version_set.cc 134.5 KB
Newer Older
1
//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
S
Siying Dong 已提交
2 3 4
//  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).
5
//
J
jorlow@chromium.org 已提交
6 7 8 9 10
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.

#include "db/version_set.h"
I
Igor Canadi 已提交
11

L
liuhuahang 已提交
12
#ifndef __STDC_FORMAT_MACROS
I
Igor Canadi 已提交
13
#define __STDC_FORMAT_MACROS
L
liuhuahang 已提交
14 15
#endif

I
Igor Canadi 已提交
16
#include <inttypes.h>
17
#include <stdio.h>
J
jorlow@chromium.org 已提交
18
#include <algorithm>
19
#include <climits>
I
Igor Canadi 已提交
20
#include <map>
I
Igor Canadi 已提交
21
#include <set>
22
#include <string>
23
#include <unordered_map>
24
#include <vector>
25
#include "db/compaction.h"
26
#include "db/internal_stats.h"
J
jorlow@chromium.org 已提交
27 28 29
#include "db/log_reader.h"
#include "db/log_writer.h"
#include "db/memtable.h"
30
#include "db/merge_context.h"
31
#include "db/merge_helper.h"
32
#include "db/pinned_iterators_manager.h"
J
jorlow@chromium.org 已提交
33
#include "db/table_cache.h"
S
sdong 已提交
34
#include "db/version_builder.h"
35
#include "monitoring/file_read_sample.h"
36
#include "monitoring/perf_context_imp.h"
37 38
#include "rocksdb/env.h"
#include "rocksdb/merge_operator.h"
39
#include "rocksdb/write_buffer_manager.h"
40 41
#include "table/format.h"
#include "table/get_context.h"
S
sdong 已提交
42
#include "table/internal_iterator.h"
43
#include "table/merging_iterator.h"
44
#include "table/meta_blocks.h"
45 46 47
#include "table/plain_table_factory.h"
#include "table/table_reader.h"
#include "table/two_level_iterator.h"
J
jorlow@chromium.org 已提交
48
#include "util/coding.h"
49
#include "util/file_reader_writer.h"
50
#include "util/filename.h"
51
#include "util/stop_watch.h"
52
#include "util/string_util.h"
S
sdong 已提交
53
#include "util/sync_point.h"
J
jorlow@chromium.org 已提交
54

55
namespace rocksdb {
J
jorlow@chromium.org 已提交
56

57 58
namespace {

59
// Find File in LevelFilesBrief data structure
60 61
// Within an index range defined by left and right
int FindFileInRange(const InternalKeyComparator& icmp,
62
    const LevelFilesBrief& file_level,
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 88 89
    const Slice& key,
    uint32_t left,
    uint32_t right) {
  while (left < right) {
    uint32_t mid = (left + right) / 2;
    const FdWithKeyRange& f = file_level.files[mid];
    if (icmp.InternalKeyComparator::Compare(f.largest_key, key) < 0) {
      // Key at "mid.largest" is < "target".  Therefore all
      // files at or before "mid" are uninteresting.
      left = mid + 1;
    } else {
      // Key at "mid.largest" is >= "target".  Therefore all files
      // after "mid" are uninteresting.
      right = mid;
    }
  }
  return right;
}

// Class to help choose the next file to search for the particular key.
// Searches and returns files level by level.
// We can search level-by-level since entries never hop across
// levels. Therefore we are guaranteed that if we find data
// in a smaller level, later levels are irrelevant (unless we
// are MergeInProgress).
class FilePicker {
 public:
90 91 92 93 94
  FilePicker(std::vector<FileMetaData*>* files, const Slice& user_key,
             const Slice& ikey, autovector<LevelFilesBrief>* file_levels,
             unsigned int num_levels, FileIndexer* file_indexer,
             const Comparator* user_comparator,
             const InternalKeyComparator* internal_comparator)
95
      : num_levels_(num_levels),
96
        curr_level_(static_cast<unsigned int>(-1)),
97
        returned_file_level_(static_cast<unsigned int>(-1)),
98
        hit_file_level_(static_cast<unsigned int>(-1)),
99 100
        search_left_bound_(0),
        search_right_bound_(FileIndexer::kLevelMaxIndex),
101
#ifndef NDEBUG
102
        files_(files),
103
#endif
104
        level_files_brief_(file_levels),
105
        is_hit_file_last_in_level_(false),
106 107 108 109 110 111 112 113 114
        user_key_(user_key),
        ikey_(ikey),
        file_indexer_(file_indexer),
        user_comparator_(user_comparator),
        internal_comparator_(internal_comparator) {
    // Setup member variables to search first level.
    search_ended_ = !PrepareNextLevel();
    if (!search_ended_) {
      // Prefetch Level 0 table data to avoid cache miss if possible.
115 116
      for (unsigned int i = 0; i < (*level_files_brief_)[0].num_files; ++i) {
        auto* r = (*level_files_brief_)[0].files[i].fd.table_reader;
117 118 119 120 121 122 123
        if (r) {
          r->Prepare(ikey);
        }
      }
    }
  }

