUIRES: Mutex held

db/external_sst_file_test.cc

@@ -388,15 +388,17 @@ class SstFileWriterCollector : public TablePropertiesCollector {
   const char* Name() const override { return name_.c_str(); }
 
   Status Finish(UserCollectedProperties* properties) override {
+    std::string count = std::to_string(count_);
     *properties = UserCollectedProperties{
         {prefix_ + "_SstFileWriterCollector", "YES"},
-        {prefix_ + "_Count", std::to_string(count_)},
+        {prefix_ + "_Count", count},
     };
     return Status::OK();
   }
 
-  Status AddUserKey(const Slice& user_key, const Slice& value, EntryType type,
-                    SequenceNumber seq, uint64_t file_size) override {
+  Status AddUserKey(const Slice& /*user_key*/, const Slice& /*value*/,
+                    EntryType /*type*/, SequenceNumber /*seq*/,
+                    uint64_t /*file_size*/) override {
     ++count_;
     return Status::OK();
   }
@@ -416,7 +418,7 @@ class SstFileWriterCollectorFactory : public TablePropertiesCollectorFactory {
   explicit SstFileWriterCollectorFactory(std::string prefix)
       : prefix_(prefix), num_created_(0) {}
   virtual TablePropertiesCollector* CreateTablePropertiesCollector(
-      TablePropertiesCollectorFactory::Context context) override {
+      TablePropertiesCollectorFactory::Context /*context*/) override {
     num_created_++;
     return new SstFileWriterCollector(prefix_);
   }
@@ -1797,7 +1799,7 @@ TEST_F(ExternalSSTFileTest, FileWithCFInfo) {
 
 class TestIngestExternalFileListener : public EventListener {
  public:
-  void OnExternalFileIngested(DB* db,
+  void OnExternalFileIngested(DB* /*db*/,
                               const ExternalFileIngestionInfo& info) override {
     ingested_files.push_back(info);
   }

db/fault_injection_test.cc

@@ -228,16 +228,9 @@ class FaultInjectionTest : public testing::Test,
     return Status::OK();
   }
 
-#ifdef ROCKSDB_UBSAN_RUN
-#if defined(__clang__)
-__attribute__((__no_sanitize__("shift"), no_sanitize("signed-integer-overflow")))
-#elif defined(__GNUC__)
-__attribute__((__no_sanitize_undefined__))
-#endif
-#endif
   // Return the ith key
   Slice Key(int i, std::string* storage) const {
-    int num = i;
+    unsigned long long num = i;
     if (!sequential_order_) {
       // random transfer
       const int m = 0x5bd1e995;
@@ -245,7 +238,7 @@ __attribute__((__no_sanitize_undefined__))
       num ^= num << 24;
     }
     char buf[100];
-    snprintf(buf, sizeof(buf), "%016d", num);
+    snprintf(buf, sizeof(buf), "%016d", static_cast<int>(num));
     storage->assign(buf, strlen(buf));
     return Slice(*storage);
   }

db/file_indexer_test.cc

@@ -36,10 +36,10 @@ class IntComparator : public Comparator {
 
   const char* Name() const override { return "IntComparator"; }
 
-  void FindShortestSeparator(std::string* start,
-                             const Slice& limit) const override {}
+  void FindShortestSeparator(std::string* /*start*/,
+                             const Slice& /*limit*/) const override {}
 
-  void FindShortSuccessor(std::string* key) const override {}
+  void FindShortSuccessor(std::string* /*key*/) const override {}
 };
 
 class FileIndexerTest : public testing::Test {

db/flush_job.cc

@@ -54,6 +54,32 @@
 
 namespace rocksdb {
 
+const char* GetFlushReasonString (FlushReason flush_reason) {
+  switch (flush_reason) {
+    case FlushReason::kUnknown:
+      return "Unknown";
+    case FlushReason::kGetLiveFiles:
+      return "Get Live Files";
+    case FlushReason::kShutDown:
+      return "Shut down";
+    case FlushReason::kExternalFileIngestion:
+      return "External File Ingestion";
+    case FlushReason::kManualCompaction:
+      return "Manual Compaction";
+    case FlushReason::kWriteBufferManager:
+      return "Write Buffer Manager";
+    case FlushReason::kWriteBufferFull:
+      return "Write Buffer Full";
+    case FlushReason::kTest:
+      return "Test";
+    case FlushReason::kSuperVersionChange:
+      return "SuperVersion Change";
+    default:
+      return "Invalid";
+  }
+}
+
+
 FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
                    const ImmutableDBOptions& db_options,
                    const MutableCFOptions& mutable_cf_options,
@@ -278,12 +304,13 @@ Status FlushJob::WriteLevel0Table() {
       total_memory_usage += m->ApproximateMemoryUsage();
     }
 
-    event_logger_->Log() << "job" << job_context_->job_id << "event"
-                         << "flush_started"
-                         << "num_memtables" << mems_.size() << "num_entries"
-                         << total_num_entries << "num_deletes"
-                         << total_num_deletes << "memory_usage"
-                         << total_memory_usage;
+    event_logger_->Log()
+        << "job" << job_context_->job_id << "event"
+        << "flush_started"
+        << "num_memtables" << mems_.size() << "num_entries" << total_num_entries
+        << "num_deletes" << total_num_deletes << "memory_usage"
+        << total_memory_usage << "flush_reason"
+        << GetFlushReasonString(cfd_->GetFlushReason());
 
     {
       ScopedArenaIterator iter(

db/forward_iterator.cc

@@ -25,15 +25,15 @@
 namespace rocksdb {
 
 // Usage:
-//     LevelIterator iter;
+//     ForwardLevelIterator iter;
 //     iter.SetFileIndex(file_index);
 //     iter.Seek(target);
 //     iter.Next()
-class LevelIterator : public InternalIterator {
+class ForwardLevelIterator : public InternalIterator {
  public:
-  LevelIterator(const ColumnFamilyData* const cfd,
-                const ReadOptions& read_options,
-                const std::vector<FileMetaData*>& files)
+  ForwardLevelIterator(const ColumnFamilyData* const cfd,
+                       const ReadOptions& read_options,
+                       const std::vector<FileMetaData*>& files)
       : cfd_(cfd),
         read_options_(read_options),
         files_(files),
@@ -42,7 +42,7 @@ class LevelIterator : public InternalIterator {
         file_iter_(nullptr),
         pinned_iters_mgr_(nullptr) {}
 
-  ~LevelIterator() {
+  ~ForwardLevelIterator() {
     // Reset current pointer
     if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
       pinned_iters_mgr_->PinIterator(file_iter_);
@@ -84,11 +84,11 @@ class LevelIterator : public InternalIterator {
     }
   }
   void SeekToLast() override {
-    status_ = Status::NotSupported("LevelIterator::SeekToLast()");
+    status_ = Status::NotSupported("ForwardLevelIterator::SeekToLast()");
     valid_ = false;
   }
   void Prev() override {
-    status_ = Status::NotSupported("LevelIterator::Prev()");
+    status_ = Status::NotSupported("ForwardLevelIterator::Prev()");
     valid_ = false;
   }
   bool Valid() const override {
@@ -104,8 +104,8 @@ class LevelIterator : public InternalIterator {
     file_iter_->Seek(internal_key);
     valid_ = file_iter_->Valid();
   }
-  void SeekForPrev(const Slice& internal_key) override {
-    status_ = Status::NotSupported("LevelIterator::SeekForPrev()");
+  void SeekForPrev(const Slice& /*internal_key*/) override {
+    status_ = Status::NotSupported("ForwardLevelIterator::SeekForPrev()");
     valid_ = false;
   }
   void Next() override {
@@ -196,38 +196,60 @@ 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()) {
+void ForwardIterator::SVCleanup(DBImpl* db, SuperVersion* sv,
+                                bool background_purge_on_iterator_cleanup) {
+  if (sv->Unref()) {
     // Job id == 0 means that this is not our background process, but rather
     // user thread
     JobContext job_context(0);
-    db_->mutex_.Lock();
-    sv_->Cleanup();
-    db_->FindObsoleteFiles(&job_context, false, true);
-    if (read_options_.background_purge_on_iterator_cleanup) {
-      db_->ScheduleBgLogWriterClose(&job_context);
-    }
-    db_->mutex_.Unlock();
-    if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
-      pinned_iters_mgr_->PinPtr(sv_, &ReleaseSuperVersionFunc);
-    } else {
-      delete sv_;
+    db->mutex_.Lock();
+    sv->Cleanup();
+    db->FindObsoleteFiles(&job_context, false, true);
+    if (background_purge_on_iterator_cleanup) {
+      db->ScheduleBgLogWriterClose(&job_context);
     }
+    db->mutex_.Unlock();
+    delete sv;
     if (job_context.HaveSomethingToDelete()) {
-      db_->PurgeObsoleteFiles(
-          job_context, read_options_.background_purge_on_iterator_cleanup);
+      db->PurgeObsoleteFiles(job_context, background_purge_on_iterator_cleanup);
     }
     job_context.Clean();
   }
 }
 
+namespace {
+struct SVCleanupParams {
+  DBImpl* db;
+  SuperVersion* sv;
+  bool background_purge_on_iterator_cleanup;
+};
+}
+
+// Used in PinnedIteratorsManager to release pinned SuperVersion
+void ForwardIterator::DeferredSVCleanup(void* arg) {
+  auto d = reinterpret_cast<SVCleanupParams*>(arg);
+  ForwardIterator::SVCleanup(
+    d->db, d->sv, d->background_purge_on_iterator_cleanup);
+  delete d;
+}
+
+void ForwardIterator::SVCleanup() {
+  if (sv_ == nullptr) {
+    return;
+  }
+  if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
+    // pinned_iters_mgr_ tells us to make sure that all visited key-value slices
+    // are alive until pinned_iters_mgr_->ReleasePinnedData() is called.
+    // The slices may point into some memtables owned by sv_, so we need to keep
+    // sv_ referenced until pinned_iters_mgr_ unpins everything.
+    auto p = new SVCleanupParams{
+      db_, sv_, read_options_.background_purge_on_iterator_cleanup};
+    pinned_iters_mgr_->PinPtr(p, &ForwardIterator::DeferredSVCleanup);
+  } else {
+    SVCleanup(db_, sv_, read_options_.background_purge_on_iterator_cleanup);
+  }
+}
+
 void ForwardIterator::Cleanup(bool release_sv) {
   if (mutable_iter_ != nullptr) {
     DeleteIterator(mutable_iter_, true /* is_arena */);
@@ -361,14 +383,13 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,
 
       if (!l0_iters_[i]->status().ok()) {
         immutable_status_ = l0_iters_[i]->status();
-      } else if (l0_iters_[i]->Valid()) {
-        if (!IsOverUpperBound(l0_iters_[i]->key())) {
-          immutable_min_heap_.push(l0_iters_[i]);
-        } else {
-          has_iter_trimmed_for_upper_bound_ = true;
-          DeleteIterator(l0_iters_[i]);
-          l0_iters_[i] = nullptr;
-        }
+      } else if (l0_iters_[i]->Valid() &&
+                 !IsOverUpperBound(l0_iters_[i]->key())) {
+        immutable_min_heap_.push(l0_iters_[i]);
+      } else {
+        has_iter_trimmed_for_upper_bound_ = true;
+        DeleteIterator(l0_iters_[i]);
+        l0_iters_[i] = nullptr;
       }
     }
 
