Perf Level By Bit Field

CMakeLists.txt

@@ -189,7 +189,7 @@ if(MSVC)
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /W4 /wd4127 /wd4800 /wd4996 /wd4351 /wd4100 /wd4204 /wd4324")
 else()
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -W -Wextra -Wall")
- set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wsign-compare -Wshadow -Wno-unused-parameter -Wno-unused-variable -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers -Wno-strict-aliasing")
+ set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wsign-compare -Wshadow -Wno-unused-parameter -Wno-unused-variable -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers -Wno-strict-aliasing -w")
  if(MINGW)
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-format")
  add_definitions(-D_POSIX_C_SOURCE=1)
@@ -584,6 +584,7 @@ set(SOURCES
  monitoring/instrumented_mutex.cc
  monitoring/iostats_context.cc
  monitoring/perf_context.cc
+ monitoring/perf_flags.cc
  monitoring/perf_level.cc
  monitoring/persistent_stats_history.cc
  monitoring/statistics.cc

db/c.cc

@@ -47,6 +47,9 @@
 #include <unordered_set>
 #include <map>
 
+using rocksdb::PerfContext;
+using rocksdb::PerfLevel;
+using rocksdb::PerfFlags;
 using rocksdb::BytewiseComparator;
 using rocksdb::Cache;
 using rocksdb::ColumnFamilyDescriptor;
@@ -113,8 +116,9 @@ using rocksdb::Transaction;
 using rocksdb::Checkpoint;
 using rocksdb::TransactionLogIterator;
 using rocksdb::BatchResult;
-using rocksdb::PerfLevel;
-using rocksdb::PerfContext;
+using rocksdb::WritableFile;
+using rocksdb::WriteBatch;
+using rocksdb::WriteBatchWithIndex;
 using rocksdb::MemoryUtil;
 
 using std::shared_ptr;
@@ -2740,7 +2744,7 @@ rocksdb_ratelimiter_t* rocksdb_writeampbasedratelimiter_create(
  return rate_limiter;
 }
 
-void rocksdb_ratelimiter_destroy(rocksdb_ratelimiter_t *limiter) {
+void rocksdb_ratelimiter_destroy(rocksdb_ratelimiter_t* limiter) {
  delete limiter;
 }
 
@@ -2749,8 +2753,14 @@ void rocksdb_set_perf_level(int v) {
  SetPerfLevel(level);
 }
 
+void rocksdb_set_perf_flags_by_mask(uint8_t* flags_bytes) {
+ PerfFlags flags;
+ memmove(&flags,flags_bytes,sizeof(PerfFlags));
+ SetPerfFlags(flags);
+}
+
 rocksdb_perfcontext_t* rocksdb_perfcontext_create() {
- rocksdb_perfcontext_t* context = new rocksdb_perfcontext_t;
+ auto* context = new rocksdb_perfcontext_t;
  context->rep = rocksdb::get_perf_context();
  return context;
 }

db/compaction/compaction_job.cc

@@ -7,8 +7,10 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
-#include <cinttypes>
+#include "db/compaction/compaction_job.h"
+
 #include <algorithm>
+#include <cinttypes>
 #include <functional>
 #include <list>
 #include <memory>
@@ -19,7 +21,6 @@
 #include <vector>
 
 #include "db/builder.h"
-#include "db/compaction/compaction_job.h"
 #include "db/db_impl/db_impl.h"
 #include "db/db_iter.h"
 #include "db/dbformat.h"

db/db_impl/db_impl.cc

@@ -33,9 +33,9 @@
 #include "db/error_handler.h"
 #include "db/event_helpers.h"
 #include "db/external_sst_file_ingestion_job.h"
-#include "db/import_column_family_job.h"
 #include "db/flush_job.h"
 #include "db/forward_iterator.h"
+#include "db/import_column_family_job.h"
 #include "db/job_context.h"
 #include "db/log_reader.h"
 #include "db/log_writer.h"

db/db_impl/db_impl_compaction_flush.cc

@@ -6,11 +6,10 @@
 // Copyright (c) 2011 The LevelDB Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
-#include "db/db_impl/db_impl.h"
-
 #include <cinttypes>
 
 #include "db/builder.h"
+#include "db/db_impl/db_impl.h"
 #include "db/error_handler.h"
 #include "db/event_helpers.h"
 #include "file/sst_file_manager_impl.h"

db/db_iter.cc

@@ -8,9 +8,10 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/db_iter.h"
-#include <string>
+
 #include <iostream>
 #include <limits>
+#include <string>
 