124
  int GetCurrentLevel() const { return curr_level_; }
125

126 127 128 129 130
  FdWithKeyRange* GetNextFile() {
    while (!search_ended_) {  // Loops over different levels.
      while (curr_index_in_curr_level_ < curr_file_level_->num_files) {
        // Loops over all files in current level.
        FdWithKeyRange* f = &curr_file_level_->files[curr_index_in_curr_level_];
131
        hit_file_level_ = curr_level_;
132 133
        is_hit_file_last_in_level_ =
            curr_index_in_curr_level_ == curr_file_level_->num_files - 1;
134 135 136 137
        int cmp_largest = -1;

        // Do key range filtering of files or/and fractional cascading if:
        // (1) not all the files are in level 0, or
赵星宇 已提交
138 139
        // (2) there are more than 3 current level files
        // If there are only 3 or less current level files in the system, we skip
140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197
        // the key range filtering. In this case, more likely, the system is
        // highly tuned to minimize number of tables queried by each query,
        // so it is unlikely that key range filtering is more efficient than
        // querying the files.
        if (num_levels_ > 1 || curr_file_level_->num_files > 3) {
          // Check if key is within a file's range. If search left bound and
          // right bound point to the same find, we are sure key falls in
          // range.
          assert(
              curr_level_ == 0 ||
              curr_index_in_curr_level_ == start_index_in_curr_level_ ||
              user_comparator_->Compare(user_key_,
                ExtractUserKey(f->smallest_key)) <= 0);

          int cmp_smallest = user_comparator_->Compare(user_key_,
              ExtractUserKey(f->smallest_key));
          if (cmp_smallest >= 0) {
            cmp_largest = user_comparator_->Compare(user_key_,
                ExtractUserKey(f->largest_key));
          }

          // Setup file search bound for the next level based on the
          // comparison results
          if (curr_level_ > 0) {
            file_indexer_->GetNextLevelIndex(curr_level_,
                                            curr_index_in_curr_level_,
                                            cmp_smallest, cmp_largest,
                                            &search_left_bound_,
                                            &search_right_bound_);
          }
          // Key falls out of current file's range
          if (cmp_smallest < 0 || cmp_largest > 0) {
            if (curr_level_ == 0) {
              ++curr_index_in_curr_level_;
              continue;
            } else {
              // Search next level.
              break;
            }
          }
        }
#ifndef NDEBUG
        // Sanity check to make sure that the files are correctly sorted
        if (prev_file_) {
          if (curr_level_ != 0) {
            int comp_sign = internal_comparator_->Compare(
                prev_file_->largest_key, f->smallest_key);
            assert(comp_sign < 0);
          } else {
            // level == 0, the current file cannot be newer than the previous
            // one. Use compressed data structure, has no attribute seqNo
            assert(curr_index_in_curr_level_ > 0);
            assert(!NewestFirstBySeqNo(files_[0][curr_index_in_curr_level_],
                  files_[0][curr_index_in_curr_level_-1]));
          }
        }
        prev_file_ = f;
#endif
198
        returned_file_level_ = curr_level_;
199 200 201 202 203 204 205 206 207 208 209 210 211 212 213
        if (curr_level_ > 0 && cmp_largest < 0) {
          // No more files to search in this level.
          search_ended_ = !PrepareNextLevel();
        } else {
          ++curr_index_in_curr_level_;
        }
        return f;
      }
      // Start searching next level.
      search_ended_ = !PrepareNextLevel();
    }
    // Search ended.
    return nullptr;
  }

214 215 216 217
  // getter for current file level
  // for GET_HIT_L0, GET_HIT_L1 & GET_HIT_L2_AND_UP counts
  unsigned int GetHitFileLevel() { return hit_file_level_; }

218 219 220 221
  // Returns true if the most recent "hit file" (i.e., one returned by
  // GetNextFile()) is at the last index in its level.
  bool IsHitFileLastInLevel() { return is_hit_file_last_in_level_; }

222 223 224
 private:
  unsigned int num_levels_;
  unsigned int curr_level_;
225
  unsigned int returned_file_level_;
226
  unsigned int hit_file_level_;
227 228
  int32_t search_left_bound_;
  int32_t search_right_bound_;
229
#ifndef NDEBUG
230
  std::vector<FileMetaData*>* files_;
231
#endif
232
  autovector<LevelFilesBrief>* level_files_brief_;
233
  bool search_ended_;
234
  bool is_hit_file_last_in_level_;
235
  LevelFilesBrief* curr_file_level_;
236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
  unsigned int curr_index_in_curr_level_;
  unsigned int start_index_in_curr_level_;
  Slice user_key_;
  Slice ikey_;
  FileIndexer* file_indexer_;
  const Comparator* user_comparator_;
  const InternalKeyComparator* internal_comparator_;
#ifndef NDEBUG
  FdWithKeyRange* prev_file_;
#endif

