New API for background work in single thread pool

Summary:
Previously users could set `max_background_flushes=0` to force rocksdb to use a single thread pool for both background flushes and compactions. That'll no longer be possible since I'm going to deprecate `max_background_flushes` and `max_background_compactions` in favor of a single option. This diff introduces a new way to force a single thread pool: when high-pri pool has zero threads, all background jobs will be submitted to low-pri pool.

Note the majority of the code change is adding `Env::GetBackgroundThreads()`, which is necessary to check whether the user has provided a zero-sized thread pool.
Closes https://github.com/facebook/rocksdb/pull/2204

Differential Revision: D4936256

Pulled By: ajkr

fbshipit-source-id: 929a07a0c0705f7766f5339cd013ff74e90d6e01

HISTORY.md

@@ -1,5 +1,8 @@
 # Rocksdb Change Log
 ## Unreleased
+### Public API Change
+*Scheduling flushes and compactions in the same thread pool is no longer supported by setting `max_background_flushes=0`. Instead, users can achieve this by configuring their high-pri thread pool to have zero threads.
+
 ### New Features
 * Change ticker/histogram statistics implementations to use core-local storage. This improves aggregation speed compared to our previous thread-local approach, particularly for applications with many threads.
 

db/compaction_job_stats_test.cc

@@ -656,7 +656,6 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
   Options options;
   options.listeners.emplace_back(stats_checker);
   options.create_if_missing = true;
-  options.max_background_flushes = 0;
   // just enough setting to hold off auto-compaction.
   options.level0_file_num_compaction_trigger = kTestScale + 1;
   options.num_levels = 3;
@@ -878,7 +877,6 @@ TEST_P(CompactionJobStatsTest, DeletionStatsTest) {
   Options options;
   options.listeners.emplace_back(stats_checker);
   options.create_if_missing = true;
-  options.max_background_flushes = 0;
   options.level0_file_num_compaction_trigger = kTestScale+1;
   options.num_levels = 3;
   options.compression = kNoCompression;

db/db_compaction_test.cc

@@ -1977,7 +1977,6 @@ TEST_P(DBCompactionTestWithParam, ManualCompaction) {
 
     if (iter == 0) {
       options = CurrentOptions();
-      options.max_background_flushes = 0;
       options.num_levels = 3;
       options.create_if_missing = true;
       options.statistics = rocksdb::CreateDBStatistics();

db/db_flush_test.cc

@@ -91,6 +91,43 @@ TEST_F(DBFlushTest, SyncFail) {
   Destroy(options);
 }
 
+TEST_F(DBFlushTest, FlushInLowPriThreadPool) {
+  // Verify setting an empty high-pri (flush) thread pool causes flushes to be
+  // scheduled in the low-pri (compaction) thread pool.
+  Options options = CurrentOptions();
+  options.level0_file_num_compaction_trigger = 4;
+  options.memtable_factory.reset(new SpecialSkipListFactory(1));
+  Reopen(options);
+  env_->SetBackgroundThreads(0, Env::HIGH);
+
+  std::thread::id tid;
+  int num_flushes = 0, num_compactions = 0;
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BGWorkFlush", [&](void* arg) {
+        if (tid == std::thread::id()) {
+          tid = std::this_thread::get_id();
+        } else {
+          ASSERT_EQ(tid, std::this_thread::get_id());
+        }
+        ++num_flushes;
+      });
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BGWorkCompaction", [&](void* arg) {
+        ASSERT_EQ(tid, std::this_thread::get_id());
+        ++num_compactions;
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put("key", "val"));
+  for (int i = 0; i < 4; ++i) {
+    ASSERT_OK(Put("key", "val"));
+    dbfull()->TEST_WaitForFlushMemTable();
+  }
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(4, num_flushes);
+  ASSERT_EQ(1, num_compactions);
+}
+
 TEST_P(DBFlushDirectIOTest, DirectIO) {
   Options options;
   options.create_if_missing = true;