 #include "db/dbformat.h"
 #include "db/merge_context.h"

db/db_iter_test.cc

@@ -420,7 +420,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
  db_iter->SeekToLast();
 
  ASSERT_TRUE(db_iter->Valid());
- ASSERT_EQ(static_cast<int>(get_perf_context()->internal_key_skipped_count), 1);
+ ASSERT_EQ(static_cast<int>(get_perf_context()->internal_key_skipped_count),
+ 1);
  ASSERT_EQ(db_iter->key().ToString(), "b");
 
  SetPerfLevel(kDisable);
@@ -543,7 +544,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
  db_iter->SeekToLast();
 
  ASSERT_TRUE(db_iter->Valid());
- ASSERT_EQ(static_cast<int>(get_perf_context()->internal_delete_skipped_count), 0);
+ ASSERT_EQ(
+ static_cast<int>(get_perf_context()->internal_delete_skipped_count), 0);
  ASSERT_EQ(db_iter->key().ToString(), "b");
 
  SetPerfLevel(kDisable);

db/dbformat.cc

@@ -8,8 +8,10 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 #include "db/dbformat.h"
 
-#include <cinttypes>
 #include <stdio.h>
+
+#include <cinttypes>
+
 #include "monitoring/perf_context_imp.h"
 #include "port/port.h"
 #include "util/coding.h"

db/dbformat.h

@@ -9,9 +9,11 @@
 
 #pragma once
 #include <stdio.h>
+
 #include <memory>
 #include <string>
 #include <utility>
+
 #include "db/lookup_key.h"
 #include "db/merge_context.h"
 #include "logging/logging.h"

db/flush_job.cc

@@ -9,9 +9,8 @@
 
 #include "db/flush_job.h"
 
-#include <cinttypes>
-
 #include <algorithm>
+#include <cinttypes>
 #include <vector>
 
 #include "db/builder.h"

db/perf_context_test.cc

@@ -3,6 +3,8 @@
 // COPYING file in the root directory) and Apache 2.0 License
 // (found in the LICENSE.Apache file in the root directory).
 //
+#include "rocksdb/perf_context.h"
+
 #include <algorithm>
 #include <iostream>
 #include <thread>
@@ -15,7 +17,6 @@
 #include "port/port.h"
 #include "rocksdb/db.h"
 #include "rocksdb/memtablerep.h"
-#include "rocksdb/perf_context.h"
 #include "rocksdb/slice_transform.h"
 #include "test_util/testharness.h"
 #include "util/stop_watch.h"

db/table_cache.cc

@@ -14,7 +14,6 @@
 #include "db/snapshot_impl.h"
 #include "db/version_edit.h"
 #include "file/filename.h"
-
 #include "monitoring/perf_context_imp.h"
 #include "rocksdb/statistics.h"
 #include "table/block_based/block_based_table_reader.h"

db/version_set.cc

@@ -10,6 +10,7 @@
 #include "db/version_set.h"
 
 #include <stdio.h>
+
 #include <algorithm>
 #include <array>
 #include <cinttypes>
@@ -19,6 +20,7 @@
 #include <string>
 #include <unordered_map>
 #include <vector>
+
 #include "compaction/compaction.h"
 #include "db/internal_stats.h"
 #include "db/log_reader.h"

include/rocksdb/c.h

@@ -1120,6 +1120,7 @@ enum {
 };
 
 extern ROCKSDB_LIBRARY_API void rocksdb_set_perf_level(int);
+extern ROCKSDB_LIBRARY_API void rocksdb_set_perf_flags_by_mask(uint8_t* flags_bytes);
 extern ROCKSDB_LIBRARY_API rocksdb_perfcontext_t* rocksdb_perfcontext_create();
 extern ROCKSDB_LIBRARY_API void rocksdb_perfcontext_reset(
  rocksdb_perfcontext_t* context);

include/rocksdb/perf_context.h

@@ -6,11 +6,12 @@
 #pragma once
 
 #include <stdint.h>
+
 #include <map>
 #include <string>
 
+#include "rocksdb/perf_flags.h"
 #include "rocksdb/perf_level.h"
-
 namespace rocksdb {
 
