db_impl.cc 46.7 KB
Newer Older
J
jorlow@chromium.org 已提交
1 2 3 4 5 6 7 8 9 10 11 12
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.

#include "db/db_impl.h"

#include <algorithm>
#include <set>
#include <string>
#include <stdint.h>
#include <stdio.h>
#include <vector>
H
heyongqiang 已提交
13
#include <algorithm>
J
jorlow@chromium.org 已提交
14 15 16 17 18 19 20 21 22 23
#include "db/builder.h"
#include "db/db_iter.h"
#include "db/dbformat.h"
#include "db/filename.h"
#include "db/log_reader.h"
#include "db/log_writer.h"
#include "db/memtable.h"
#include "db/table_cache.h"
#include "db/version_set.h"
#include "db/write_batch_internal.h"
24 25 26 27 28
#include "leveldb/db.h"
#include "leveldb/env.h"
#include "leveldb/status.h"
#include "leveldb/table.h"
#include "leveldb/table_builder.h"
J
jorlow@chromium.org 已提交
29 30 31 32 33 34 35
#include "port/port.h"
#include "table/block.h"
#include "table/merger.h"
#include "table/two_level_iterator.h"
#include "util/coding.h"
#include "util/logging.h"
#include "util/mutexlock.h"
36
#include "util/build_version.h"
J
jorlow@chromium.org 已提交
37 38 39

namespace leveldb {

40 41 42 43 44
// Information kept for every waiting writer
struct DBImpl::Writer {
  Status status;
  WriteBatch* batch;
  bool sync;
H
heyongqiang 已提交
45
  bool disableWAL;
46 47 48 49 50 51
  bool done;
  port::CondVar cv;

  explicit Writer(port::Mutex* mu) : cv(mu) { }
};

J
jorlow@chromium.org 已提交
52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87
struct DBImpl::CompactionState {
  Compaction* const compaction;

  // Sequence numbers < smallest_snapshot are not significant since we
  // will never have to service a snapshot below smallest_snapshot.
  // Therefore if we have seen a sequence number S <= smallest_snapshot,
  // we can drop all entries for the same key with sequence numbers < S.
  SequenceNumber smallest_snapshot;

  // Files produced by compaction
  struct Output {
    uint64_t number;
    uint64_t file_size;
    InternalKey smallest, largest;
  };
  std::vector<Output> outputs;

  // State kept for output being generated
  WritableFile* outfile;
  TableBuilder* builder;

  uint64_t total_bytes;

  Output* current_output() { return &outputs[outputs.size()-1]; }

  explicit CompactionState(Compaction* c)
      : compaction(c),
        outfile(NULL),
        builder(NULL),
        total_bytes(0) {
  }
};

// Fix user-supplied options to be reasonable
template <class T,class V>
static void ClipToRange(T* ptr, V minvalue, V maxvalue) {
D
dgrogan@chromium.org 已提交
88 89
  if (static_cast<V>(*ptr) > maxvalue) *ptr = maxvalue;
  if (static_cast<V>(*ptr) < minvalue) *ptr = minvalue;
J
jorlow@chromium.org 已提交
90 91 92
}
Options SanitizeOptions(const std::string& dbname,
                        const InternalKeyComparator* icmp,
S
Sanjay Ghemawat 已提交
93
                        const InternalFilterPolicy* ipolicy,
J
jorlow@chromium.org 已提交
94 95 96
                        const Options& src) {
  Options result = src;
  result.comparator = icmp;
S
Sanjay Ghemawat 已提交
97 98 99 100
  result.filter_policy = (src.filter_policy != NULL) ? ipolicy : NULL;
  ClipToRange(&result.max_open_files,            20,     50000);
  ClipToRange(&result.write_buffer_size,         64<<10, 1<<30);
  ClipToRange(&result.block_size,                1<<10,  4<<20);
J
jorlow@chromium.org 已提交
101 102 103
  if (result.info_log == NULL) {
    // Open a log file in the same directory as the db
    src.env->CreateDir(dbname);  // In case it does not exist
H
heyongqiang 已提交
104 105
    src.env->RenameFile(InfoLogFileName(dbname),
        OldInfoLogFileName(dbname, src.env->NowMicros()));
106
    Status s = src.env->NewLogger(InfoLogFileName(dbname), &result.info_log);
J
jorlow@chromium.org 已提交
107 108
    if (!s.ok()) {
      // No place suitable for logging
109
      result.info_log = NULL;
J
jorlow@chromium.org 已提交
110 111
    }
  }
112 113 114
  if (result.block_cache == NULL) {
    result.block_cache = NewLRUCache(8 << 20);
  }
J
jorlow@chromium.org 已提交
115 116 117 118 119 120
  return result;
}

DBImpl::DBImpl(const Options& options, const std::string& dbname)
    : env_(options.env),
      internal_comparator_(options.comparator),
S
Sanjay Ghemawat 已提交
121 122 123
      internal_filter_policy_(options.filter_policy),
      options_(SanitizeOptions(
          dbname, &internal_comparator_, &internal_filter_policy_, options)),
J
jorlow@chromium.org 已提交
124
      owns_info_log_(options_.info_log != options.info_log),
125
      owns_cache_(options_.block_cache != options.block_cache),
J
jorlow@chromium.org 已提交
126 127 128 129 130
      dbname_(dbname),
      db_lock_(NULL),
      shutting_down_(NULL),
      bg_cv_(&mutex_),
      mem_(new MemTable(internal_comparator_)),
131
      imm_(NULL),
J
jorlow@chromium.org 已提交
132
      logfile_(NULL),
133
      logfile_number_(0),
J
jorlow@chromium.org 已提交
134
      log_(NULL),
135
      tmp_batch_(new WriteBatch),
J
jorlow@chromium.org 已提交
136
      bg_compaction_scheduled_(false),
137
      bg_logstats_scheduled_(false),
138 139
      manual_compaction_(NULL),
      logger_(NULL) {
140
  mem_->Ref();
141 142
  has_imm_.Release_Store(NULL);

143
  stats_ = new CompactionStats[options.num_levels];
J
jorlow@chromium.org 已提交
144
  // Reserve ten files or so for other uses and give the rest to TableCache.
145
  const int table_cache_size = options_.max_open_files - 10;
J
jorlow@chromium.org 已提交
146 147 148 149
  table_cache_ = new TableCache(dbname_, &options_, table_cache_size);

  versions_ = new VersionSet(dbname_, &options_, table_cache_,
                             &internal_comparator_);
150

151 152
  options_.Dump(options_.info_log);

153 154 155 156 157 158 159 160 161 162 163 164 165
#ifdef USE_SCRIBE
  logger_ = new ScribeLogger("localhost", 1456);
#endif

  char name[100];
  Status st = env_->GetHostName(name, 100);
  if(st.ok()) {
    host_name_ = name;
  } else {
    Log(options_.info_log, "Can't get hostname, use localhost as host name.");
    host_name_ = "localhost";
  }
  last_log_ts = 0;
J
jorlow@chromium.org 已提交
166 167 168 169 170 171
}