@@ -395,15 +416,14 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,
 
         if (!level_iters_[level - 1]->status().ok()) {
           immutable_status_ = level_iters_[level - 1]->status();
-        } else if (level_iters_[level - 1]->Valid()) {
-          if (!IsOverUpperBound(level_iters_[level - 1]->key())) {
-            immutable_min_heap_.push(level_iters_[level - 1]);
-          } else {
-            // Nothing in this level is interesting. Remove.
-            has_iter_trimmed_for_upper_bound_ = true;
-            DeleteIterator(level_iters_[level - 1]);
-            level_iters_[level - 1] = nullptr;
-          }
+        } else if (level_iters_[level - 1]->Valid() &&
+                   !IsOverUpperBound(level_iters_[level - 1]->key())) {
+          immutable_min_heap_.push(level_iters_[level - 1]);
+        } else {
+          // Nothing in this level is interesting. Remove.
+          has_iter_trimmed_for_upper_bound_ = true;
+          DeleteIterator(level_iters_[level - 1]);
+          level_iters_[level - 1] = nullptr;
         }
       }
     }
@@ -541,7 +561,7 @@ void ForwardIterator::UpdateChildrenPinnedItersMgr() {
   }
 
   // Set PinnedIteratorsManager for L1+ levels iterators.
-  for (LevelIterator* child_iter : level_iters_) {
+  for (ForwardLevelIterator* child_iter : level_iters_) {
     if (child_iter) {
       child_iter->SetPinnedItersMgr(pinned_iters_mgr_);
     }
@@ -703,7 +723,7 @@ void ForwardIterator::BuildLevelIterators(const VersionStorageInfo* vstorage) {
       }
     } else {
       level_iters_.push_back(
-          new LevelIterator(cfd_, read_options_, level_files));
+          new ForwardLevelIterator(cfd_, read_options_, level_files));
     }
   }
 }

db/forward_iterator.h

@@ -23,7 +23,7 @@ class DBImpl;
 class Env;
 struct SuperVersion;
 class ColumnFamilyData;
-class LevelIterator;
+class ForwardLevelIterator;
 class VersionStorageInfo;
 struct FileMetaData;
 
@@ -55,7 +55,7 @@ class ForwardIterator : public InternalIterator {
                   ColumnFamilyData* cfd, SuperVersion* current_sv = nullptr);
   virtual ~ForwardIterator();
 
-  void SeekForPrev(const Slice& target) override {
+  void SeekForPrev(const Slice& /*target*/) override {
     status_ = Status::NotSupported("ForwardIterator::SeekForPrev()");
     valid_ = false;
   }
@@ -85,7 +85,14 @@ class ForwardIterator : public InternalIterator {
 
  private:
   void Cleanup(bool release_sv);
+  // Unreference and, if needed, clean up the current SuperVersion. This is
+  // either done immediately or deferred until this iterator is unpinned by
+  // PinnedIteratorsManager.
   void SVCleanup();
+  static void SVCleanup(
+    DBImpl* db, SuperVersion* sv, bool background_purge_on_iterator_cleanup);
+  static void DeferredSVCleanup(void* arg);
+
   void RebuildIterators(bool refresh_sv);
   void RenewIterators();
   void BuildLevelIterators(const VersionStorageInfo* vstorage);
@@ -119,7 +126,7 @@ class ForwardIterator : public InternalIterator {
   InternalIterator* mutable_iter_;
   std::vector<InternalIterator*> imm_iters_;
   std::vector<InternalIterator*> l0_iters_;
-  std::vector<LevelIterator*> level_iters_;
+  std::vector<ForwardLevelIterator*> level_iters_;
   InternalIterator* current_;
   bool valid_;
 

db/internal_stats.cc

@@ -233,6 +233,7 @@ static const std::string estimate_live_data_size = "estimate-live-data-size";
 static const std::string min_log_number_to_keep = "min-log-number-to-keep";
 static const std::string base_level = "base-level";
 static const std::string total_sst_files_size = "total-sst-files-size";
+static const std::string live_sst_files_size = "live-sst-files-size";
 static const std::string estimate_pending_comp_bytes =
     "estimate-pending-compaction-bytes";
 static const std::string aggregated_table_properties =
@@ -307,6 +308,8 @@ const std::string DB::Properties::kMinLogNumberToKeep =
     rocksdb_prefix + min_log_number_to_keep;
 const std::string DB::Properties::kTotalSstFilesSize =
                       rocksdb_prefix + total_sst_files_size;
+const std::string DB::Properties::kLiveSstFilesSize =
+    rocksdb_prefix + live_sst_files_size;
 const std::string DB::Properties::kBaseLevel = rocksdb_prefix + base_level;
 const std::string DB::Properties::kEstimatePendingCompactionBytes =
     rocksdb_prefix + estimate_pending_comp_bytes;
@@ -405,6 +408,8 @@ const std::unordered_map<std::string, DBPropertyInfo>
          {false, nullptr, &InternalStats::HandleBaseLevel, nullptr}},
         {DB::Properties::kTotalSstFilesSize,
          {false, nullptr, &InternalStats::HandleTotalSstFilesSize, nullptr}},
+        {DB::Properties::kLiveSstFilesSize,
+         {false, nullptr, &InternalStats::HandleLiveSstFilesSize, nullptr}},
         {DB::Properties::kEstimatePendingCompactionBytes,
          {false, nullptr, &InternalStats::HandleEstimatePendingCompactionBytes,
           nullptr}},
@@ -442,7 +447,7 @@ bool InternalStats::GetStringProperty(const DBPropertyInfo& property_info,
 }
 
 bool InternalStats::GetMapProperty(const DBPropertyInfo& property_info,
-                                   const Slice& property,
+                                   const Slice& /*property*/,
                                    std::map<std::string, std::string>* value) {
   assert(value != nullptr);
   assert(property_info.handle_map != nullptr);
@@ -494,7 +499,7 @@ bool InternalStats::HandleCompressionRatioAtLevelPrefix(std::string* value,
   return true;
 }
 
-bool InternalStats::HandleLevelStats(std::string* value, Slice suffix) {
+bool InternalStats::HandleLevelStats(std::string* value, Slice /*suffix*/) {
   char buf[1000];
   const auto* vstorage = cfd_->current()->storage_info();
   snprintf(buf, sizeof(buf),
@@ -527,35 +532,36 @@ bool InternalStats::HandleCFMapStats(
   return true;
 }
 
-bool InternalStats::HandleCFStats(std::string* value, Slice suffix) {
+bool InternalStats::HandleCFStats(std::string* value, Slice /*suffix*/) {
   DumpCFStats(value);
   return true;
 }
 
 bool InternalStats::HandleCFStatsNoFileHistogram(std::string* value,
-                                                 Slice suffix) {
+                                                 Slice /*suffix*/) {
   DumpCFStatsNoFileHistogram(value);
   return true;
 }
 
-bool InternalStats::HandleCFFileHistogram(std::string* value, Slice suffix) {
+bool InternalStats::HandleCFFileHistogram(std::string* value,
+                                          Slice /*suffix*/) {
   DumpCFFileHistogram(value);
   return true;
 }
 
-bool InternalStats::HandleDBStats(std::string* value, Slice suffix) {
+bool InternalStats::HandleDBStats(std::string* value, Slice /*suffix*/) {
   DumpDBStats(value);
   return true;
 }
 
-bool InternalStats::HandleSsTables(std::string* value, Slice suffix) {
+bool InternalStats::HandleSsTables(std::string* value, Slice /*suffix*/) {
   auto* current = cfd_->current();
   *value = current->DebugString(true, true);
   return true;
 }
 
 bool InternalStats::HandleAggregatedTableProperties(std::string* value,
-                                                    Slice suffix) {
+                                                    Slice /*suffix*/) {
   std::shared_ptr<const TableProperties> tp;
   auto s = cfd_->current()->GetAggregatedTableProperties(&tp);
   if (!s.ok()) {
@@ -582,34 +588,34 @@ bool InternalStats::HandleAggregatedTablePropertiesAtLevel(std::string* value,
   return true;
 }
 
-bool InternalStats::HandleNumImmutableMemTable(uint64_t* value, DBImpl* db,
-                                               Version* version) {
+bool InternalStats::HandleNumImmutableMemTable(uint64_t* value, DBImpl* /*db*/,
+                                               Version* /*version*/) {
   *value = cfd_->imm()->NumNotFlushed();
   return true;
 }
 
 bool InternalStats::HandleNumImmutableMemTableFlushed(uint64_t* value,
-                                                      DBImpl* db,
-                                                      Version* version) {
+                                                      DBImpl* /*db*/,
+                                                      Version* /*version*/) {
   *value = cfd_->imm()->NumFlushed();
   return true;
 }
 
-bool InternalStats::HandleMemTableFlushPending(uint64_t* value, DBImpl* db,
-                                               Version* version) {
+bool InternalStats::HandleMemTableFlushPending(uint64_t* value, DBImpl* /*db*/,
+                                               Version* /*version*/) {
   // Return number of mem tables that are ready to flush (made immutable)
   *value = (cfd_->imm()->IsFlushPending() ? 1 : 0);
   return true;
 }
 
 bool InternalStats::HandleNumRunningFlushes(uint64_t* value, DBImpl* db,
-                                            Version* version) {
+                                            Version* /*version*/) {
   *value = db->num_running_flushes();
   return true;
 }
 
-bool InternalStats::HandleCompactionPending(uint64_t* value, DBImpl* db,
-                                            Version* version) {
+bool InternalStats::HandleCompactionPending(uint64_t* value, DBImpl* /*db*/,
+                                            Version* /*version*/) {
   // 1 if the system already determines at least one compaction is needed.
   // 0 otherwise,
   const auto* vstorage = cfd_->current()->storage_info();
@@ -618,70 +624,74 @@ bool InternalStats::HandleCompactionPending(uint64_t* value, DBImpl* db,
 }
 
 bool InternalStats::HandleNumRunningCompactions(uint64_t* value, DBImpl* db,
-                                                Version* version) {
+                                                Version* /*version*/) {
   *value = db->num_running_compactions_;
   return true;
 }
 
