Make writable_file_max_buffer_size dynamic

Summary:
The DBOptions::writable_file_max_buffer_size can be changed dynamically.
Closes facebook#3053

Differential Revision: D6152720

Pulled By: shligit

fbshipit-source-id: aa0c0cfcfae6a54eb17faadb148d904797c68681

Author: shligit

HISTORY.md

@@ -6,6 +6,7 @@
 * Return an error on write if write_options.sync = true and write_options.disableWAL = true to warn user of inconsistent options. Previously we will not write to WAL and not respecting the sync options in this case.
 
 ### New Features
+* `DBOptions::writable_file_max_buffer_size` can now be changed dynamically.
 * `DBOptions::bytes_per_sync` and `DBOptions::wal_bytes_per_sync` can now be changed dynamically, `DBOptions::wal_bytes_per_sync` will flush all memtables and switch to a new WAL file.
 * Support dynamic adjustment of rate limit according to demand for background I/O. It can be enabled by passing `true` to the `auto_tuned` parameter in `NewGenericRateLimiter()`. The value passed as `rate_bytes_per_sec` will still be respected as an upper-bound.
 * Support dynamically changing `ColumnFamilyOptions::compaction_options_fifo`.

db/c.cc

@@ -2276,6 +2276,11 @@ void rocksdb_options_set_bytes_per_sync(
   opt->rep.bytes_per_sync = v;
 }
 
+void rocksdb_options_set_writable_file_max_buffer_size(rocksdb_options_t* opt,
+                                                       uint64_t v) {
+  opt->rep.writable_file_max_buffer_size = v;
+}
+
 void rocksdb_options_set_allow_concurrent_memtable_write(rocksdb_options_t* opt,
                                                          unsigned char v) {
   opt->rep.allow_concurrent_memtable_write = v;

db/db_impl.cc

@@ -574,6 +574,7 @@ Status DBImpl::SetDBOptions(
       env_options_for_compaction_ = env_->OptimizeForCompactionTableWrite(
                                             env_options_for_compaction_,
                                             immutable_db_options_);
+      versions_->ChangeEnvOptions(mutable_db_options_);
       write_thread_.EnterUnbatched(&w, &mutex_);
       if (total_log_size_ > GetMaxTotalWalSize() || wal_changed) {
         Status purge_wal_status = SwitchWAL(&write_context);

db/db_options_test.cc

@@ -208,6 +208,58 @@ TEST_F(DBOptionsTest, SetWalBytesPerSync) {
   ASSERT_GT(low_bytes_per_sync, counter);
 }
 
+TEST_F(DBOptionsTest, WritableFileMaxBufferSize) {
+  Options options;
+  options.create_if_missing = true;
+  options.writable_file_max_buffer_size = 1024 * 1024;
+  options.level0_file_num_compaction_trigger = 3;
+  options.max_manifest_file_size = 1;
+  options.env = env_;
+  int buffer_size = 1024 * 1024;
+  Reopen(options);
+  ASSERT_EQ(buffer_size,
+            dbfull()->GetDBOptions().writable_file_max_buffer_size);
+
+  std::atomic<int> match_cnt(0);
+  std::atomic<int> unmatch_cnt(0);
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "WritableFileWriter::WritableFileWriter:0", [&](void* arg) {
+        int value = static_cast<int>(reinterpret_cast<uintptr_t>(arg));
+        if (value == buffer_size) {
+          match_cnt++;
+        } else {
+          unmatch_cnt++;
+        }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  int i = 0;
+  for (; i < 3; i++) {
+    ASSERT_OK(Put("foo", ToString(i)));
+    ASSERT_OK(Put("bar", ToString(i)));
+    Flush();
+  }
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(unmatch_cnt, 0);
+  ASSERT_GE(match_cnt, 11);
+
+  buffer_size = 512 * 1024;
+  match_cnt = 0;
+  unmatch_cnt = 0;
+  ASSERT_OK(
+      dbfull()->SetDBOptions({{"writable_file_max_buffer_size", "524288"}}));
+  ASSERT_EQ(buffer_size,
+            dbfull()->GetDBOptions().writable_file_max_buffer_size);
+  i = 0;
+  for (; i < 3; i++) {
+    ASSERT_OK(Put("foo", ToString(i)));
+    ASSERT_OK(Put("bar", ToString(i)));
+    Flush();
+  }
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(unmatch_cnt, 0);
+  ASSERT_GE(match_cnt, 11);
+}
+
 TEST_F(DBOptionsTest, SetOptionsAndReopen) {
   Random rnd(1044);
   auto rand_opts = GetRandomizedMutableCFOptionsMap(&rnd);

