db_impl.h 13.3 KB
Newer Older
J
jorlow@chromium.org 已提交
1 2 3 4 5 6 7
// 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.

#ifndef STORAGE_LEVELDB_DB_DB_IMPL_H_
#define STORAGE_LEVELDB_DB_DB_IMPL_H_

H
Haobo Xu 已提交
8
#include <atomic>
9
#include <deque>
J
jorlow@chromium.org 已提交
10
#include <set>
11
#include <vector>
J
jorlow@chromium.org 已提交
12
#include "db/dbformat.h"
13
#include "db/log_file.h"
J
jorlow@chromium.org 已提交
14 15
#include "db/log_writer.h"
#include "db/snapshot.h"
16 17
#include "leveldb/db.h"
#include "leveldb/env.h"
J
jorlow@chromium.org 已提交
18
#include "port/port.h"
19
#include "util/stats_logger.h"
20
#include "memtablelist.h"
21 22 23 24 25

#ifdef USE_SCRIBE
#include "scribe/scribe_logger.h"
#endif

J
jorlow@chromium.org 已提交
26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
namespace leveldb {

class MemTable;
class TableCache;
class Version;
class VersionEdit;
class VersionSet;

class DBImpl : public DB {
 public:
  DBImpl(const Options& options, const std::string& dbname);
  virtual ~DBImpl();

  // Implementations of the DB interface
  virtual Status Put(const WriteOptions&, const Slice& key, const Slice& value);
41 42
  virtual Status Merge(const WriteOptions&, const Slice& key,
                       const Slice& value);
J
jorlow@chromium.org 已提交
43 44 45 46 47
  virtual Status Delete(const WriteOptions&, const Slice& key);
  virtual Status Write(const WriteOptions& options, WriteBatch* updates);
  virtual Status Get(const ReadOptions& options,
                     const Slice& key,
                     std::string* value);
48 49 50
  virtual std::vector<Status> MultiGet(const ReadOptions& options,
                                       const std::vector<Slice>& keys,
                                       std::vector<std::string>* values);
J
jorlow@chromium.org 已提交
51 52 53
  virtual Iterator* NewIterator(const ReadOptions&);
  virtual const Snapshot* GetSnapshot();
  virtual void ReleaseSnapshot(const Snapshot* snapshot);
54
  virtual bool GetProperty(const Slice& property, std::string* value);
J
jorlow@chromium.org 已提交
55
  virtual void GetApproximateSizes(const Range* range, int n, uint64_t* sizes);
G
Gabor Cselle 已提交
56
  virtual void CompactRange(const Slice* begin, const Slice* end);
57 58 59
  virtual int NumberLevels();
  virtual int MaxMemCompactionLevel();
  virtual int Level0StopWriteTrigger();
H
heyongqiang 已提交
60
  virtual Status Flush(const FlushOptions& options);
61 62
  virtual Status DisableFileDeletions();
  virtual Status EnableFileDeletions();
63
  virtual Status GetLiveFiles(std::vector<std::string>&,
64
                              uint64_t* manifest_file_size);
65
  virtual SequenceNumber GetLatestSequenceNumber();
66
  virtual Status GetUpdatesSince(SequenceNumber seq_number,
67
                                 unique_ptr<TransactionLogIterator>* iter);
68

J
jorlow@chromium.org 已提交
69 70
  // Extra methods (for testing) that are not in the public DB interface

71
  // Compact any files in the named level that overlap [*begin, *end]
G
Gabor Cselle 已提交
72
  void TEST_CompactRange(int level, const Slice* begin, const Slice* end);
J
jorlow@chromium.org 已提交
73 74 75 76

  // Force current memtable contents to be compacted.
  Status TEST_CompactMemTable();

77 78 79 80 81 82
  // Wait for memtable compaction
  Status TEST_WaitForCompactMemTable();

  // Wait for any compaction
  Status TEST_WaitForCompact();

J
jorlow@chromium.org 已提交
83 84 85 86 87
  // Return an internal iterator over the current state of the database.
  // The keys of this iterator are internal keys (see format.h).
  // The returned iterator should be deleted when no longer needed.
  Iterator* TEST_NewInternalIterator();

88 89
  // Return the maximum overlapping data (in bytes) at next level for any
  // file at a level >= 1.
J
jorlow@chromium.org 已提交
90
  int64_t TEST_MaxNextLevelOverlappingBytes();
91

92 93 94
  // Simulate a db crash, no elegant closing of database.
  void TEST_Destroy_DBImpl();