-bool InternalStats::HandleBackgroundErrors(uint64_t* value, DBImpl* db,
-                                           Version* version) {
+bool InternalStats::HandleBackgroundErrors(uint64_t* value, DBImpl* /*db*/,
+                                           Version* /*version*/) {
   // Accumulated number of  errors in background flushes or compactions.
   *value = GetBackgroundErrorCount();
   return true;
 }
 
-bool InternalStats::HandleCurSizeActiveMemTable(uint64_t* value, DBImpl* db,
-                                                Version* version) {
+bool InternalStats::HandleCurSizeActiveMemTable(uint64_t* value, DBImpl* /*db*/,
+                                                Version* /*version*/) {
   // Current size of the active memtable
   *value = cfd_->mem()->ApproximateMemoryUsage();
   return true;
 }
 
-bool InternalStats::HandleCurSizeAllMemTables(uint64_t* value, DBImpl* db,
-                                              Version* version) {
+bool InternalStats::HandleCurSizeAllMemTables(uint64_t* value, DBImpl* /*db*/,
+                                              Version* /*version*/) {
   // Current size of the active memtable + immutable memtables
   *value = cfd_->mem()->ApproximateMemoryUsage() +
            cfd_->imm()->ApproximateUnflushedMemTablesMemoryUsage();
   return true;
 }
 
-bool InternalStats::HandleSizeAllMemTables(uint64_t* value, DBImpl* db,
-                                           Version* version) {
+bool InternalStats::HandleSizeAllMemTables(uint64_t* value, DBImpl* /*db*/,
+                                           Version* /*version*/) {
   *value = cfd_->mem()->ApproximateMemoryUsage() +
            cfd_->imm()->ApproximateMemoryUsage();
   return true;
 }
 
-bool InternalStats::HandleNumEntriesActiveMemTable(uint64_t* value, DBImpl* db,
-                                                   Version* version) {
+bool InternalStats::HandleNumEntriesActiveMemTable(uint64_t* value,
+                                                   DBImpl* /*db*/,
+                                                   Version* /*version*/) {
   // Current number of entires in the active memtable
   *value = cfd_->mem()->num_entries();
   return true;
 }
 
-bool InternalStats::HandleNumEntriesImmMemTables(uint64_t* value, DBImpl* db,
-                                                 Version* version) {
+bool InternalStats::HandleNumEntriesImmMemTables(uint64_t* value,
+                                                 DBImpl* /*db*/,
+                                                 Version* /*version*/) {
   // Current number of entries in the immutable memtables
   *value = cfd_->imm()->current()->GetTotalNumEntries();
   return true;
 }
 
-bool InternalStats::HandleNumDeletesActiveMemTable(uint64_t* value, DBImpl* db,
-                                                   Version* version) {
+bool InternalStats::HandleNumDeletesActiveMemTable(uint64_t* value,
+                                                   DBImpl* /*db*/,
+                                                   Version* /*version*/) {
   // Current number of entires in the active memtable
   *value = cfd_->mem()->num_deletes();
   return true;
 }
 
-bool InternalStats::HandleNumDeletesImmMemTables(uint64_t* value, DBImpl* db,
-                                                 Version* version) {
+bool InternalStats::HandleNumDeletesImmMemTables(uint64_t* value,
+                                                 DBImpl* /*db*/,
+                                                 Version* /*version*/) {
   // Current number of entries in the immutable memtables
   *value = cfd_->imm()->current()->GetTotalNumDeletes();
   return true;
 }
 
-bool InternalStats::HandleEstimateNumKeys(uint64_t* value, DBImpl* db,
-                                          Version* version) {
+bool InternalStats::HandleEstimateNumKeys(uint64_t* value, DBImpl* /*db*/,
+                                          Version* /*version*/) {
   // Estimate number of entries in the column family:
   // Use estimated entries in tables + total entries in memtables.
   const auto* vstorage = cfd_->current()->storage_info();
@@ -697,77 +707,85 @@ bool InternalStats::HandleEstimateNumKeys(uint64_t* value, DBImpl* db,
 }
 
 bool InternalStats::HandleNumSnapshots(uint64_t* value, DBImpl* db,
-                                       Version* version) {
+                                       Version* /*version*/) {
   *value = db->snapshots().count();
   return true;
 }
 
 bool InternalStats::HandleOldestSnapshotTime(uint64_t* value, DBImpl* db,
-                                             Version* version) {
+                                             Version* /*version*/) {
   *value = static_cast<uint64_t>(db->snapshots().GetOldestSnapshotTime());
   return true;
 }
 
-bool InternalStats::HandleNumLiveVersions(uint64_t* value, DBImpl* db,
-                                          Version* version) {
+bool InternalStats::HandleNumLiveVersions(uint64_t* value, DBImpl* /*db*/,
+                                          Version* /*version*/) {
   *value = cfd_->GetNumLiveVersions();
   return true;
 }
 
-bool InternalStats::HandleCurrentSuperVersionNumber(uint64_t* value, DBImpl* db,
-                                                    Version* version) {
+bool InternalStats::HandleCurrentSuperVersionNumber(uint64_t* value,
+                                                    DBImpl* /*db*/,
+                                                    Version* /*version*/) {
   *value = cfd_->GetSuperVersionNumber();
   return true;
 }
 
 bool InternalStats::HandleIsFileDeletionsEnabled(uint64_t* value, DBImpl* db,
-                                                 Version* version) {
+                                                 Version* /*version*/) {
   *value = db->IsFileDeletionsEnabled();
   return true;
 }
 
-bool InternalStats::HandleBaseLevel(uint64_t* value, DBImpl* db,
-                                    Version* version) {
+bool InternalStats::HandleBaseLevel(uint64_t* value, DBImpl* /*db*/,
+                                    Version* /*version*/) {
   const auto* vstorage = cfd_->current()->storage_info();
   *value = vstorage->base_level();
   return true;
 }
 
-bool InternalStats::HandleTotalSstFilesSize(uint64_t* value, DBImpl* db,
-                                            Version* version) {
+bool InternalStats::HandleTotalSstFilesSize(uint64_t* value, DBImpl* /*db*/,
+                                            Version* /*version*/) {
   *value = cfd_->GetTotalSstFilesSize();
   return true;
 }
 
+bool InternalStats::HandleLiveSstFilesSize(uint64_t* value, DBImpl* /*db*/,
+                                           Version* /*version*/) {
+  *value = cfd_->GetLiveSstFilesSize();
+  return true;
+}
+
 bool InternalStats::HandleEstimatePendingCompactionBytes(uint64_t* value,
-                                                         DBImpl* db,
-                                                         Version* version) {
+                                                         DBImpl* /*db*/,
+                                                         Version* /*version*/) {
   const auto* vstorage = cfd_->current()->storage_info();
   *value = vstorage->estimated_compaction_needed_bytes();
   return true;
 }
 
-bool InternalStats::HandleEstimateTableReadersMem(uint64_t* value, DBImpl* db,
+bool InternalStats::HandleEstimateTableReadersMem(uint64_t* value,
+                                                  DBImpl* /*db*/,
                                                   Version* version) {
   *value = (version == nullptr) ? 0 : version->GetMemoryUsageByTableReaders();
   return true;
 }
 
-bool InternalStats::HandleEstimateLiveDataSize(uint64_t* value, DBImpl* db,
-                                               Version* version) {
+bool InternalStats::HandleEstimateLiveDataSize(uint64_t* value, DBImpl* /*db*/,
+                                               Version* /*version*/) {
   const auto* vstorage = cfd_->current()->storage_info();
   *value = vstorage->EstimateLiveDataSize();
   return true;
 }
 
 bool InternalStats::HandleMinLogNumberToKeep(uint64_t* value, DBImpl* db,
-                                             Version* version) {
+                                             Version* /*version*/) {
   *value = db->MinLogNumberToKeep();
   return true;
 }
 
 bool InternalStats::HandleActualDelayedWriteRate(uint64_t* value, DBImpl* db,
-                                                 Version* version) {
+                                                 Version* /*version*/) {
   const WriteController& wc = db->write_controller();
   if (!wc.NeedsDelay()) {
     *value = 0;
@@ -778,7 +796,7 @@ bool InternalStats::HandleActualDelayedWriteRate(uint64_t* value, DBImpl* db,
 }
 
 bool InternalStats::HandleIsWriteStopped(uint64_t* value, DBImpl* db,
-                                         Version* version) {
+                                         Version* /*version*/) {
   *value = db->write_controller().IsStopped() ? 1 : 0;
   return true;
 }

db/internal_stats.h

@@ -467,6 +467,7 @@ class InternalStats {
                                     Version* version);
   bool HandleBaseLevel(uint64_t* value, DBImpl* db, Version* version);
   bool HandleTotalSstFilesSize(uint64_t* value, DBImpl* db, Version* version);
+  bool HandleLiveSstFilesSize(uint64_t* value, DBImpl* db, Version* version);
   bool HandleEstimatePendingCompactionBytes(uint64_t* value, DBImpl* db,
                                             Version* version);
   bool HandleEstimateTableReadersMem(uint64_t* value, DBImpl* db,

db/listener_test.cc

@@ -46,11 +46,11 @@ class EventListenerTest : public DBTestBase {
 };
 
 struct TestPropertiesCollector : public rocksdb::TablePropertiesCollector {
-  virtual rocksdb::Status AddUserKey(const rocksdb::Slice& key,
-                                     const rocksdb::Slice& value,
-                                     rocksdb::EntryType type,
-                                     rocksdb::SequenceNumber seq,
-                                     uint64_t file_size) override {
+  virtual rocksdb::Status AddUserKey(const rocksdb::Slice& /*key*/,
+                                     const rocksdb::Slice& /*value*/,
+                                     rocksdb::EntryType /*type*/,
+                                     rocksdb::SequenceNumber /*seq*/,
+                                     uint64_t /*file_size*/) override {
     return Status::OK();
   }
   virtual rocksdb::Status Finish(
@@ -73,7 +73,7 @@ struct TestPropertiesCollector : public rocksdb::TablePropertiesCollector {
 class TestPropertiesCollectorFactory : public TablePropertiesCollectorFactory {
  public:
   virtual TablePropertiesCollector* CreateTablePropertiesCollector(
-      TablePropertiesCollectorFactory::Context context) override {
+      TablePropertiesCollectorFactory::Context /*context*/) override {
     return new TestPropertiesCollector;
   }
   const char* Name() const override { return "TestTablePropertiesCollector"; }
@@ -425,7 +425,7 @@ TEST_F(EventListenerTest, DisableBGCompaction) {
 
 class TestCompactionReasonListener : public EventListener {
  public:
-  void OnCompactionCompleted(DB* db, const CompactionJobInfo& ci) override {
+  void OnCompactionCompleted(DB* /*db*/, const CompactionJobInfo& ci) override {
     std::lock_guard<std::mutex> lock(mutex_);
     compaction_reasons_.push_back(ci.compaction_reason);
   }
@@ -528,7 +528,7 @@ TEST_F(EventListenerTest, CompactionReasonUniversal) {
 
