internal_stats.cc 72.8 KB
Newer Older
S
Siying Dong 已提交
1 2 3
//  This source code is licensed under both the GPLv2 (found in the
//  COPYING file in the root directory) and Apache 2.0 License
//  (found in the LICENSE.Apache file in the root directory).
I
Igor Canadi 已提交
4
//
5 6
// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
//
I
Igor Canadi 已提交
7 8 9 10 11
// 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/internal_stats.h"
L
liuhuahang 已提交
12

13
#include <algorithm>
14
#include <cinttypes>
15
#include <cstddef>
Y
Yi Wu 已提交
16
#include <limits>
17
#include <sstream>
Y
Yi Wu 已提交
18
#include <string>
19
#include <utility>
I
Igor Canadi 已提交
20
#include <vector>
21

22 23
#include "cache/cache_entry_roles.h"
#include "cache/cache_entry_stats.h"
Y
Yi Wu 已提交
24
#include "db/column_family.h"
25
#include "db/db_impl/db_impl.h"
26
#include "port/port.h"
27
#include "rocksdb/system_clock.h"
28
#include "rocksdb/table.h"
29
#include "table/block_based/cachable_entry.h"
30
#include "util/string_util.h"
I
Igor Canadi 已提交
31

32
namespace ROCKSDB_NAMESPACE {
I
Igor Canadi 已提交
33

34
#ifndef ROCKSDB_LITE
35 36 37 38 39 40

const std::map<LevelStatType, LevelStat> InternalStats::compaction_level_stats =
    {
        {LevelStatType::NUM_FILES, LevelStat{"NumFiles", "Files"}},
        {LevelStatType::COMPACTED_FILES,
         LevelStat{"CompactedFiles", "CompactedFiles"}},
41
        {LevelStatType::SIZE_BYTES, LevelStat{"SizeBytes", "Size"}},
42
        {LevelStatType::SCORE, LevelStat{"Score", "Score"}},
43 44 45 46 47 48
        {LevelStatType::READ_GB, LevelStat{"ReadGB", "Read(GB)"}},
        {LevelStatType::RN_GB, LevelStat{"RnGB", "Rn(GB)"}},
        {LevelStatType::RNP1_GB, LevelStat{"Rnp1GB", "Rnp1(GB)"}},
        {LevelStatType::WRITE_GB, LevelStat{"WriteGB", "Write(GB)"}},
        {LevelStatType::W_NEW_GB, LevelStat{"WnewGB", "Wnew(GB)"}},
        {LevelStatType::MOVED_GB, LevelStat{"MovedGB", "Moved(GB)"}},
49
        {LevelStatType::WRITE_AMP, LevelStat{"WriteAmp", "W-Amp"}},
50 51 52
        {LevelStatType::READ_MBPS, LevelStat{"ReadMBps", "Rd(MB/s)"}},
        {LevelStatType::WRITE_MBPS, LevelStat{"WriteMBps", "Wr(MB/s)"}},
        {LevelStatType::COMP_SEC, LevelStat{"CompSec", "Comp(sec)"}},
53 54
        {LevelStatType::COMP_CPU_SEC,
         LevelStat{"CompMergeCPU", "CompMergeCPU(sec)"}},
55 56
        {LevelStatType::COMP_COUNT, LevelStat{"CompCount", "Comp(cnt)"}},
        {LevelStatType::AVG_SEC, LevelStat{"AvgSec", "Avg(sec)"}},
57 58
        {LevelStatType::KEY_IN, LevelStat{"KeyIn", "KeyIn"}},
        {LevelStatType::KEY_DROP, LevelStat{"KeyDrop", "KeyDrop"}},
59
        {LevelStatType::R_BLOB_GB, LevelStat{"RblobGB", "Rblob(GB)"}},
60
        {LevelStatType::W_BLOB_GB, LevelStat{"WblobGB", "Wblob(GB)"}},
61 62
};

L
Lei Jin 已提交
63 64 65
namespace {
const double kMB = 1048576.0;
const double kGB = kMB * 1024;
66
const double kMicrosInSec = 1000000.0;
L
Lei Jin 已提交
67

68 69
void PrintLevelStatsHeader(char* buf, size_t len, const std::string& cf_name,
                           const std::string& group_by) {
70 71
  int written_size =
      snprintf(buf, len, "\n** Compaction Stats [%s] **\n", cf_name.c_str());
72
  written_size = std::min(written_size, static_cast<int>(len));
73 74 75 76 77
  auto hdr = [](LevelStatType t) {
    return InternalStats::compaction_level_stats.at(t).header_name.c_str();
  };
  int line_size = snprintf(
      buf + written_size, len - written_size,
78 79
      "%s    %s   %s     %s %s  %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s "
      "%s\n",
80
      // Note that we skip COMPACTED_FILES and merge it with Files column
81 82 83 84 85 86 87 88 89
      group_by.c_str(), hdr(LevelStatType::NUM_FILES),
      hdr(LevelStatType::SIZE_BYTES), hdr(LevelStatType::SCORE),
      hdr(LevelStatType::READ_GB), hdr(LevelStatType::RN_GB),
      hdr(LevelStatType::RNP1_GB), hdr(LevelStatType::WRITE_GB),
      hdr(LevelStatType::W_NEW_GB), hdr(LevelStatType::MOVED_GB),
      hdr(LevelStatType::WRITE_AMP), hdr(LevelStatType::READ_MBPS),
      hdr(LevelStatType::WRITE_MBPS), hdr(LevelStatType::COMP_SEC),
      hdr(LevelStatType::COMP_CPU_SEC), hdr(LevelStatType::COMP_COUNT),
      hdr(LevelStatType::AVG_SEC), hdr(LevelStatType::KEY_IN),
90 91
      hdr(LevelStatType::KEY_DROP), hdr(LevelStatType::R_BLOB_GB),
      hdr(LevelStatType::W_BLOB_GB));
92 93

  written_size += line_size;
94
  written_size = std::min(written_size, static_cast<int>(len));
95 96
  snprintf(buf + written_size, len - written_size, "%s\n",
           std::string(line_size, '-').c_str());
L
Lei Jin 已提交
97 98
}

99 100 101 102
void PrepareLevelStats(std::map<LevelStatType, double>* level_stats,
                       int num_files, int being_compacted,
                       double total_file_size, double score, double w_amp,
                       const InternalStats::CompactionStats& stats) {
103 104 105
  const uint64_t bytes_read = stats.bytes_read_non_output_levels +
                              stats.bytes_read_output_level +
                              stats.bytes_read_blob;
106 107 108
  const uint64_t bytes_written = stats.bytes_written + stats.bytes_written_blob;
  const int64_t bytes_new = stats.bytes_written - stats.bytes_read_output_level;
  const double elapsed = (stats.micros + 1) / kMicrosInSec;
L
Lei Jin 已提交
109

110 111
  (*level_stats)[LevelStatType::NUM_FILES] = num_files;
  (*level_stats)[LevelStatType::COMPACTED_FILES] = being_compacted;
112
  (*level_stats)[LevelStatType::SIZE_BYTES] = total_file_size;
113 114 115 116 117 118 119 120 121 122
  (*level_stats)[LevelStatType::SCORE] = score;
  (*level_stats)[LevelStatType::READ_GB] = bytes_read / kGB;
  (*level_stats)[LevelStatType::RN_GB] =
      stats.bytes_read_non_output_levels / kGB;
  (*level_stats)[LevelStatType::RNP1_GB] = stats.bytes_read_output_level / kGB;
  (*level_stats)[LevelStatType::WRITE_GB] = stats.bytes_written / kGB;
  (*level_stats)[LevelStatType::W_NEW_GB] = bytes_new / kGB;
  (*level_stats)[LevelStatType::MOVED_GB] = stats.bytes_moved / kGB;
  (*level_stats)[LevelStatType::WRITE_AMP] = w_amp;
  (*level_stats)[LevelStatType::READ_MBPS] = bytes_read / kMB / elapsed;
123
  (*level_stats)[LevelStatType::WRITE_MBPS] = bytes_written / kMB / elapsed;
124
  (*level_stats)[LevelStatType::COMP_SEC] = stats.micros / kMicrosInSec;
125
  (*level_stats)[LevelStatType::COMP_CPU_SEC] = stats.cpu_micros / kMicrosInSec;
126 127 128 129 130 131 132
  (*level_stats)[LevelStatType::COMP_COUNT] = stats.count;
  (*level_stats)[LevelStatType::AVG_SEC] =
      stats.count == 0 ? 0 : stats.micros / kMicrosInSec / stats.count;
  (*level_stats)[LevelStatType::KEY_IN] =
      static_cast<double>(stats.num_input_records);
  (*level_stats)[LevelStatType::KEY_DROP] =
      static_cast<double>(stats.num_dropped_records);
133
  (*level_stats)[LevelStatType::R_BLOB_GB] = stats.bytes_read_blob / kGB;
134
  (*level_stats)[LevelStatType::W_BLOB_GB] = stats.bytes_written_blob / kGB;
135 136 137 138
}

void PrintLevelStats(char* buf, size_t len, const std::string& name,
                     const std::map<LevelStatType, double>& stat_value) {
139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
  snprintf(
      buf, len,
      "%4s "      /*  Level */
      "%6d/%-3d " /*  Files */
      "%8s "      /*  Size */
      "%5.1f "    /*  Score */
      "%8.1f "    /*  Read(GB) */
      "%7.1f "    /*  Rn(GB) */
      "%8.1f "    /*  Rnp1(GB) */
      "%9.1f "    /*  Write(GB) */
      "%8.1f "    /*  Wnew(GB) */
      "%9.1f "    /*  Moved(GB) */
      "%5.1f "    /*  W-Amp */
      "%8.1f "    /*  Rd(MB/s) */
      "%8.1f "    /*  Wr(MB/s) */
154 155
      "%9.2f "    /*  Comp(sec) */
      "%17.2f "   /*  CompMergeCPU(sec) */
156 157 158
      "%9d "      /*  Comp(cnt) */
      "%8.3f "    /*  Avg(sec) */
      "%7s "      /*  KeyIn */
159
      "%6s "      /*  KeyDrop */
160
      "%9.1f "    /*  Rblob(GB) */
161
      "%9.1f\n",  /*  Wblob(GB) */
162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
      name.c_str(), static_cast<int>(stat_value.at(LevelStatType::NUM_FILES)),
      static_cast<int>(stat_value.at(LevelStatType::COMPACTED_FILES)),
      BytesToHumanString(
          static_cast<uint64_t>(stat_value.at(LevelStatType::SIZE_BYTES)))
          .c_str(),
      stat_value.at(LevelStatType::SCORE),
      stat_value.at(LevelStatType::READ_GB),
      stat_value.at(LevelStatType::RN_GB),
      stat_value.at(LevelStatType::RNP1_GB),
      stat_value.at(LevelStatType::WRITE_GB),
      stat_value.at(LevelStatType::W_NEW_GB),
      stat_value.at(LevelStatType::MOVED_GB),
      stat_value.at(LevelStatType::WRITE_AMP),
      stat_value.at(LevelStatType::READ_MBPS),
      stat_value.at(LevelStatType::WRITE_MBPS),
      stat_value.at(LevelStatType::COMP_SEC),
178
      stat_value.at(LevelStatType::COMP_CPU_SEC),
179 180 181 182 183 184 185
      static_cast<int>(stat_value.at(LevelStatType::COMP_COUNT)),
      stat_value.at(LevelStatType::AVG_SEC),
      NumberToHumanString(
          static_cast<std::int64_t>(stat_value.at(LevelStatType::KEY_IN)))
          .c_str(),
      NumberToHumanString(
          static_cast<std::int64_t>(stat_value.at(LevelStatType::KEY_DROP)))
186
          .c_str(),
187
      stat_value.at(LevelStatType::R_BLOB_GB),
188
      stat_value.at(LevelStatType::W_BLOB_GB));
189 190 191 192 193 194 195 196 197 198
}

void PrintLevelStats(char* buf, size_t len, const std::string& name,
                     int num_files, int being_compacted, double total_file_size,
                     double score, double w_amp,
                     const InternalStats::CompactionStats& stats) {
  std::map<LevelStatType, double> level_stats;
  PrepareLevelStats(&level_stats, num_files, being_compacted, total_file_size,
                    score, w_amp, stats);
  PrintLevelStats(buf, len, name, level_stats);
L
Lei Jin 已提交
199
}
200 201 202 203 204 205 206 207 208 209 210 211 212