A
Abhishek Kona 已提交
95 96
  // Return the current manifest file no.
  uint64_t TEST_Current_Manifest_FileNo();
97 98 99 100

  // Trigger's a background call for testing.
  void TEST_PurgeObsoleteteWAL();

101
 protected:
H
heyongqiang 已提交
102 103
  Env* const env_;
  const std::string dbname_;
104
  unique_ptr<VersionSet> versions_;
H
heyongqiang 已提交
105 106 107 108 109 110
  const InternalKeyComparator internal_comparator_;
  const Options options_;  // options_.comparator == &internal_comparator_

  const Comparator* user_comparator() const {
    return internal_comparator_.user_comparator();
  }
111

112 113
  MemTable* GetMemTable() {
    return mem_;
A
Abhishek Kona 已提交
114
  }
H
heyongqiang 已提交
115

116 117 118
  Iterator* NewInternalIterator(const ReadOptions&,
                                SequenceNumber* latest_snapshot);

J
jorlow@chromium.org 已提交
119 120
 private:
  friend class DB;
121 122
  struct CompactionState;
  struct Writer;
D
Dhruba Borthakur 已提交
123
  struct DeletionState;
J
jorlow@chromium.org 已提交
124 125 126 127 128 129

  Status NewDB();

  // Recover the descriptor from persistent storage.  May do a significant
  // amount of work to recover recently logged updates.  Any changes to
  // be made to the descriptor are added to *edit.
130
  Status Recover(VersionEdit* edit, MemTable* external_table = nullptr,
H
heyongqiang 已提交
131
      bool error_if_log_file_exist = false);
J
jorlow@chromium.org 已提交
132 133 134

  void MaybeIgnoreError(Status* s) const;

135 136
  const Status CreateArchivalDirectory();

J
jorlow@chromium.org 已提交
137 138 139 140 141
  // Delete any unneeded files and stale in-memory entries.
  void DeleteObsoleteFiles();

  // Compact the in-memory write buffer to disk.  Switches to a new
  // log-file/memtable and writes a new descriptor iff successful.
142
  Status CompactMemTable(bool* madeProgress = nullptr);
J
jorlow@chromium.org 已提交
143 144 145

  Status RecoverLogFile(uint64_t log_number,
                        VersionEdit* edit,
146 147
                        SequenceNumber* max_sequence,
                        MemTable* external_table);
J
jorlow@chromium.org 已提交
148

149 150 151 152 153 154 155 156
  // The following two methods are used to flush a memtable to
  // storage. The first one is used atdatabase RecoveryTime (when the
  // database is opened) and is heavyweight because it holds the mutex
  // for the entire period. The second method WriteLevel0Table supports
  // concurrent flush memtables to storage.
  Status WriteLevel0TableForRecovery(MemTable* mem, VersionEdit* edit);
  Status WriteLevel0Table(MemTable* mem, VersionEdit* edit,
                                uint64_t* filenumber);
J
jorlow@chromium.org 已提交
157

158
  Status MakeRoomForWrite(bool force /* compact even if there is room? */);
159
  WriteBatch* BuildBatchGroup(Writer** last_writer);
J
jorlow@chromium.org 已提交
160

H
heyongqiang 已提交
161 162 163 164 165 166
  // Force current memtable contents to be flushed.
  Status FlushMemTable(const FlushOptions& options);

  // Wait for memtable compaction
  Status WaitForCompactMemTable();

167
  void MaybeScheduleLogDBDeployStats();
168 169
  static void BGLogDBDeployStats(void* db);
  void LogDBDeployStats();
170

J
jorlow@chromium.org 已提交
171 172 173
  void MaybeScheduleCompaction();
  static void BGWork(void* db);
  void BackgroundCall();
174
  Status BackgroundCompaction(bool* madeProgress, DeletionState& deletion_state);
J
jorlow@chromium.org 已提交
175 176 177 178 179 180
  void CleanupCompaction(CompactionState* compact);
  Status DoCompactionWork(CompactionState* compact);

  Status OpenCompactionOutputFile(CompactionState* compact);
  Status FinishCompactionOutputFile(CompactionState* compact, Iterator* input);
  Status InstallCompactionResults(CompactionState* compact);
181 182
  void AllocateCompactionOutputFileNumbers(CompactionState* compact);
  void ReleaseCompactionUnusedFileNumbers(CompactionState* compact);
A
Abhishek Kona 已提交
183

J
jorlow@chromium.org 已提交
184

D
Dhruba Borthakur 已提交
185 186 187 188 189 190 191 192 193 194 195 196
  // Returns the list of live files in 'live' and the list
  // of all files in the filesystem in 'allfiles'.
  void FindObsoleteFiles(DeletionState& deletion_state);