  // Setup local variables to search next level.
  // Returns false if there are no more levels to search.
  bool PrepareNextLevel() {
    curr_level_++;
    while (curr_level_ < num_levels_) {
252
      curr_file_level_ = &(*level_files_brief_)[curr_level_];
253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
      if (curr_file_level_->num_files == 0) {
        // When current level is empty, the search bound generated from upper
        // level must be [0, -1] or [0, FileIndexer::kLevelMaxIndex] if it is
        // also empty.
        assert(search_left_bound_ == 0);
        assert(search_right_bound_ == -1 ||
               search_right_bound_ == FileIndexer::kLevelMaxIndex);
        // Since current level is empty, it will need to search all files in
        // the next level
        search_left_bound_ = 0;
        search_right_bound_ = FileIndexer::kLevelMaxIndex;
        curr_level_++;
        continue;
      }

      // Some files may overlap each other. We find
      // all files that overlap user_key and process them in order from
      // newest to oldest. In the context of merge-operator, this can occur at
      // any level. Otherwise, it only occurs at Level-0 (since Put/Deletes
      // are always compacted into a single entry).
      int32_t start_index;
      if (curr_level_ == 0) {
        // On Level-0, we read through all files to check for overlap.
        start_index = 0;
      } else {
        // On Level-n (n>=1), files are sorted. Binary search to find the
        // earliest file whose largest key >= ikey. Search left bound and
        // right bound are used to narrow the range.
        if (search_left_bound_ == search_right_bound_) {
          start_index = search_left_bound_;
        } else if (search_left_bound_ < search_right_bound_) {
          if (search_right_bound_ == FileIndexer::kLevelMaxIndex) {
285 286
            search_right_bound_ =
                static_cast<int32_t>(curr_file_level_->num_files) - 1;
287
          }
288 289 290 291
          start_index =
              FindFileInRange(*internal_comparator_, *curr_file_level_, ikey_,
                              static_cast<uint32_t>(search_left_bound_),
                              static_cast<uint32_t>(search_right_bound_));
292 293
        } else {
          // search_left_bound > search_right_bound, key does not exist in
C
clark.kang 已提交
294
          // this level. Since no comparison is done in this level, it will
295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314
          // need to search all files in the next level.
          search_left_bound_ = 0;
          search_right_bound_ = FileIndexer::kLevelMaxIndex;
          curr_level_++;
          continue;
        }
      }
      start_index_in_curr_level_ = start_index;
      curr_index_in_curr_level_ = start_index;
#ifndef NDEBUG
      prev_file_ = nullptr;
#endif
      return true;
    }
    // curr_level_ = num_levels_. So, no more levels to search.
    return false;
  }
};
}  // anonymous namespace

S
sdong 已提交
315 316
VersionStorageInfo::~VersionStorageInfo() { delete[] files_; }

J
jorlow@chromium.org 已提交
317 318
Version::~Version() {
  assert(refs_ == 0);
319 320 321 322 323 324

  // Remove from linked list
  prev_->next_ = next_;
  next_->prev_ = prev_;

  // Drop references to files
S
sdong 已提交
325 326 327
  for (int level = 0; level < storage_info_.num_levels_; level++) {
    for (size_t i = 0; i < storage_info_.files_[level].size(); i++) {
      FileMetaData* f = storage_info_.files_[level][i];
328
      assert(f->refs > 0);
J
jorlow@chromium.org 已提交
329 330
      f->refs--;
      if (f->refs <= 0) {
I
Igor Canadi 已提交
331
        vset_->obsolete_files_.push_back(f);
J
jorlow@chromium.org 已提交
332 333 334 335 336
      }
    }
  }
}

337
int FindFile(const InternalKeyComparator& icmp,
338
             const LevelFilesBrief& file_level,
339
             const Slice& key) {
340 341
  return FindFileInRange(icmp, file_level, key, 0,
                         static_cast<uint32_t>(file_level.num_files));
342 343
}

344
void DoGenerateLevelFilesBrief(LevelFilesBrief* file_level,
F
Feng Zhu 已提交
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
        const std::vector<FileMetaData*>& files,
        Arena* arena) {
  assert(file_level);
  assert(arena);

  size_t num = files.size();
  file_level->num_files = num;
  char* mem = arena->AllocateAligned(num * sizeof(FdWithKeyRange));
  file_level->files = new (mem)FdWithKeyRange[num];

  for (size_t i = 0; i < num; i++) {
    Slice smallest_key = files[i]->smallest.Encode();
    Slice largest_key = files[i]->largest.Encode();