// Assumes that trailing numbers represent an optional argument. This requires
// property names to not end with numbers.
std::pair<Slice, Slice> GetPropertyNameAndArg(const Slice& property) {
  Slice name = property, arg = property;
  size_t sfx_len = 0;
  while (sfx_len < property.size() &&
         isdigit(property[property.size() - sfx_len - 1])) {
    ++sfx_len;
  }
  name.remove_suffix(sfx_len);
  arg.remove_prefix(property.size() - sfx_len);
  return {name, arg};
L
Lei Jin 已提交
213
}
214
}  // anonymous namespace
L
Lei Jin 已提交
215

216 217 218
static const std::string rocksdb_prefix = "rocksdb.";

static const std::string num_files_at_level_prefix = "num-files-at-level";
219 220
static const std::string compression_ratio_at_level_prefix =
    "compression-ratio-at-level";
221
static const std::string allstats = "stats";
222 223
static const std::string sstables = "sstables";
static const std::string cfstats = "cfstats";
224 225 226
static const std::string cfstats_no_file_histogram =
    "cfstats-no-file-histogram";
static const std::string cf_file_histogram = "cf-file-histogram";
227 228
static const std::string dbstats = "dbstats";
static const std::string levelstats = "levelstats";
229
static const std::string block_cache_entry_stats = "block-cache-entry-stats";
230
static const std::string num_immutable_mem_table = "num-immutable-mem-table";
231 232
static const std::string num_immutable_mem_table_flushed =
    "num-immutable-mem-table-flushed";
233 234 235 236
static const std::string mem_table_flush_pending = "mem-table-flush-pending";
static const std::string compaction_pending = "compaction-pending";
static const std::string background_errors = "background-errors";
static const std::string cur_size_active_mem_table =
237
    "cur-size-active-mem-table";
238 239
static const std::string cur_size_all_mem_tables = "cur-size-all-mem-tables";
static const std::string size_all_mem_tables = "size-all-mem-tables";
240
static const std::string num_entries_active_mem_table =
241
    "num-entries-active-mem-table";
242
static const std::string num_entries_imm_mem_tables =
243
    "num-entries-imm-mem-tables";
244
static const std::string num_deletes_active_mem_table =
245
    "num-deletes-active-mem-table";
246
static const std::string num_deletes_imm_mem_tables =
247
    "num-deletes-imm-mem-tables";
248 249
static const std::string estimate_num_keys = "estimate-num-keys";
static const std::string estimate_table_readers_mem =
250
    "estimate-table-readers-mem";
251
static const std::string is_file_deletions_enabled =
252
    "is-file-deletions-enabled";
253 254
static const std::string num_snapshots = "num-snapshots";
static const std::string oldest_snapshot_time = "oldest-snapshot-time";
255
static const std::string oldest_snapshot_sequence = "oldest-snapshot-sequence";
256
static const std::string num_live_versions = "num-live-versions";
257 258
static const std::string current_version_number =
    "current-super-version-number";
A
Andres Notzli 已提交
259
static const std::string estimate_live_data_size = "estimate-live-data-size";
260
static const std::string min_log_number_to_keep_str = "min-log-number-to-keep";
261 262
static const std::string min_obsolete_sst_number_to_keep_str =
    "min-obsolete-sst-number-to-keep";
263
static const std::string base_level_str = "base-level";
264
static const std::string total_sst_files_size = "total-sst-files-size";
265
static const std::string live_sst_files_size = "live-sst-files-size";
266 267
static const std::string live_sst_files_size_at_temperature =
    "live-sst-files-size-at-temperature";
268 269
static const std::string estimate_pending_comp_bytes =
    "estimate-pending-compaction-bytes";
270 271 272 273
static const std::string aggregated_table_properties =
    "aggregated-table-properties";
static const std::string aggregated_table_properties_at_level =
    aggregated_table_properties + "-at-level";
274 275
static const std::string num_running_compactions = "num-running-compactions";
static const std::string num_running_flushes = "num-running-flushes";
276 277 278
static const std::string actual_delayed_write_rate =
    "actual-delayed-write-rate";
static const std::string is_write_stopped = "is-write-stopped";
Y
Yi Wu 已提交
279
static const std::string estimate_oldest_key_time = "estimate-oldest-key-time";
Y
Yi Wu 已提交
280 281 282
static const std::string block_cache_capacity = "block-cache-capacity";
static const std::string block_cache_usage = "block-cache-usage";
static const std::string block_cache_pinned_usage = "block-cache-pinned-usage";
283
static const std::string options_statistics = "options-statistics";
284 285 286 287
static const std::string num_blob_files = "num-blob-files";
static const std::string blob_stats = "blob-stats";
static const std::string total_blob_file_size = "total-blob-file-size";
static const std::string live_blob_file_size = "live-blob-file-size";
288 289

const std::string DB::Properties::kNumFilesAtLevelPrefix =
290
    rocksdb_prefix + num_files_at_level_prefix;
291
const std::string DB::Properties::kCompressionRatioAtLevelPrefix =
292
    rocksdb_prefix + compression_ratio_at_level_prefix;
293
const std::string DB::Properties::kStats = rocksdb_prefix + allstats;
294 295
const std::string DB::Properties::kSSTables = rocksdb_prefix + sstables;
const std::string DB::Properties::kCFStats = rocksdb_prefix + cfstats;
296 297 298 299
const std::string DB::Properties::kCFStatsNoFileHistogram =
    rocksdb_prefix + cfstats_no_file_histogram;
const std::string DB::Properties::kCFFileHistogram =
    rocksdb_prefix + cf_file_histogram;
300
const std::string DB::Properties::kDBStats = rocksdb_prefix + dbstats;
301
const std::string DB::Properties::kLevelStats = rocksdb_prefix + levelstats;
302 303
const std::string DB::Properties::kBlockCacheEntryStats =
    rocksdb_prefix + block_cache_entry_stats;
304
const std::string DB::Properties::kNumImmutableMemTable =
305
    rocksdb_prefix + num_immutable_mem_table;
306
const std::string DB::Properties::kNumImmutableMemTableFlushed =
307
    rocksdb_prefix + num_immutable_mem_table_flushed;
308
const std::string DB::Properties::kMemTableFlushPending =
309
    rocksdb_prefix + mem_table_flush_pending;
310
const std::string DB::Properties::kCompactionPending =
311
    rocksdb_prefix + compaction_pending;
312 313 314 315
const std::string DB::Properties::kNumRunningCompactions =
    rocksdb_prefix + num_running_compactions;
const std::string DB::Properties::kNumRunningFlushes =
    rocksdb_prefix + num_running_flushes;
316
const std::string DB::Properties::kBackgroundErrors =
317
    rocksdb_prefix + background_errors;
318
const std::string DB::Properties::kCurSizeActiveMemTable =
319
    rocksdb_prefix + cur_size_active_mem_table;
320
const std::string DB::Properties::kCurSizeAllMemTables =
321
    rocksdb_prefix + cur_size_all_mem_tables;
322 323
const std::string DB::Properties::kSizeAllMemTables =
    rocksdb_prefix + size_all_mem_tables;
324
const std::string DB::Properties::kNumEntriesActiveMemTable =
325
    rocksdb_prefix + num_entries_active_mem_table;
326
const std::string DB::Properties::kNumEntriesImmMemTables =
327
    rocksdb_prefix + num_entries_imm_mem_tables;
328
const std::string DB::Properties::kNumDeletesActiveMemTable =
329
    rocksdb_prefix + num_deletes_active_mem_table;
330
const std::string DB::Properties::kNumDeletesImmMemTables =
331
    rocksdb_prefix + num_deletes_imm_mem_tables;
332
const std::string DB::Properties::kEstimateNumKeys =
333
    rocksdb_prefix + estimate_num_keys;
334
const std::string DB::Properties::kEstimateTableReadersMem =
335
    rocksdb_prefix + estimate_table_readers_mem;
336
const std::string DB::Properties::kIsFileDeletionsEnabled =
337
    rocksdb_prefix + is_file_deletions_enabled;
338
const std::string DB::Properties::kNumSnapshots =
339
    rocksdb_prefix + num_snapshots;
340
const std::string DB::Properties::kOldestSnapshotTime =
341
    rocksdb_prefix + oldest_snapshot_time;
342 343
const std::string DB::Properties::kOldestSnapshotSequence =
    rocksdb_prefix + oldest_snapshot_sequence;
344
const std::string DB::Properties::kNumLiveVersions =
345
    rocksdb_prefix + num_live_versions;
346
const std::string DB::Properties::kCurrentSuperVersionNumber =
347
    rocksdb_prefix + current_version_number;
A
Andres Notzli 已提交
348
const std::string DB::Properties::kEstimateLiveDataSize =
349
    rocksdb_prefix + estimate_live_data_size;
350
const std::string DB::Properties::kMinLogNumberToKeep =
351
    rocksdb_prefix + min_log_number_to_keep_str;
352 353
const std::string DB::Properties::kMinObsoleteSstNumberToKeep =
    rocksdb_prefix + min_obsolete_sst_number_to_keep_str;
354
const std::string DB::Properties::kTotalSstFilesSize =
355
    rocksdb_prefix + total_sst_files_size;
356 357
const std::string DB::Properties::kLiveSstFilesSize =
    rocksdb_prefix + live_sst_files_size;
358
const std::string DB::Properties::kBaseLevel = rocksdb_prefix + base_level_str;
359 360
const std::string DB::Properties::kEstimatePendingCompactionBytes =
    rocksdb_prefix + estimate_pending_comp_bytes;
361 362 363 364
const std::string DB::Properties::kAggregatedTableProperties =
    rocksdb_prefix + aggregated_table_properties;
const std::string DB::Properties::kAggregatedTablePropertiesAtLevel =
    rocksdb_prefix + aggregated_table_properties_at_level;
365 366 367 368
const std::string DB::Properties::kActualDelayedWriteRate =
    rocksdb_prefix + actual_delayed_write_rate;
const std::string DB::Properties::kIsWriteStopped =
    rocksdb_prefix + is_write_stopped;
Y
Yi Wu 已提交
369 370
const std::string DB::Properties::kEstimateOldestKeyTime =
    rocksdb_prefix + estimate_oldest_key_time;
Y
Yi Wu 已提交
371 372 373 374 375 376
const std::string DB::Properties::kBlockCacheCapacity =
    rocksdb_prefix + block_cache_capacity;
const std::string DB::Properties::kBlockCacheUsage =
    rocksdb_prefix + block_cache_usage;
const std::string DB::Properties::kBlockCachePinnedUsage =
    rocksdb_prefix + block_cache_pinned_usage;
377 378
const std::string DB::Properties::kOptionsStatistics =
    rocksdb_prefix + options_statistics;
379 380
const std::string DB::Properties::kLiveSstFilesSizeAtTemperature =
    rocksdb_prefix + live_sst_files_size_at_temperature;
381 382 383 384 385 386 387
const std::string DB::Properties::kNumBlobFiles =
    rocksdb_prefix + num_blob_files;
const std::string DB::Properties::kBlobStats = rocksdb_prefix + blob_stats;
const std::string DB::Properties::kTotalBlobFileSize =
    rocksdb_prefix + total_blob_file_size;
const std::string DB::Properties::kLiveBlobFileSize =
    rocksdb_prefix + live_blob_file_size;
388