DBImpl::~DBImpl() {
  // Wait for background work to finish
  mutex_.Lock();
  shutting_down_.Release_Store(this);  // Any non-NULL value is ok
172
  while (bg_compaction_scheduled_ || bg_logstats_scheduled_) {
H
hans@chromium.org 已提交
173
    bg_cv_.Wait();
J
jorlow@chromium.org 已提交
174 175 176 177 178 179 180 181
  }
  mutex_.Unlock();

  if (db_lock_ != NULL) {
    env_->UnlockFile(db_lock_);
  }

  delete versions_;
182 183
  if (mem_ != NULL) mem_->Unref();
  if (imm_ != NULL) imm_->Unref();
184
  delete tmp_batch_;
J
jorlow@chromium.org 已提交
185 186 187
  delete log_;
  delete logfile_;
  delete table_cache_;
188
  delete[] stats_;
J
jorlow@chromium.org 已提交
189 190 191 192

  if (owns_info_log_) {
    delete options_.info_log;
  }
193 194 195
  if (owns_cache_) {
    delete options_.block_cache;
  }
196 197

  delete logger_;
J
jorlow@chromium.org 已提交
198 199 200
}

Status DBImpl::NewDB() {
201
  VersionEdit new_db(NumberLevels());
J
jorlow@chromium.org 已提交
202
  new_db.SetComparatorName(user_comparator()->Name());
203
  new_db.SetLogNumber(0);
J
jorlow@chromium.org 已提交
204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
  new_db.SetNextFile(2);
  new_db.SetLastSequence(0);

  const std::string manifest = DescriptorFileName(dbname_, 1);
  WritableFile* file;
  Status s = env_->NewWritableFile(manifest, &file);
  if (!s.ok()) {
    return s;
  }
  {
    log::Writer log(file);
    std::string record;
    new_db.EncodeTo(&record);
    s = log.AddRecord(record);
    if (s.ok()) {
      s = file->Close();
    }
  }
  delete file;
  if (s.ok()) {
    // Make "CURRENT" file that points to the new manifest file.
    s = SetCurrentFile(env_, dbname_, 1);
  } else {
    env_->DeleteFile(manifest);
  }
  return s;
}

void DBImpl::MaybeIgnoreError(Status* s) const {
  if (s->ok() || options_.paranoid_checks) {
    // No change needed
  } else {
236
    Log(options_.info_log, "Ignoring error %s", s->ToString().c_str());
J
jorlow@chromium.org 已提交
237 238 239 240 241 242 243 244 245 246 247 248 249
    *s = Status::OK();
  }
}

void DBImpl::DeleteObsoleteFiles() {
  // Make a set of all of the live files
  std::set<uint64_t> live = pending_outputs_;
  versions_->AddLiveFiles(&live);

  std::vector<std::string> filenames;
  env_->GetChildren(dbname_, &filenames); // Ignoring errors on purpose
  uint64_t number;
  FileType type;
H
heyongqiang 已提交
250
  std::vector<uint64_t> old_log_files_ts;
D
dgrogan@chromium.org 已提交
251 252
  for (size_t i = 0; i < filenames.size(); i++) {
    if (ParseFileName(filenames[i], &number, &type)) {
J
jorlow@chromium.org 已提交
253 254 255
      bool keep = true;
      switch (type) {
        case kLogFile:
256
          keep = ((number >= versions_->LogNumber()) ||
257
                  (number == versions_->PrevLogNumber()));
J
jorlow@chromium.org 已提交
258 259 260 261 262 263 264 265 266 267 268 269 270 271
          break;
        case kDescriptorFile:
          // Keep my manifest file, and any newer incarnations'
          // (in case there is a race that allows other incarnations)
          keep = (number >= versions_->ManifestFileNumber());
          break;
        case kTableFile:
          keep = (live.find(number) != live.end());
          break;
        case kTempFile:
          // Any temp files that are currently being written to must
          // be recorded in pending_outputs_, which is inserted into "live"
          keep = (live.find(number) != live.end());
          break;
H
heyongqiang 已提交
272 273 274 275 276 277
        case kInfoLogFile:
          keep = true;
          if (number != 0) {
            old_log_files_ts.push_back(number);
          }
          break;
J
jorlow@chromium.org 已提交
278 279 280 281 282 283 284 285 286 287
        case kCurrentFile:
        case kDBLockFile:
          keep = true;
          break;
      }

      if (!keep) {
        if (type == kTableFile) {
          table_cache_->Evict(number);
        }
288
        Log(options_.info_log, "Delete type=%d #%lld\n",
J
jorlow@chromium.org 已提交
289 290
            int(type),
            static_cast<unsigned long long>(number));
H
heyongqiang 已提交
291 292 293 294 295 296
        Status st = env_->DeleteFile(dbname_ + "/" + filenames[i]);
        if(!st.ok()) {
          Log(options_.info_log, "Delete type=%d #%lld FAILED\n",
              int(type),
              static_cast<unsigned long long>(number));
        }
J
jorlow@chromium.org 已提交
297 298 299
      }
    }
  }
H
heyongqiang 已提交
300 301 302 303 304 305 306 307 308 309 310 311 312 313

  // Delete old log files.
  int old_log_file_count = old_log_files_ts.size();
  if (old_log_file_count >= KEEP_LOG_FILE_NUM) {
    std::sort(old_log_files_ts.begin(), old_log_files_ts.end());
    for (int i = 0; i >= (old_log_file_count - KEEP_LOG_FILE_NUM); i++) {
      uint64_t ts = old_log_files_ts.at(i);
      std::string to_delete = OldInfoLogFileName(dbname_, ts);
      Log(options_.info_log, "Delete type=%d #%lld\n",
          int(kInfoLogFile),
          static_cast<unsigned long long>(ts));
      env_->DeleteFile(dbname_ + "/" + to_delete);
    }
  }
J
jorlow@chromium.org 已提交
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345
}

Status DBImpl::Recover(VersionEdit* edit) {
  mutex_.AssertHeld();

  // Ignore error from CreateDir since the creation of the DB is
  // committed only when the descriptor is created, and this directory
  // may already exist from a previous failed creation attempt.
  env_->CreateDir(dbname_);
  assert(db_lock_ == NULL);
  Status s = env_->LockFile(LockFileName(dbname_), &db_lock_);
  if (!s.ok()) {
    return s;
  }

  if (!env_->FileExists(CurrentFileName(dbname_))) {
    if (options_.create_if_missing) {
      s = NewDB();
      if (!s.ok()) {
        return s;
      }
    } else {
      return Status::InvalidArgument(
          dbname_, "does not exist (create_if_missing is false)");
    }
  } else {
    if (options_.error_if_exists) {
      return Status::InvalidArgument(
          dbname_, "exists (error_if_exists is true)");
    }
  }

346
  s = versions_->Recover();
J
jorlow@chromium.org 已提交
347 348
  if (s.ok()) {
    SequenceNumber max_sequence(0);
349 350 351 352 353 354 355 356 357 358 359 360 361 362

    // Recover from all newer log files than the ones named in the
    // descriptor (new log files may have been added by the previous
    // incarnation without registering them in the descriptor).
    //
    // Note that PrevLogNumber() is no longer used, but we pay
    // attention to it in case we are recovering a database
    // produced by an older version of leveldb.
    const uint64_t min_log = versions_->LogNumber();
    const uint64_t prev_log = versions_->PrevLogNumber();
    std::vector<std::string> filenames;
    s = env_->GetChildren(dbname_, &filenames);
    if (!s.ok()) {
      return s;
363
    }
364 365 366 367 368 369 370 371 372
    uint64_t number;
    FileType type;
    std::vector<uint64_t> logs;
    for (size_t i = 0; i < filenames.size(); i++) {
      if (ParseFileName(filenames[i], &number, &type)
          && type == kLogFile
          && ((number >= min_log) || (number == prev_log))) {
        logs.push_back(number);
      }
J
jorlow@chromium.org 已提交
373
    }
374 375 376 377 378

    // Recover in the order in which the logs were generated
    std::sort(logs.begin(), logs.end());
    for (size_t i = 0; i < logs.size(); i++) {
      s = RecoverLogFile(logs[i], edit, &max_sequence);
379 380 381 382 383

      // The previous incarnation may not have written any MANIFEST
      // records after allocating this log number.  So we manually
      // update the file number allocation counter in VersionSet.
      versions_->MarkFileNumberUsed(logs[i]);
384 385
    }

J
jorlow@chromium.org 已提交
386
    if (s.ok()) {
387 388 389
      if (versions_->LastSequence() < max_sequence) {
        versions_->SetLastSequence(max_sequence);
      }
J
jorlow@chromium.org 已提交
390 391 392 393 394 395 396 397 398 399 400
    }
  }

  return s;
}

Status DBImpl::RecoverLogFile(uint64_t log_number,
                              VersionEdit* edit,
                              SequenceNumber* max_sequence) {
  struct LogReporter : public log::Reader::Reporter {
    Env* env;
401
    Logger* info_log;
J
jorlow@chromium.org 已提交
402 403 404
    const char* fname;
    Status* status;  // NULL if options_.paranoid_checks==false
    virtual void Corruption(size_t bytes, const Status& s) {
405
      Log(info_log, "%s%s: dropping %d bytes; %s",
J
jorlow@chromium.org 已提交
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
          (this->status == NULL ? "(ignoring error) " : ""),
          fname, static_cast<int>(bytes), s.ToString().c_str());
      if (this->status != NULL && this->status->ok()) *this->status = s;
    }
  };

  mutex_.AssertHeld();