db/db_impl_compaction_flush.cc

@@ -983,8 +983,9 @@ void DBImpl::MaybeScheduleFlushOrCompaction() {
     // DB is being deleted; no more background compactions
     return;
   }
-
-  while (unscheduled_flushes_ > 0 &&
+  bool is_flush_pool_empty =
+    env_->GetBackgroundThreads(Env::Priority::HIGH) == 0;
+  while (!is_flush_pool_empty && unscheduled_flushes_ > 0 &&
          bg_flush_scheduled_ < immutable_db_options_.max_background_flushes) {
     unscheduled_flushes_--;
     bg_flush_scheduled_++;
@@ -993,9 +994,9 @@ void DBImpl::MaybeScheduleFlushOrCompaction() {
 
   auto bg_compactions_allowed = BGCompactionsAllowed();
 
-  // special case -- if max_background_flushes == 0, then schedule flush on a
-  // compaction thread
-  if (immutable_db_options_.max_background_flushes == 0) {
+  // special case -- if high-pri (flush) thread pool is empty, then schedule
+  // flushes in low-pri (compaction) thread pool.
+  if (is_flush_pool_empty) {
     while (unscheduled_flushes_ > 0 &&
            bg_flush_scheduled_ + bg_compaction_scheduled_ <
                bg_compactions_allowed) {

db/db_properties_test.cc

@@ -940,7 +940,6 @@ TEST_F(DBPropertiesTest, EstimateCompressionRatio) {
   Options options = CurrentOptions();
   options.compression_per_level = {kNoCompression, kSnappyCompression};
   options.disable_auto_compactions = true;
-  options.max_background_flushes = 0;
   options.num_levels = 2;
   Reopen(options);
 
@@ -1068,7 +1067,6 @@ class CountingDeleteTabPropCollectorFactory
 TEST_F(DBPropertiesTest, GetUserDefinedTableProperties) {
   Options options = CurrentOptions();
   options.level0_file_num_compaction_trigger = (1 << 30);
-  options.max_background_flushes = 0;
   options.table_properties_collector_factories.resize(1);
   std::shared_ptr<CountingUserTblPropCollectorFactory> collector_factory =
       std::make_shared<CountingUserTblPropCollectorFactory>(0);
@@ -1109,7 +1107,6 @@ TEST_F(DBPropertiesTest, GetUserDefinedTableProperties) {
 TEST_F(DBPropertiesTest, UserDefinedTablePropertiesContext) {
   Options options = CurrentOptions();
   options.level0_file_num_compaction_trigger = 3;
-  options.max_background_flushes = 0;
   options.table_properties_collector_factories.resize(1);
   std::shared_ptr<CountingUserTblPropCollectorFactory> collector_factory =
       std::make_shared<CountingUserTblPropCollectorFactory>(1);

db/db_test.cc

@@ -1441,7 +1441,6 @@ TEST_F(DBTest, UnremovableSingleDelete) {
 #ifndef ROCKSDB_LITE
 TEST_F(DBTest, DeletionMarkers1) {
   Options options = CurrentOptions();
-  options.max_background_flushes = 0;
   CreateAndReopenWithCF({"pikachu"}, options);
   Put(1, "foo", "v1");
   ASSERT_OK(Flush(1));
@@ -3075,9 +3074,9 @@ TEST_F(DBTest, DynamicMemtableOptions) {
                  Env::Priority::LOW);
   // Start from scratch and disable compaction/flush. Flush can only happen
   // during compaction but trigger is pretty high
-  options.max_background_flushes = 0;
   options.disable_auto_compactions = true;
   DestroyAndReopen(options);
+  env_->SetBackgroundThreads(0, Env::HIGH);
 
   // Put until writes are stopped, bounded by 256 puts. We should see stop at
   // ~128KB
@@ -3354,7 +3353,6 @@ TEST_P(DBTestWithParam, ThreadStatusSingleCompaction) {
 