389 390 391
const std::unordered_map<std::string, DBPropertyInfo>
    InternalStats::ppt_name_to_info = {
        {DB::Properties::kNumFilesAtLevelPrefix,
392 393
         {false, &InternalStats::HandleNumFilesAtLevel, nullptr, nullptr,
          nullptr}},
394 395
        {DB::Properties::kCompressionRatioAtLevelPrefix,
         {false, &InternalStats::HandleCompressionRatioAtLevelPrefix, nullptr,
396
          nullptr, nullptr}},
397
        {DB::Properties::kLevelStats,
398
         {false, &InternalStats::HandleLevelStats, nullptr, nullptr, nullptr}},
399
        {DB::Properties::kStats,
400
         {false, &InternalStats::HandleStats, nullptr, nullptr, nullptr}},
401 402
        {DB::Properties::kCFStats,
         {false, &InternalStats::HandleCFStats, nullptr,
403
          &InternalStats::HandleCFMapStats, nullptr}},
404
        {DB::Properties::kCFStatsNoFileHistogram,
405
         {false, &InternalStats::HandleCFStatsNoFileHistogram, nullptr, nullptr,
406 407
          nullptr}},
        {DB::Properties::kCFFileHistogram,
408 409
         {false, &InternalStats::HandleCFFileHistogram, nullptr, nullptr,
          nullptr}},
410
        {DB::Properties::kDBStats,
411
         {false, &InternalStats::HandleDBStats, nullptr, nullptr, nullptr}},
412
        {DB::Properties::kBlockCacheEntryStats,
413
         {true, &InternalStats::HandleBlockCacheEntryStats, nullptr,
414
          &InternalStats::HandleBlockCacheEntryStatsMap, nullptr}},
415
        {DB::Properties::kSSTables,
416
         {false, &InternalStats::HandleSsTables, nullptr, nullptr, nullptr}},
417 418
        {DB::Properties::kAggregatedTableProperties,
         {false, &InternalStats::HandleAggregatedTableProperties, nullptr,
419
          &InternalStats::HandleAggregatedTablePropertiesMap, nullptr}},
420 421
        {DB::Properties::kAggregatedTablePropertiesAtLevel,
         {false, &InternalStats::HandleAggregatedTablePropertiesAtLevel,
422 423
          nullptr, &InternalStats::HandleAggregatedTablePropertiesAtLevelMap,
          nullptr}},
424
        {DB::Properties::kNumImmutableMemTable,
425 426
         {false, nullptr, &InternalStats::HandleNumImmutableMemTable, nullptr,
          nullptr}},
427 428
        {DB::Properties::kNumImmutableMemTableFlushed,
         {false, nullptr, &InternalStats::HandleNumImmutableMemTableFlushed,
429
          nullptr, nullptr}},
430
        {DB::Properties::kMemTableFlushPending,
431 432
         {false, nullptr, &InternalStats::HandleMemTableFlushPending, nullptr,
          nullptr}},
433
        {DB::Properties::kCompactionPending,
434 435
         {false, nullptr, &InternalStats::HandleCompactionPending, nullptr,
          nullptr}},
436
        {DB::Properties::kBackgroundErrors,
437 438
         {false, nullptr, &InternalStats::HandleBackgroundErrors, nullptr,
          nullptr}},
439
        {DB::Properties::kCurSizeActiveMemTable,
440
         {false, nullptr, &InternalStats::HandleCurSizeActiveMemTable, nullptr,
441 442
          nullptr}},
        {DB::Properties::kCurSizeAllMemTables,
443 444
         {false, nullptr, &InternalStats::HandleCurSizeAllMemTables, nullptr,
          nullptr}},
445
        {DB::Properties::kSizeAllMemTables,
446 447
         {false, nullptr, &InternalStats::HandleSizeAllMemTables, nullptr,
          nullptr}},
448 449
        {DB::Properties::kNumEntriesActiveMemTable,
         {false, nullptr, &InternalStats::HandleNumEntriesActiveMemTable,
450
          nullptr, nullptr}},
451
        {DB::Properties::kNumEntriesImmMemTables,
452
         {false, nullptr, &InternalStats::HandleNumEntriesImmMemTables, nullptr,
453 454 455
          nullptr}},
        {DB::Properties::kNumDeletesActiveMemTable,
         {false, nullptr, &InternalStats::HandleNumDeletesActiveMemTable,
456
          nullptr, nullptr}},
457
        {DB::Properties::kNumDeletesImmMemTables,
458
         {false, nullptr, &InternalStats::HandleNumDeletesImmMemTables, nullptr,
459 460
          nullptr}},
        {DB::Properties::kEstimateNumKeys,
461 462
         {false, nullptr, &InternalStats::HandleEstimateNumKeys, nullptr,
          nullptr}},
463
        {DB::Properties::kEstimateTableReadersMem,
464
         {true, nullptr, &InternalStats::HandleEstimateTableReadersMem, nullptr,
465 466
          nullptr}},
        {DB::Properties::kIsFileDeletionsEnabled,
467
         {false, nullptr, &InternalStats::HandleIsFileDeletionsEnabled, nullptr,
468 469
          nullptr}},
        {DB::Properties::kNumSnapshots,
470 471
         {false, nullptr, &InternalStats::HandleNumSnapshots, nullptr,
          nullptr}},
472
        {DB::Properties::kOldestSnapshotTime,
473 474
         {false, nullptr, &InternalStats::HandleOldestSnapshotTime, nullptr,
          nullptr}},
475 476 477
        {DB::Properties::kOldestSnapshotSequence,
         {false, nullptr, &InternalStats::HandleOldestSnapshotSequence, nullptr,
          nullptr}},
478
        {DB::Properties::kNumLiveVersions,
479 480
         {false, nullptr, &InternalStats::HandleNumLiveVersions, nullptr,
          nullptr}},
481 482
        {DB::Properties::kCurrentSuperVersionNumber,
         {false, nullptr, &InternalStats::HandleCurrentSuperVersionNumber,
483
          nullptr, nullptr}},
484
        {DB::Properties::kEstimateLiveDataSize,
485 486
         {true, nullptr, &InternalStats::HandleEstimateLiveDataSize, nullptr,
          nullptr}},
487
        {DB::Properties::kMinLogNumberToKeep,
488 489
         {false, nullptr, &InternalStats::HandleMinLogNumberToKeep, nullptr,
          nullptr}},
490 491 492
        {DB::Properties::kMinObsoleteSstNumberToKeep,
         {false, nullptr, &InternalStats::HandleMinObsoleteSstNumberToKeep,
          nullptr, nullptr}},
493
        {DB::Properties::kBaseLevel,
494
         {false, nullptr, &InternalStats::HandleBaseLevel, nullptr, nullptr}},
495
        {DB::Properties::kTotalSstFilesSize,
496 497
         {false, nullptr, &InternalStats::HandleTotalSstFilesSize, nullptr,
          nullptr}},
498
        {DB::Properties::kLiveSstFilesSize,
499 500
         {false, nullptr, &InternalStats::HandleLiveSstFilesSize, nullptr,
          nullptr}},
501 502 503
        {DB::Properties::kLiveSstFilesSizeAtTemperature,
         {true, &InternalStats::HandleLiveSstFilesSizeAtTemperature, nullptr,
          nullptr, nullptr}},
504 505
        {DB::Properties::kEstimatePendingCompactionBytes,
         {false, nullptr, &InternalStats::HandleEstimatePendingCompactionBytes,
506
          nullptr, nullptr}},
507
        {DB::Properties::kNumRunningFlushes,
508 509
         {false, nullptr, &InternalStats::HandleNumRunningFlushes, nullptr,
          nullptr}},
510
        {DB::Properties::kNumRunningCompactions,
511
         {false, nullptr, &InternalStats::HandleNumRunningCompactions, nullptr,
512
          nullptr}},
513
        {DB::Properties::kActualDelayedWriteRate,
514
         {false, nullptr, &InternalStats::HandleActualDelayedWriteRate, nullptr,
515 516
          nullptr}},
        {DB::Properties::kIsWriteStopped,
517 518
         {false, nullptr, &InternalStats::HandleIsWriteStopped, nullptr,
          nullptr}},
Y
Yi Wu 已提交
519
        {DB::Properties::kEstimateOldestKeyTime,
520
         {false, nullptr, &InternalStats::HandleEstimateOldestKeyTime, nullptr,
Y
Yi Wu 已提交
521
          nullptr}},
Y
Yi Wu 已提交
522
        {DB::Properties::kBlockCacheCapacity,
523 524
         {false, nullptr, &InternalStats::HandleBlockCacheCapacity, nullptr,
          nullptr}},
Y
Yi Wu 已提交
525
        {DB::Properties::kBlockCacheUsage,
526 527
         {false, nullptr, &InternalStats::HandleBlockCacheUsage, nullptr,
          nullptr}},
Y
Yi Wu 已提交
528
        {DB::Properties::kBlockCachePinnedUsage,
529
         {false, nullptr, &InternalStats::HandleBlockCachePinnedUsage, nullptr,
Y
Yi Wu 已提交
530
          nullptr}},
531
        {DB::Properties::kOptionsStatistics,
532
         {true, nullptr, nullptr, nullptr,
533
          &DBImpl::GetPropertyHandleOptionsStatistics}},
534 535 536 537 538 539 540 541 542 543 544
        {DB::Properties::kNumBlobFiles,
         {false, nullptr, &InternalStats::HandleNumBlobFiles, nullptr,
          nullptr}},
        {DB::Properties::kBlobStats,
         {false, &InternalStats::HandleBlobStats, nullptr, nullptr, nullptr}},
        {DB::Properties::kTotalBlobFileSize,
         {false, nullptr, &InternalStats::HandleTotalBlobFileSize, nullptr,
          nullptr}},
        {DB::Properties::kLiveBlobFileSize,
         {false, nullptr, &InternalStats::HandleLiveBlobFileSize, nullptr,
          nullptr}},
545 546
};

547 548 549 550 551 552 553 554 555 556 557 558
InternalStats::InternalStats(int num_levels, SystemClock* clock,
                             ColumnFamilyData* cfd)
    : db_stats_{},
      cf_stats_value_{},
      cf_stats_count_{},
      comp_stats_(num_levels),
      comp_stats_by_pri_(Env::Priority::TOTAL),
      file_read_latency_(num_levels),
      bg_error_count_(0),
      number_levels_(num_levels),
      clock_(clock),
      cfd_(cfd),
559 560 561 562 563 564 565 566 567 568
      started_at_(clock->NowMicros()) {
  Cache* block_cache = nullptr;
  bool ok = GetBlockCacheForStats(&block_cache);
  if (ok) {
    assert(block_cache);
    // Extract or create stats collector. Could fail in rare cases.
    Status s = CacheEntryStatsCollector<CacheEntryRoleStats>::GetShared(
        block_cache, clock_, &cache_entry_stats_collector_);
    if (s.ok()) {
      assert(cache_entry_stats_collector_);
569
    } else {
570
      assert(!cache_entry_stats_collector_);
571
    }
572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592
  } else {
    assert(!block_cache);
  }
}

void InternalStats::TEST_GetCacheEntryRoleStats(CacheEntryRoleStats* stats,
                                                bool foreground) {
  CollectCacheEntryStats(foreground);
  if (cache_entry_stats_collector_) {
    cache_entry_stats_collector_->GetStats(stats);
  }
}

void InternalStats::CollectCacheEntryStats(bool foreground) {
  // This function is safe to call from any thread because
  // cache_entry_stats_collector_ field is const after constructor
  // and ->GetStats does its own synchronization, which also suffices for
  // cache_entry_stats_.