  // Open the log file
  std::string fname = LogFileName(dbname_, log_number);
  SequentialFile* file;
  Status status = env_->NewSequentialFile(fname, &file);
  if (!status.ok()) {
    MaybeIgnoreError(&status);
    return status;
  }

  // Create the log reader.
  LogReporter reporter;
  reporter.env = env_;
  reporter.info_log = options_.info_log;
  reporter.fname = fname.c_str();
  reporter.status = (options_.paranoid_checks ? &status : NULL);
  // We intentially make log::Reader do checksumming even if
  // paranoid_checks==false so that corruptions cause entire commits
  // to be skipped instead of propagating bad information (like overly
  // large sequence numbers).
433 434
  log::Reader reader(file, &reporter, true/*checksum*/,
                     0/*initial_offset*/);
435
  Log(options_.info_log, "Recovering log #%llu",
J
jorlow@chromium.org 已提交
436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453
      (unsigned long long) log_number);

  // Read all the records and add to a memtable
  std::string scratch;
  Slice record;
  WriteBatch batch;
  MemTable* mem = NULL;
  while (reader.ReadRecord(&record, &scratch) &&
         status.ok()) {
    if (record.size() < 12) {
      reporter.Corruption(
          record.size(), Status::Corruption("log record too small"));
      continue;
    }
    WriteBatchInternal::SetContents(&batch, record);

    if (mem == NULL) {
      mem = new MemTable(internal_comparator_);
454
      mem->Ref();
J
jorlow@chromium.org 已提交
455 456 457 458 459 460 461 462 463 464 465 466 467 468
    }
    status = WriteBatchInternal::InsertInto(&batch, mem);
    MaybeIgnoreError(&status);
    if (!status.ok()) {
      break;
    }
    const SequenceNumber last_seq =
        WriteBatchInternal::Sequence(&batch) +
        WriteBatchInternal::Count(&batch) - 1;
    if (last_seq > *max_sequence) {
      *max_sequence = last_seq;
    }

    if (mem->ApproximateMemoryUsage() > options_.write_buffer_size) {
469
      status = WriteLevel0Table(mem, edit, NULL);
J
jorlow@chromium.org 已提交
470 471 472 473 474
      if (!status.ok()) {
        // Reflect errors immediately so that conditions like full
        // file-systems cause the DB::Open() to fail.
        break;
      }
475
      mem->Unref();
J
jorlow@chromium.org 已提交
476 477 478 479 480
      mem = NULL;
    }
  }

  if (status.ok() && mem != NULL) {
481
    status = WriteLevel0Table(mem, edit, NULL);
J
jorlow@chromium.org 已提交
482 483 484 485
    // Reflect errors immediately so that conditions like full
    // file-systems cause the DB::Open() to fail.
  }

486
  if (mem != NULL) mem->Unref();
J
jorlow@chromium.org 已提交
487 488 489 490
  delete file;
  return status;
}

491 492
Status DBImpl::WriteLevel0Table(MemTable* mem, VersionEdit* edit,
                                Version* base) {
J
jorlow@chromium.org 已提交
493
  mutex_.AssertHeld();
494
  const uint64_t start_micros = env_->NowMicros();
J
jorlow@chromium.org 已提交
495 496 497 498
  FileMetaData meta;
  meta.number = versions_->NewFileNumber();
  pending_outputs_.insert(meta.number);
  Iterator* iter = mem->NewIterator();
499
  Log(options_.info_log, "Level-0 table #%llu: started",
J
jorlow@chromium.org 已提交
500
      (unsigned long long) meta.number);
501 502 503 504

  Status s;
  {
    mutex_.Unlock();
505
    s = BuildTable(dbname_, env_, options_, table_cache_, iter, &meta);
506 507 508
    mutex_.Lock();
  }

509
  Log(options_.info_log, "Level-0 table #%llu: %lld bytes %s",
J
jorlow@chromium.org 已提交
510 511 512 513 514
      (unsigned long long) meta.number,
      (unsigned long long) meta.file_size,
      s.ToString().c_str());
  delete iter;
  pending_outputs_.erase(meta.number);
515

516 517 518 519 520

  // Note that if file_size is zero, the file has been deleted and
  // should not be added to the manifest.
  int level = 0;
  if (s.ok() && meta.file_size > 0) {
521 522
    const Slice min_user_key = meta.smallest.user_key();
    const Slice max_user_key = meta.largest.user_key();
G
Gabor Cselle 已提交
523 524
    if (base != NULL) {
      level = base->PickLevelForMemTableOutput(min_user_key, max_user_key);
525 526 527 528 529
    }
    edit->AddFile(level, meta.number, meta.file_size,
                  meta.smallest, meta.largest);
  }

530 531 532
  CompactionStats stats;
  stats.micros = env_->NowMicros() - start_micros;
  stats.bytes_written = meta.file_size;
533
  stats_[level].Add(stats);
J
jorlow@chromium.org 已提交
534 535 536 537 538
  return s;
}

Status DBImpl::CompactMemTable() {
  mutex_.AssertHeld();
539
  assert(imm_ != NULL);
J
jorlow@chromium.org 已提交
540 541

  // Save the contents of the memtable as a new Table
542
  VersionEdit edit(NumberLevels());
543 544 545 546 547 548 549 550
  Version* base = versions_->current();
  base->Ref();
  Status s = WriteLevel0Table(imm_, &edit, base);
  base->Unref();

  if (s.ok() && shutting_down_.Acquire_Load()) {
    s = Status::IOError("Deleting DB during memtable compaction");
  }
J
jorlow@chromium.org 已提交
551

552
  // Replace immutable memtable with the generated Table
J
jorlow@chromium.org 已提交
553
  if (s.ok()) {
554
    edit.SetPrevLogNumber(0);
555
    edit.SetLogNumber(logfile_number_);  // Earlier logs no longer needed
556
    s = versions_->LogAndApply(&edit, &mutex_);
J
jorlow@chromium.org 已提交
557 558 559 560
  }

  if (s.ok()) {
    // Commit to the new state
561
    imm_->Unref();
562 563
    imm_ = NULL;
    has_imm_.Release_Store(NULL);
J
jorlow@chromium.org 已提交
564
    DeleteObsoleteFiles();
565
    MaybeScheduleLogDBDeployStats();
J
jorlow@chromium.org 已提交
566
  }
567

J
jorlow@chromium.org 已提交
568 569 570
  return s;
}

G
Gabor Cselle 已提交
571 572 573 574 575
void DBImpl::CompactRange(const Slice* begin, const Slice* end) {
  int max_level_with_files = 1;
  {
    MutexLock l(&mutex_);
    Version* base = versions_->current();
576
    for (int level = 1; level < NumberLevels(); level++) {
G
Gabor Cselle 已提交
577 578 579 580 581 582 583 584 585 586 587
      if (base->OverlapInLevel(level, begin, end)) {
        max_level_with_files = level;
      }
    }
  }
  TEST_CompactMemTable(); // TODO(sanjay): Skip if memtable does not overlap
  for (int level = 0; level < max_level_with_files; level++) {
    TEST_CompactRange(level, begin, end);
  }
}

588
int DBImpl::NumberLevels() {
589
  return options_.num_levels;
590 591 592
}

int DBImpl::MaxMemCompactionLevel() {
593
  return options_.max_mem_compaction_level;
594 595 596
}

int DBImpl::Level0StopWriteTrigger() {
597
  return options_.level0_stop_writes_trigger;
598 599
}

H
heyongqiang 已提交
600 601 602 603 604
Status DBImpl::Flush(const FlushOptions& options) {
  Status status = FlushMemTable(options);
  return status;
}