  // Diffs the files listed in filenames and those that do not
  // belong to live files are posibly removed. If the removed file
  // is a sst file, then it returns the file number in files_to_evict.
  void PurgeObsoleteFiles(DeletionState& deletion_state);

  // Removes the file listed in files_to_evict from the table_cache
  void EvictObsoleteFiles(DeletionState& deletion_state);

197
  void PurgeObsoleteWALFiles();
198 199 200 201 202 203 204 205 206 207

  Status ListAllWALFiles(const std::string& path,
                         std::vector<LogFile>* logFiles,
                         WalFileType type);

  //  Find's all the log files which contain updates with seq no.
  //  Greater Than or Equal to the requested SequenceNumber
  Status FindProbableWALFiles(std::vector<LogFile>* const allLogs,
                              std::vector<LogFile>* const result,
                              const SequenceNumber target);
208 209
  //  return true if
  bool CheckFileExistsAndEmpty(const LogFile& file);
210 211 212 213

  Status ReadFirstRecord(const LogFile& file, WriteBatch* const result);

  Status ReadFirstLine(const std::string& fname, WriteBatch* const batch);
214

215 216 217 218
  void PrintHistogram(Histograms histogram_type, std::string name);

  void PrintStatistics();

219 220 221
  // dump leveldb.stats to LOG
  void MaybeDumpStats();

J
jorlow@chromium.org 已提交
222
  // Constant after construction
S
Sanjay Ghemawat 已提交
223
  const InternalFilterPolicy internal_filter_policy_;
J
jorlow@chromium.org 已提交
224 225 226
  bool owns_info_log_;

  // table_cache_ provides its own synchronization
227
  unique_ptr<TableCache> table_cache_;
J
jorlow@chromium.org 已提交
228

229
  // Lock over the persistent DB state.  Non-nullptr iff successfully acquired.
J
jorlow@chromium.org 已提交
230 231 232 233 234
  FileLock* db_lock_;

  // State below is protected by mutex_
  port::Mutex mutex_;
  port::AtomicPointer shutting_down_;
H
hans@chromium.org 已提交
235
  port::CondVar bg_cv_;          // Signalled when background work finishes
J
jorlow@chromium.org 已提交
236
  MemTable* mem_;
237
  MemTableList imm_;             // Memtable that are not changing
238
  uint64_t logfile_number_;
239
  unique_ptr<log::Writer> log_;
240

241 242
  std::string host_name_;

243 244
  // Queue of writers.
  std::deque<Writer*> writers_;
245
  WriteBatch tmp_batch_;
246

J
jorlow@chromium.org 已提交
247 248 249 250 251 252
  SnapshotList snapshots_;

  // Set of table files to protect from deletion because they are
  // part of ongoing compactions.
  std::set<uint64_t> pending_outputs_;

253 254
  // count how many background compaction been scheduled or is running?
  int bg_compaction_scheduled_;
J
jorlow@chromium.org 已提交
255

256 257 258
  // Has a background stats log thread scheduled?
  bool bg_logstats_scheduled_;

H
hans@chromium.org 已提交
259 260 261
  // Information for a manual compaction
  struct ManualCompaction {
    int level;
G
Gabor Cselle 已提交
262
    bool done;
263
    bool in_progress;           // compaction request being processed?
264 265
    const InternalKey* begin;   // nullptr means beginning of key range
    const InternalKey* end;     // nullptr means end of key range
G
Gabor Cselle 已提交
266
    InternalKey tmp_storage;    // Used to keep track of compaction progress
H
hans@chromium.org 已提交
267 268
  };
  ManualCompaction* manual_compaction_;
J
jorlow@chromium.org 已提交
269 270 271 272

  // Have we encountered a background error in paranoid mode?
  Status bg_error_;

273
  std::unique_ptr<StatsLogger> logger_;
274

275
  int64_t volatile last_log_ts;
276

277 278 279
  // shall we disable deletion of obsolete files
  bool disable_delete_obsolete_files_;

280 281 282
  // last time when DeleteObsoleteFiles was invoked
  uint64_t delete_obsolete_files_last_run_;

283 284 285
  // last time when PurgeObsoleteWALFiles ran.
  uint64_t purge_wal_files_last_run_;

286
  // last time stats were dumped to LOG
H
Haobo Xu 已提交
287
  std::atomic<uint64_t> last_stats_dump_time_microsec_;
288

M
Mark Callaghan 已提交
289 290 291 292
  // These count the number of microseconds for which MakeRoomForWrite stalls.
  uint64_t stall_level0_slowdown_;
  uint64_t stall_memtable_compaction_;
  uint64_t stall_level0_num_files_;
293
  std::vector<uint64_t> stall_leveln_slowdown_;
M
Mark Callaghan 已提交
294 295 296 297

  // Time at which this instance was started.
  const uint64_t started_at_;

298 299
  bool flush_on_destroy_; // Used when disableWAL is true.