    // Copy key slice to sequential memory
    size_t smallest_size = smallest_key.size();
    size_t largest_size = largest_key.size();
    mem = arena->AllocateAligned(smallest_size + largest_size);
    memcpy(mem, smallest_key.data(), smallest_size);
    memcpy(mem + smallest_size, largest_key.data(), largest_size);

    FdWithKeyRange& f = file_level->files[i];
    f.fd = files[i]->fd;
368
    f.file_metadata = files[i];
F
Feng Zhu 已提交
369 370 371 372 373
    f.smallest_key = Slice(mem, smallest_size);
    f.largest_key = Slice(mem + smallest_size, largest_size);
  }
}

G
Gabor Cselle 已提交
374
static bool AfterFile(const Comparator* ucmp,
375
                      const Slice* user_key, const FdWithKeyRange* f) {
A
Abhishek Kona 已提交
376 377
  // nullptr user_key occurs before all keys and is therefore never after *f
  return (user_key != nullptr &&
378
          ucmp->Compare(*user_key, ExtractUserKey(f->largest_key)) > 0);
G
Gabor Cselle 已提交
379 380 381
}

static bool BeforeFile(const Comparator* ucmp,
382
                       const Slice* user_key, const FdWithKeyRange* f) {
A
Abhishek Kona 已提交
383 384
  // nullptr user_key occurs after all keys and is therefore never before *f
  return (user_key != nullptr &&
385
          ucmp->Compare(*user_key, ExtractUserKey(f->smallest_key)) < 0);
G
Gabor Cselle 已提交
386 387
}

388 389
bool SomeFileOverlapsRange(
    const InternalKeyComparator& icmp,
G
Gabor Cselle 已提交
390
    bool disjoint_sorted_files,
391
    const LevelFilesBrief& file_level,
G
Gabor Cselle 已提交
392 393 394 395 396
    const Slice* smallest_user_key,
    const Slice* largest_user_key) {
  const Comparator* ucmp = icmp.user_comparator();
  if (!disjoint_sorted_files) {
    // Need to check against all files
397 398
    for (size_t i = 0; i < file_level.num_files; i++) {
      const FdWithKeyRange* f = &(file_level.files[i]);
G
Gabor Cselle 已提交
399 400 401 402 403 404 405 406 407 408 409 410
      if (AfterFile(ucmp, smallest_user_key, f) ||
          BeforeFile(ucmp, largest_user_key, f)) {
        // No overlap
      } else {
        return true;  // Overlap
      }
    }
    return false;
  }

  // Binary search over file list
  uint32_t index = 0;
A
Abhishek Kona 已提交
411
  if (smallest_user_key != nullptr) {
G
Gabor Cselle 已提交
412
    // Find the earliest possible internal key for smallest_user_key
413 414
    InternalKey small;
    small.SetMaxPossibleForUserKey(*smallest_user_key);
415
    index = FindFile(icmp, file_level, small.Encode());
G
Gabor Cselle 已提交
416 417
  }

418
  if (index >= file_level.num_files) {
G
Gabor Cselle 已提交
419 420 421 422
    // beginning of range is after all files, so no overlap.
    return false;
  }

423
  return !BeforeFile(ucmp, largest_user_key, &file_level.files[index]);
424 425
}

426 427
namespace {

J
jorlow@chromium.org 已提交
428 429 430
// An internal iterator.  For a given version/level pair, yields
// information about the files in the level.  For a given entry, key()
// is the largest key that occurs in the file, and value() is an
J
jorlow@chromium.org 已提交
431 432
// 16-byte value containing the file number and file size, both
// encoded using EncodeFixed64.
S
sdong 已提交
433
class LevelFileNumIterator : public InternalIterator {
J
jorlow@chromium.org 已提交
434
 public:
435
  LevelFileNumIterator(const InternalKeyComparator& icmp,
436
                       const LevelFilesBrief* flevel, bool should_sample)
437
      : icmp_(icmp),
F
Feng Zhu 已提交
438
        flevel_(flevel),
439
        index_(static_cast<uint32_t>(flevel->num_files)),
440 441
        current_value_(0, 0, 0),  // Marks as invalid
        should_sample_(should_sample) {}
I
Igor Sugak 已提交
442 443
  virtual bool Valid() const override { return index_ < flevel_->num_files; }
  virtual void Seek(const Slice& target) override {
F
Feng Zhu 已提交
444
    index_ = FindFile(icmp_, *flevel_, target);
J
jorlow@chromium.org 已提交
445
  }
A
Aaron Gao 已提交
446 447 448 449
  virtual void SeekForPrev(const Slice& target) override {
    SeekForPrevImpl(target, &icmp_);
  }