G
Gabor Cselle 已提交
605
void DBImpl::TEST_CompactRange(int level, const Slice* begin,const Slice* end) {
606 607
  assert(level >= 0);

G
Gabor Cselle 已提交
608 609
  InternalKey begin_storage, end_storage;

H
hans@chromium.org 已提交
610 611
  ManualCompaction manual;
  manual.level = level;
G
Gabor Cselle 已提交
612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635
  manual.done = false;
  if (begin == NULL) {
    manual.begin = NULL;
  } else {
    begin_storage = InternalKey(*begin, kMaxSequenceNumber, kValueTypeForSeek);
    manual.begin = &begin_storage;
  }
  if (end == NULL) {
    manual.end = NULL;
  } else {
    end_storage = InternalKey(*end, 0, static_cast<ValueType>(0));
    manual.end = &end_storage;
  }

  MutexLock l(&mutex_);
  while (!manual.done) {
    while (manual_compaction_ != NULL) {
      bg_cv_.Wait();
    }
    manual_compaction_ = &manual;
    MaybeScheduleCompaction();
    while (manual_compaction_ == &manual) {
      bg_cv_.Wait();
    }
H
hans@chromium.org 已提交
636
  }
J
jorlow@chromium.org 已提交
637 638
}

H
heyongqiang 已提交
639
Status DBImpl::FlushMemTable(const FlushOptions& options) {
640 641
  // NULL batch means just wait for earlier writes to be done
  Status s = Write(WriteOptions(), NULL);
H
heyongqiang 已提交
642
  if (s.ok() && options.wait) {
643
    // Wait until the compaction completes
H
heyongqiang 已提交
644
    s = WaitForCompactMemTable();
645 646
  }
  return s;
J
jorlow@chromium.org 已提交
647 648
}

H
heyongqiang 已提交
649
Status DBImpl::WaitForCompactMemTable() {
650 651 652 653 654 655 656 657 658 659
  Status s;
  // Wait until the compaction completes
  MutexLock l(&mutex_);
  while (imm_ != NULL && bg_error_.ok()) {
    bg_cv_.Wait();
  }
  if (imm_ != NULL) {
    s = bg_error_;
  }
  return s;
H
heyongqiang 已提交
660 661 662 663 664 665
}

Status DBImpl::TEST_CompactMemTable() {
  return FlushMemTable(FlushOptions());
}

666
Status DBImpl::TEST_WaitForCompactMemTable() {
667
  return WaitForCompactMemTable();
668 669 670
}

Status DBImpl::TEST_WaitForCompact() {
671 672 673 674 675 676
  // Wait until the compaction completes
  MutexLock l(&mutex_);
  while (bg_compaction_scheduled_ && bg_error_.ok()) {
    bg_cv_.Wait();
  }
  return bg_error_;
677 678
}

J
jorlow@chromium.org 已提交
679 680 681 682 683 684
void DBImpl::MaybeScheduleCompaction() {
  mutex_.AssertHeld();
  if (bg_compaction_scheduled_) {
    // Already scheduled
  } else if (shutting_down_.Acquire_Load()) {
    // DB is being deleted; no more background compactions
H
hans@chromium.org 已提交
685 686 687
  } else if (imm_ == NULL &&
             manual_compaction_ == NULL &&
             !versions_->NeedsCompaction()) {
J
jorlow@chromium.org 已提交
688 689 690 691 692 693 694 695 696 697 698 699 700 701
    // No work to be done
  } else {
    bg_compaction_scheduled_ = true;
    env_->Schedule(&DBImpl::BGWork, this);
  }
}

void DBImpl::BGWork(void* db) {
  reinterpret_cast<DBImpl*>(db)->BackgroundCall();
}

void DBImpl::BackgroundCall() {
  MutexLock l(&mutex_);
  assert(bg_compaction_scheduled_);
H
hans@chromium.org 已提交
702
  if (!shutting_down_.Acquire_Load()) {
703 704 705 706 707 708 709 710 711 712 713 714 715
    Status s = BackgroundCompaction();
    if (!s.ok()) {
      // Wait a little bit before retrying background compaction in
      // case this is an environmental problem and we do not want to
      // chew up resources for failed compactions for the duration of
      // the problem.
      bg_cv_.SignalAll();  // In case a waiter can proceed despite the error
      Log(options_.info_log, "Waiting after background compaction error: %s",
          s.ToString().c_str());
      mutex_.Unlock();
      env_->SleepForMicroseconds(1000000);
      mutex_.Lock();
    }
J
jorlow@chromium.org 已提交
716
  }
717

J
jorlow@chromium.org 已提交
718 719
  bg_compaction_scheduled_ = false;

720 721
  MaybeScheduleLogDBDeployStats();

J
jorlow@chromium.org 已提交
722 723 724
  // Previous compaction may have produced too many files in a level,
  // so reschedule another compaction if needed.
  MaybeScheduleCompaction();
H
hans@chromium.org 已提交
725
  bg_cv_.SignalAll();
J
jorlow@chromium.org 已提交
726 727
}

728
Status DBImpl::BackgroundCompaction() {
J
jorlow@chromium.org 已提交
729
  mutex_.AssertHeld();
730 731

  if (imm_ != NULL) {
732
    return CompactMemTable();
733 734
  }

H
hans@chromium.org 已提交
735 736
  Compaction* c;
  bool is_manual = (manual_compaction_ != NULL);
G
Gabor Cselle 已提交
737
  InternalKey manual_end;
H
hans@chromium.org 已提交
738
  if (is_manual) {
G
Gabor Cselle 已提交
739 740 741 742 743 744 745 746
    ManualCompaction* m = manual_compaction_;
    c = versions_->CompactRange(m->level, m->begin, m->end);
    m->done = (c == NULL);
    if (c != NULL) {
      manual_end = c->input(0, c->num_input_files(0) - 1)->largest;
    }
    Log(options_.info_log,
        "Manual compaction at level-%d from %s .. %s; will stop at %s\n",
H
hans@chromium.org 已提交
747
        m->level,
G
Gabor Cselle 已提交
748 749 750
        (m->begin ? m->begin->DebugString().c_str() : "(begin)"),
        (m->end ? m->end->DebugString().c_str() : "(end)"),
        (m->done ? "(end)" : manual_end.DebugString().c_str()));
H
hans@chromium.org 已提交
751 752
  } else {
    c = versions_->PickCompaction();
J
jorlow@chromium.org 已提交
753 754 755
  }

  Status status;
H
hans@chromium.org 已提交
756 757 758
  if (c == NULL) {
    // Nothing to do
  } else if (!is_manual && c->IsTrivialMove()) {
J
jorlow@chromium.org 已提交
759
    // Move file to next level
760
    assert(c->num_input_files(0) == 1);
J
jorlow@chromium.org 已提交
761 762 763 764
    FileMetaData* f = c->input(0, 0);
    c->edit()->DeleteFile(c->level(), f->number);
    c->edit()->AddFile(c->level() + 1, f->number, f->file_size,
                       f->smallest, f->largest);
765
    status = versions_->LogAndApply(c->edit(), &mutex_);
H
hans@chromium.org 已提交
766
    VersionSet::LevelSummaryStorage tmp;
767
    Log(options_.info_log, "Moved #%lld to level-%d %lld bytes %s: %s\n",
J
jorlow@chromium.org 已提交
768 769 770
        static_cast<unsigned long long>(f->number),
        c->level() + 1,
        static_cast<unsigned long long>(f->file_size),
H
hans@chromium.org 已提交
771 772
        status.ToString().c_str(),
        versions_->LevelSummary(&tmp));
J
jorlow@chromium.org 已提交
773 774 775 776
  } else {
    CompactionState* compact = new CompactionState(c);
    status = DoCompactionWork(compact);
    CleanupCompaction(compact);
777 778
    c->ReleaseInputs();
    DeleteObsoleteFiles();
J
jorlow@chromium.org 已提交
779 780 781 782 783 784 785 786
  }
  delete c;