   ASSERT_GT(listener->compaction_reasons_.size(), 0);
   for (auto compaction_reason : listener->compaction_reasons_) {
-    ASSERT_EQ(compaction_reason, CompactionReason::kUniversalSortedRunNum);
+    ASSERT_EQ(compaction_reason, CompactionReason::kUniversalSizeRatio);
   }
 
   options.level0_file_num_compaction_trigger = 8;
@@ -807,7 +807,8 @@ class BackgroundErrorListener : public EventListener {
  public:
   BackgroundErrorListener(SpecialEnv* env) : env_(env), counter_(0) {}
 
-  void OnBackgroundError(BackgroundErrorReason reason, Status* bg_error) override {
+  void OnBackgroundError(BackgroundErrorReason /*reason*/,
+                         Status* bg_error) override {
     if (counter_ == 0) {
       // suppress the first error and disable write-dropping such that a retry
       // can succeed.

db/log_reader.h

@@ -54,6 +54,7 @@ class Reader {
   // The Reader will start reading at the first record located at physical
   // position >= initial_offset within the file.
   Reader(std::shared_ptr<Logger> info_log,
+  // @lint-ignore TXT2 T25377293 Grandfathered in
 	 unique_ptr<SequentialFileReader>&& file,
          Reporter* reporter, bool checksum, uint64_t initial_offset,
          uint64_t log_num);

db/malloc_stats.cc

@@ -41,8 +41,8 @@ static void GetJemallocStatus(void* mstat_arg, const char* status) {
 }
 #endif  // ROCKSDB_JEMALLOC
 
-void DumpMallocStats(std::string* stats) {
 #ifdef ROCKSDB_JEMALLOC
+void DumpMallocStats(std::string* stats) {
   MallocStatus mstat;
   const unsigned int kMallocStatusLen = 1000000;
   std::unique_ptr<char[]> buf{new char[kMallocStatusLen + 1]};
@@ -50,8 +50,10 @@ void DumpMallocStats(std::string* stats) {
   mstat.end = buf.get() + kMallocStatusLen;
   je_malloc_stats_print(GetJemallocStatus, &mstat, "");
   stats->append(buf.get());
-#endif  // ROCKSDB_JEMALLOC
 }
+#else
+void DumpMallocStats(std::string*) {}
+#endif  // ROCKSDB_JEMALLOC
 
 }
 #endif  // !ROCKSDB_LITE

db/manual_compaction_test.cc

@@ -46,9 +46,9 @@ class DestroyAllCompactionFilter : public CompactionFilter {
  public:
   DestroyAllCompactionFilter() {}
 
-  virtual bool Filter(int level, const Slice& key, const Slice& existing_value,
-                      std::string* new_value,
-                      bool* value_changed) const override {
+  virtual bool Filter(int /*level*/, const Slice& /*key*/,
+                      const Slice& existing_value, std::string* /*new_value*/,
+                      bool* /*value_changed*/) const override {
     return existing_value.ToString() == "destroy";
   }
 

db/memtable.cc

@@ -225,7 +225,7 @@ int MemTable::KeyComparator::operator()(const char* prefix_len_key1,
   // Internal keys are encoded as length-prefixed strings.
   Slice k1 = GetLengthPrefixedSlice(prefix_len_key1);
   Slice k2 = GetLengthPrefixedSlice(prefix_len_key2);
-  return comparator.Compare(k1, k2);
+  return comparator.CompareKeySeq(k1, k2);
 }
 
 int MemTable::KeyComparator::operator()(const char* prefix_len_key,
@@ -233,16 +233,16 @@ int MemTable::KeyComparator::operator()(const char* prefix_len_key,
     const {
   // Internal keys are encoded as length-prefixed strings.
   Slice a = GetLengthPrefixedSlice(prefix_len_key);
-  return comparator.Compare(a, key);
+  return comparator.CompareKeySeq(a, key);
 }
 
-void MemTableRep::InsertConcurrently(KeyHandle handle) {
+void MemTableRep::InsertConcurrently(KeyHandle /*handle*/) {
 #ifndef ROCKSDB_LITE
-    throw std::runtime_error("concurrent insert not supported");
+  throw std::runtime_error("concurrent insert not supported");
 #else
-    abort();
+  abort();
 #endif
-  }
+}
 
 Slice MemTableRep::UserKey(const char* key) const {
   Slice slice = GetLengthPrefixedSlice(key);
@@ -289,8 +289,7 @@ class MemTableIterator : public InternalIterator {
 #ifndef NDEBUG
     // Assert that the MemTableIterator is never deleted while
     // Pinning is Enabled.
-    assert(!pinned_iters_mgr_ ||
-           (pinned_iters_mgr_ && !pinned_iters_mgr_->PinningEnabled()));
+    assert(!pinned_iters_mgr_ || !pinned_iters_mgr_->PinningEnabled());
 #endif
     if (arena_mode_) {
       iter_->~Iterator();
@@ -444,7 +443,7 @@ MemTable::MemTableStats MemTable::ApproximateStats(const Slice& start_ikey,
   return {entry_count * (data_size / n), entry_count};
 }
 
-void MemTable::Add(SequenceNumber s, ValueType type,
+bool MemTable::Add(SequenceNumber s, ValueType type,
                    const Slice& key, /* user key */
                    const Slice& value, bool allow_concurrent,
                    MemTablePostProcessInfo* post_process_info) {
@@ -479,9 +478,15 @@ void MemTable::Add(SequenceNumber s, ValueType type,
     if (insert_with_hint_prefix_extractor_ != nullptr &&
         insert_with_hint_prefix_extractor_->InDomain(key_slice)) {
       Slice prefix = insert_with_hint_prefix_extractor_->Transform(key_slice);
-      table->InsertWithHint(handle, &insert_hints_[prefix]);
+      bool res = table->InsertKeyWithHint(handle, &insert_hints_[prefix]);
+      if (!res) {
+        return res;
+      }
     } else {
-      table->Insert(handle);
+      bool res = table->InsertKey(handle);
+      if (!res) {
+        return res;
+      }
     }
 
     // this is a bit ugly, but is the way to avoid locked instructions
@@ -514,7 +519,10 @@ void MemTable::Add(SequenceNumber s, ValueType type,
     assert(post_process_info == nullptr);
     UpdateFlushState();
   } else {
-    table->InsertConcurrently(handle);
+    bool res = table->InsertKeyConcurrently(handle);
+    if (!res) {
+      return res;
+    }
 
     assert(post_process_info != nullptr);
     post_process_info->num_entries++;
@@ -544,6 +552,7 @@ void MemTable::Add(SequenceNumber s, ValueType type,
     is_range_del_table_empty_ = false;
   }
   UpdateOldestKeyTime();
+  return true;
 }
 
 // Callback from MemTable::Get()
@@ -579,11 +588,12 @@ struct Saver {
 
 static bool SaveValue(void* arg, const char* entry) {
   Saver* s = reinterpret_cast<Saver*>(arg);
+  assert(s != nullptr);
   MergeContext* merge_context = s->merge_context;
   RangeDelAggregator* range_del_agg = s->range_del_agg;
   const MergeOperator* merge_operator = s->merge_operator;
 
-  assert(s != nullptr && merge_context != nullptr && range_del_agg != nullptr);
+  assert(merge_context != nullptr && range_del_agg != nullptr);
 
   // entry format is:
   //    klength  varint32
@@ -799,8 +809,9 @@ void MemTable::Update(SequenceNumber seq,
       // Correct user key
       const uint64_t tag = DecodeFixed64(key_ptr + key_length - 8);
       ValueType type;
-      SequenceNumber unused;
-      UnPackSequenceAndType(tag, &unused, &type);
+      SequenceNumber existing_seq;
+      UnPackSequenceAndType(tag, &existing_seq, &type);
+      assert(existing_seq != seq);
       if (type == kTypeValue) {
         Slice prev_value = GetLengthPrefixedSlice(key_ptr + key_length);
         uint32_t prev_size = static_cast<uint32_t>(prev_value.size());
@@ -823,7 +834,10 @@ void MemTable::Update(SequenceNumber seq,
   }
 
   // key doesn't exist
-  Add(seq, kTypeValue, key, value);
+  bool add_res __attribute__((__unused__));
+  add_res = Add(seq, kTypeValue, key, value);
+  // We already checked unused != seq above. In that case, Add should not fail.
+  assert(add_res);
 }
 
 bool MemTable::UpdateCallback(SequenceNumber seq,

db/memtable.h

@@ -167,7 +167,10 @@ class MemTable {
   //
   // REQUIRES: if allow_concurrent = false, external synchronization to prevent
   // simultaneous operations on the same MemTable.
-  void Add(SequenceNumber seq, ValueType type, const Slice& key,
+  //
+  // Returns false if MemTableRepFactory::CanHandleDuplicatedKey() is true and
+  // the <key, seq> already exists.
+  bool Add(SequenceNumber seq, ValueType type, const Slice& key,
            const Slice& value, bool allow_concurrent = false,
            MemTablePostProcessInfo* post_process_info = nullptr);
 

db/memtable_list.cc

@@ -154,7 +154,7 @@ bool MemTableListVersion::GetFromList(
 }
 
 Status MemTableListVersion::AddRangeTombstoneIterators(
-    const ReadOptions& read_opts, Arena* arena,
+    const ReadOptions& read_opts, Arena* /*arena*/,
     RangeDelAggregator* range_del_agg) {
   assert(range_del_agg != nullptr);
   for (auto& m : memlist_) {
@@ -300,7 +300,7 @@ void MemTableList::PickMemtablesToFlush(autovector<MemTable*>* ret) {
 }
 
 void MemTableList::RollbackMemtableFlush(const autovector<MemTable*>& mems,
-                                         uint64_t file_number) {
+                                         uint64_t /*file_number*/) {
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_MEMTABLE_ROLLBACK);
   assert(!mems.empty());

db/merge_helper.cc

@@ -5,12 +5,12 @@
 
 #include "db/merge_helper.h"
 