  if (!cache_entry_stats_collector_) {
    return;  // nothing to do (e.g. no block cache)
593
  }
594 595 596 597 598 599 600

  // For "background" collections, strictly cap the collection time by
  // expanding effective cache TTL. For foreground, be more aggressive about
  // getting latest data.
  int min_interval_seconds = foreground ? 10 : 180;
  // 1/500 = max of 0.2% of one CPU thread
  int min_interval_factor = foreground ? 10 : 500;
601 602
  cache_entry_stats_collector_->CollectStats(min_interval_seconds,
                                             min_interval_factor);
603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627
}

std::function<void(const Slice&, void*, size_t, Cache::DeleterFn)>
InternalStats::CacheEntryRoleStats::GetEntryCallback() {
  return [&](const Slice& /*key*/, void* /*value*/, size_t charge,
             Cache::DeleterFn deleter) {
    auto e = role_map_.find(deleter);
    size_t role_idx;
    if (e == role_map_.end()) {
      role_idx = static_cast<size_t>(CacheEntryRole::kMisc);
    } else {
      role_idx = static_cast<size_t>(e->second);
    }
    entry_counts[role_idx]++;
    total_charges[role_idx] += charge;
  };
}

void InternalStats::CacheEntryRoleStats::BeginCollection(
    Cache* cache, SystemClock*, uint64_t start_time_micros) {
  Clear();
  last_start_time_micros_ = start_time_micros;
  ++collection_count;
  role_map_ = CopyCacheDeleterRoleMap();
  std::ostringstream str;
628 629
  str << cache->Name() << "@" << static_cast<void*>(cache) << "#"
      << port::GetProcessID();
630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693
  cache_id = str.str();
  cache_capacity = cache->GetCapacity();
}

void InternalStats::CacheEntryRoleStats::EndCollection(
    Cache*, SystemClock*, uint64_t end_time_micros) {
  last_end_time_micros_ = end_time_micros;
}

void InternalStats::CacheEntryRoleStats::SkippedCollection() {
  ++copies_of_last_collection;
}

uint64_t InternalStats::CacheEntryRoleStats::GetLastDurationMicros() const {
  if (last_end_time_micros_ > last_start_time_micros_) {
    return last_end_time_micros_ - last_start_time_micros_;
  } else {
    return 0U;
  }
}

std::string InternalStats::CacheEntryRoleStats::ToString(
    SystemClock* clock) const {
  std::ostringstream str;
  str << "Block cache " << cache_id
      << " capacity: " << BytesToHumanString(cache_capacity)
      << " collections: " << collection_count
      << " last_copies: " << copies_of_last_collection
      << " last_secs: " << (GetLastDurationMicros() / 1000000.0)
      << " secs_since: "
      << ((clock->NowMicros() - last_end_time_micros_) / 1000000U) << "\n";
  str << "Block cache entry stats(count,size,portion):";
  for (size_t i = 0; i < kNumCacheEntryRoles; ++i) {
    if (entry_counts[i] > 0) {
      str << " " << kCacheEntryRoleToCamelString[i] << "(" << entry_counts[i]
          << "," << BytesToHumanString(total_charges[i]) << ","
          << (100.0 * total_charges[i] / cache_capacity) << "%)";
    }
  }
  str << "\n";
  return str.str();
}

void InternalStats::CacheEntryRoleStats::ToMap(
    std::map<std::string, std::string>* values, SystemClock* clock) const {
  values->clear();
  auto& v = *values;
  v["id"] = cache_id;
  v["capacity"] = ROCKSDB_NAMESPACE::ToString(cache_capacity);
  v["secs_for_last_collection"] =
      ROCKSDB_NAMESPACE::ToString(GetLastDurationMicros() / 1000000.0);
  v["secs_since_last_collection"] = ROCKSDB_NAMESPACE::ToString(
      (clock->NowMicros() - last_end_time_micros_) / 1000000U);
  for (size_t i = 0; i < kNumCacheEntryRoles; ++i) {
    std::string role = kCacheEntryRoleToHyphenString[i];
    v["count." + role] = ROCKSDB_NAMESPACE::ToString(entry_counts[i]);
    v["bytes." + role] = ROCKSDB_NAMESPACE::ToString(total_charges[i]);
    v["percent." + role] =
        ROCKSDB_NAMESPACE::ToString(100.0 * total_charges[i] / cache_capacity);
  }
}

bool InternalStats::HandleBlockCacheEntryStats(std::string* value,
                                               Slice /*suffix*/) {
694
  if (!cache_entry_stats_collector_) {
695 696
    return false;
  }
697 698 699 700
  CollectCacheEntryStats(/*foreground*/ true);
  CacheEntryRoleStats stats;
  cache_entry_stats_collector_->GetStats(&stats);
  *value = stats.ToString(clock_);
701 702 703 704 705
  return true;
}

bool InternalStats::HandleBlockCacheEntryStatsMap(
    std::map<std::string, std::string>* values, Slice /*suffix*/) {
706
  if (!cache_entry_stats_collector_) {
707 708
    return false;
  }
709 710 711 712
  CollectCacheEntryStats(/*foreground*/ true);
  CacheEntryRoleStats stats;
  cache_entry_stats_collector_->GetStats(&stats);
  stats.ToMap(values, clock_);
713 714 715
  return true;
}

716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737
bool InternalStats::HandleLiveSstFilesSizeAtTemperature(std::string* value,
                                                        Slice suffix) {
  uint64_t temperature;
  bool ok = ConsumeDecimalNumber(&suffix, &temperature) && suffix.empty();
  if (!ok) {
    return false;
  }

  uint64_t size = 0;
  const auto* vstorage = cfd_->current()->storage_info();
  for (int level = 0; level < vstorage->num_levels(); level++) {
    for (const auto& file_meta : vstorage->LevelFiles(level)) {
      if (static_cast<uint8_t>(file_meta->temperature) == temperature) {
        size += file_meta->fd.GetFileSize();
      }
    }
  }

  *value = ToString(size);
  return true;
}

738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754
bool InternalStats::HandleNumBlobFiles(uint64_t* value, DBImpl* /*db*/,
                                       Version* /*version*/) {
  const auto* vstorage = cfd_->current()->storage_info();
  const auto& blob_files = vstorage->GetBlobFiles();
  *value = blob_files.size();
  return true;
}

bool InternalStats::HandleBlobStats(std::string* value, Slice /*suffix*/) {
  std::ostringstream oss;
  auto* current_version = cfd_->current();
  const auto& blob_files = current_version->storage_info()->GetBlobFiles();
  uint64_t current_num_blob_files = blob_files.size();
  uint64_t current_file_size = 0;
  uint64_t current_garbage_size = 0;
  for (const auto& pair : blob_files) {
    const auto& meta = pair.second;
755
    current_file_size += meta->GetBlobFileSize();
756 757
    current_garbage_size += meta->GetGarbageBlobBytes();
  }
758 759 760 761
  oss << "Number of blob files: " << current_num_blob_files
      << "\nTotal size of blob files: " << current_file_size
      << "\nTotal size of garbage in blob files: " << current_garbage_size
      << '\n';
762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778
  value->append(oss.str());
  return true;
}

bool InternalStats::HandleTotalBlobFileSize(uint64_t* value, DBImpl* /*db*/,
                                            Version* /*version*/) {
  *value = cfd_->GetTotalBlobFileSize();
  return true;
}

bool InternalStats::HandleLiveBlobFileSize(uint64_t* value, DBImpl* /*db*/,
                                           Version* /*version*/) {
  const auto* vstorage = cfd_->current()->storage_info();
  *value = vstorage->GetTotalBlobFileSize();
  return true;
}

779 780 781 782 783
const DBPropertyInfo* GetPropertyInfo(const Slice& property) {
  std::string ppt_name = GetPropertyNameAndArg(property).first.ToString();
  auto ppt_info_iter = InternalStats::ppt_name_to_info.find(ppt_name);
  if (ppt_info_iter == InternalStats::ppt_name_to_info.end()) {
    return nullptr;
784
  }
785 786
  return &ppt_info_iter->second;
}
787

788 789 790 791 792 793 794
bool InternalStats::GetStringProperty(const DBPropertyInfo& property_info,
                                      const Slice& property,
                                      std::string* value) {
  assert(value != nullptr);
  assert(property_info.handle_string != nullptr);
  Slice arg = GetPropertyNameAndArg(property).second;
  return (this->*(property_info.handle_string))(value, arg);
795 796
}

797
bool InternalStats::GetMapProperty(const DBPropertyInfo& property_info,
798
                                   const Slice& property,
799
                                   std::map<std::string, std::string>* value) {
800 801
  assert(value != nullptr);
  assert(property_info.handle_map != nullptr);
802 803
  Slice arg = GetPropertyNameAndArg(property).second;
  return (this->*(property_info.handle_map))(value, arg);
804 805
}

806 807
bool InternalStats::GetIntProperty(const DBPropertyInfo& property_info,
                                   uint64_t* value, DBImpl* db) {
808
  assert(value != nullptr);
809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824
  assert(property_info.handle_int != nullptr &&
         !property_info.need_out_of_mutex);
  db->mutex_.AssertHeld();
  return (this->*(property_info.handle_int))(value, db, nullptr /* version */);
}

bool InternalStats::GetIntPropertyOutOfMutex(
    const DBPropertyInfo& property_info, Version* version, uint64_t* value) {
  assert(value != nullptr);
  assert(property_info.handle_int != nullptr &&
         property_info.need_out_of_mutex);
  return (this->*(property_info.handle_int))(value, nullptr /* db */, version);
}

bool InternalStats::HandleNumFilesAtLevel(std::string* value, Slice suffix) {
  uint64_t level;
A
Andres Notzli 已提交
825
  const auto* vstorage = cfd_->current()->storage_info();
826 827 828 829 830 831 832 833 834 835 836 837
  bool ok = ConsumeDecimalNumber(&suffix, &level) && suffix.empty();
  if (!ok || static_cast<int>(level) >= number_levels_) {
    return false;
  } else {
    char buf[100];
    snprintf(buf, sizeof(buf), "%d",
             vstorage->NumLevelFiles(static_cast<int>(level)));
    *value = buf;
    return true;
  }
}

838 839 840 841 842 843 844 845 846 847 848 849 850
bool InternalStats::HandleCompressionRatioAtLevelPrefix(std::string* value,
                                                        Slice suffix) {
  uint64_t level;
  const auto* vstorage = cfd_->current()->storage_info();
  bool ok = ConsumeDecimalNumber(&suffix, &level) && suffix.empty();
  if (!ok || level >= static_cast<uint64_t>(number_levels_)) {
    return false;
  }
  *value = ToString(
      vstorage->GetEstimatedCompressionRatioAtLevel(static_cast<int>(level)));
  return true;
}

A
Andrew Kryczka 已提交
851
bool InternalStats::HandleLevelStats(std::string* value, Slice /*suffix*/) {
852 853 854 855 856 857
  char buf[1000];
  const auto* vstorage = cfd_->current()->storage_info();
  snprintf(buf, sizeof(buf),
           "Level Files Size(MB)\n"
           "--------------------\n");
  value->append(buf);
A
Andres Notzli 已提交
858

859 860 861 862 863
  for (int level = 0; level < number_levels_; level++) {
    snprintf(buf, sizeof(buf), "%3d %8d %8.0f\n", level,
             vstorage->NumLevelFiles(level),
             vstorage->NumLevelBytes(level) / kMB);
    value->append(buf);
864
  }
865
  return true;
866
}
867