  if (status.ok()) {
    // Done
  } else if (shutting_down_.Acquire_Load()) {
    // Ignore compaction errors found during shutting down
  } else {
787
    Log(options_.info_log,
J
jorlow@chromium.org 已提交
788 789 790 791 792
        "Compaction error: %s", status.ToString().c_str());
    if (options_.paranoid_checks && bg_error_.ok()) {
      bg_error_ = status;
    }
  }
H
hans@chromium.org 已提交
793 794

  if (is_manual) {
G
Gabor Cselle 已提交
795
    ManualCompaction* m = manual_compaction_;
796 797 798
    if (!status.ok()) {
      m->done = true;
    }
G
Gabor Cselle 已提交
799 800 801 802 803 804
    if (!m->done) {
      // We only compacted part of the requested range.  Update *m
      // to the range that is left to be compacted.
      m->tmp_storage = manual_end;
      m->begin = &m->tmp_storage;
    }
H
hans@chromium.org 已提交
805 806
    manual_compaction_ = NULL;
  }
807
  return status;
J
jorlow@chromium.org 已提交
808 809 810 811 812 813 814 815 816 817 818 819
}

void DBImpl::CleanupCompaction(CompactionState* compact) {
  mutex_.AssertHeld();
  if (compact->builder != NULL) {
    // May happen if we get a shutdown call in the middle of compaction
    compact->builder->Abandon();
    delete compact->builder;
  } else {
    assert(compact->outfile == NULL);
  }
  delete compact->outfile;
D
dgrogan@chromium.org 已提交
820
  for (size_t i = 0; i < compact->outputs.size(); i++) {
J
jorlow@chromium.org 已提交
821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
    const CompactionState::Output& out = compact->outputs[i];
    pending_outputs_.erase(out.number);
  }
  delete compact;
}

Status DBImpl::OpenCompactionOutputFile(CompactionState* compact) {
  assert(compact != NULL);
  assert(compact->builder == NULL);
  uint64_t file_number;
  {
    mutex_.Lock();
    file_number = versions_->NewFileNumber();
    pending_outputs_.insert(file_number);
    CompactionState::Output out;
    out.number = file_number;
    out.smallest.Clear();
    out.largest.Clear();
    compact->outputs.push_back(out);
    mutex_.Unlock();
  }

  // Make the output file
  std::string fname = TableFileName(dbname_, file_number);
  Status s = env_->NewWritableFile(fname, &compact->outfile);
  if (s.ok()) {
    compact->builder = new TableBuilder(options_, compact->outfile);
  }
  return s;
}

Status DBImpl::FinishCompactionOutputFile(CompactionState* compact,
                                          Iterator* input) {
  assert(compact != NULL);
  assert(compact->outfile != NULL);
  assert(compact->builder != NULL);

  const uint64_t output_number = compact->current_output()->number;
  assert(output_number != 0);

  // Check for iterator errors
  Status s = input->status();
  const uint64_t current_entries = compact->builder->NumEntries();
  if (s.ok()) {
    s = compact->builder->Finish();
  } else {
    compact->builder->Abandon();
  }
  const uint64_t current_bytes = compact->builder->FileSize();
  compact->current_output()->file_size = current_bytes;
  compact->total_bytes += current_bytes;
  delete compact->builder;
  compact->builder = NULL;

  // Finish and check for file errors
876
  if (s.ok() && !options_.disableDataSync) {
877 878 879 880 881
    if (options_.use_fsync) {
      s = compact->outfile->Fsync();
    } else {
      s = compact->outfile->Sync();
    }
J
jorlow@chromium.org 已提交
882 883 884 885 886 887 888 889 890
  }
  if (s.ok()) {
    s = compact->outfile->Close();
  }
  delete compact->outfile;
  compact->outfile = NULL;

  if (s.ok() && current_entries > 0) {
    // Verify that the table is usable
J
jorlow@chromium.org 已提交
891 892 893
    Iterator* iter = table_cache_->NewIterator(ReadOptions(),
                                               output_number,
                                               current_bytes);
J
jorlow@chromium.org 已提交
894 895 896
    s = iter->status();
    delete iter;
    if (s.ok()) {
897
      Log(options_.info_log,
J
jorlow@chromium.org 已提交
898 899 900 901 902 903 904 905 906 907 908 909
          "Generated table #%llu: %lld keys, %lld bytes",
          (unsigned long long) output_number,
          (unsigned long long) current_entries,
          (unsigned long long) current_bytes);
    }
  }
  return s;
}


Status DBImpl::InstallCompactionResults(CompactionState* compact) {
  mutex_.AssertHeld();
910
  Log(options_.info_log,  "Compacted %d@%d + %d@%d files => %lld bytes",
J
jorlow@chromium.org 已提交
911 912 913 914 915 916 917 918 919
      compact->compaction->num_input_files(0),
      compact->compaction->level(),
      compact->compaction->num_input_files(1),
      compact->compaction->level() + 1,
      static_cast<long long>(compact->total_bytes));

  // Add compaction outputs
  compact->compaction->AddInputDeletions(compact->compaction->edit());
  const int level = compact->compaction->level();
D
dgrogan@chromium.org 已提交
920
  for (size_t i = 0; i < compact->outputs.size(); i++) {
J
jorlow@chromium.org 已提交
921 922 923 924 925
    const CompactionState::Output& out = compact->outputs[i];
    compact->compaction->edit()->AddFile(
        level + 1,
        out.number, out.file_size, out.smallest, out.largest);
  }
926
  return versions_->LogAndApply(compact->compaction->edit(), &mutex_);
J
jorlow@chromium.org 已提交
927 928 929
}

Status DBImpl::DoCompactionWork(CompactionState* compact) {
930 931 932
  const uint64_t start_micros = env_->NowMicros();
  int64_t imm_micros = 0;  // Micros spent doing imm_ compactions

933
  Log(options_.info_log,  "Compacting %d@%d + %d@%d files",
J
jorlow@chromium.org 已提交
934 935 936 937
      compact->compaction->num_input_files(0),
      compact->compaction->level(),
      compact->compaction->num_input_files(1),
      compact->compaction->level() + 1);
H
heyongqiang 已提交
938 939 940
  char scratch[200];
  compact->compaction->Summary(scratch, 256);
  Log(options_.info_log, "Compaction start summary: %s\n", scratch);
J
jorlow@chromium.org 已提交
941 942 943 944 945

  assert(versions_->NumLevelFiles(compact->compaction->level()) > 0);
  assert(compact->builder == NULL);
  assert(compact->outfile == NULL);
  if (snapshots_.empty()) {
946
    compact->smallest_snapshot = versions_->LastSequence();
J
jorlow@chromium.org 已提交
947 948 949 950 951 952 953 954 955 956 957 958 959 960 961
  } else {
    compact->smallest_snapshot = snapshots_.oldest()->number_;
  }

  // Release mutex while we're actually doing the compaction work
  mutex_.Unlock();

  Iterator* input = versions_->MakeInputIterator(compact->compaction);
  input->SeekToFirst();
  Status status;
  ParsedInternalKey ikey;
  std::string current_user_key;
  bool has_current_user_key = false;
  SequenceNumber last_sequence_for_key = kMaxSequenceNumber;
  for (; input->Valid() && !shutting_down_.Acquire_Load(); ) {
962 963 964 965 966 967
    // Prioritize immutable compaction work
    if (has_imm_.NoBarrier_Load() != NULL) {
      const uint64_t imm_start = env_->NowMicros();
      mutex_.Lock();
      if (imm_ != NULL) {
        CompactMemTable();
H
hans@chromium.org 已提交
968
        bg_cv_.SignalAll();  // Wakeup MakeRoomForWrite() if necessary
969 970 971 972 973
      }
      mutex_.Unlock();
      imm_micros += (env_->NowMicros() - imm_start);
    }

J
jorlow@chromium.org 已提交
974
    Slice key = input->key();
975
    if (compact->compaction->ShouldStopBefore(key) &&
976 977 978 979 980 981 982 983
        compact->builder != NULL) {
      status = FinishCompactionOutputFile(compact, input);
      if (!status.ok()) {
        break;
      }
    }