I
Igor Sugak 已提交
450 451
  virtual void SeekToFirst() override { index_ = 0; }
  virtual void SeekToLast() override {
452 453 454
    index_ = (flevel_->num_files == 0)
                 ? 0
                 : static_cast<uint32_t>(flevel_->num_files) - 1;
J
jorlow@chromium.org 已提交
455
  }
I
Igor Sugak 已提交
456
  virtual void Next() override {
J
jorlow@chromium.org 已提交
457 458 459
    assert(Valid());
    index_++;
  }
I
Igor Sugak 已提交
460
  virtual void Prev() override {
J
jorlow@chromium.org 已提交
461 462
    assert(Valid());
    if (index_ == 0) {
463
      index_ = static_cast<uint32_t>(flevel_->num_files);  // Marks as invalid
J
jorlow@chromium.org 已提交
464 465 466 467
    } else {
      index_--;
    }
  }
I
Igor Sugak 已提交
468
  Slice key() const override {
J
jorlow@chromium.org 已提交
469
    assert(Valid());
F
Feng Zhu 已提交
470
    return flevel_->files[index_].largest_key;
J
jorlow@chromium.org 已提交
471
  }
I
Igor Sugak 已提交
472
  Slice value() const override {
J
jorlow@chromium.org 已提交
473
    assert(Valid());
F
Feng Zhu 已提交
474 475

    auto file_meta = flevel_->files[index_];
476 477 478
    if (should_sample_) {
      sample_file_read_inc(file_meta.file_metadata);
    }
F
Feng Zhu 已提交
479
    current_value_ = file_meta.fd;
480
    return Slice(reinterpret_cast<const char*>(&current_value_),
481
                 sizeof(FileDescriptor));
J
jorlow@chromium.org 已提交
482
  }
I
Igor Sugak 已提交
483 484
  virtual Status status() const override { return Status::OK(); }

J
jorlow@chromium.org 已提交
485 486
 private:
  const InternalKeyComparator icmp_;
487
  const LevelFilesBrief* flevel_;
D
dgrogan@chromium.org 已提交
488
  uint32_t index_;
489
  mutable FileDescriptor current_value_;
490
  bool should_sample_;
J
jorlow@chromium.org 已提交
491 492
};

493
class LevelFileIteratorState : public TwoLevelIteratorState {
L
Lei Jin 已提交
494
 public:
495
  // @param skip_filters Disables loading/accessing the filter block
L
Lei Jin 已提交
496
  LevelFileIteratorState(TableCache* table_cache,
497 498 499 500
                         const ReadOptions& read_options,
                         const EnvOptions& env_options,
                         const InternalKeyComparator& icomparator,
                         HistogramImpl* file_read_hist, bool for_compaction,
501 502
                         bool prefix_enabled, bool skip_filters, int level,
                         RangeDelAggregator* range_del_agg)
503 504 505 506 507 508
      : TwoLevelIteratorState(prefix_enabled),
        table_cache_(table_cache),
        read_options_(read_options),
        env_options_(env_options),
        icomparator_(icomparator),
        file_read_hist_(file_read_hist),
509
        for_compaction_(for_compaction),
510
        skip_filters_(skip_filters),
511 512
        level_(level),
        range_del_agg_(range_del_agg) {}
L
Lei Jin 已提交
513

S
sdong 已提交
514
  InternalIterator* NewSecondaryIterator(const Slice& meta_handle) override {
515
    if (meta_handle.size() != sizeof(FileDescriptor)) {
S
sdong 已提交
516
      return NewErrorInternalIterator(
L
Lei Jin 已提交
517 518
          Status::Corruption("FileReader invoked with unexpected value"));
    }
A
Andrew Kryczka 已提交
519 520 521
    const FileDescriptor* fd =
        reinterpret_cast<const FileDescriptor*>(meta_handle.data());
    return table_cache_->NewIterator(
522
        read_options_, env_options_, icomparator_, *fd, range_del_agg_,
A
Andrew Kryczka 已提交
523
        nullptr /* don't need reference to table */, file_read_hist_,
524
        for_compaction_, nullptr /* arena */, skip_filters_, level_);
J
jorlow@chromium.org 已提交
525 526
  }

527 528 529
  bool PrefixMayMatch(const Slice& internal_key) override {
    return true;
  }
L
Lei Jin 已提交
530

531 532 533 534 535 536 537
  bool KeyReachedUpperBound(const Slice& internal_key) override {
    return read_options_.iterate_upper_bound != nullptr &&
           icomparator_.user_comparator()->Compare(
               ExtractUserKey(internal_key),
               *read_options_.iterate_upper_bound) >= 0;
  }