868 869 870 871 872 873 874 875 876 877
bool InternalStats::HandleStats(std::string* value, Slice suffix) {
  if (!HandleCFStats(value, suffix)) {
    return false;
  }
  if (!HandleDBStats(value, suffix)) {
    return false;
  }
  return true;
}

878
bool InternalStats::HandleCFMapStats(
879
    std::map<std::string, std::string>* cf_stats, Slice /*suffix*/) {
880 881 882 883
  DumpCFMapStats(cf_stats);
  return true;
}

A
Andrew Kryczka 已提交
884
bool InternalStats::HandleCFStats(std::string* value, Slice /*suffix*/) {
885 886 887 888
  DumpCFStats(value);
  return true;
}

889
bool InternalStats::HandleCFStatsNoFileHistogram(std::string* value,
A
Andrew Kryczka 已提交
890
                                                 Slice /*suffix*/) {
891 892 893 894
  DumpCFStatsNoFileHistogram(value);
  return true;
}

A
Andrew Kryczka 已提交
895 896
bool InternalStats::HandleCFFileHistogram(std::string* value,
                                          Slice /*suffix*/) {
897 898 899 900
  DumpCFFileHistogram(value);
  return true;
}

A
Andrew Kryczka 已提交
901
bool InternalStats::HandleDBStats(std::string* value, Slice /*suffix*/) {
902 903 904 905
  DumpDBStats(value);
  return true;
}

A
Andrew Kryczka 已提交
906
bool InternalStats::HandleSsTables(std::string* value, Slice /*suffix*/) {
S
sdong 已提交
907
  auto* current = cfd_->current();
908
  *value = current->DebugString(true, true);
909 910
  return true;
}
I
Igor Canadi 已提交
911

912
bool InternalStats::HandleAggregatedTableProperties(std::string* value,
A
Andrew Kryczka 已提交
913
                                                    Slice /*suffix*/) {
914 915 916 917
  std::shared_ptr<const TableProperties> tp;
  auto s = cfd_->current()->GetAggregatedTableProperties(&tp);
  if (!s.ok()) {
    return false;
918
  }
919 920
  *value = tp->ToString();
  return true;
921 922
}

923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943
static std::map<std::string, std::string> MapUint64ValuesToString(
    const std::map<std::string, uint64_t>& from) {
  std::map<std::string, std::string> to;
  for (const auto& e : from) {
    to[e.first] = ToString(e.second);
  }
  return to;
}

bool InternalStats::HandleAggregatedTablePropertiesMap(
    std::map<std::string, std::string>* values, Slice /*suffix*/) {
  std::shared_ptr<const TableProperties> tp;
  auto s = cfd_->current()->GetAggregatedTableProperties(&tp);
  if (!s.ok()) {
    return false;
  }
  *values = MapUint64ValuesToString(tp->GetAggregatablePropertiesAsMap());
  return true;
}

bool InternalStats::HandleAggregatedTablePropertiesAtLevel(std::string* values,
944 945 946 947 948 949 950 951 952 953 954 955
                                                           Slice suffix) {
  uint64_t level;
  bool ok = ConsumeDecimalNumber(&suffix, &level) && suffix.empty();
  if (!ok || static_cast<int>(level) >= number_levels_) {
    return false;
  }
  std::shared_ptr<const TableProperties> tp;
  auto s = cfd_->current()->GetAggregatedTableProperties(
      &tp, static_cast<int>(level));
  if (!s.ok()) {
    return false;
  }
956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973
  *values = tp->ToString();
  return true;
}

bool InternalStats::HandleAggregatedTablePropertiesAtLevelMap(
    std::map<std::string, std::string>* values, Slice suffix) {
  uint64_t level;
  bool ok = ConsumeDecimalNumber(&suffix, &level) && suffix.empty();
  if (!ok || static_cast<int>(level) >= number_levels_) {
    return false;
  }
  std::shared_ptr<const TableProperties> tp;
  auto s = cfd_->current()->GetAggregatedTableProperties(
      &tp, static_cast<int>(level));
  if (!s.ok()) {
    return false;
  }
  *values = MapUint64ValuesToString(tp->GetAggregatablePropertiesAsMap());
974 975 976
  return true;
}

A
Andrew Kryczka 已提交
977 978
bool InternalStats::HandleNumImmutableMemTable(uint64_t* value, DBImpl* /*db*/,
                                               Version* /*version*/) {
979 980 981 982 983
  *value = cfd_->imm()->NumNotFlushed();
  return true;
}

bool InternalStats::HandleNumImmutableMemTableFlushed(uint64_t* value,
A
Andrew Kryczka 已提交
984 985
                                                      DBImpl* /*db*/,
                                                      Version* /*version*/) {
986 987 988 989
  *value = cfd_->imm()->NumFlushed();
  return true;
}

A
Andrew Kryczka 已提交
990 991
bool InternalStats::HandleMemTableFlushPending(uint64_t* value, DBImpl* /*db*/,
                                               Version* /*version*/) {
992 993 994 995 996
  *value = (cfd_->imm()->IsFlushPending() ? 1 : 0);
  return true;
}

bool InternalStats::HandleNumRunningFlushes(uint64_t* value, DBImpl* db,
A
Andrew Kryczka 已提交
997
                                            Version* /*version*/) {
998 999 1000 1001
  *value = db->num_running_flushes();
  return true;
}

A
Andrew Kryczka 已提交
1002 1003
bool InternalStats::HandleCompactionPending(uint64_t* value, DBImpl* /*db*/,
                                            Version* /*version*/) {
1004 1005
  // 1 if the system already determines at least one compaction is needed.
  // 0 otherwise,
S
sdong 已提交
1006
  const auto* vstorage = cfd_->current()->storage_info();
1007 1008 1009
  *value = (cfd_->compaction_picker()->NeedsCompaction(vstorage) ? 1 : 0);
  return true;
}
1010

1011
bool InternalStats::HandleNumRunningCompactions(uint64_t* value, DBImpl* db,
A
Andrew Kryczka 已提交
1012
                                                Version* /*version*/) {
1013 1014 1015 1016
  *value = db->num_running_compactions_;
  return true;
}

A
Andrew Kryczka 已提交
1017 1018
bool InternalStats::HandleBackgroundErrors(uint64_t* value, DBImpl* /*db*/,
                                           Version* /*version*/) {
1019 1020 1021 1022 1023
  // Accumulated number of  errors in background flushes or compactions.
  *value = GetBackgroundErrorCount();
  return true;
}

A
Andrew Kryczka 已提交
1024 1025
bool InternalStats::HandleCurSizeActiveMemTable(uint64_t* value, DBImpl* /*db*/,
                                                Version* /*version*/) {
1026
  // Current size of the active memtable
1027 1028
  // Using ApproximateMemoryUsageFast to avoid the need for synchronization
  *value = cfd_->mem()->ApproximateMemoryUsageFast();
1029 1030 1031
  return true;
}

A
Andrew Kryczka 已提交
1032 1033
bool InternalStats::HandleCurSizeAllMemTables(uint64_t* value, DBImpl* /*db*/,
                                              Version* /*version*/) {
1034
  // Current size of the active memtable + immutable memtables
1035 1036
  // Using ApproximateMemoryUsageFast to avoid the need for synchronization
  *value = cfd_->mem()->ApproximateMemoryUsageFast() +
1037 1038 1039 1040
           cfd_->imm()->ApproximateUnflushedMemTablesMemoryUsage();
  return true;
}

A
Andrew Kryczka 已提交
1041 1042
bool InternalStats::HandleSizeAllMemTables(uint64_t* value, DBImpl* /*db*/,
                                           Version* /*version*/) {
1043 1044
  // Using ApproximateMemoryUsageFast to avoid the need for synchronization
  *value = cfd_->mem()->ApproximateMemoryUsageFast() +
1045 1046 1047 1048
           cfd_->imm()->ApproximateMemoryUsage();
  return true;
}

A
Andrew Kryczka 已提交
1049 1050 1051
bool InternalStats::HandleNumEntriesActiveMemTable(uint64_t* value,
                                                   DBImpl* /*db*/,
                                                   Version* /*version*/) {
1052 1053 1054 1055 1056
  // Current number of entires in the active memtable
  *value = cfd_->mem()->num_entries();
  return true;
}

A
Andrew Kryczka 已提交
1057 1058 1059
bool InternalStats::HandleNumEntriesImmMemTables(uint64_t* value,
                                                 DBImpl* /*db*/,
                                                 Version* /*version*/) {
1060 1061 1062 1063 1064
  // Current number of entries in the immutable memtables
  *value = cfd_->imm()->current()->GetTotalNumEntries();
  return true;
}

A
Andrew Kryczka 已提交
1065 1066 1067
bool InternalStats::HandleNumDeletesActiveMemTable(uint64_t* value,
                                                   DBImpl* /*db*/,
                                                   Version* /*version*/) {
1068 1069 1070 1071 1072
  // Current number of entires in the active memtable
  *value = cfd_->mem()->num_deletes();
  return true;
}

A
Andrew Kryczka 已提交
1073 1074 1075
bool InternalStats::HandleNumDeletesImmMemTables(uint64_t* value,
                                                 DBImpl* /*db*/,
                                                 Version* /*version*/) {
1076 1077 1078 1079 1080
  // Current number of entries in the immutable memtables
  *value = cfd_->imm()->current()->GetTotalNumDeletes();
  return true;
}

A
Andrew Kryczka 已提交
1081 1082
bool InternalStats::HandleEstimateNumKeys(uint64_t* value, DBImpl* /*db*/,
                                          Version* /*version*/) {
1083 1084 1085
  // Estimate number of entries in the column family:
  // Use estimated entries in tables + total entries in memtables.
  const auto* vstorage = cfd_->current()->storage_info();
1086 1087 1088 1089 1090 1091 1092 1093
  uint64_t estimate_keys = cfd_->mem()->num_entries() +
                           cfd_->imm()->current()->GetTotalNumEntries() +
                           vstorage->GetEstimatedActiveKeys();
  uint64_t estimate_deletes =
      cfd_->mem()->num_deletes() + cfd_->imm()->current()->GetTotalNumDeletes();
  *value = estimate_keys > estimate_deletes * 2
               ? estimate_keys - (estimate_deletes * 2)
               : 0;
1094 1095 1096 1097
  return true;
}

bool InternalStats::HandleNumSnapshots(uint64_t* value, DBImpl* db,
A
Andrew Kryczka 已提交
1098
                                       Version* /*version*/) {
1099 1100 1101 1102 1103
  *value = db->snapshots().count();
  return true;
}

bool InternalStats::HandleOldestSnapshotTime(uint64_t* value, DBImpl* db,
A
Andrew Kryczka 已提交
1104
                                             Version* /*version*/) {
1105 1106 1107 1108
  *value = static_cast<uint64_t>(db->snapshots().GetOldestSnapshotTime());
  return true;
}

1109 1110 1111 1112 1113 1114
bool InternalStats::HandleOldestSnapshotSequence(uint64_t* value, DBImpl* db,
                                                 Version* /*version*/) {
  *value = static_cast<uint64_t>(db->snapshots().GetOldestSnapshotSequence());
  return true;
}

A
Andrew Kryczka 已提交
1115 1116
bool InternalStats::HandleNumLiveVersions(uint64_t* value, DBImpl* /*db*/,
                                          Version* /*version*/) {
1117 1118 1119 1120
  *value = cfd_->GetNumLiveVersions();
  return true;
}