db/version_set.cc

@@ -576,7 +576,7 @@ Status Version::GetTableProperties(std::shared_ptr<const TableProperties>* tp,
   auto table_cache = cfd_->table_cache();
   auto ioptions = cfd_->ioptions();
   Status s = table_cache->GetTableProperties(
-      vset_->env_options_, cfd_->internal_comparator(), file_meta->fd,
+      env_options_, cfd_->internal_comparator(), file_meta->fd,
       tp, true /* no io */);
   if (s.ok()) {
     return s;
@@ -599,7 +599,7 @@ Status Version::GetTableProperties(std::shared_ptr<const TableProperties>* tp,
       TableFileName(vset_->db_options_->db_paths, file_meta->fd.GetNumber(),
                     file_meta->fd.GetPathId());
   }
-  s = ioptions->env->NewRandomAccessFile(file_name, &file, vset_->env_options_);
+  s = ioptions->env->NewRandomAccessFile(file_name, &file, env_options_);
   if (!s.ok()) {
     return s;
   }
@@ -711,7 +711,7 @@ size_t Version::GetMemoryUsageByTableReaders() {
   for (auto& file_level : storage_info_.level_files_brief_) {
     for (size_t i = 0; i < file_level.num_files; i++) {
       total_usage += cfd_->table_cache()->GetMemoryUsageByTableReader(
-          vset_->env_options_, cfd_->internal_comparator(),
+          env_options_, cfd_->internal_comparator(),
           file_level.files[i].fd);
     }
   }
@@ -936,7 +936,7 @@ VersionStorageInfo::VersionStorageInfo(
 }
 
 Version::Version(ColumnFamilyData* column_family_data, VersionSet* vset,
-                 uint64_t version_number)
+                 const EnvOptions& env_opt, uint64_t version_number)
     : env_(vset->env_),
       cfd_(column_family_data),
       info_log_((cfd_ == nullptr) ? nullptr : cfd_->ioptions()->info_log),
@@ -959,6 +959,7 @@ Version::Version(ColumnFamilyData* column_family_data, VersionSet* vset,
       next_(this),
       prev_(this),
       refs_(0),
+      env_options_(env_opt),
       version_number_(version_number) {}
 
 void Version::Get(const ReadOptions& read_options, const LookupKey& k,
@@ -2532,7 +2533,8 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
     LogAndApplyCFHelper(w.edit_list.front());
     batch_edits.push_back(w.edit_list.front());
   } else {
-    v = new Version(column_family_data, this, current_version_number_++);
+    v = new Version(column_family_data, this, env_options_,
+                    current_version_number_++);
     builder_guard.reset(new BaseReferencedVersionBuilder(column_family_data));
     auto* builder = builder_guard->version_builder();
     for (const auto& writer : manifest_writers_) {
@@ -2577,7 +2579,7 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
   // Unlock during expensive operations. New writes cannot get here
   // because &w is ensuring that all new writes get queued.
   {
-
+    EnvOptions opt_env_opts = env_->OptimizeForManifestWrite(env_options_);
     mu->Unlock();
 
     TEST_SYNC_POINT("VersionSet::LogAndApply:WriteManifest");
@@ -2599,7 +2601,6 @@ Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
       ROCKS_LOG_INFO(db_options_->info_log, "Creating manifest %" PRIu64 "\n",
                      pending_manifest_file_number_);
       unique_ptr<WritableFile> descriptor_file;
-      EnvOptions opt_env_opts = env_->OptimizeForManifestWrite(env_options_);
       s = NewWritableFile(
           env_, DescriptorFileName(dbname_, pending_manifest_file_number_),
           &descriptor_file, opt_env_opts);
@@ -3064,7 +3065,8 @@ Status VersionSet::Recover(
             false /* prefetch_index_and_filter_in_cache */);
       }
 