300 301 302
  // Per level compaction stats.  stats_[level] stores the stats for
  // compactions that produced data for the specified "level".
  struct CompactionStats {
A
Abhishek Kona 已提交
303
    uint64_t micros;
M
Mark Callaghan 已提交
304 305 306 307 308 309 310 311

    // Bytes read from level N during compaction between levels N and N+1
    int64_t bytes_readn;

    // Bytes read from level N+1 during compaction between levels N and N+1
    int64_t bytes_readnp1;

    // Total bytes written during compaction between levels N and N+1
312 313
    int64_t bytes_written;

M
Mark Callaghan 已提交
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329
    // Files read from level N during compaction between levels N and N+1
    int     files_in_leveln;

    // Files read from level N+1 during compaction between levels N and N+1
    int     files_in_levelnp1;

    // Files written during compaction between levels N and N+1
    int     files_out_levelnp1;

    // Number of compactions done
    int     count;

    CompactionStats() : micros(0), bytes_readn(0), bytes_readnp1(0),
                        bytes_written(0), files_in_leveln(0),
                        files_in_levelnp1(0), files_out_levelnp1(0),
                        count(0) { }
330 331 332

    void Add(const CompactionStats& c) {
      this->micros += c.micros;
M
Mark Callaghan 已提交
333 334
      this->bytes_readn += c.bytes_readn;
      this->bytes_readnp1 += c.bytes_readnp1;
335
      this->bytes_written += c.bytes_written;
M
Mark Callaghan 已提交
336 337 338 339
      this->files_in_leveln += c.files_in_leveln;
      this->files_in_levelnp1 += c.files_in_levelnp1;
      this->files_out_levelnp1 += c.files_out_levelnp1;
      this->count += 1;
340 341
    }
  };
M
Mark Callaghan 已提交
342

343
  std::vector<CompactionStats> stats_;
344

345 346 347 348 349 350 351 352 353 354 355 356 357
  // Used to compute per-interval statistics
  struct StatsSnapshot {
    uint64_t bytes_read_;
    uint64_t bytes_written_;
    uint64_t bytes_new_;
    double   seconds_up_;

    StatsSnapshot() : bytes_read_(0), bytes_written_(0),
                      bytes_new_(0), seconds_up_(0) {}
  };

  StatsSnapshot last_stats_;

H
heyongqiang 已提交
358
  static const int KEEP_LOG_FILE_NUM = 1000;
H
heyongqiang 已提交
359
  std::string db_absolute_path_;
H
heyongqiang 已提交
360

361 362 363
  // count of the number of contiguous delaying writes
  int delayed_writes_;

364 365 366 367
  // store the last flushed sequence.
  // Used by transaction log iterator.
  SequenceNumber last_flushed_sequence_;

368
  // The options to access storage files
H
Haobo Xu 已提交
369
  const EnvOptions storage_options_;
370

J
jorlow@chromium.org 已提交
371 372 373 374
  // No copying allowed
  DBImpl(const DBImpl&);
  void operator=(const DBImpl&);

375 376
  // dump the delayed_writes_ to the log file and reset counter.
  void DelayLoggingAndReset();
377

378 379 380 381 382 383
  // Return the earliest snapshot where seqno is visible.
  // Store the snapshot right before that, if any, in prev_snapshot
  inline SequenceNumber findEarliestVisibleSnapshot(
    SequenceNumber in,
    std::vector<SequenceNumber>& snapshots,
    SequenceNumber* prev_snapshot);
J
jorlow@chromium.org 已提交
384 385 386 387 388 389
};

// Sanitize db options.  The caller should delete result.info_log if
// it is not equal to src.info_log.
extern Options SanitizeOptions(const std::string& db,
                               const InternalKeyComparator* icmp,
S
Sanjay Ghemawat 已提交
390
                               const InternalFilterPolicy* ipolicy,
J
jorlow@chromium.org 已提交
391 392
                               const Options& src);

H
Hans Wennborg 已提交
393
}  // namespace leveldb
J
jorlow@chromium.org 已提交
394 395

#endif  // STORAGE_LEVELDB_DB_DB_IMPL_H_