 TEST_P(DBTestWithParam, PreShutdownManualCompaction) {
   Options options = CurrentOptions();
-  options.max_background_flushes = 0;
   options.max_subcompactions = max_subcompactions_;
   CreateAndReopenWithCF({"pikachu"}, options);
 
@@ -3393,7 +3391,6 @@ TEST_P(DBTestWithParam, PreShutdownManualCompaction) {
 
     if (iter == 0) {
       options = CurrentOptions();
-      options.max_background_flushes = 0;
       options.num_levels = 3;
       options.create_if_missing = true;
       DestroyAndReopen(options);
@@ -3404,7 +3401,6 @@ TEST_P(DBTestWithParam, PreShutdownManualCompaction) {
 
 TEST_F(DBTest, PreShutdownFlush) {
   Options options = CurrentOptions();
-  options.max_background_flushes = 0;
   CreateAndReopenWithCF({"pikachu"}, options);
   ASSERT_OK(Put(1, "key", "value"));
   CancelAllBackgroundWork(db_);

env/env_posix.cc

@@ -763,6 +763,11 @@ class PosixEnv : public Env {
     thread_pools_[pri].SetBackgroundThreads(num);
   }
 
+  virtual int GetBackgroundThreads(Priority pri) override {
+    assert(pri >= Priority::LOW && pri <= Priority::HIGH);
+    return thread_pools_[pri].GetBackgroundThreads();
+  }
+
   // Allow increasing the number of worker threads.
   virtual void IncBackgroundThreadsIfNeeded(int num, Priority pri) override {
     assert(pri >= Priority::LOW && pri <= Priority::HIGH);

hdfs/env_hdfs.h

@@ -151,6 +151,10 @@ class HdfsEnv : public Env {
     posixEnv->SetBackgroundThreads(number, pri);
   }
 
+  virtual int GetBackgroundThreads(Priority pri = LOW) {
+    return posixEnv->GetBackgroundThreads(pri);
+  }
+
   virtual void IncBackgroundThreadsIfNeeded(int number, Priority pri) override {
     posixEnv->IncBackgroundThreadsIfNeeded(number, pri);
   }
@@ -358,6 +362,7 @@ class HdfsEnv : public Env {
   }
 
   virtual void SetBackgroundThreads(int number, Priority pri = LOW) override {}
+  virtual int GetBackgroundThreads(Priority pri = LOW) override { return 0; }
   virtual void IncBackgroundThreadsIfNeeded(int number, Priority pri) override {
   }
   virtual std::string TimeToString(uint64_t number) override { return ""; }

include/rocksdb/env.h

@@ -364,6 +364,7 @@ class Env {
   // for this environment. 'LOW' is the default pool.
   // default number: 1
   virtual void SetBackgroundThreads(int number, Priority pri = LOW) = 0;
+  virtual int GetBackgroundThreads(Priority pri = LOW) = 0;
 
   // Enlarge number of background worker threads of a specific thread pool
   // for this environment if it is smaller than specified. 'LOW' is the default
@@ -1027,6 +1028,9 @@ class EnvWrapper : public Env {
   void SetBackgroundThreads(int num, Priority pri) override {
     return target_->SetBackgroundThreads(num, pri);
   }
+  int GetBackgroundThreads(Priority pri) override {
+    return target_->GetBackgroundThreads(pri);
+  }
 
   void IncBackgroundThreadsIfNeeded(int num, Priority pri) override {
     return target_->IncBackgroundThreadsIfNeeded(num, pri);

include/rocksdb/threadpool.h

@@ -31,6 +31,7 @@ class ThreadPool {
   // Set the number of background threads that will be executing the
   // scheduled jobs.
   virtual void SetBackgroundThreads(int num) = 0;
+  virtual int GetBackgroundThreads() = 0;
 
   // Get the number of jobs scheduled in the ThreadPool queue.
   virtual unsigned int GetQueueLen() const = 0;

port/win/env_win.cc

@@ -879,6 +879,11 @@ void WinEnvThreads::SetBackgroundThreads(int num, Env::Priority pri) {
   thread_pools_[pri].SetBackgroundThreads(num);
 }
 