-#include <stdio.h>
 #include <string>
 
 #include "db/dbformat.h"
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/statistics.h"
+#include "port/likely.h"
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/merge_operator.h"
@@ -22,7 +22,8 @@ MergeHelper::MergeHelper(Env* env, const Comparator* user_comparator,
                          const MergeOperator* user_merge_operator,
                          const CompactionFilter* compaction_filter,
                          Logger* logger, bool assert_valid_internal_key,
-                         SequenceNumber latest_snapshot, int level,
+                         SequenceNumber latest_snapshot,
+                         const SnapshotChecker* snapshot_checker, int level,
                          Statistics* stats,
                          const std::atomic<bool>* shutting_down)
     : env_(env),
@@ -34,6 +35,7 @@ MergeHelper::MergeHelper(Env* env, const Comparator* user_comparator,
       assert_valid_internal_key_(assert_valid_internal_key),
       allow_single_operand_(false),
       latest_snapshot_(latest_snapshot),
+      snapshot_checker_(snapshot_checker),
       level_(level),
       keys_(),
       filter_timer_(env_),
@@ -158,7 +160,10 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
       // hit a different user key, stop right here
       hit_the_next_user_key = true;
       break;
-    } else if (stop_before && ikey.sequence <= stop_before) {
+    } else if (stop_before > 0 && ikey.sequence <= stop_before &&
+               LIKELY(snapshot_checker_ == nullptr ||
+                      snapshot_checker_->IsInSnapshot(ikey.sequence,
+                                                      stop_before))) {
       // hit an entry that's visible by the previous snapshot, can't touch that
       break;
     }

db/merge_helper.h

@@ -13,6 +13,7 @@
 #include "db/dbformat.h"
 #include "db/merge_context.h"
 #include "db/range_del_aggregator.h"
+#include "db/snapshot_checker.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/env.h"
 #include "rocksdb/slice.h"
@@ -33,7 +34,8 @@ class MergeHelper {
               const MergeOperator* user_merge_operator,
               const CompactionFilter* compaction_filter, Logger* logger,
               bool assert_valid_internal_key, SequenceNumber latest_snapshot,
-              int level = 0, Statistics* stats = nullptr,
+              const SnapshotChecker* snapshot_checker = nullptr, int level = 0,
+              Statistics* stats = nullptr,
               const std::atomic<bool>* shutting_down = nullptr);
 
   // Wrapper around MergeOperator::FullMergeV2() that records perf statistics.
@@ -145,6 +147,7 @@ class MergeHelper {
   bool assert_valid_internal_key_; // enforce no internal key corruption?
   bool allow_single_operand_;
   SequenceNumber latest_snapshot_;
+  const SnapshotChecker* const snapshot_checker_;
   int level_;
 
   // the scratch area that holds the result of MergeUntil

db/merge_test.cc

@@ -504,7 +504,7 @@ void runTest(int argc, const std::string& dbname, const bool use_ttl = false) {
 }
 }  // namespace
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* /*argv*/ []) {
   //TODO: Make this test like a general rocksdb unit-test
   rocksdb::port::InstallStackTraceHandler();
   runTest(argc, test::TmpDir() + "/merge_testdb");

db/plain_table_db_test.cc

@@ -327,7 +327,7 @@ class TestPlainTableFactory : public PlainTableFactory {
       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 {
+      bool /*prefetch_index_and_filter_in_cache*/) const override {
     TableProperties* props = nullptr;
     auto s =
         ReadTableProperties(file.get(), file_size, kPlainTableMagicNumber,

db/prefix_test.cc

@@ -126,10 +126,10 @@ class TestKeyComparator : public Comparator {
     return "TestKeyComparator";
   }
 
-  virtual void FindShortestSeparator(std::string* start,
-                                     const Slice& limit) const override {}
+  virtual void FindShortestSeparator(std::string* /*start*/,
+                                     const Slice& /*limit*/) const override {}
 
-  virtual void FindShortSuccessor(std::string* key) const override {}
+  virtual void FindShortSuccessor(std::string* /*key*/) const override {}
 };
 
 namespace {

db/snapshot_checker.h

@@ -19,8 +19,8 @@ class SnapshotChecker {
 class DisableGCSnapshotChecker : public SnapshotChecker {
  public:
   virtual ~DisableGCSnapshotChecker() {}
-  virtual bool IsInSnapshot(SequenceNumber sequence,
-                            SequenceNumber snapshot_sequence) const {
+  virtual bool IsInSnapshot(SequenceNumber /*sequence*/,
+                            SequenceNumber /*snapshot_sequence*/) const {
     // By returning false, we prevent all the values from being GCed
     return false;
   }

db/snapshot_impl.h

@@ -45,6 +45,10 @@ class SnapshotList {
     list_.prev_ = &list_;
     list_.next_ = &list_;
     list_.number_ = 0xFFFFFFFFL;      // placeholder marker, for debugging
+    // Set all the variables to make UBSAN happy.
+    list_.list_ = nullptr;
+    list_.unix_time_ = 0;
+    list_.is_write_conflict_boundary_ = false;
     count_ = 0;
   }
 

db/table_cache.cc

@@ -30,7 +30,7 @@ namespace rocksdb {
 namespace {
 
 template <class T>
-static void DeleteEntry(const Slice& key, void* value) {
+static void DeleteEntry(const Slice& /*key*/, void* value) {
   T* typed_value = reinterpret_cast<T*>(value);
   delete typed_value;
 }
@@ -41,7 +41,7 @@ static void UnrefEntry(void* arg1, void* arg2) {
   cache->Release(h);
 }
 
-static void DeleteTableReader(void* arg1, void* arg2) {
+static void DeleteTableReader(void* arg1, void* /*arg2*/) {
   TableReader* table_reader = reinterpret_cast<TableReader*>(arg1);
   delete table_reader;
 }
@@ -272,9 +272,8 @@ InternalIterator* TableCache::NewRangeTombstoneIterator(
     const InternalKeyComparator& icomparator, const FileDescriptor& fd,
     HistogramImpl* file_read_hist, bool skip_filters, int level) {
   Status s;
-  TableReader* table_reader = nullptr;
   Cache::Handle* handle = nullptr;
-  table_reader = fd.table_reader;
+  TableReader* table_reader = fd.table_reader;
   if (table_reader == nullptr) {
     s = FindTable(env_options, icomparator, fd, &handle,
                   options.read_tier == kBlockCacheTier /* no_io */,

db/table_properties_collector.cc

@@ -12,8 +12,8 @@
 namespace rocksdb {
 
 Status InternalKeyPropertiesCollector::InternalAdd(const Slice& key,
-                                                   const Slice& value,
-                                                   uint64_t file_size) {
+                                                   const Slice& /*value*/,
+                                                   uint64_t /*file_size*/) {
   ParsedInternalKey ikey;
   if (!ParseInternalKey(key, &ikey)) {
     return Status::InvalidArgument("Invalid internal key");
@@ -60,7 +60,7 @@ InternalKeyPropertiesCollector::GetReadableProperties() const {
 namespace {
 
 uint64_t GetUint64Property(const UserCollectedProperties& props,
-                           const std::string property_name,
+                           const std::string& property_name,
                            bool* property_present) {
   auto pos = props.find(property_name);
   if (pos == props.end()) {

db/table_properties_collector.h

@@ -73,7 +73,7 @@ class InternalKeyPropertiesCollectorFactory
     : public IntTblPropCollectorFactory {
  public:
   virtual IntTblPropCollector* CreateIntTblPropCollector(
-      uint32_t column_family_id) override {
+      uint32_t /*column_family_id*/) override {
     return new InternalKeyPropertiesCollector();
   }
 

db/table_properties_collector_test.cc

@@ -82,8 +82,9 @@ class RegularKeysStartWithA: public TablePropertiesCollector {
      return Status::OK();
   }
 
-  Status AddUserKey(const Slice& user_key, const Slice& value, EntryType type,
-                    SequenceNumber seq, uint64_t file_size) override {
+  Status AddUserKey(const Slice& user_key, const Slice& /*value*/,
+                    EntryType type, SequenceNumber /*seq*/,
+                    uint64_t file_size) override {
     // simply asssume all user keys are not empty.
     if (user_key.data()[0] == 'A') {
       ++count_;
@@ -133,7 +134,7 @@ class RegularKeysStartWithABackwardCompatible
     return Status::OK();
   }
 
-  Status Add(const Slice& user_key, const Slice& value) override {
+  Status Add(const Slice& user_key, const Slice& /*value*/) override {
     // simply asssume all user keys are not empty.
     if (user_key.data()[0] == 'A') {
       ++count_;
@@ -161,8 +162,8 @@ class RegularKeysStartWithAInternal : public IntTblPropCollector {
     return Status::OK();
   }
 
-  Status InternalAdd(const Slice& user_key, const Slice& value,
-                     uint64_t file_size) override {
+  Status InternalAdd(const Slice& user_key, const Slice& /*value*/,
+                     uint64_t /*file_size*/) override {
     // simply asssume all user keys are not empty.
     if (user_key.data()[0] == 'A') {
       ++count_;
@@ -193,7 +194,7 @@ class RegularKeysStartWithAFactory : public IntTblPropCollectorFactory,
     }
   }
   virtual IntTblPropCollector* CreateIntTblPropCollector(
-      uint32_t column_family_id) override {
+      uint32_t /*column_family_id*/) override {
     return new RegularKeysStartWithAInternal();
   }
   const char* Name() const override { return "RegularKeysStartWithA"; }
@@ -203,7 +204,7 @@ class RegularKeysStartWithAFactory : public IntTblPropCollectorFactory,
 
 class FlushBlockEveryThreePolicy : public FlushBlockPolicy {
  public:
-  virtual bool Update(const Slice& key, const Slice& value) override {
+  virtual bool Update(const Slice& /*key*/, const Slice& /*value*/) override {
     return (++count_ % 3U == 0);
   }
 
@@ -220,8 +221,8 @@ class FlushBlockEveryThreePolicyFactory : public FlushBlockPolicyFactory {
   }
 
   FlushBlockPolicy* NewFlushBlockPolicy(
-      const BlockBasedTableOptions& table_options,
-      const BlockBuilder& data_block_builder) const override {
+      const BlockBasedTableOptions& /*table_options*/,
+      const BlockBuilder& /*data_block_builder*/) const override {
     return new FlushBlockEveryThreePolicy;
   }
 };

db/transaction_log_impl.cc

@@ -269,16 +269,18 @@ void TransactionLogIteratorImpl::UpdateCurrentWriteBatch(const Slice& record) {
       return Status::OK();
     }
 
-    Status PutCF(uint32_t cf, const Slice& key, const Slice& val) override {
+    Status PutCF(uint32_t /*cf*/, const Slice& /*key*/,
+                 const Slice& /*val*/) override {
       return Status::OK();
     }
-    Status DeleteCF(uint32_t cf, const Slice& key) override {
+    Status DeleteCF(uint32_t /*cf*/, const Slice& /*key*/) override {
       return Status::OK();
     }
-    Status SingleDeleteCF(uint32_t cf, const Slice& key) override {
+    Status SingleDeleteCF(uint32_t /*cf*/, const Slice& /*key*/) override {
       return Status::OK();
     }
-    Status MergeCF(uint32_t cf, const Slice& key, const Slice& val) override {
+    Status MergeCF(uint32_t /*cf*/, const Slice& /*key*/,
+                   const Slice& /*val*/) override {
       return Status::OK();
     }
     Status MarkBeginPrepare() override { return Status::OK(); }

db/version_builder.cc

@@ -199,7 +199,7 @@ class VersionBuilder::Rep {
     }
   }
 
-  void CheckConsistencyForDeletes(VersionEdit* edit, uint64_t number,
+  void CheckConsistencyForDeletes(VersionEdit* /*edit*/, uint64_t number,
                                   int level) {
 #ifdef NDEBUG
     if (!base_vstorage_->force_consistency_checks()) {

db/version_edit.cc

@@ -198,7 +198,7 @@ static bool GetInternalKey(Slice* input, InternalKey* dst) {
   }
 }
 