A
Andrew Kryczka 已提交
1121 1122 1123
bool InternalStats::HandleCurrentSuperVersionNumber(uint64_t* value,
                                                    DBImpl* /*db*/,
                                                    Version* /*version*/) {
1124 1125 1126 1127
  *value = cfd_->GetSuperVersionNumber();
  return true;
}

1128
bool InternalStats::HandleIsFileDeletionsEnabled(uint64_t* value, DBImpl* db,
A
Andrew Kryczka 已提交
1129
                                                 Version* /*version*/) {
1130
  *value = db->IsFileDeletionsEnabled() ? 1 : 0;
1131 1132 1133
  return true;
}

A
Andrew Kryczka 已提交
1134 1135
bool InternalStats::HandleBaseLevel(uint64_t* value, DBImpl* /*db*/,
                                    Version* /*version*/) {
1136 1137 1138 1139 1140
  const auto* vstorage = cfd_->current()->storage_info();
  *value = vstorage->base_level();
  return true;
}

1141 1142
bool InternalStats::HandleTotalSstFilesSize(uint64_t* value, DBImpl* /*db*/,
                                            Version* /*version*/) {
1143 1144 1145 1146
  *value = cfd_->GetTotalSstFilesSize();
  return true;
}

1147 1148 1149 1150 1151 1152
bool InternalStats::HandleLiveSstFilesSize(uint64_t* value, DBImpl* /*db*/,
                                           Version* /*version*/) {
  *value = cfd_->GetLiveSstFilesSize();
  return true;
}

1153
bool InternalStats::HandleEstimatePendingCompactionBytes(uint64_t* value,
A
Andrew Kryczka 已提交
1154 1155
                                                         DBImpl* /*db*/,
                                                         Version* /*version*/) {
1156 1157 1158 1159 1160
  const auto* vstorage = cfd_->current()->storage_info();
  *value = vstorage->estimated_compaction_needed_bytes();
  return true;
}

A
Andrew Kryczka 已提交
1161 1162
bool InternalStats::HandleEstimateTableReadersMem(uint64_t* value,
                                                  DBImpl* /*db*/,
1163 1164 1165 1166 1167
                                                  Version* version) {
  *value = (version == nullptr) ? 0 : version->GetMemoryUsageByTableReaders();
  return true;
}

A
Andrew Kryczka 已提交
1168
bool InternalStats::HandleEstimateLiveDataSize(uint64_t* value, DBImpl* /*db*/,
1169 1170
                                               Version* version) {
  const auto* vstorage = version->storage_info();
1171 1172
  *value = vstorage->EstimateLiveDataSize();
  return true;
I
Igor Canadi 已提交
1173 1174
}

1175
bool InternalStats::HandleMinLogNumberToKeep(uint64_t* value, DBImpl* db,
A
Andrew Kryczka 已提交
1176
                                             Version* /*version*/) {
1177 1178 1179 1180
  *value = db->MinLogNumberToKeep();
  return true;
}

1181 1182 1183 1184 1185 1186 1187
bool InternalStats::HandleMinObsoleteSstNumberToKeep(uint64_t* value,
                                                     DBImpl* db,
                                                     Version* /*version*/) {
  *value = db->MinObsoleteSstNumberToKeep();
  return true;
}

1188
bool InternalStats::HandleActualDelayedWriteRate(uint64_t* value, DBImpl* db,
A
Andrew Kryczka 已提交
1189
                                                 Version* /*version*/) {
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
  const WriteController& wc = db->write_controller();
  if (!wc.NeedsDelay()) {
    *value = 0;
  } else {
    *value = wc.delayed_write_rate();
  }
  return true;
}

bool InternalStats::HandleIsWriteStopped(uint64_t* value, DBImpl* db,
A
Andrew Kryczka 已提交
1200
                                         Version* /*version*/) {
1201 1202 1203 1204
  *value = db->write_controller().IsStopped() ? 1 : 0;
  return true;
}

Y
Yi Wu 已提交
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223
bool InternalStats::HandleEstimateOldestKeyTime(uint64_t* value, DBImpl* /*db*/,
                                                Version* /*version*/) {
  // TODO(yiwu): The property is currently available for fifo compaction
  // with allow_compaction = false. This is because we don't propagate
  // oldest_key_time on compaction.
  if (cfd_->ioptions()->compaction_style != kCompactionStyleFIFO ||
      cfd_->GetCurrentMutableCFOptions()
          ->compaction_options_fifo.allow_compaction) {
    return false;
  }

  TablePropertiesCollection collection;
  auto s = cfd_->current()->GetPropertiesOfAllTables(&collection);
  if (!s.ok()) {
    return false;
  }
  *value = std::numeric_limits<uint64_t>::max();
  for (auto& p : collection) {
    *value = std::min(*value, p.second->oldest_key_time);
1224 1225 1226
    if (*value == 0) {
      break;
    }
Y
Yi Wu 已提交
1227
  }
1228 1229 1230 1231 1232
  if (*value > 0) {
    *value = std::min({cfd_->mem()->ApproximateOldestKeyTime(),
                       cfd_->imm()->ApproximateOldestKeyTime(), *value});
  }
  return *value > 0 && *value < std::numeric_limits<uint64_t>::max();
Y
Yi Wu 已提交
1233 1234
}

1235
bool InternalStats::GetBlockCacheForStats(Cache** block_cache) {
Y
Yi Wu 已提交
1236
  assert(block_cache != nullptr);
1237
  auto* table_factory = cfd_->ioptions()->table_factory.get();
Y
Yi Wu 已提交
1238
  assert(table_factory != nullptr);
1239 1240 1241
  *block_cache =
      table_factory->GetOptions<Cache>(TableFactory::kBlockCacheOpts());
  return *block_cache != nullptr;
Y
Yi Wu 已提交
1242 1243 1244 1245 1246
}

bool InternalStats::HandleBlockCacheCapacity(uint64_t* value, DBImpl* /*db*/,
                                             Version* /*version*/) {
  Cache* block_cache;
1247
  bool ok = GetBlockCacheForStats(&block_cache);
Y
Yi Wu 已提交
1248 1249 1250 1251 1252 1253 1254 1255 1256 1257
  if (!ok) {
    return false;
  }
  *value = static_cast<uint64_t>(block_cache->GetCapacity());
  return true;
}

bool InternalStats::HandleBlockCacheUsage(uint64_t* value, DBImpl* /*db*/,
                                          Version* /*version*/) {
  Cache* block_cache;
1258
  bool ok = GetBlockCacheForStats(&block_cache);
Y
Yi Wu 已提交
1259 1260 1261 1262 1263 1264 1265 1266 1267 1268
  if (!ok) {
    return false;
  }
  *value = static_cast<uint64_t>(block_cache->GetUsage());
  return true;
}

bool InternalStats::HandleBlockCachePinnedUsage(uint64_t* value, DBImpl* /*db*/,
                                                Version* /*version*/) {
  Cache* block_cache;
1269
  bool ok = GetBlockCacheForStats(&block_cache);
Y
Yi Wu 已提交
1270 1271 1272 1273 1274 1275 1276
  if (!ok) {
    return false;
  }
  *value = static_cast<uint64_t>(block_cache->GetPinnedUsage());
  return true;
}

1277 1278 1279
void InternalStats::DumpDBStats(std::string* value) {
  char buf[1000];
  // DB-level stats, only available from default column family
1280
  double seconds_up = (clock_->NowMicros() - started_at_ + 1) / kMicrosInSec;
1281 1282 1283 1284 1285 1286
  double interval_seconds_up = seconds_up - db_stats_snapshot_.seconds_up;
  snprintf(buf, sizeof(buf),
           "\n** DB Stats **\nUptime(secs): %.1f total, %.1f interval\n",
           seconds_up, interval_seconds_up);
  value->append(buf);
  // Cumulative
1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
  uint64_t user_bytes_written =
      GetDBStats(InternalStats::kIntStatsBytesWritten);
  uint64_t num_keys_written =
      GetDBStats(InternalStats::kIntStatsNumKeysWritten);
  uint64_t write_other = GetDBStats(InternalStats::kIntStatsWriteDoneByOther);
  uint64_t write_self = GetDBStats(InternalStats::kIntStatsWriteDoneBySelf);
  uint64_t wal_bytes = GetDBStats(InternalStats::kIntStatsWalFileBytes);
  uint64_t wal_synced = GetDBStats(InternalStats::kIntStatsWalFileSynced);
  uint64_t write_with_wal = GetDBStats(InternalStats::kIntStatsWriteWithWal);
  uint64_t write_stall_micros =
      GetDBStats(InternalStats::kIntStatsWriteStallMicros);
1298

1299 1300 1301
  const int kHumanMicrosLen = 32;
  char human_micros[kHumanMicrosLen];

1302
  // Data
S
sdong 已提交
1303 1304
  // writes: total number of write requests.
  // keys: total number of key updates issued by all the write requests
1305 1306
  // commit groups: number of group commits issued to the DB. Each group can
  //                contain one or more writes.
S
sdong 已提交
1307
  // so writes/keys is the average number of put in multi-put or put
1308
  // writes/groups is the average group commit size.
S
sdong 已提交
1309 1310
  //
  // The format is the same for interval stats.
1311
  snprintf(buf, sizeof(buf),
1312 1313
           "Cumulative writes: %s writes, %s keys, %s commit groups, "
           "%.1f writes per commit group, ingest: %.2f GB, %.2f MB/s\n",
1314 1315 1316
           NumberToHumanString(write_other + write_self).c_str(),
           NumberToHumanString(num_keys_written).c_str(),
           NumberToHumanString(write_self).c_str(),
1317
           (write_other + write_self) / static_cast<double>(write_self + 1),
1318
           user_bytes_written / kGB, user_bytes_written / kMB / seconds_up);
1319 1320 1321
  value->append(buf);
  // WAL
  snprintf(buf, sizeof(buf),
1322 1323 1324 1325
           "Cumulative WAL: %s writes, %s syncs, "
           "%.2f writes per sync, written: %.2f GB, %.2f MB/s\n",
           NumberToHumanString(write_with_wal).c_str(),
           NumberToHumanString(wal_synced).c_str(),
1326
           write_with_wal / static_cast<double>(wal_synced + 1),
1327 1328 1329 1330
           wal_bytes / kGB, wal_bytes / kMB / seconds_up);
  value->append(buf);
  // Stall
  AppendHumanMicros(write_stall_micros, human_micros, kHumanMicrosLen, true);
1331
  snprintf(buf, sizeof(buf), "Cumulative stall: %s, %.1f percent\n",
1332 1333 1334
           human_micros,
           // 10000 = divide by 1M to get secs, then multiply by 100 for pct
           write_stall_micros / 10000.0 / std::max(seconds_up, 0.001));
1335 1336 1337 1338 1339
  value->append(buf);

  // Interval
  uint64_t interval_write_other = write_other - db_stats_snapshot_.write_other;
  uint64_t interval_write_self = write_self - db_stats_snapshot_.write_self;
S
sdong 已提交
1340 1341
  uint64_t interval_num_keys_written =
      num_keys_written - db_stats_snapshot_.num_keys_written;
1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354
  snprintf(
      buf, sizeof(buf),
      "Interval writes: %s writes, %s keys, %s commit groups, "
      "%.1f writes per commit group, ingest: %.2f MB, %.2f MB/s\n",
      NumberToHumanString(interval_write_other + interval_write_self).c_str(),
      NumberToHumanString(interval_num_keys_written).c_str(),
      NumberToHumanString(interval_write_self).c_str(),
      static_cast<double>(interval_write_other + interval_write_self) /
          (interval_write_self + 1),
      (user_bytes_written - db_stats_snapshot_.ingest_bytes) / kMB,
      (user_bytes_written - db_stats_snapshot_.ingest_bytes) / kMB /
          std::max(interval_seconds_up, 0.001)),
      value->append(buf);
1355 1356 1357 1358 1359 1360

  uint64_t interval_write_with_wal =
      write_with_wal - db_stats_snapshot_.write_with_wal;
  uint64_t interval_wal_synced = wal_synced - db_stats_snapshot_.wal_synced;
  uint64_t interval_wal_bytes = wal_bytes - db_stats_snapshot_.wal_bytes;