-      Version* v = new Version(cfd, this, current_version_number_++);
+      Version* v =
+          new Version(cfd, this, env_options_, current_version_number_++);
       builder->SaveTo(v->storage_info());
 
       // Install recovered version
@@ -3422,7 +3424,8 @@ Status VersionSet::DumpManifest(Options& options, std::string& dscname,
       assert(builders_iter != builders.end());
       auto builder = builders_iter->second->version_builder();
 
-      Version* v = new Version(cfd, this, current_version_number_++);
+      Version* v =
+          new Version(cfd, this, env_options_, current_version_number_++);
       builder->SaveTo(v->storage_info());
       v->PrepareApply(*cfd->GetLatestMutableCFOptions(), false);
 
@@ -3634,7 +3637,7 @@ uint64_t VersionSet::ApproximateSize(Version* v, const FdWithKeyRange& f,
     // approximate offset of "key" within the table.
     TableReader* table_reader_ptr;
     InternalIterator* iter = v->cfd_->table_cache()->NewIterator(
-        ReadOptions(), env_options_, v->cfd_->internal_comparator(), f.fd,
+        ReadOptions(), v->env_options_, v->cfd_->internal_comparator(), f.fd,
         nullptr /* range_del_agg */, &table_reader_ptr);
     if (table_reader_ptr != nullptr) {
       result = table_reader_ptr->ApproximateOffsetOf(key);
@@ -3865,15 +3868,16 @@ ColumnFamilyData* VersionSet::CreateColumnFamily(
     const ColumnFamilyOptions& cf_options, VersionEdit* edit) {
   assert(edit->is_column_family_add_);
 
-  Version* dummy_versions = new Version(nullptr, this);
+  Version* dummy_versions = new Version(nullptr, this, env_options_);
   // Ref() dummy version once so that later we can call Unref() to delete it
   // by avoiding calling "delete" explicitly (~Version is private)
   dummy_versions->Ref();
   auto new_cfd = column_family_set_->CreateColumnFamily(
       edit->column_family_name_, edit->column_family_, dummy_versions,
       cf_options);
 
-  Version* v = new Version(new_cfd, this, current_version_number_++);
+  Version* v =
+      new Version(new_cfd, this, env_options_, current_version_number_++);
 
   // Fill level target base information.
   v->storage_info()->CalculateBaseBytes(*new_cfd->ioptions(),

db/version_set.h

@@ -663,12 +663,14 @@ class Version {
   Version* next_;               // Next version in linked list
   Version* prev_;               // Previous version in linked list
   int refs_;                    // Number of live refs to this version
+  const EnvOptions env_options_;
 
   // A version number that uniquely represents this version. This is
   // used for debugging and logging purposes only.
   uint64_t version_number_;
 
-  Version(ColumnFamilyData* cfd, VersionSet* vset, uint64_t version_number = 0);
+  Version(ColumnFamilyData* cfd, VersionSet* vset, const EnvOptions& env_opt,
+          uint64_t version_number = 0);
 
   ~Version();
 
@@ -844,6 +846,10 @@ class VersionSet {
 
   ColumnFamilySet* GetColumnFamilySet() { return column_family_set_.get(); }
   const EnvOptions& env_options() { return env_options_; }
+  void ChangeEnvOptions(const MutableDBOptions& new_options) {
+    env_options_.writable_file_max_buffer_size =
+        new_options.writable_file_max_buffer_size;
+  }
 
   static uint64_t GetNumLiveVersions(Version* dummy_versions);
 
@@ -908,7 +914,7 @@ class VersionSet {
   std::vector<std::string> obsolete_manifests_;
 
   // env options for all reads and writes except compactions
-  const EnvOptions& env_options_;
+  EnvOptions env_options_;
 
   // env options used for compactions. This is a copy of
   // env_options_ but with readaheads set to readahead_compactions_.