L
Lei Jin 已提交
538 539 540 541 542
 private:
  TableCache* table_cache_;
  const ReadOptions read_options_;
  const EnvOptions& env_options_;
  const InternalKeyComparator& icomparator_;
543
  HistogramImpl* file_read_hist_;
L
Lei Jin 已提交
544
  bool for_compaction_;
545
  bool skip_filters_;
546
  int level_;
547
  RangeDelAggregator* range_del_agg_;
L
Lei Jin 已提交
548
};
T
Tyler Harter 已提交
549

S
sdong 已提交
550 551
// A wrapper of version builder which references the current version in
// constructor and unref it in the destructor.
552
// Both of the constructor and destructor need to be called inside DB Mutex.
S
sdong 已提交
553 554 555
class BaseReferencedVersionBuilder {
 public:
  explicit BaseReferencedVersionBuilder(ColumnFamilyData* cfd)
556
      : version_builder_(new VersionBuilder(
557
            cfd->current()->version_set()->env_options(), cfd->table_cache(),
558
            cfd->current()->storage_info(), cfd->ioptions()->info_log)),
S
sdong 已提交
559 560 561
        version_(cfd->current()) {
    version_->Ref();
  }
562 563 564 565
  ~BaseReferencedVersionBuilder() {
    delete version_builder_;
    version_->Unref();
  }
566
  VersionBuilder* version_builder() { return version_builder_; }
S
sdong 已提交
567 568

 private:
569
  VersionBuilder* version_builder_;
S
sdong 已提交
570 571
  Version* version_;
};
572 573
}  // anonymous namespace

574 575
Status Version::GetTableProperties(std::shared_ptr<const TableProperties>* tp,
                                   const FileMetaData* file_meta,
576
                                   const std::string* fname) const {
I
Igor Canadi 已提交
577
  auto table_cache = cfd_->table_cache();
578
  auto ioptions = cfd_->ioptions();
579
  Status s = table_cache->GetTableProperties(
580
      vset_->env_options_, cfd_->internal_comparator(), file_meta->fd,
581 582 583 584 585 586 587 588 589 590 591 592 593 594
      tp, true /* no io */);
  if (s.ok()) {
    return s;
  }

  // We only ignore error type `Incomplete` since it's by design that we
  // disallow table when it's not in table cache.
  if (!s.IsIncomplete()) {
    return s;
  }

  // 2. Table is not present in table cache, we'll read the table properties
  // directly from the properties block in the file.
  std::unique_ptr<RandomAccessFile> file;
595
  std::string file_name;
596
  if (fname != nullptr) {
597
    file_name = *fname;
598
  } else {
599 600 601
    file_name =
      TableFileName(vset_->db_options_->db_paths, file_meta->fd.GetNumber(),
                    file_meta->fd.GetPathId());
602
  }
603
  s = ioptions->env->NewRandomAccessFile(file_name, &file, vset_->env_options_);
604 605 606 607 608 609 610
  if (!s.ok()) {
    return s;
  }

  TableProperties* raw_table_properties;
  // By setting the magic number to kInvalidTableMagicNumber, we can by
  // pass the magic number check in the footer.
611
  std::unique_ptr<RandomAccessFileReader> file_reader(
612
      new RandomAccessFileReader(std::move(file), file_name));
613
  s = ReadTableProperties(
614
      file_reader.get(), file_meta->fd.GetFileSize(),
615
      Footer::kInvalidTableMagicNumber /* table's magic number */, *ioptions, &raw_table_properties);
616 617 618
  if (!s.ok()) {
    return s;
  }
619
  RecordTick(ioptions->statistics, NUMBER_DIRECT_LOAD_TABLE_PROPERTIES);
620 621 622 623 624 625

  *tp = std::shared_ptr<const TableProperties>(raw_table_properties);
  return s;
}

Status Version::GetPropertiesOfAllTables(TablePropertiesCollection* props) {
626
  Status s;
S
sdong 已提交
627
  for (int level = 0; level < storage_info_.num_levels_; level++) {
628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650
    s = GetPropertiesOfAllTables(props, level);
    if (!s.ok()) {
      return s;
    }
  }

  return Status::OK();
}

Status Version::GetPropertiesOfAllTables(TablePropertiesCollection* props,
                                         int level) {
  for (const auto& file_meta : storage_info_.files_[level]) {
    auto fname =
        TableFileName(vset_->db_options_->db_paths, file_meta->fd.GetNumber(),
                      file_meta->fd.GetPathId());
    // 1. If the table is already present in table cache, load table
    // properties from there.
    std::shared_ptr<const TableProperties> table_properties;
    Status s = GetTableProperties(&table_properties, file_meta, &fname);
    if (s.ok()) {
      props->insert({fname, table_properties});
    } else {
      return s;
651 652 653 654 655 656
    }
  }