-bool VersionEdit::GetLevel(Slice* input, int* level, const char** msg) {
+bool VersionEdit::GetLevel(Slice* input, int* level, const char** /*msg*/) {
   uint32_t v;
   if (GetVarint32(input, &v)) {
     *level = v;

db/version_set.cc

@@ -424,129 +424,261 @@ bool SomeFileOverlapsRange(
 }
 
 namespace {
-
-// An internal iterator.  For a given version/level pair, yields
-// information about the files in the level.  For a given entry, key()
-// is the largest key that occurs in the file, and value() is an
-// 16-byte value containing the file number and file size, both
-// encoded using EncodeFixed64.
-class LevelFileNumIterator : public InternalIterator {
+class LevelIterator final : public InternalIterator {
  public:
-  LevelFileNumIterator(const InternalKeyComparator& icmp,
-                       const LevelFilesBrief* flevel, bool should_sample)
-      : icmp_(icmp),
-        flevel_(flevel),
-        index_(static_cast<uint32_t>(flevel->num_files)),
-        current_value_(0, 0, 0),  // Marks as invalid
-        should_sample_(should_sample) {}
-  virtual bool Valid() const override { return index_ < flevel_->num_files; }
-  virtual void Seek(const Slice& target) override {
-    index_ = FindFile(icmp_, *flevel_, target);
-  }
-  virtual void SeekForPrev(const Slice& target) override {
-    SeekForPrevImpl(target, &icmp_);
-  }
-
-  virtual void SeekToFirst() override { index_ = 0; }
-  virtual void SeekToLast() override {
-    index_ = (flevel_->num_files == 0)
-                 ? 0
-                 : static_cast<uint32_t>(flevel_->num_files) - 1;
-  }
-  virtual void Next() override {
-    assert(Valid());
-    index_++;
-  }
-  virtual void Prev() override {
-    assert(Valid());
-    if (index_ == 0) {
-      index_ = static_cast<uint32_t>(flevel_->num_files);  // Marks as invalid
-    } else {
-      index_--;
-    }
-  }
-  Slice key() const override {
-    assert(Valid());
-    return flevel_->files[index_].largest_key;
-  }
-  Slice value() const override {
-    assert(Valid());
-
-    auto file_meta = flevel_->files[index_];
-    if (should_sample_) {
-      sample_file_read_inc(file_meta.file_metadata);
-    }
-    current_value_ = file_meta.fd;
-    return Slice(reinterpret_cast<const char*>(&current_value_),
-                 sizeof(FileDescriptor));
-  }
-  virtual Status status() const override { return Status::OK(); }
-
- private:
-  const InternalKeyComparator icmp_;
-  const LevelFilesBrief* flevel_;
-  uint32_t index_;
-  mutable FileDescriptor current_value_;
-  bool should_sample_;
-};
-
-class LevelFileIteratorState : public TwoLevelIteratorState {
- public:
-  // @param skip_filters Disables loading/accessing the filter block
-  LevelFileIteratorState(TableCache* table_cache,
-                         const ReadOptions& read_options,
-                         const EnvOptions& env_options,
-                         const InternalKeyComparator& icomparator,
-                         HistogramImpl* file_read_hist, bool for_compaction,
-                         bool prefix_enabled, bool skip_filters, int level,
-                         RangeDelAggregator* range_del_agg)
-      : TwoLevelIteratorState(prefix_enabled),
-        table_cache_(table_cache),
+  LevelIterator(TableCache* table_cache, const ReadOptions& read_options,
+                const EnvOptions& env_options,
+                const InternalKeyComparator& icomparator,
+                const LevelFilesBrief* flevel, bool should_sample,
+                HistogramImpl* file_read_hist, bool for_compaction,
+                bool skip_filters, int level, RangeDelAggregator* range_del_agg)
+      : table_cache_(table_cache),
         read_options_(read_options),
         env_options_(env_options),
         icomparator_(icomparator),
+        flevel_(flevel),
         file_read_hist_(file_read_hist),
+        should_sample_(should_sample),
         for_compaction_(for_compaction),
         skip_filters_(skip_filters),
+        file_index_(flevel_->num_files),
         level_(level),
-        range_del_agg_(range_del_agg) {}
+        range_del_agg_(range_del_agg),
+        pinned_iters_mgr_(nullptr) {
+    // Empty level is not supported.
+    assert(flevel_ != nullptr && flevel_->num_files > 0);
+  }
 
-  InternalIterator* NewSecondaryIterator(const Slice& meta_handle) override {
-    if (meta_handle.size() != sizeof(FileDescriptor)) {
-      return NewErrorInternalIterator(
-          Status::Corruption("FileReader invoked with unexpected value"));
+  virtual ~LevelIterator() { delete file_iter_.Set(nullptr); }
+
+  virtual void Seek(const Slice& target) override;
+  virtual void SeekForPrev(const Slice& target) override;
+  virtual void SeekToFirst() override;
+  virtual void SeekToLast() override;
+  virtual void Next() override;
+  virtual void Prev() override;
+
+  virtual bool Valid() const override { return file_iter_.Valid(); }
+  virtual Slice key() const override {
+    assert(Valid());
+    return file_iter_.key();
+  }
+  virtual Slice value() const override {
+    assert(Valid());
+    return file_iter_.value();
+  }
+  virtual Status status() const override {
+    // It'd be nice if status() returned a const Status& instead of a Status
+    if (!status_.ok()) {
+      return status_;
+    } else if (file_iter_.iter() != nullptr) {
+      return file_iter_.status();
     }
-    const FileDescriptor* fd =
-        reinterpret_cast<const FileDescriptor*>(meta_handle.data());
-    return table_cache_->NewIterator(
-        read_options_, env_options_, icomparator_, *fd, range_del_agg_,
-        nullptr /* don't need reference to table */, file_read_hist_,
-        for_compaction_, nullptr /* arena */, skip_filters_, level_);
+    return Status::OK();
+  }
+  virtual void SetPinnedItersMgr(
+      PinnedIteratorsManager* pinned_iters_mgr) override {
+    pinned_iters_mgr_ = pinned_iters_mgr;
+    if (file_iter_.iter()) {
+      file_iter_.SetPinnedItersMgr(pinned_iters_mgr);
+    }
+  }
+  virtual bool IsKeyPinned() const override {
+    return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
+           file_iter_.iter() && file_iter_.IsKeyPinned();
+  }
+  virtual bool IsValuePinned() const override {
+    return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
+           file_iter_.iter() && file_iter_.IsValuePinned();
   }
 
-  bool PrefixMayMatch(const Slice& internal_key) override {
-    return true;
+ private:
+  void SkipEmptyFileForward();
+  void SkipEmptyFileBackward();
+  void SetFileIterator(InternalIterator* iter);
+  void InitFileIterator(size_t new_file_index);
+
+  const Slice& file_smallest_key(size_t file_index) {
+    assert(file_index < flevel_->num_files);
+    return flevel_->files[file_index].smallest_key;
   }
 
-  bool KeyReachedUpperBound(const Slice& internal_key) override {
+  bool KeyReachedUpperBound(const Slice& internal_key) {
     return read_options_.iterate_upper_bound != nullptr &&
            icomparator_.user_comparator()->Compare(
                ExtractUserKey(internal_key),
                *read_options_.iterate_upper_bound) >= 0;
   }
 
- private:
+  InternalIterator* NewFileIterator() {
+    assert(file_index_ < flevel_->num_files);
+    auto file_meta = flevel_->files[file_index_];
+    if (should_sample_) {
+      sample_file_read_inc(file_meta.file_metadata);
+    }
+
+    return table_cache_->NewIterator(
+        read_options_, env_options_, icomparator_, file_meta.fd, range_del_agg_,
+        nullptr /* don't need reference to table */, file_read_hist_,
+        for_compaction_, nullptr /* arena */, skip_filters_, level_);
+  }
+
   TableCache* table_cache_;
   const ReadOptions read_options_;
   const EnvOptions& env_options_;
   const InternalKeyComparator& icomparator_;
+  const LevelFilesBrief* flevel_;
+  mutable FileDescriptor current_value_;
+
   HistogramImpl* file_read_hist_;
+  bool should_sample_;
   bool for_compaction_;
   bool skip_filters_;
+  size_t file_index_;
   int level_;
   RangeDelAggregator* range_del_agg_;
+  IteratorWrapper file_iter_;  // May be nullptr
+  PinnedIteratorsManager* pinned_iters_mgr_;
+  Status status_;
 };
 