 // A thread local context for gathering performance counter efficiently

include/rocksdb/perf_flags.h

@@ -0,0 +1,110 @@
+#pragma once
+#include <stdint.h>
+
+namespace rocksdb {
+struct PerfFlags {
+ // represent original Level 2
+ // TODO: think a better design for context_by_level structure
+ uint8_t enable_perf_context_by_level_count_bit : 1; // 2
+ uint8_t enable_user_key_comparison_count_bit : 1; // 2
+ uint8_t enable_block_cache_hit_count_bit : 1; // 2
+ uint8_t enable_block_read_count_bit : 1; // 2
+ uint8_t enable_block_read_byte_bit : 1; // 2
+ uint8_t enable_block_cache_index_hit_count_bit : 1; // 2
+ uint8_t enable_index_block_read_count_bit : 1; // 2
+ uint8_t enable_block_cache_filter_hit_count_bit : 1; // 2
+ uint8_t enable_filter_block_read_count_bit : 1; // 2
+ uint8_t enable_compression_dict_block_read_count_bit : 1; // 2
+ uint8_t enable_get_read_bytes_bit : 1; // 2
+ uint8_t enable_multiget_read_bytes_bit : 1; // 2
+ uint8_t enable_iter_read_bytes_bit : 1; // 2
+ uint8_t enable_internal_key_skipped_count_bit : 1; // 2
+ uint8_t enable_internal_delete_skipped_count_bit : 1; // 2
+ uint8_t enable_internal_recent_skipped_count_bit : 1; // 2
+ uint8_t enable_internal_merge_count_bit : 1; // 2
+ uint8_t enable_get_from_memtable_count_bit : 1; // 2
+ uint8_t enable_seek_on_memtable_count_bit : 1; // 2
+ uint8_t enable_next_on_memtable_count_bit : 1; // 2
+ uint8_t enable_prev_on_memtable_count_bit : 1; // 2
+ uint8_t enable_seek_child_seek_count_bit : 1; // 2
+ uint8_t enable_bloom_memtable_hit_count_bit : 1; // 2
+ uint8_t enable_bloom_memtable_miss_count_bit : 1; // 2
+ uint8_t enable_bloom_sst_hit_count_bit : 1; // 2
+ uint8_t enable_bloom_sst_miss_count_bit : 1; // 2
+ uint8_t enable_key_lock_wait_count_bit : 1; // 2
+
+ // represent original Level 3
+ uint8_t enable_measure_cpu_time_bit : 1; // 3
+ uint8_t enable_block_read_time_bit : 1; // 3
+ uint8_t enable_block_checksum_time_bit : 1; // 3
+ uint8_t enable_block_decompress_time_bit : 1; // 3
+ uint8_t enable_get_snapshot_time_bit : 1; // 3
+ uint8_t enable_get_from_memtable_time_bit : 1; // 3
+ uint8_t enable_get_post_process_time_bit : 1; // 3
+ uint8_t enable_get_from_output_files_time_bit : 1; // 3
+ uint8_t enable_seek_on_memtable_time_bit : 1; // 3
+ uint8_t enable_seek_child_seek_time_bit : 1; // 3
+ uint8_t enable_seek_min_heap_time_bit : 1; // 3
+ uint8_t enable_seek_max_heap_time_bit : 1; // 3
+ uint8_t enable_seek_internal_seek_time_bit : 1; // 3
+ uint8_t enable_find_next_user_entry_time_bit : 1; // 3
+ uint8_t enable_write_wal_time_bit : 1; // 3
+ uint8_t enable_write_memtable_time_bit : 1; // 3
+ uint8_t enable_write_delay_time_bit : 1; // 3
+ uint8_t enable_write_scheduling_flushes_compactions_time_bit : 1; // 3
+ uint8_t enable_write_pre_and_post_process_time_bit : 1; // 3
+ uint8_t enable_write_thread_wait_nanos_bit : 1; // 3
+ uint8_t enable_merge_operator_time_nanos_bit : 1; // 3
+ uint8_t enable_read_index_block_nanos_bit : 1; // 3
+ uint8_t enable_read_filter_block_nanos_bit : 1; // 3
+ uint8_t enable_new_table_block_iter_nanos_bit : 1; // 3
+ uint8_t enable_new_table_iterator_nanos_bit : 1; // 3
+ uint8_t enable_block_seek_nanos_bit : 1; // 3
+ uint8_t enable_find_table_nanos_bit : 1; // 3
+ uint8_t enable_key_lock_wait_time_bit : 1; // 3
+ uint8_t enable_env_new_sequential_file_nanos_bit : 1; // 3
+ uint8_t enable_env_new_random_access_file_nanos_bit : 1; // 3
+ uint8_t enable_env_new_writable_file_nanos_bit : 1; // 3
+ uint8_t enable_env_reuse_writable_file_nanos_bit : 1; // 3
+ uint8_t enable_env_new_random_rw_file_nanos_bit : 1; // 3
+ uint8_t enable_env_new_directory_nanos_bit : 1; // 3
+ uint8_t enable_env_file_exists_nanos_bit : 1; // 3
+ uint8_t enable_env_get_children_nanos_bit : 1; // 3
+ uint8_t enable_env_get_children_file_attributes_nanos_bit : 1; // 3
+ uint8_t enable_env_delete_file_nanos_bit : 1; // 3
+ uint8_t enable_env_create_dir_nanos_bit : 1; // 3
+ uint8_t enable_env_create_dir_if_missing_nanos_bit : 1; // 3
+ uint8_t enable_env_delete_dir_nanos_bit : 1; // 3
+ uint8_t enable_env_get_file_size_nanos_bit : 1; // 3
+ uint8_t enable_env_get_file_modification_time_nanos_bit : 1; // 3
+ uint8_t enable_env_rename_file_nanos_bit : 1; // 3
+ uint8_t enable_env_link_file_nanos_bit : 1; // 3
+ uint8_t enable_env_lock_file_nanos_bit : 1; // 3
+ uint8_t enable_env_unlock_file_nanos_bit : 1; // 3
+ uint8_t enable_env_new_logger_nanos_bit : 1; // 3
+ uint8_t enable_encrypt_data_nanos_bit : 1; // 3
+ uint8_t enable_decrypt_data_nanos_bit : 1; // 3
+
+ // represent original Level 4
+ // TODO: think a better design for iostats
+ uint8_t enable_iostats_cpu_timer_bit : 1; // 4
+ uint8_t enable_get_cpu_nanos_bit : 1; // 4
+ uint8_t enable_iter_next_cpu_nanos_bit : 1; // 4
+ uint8_t enable_iter_prev_cpu_nanos_bit : 1; // 4
+ uint8_t enable_iter_seek_cpu_nanos_bit : 1; // 4
+
+ // represent original Level 5
+ uint8_t enable_db_mutex_lock_nanos_bit : 1; // 5
+ uint8_t enable_db_condition_wait_nanos_bit : 1; // 5
+};
+extern const PerfFlags PERF_LEVEL2;
+extern const PerfFlags PERF_LEVEL3;
+extern const PerfFlags PERF_LEVEL4;
+extern const PerfFlags PERF_LEVEL5;
+// set the perf flags for current thread
+void SetPerfFlags(PerfFlags pbf);
+
+// get current perf flags for current thread
+PerfFlags GetPerfFlags();
+
+} // namespace rocksdb