  return Status::OK();
}

657
Status Version::GetPropertiesOfTablesInRange(
658
    const Range* range, std::size_t n, TablePropertiesCollection* props) const {
659
  for (int level = 0; level < storage_info_.num_non_empty_levels(); level++) {
660
    for (decltype(n) i = 0; i < n; i++) {
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
      // Convert user_key into a corresponding internal key.
      InternalKey k1(range[i].start, kMaxSequenceNumber, kValueTypeForSeek);
      InternalKey k2(range[i].limit, kMaxSequenceNumber, kValueTypeForSeek);
      std::vector<FileMetaData*> files;
      storage_info_.GetOverlappingInputs(level, &k1, &k2, &files, -1, nullptr,
                                         false);
      for (const auto& file_meta : files) {
        auto fname =
            TableFileName(vset_->db_options_->db_paths,
                          file_meta->fd.GetNumber(), file_meta->fd.GetPathId());
        if (props->count(fname) == 0) {
          // 1. If the table is already present in table cache, load table
          // properties from there.
          std::shared_ptr<const TableProperties> table_properties;
          Status s = GetTableProperties(&table_properties, file_meta, &fname);
          if (s.ok()) {
            props->insert({fname, table_properties});
          } else {
            return s;
          }
        }
      }
    }
  }

  return Status::OK();
}

689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709
Status Version::GetAggregatedTableProperties(
    std::shared_ptr<const TableProperties>* tp, int level) {
  TablePropertiesCollection props;
  Status s;
  if (level < 0) {
    s = GetPropertiesOfAllTables(&props);
  } else {
    s = GetPropertiesOfAllTables(&props, level);
  }
  if (!s.ok()) {
    return s;
  }

  auto* new_tp = new TableProperties();
  for (const auto& item : props) {
    new_tp->Add(*item.second);
  }
  tp->reset(new_tp);
  return Status::OK();
}

710 711
size_t Version::GetMemoryUsageByTableReaders() {
  size_t total_usage = 0;
S
sdong 已提交
712
  for (auto& file_level : storage_info_.level_files_brief_) {
713 714
    for (size_t i = 0; i < file_level.num_files; i++) {
      total_usage += cfd_->table_cache()->GetMemoryUsageByTableReader(
715
          vset_->env_options_, cfd_->internal_comparator(),
716 717 718 719 720 721
          file_level.files[i].fd);
    }
  }
  return total_usage;
}

722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747
void Version::GetColumnFamilyMetaData(ColumnFamilyMetaData* cf_meta) {
  assert(cf_meta);
  assert(cfd_);

  cf_meta->name = cfd_->GetName();
  cf_meta->size = 0;
  cf_meta->file_count = 0;
  cf_meta->levels.clear();

  auto* ioptions = cfd_->ioptions();
  auto* vstorage = storage_info();

  for (int level = 0; level < cfd_->NumberLevels(); level++) {
    uint64_t level_size = 0;
    cf_meta->file_count += vstorage->LevelFiles(level).size();
    std::vector<SstFileMetaData> files;
    for (const auto& file : vstorage->LevelFiles(level)) {
      uint32_t path_id = file->fd.GetPathId();
      std::string file_path;
      if (path_id < ioptions->db_paths.size()) {
        file_path = ioptions->db_paths[path_id].path;
      } else {
        assert(!ioptions->db_paths.empty());
        file_path = ioptions->db_paths.back().path;
      }
      files.emplace_back(
748 749
          MakeTableFileName("", file->fd.GetNumber()), file_path,
          file->fd.GetFileSize(), file->smallest_seqno, file->largest_seqno,
750 751
          file->smallest.user_key().ToString(),
          file->largest.user_key().ToString(),
752
          file->stats.num_reads_sampled.load(std::memory_order_relaxed),
753 754 755 756 757 758 759 760 761 762
          file->being_compacted);
      level_size += file->fd.GetFileSize();
    }
    cf_meta->levels.emplace_back(
        level, level_size, std::move(files));
    cf_meta->size += level_size;
  }
}


S
sdong 已提交
763
uint64_t VersionStorageInfo::GetEstimatedActiveKeys() const {
764 765
  // Estimation will be inaccurate when:
  // (1) there exist merge keys
S
sdong 已提交
766 767
  // (2) keys are directly overwritten
  // (3) deletion on non-existing keys
768
  // (4) low number of samples
769
  if (current_num_samples_ == 0) {
770 771 772
    return 0;
  }

773
  if (current_num_non_deletions_ <= current_num_deletions_) {
774 775 776
    return 0;
  }