    // Handle key/value, add to state, etc.
J
jorlow@chromium.org 已提交
984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018
    bool drop = false;
    if (!ParseInternalKey(key, &ikey)) {
      // Do not hide error keys
      current_user_key.clear();
      has_current_user_key = false;
      last_sequence_for_key = kMaxSequenceNumber;
    } else {
      if (!has_current_user_key ||
          user_comparator()->Compare(ikey.user_key,
                                     Slice(current_user_key)) != 0) {
        // First occurrence of this user key
        current_user_key.assign(ikey.user_key.data(), ikey.user_key.size());
        has_current_user_key = true;
        last_sequence_for_key = kMaxSequenceNumber;
      }

      if (last_sequence_for_key <= compact->smallest_snapshot) {
        // Hidden by an newer entry for same user key
        drop = true;    // (A)
      } else if (ikey.type == kTypeDeletion &&
                 ikey.sequence <= compact->smallest_snapshot &&
                 compact->compaction->IsBaseLevelForKey(ikey.user_key)) {
        // For this user key:
        // (1) there is no data in higher levels
        // (2) data in lower levels will have larger sequence numbers
        // (3) data in layers that are being compacted here and have
        //     smaller sequence numbers will be dropped in the next
        //     few iterations of this loop (by rule (A) above).
        // Therefore this deletion marker is obsolete and can be dropped.
        drop = true;
      }

      last_sequence_for_key = ikey.sequence;
    }
#if 0
1019
    Log(options_.info_log,
J
jorlow@chromium.org 已提交
1020 1021 1022
        "  Compact: %s, seq %d, type: %d %d, drop: %d, is_base: %d, "
        "%d smallest_snapshot: %d",
        ikey.user_key.ToString().c_str(),
D
dgrogan@chromium.org 已提交
1023
        (int)ikey.sequence, ikey.type, kTypeValue, drop,
J
jorlow@chromium.org 已提交
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
        compact->compaction->IsBaseLevelForKey(ikey.user_key),
        (int)last_sequence_for_key, (int)compact->smallest_snapshot);
#endif

    if (!drop) {
      // Open output file if necessary
      if (compact->builder == NULL) {
        status = OpenCompactionOutputFile(compact);
        if (!status.ok()) {
          break;
        }
      }
      if (compact->builder->NumEntries() == 0) {
        compact->current_output()->smallest.DecodeFrom(key);
      }
      compact->current_output()->largest.DecodeFrom(key);
D
dgrogan@chromium.org 已提交
1040
      compact->builder->Add(key, input->value());
J
jorlow@chromium.org 已提交
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066

      // Close output file if it is big enough
      if (compact->builder->FileSize() >=
          compact->compaction->MaxOutputFileSize()) {
        status = FinishCompactionOutputFile(compact, input);
        if (!status.ok()) {
          break;
        }
      }
    }

    input->Next();
  }

  if (status.ok() && shutting_down_.Acquire_Load()) {
    status = Status::IOError("Deleting DB during compaction");
  }
  if (status.ok() && compact->builder != NULL) {
    status = FinishCompactionOutputFile(compact, input);
  }
  if (status.ok()) {
    status = input->status();
  }
  delete input;
  input = NULL;

1067 1068 1069 1070 1071 1072 1073
  CompactionStats stats;
  stats.micros = env_->NowMicros() - start_micros - imm_micros;
  for (int which = 0; which < 2; which++) {
    for (int i = 0; i < compact->compaction->num_input_files(which); i++) {
      stats.bytes_read += compact->compaction->input(which, i)->file_size;
    }
  }
D
dgrogan@chromium.org 已提交
1074
  for (size_t i = 0; i < compact->outputs.size(); i++) {
1075 1076 1077
    stats.bytes_written += compact->outputs[i].file_size;
  }

J
jorlow@chromium.org 已提交
1078
  mutex_.Lock();
1079
  stats_[compact->compaction->level() + 1].Add(stats);
J
jorlow@chromium.org 已提交
1080 1081 1082 1083

  if (status.ok()) {
    status = InstallCompactionResults(compact);
  }
1084
  VersionSet::LevelSummaryStorage tmp;
1085
  Log(options_.info_log,
1086
      "compacted to: %s", versions_->LevelSummary(&tmp));
J
jorlow@chromium.org 已提交
1087 1088 1089
  return status;
}

1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
namespace {
struct IterState {
  port::Mutex* mu;
  Version* version;
  MemTable* mem;
  MemTable* imm;
};

static void CleanupIteratorState(void* arg1, void* arg2) {
  IterState* state = reinterpret_cast<IterState*>(arg1);
  state->mu->Lock();
  state->mem->Unref();
  if (state->imm != NULL) state->imm->Unref();
  state->version->Unref();
  state->mu->Unlock();
  delete state;
}
H
Hans Wennborg 已提交
1107
}  // namespace
1108

J
jorlow@chromium.org 已提交
1109 1110
Iterator* DBImpl::NewInternalIterator(const ReadOptions& options,
                                      SequenceNumber* latest_snapshot) {
1111
  IterState* cleanup = new IterState;
J
jorlow@chromium.org 已提交
1112
  mutex_.Lock();
1113
  *latest_snapshot = versions_->LastSequence();
J
jorlow@chromium.org 已提交
1114 1115 1116 1117

  // Collect together all needed child iterators
  std::vector<Iterator*> list;
  list.push_back(mem_->NewIterator());
1118
  mem_->Ref();
1119 1120
  if (imm_ != NULL) {
    list.push_back(imm_->NewIterator());
1121
    imm_->Ref();
1122
  }
J
jorlow@chromium.org 已提交
1123 1124 1125 1126
  versions_->current()->AddIterators(options, &list);
  Iterator* internal_iter =
      NewMergingIterator(&internal_comparator_, &list[0], list.size());
  versions_->current()->Ref();
1127 1128 1129 1130 1131 1132

  cleanup->mu = &mutex_;
  cleanup->mem = mem_;
  cleanup->imm = imm_;
  cleanup->version = versions_->current();
  internal_iter->RegisterCleanup(CleanupIteratorState, cleanup, NULL);
J
jorlow@chromium.org 已提交
1133 1134 1135 1136 1137 1138 1139 1140 1141 1142

  mutex_.Unlock();
  return internal_iter;
}

Iterator* DBImpl::TEST_NewInternalIterator() {
  SequenceNumber ignored;
  return NewInternalIterator(ReadOptions(), &ignored);
}

J
jorlow@chromium.org 已提交
1143
int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes() {
1144 1145 1146 1147
  MutexLock l(&mutex_);
  return versions_->MaxNextLevelOverlappingBytes();
}

J
jorlow@chromium.org 已提交
1148 1149 1150
Status DBImpl::Get(const ReadOptions& options,
                   const Slice& key,
                   std::string* value) {
1151 1152 1153 1154 1155 1156 1157
  Status s;
  MutexLock l(&mutex_);
  SequenceNumber snapshot;
  if (options.snapshot != NULL) {
    snapshot = reinterpret_cast<const SnapshotImpl*>(options.snapshot)->number_;
  } else {
    snapshot = versions_->LastSequence();
J
jorlow@chromium.org 已提交
1158
  }
1159

1160 1161
  MemTable* mem = mem_;
  MemTable* imm = imm_;
1162
  Version* current = versions_->current();
1163 1164
  mem->Ref();
  if (imm != NULL) imm->Ref();
1165
  current->Ref();
1166 1167

  bool have_stat_update = false;
1168
  Version::GetStats stats;
1169 1170 1171

  // Unlock while reading from files and memtables
  {
1172
    mutex_.Unlock();
1173 1174
    // First look in the memtable, then in the immutable memtable (if any).
    LookupKey lkey(key, snapshot);
1175
    if (mem->Get(lkey, value, &s)) {
1176
      // Done
1177
    } else if (imm != NULL && imm->Get(lkey, value, &s)) {
1178 1179 1180 1181 1182
      // Done
    } else {
      s = current->Get(options, lkey, value, &stats);
      have_stat_update = true;
    }
1183 1184
    mutex_.Lock();
  }
1185 1186

  if (have_stat_update && current->UpdateStats(stats)) {
1187 1188
    MaybeScheduleCompaction();
  }
1189 1190
  mem->Unref();
  if (imm != NULL) imm->Unref();
1191 1192
  current->Unref();
  return s;
J
jorlow@chromium.org 已提交
1193 1194 1195 1196 1197
}