1361 1362 1363
  snprintf(
      buf, sizeof(buf),
      "Interval WAL: %s writes, %s syncs, "
1364
      "%.2f writes per sync, written: %.2f GB, %.2f MB/s\n",
1365 1366 1367 1368 1369
      NumberToHumanString(interval_write_with_wal).c_str(),
      NumberToHumanString(interval_wal_synced).c_str(),
      interval_write_with_wal / static_cast<double>(interval_wal_synced + 1),
      interval_wal_bytes / kGB,
      interval_wal_bytes / kMB / std::max(interval_seconds_up, 0.001));
1370 1371 1372
  value->append(buf);

  // Stall
1373 1374 1375
  AppendHumanMicros(write_stall_micros - db_stats_snapshot_.write_stall_micros,
                    human_micros, kHumanMicrosLen, true);
  snprintf(buf, sizeof(buf), "Interval stall: %s, %.1f percent\n", human_micros,
1376 1377 1378
           // 10000 = divide by 1M to get secs, then multiply by 100 for pct
           (write_stall_micros - db_stats_snapshot_.write_stall_micros) /
               10000.0 / std::max(interval_seconds_up, 0.001));
1379 1380 1381 1382 1383 1384
  value->append(buf);

  db_stats_snapshot_.seconds_up = seconds_up;
  db_stats_snapshot_.ingest_bytes = user_bytes_written;
  db_stats_snapshot_.write_other = write_other;
  db_stats_snapshot_.write_self = write_self;
S
sdong 已提交
1385
  db_stats_snapshot_.num_keys_written = num_keys_written;
1386 1387 1388
  db_stats_snapshot_.wal_bytes = wal_bytes;
  db_stats_snapshot_.wal_synced = wal_synced;
  db_stats_snapshot_.write_with_wal = write_with_wal;
S
sdong 已提交
1389
  db_stats_snapshot_.write_stall_micros = write_stall_micros;
1390 1391
}

1392
/**
1393 1394 1395 1396 1397 1398
 * Dump Compaction Level stats to a map of stat name with "compaction." prefix
 * to value in double as string. The level in stat name is represented with
 * a prefix "Lx" where "x" is the level number. A special level "Sum"
 * represents the sum of a stat for all levels.
 * The result also contains IO stall counters which keys start with "io_stalls."
 * and values represent uint64 encoded as strings.
1399
 */
1400
void InternalStats::DumpCFMapStats(
1401
    std::map<std::string, std::string>* cf_stats) {
1402
  const VersionStorageInfo* vstorage = cfd_->current()->storage_info();
1403
  CompactionStats compaction_stats_sum;
1404
  std::map<int, std::map<LevelStatType, double>> levels_stats;
1405
  DumpCFMapStats(vstorage, &levels_stats, &compaction_stats_sum);
1406 1407 1408 1409 1410 1411
  for (auto const& level_ent : levels_stats) {
    auto level_str =
        level_ent.first == -1 ? "Sum" : "L" + ToString(level_ent.first);
    for (auto const& stat_ent : level_ent.second) {
      auto stat_type = stat_ent.first;
      auto key_str =
1412
          "compaction." + level_str + "." +
1413
          InternalStats::compaction_level_stats.at(stat_type).property_name;
1414
      (*cf_stats)[key_str] = std::to_string(stat_ent.second);
1415 1416
    }
  }
1417 1418

  DumpCFMapStatsIOStalls(cf_stats);
1419 1420
}

1421
void InternalStats::DumpCFMapStats(
1422
    const VersionStorageInfo* vstorage,
1423 1424
    std::map<int, std::map<LevelStatType, double>>* levels_stats,
    CompactionStats* compaction_stats_sum) {
1425
  assert(vstorage);
1426 1427

  int num_levels_to_check =
1428
      (cfd_->ioptions()->compaction_style != kCompactionStyleFIFO)
1429
          ? vstorage->num_levels() - 1
1430
          : 1;
S
sdong 已提交
1431

1432
  // Compaction scores are sorted based on its value. Restore them to the
1433 1434 1435
  // level order
  std::vector<double> compaction_score(number_levels_, 0);
  for (int i = 0; i < num_levels_to_check; ++i) {
1436 1437
    compaction_score[vstorage->CompactionScoreLevel(i)] =
        vstorage->CompactionScore(i);
1438 1439 1440
  }
  // Count # of files being compacted for each level
  std::vector<int> files_being_compacted(number_levels_, 0);
1441
  for (int level = 0; level < number_levels_; ++level) {
1442
    for (auto* f : vstorage->LevelFiles(level)) {
1443 1444 1445 1446 1447 1448 1449 1450 1451
      if (f->being_compacted) {
        ++files_being_compacted[level];
      }
    }
  }

  int total_files = 0;
  int total_files_being_compacted = 0;
  double total_file_size = 0;
1452 1453 1454
  uint64_t flush_ingest = cf_stats_value_[BYTES_FLUSHED];
  uint64_t add_file_ingest = cf_stats_value_[BYTES_INGESTED_ADD_FILE];
  uint64_t curr_ingest = flush_ingest + add_file_ingest;
1455
  for (int level = 0; level < number_levels_; level++) {
S
sdong 已提交
1456
    int files = vstorage->NumLevelFiles(level);
1457 1458 1459
    total_files += files;
    total_files_being_compacted += files_being_compacted[level];
    if (comp_stats_[level].micros > 0 || files > 0) {
1460
      compaction_stats_sum->Add(comp_stats_[level]);
S
sdong 已提交
1461
      total_file_size += vstorage->NumLevelBytes(level);
1462 1463 1464 1465
      uint64_t input_bytes;
      if (level == 0) {
        input_bytes = curr_ingest;
      } else {
1466 1467
        input_bytes = comp_stats_[level].bytes_read_non_output_levels +
                      comp_stats_[level].bytes_read_blob;
1468
      }
1469
      double w_amp =
1470
          (input_bytes == 0)
1471
              ? 0.0
1472 1473
              : static_cast<double>(comp_stats_[level].bytes_written +
                                    comp_stats_[level].bytes_written_blob) /
1474
                    input_bytes;
1475 1476 1477 1478 1479
      std::map<LevelStatType, double> level_stats;
      PrepareLevelStats(&level_stats, files, files_being_compacted[level],
                        static_cast<double>(vstorage->NumLevelBytes(level)),
                        compaction_score[level], w_amp, comp_stats_[level]);
      (*levels_stats)[level] = level_stats;
1480 1481
    }
  }
1482
  // Cumulative summary
1483 1484
  double w_amp = (compaction_stats_sum->bytes_written +
                  compaction_stats_sum->bytes_written_blob) /
1485 1486 1487 1488 1489 1490 1491 1492
                 static_cast<double>(curr_ingest + 1);
  // Stats summary across levels
  std::map<LevelStatType, double> sum_stats;
  PrepareLevelStats(&sum_stats, total_files, total_files_being_compacted,
                    total_file_size, 0, w_amp, *compaction_stats_sum);
  (*levels_stats)[-1] = sum_stats;  //  -1 is for the Sum level
}

1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506
void InternalStats::DumpCFMapStatsByPriority(
    std::map<int, std::map<LevelStatType, double>>* priorities_stats) {
  for (size_t priority = 0; priority < comp_stats_by_pri_.size(); priority++) {
    if (comp_stats_by_pri_[priority].micros > 0) {
      std::map<LevelStatType, double> priority_stats;
      PrepareLevelStats(&priority_stats, 0 /* num_files */,
                        0 /* being_compacted */, 0 /* total_file_size */,
                        0 /* compaction_score */, 0 /* w_amp */,
                        comp_stats_by_pri_[priority]);
      (*priorities_stats)[static_cast<int>(priority)] = priority_stats;
    }
  }
}

1507 1508 1509
void InternalStats::DumpCFMapStatsIOStalls(
    std::map<std::string, std::string>* cf_stats) {
  (*cf_stats)["io_stalls.level0_slowdown"] =
1510
      std::to_string(cf_stats_count_[L0_FILE_COUNT_LIMIT_SLOWDOWNS]);
1511
  (*cf_stats)["io_stalls.level0_slowdown_with_compaction"] =
1512
      std::to_string(cf_stats_count_[LOCKED_L0_FILE_COUNT_LIMIT_SLOWDOWNS]);
1513
  (*cf_stats)["io_stalls.level0_numfiles"] =
1514
      std::to_string(cf_stats_count_[L0_FILE_COUNT_LIMIT_STOPS]);
1515
  (*cf_stats)["io_stalls.level0_numfiles_with_compaction"] =
1516
      std::to_string(cf_stats_count_[LOCKED_L0_FILE_COUNT_LIMIT_STOPS]);
1517
  (*cf_stats)["io_stalls.stop_for_pending_compaction_bytes"] =
1518
      std::to_string(cf_stats_count_[PENDING_COMPACTION_BYTES_LIMIT_STOPS]);
1519
  (*cf_stats)["io_stalls.slowdown_for_pending_compaction_bytes"] =
1520
      std::to_string(cf_stats_count_[PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS]);
1521
  (*cf_stats)["io_stalls.memtable_compaction"] =
1522
      std::to_string(cf_stats_count_[MEMTABLE_LIMIT_STOPS]);
1523
  (*cf_stats)["io_stalls.memtable_slowdown"] =
1524
      std::to_string(cf_stats_count_[MEMTABLE_LIMIT_SLOWDOWNS]);
1525

1526 1527 1528
  uint64_t total_stop = cf_stats_count_[L0_FILE_COUNT_LIMIT_STOPS] +
                        cf_stats_count_[PENDING_COMPACTION_BYTES_LIMIT_STOPS] +
                        cf_stats_count_[MEMTABLE_LIMIT_STOPS];
1529 1530

  uint64_t total_slowdown =
1531 1532 1533
      cf_stats_count_[L0_FILE_COUNT_LIMIT_SLOWDOWNS] +
      cf_stats_count_[PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS] +
      cf_stats_count_[MEMTABLE_LIMIT_SLOWDOWNS];
1534 1535 1536 1537 1538

  (*cf_stats)["io_stalls.total_stop"] = std::to_string(total_stop);
  (*cf_stats)["io_stalls.total_slowdown"] = std::to_string(total_slowdown);
}

1539
void InternalStats::DumpCFStats(std::string* value) {
1540 1541 1542 1543 1544
  DumpCFStatsNoFileHistogram(value);
  DumpCFFileHistogram(value);
}

void InternalStats::DumpCFStatsNoFileHistogram(std::string* value) {
J
Jonathan Lee 已提交
1545
  char buf[2000];
1546
  // Per-ColumnFamily stats
1547
  PrintLevelStatsHeader(buf, sizeof(buf), cfd_->GetName(), "Level");
1548 1549 1550
  value->append(buf);

  // Print stats for each level
1551
  const VersionStorageInfo* vstorage = cfd_->current()->storage_info();
1552
  std::map<int, std::map<LevelStatType, double>> levels_stats;
1553
  CompactionStats compaction_stats_sum;
1554
  DumpCFMapStats(vstorage, &levels_stats, &compaction_stats_sum);
1555
  for (int l = 0; l < number_levels_; ++l) {
1556 1557 1558 1559 1560
    if (levels_stats.find(l) != levels_stats.end()) {
      PrintLevelStats(buf, sizeof(buf), "L" + ToString(l), levels_stats[l]);
      value->append(buf);
    }
  }
1561