+int WinEnvThreads::GetBackgroundThreads(Env::Priority pri) {
+  assert(pri >= Env::Priority::LOW && pri <= Env::Priority::HIGH);
+  return thread_pools_[pri].GetBackgroundThreads();
+}
+
 void WinEnvThreads::IncBackgroundThreadsIfNeeded(int num, Env::Priority pri) {
   assert(pri >= Env::Priority::LOW && pri <= Env::Priority::HIGH);
   thread_pools_[pri].IncBackgroundThreadsIfNeeded(num);
@@ -1056,6 +1061,10 @@ void  WinEnv::SetBackgroundThreads(int num, Env::Priority pri) {
   return winenv_threads_.SetBackgroundThreads(num, pri);
 }
 
+int WinEnv::GetBackgroundThreads(Env::Priority pri) {
+  return winenv_threads_.GetBackgroundThreads(pri);
+}
+
 void  WinEnv::IncBackgroundThreadsIfNeeded(int num, Env::Priority pri) {
   return winenv_threads_.IncBackgroundThreadsIfNeeded(num, pri);
 }

port/win/env_win.h

@@ -66,6 +66,7 @@ public:
 
   // Allow increasing the number of worker threads.
   void SetBackgroundThreads(int num, Env::Priority pri);
+  int GetBackgroundThreads(Env::Priority pri);
 
   void IncBackgroundThreadsIfNeeded(int num, Env::Priority pri);
 
@@ -276,6 +277,7 @@ public:
 
   // Allow increasing the number of worker threads.
   void SetBackgroundThreads(int num, Env::Priority pri) override;
+  int GetBackgroundThreads(Env::Priority pri) override;
 
   void IncBackgroundThreadsIfNeeded(int num, Env::Priority pri) override;
 

util/threadpool_imp.cc

@@ -47,6 +47,7 @@ struct ThreadPoolImpl::Impl {
   void JoinThreads(bool wait_for_jobs_to_complete);
 
   void SetBackgroundThreadsInternal(int num, bool allow_reduce);
+  int GetBackgroundThreads();
 
   unsigned int GetQueueLen() const {
     return queue_len_.load(std::memory_order_relaxed);
@@ -275,12 +276,17 @@ void ThreadPoolImpl::Impl::SetBackgroundThreadsInternal(int num,
   }
   if (num > total_threads_limit_ ||
       (num < total_threads_limit_ && allow_reduce)) {
-    total_threads_limit_ = std::max(1, num);
+    total_threads_limit_ = std::max(0, num);
     WakeUpAllThreads();
     StartBGThreads();
   }
 }
 
+int ThreadPoolImpl::Impl::GetBackgroundThreads() {
+  std::unique_lock<std::mutex> lock(mu_);
+  return total_threads_limit_;
+}
+
 void ThreadPoolImpl::Impl::StartBGThreads() {
   // Start background thread if necessary
   while ((int)bgthreads_.size() < total_threads_limit_) {
@@ -384,6 +390,10 @@ void ThreadPoolImpl::SetBackgroundThreads(int num) {
   impl_->SetBackgroundThreadsInternal(num, true);
 }
 
+int ThreadPoolImpl::GetBackgroundThreads() {
+  return impl_->GetBackgroundThreads();
+}
+
 unsigned int ThreadPoolImpl::GetQueueLen() const {
   return impl_->GetQueueLen();
 }

util/threadpool_imp.h

@@ -38,6 +38,8 @@ class ThreadPoolImpl : public ThreadPool {
   // Set the number of background threads that will be executing the
   // scheduled jobs.
   void SetBackgroundThreads(int num) override;
+  int GetBackgroundThreads() override;
+
   // Get the number of jobs scheduled in the ThreadPool queue.
   unsigned int GetQueueLen() const override;