include/rocksdb/perf_level.h

@@ -21,13 +21,14 @@ enum PerfLevel : unsigned char {
  // time (neither wall time nor CPU time) for mutexes.
  kEnableTimeAndCPUTimeExceptForMutex = 4,
  kEnableTime = 5, // enable count and time stats
- kOutOfBounds = 6 // N.B. Must always be the last value!
+ kCustomFlags = 6, // could not measured by level because is a custom bit flags configuration
+ kOutOfBounds = 7, // N.B. Must always be the last value!
 };
 
 // set the perf stats level for current thread
 void SetPerfLevel(PerfLevel level);
 
-// get current perf stats level for current thread
+// get current estimated perf stats level for current thread
 PerfLevel GetPerfLevel();
 
 } // namespace rocksdb

monitoring/iostats_context_imp.h

@@ -4,6 +4,7 @@
 // (found in the LICENSE.Apache file in the root directory).
 //
 #pragma once
+#include "monitoring/perf_flags_imp.h"
 #include "monitoring/perf_step_timer.h"
 #include "rocksdb/iostats_context.h"
 
@@ -34,14 +35,13 @@ extern __thread IOStatsContext iostats_context;
 
 // Declare and set start time of the timer
 #define IOSTATS_TIMER_GUARD(metric) \
- PerfStepTimer iostats_step_timer_##metric(&(iostats_context.metric)); \
+ PerfStepTimer iostats_step_timer_##metric(&(iostats_context.metric),(bool)perf_flags.enable_iostats_cpu_timer_bit); \
  iostats_step_timer_##metric.Start();
 
 // Declare and set start time of the timer
 #define IOSTATS_CPU_TIMER_GUARD(metric, env) \
  PerfStepTimer iostats_step_timer_##metric( \
- &(iostats_context.metric), env, true, \
- PerfLevel::kEnableTimeAndCPUTimeExceptForMutex); \
+ &(iostats_context.metric),(bool)perf_flags.enable_iostats_cpu_timer_bit, env, true); \
  iostats_step_timer_##metric.Start();
 
 #else // ROCKSDB_SUPPORT_THREAD_LOCAL