env/env.cc

@@ -333,6 +333,8 @@ EnvOptions Env::OptimizeForLogWrite(const EnvOptions& env_options,
                                     const DBOptions& db_options) const {
   EnvOptions optimized_env_options(env_options);
   optimized_env_options.bytes_per_sync = db_options.wal_bytes_per_sync;
+  optimized_env_options.writable_file_max_buffer_size =
+      db_options.writable_file_max_buffer_size;
   return optimized_env_options;
 }
 

env/env_posix.cc

@@ -832,6 +832,8 @@ class PosixEnv : public Env {
     // breaks TransactionLogIteratorStallAtLastRecord unit test. Fix the unit
     // test and make this false
     optimized.fallocate_with_keep_size = true;
+    optimized.writable_file_max_buffer_size =
+        db_options.writable_file_max_buffer_size;
     return optimized;
   }
 

include/rocksdb/c.h

@@ -857,6 +857,8 @@ extern ROCKSDB_LIBRARY_API void rocksdb_options_set_bytes_per_sync(
 extern ROCKSDB_LIBRARY_API void rocksdb_options_set_wal_bytes_per_sync(
         rocksdb_options_t*, uint64_t);
 extern ROCKSDB_LIBRARY_API void
+rocksdb_options_set_writable_file_max_buffer_size(rocksdb_options_t*, uint64_t);
+extern ROCKSDB_LIBRARY_API void
 rocksdb_options_set_allow_concurrent_memtable_write(rocksdb_options_t*,
                                                     unsigned char);
 extern ROCKSDB_LIBRARY_API void

options/db_options.cc

@@ -64,7 +64,6 @@ ImmutableDBOptions::ImmutableDBOptions(const DBOptions& options)
           options.new_table_reader_for_compaction_inputs),
       compaction_readahead_size(options.compaction_readahead_size),
       random_access_max_buffer_size(options.random_access_max_buffer_size),
-      writable_file_max_buffer_size(options.writable_file_max_buffer_size),
       use_adaptive_mutex(options.use_adaptive_mutex),
       listeners(options.listeners),
       enable_thread_tracking(options.enable_thread_tracking),
@@ -175,9 +174,6 @@ void ImmutableDBOptions::Dump(Logger* log) const {
   ROCKS_LOG_HEADER(
       log, "          Options.random_access_max_buffer_size: %" ROCKSDB_PRIszt,
       random_access_max_buffer_size);
-  ROCKS_LOG_HEADER(
-      log, "          Options.writable_file_max_buffer_size: %" ROCKSDB_PRIszt,
-      writable_file_max_buffer_size);
   ROCKS_LOG_HEADER(log, "                     Options.use_adaptive_mutex: %d",
                    use_adaptive_mutex);
   ROCKS_LOG_HEADER(log, "                           Options.rate_limiter: %p",
@@ -230,6 +226,7 @@ MutableDBOptions::MutableDBOptions()
       base_background_compactions(-1),
       max_background_compactions(-1),
       avoid_flush_during_shutdown(false),
+      writable_file_max_buffer_size(1024 * 1024),
       delayed_write_rate(2 * 1024U * 1024U),
       max_total_wal_size(0),
       delete_obsolete_files_period_micros(6ULL * 60 * 60 * 1000000),
@@ -243,6 +240,7 @@ MutableDBOptions::MutableDBOptions(const DBOptions& options)
       base_background_compactions(options.base_background_compactions),
       max_background_compactions(options.max_background_compactions),
       avoid_flush_during_shutdown(options.avoid_flush_during_shutdown),
+      writable_file_max_buffer_size(options.writable_file_max_buffer_size),
       delayed_write_rate(options.delayed_write_rate),
       max_total_wal_size(options.max_total_wal_size),
       delete_obsolete_files_period_micros(
@@ -259,6 +257,9 @@ void MutableDBOptions::Dump(Logger* log) const {
                    max_background_compactions);
   ROCKS_LOG_HEADER(log, "            Options.avoid_flush_during_shutdown: %d",
                    avoid_flush_during_shutdown);
+  ROCKS_LOG_HEADER(
+      log, "          Options.writable_file_max_buffer_size: %" ROCKSDB_PRIszt,
+      writable_file_max_buffer_size);
   ROCKS_LOG_HEADER(log, "            Options.delayed_write_rate : %" PRIu64,
                    delayed_write_rate);
   ROCKS_LOG_HEADER(log, "            Options.max_total_wal_size: %" PRIu64,