Iterator* DBImpl::NewIterator(const ReadOptions& options) {
  SequenceNumber latest_snapshot;
  Iterator* internal_iter = NewInternalIterator(options, &latest_snapshot);
1198 1199 1200 1201 1202
  return NewDBIterator(
      &dbname_, env_, user_comparator(), internal_iter,
      (options.snapshot != NULL
       ? reinterpret_cast<const SnapshotImpl*>(options.snapshot)->number_
       : latest_snapshot));
J
jorlow@chromium.org 已提交
1203 1204 1205 1206
}

const Snapshot* DBImpl::GetSnapshot() {
  MutexLock l(&mutex_);
1207
  return snapshots_.New(versions_->LastSequence());
J
jorlow@chromium.org 已提交
1208 1209 1210 1211
}

void DBImpl::ReleaseSnapshot(const Snapshot* s) {
  MutexLock l(&mutex_);
1212
  snapshots_.Delete(reinterpret_cast<const SnapshotImpl*>(s));
J
jorlow@chromium.org 已提交
1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223
}

// Convenience methods
Status DBImpl::Put(const WriteOptions& o, const Slice& key, const Slice& val) {
  return DB::Put(o, key, val);
}

Status DBImpl::Delete(const WriteOptions& options, const Slice& key) {
  return DB::Delete(options, key);
}

1224 1225 1226 1227
Status DBImpl::Write(const WriteOptions& options, WriteBatch* my_batch) {
  Writer w(&mutex_);
  w.batch = my_batch;
  w.sync = options.sync;
H
heyongqiang 已提交
1228
  w.disableWAL = options.disableWAL;
1229
  w.done = false;
1230

D
dgrogan@chromium.org 已提交
1231
  MutexLock l(&mutex_);
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
  writers_.push_back(&w);
  while (!w.done && &w != writers_.front()) {
    w.cv.Wait();
  }
  if (w.done) {
    return w.status;
  }

  // May temporarily unlock and wait.
  Status status = MakeRoomForWrite(my_batch == NULL);
D
dgrogan@chromium.org 已提交
1242
  uint64_t last_sequence = versions_->LastSequence();
1243 1244 1245
  Writer* last_writer = &w;
  if (status.ok() && my_batch != NULL) {  // NULL batch is for compactions
    WriteBatch* updates = BuildBatchGroup(&last_writer);
D
dgrogan@chromium.org 已提交
1246 1247 1248
    WriteBatchInternal::SetSequence(updates, last_sequence + 1);
    last_sequence += WriteBatchInternal::Count(updates);

1249 1250 1251 1252
    // Add to log and apply to memtable.  We can release the lock
    // during this phase since &w is currently responsible for logging
    // and protects against concurrent loggers and concurrent writes
    // into mem_.
1253 1254
    {
      mutex_.Unlock();
H
heyongqiang 已提交
1255 1256 1257
      if (!options.disableWAL) {
        status = log_->AddRecord(WriteBatchInternal::Contents(updates));
        if (status.ok() && options.sync) {
1258 1259 1260 1261 1262
          if (options_.use_fsync) {
            status = logfile_->Fsync();
          } else {
            status = logfile_->Sync();
          }
H
heyongqiang 已提交
1263
        }
1264 1265 1266 1267 1268
      }
      if (status.ok()) {
        status = WriteBatchInternal::InsertInto(updates, mem_);
      }
      mutex_.Lock();
J
jorlow@chromium.org 已提交
1269
    }
1270
    if (updates == tmp_batch_) tmp_batch_->Clear();
1271 1272

    versions_->SetLastSequence(last_sequence);
J
jorlow@chromium.org 已提交
1273
  }
1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290

  while (true) {
    Writer* ready = writers_.front();
    writers_.pop_front();
    if (ready != &w) {
      ready->status = status;
      ready->done = true;
      ready->cv.Signal();
    }
    if (ready == last_writer) break;
  }

  // Notify new head of write queue
  if (!writers_.empty()) {
    writers_.front()->cv.Signal();
  }

J
jorlow@chromium.org 已提交
1291 1292 1293
  return status;
}

1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321
// REQUIRES: Writer list must be non-empty
// REQUIRES: First writer must have a non-NULL batch
WriteBatch* DBImpl::BuildBatchGroup(Writer** last_writer) {
  assert(!writers_.empty());
  Writer* first = writers_.front();
  WriteBatch* result = first->batch;
  assert(result != NULL);

  size_t size = WriteBatchInternal::ByteSize(first->batch);

  // Allow the group to grow up to a maximum size, but if the
  // original write is small, limit the growth so we do not slow
  // down the small write too much.
  size_t max_size = 1 << 20;
  if (size <= (128<<10)) {
    max_size = size + (128<<10);
  }

  *last_writer = first;
  std::deque<Writer*>::iterator iter = writers_.begin();
  ++iter;  // Advance past "first"
  for (; iter != writers_.end(); ++iter) {
    Writer* w = *iter;
    if (w->sync && !first->sync) {
      // Do not include a sync write into a batch handled by a non-sync write.
      break;
    }

H
heyongqiang 已提交
1322 1323 1324 1325 1326 1327
    if (!w->disableWAL && first->disableWAL) {
      // Do not include a write that needs WAL into a batch that has
      // WAL disabled.
      break;
    }

1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
    if (w->batch != NULL) {
      size += WriteBatchInternal::ByteSize(w->batch);
      if (size > max_size) {
        // Do not make batch too big
        break;
      }

      // Append to *reuslt
      if (result == first->batch) {
        // Switch to temporary batch instead of disturbing caller's batch
        result = tmp_batch_;
        assert(WriteBatchInternal::Count(result) == 0);
        WriteBatchInternal::Append(result, first->batch);
      }
      WriteBatchInternal::Append(result, w->batch);
    }
    *last_writer = w;
  }
  return result;
}

1349
// REQUIRES: mutex_ is held
1350
// REQUIRES: this thread is currently at the front of the writer queue
1351 1352
Status DBImpl::MakeRoomForWrite(bool force) {
  mutex_.AssertHeld();
1353
  assert(!writers_.empty());
1354
  bool allow_delay = !force;
1355
  Status s;
1356

1357 1358 1359 1360 1361
  while (true) {
    if (!bg_error_.ok()) {
      // Yield previous error
      s = bg_error_;
      break;
1362 1363
    } else if (
        allow_delay &&
1364 1365
        versions_->NumLevelFiles(0) >=
					options_.level0_slowdown_writes_trigger) {
1366 1367 1368 1369 1370 1371 1372 1373 1374
      // We are getting close to hitting a hard limit on the number of
      // L0 files.  Rather than delaying a single write by several
      // seconds when we hit the hard limit, start delaying each
      // individual write by 1ms to reduce latency variance.  Also,
      // this delay hands over some CPU to the compaction thread in
      // case it is sharing the same core as the writer.
      mutex_.Unlock();
      env_->SleepForMicroseconds(1000);
      allow_delay = false;  // Do not delay a single write more than once
1375
      Log(options_.info_log, "delaying write...\n");
1376
      mutex_.Lock();
1377 1378 1379 1380 1381 1382 1383
    } else if (!force &&
               (mem_->ApproximateMemoryUsage() <= options_.write_buffer_size)) {
      // There is room in current memtable
      break;
    } else if (imm_ != NULL) {
      // We have filled up the current memtable, but the previous
      // one is still being compacted, so we wait.
1384
      Log(options_.info_log, "wait for memtable compaction...\n");
H
hans@chromium.org 已提交
1385
      bg_cv_.Wait();
1386 1387
    } else if (versions_->NumLevelFiles(0) >=
		options_.level0_stop_writes_trigger) {
1388
      // There are too many level-0 files.
1389
      Log(options_.info_log, "waiting...\n");
H
hans@chromium.org 已提交
1390
      bg_cv_.Wait();
1391 1392 1393 1394 1395 1396 1397
    } else {
      // Attempt to switch to a new memtable and trigger compaction of old
      assert(versions_->PrevLogNumber() == 0);
      uint64_t new_log_number = versions_->NewFileNumber();
      WritableFile* lfile = NULL;
      s = env_->NewWritableFile(LogFileName(dbname_, new_log_number), &lfile);
      if (!s.ok()) {
H
heyongqiang 已提交
1398 1399
        // Avoid chewing through file number space in a tight loop.
	versions_->ReuseFileNumber(new_log_number);
1400 1401 1402 1403 1404
        break;
      }
      delete log_;
      delete logfile_;
      logfile_ = lfile;
1405
      logfile_number_ = new_log_number;
1406 1407 1408 1409
      log_ = new log::Writer(lfile);
      imm_ = mem_;
      has_imm_.Release_Store(imm_);
      mem_ = new MemTable(internal_comparator_);
1410
      mem_->Ref();
1411 1412 1413 1414 1415 1416 1417 1418 1419 1420
      force = false;   // Do not force another compaction if have room
      MaybeScheduleCompaction();
    }
  }
  return s;
}

bool DBImpl::GetProperty(const Slice& property, std::string* value) {
  value->clear();