1562 1563 1564 1565 1566 1567
  // Print sum of level stats
  PrintLevelStats(buf, sizeof(buf), "Sum", levels_stats[-1]);
  value->append(buf);

  uint64_t flush_ingest = cf_stats_value_[BYTES_FLUSHED];
  uint64_t add_file_ingest = cf_stats_value_[BYTES_INGESTED_ADD_FILE];
1568 1569 1570 1571
  uint64_t ingest_files_addfile = cf_stats_value_[INGESTED_NUM_FILES_TOTAL];
  uint64_t ingest_l0_files_addfile =
      cf_stats_value_[INGESTED_LEVEL0_NUM_FILES_TOTAL];
  uint64_t ingest_keys_addfile = cf_stats_value_[INGESTED_NUM_KEYS_TOTAL];
1572
  // Cumulative summary
1573
  uint64_t total_stall_count =
1574 1575 1576 1577 1578 1579
      cf_stats_count_[L0_FILE_COUNT_LIMIT_SLOWDOWNS] +
      cf_stats_count_[L0_FILE_COUNT_LIMIT_STOPS] +
      cf_stats_count_[PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS] +
      cf_stats_count_[PENDING_COMPACTION_BYTES_LIMIT_STOPS] +
      cf_stats_count_[MEMTABLE_LIMIT_STOPS] +
      cf_stats_count_[MEMTABLE_LIMIT_SLOWDOWNS];
1580
  // Interval summary
1581 1582 1583
  uint64_t interval_flush_ingest =
      flush_ingest - cf_stats_snapshot_.ingest_bytes_flush;
  uint64_t interval_add_file_inget =
1584
      add_file_ingest - cf_stats_snapshot_.ingest_bytes_addfile;
1585
  uint64_t interval_ingest =
1586
      interval_flush_ingest + interval_add_file_inget + 1;
1587
  CompactionStats interval_stats(compaction_stats_sum);
1588
  interval_stats.Subtract(cf_stats_snapshot_.comp_stats);
1589
  double w_amp =
1590 1591
      (interval_stats.bytes_written + interval_stats.bytes_written_blob) /
      static_cast<double>(interval_ingest);
1592
  PrintLevelStats(buf, sizeof(buf), "Int", 0, 0, 0, 0, w_amp, interval_stats);
1593 1594
  value->append(buf);

1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609
  PrintLevelStatsHeader(buf, sizeof(buf), cfd_->GetName(), "Priority");
  value->append(buf);
  std::map<int, std::map<LevelStatType, double>> priorities_stats;
  DumpCFMapStatsByPriority(&priorities_stats);
  for (size_t priority = 0; priority < comp_stats_by_pri_.size(); ++priority) {
    if (priorities_stats.find(static_cast<int>(priority)) !=
        priorities_stats.end()) {
      PrintLevelStats(
          buf, sizeof(buf),
          Env::PriorityToString(static_cast<Env::Priority>(priority)),
          priorities_stats[static_cast<int>(priority)]);
      value->append(buf);
    }
  }

1610 1611 1612 1613 1614 1615
  snprintf(buf, sizeof(buf),
           "\nBlob file count: %" ROCKSDB_PRIszt ", total size: %.1f GB\n\n",
           vstorage->GetBlobFiles().size(),
           vstorage->GetTotalBlobFileSize() / kGB);
  value->append(buf);

1616 1617
  uint64_t now_micros = clock_->NowMicros();
  double seconds_up = (now_micros - started_at_ + 1) / kMicrosInSec;
1618 1619 1620 1621
  double interval_seconds_up = seconds_up - cf_stats_snapshot_.seconds_up;
  snprintf(buf, sizeof(buf), "Uptime(secs): %.1f total, %.1f interval\n",
           seconds_up, interval_seconds_up);
  value->append(buf);
1622 1623
  snprintf(buf, sizeof(buf), "Flush(GB): cumulative %.3f, interval %.3f\n",
           flush_ingest / kGB, interval_flush_ingest / kGB);
1624
  value->append(buf);
1625 1626
  snprintf(buf, sizeof(buf), "AddFile(GB): cumulative %.3f, interval %.3f\n",
           add_file_ingest / kGB, interval_add_file_inget / kGB);
1627 1628
  value->append(buf);

1629 1630
  uint64_t interval_ingest_files_addfile =
      ingest_files_addfile - cf_stats_snapshot_.ingest_files_addfile;
1631 1632 1633
  snprintf(buf, sizeof(buf),
           "AddFile(Total Files): cumulative %" PRIu64 ", interval %" PRIu64
           "\n",
1634 1635 1636 1637 1638 1639
           ingest_files_addfile, interval_ingest_files_addfile);
  value->append(buf);

  uint64_t interval_ingest_l0_files_addfile =
      ingest_l0_files_addfile - cf_stats_snapshot_.ingest_l0_files_addfile;
  snprintf(buf, sizeof(buf),
I
Islam AbdelRahman 已提交
1640
           "AddFile(L0 Files): cumulative %" PRIu64 ", interval %" PRIu64 "\n",
1641 1642 1643 1644 1645
           ingest_l0_files_addfile, interval_ingest_l0_files_addfile);
  value->append(buf);

  uint64_t interval_ingest_keys_addfile =
      ingest_keys_addfile - cf_stats_snapshot_.ingest_keys_addfile;
I
Islam AbdelRahman 已提交
1646 1647
  snprintf(buf, sizeof(buf),
           "AddFile(Keys): cumulative %" PRIu64 ", interval %" PRIu64 "\n",
1648 1649 1650
           ingest_keys_addfile, interval_ingest_keys_addfile);
  value->append(buf);

1651 1652 1653 1654 1655 1656
  // Compact
  uint64_t compact_bytes_read = 0;
  uint64_t compact_bytes_write = 0;
  uint64_t compact_micros = 0;
  for (int level = 0; level < number_levels_; level++) {
    compact_bytes_read += comp_stats_[level].bytes_read_output_level +
1657 1658
                          comp_stats_[level].bytes_read_non_output_levels +
                          comp_stats_[level].bytes_read_blob;
1659 1660
    compact_bytes_write += comp_stats_[level].bytes_written +
                           comp_stats_[level].bytes_written_blob;
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
    compact_micros += comp_stats_[level].micros;
  }

  snprintf(buf, sizeof(buf),
           "Cumulative compaction: %.2f GB write, %.2f MB/s write, "
           "%.2f GB read, %.2f MB/s read, %.1f seconds\n",
           compact_bytes_write / kGB, compact_bytes_write / kMB / seconds_up,
           compact_bytes_read / kGB, compact_bytes_read / kMB / seconds_up,
           compact_micros / kMicrosInSec);
  value->append(buf);

  // Compaction interval
  uint64_t interval_compact_bytes_write =
      compact_bytes_write - cf_stats_snapshot_.compact_bytes_write;
  uint64_t interval_compact_bytes_read =
      compact_bytes_read - cf_stats_snapshot_.compact_bytes_read;
  uint64_t interval_compact_micros =
      compact_micros - cf_stats_snapshot_.compact_micros;

  snprintf(
      buf, sizeof(buf),
      "Interval compaction: %.2f GB write, %.2f MB/s write, "
      "%.2f GB read, %.2f MB/s read, %.1f seconds\n",
      interval_compact_bytes_write / kGB,
      interval_compact_bytes_write / kMB / std::max(interval_seconds_up, 0.001),
      interval_compact_bytes_read / kGB,
      interval_compact_bytes_read / kMB / std::max(interval_seconds_up, 0.001),
      interval_compact_micros / kMicrosInSec);
  value->append(buf);
  cf_stats_snapshot_.compact_bytes_write = compact_bytes_write;
  cf_stats_snapshot_.compact_bytes_read = compact_bytes_read;
  cf_stats_snapshot_.compact_micros = compact_micros;

1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719
  snprintf(buf, sizeof(buf),
           "Stalls(count): %" PRIu64
           " level0_slowdown, "
           "%" PRIu64
           " level0_slowdown_with_compaction, "
           "%" PRIu64
           " level0_numfiles, "
           "%" PRIu64
           " level0_numfiles_with_compaction, "
           "%" PRIu64
           " stop for pending_compaction_bytes, "
           "%" PRIu64
           " slowdown for pending_compaction_bytes, "
           "%" PRIu64
           " memtable_compaction, "
           "%" PRIu64
           " memtable_slowdown, "
           "interval %" PRIu64 " total count\n",
           cf_stats_count_[L0_FILE_COUNT_LIMIT_SLOWDOWNS],
           cf_stats_count_[LOCKED_L0_FILE_COUNT_LIMIT_SLOWDOWNS],
           cf_stats_count_[L0_FILE_COUNT_LIMIT_STOPS],
           cf_stats_count_[LOCKED_L0_FILE_COUNT_LIMIT_STOPS],
           cf_stats_count_[PENDING_COMPACTION_BYTES_LIMIT_STOPS],
           cf_stats_count_[PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS],
           cf_stats_count_[MEMTABLE_LIMIT_STOPS],
           cf_stats_count_[MEMTABLE_LIMIT_SLOWDOWNS],
1720
           total_stall_count - cf_stats_snapshot_.stall_count);
1721 1722
  value->append(buf);

1723
  cf_stats_snapshot_.seconds_up = seconds_up;
1724
  cf_stats_snapshot_.ingest_bytes_flush = flush_ingest;
1725 1726 1727 1728
  cf_stats_snapshot_.ingest_bytes_addfile = add_file_ingest;
  cf_stats_snapshot_.ingest_files_addfile = ingest_files_addfile;
  cf_stats_snapshot_.ingest_l0_files_addfile = ingest_l0_files_addfile;
  cf_stats_snapshot_.ingest_keys_addfile = ingest_keys_addfile;
1729
  cf_stats_snapshot_.comp_stats = compaction_stats_sum;
1730
  cf_stats_snapshot_.stall_count = total_stall_count;
1731

1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745
  // Do not gather cache entry stats during CFStats because DB
  // mutex is held. Only dump last cached collection (rely on DB
  // periodic stats dump to update)
  if (cache_entry_stats_collector_) {
    CacheEntryRoleStats stats;
    // thread safe
    cache_entry_stats_collector_->GetStats(&stats);

    constexpr uint64_t kDayInMicros = uint64_t{86400} * 1000000U;

    // Skip if stats are extremely old (> 1 day, incl not yet populated)
    if (now_micros - stats.last_end_time_micros_ < kDayInMicros) {
      value->append(stats.ToString(clock_));
    }
1746
  }
1747 1748
}

1749
void InternalStats::DumpCFFileHistogram(std::string* value) {
1750 1751 1752 1753 1754 1755
  assert(value);
  assert(cfd_);

  std::ostringstream oss;
  oss << "\n** File Read Latency Histogram By Level [" << cfd_->GetName()
      << "] **\n";
1756 1757 1758

  for (int level = 0; level < number_levels_; level++) {
    if (!file_read_latency_[level].Empty()) {
1759 1760
      oss << "** Level " << level << " read latency histogram (micros):\n"
          << file_read_latency_[level].ToString() << '\n';
1761 1762
    }
  }
1763 1764 1765 1766 1767 1768

  if (!blob_file_read_latency_.Empty()) {
    oss << "** Blob file read latency histogram (micros):\n"
        << blob_file_read_latency_.ToString() << '\n';
  }

1769
  value->append(oss.str());
1770
}
1771 1772 1773

#else

1774 1775 1776
const DBPropertyInfo* GetPropertyInfo(const Slice& /*property*/) {
  return nullptr;
}
1777 1778 1779

#endif  // !ROCKSDB_LITE

1780
}  // namespace ROCKSDB_NAMESPACE