options/db_options.h

@@ -57,7 +57,6 @@ struct ImmutableDBOptions {
   bool new_table_reader_for_compaction_inputs;
   size_t compaction_readahead_size;
   size_t random_access_max_buffer_size;
-  size_t writable_file_max_buffer_size;
   bool use_adaptive_mutex;
   std::vector<std::shared_ptr<EventListener>> listeners;
   bool enable_thread_tracking;
@@ -93,6 +92,7 @@ struct MutableDBOptions {
   int base_background_compactions;
   int max_background_compactions;
   bool avoid_flush_during_shutdown;
+  size_t writable_file_max_buffer_size;
   uint64_t delayed_write_rate;
   uint64_t max_total_wal_size;
   uint64_t delete_obsolete_files_period_micros;

options/options_helper.cc

@@ -91,7 +91,7 @@ DBOptions BuildDBOptions(const ImmutableDBOptions& immutable_db_options,
   options.random_access_max_buffer_size =
       immutable_db_options.random_access_max_buffer_size;
   options.writable_file_max_buffer_size =
-      immutable_db_options.writable_file_max_buffer_size;
+      mutable_db_options.writable_file_max_buffer_size;
   options.use_adaptive_mutex = immutable_db_options.use_adaptive_mutex;
   options.listeners = immutable_db_options.listeners;
   options.enable_thread_tracking = immutable_db_options.enable_thread_tracking;

options/options_helper.h

@@ -220,9 +220,6 @@ static std::unordered_map<std::string, OptionTypeInfo> db_options_type_info = {
     {"random_access_max_buffer_size",
      {offsetof(struct DBOptions, random_access_max_buffer_size),
       OptionType::kSizeT, OptionVerificationType::kNormal, false, 0}},
-    {"writable_file_max_buffer_size",
-     {offsetof(struct DBOptions, writable_file_max_buffer_size),
-      OptionType::kSizeT, OptionVerificationType::kNormal, false, 0}},
     {"use_adaptive_mutex",
      {offsetof(struct DBOptions, use_adaptive_mutex), OptionType::kBoolean,
       OptionVerificationType::kNormal, false, 0}},
@@ -351,6 +348,10 @@ static std::unordered_map<std::string, OptionTypeInfo> db_options_type_info = {
      {offsetof(struct DBOptions, avoid_flush_during_shutdown),
       OptionType::kBoolean, OptionVerificationType::kNormal, true,
       offsetof(struct MutableDBOptions, avoid_flush_during_shutdown)}},
+    {"writable_file_max_buffer_size",
+     {offsetof(struct DBOptions, writable_file_max_buffer_size),
+      OptionType::kSizeT, OptionVerificationType::kNormal, true,
+      offsetof(struct MutableDBOptions, writable_file_max_buffer_size)}},
     {"allow_ingest_behind",
      {offsetof(struct DBOptions, allow_ingest_behind), OptionType::kBoolean,
       OptionVerificationType::kNormal, false,

port/win/env_win.cc

@@ -786,6 +786,8 @@ EnvOptions WinEnvIO::OptimizeForLogWrite(const EnvOptions& env_options,
   // breaks TransactionLogIteratorStallAtLastRecord unit test. Fix the unit
   // test and make this false
   optimized.fallocate_with_keep_size = true;
+  optimized.writable_file_max_buffer_size =
+      db_options.writable_file_max_buffer_size;
   return optimized;
 }
 

util/file_reader_writer.h

@@ -13,6 +13,7 @@
 #include "rocksdb/env.h"
 #include "rocksdb/rate_limiter.h"
 #include "util/aligned_buffer.h"
+#include "util/sync_point.h"
 
 namespace rocksdb {
 
@@ -151,6 +152,8 @@ class WritableFileWriter {
         bytes_per_sync_(options.bytes_per_sync),
         rate_limiter_(options.rate_limiter),
         stats_(stats) {
+    TEST_SYNC_POINT_CALLBACK("WritableFileWriter::WritableFileWriter:0",
+                             reinterpret_cast<void*>(max_buffer_size_));
     buf_.Alignment(writable_file_->GetRequiredBufferAlignment());
     buf_.AllocateNewBuffer(std::min((size_t)65536, max_buffer_size_));
   }