J
jorlow@chromium.org 已提交
1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
  MutexLock l(&mutex_);
  Slice in = property;
  Slice prefix("leveldb.");
  if (!in.starts_with(prefix)) return false;
  in.remove_prefix(prefix.size());

  if (in.starts_with("num-files-at-level")) {
    in.remove_prefix(strlen("num-files-at-level"));
    uint64_t level;
    bool ok = ConsumeDecimalNumber(&in, &level) && in.empty();
1431
    if (!ok || (int)level >= NumberLevels()) {
J
jorlow@chromium.org 已提交
1432 1433
      return false;
    } else {
1434
      char buf[100];
D
dgrogan@chromium.org 已提交
1435 1436
      snprintf(buf, sizeof(buf), "%d",
               versions_->NumLevelFiles(static_cast<int>(level)));
1437
      *value = buf;
J
jorlow@chromium.org 已提交
1438 1439
      return true;
    }
1440 1441 1442 1443 1444 1445 1446 1447
  } else if (in == "stats") {
    char buf[200];
    snprintf(buf, sizeof(buf),
             "                               Compactions\n"
             "Level  Files Size(MB) Time(sec) Read(MB) Write(MB)\n"
             "--------------------------------------------------\n"
             );
    value->append(buf);
1448
    for (int level = 0; level < NumberLevels(); level++) {
1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463
      int files = versions_->NumLevelFiles(level);
      if (stats_[level].micros > 0 || files > 0) {
        snprintf(
            buf, sizeof(buf),
            "%3d %8d %8.0f %9.0f %8.0f %9.0f\n",
            level,
            files,
            versions_->NumLevelBytes(level) / 1048576.0,
            stats_[level].micros / 1e6,
            stats_[level].bytes_read / 1048576.0,
            stats_[level].bytes_written / 1048576.0);
        value->append(buf);
      }
    }
    return true;
G
Gabor Cselle 已提交
1464 1465 1466
  } else if (in == "sstables") {
    *value = versions_->current()->DebugString();
    return true;
J
jorlow@chromium.org 已提交
1467
  }
1468

J
jorlow@chromium.org 已提交
1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519
  return false;
}

void DBImpl::GetApproximateSizes(
    const Range* range, int n,
    uint64_t* sizes) {
  // TODO(opt): better implementation
  Version* v;
  {
    MutexLock l(&mutex_);
    versions_->current()->Ref();
    v = versions_->current();
  }

  for (int i = 0; i < n; i++) {
    // Convert user_key into a corresponding internal key.
    InternalKey k1(range[i].start, kMaxSequenceNumber, kValueTypeForSeek);
    InternalKey k2(range[i].limit, kMaxSequenceNumber, kValueTypeForSeek);
    uint64_t start = versions_->ApproximateOffsetOf(v, k1);
    uint64_t limit = versions_->ApproximateOffsetOf(v, k2);
    sizes[i] = (limit >= start ? limit - start : 0);
  }

  {
    MutexLock l(&mutex_);
    v->Unref();
  }
}

// Default implementations of convenience methods that subclasses of DB
// can call if they wish
Status DB::Put(const WriteOptions& opt, const Slice& key, const Slice& value) {
  WriteBatch batch;
  batch.Put(key, value);
  return Write(opt, &batch);
}

Status DB::Delete(const WriteOptions& opt, const Slice& key) {
  WriteBatch batch;
  batch.Delete(key);
  return Write(opt, &batch);
}

DB::~DB() { }

Status DB::Open(const Options& options, const std::string& dbname,
                DB** dbptr) {
  *dbptr = NULL;

  DBImpl* impl = new DBImpl(options, dbname);
  impl->mutex_.Lock();
1520
  VersionEdit edit(impl->NumberLevels());
J
jorlow@chromium.org 已提交
1521 1522
  Status s = impl->Recover(&edit); // Handles create_if_missing, error_if_exists
  if (s.ok()) {
1523
    uint64_t new_log_number = impl->versions_->NewFileNumber();
J
jorlow@chromium.org 已提交
1524
    WritableFile* lfile;
1525
    s = options.env->NewWritableFile(LogFileName(dbname, new_log_number),
J
jorlow@chromium.org 已提交
1526 1527
                                     &lfile);
    if (s.ok()) {
1528
      edit.SetLogNumber(new_log_number);
J
jorlow@chromium.org 已提交
1529
      impl->logfile_ = lfile;
1530
      impl->logfile_number_ = new_log_number;
J
jorlow@chromium.org 已提交
1531
      impl->log_ = new log::Writer(lfile);
1532
      s = impl->versions_->LogAndApply(&edit, &impl->mutex_);
J
jorlow@chromium.org 已提交
1533 1534 1535
    }
    if (s.ok()) {
      impl->DeleteObsoleteFiles();
1536
      impl->MaybeScheduleCompaction();
1537
      impl->MaybeScheduleLogDBDeployStats();
J
jorlow@chromium.org 已提交
1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548
    }
  }
  impl->mutex_.Unlock();
  if (s.ok()) {
    *dbptr = impl;
  } else {
    delete impl;
  }
  return s;
}

1549 1550 1551
Snapshot::~Snapshot() {
}

J
jorlow@chromium.org 已提交
1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
Status DestroyDB(const std::string& dbname, const Options& options) {
  Env* env = options.env;
  std::vector<std::string> filenames;
  // Ignore error in case directory does not exist
  env->GetChildren(dbname, &filenames);
  if (filenames.empty()) {
    return Status::OK();
  }

  FileLock* lock;
1562 1563
  const std::string lockname = LockFileName(dbname);
  Status result = env->LockFile(lockname, &lock);
J
jorlow@chromium.org 已提交
1564 1565 1566
  if (result.ok()) {
    uint64_t number;
    FileType type;
D
dgrogan@chromium.org 已提交
1567
    for (size_t i = 0; i < filenames.size(); i++) {
1568
      if (ParseFileName(filenames[i], &number, &type) &&
1569
          type != kDBLockFile) {  // Lock file will be deleted at end
J
jorlow@chromium.org 已提交
1570 1571 1572 1573 1574 1575 1576
        Status del = env->DeleteFile(dbname + "/" + filenames[i]);
        if (result.ok() && !del.ok()) {
          result = del;
        }
      }
    }
    env->UnlockFile(lock);  // Ignore error since state is already gone
1577
    env->DeleteFile(lockname);
J
jorlow@chromium.org 已提交
1578 1579 1580 1581 1582
    env->DeleteDir(dbname);  // Ignore error in case dir contains other files
  }
  return result;
}

1583 1584 1585 1586 1587 1588 1589 1590 1591
//
// A global method that can dump out the build version
void printLeveldbBuildVersion() {
  printf("Git sha %s", leveldb_build_git_sha);
  printf("Git datetime %s", leveldb_build_git_datetime);
  printf("Compile time %s", leveldb_build_compile_time);
  printf("Compile date %s", leveldb_build_compile_date);
}

H
Hans Wennborg 已提交
1592
}  // namespace leveldb