+void LevelIterator::Seek(const Slice& target) {
+  size_t new_file_index = FindFile(icomparator_, *flevel_, target);
+
+  InitFileIterator(new_file_index);
+  if (file_iter_.iter() != nullptr) {
+    file_iter_.Seek(target);
+  }
+  SkipEmptyFileForward();
+}
+
+void LevelIterator::SeekForPrev(const Slice& target) {
+  size_t new_file_index = FindFile(icomparator_, *flevel_, target);
+  if (new_file_index >= flevel_->num_files) {
+    new_file_index = flevel_->num_files - 1;
+  }
+
+  InitFileIterator(new_file_index);
+  if (file_iter_.iter() != nullptr) {
+    file_iter_.SeekForPrev(target);
+    SkipEmptyFileBackward();
+  }
+}
+
+void LevelIterator::SeekToFirst() {
+  InitFileIterator(0);
+  if (file_iter_.iter() != nullptr) {
+    file_iter_.SeekToFirst();
+  }
+  SkipEmptyFileForward();
+}
+
+void LevelIterator::SeekToLast() {
+  InitFileIterator(flevel_->num_files - 1);
+  if (file_iter_.iter() != nullptr) {
+    file_iter_.SeekToLast();
+  }
+  SkipEmptyFileBackward();
+}
+
+void LevelIterator::Next() {
+  assert(Valid());
+  file_iter_.Next();
+  SkipEmptyFileForward();
+}
+
+void LevelIterator::Prev() {
+  assert(Valid());
+  file_iter_.Prev();
+  SkipEmptyFileBackward();
+}
+
+void LevelIterator::SkipEmptyFileForward() {
+  // For an error (IO error, checksum mismatch, etc), we skip the file
+  // and move to the next one and continue reading data.
+  // TODO this behavior is from LevelDB. We should revisit it.
+  while (file_iter_.iter() == nullptr ||
+         (!file_iter_.Valid() && !file_iter_.status().IsIncomplete())) {
+    if (file_iter_.iter() != nullptr && !file_iter_.Valid() &&
+        file_iter_.iter()->IsOutOfBound()) {
+      return;
+    }
+
+    // Move to next file
+    if (file_index_ >= flevel_->num_files - 1) {
+      // Already at the last file
+      SetFileIterator(nullptr);
+      return;
+    }
+    if (KeyReachedUpperBound(file_smallest_key(file_index_ + 1))) {
+      SetFileIterator(nullptr);
+      return;
+    }
+    InitFileIterator(file_index_ + 1);
+    if (file_iter_.iter() != nullptr) {
+      file_iter_.SeekToFirst();
+    }
+  }
+}
+
+void LevelIterator::SkipEmptyFileBackward() {
+  while (file_iter_.iter() == nullptr ||
+         (!file_iter_.Valid() && !file_iter_.status().IsIncomplete())) {
+    // Move to previous file
+    if (file_index_ == 0) {
+      // Already the first file
+      SetFileIterator(nullptr);
+      return;
+    }
+    InitFileIterator(file_index_ - 1);
+    if (file_iter_.iter() != nullptr) {
+      file_iter_.SeekToLast();
+    }
+  }
+}
+
+void LevelIterator::SetFileIterator(InternalIterator* iter) {
+  if (file_iter_.iter() != nullptr && status_.ok()) {
+    // TODO right now we don't invalidate the iterator even if the status is
+    // not OK. We should consider to do that so that it is harder for users to
+    // skip errors.
+    status_ = file_iter_.status();
+  }
+
+  if (pinned_iters_mgr_ && iter) {
+    iter->SetPinnedItersMgr(pinned_iters_mgr_);
+  }
+
+  InternalIterator* old_iter = file_iter_.Set(iter);
+  if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
+    pinned_iters_mgr_->PinIterator(old_iter);
+  } else {
+    delete old_iter;
+  }
+}
+
+void LevelIterator::InitFileIterator(size_t new_file_index) {
+  if (new_file_index >= flevel_->num_files) {
+    file_index_ = new_file_index;
+    SetFileIterator(nullptr);
+    return;
+  } else {
+    // If the file iterator shows incomplete, we try it again if users seek
+    // to the same file, as this time we may go to a different data block
+    // which is cached in block cache.
+    //
+    if (file_iter_.iter() != nullptr && !file_iter_.status().IsIncomplete() &&
+        new_file_index == file_index_) {
+      // file_iter_ is already constructed with this iterator, so
+      // no need to change anything
+    } else {
+      file_index_ = new_file_index;
+      InternalIterator* iter = NewFileIterator();
+      SetFileIterator(iter);
+    }
+  }
+}
+
 // A wrapper of version builder which references the current version in
 // constructor and unref it in the destructor.
 // Both of the constructor and destructor need to be called inside DB Mutex.
@@ -759,6 +891,15 @@ void Version::GetColumnFamilyMetaData(ColumnFamilyMetaData* cf_meta) {
   }
 }
 
+uint64_t Version::GetSstFilesSize() {
+  uint64_t sst_files_size = 0;
+  for (int level = 0; level < storage_info_.num_levels_; level++) {
+    for (const auto& file_meta : storage_info_.LevelFiles(level)) {
+      sst_files_size += file_meta->fd.GetFileSize();
+    }
+  }
+  return sst_files_size;
+}
 
 uint64_t VersionStorageInfo::GetEstimatedActiveKeys() const {
   // Estimation will be inaccurate when:
@@ -854,24 +995,18 @@ void Version::AddIteratorsForLevel(const ReadOptions& read_options,
         sample_file_read_inc(meta);
       }
     }
-  } else {
+  } else if (storage_info_.LevelFilesBrief(level).num_files > 0) {
     // For levels > 0, we can use a concatenating iterator that sequentially
     // walks through the non-overlapping files in the level, opening them
     // lazily.
-    auto* mem = arena->AllocateAligned(sizeof(LevelFileIteratorState));
-    auto* state = new (mem)
-        LevelFileIteratorState(cfd_->table_cache(), read_options, soptions,
-                               cfd_->internal_comparator(),
-                               cfd_->internal_stats()->GetFileReadHist(level),
-                               false /* for_compaction */,
-                               cfd_->ioptions()->prefix_extractor != nullptr,
-                               IsFilterSkipped(level), level, range_del_agg);
-    mem = arena->AllocateAligned(sizeof(LevelFileNumIterator));
-    auto* first_level_iter = new (mem) LevelFileNumIterator(
+    auto* mem = arena->AllocateAligned(sizeof(LevelIterator));
+    merge_iter_builder->AddIterator(new (mem) LevelIterator(
+        cfd_->table_cache(), read_options, soptions,
         cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level),
-        should_sample_file_read());
-    merge_iter_builder->AddIterator(
-        NewTwoLevelIterator(state, first_level_iter, arena, false));
+        should_sample_file_read(),
+        cfd_->internal_stats()->GetFileReadHist(level),
+        false /* for_compaction */, IsFilterSkipped(level), level,
+        range_del_agg));
   }
 }
 
@@ -1976,7 +2111,7 @@ void VersionStorageInfo::ExtendFileRangeOverlappingInterval(
 #endif
   *start_index = mid_index + 1;
   *end_index = mid_index;
-  int count __attribute__((unused));
+  int count __attribute__((__unused__));
   count = 0;
 
   // check backwards from 'mid' to lower indices
@@ -2772,7 +2907,7 @@ void VersionSet::LogAndApplyCFHelper(VersionEdit* edit) {
 }
 
 void VersionSet::LogAndApplyHelper(ColumnFamilyData* cfd,
-                                   VersionBuilder* builder, Version* v,
+                                   VersionBuilder* builder, Version* /*v*/,
                                    VersionEdit* edit, InstrumentedMutex* mu) {
   mu->AssertHeld();
   assert(!edit->IsColumnFamilyManipulation());
@@ -3732,17 +3867,13 @@ InternalIterator* VersionSet::MakeInputIterator(
         }
       } else {
         // Create concatenating iterator for the files from this level
-        list[num++] = NewTwoLevelIterator(
-            new LevelFileIteratorState(
-                cfd->table_cache(), read_options, env_options_compactions,
-                cfd->internal_comparator(),
-                nullptr /* no per level latency histogram */,
-                true /* for_compaction */, false /* prefix enabled */,
-                false /* skip_filters */, (int)which /* level */,
-                range_del_agg),
-            new LevelFileNumIterator(cfd->internal_comparator(),
-                                     c->input_levels(which),
-                                     false /* don't sample compaction */));
+        list[num++] = new LevelIterator(
+            cfd->table_cache(), read_options, env_options_compactions,
+            cfd->internal_comparator(), c->input_levels(which),
+            false /* should_sample */,
+            nullptr /* no per level latency histogram */,
+            true /* for_compaction */, false /* skip_filters */,
+            (int)which /* level */, range_del_agg);
       }
     }
   }

db/version_set.h

@@ -613,6 +613,8 @@ class Version {
 
   void GetColumnFamilyMetaData(ColumnFamilyMetaData* cf_meta);
 
+  uint64_t GetSstFilesSize();
+
  private:
   Env* env_;
   friend class VersionSet;
@@ -870,7 +872,7 @@ class VersionSet {
 
   struct LogReporter : public log::Reader::Reporter {
     Status* status;
-    virtual void Corruption(size_t bytes, const Status& s) override {
+    virtual void Corruption(size_t /*bytes*/, const Status& s) override {
       if (this->status->ok()) *this->status = s;
     }
   };

db/version_set_test.cc

@@ -76,7 +76,9 @@ class CountingLogger : public Logger {
  public:
   CountingLogger() : log_count(0) {}
   using Logger::Logv;
-  virtual void Logv(const char* format, va_list ap) override { log_count++; }
+  virtual void Logv(const char* /*format*/, va_list /*ap*/) override {
+    log_count++;
+  }
   int log_count;
 };
 

db/wal_manager_test.cc

@@ -73,7 +73,7 @@ class WalManagerTest : public testing::Test {
   }
 
   // NOT thread safe
-  void RollTheLog(bool archived) {
+  void RollTheLog(bool /*archived*/) {
     current_log_number_++;
     std::string fname = ArchivedLogFileName(dbname_, current_log_number_);
     unique_ptr<WritableFile> file;

db/write_batch.cc

@@ -186,7 +186,7 @@ WriteBatch::~WriteBatch() { delete save_points_; }
 
 WriteBatch::Handler::~Handler() { }
 
-void WriteBatch::Handler::LogData(const Slice& blob) {
+void WriteBatch::Handler::LogData(const Slice& /*blob*/) {
   // If the user has not specified something to do with blobs, then we ignore
   // them.
 }
@@ -402,14 +402,22 @@ Status WriteBatch::Iterate(Handler* handler) const {
   bool empty_batch = true;
   int found = 0;
   Status s;
-  while (s.ok() && !input.empty() && handler->Continue()) {
-    char tag = 0;
-    uint32_t column_family = 0;  // default
+  char tag = 0;
+  uint32_t column_family = 0;  // default
+  while ((s.ok() || UNLIKELY(s.IsTryAgain())) && !input.empty() &&
+         handler->Continue()) {
+    if (LIKELY(!s.IsTryAgain())) {
+      tag = 0;
+      column_family = 0;  // default
 
-    s = ReadRecordFromWriteBatch(&input, &tag, &column_family, &key, &value,
-                                 &blob, &xid);
-    if (!s.ok()) {
-      return s;
+      s = ReadRecordFromWriteBatch(&input, &tag, &column_family, &key, &value,
+                                   &blob, &xid);
+      if (!s.ok()) {
+        return s;
+      }
+    } else {
+      assert(s.IsTryAgain());
+      s = Status::OK();
     }
 
     switch (tag) {
@@ -418,47 +426,59 @@ Status WriteBatch::Iterate(Handler* handler) const {
         assert(content_flags_.load(std::memory_order_relaxed) &
                (ContentFlags::DEFERRED | ContentFlags::HAS_PUT));
         s = handler->PutCF(column_family, key, value);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
         break;
       case kTypeColumnFamilyDeletion:
       case kTypeDeletion:
         assert(content_flags_.load(std::memory_order_relaxed) &
                (ContentFlags::DEFERRED | ContentFlags::HAS_DELETE));
         s = handler->DeleteCF(column_family, key);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
         break;
       case kTypeColumnFamilySingleDeletion:
       case kTypeSingleDeletion:
         assert(content_flags_.load(std::memory_order_relaxed) &
                (ContentFlags::DEFERRED | ContentFlags::HAS_SINGLE_DELETE));
         s = handler->SingleDeleteCF(column_family, key);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          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);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
         break;
       case kTypeColumnFamilyMerge:
       case kTypeMerge:
         assert(content_flags_.load(std::memory_order_relaxed) &
                (ContentFlags::DEFERRED | ContentFlags::HAS_MERGE));
         s = handler->MergeCF(column_family, key, value);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
         break;
       case kTypeColumnFamilyBlobIndex:
       case kTypeBlobIndex:
         assert(content_flags_.load(std::memory_order_relaxed) &
                (ContentFlags::DEFERRED | ContentFlags::HAS_BLOB_INDEX));
         s = handler->PutBlobIndexCF(column_family, key, value);
-        found++;
+        if (LIKELY(s.ok())) {
+          found++;
+        }
         break;
       case kTypeLogData:
         handler->LogData(blob);
@@ -549,12 +569,19 @@ void WriteBatchInternal::SetSequence(WriteBatch* b, SequenceNumber seq) {
   EncodeFixed64(&b->rep_[0], seq);
 }
 
-size_t WriteBatchInternal::GetFirstOffset(WriteBatch* b) {
+size_t WriteBatchInternal::GetFirstOffset(WriteBatch* /*b*/) {
   return WriteBatchInternal::kHeader;
 }
 
 Status WriteBatchInternal::Put(WriteBatch* b, uint32_t column_family_id,
                                const Slice& key, const Slice& value) {
+  if (key.size() > size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("key is too large");
+  }
+  if (value.size() > size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("value is too large");
+  }
+
   LocalSavePoint save(b);
   WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
   if (column_family_id == 0) {
@@ -577,8 +604,33 @@ Status WriteBatch::Put(ColumnFamilyHandle* column_family, const Slice& key,
                                  value);
 }
 
+Status WriteBatchInternal::CheckSlicePartsLength(const SliceParts& key,
+                                                 const SliceParts& value) {
+  size_t total_key_bytes = 0;
+  for (int i = 0; i < key.num_parts; ++i) {
+    total_key_bytes += key.parts[i].size();
+  }
+  if (total_key_bytes >= size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("key is too large");
+  }
+
+  size_t total_value_bytes = 0;
+  for (int i = 0; i < value.num_parts; ++i) {
+    total_value_bytes += value.parts[i].size();
+  }
+  if (total_value_bytes >= size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("value is too large");
+  }
+  return Status::OK();
+}
+
 Status WriteBatchInternal::Put(WriteBatch* b, uint32_t column_family_id,
                                const SliceParts& key, const SliceParts& value) {
+  Status s = CheckSlicePartsLength(key, value);
+  if (!s.ok()) {
+    return s;
+  }
+
   LocalSavePoint save(b);
   WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
   if (column_family_id == 0) {
@@ -795,6 +847,13 @@ Status WriteBatch::DeleteRange(ColumnFamilyHandle* column_family,
 
 Status WriteBatchInternal::Merge(WriteBatch* b, uint32_t column_family_id,
                                  const Slice& key, const Slice& value) {
+  if (key.size() > size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("key is too large");
+  }
+  if (value.size() > size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("value is too large");
+  }
+
   LocalSavePoint save(b);
   WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
   if (column_family_id == 0) {
@@ -820,6 +879,11 @@ Status WriteBatch::Merge(ColumnFamilyHandle* column_family, const Slice& key,
 Status WriteBatchInternal::Merge(WriteBatch* b, uint32_t column_family_id,
                                  const SliceParts& key,
                                  const SliceParts& value) {
+  Status s = CheckSlicePartsLength(key, value);
+  if (!s.ok()) {
+    return s;
+  }
+
   LocalSavePoint save(b);
   WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
   if (column_family_id == 0) {
@@ -1084,12 +1148,23 @@ class MemTableInserter : public WriteBatch::Handler {
       MaybeAdvanceSeq();
       return seek_status;
     }
+    Status ret_status;
 
     MemTable* mem = cf_mems_->GetMemTable();
     auto* moptions = mem->GetImmutableMemTableOptions();
+    // inplace_update_support is inconsistent with snapshots, and therefore with
+    // any kind of transactions including the ones that use seq_per_batch
+    assert(!seq_per_batch_ || !moptions->inplace_update_support);
     if (!moptions->inplace_update_support) {
-      mem->Add(sequence_, value_type, key, value, concurrent_memtable_writes_,
-               get_post_process_info(mem));
+      bool mem_res =
+          mem->Add(sequence_, value_type, key, value,
+                   concurrent_memtable_writes_, get_post_process_info(mem));
+      if (UNLIKELY(!mem_res)) {
+        assert(seq_per_batch_);
+        ret_status = Status::TryAgain("key+seq exists");
+        const bool BATCH_BOUNDRY = true;
+        MaybeAdvanceSeq(BATCH_BOUNDRY);
+      }
     } else if (moptions->inplace_callback == nullptr) {
       assert(!concurrent_memtable_writes_);
       mem->Update(sequence_, key, value);
@@ -1125,11 +1200,17 @@ class MemTableInserter : public WriteBatch::Handler {
                                                  value, &merged_value);
         if (status == UpdateStatus::UPDATED_INPLACE) {
           // prev_value is updated in-place with final value.
-          mem->Add(sequence_, value_type, key, Slice(prev_buffer, prev_size));
+          bool mem_res __attribute__((__unused__));
+          mem_res = mem->Add(
+              sequence_, value_type, key, Slice(prev_buffer, prev_size));
+          assert(mem_res);
           RecordTick(moptions->statistics, NUMBER_KEYS_WRITTEN);
         } else if (status == UpdateStatus::UPDATED) {
           // merged_value contains the final value.
-          mem->Add(sequence_, value_type, key, Slice(merged_value));
+          bool mem_res __attribute__((__unused__));
+          mem_res =
+              mem->Add(sequence_, value_type, key, Slice(merged_value));
+          assert(mem_res);
           RecordTick(moptions->statistics, NUMBER_KEYS_WRITTEN);
         }
       }
@@ -1139,7 +1220,7 @@ class MemTableInserter : public WriteBatch::Handler {
     // in memtable add/update.
     MaybeAdvanceSeq();
     CheckMemtableFull();
-    return Status::OK();
+    return ret_status;
   }
 
   virtual Status PutCF(uint32_t column_family_id, const Slice& key,
@@ -1147,14 +1228,22 @@ class MemTableInserter : public WriteBatch::Handler {
     return PutCFImpl(column_family_id, key, value, kTypeValue);
   }
 
-  Status DeleteImpl(uint32_t column_family_id, const Slice& key,
+  Status DeleteImpl(uint32_t /*column_family_id*/, const Slice& key,
                     const Slice& value, ValueType delete_type) {
+    Status ret_status;
     MemTable* mem = cf_mems_->GetMemTable();
-    mem->Add(sequence_, delete_type, key, value, concurrent_memtable_writes_,
-             get_post_process_info(mem));
+    bool mem_res =
+        mem->Add(sequence_, delete_type, key, value,
+                 concurrent_memtable_writes_, get_post_process_info(mem));
+    if (UNLIKELY(!mem_res)) {
+      assert(seq_per_batch_);
+      ret_status = Status::TryAgain("key+seq exists");
+      const bool BATCH_BOUNDRY = true;
+      MaybeAdvanceSeq(BATCH_BOUNDRY);
+    }
     MaybeAdvanceSeq();
     CheckMemtableFull();
-    return Status::OK();
+    return ret_status;
   }
 
   virtual Status DeleteCF(uint32_t column_family_id,
@@ -1246,6 +1335,7 @@ class MemTableInserter : public WriteBatch::Handler {
       return seek_status;
     }
 
+    Status ret_status;
     MemTable* mem = cf_mems_->GetMemTable();
     auto* moptions = mem->GetImmutableMemTableOptions();
     bool perform_merge = false;
@@ -1301,18 +1391,30 @@ class MemTableInserter : public WriteBatch::Handler {
         perform_merge = false;
       } else {
         // 3) Add value to memtable
-        mem->Add(sequence_, kTypeValue, key, new_value);
+        bool mem_res = mem->Add(sequence_, kTypeValue, key, new_value);
+        if (UNLIKELY(!mem_res)) {
+          assert(seq_per_batch_);
+          ret_status = Status::TryAgain("key+seq exists");
+          const bool BATCH_BOUNDRY = true;
+          MaybeAdvanceSeq(BATCH_BOUNDRY);
+        }
       }
     }
 
     if (!perform_merge) {
       // Add merge operator to memtable
-      mem->Add(sequence_, kTypeMerge, key, value);
+      bool mem_res = mem->Add(sequence_, kTypeMerge, key, value);
+      if (UNLIKELY(!mem_res)) {
+        assert(seq_per_batch_);
+        ret_status = Status::TryAgain("key+seq exists");
+        const bool BATCH_BOUNDRY = true;
+        MaybeAdvanceSeq(BATCH_BOUNDRY);
+      }
     }
 
     MaybeAdvanceSeq();
     CheckMemtableFull();
-    return Status::OK();
+    return ret_status;
   }
 
   virtual Status PutBlobIndexCF(uint32_t column_family_id, const Slice& key,
@@ -1496,6 +1598,8 @@ Status WriteBatchInternal::InsertInto(
     if (!w->status.ok()) {
       return w->status;
     }
+    assert(!seq_per_batch || w->batch_cnt != 0);
+    assert(!seq_per_batch || inserter.sequence() - w->sequence == w->batch_cnt);
   }
   return Status::OK();
 }
@@ -1504,7 +1608,7 @@ Status WriteBatchInternal::InsertInto(
     WriteThread::Writer* writer, SequenceNumber sequence,
     ColumnFamilyMemTables* memtables, FlushScheduler* flush_scheduler,
     bool ignore_missing_column_families, uint64_t log_number, DB* db,
-    bool concurrent_memtable_writes, bool seq_per_batch) {
+    bool concurrent_memtable_writes, bool seq_per_batch, size_t batch_cnt) {
   assert(writer->ShouldWriteToMemtable());
   MemTableInserter inserter(sequence, memtables, flush_scheduler,
                             ignore_missing_column_families, log_number, db,
@@ -1513,6 +1617,8 @@ Status WriteBatchInternal::InsertInto(
   SetSequence(writer->batch, sequence);
   inserter.set_log_number_ref(writer->log_ref);
   Status s = writer->batch->Iterate(&inserter);
+  assert(!seq_per_batch || batch_cnt != 0);
+  assert(!seq_per_batch || inserter.sequence() - sequence == batch_cnt);
   if (concurrent_memtable_writes) {
     inserter.PostProcess();
   }