777
  uint64_t est = current_num_non_deletions_ - current_num_deletions_;
778

Y
Yueh-Hsuan Chiang 已提交
779 780 781 782 783
  uint64_t file_count = 0;
  for (int level = 0; level < num_levels_; ++level) {
    file_count += files_[level].size();
  }

784
  if (current_num_samples_ < file_count) {
785
    // casting to avoid overflowing
786
    return
D
Dmitri Smirnov 已提交
787
      static_cast<uint64_t>(
D
Dmitri Smirnov 已提交
788
        (est * static_cast<double>(file_count) / current_num_samples_)
D
Dmitri Smirnov 已提交
789
      );
790
  } else {
791
    return est;
792
  }
S
sdong 已提交
793 794
}

795 796 797 798 799 800 801 802 803 804 805 806 807 808 809
double VersionStorageInfo::GetEstimatedCompressionRatioAtLevel(
    int level) const {
  assert(level < num_levels_);
  uint64_t sum_file_size_bytes = 0;
  uint64_t sum_data_size_bytes = 0;
  for (auto* file_meta : files_[level]) {
    sum_file_size_bytes += file_meta->fd.GetFileSize();
    sum_data_size_bytes += file_meta->raw_key_size + file_meta->raw_value_size;
  }
  if (sum_file_size_bytes == 0) {
    return -1.0;
  }
  return static_cast<double>(sum_data_size_bytes) / sum_file_size_bytes;
}

810 811
void Version::AddIterators(const ReadOptions& read_options,
                           const EnvOptions& soptions,
A
Andrew Kryczka 已提交
812 813
                           MergeIteratorBuilder* merge_iter_builder,
                           RangeDelAggregator* range_del_agg) {
S
sdong 已提交
814
  assert(storage_info_.finalized_);
815

816
  for (int level = 0; level < storage_info_.num_non_empty_levels(); level++) {
A
Andrew Kryczka 已提交
817 818
    AddIteratorsForLevel(read_options, soptions, merge_iter_builder, level,
                         range_del_agg);
819 820 821 822 823 824
  }
}

void Version::AddIteratorsForLevel(const ReadOptions& read_options,
                                   const EnvOptions& soptions,
                                   MergeIteratorBuilder* merge_iter_builder,
A
Andrew Kryczka 已提交
825 826
                                   int level,
                                   RangeDelAggregator* range_del_agg) {
827 828 829 830 831 832
  assert(storage_info_.finalized_);
  if (level >= storage_info_.num_non_empty_levels()) {
    // This is an empty level
    return;
  } else if (storage_info_.LevelFilesBrief(level).num_files == 0) {
    // No files in this level
S
sdong 已提交
833 834 835
    return;
  }

836 837
  bool should_sample = should_sample_file_read();

838
  auto* arena = merge_iter_builder->GetArena();
839 840 841 842 843
  if (level == 0) {
    // Merge all level zero files together since they may overlap
    for (size_t i = 0; i < storage_info_.LevelFilesBrief(0).num_files; i++) {
      const auto& file = storage_info_.LevelFilesBrief(0).files[i];
      merge_iter_builder->AddIterator(cfd_->table_cache()->NewIterator(
844 845 846
          read_options, soptions, cfd_->internal_comparator(), file.fd,
          range_del_agg, nullptr, cfd_->internal_stats()->GetFileReadHist(0),
          false, arena, false /* skip_filters */, 0 /* level */));
847
    }
848 849 850 851 852 853 854 855 856
    if (should_sample) {
      // Count ones for every L0 files. This is done per iterator creation
      // rather than Seek(), while files in other levels are recored per seek.
      // If users execute one range query per iterator, there may be some
      // discrepancy here.
      for (FileMetaData* meta : storage_info_.LevelFiles(0)) {
        sample_file_read_inc(meta);
      }
    }
857 858 859 860 861
  } else {
    // For levels > 0, we can use a concatenating iterator that sequentially
    // walks through the non-overlapping files in the level, opening them
    // lazily.
    auto* mem = arena->AllocateAligned(sizeof(LevelFileIteratorState));
A
Andrew Kryczka 已提交
862 863 864 865 866 867 868
    auto* state = new (mem)
        LevelFileIteratorState(cfd_->table_cache(), read_options, soptions,
                               cfd_->internal_comparator(),
                               cfd_->internal_stats()->GetFileReadHist(level),
                               false /* for_compaction */,
                               cfd_->ioptions()->prefix_extractor != nullptr,
                               IsFilterSkipped(level), level, range_del_agg);
869 870
    mem = arena->AllocateAligned(sizeof(LevelFileNumIterator));
    auto* first_level_iter = new (mem) LevelFileNumIterator(
871 872
        cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level),
        should_sample_file_read());
873 874
    merge_iter_builder->AddIterator(
        NewTwoLevelIterator(state, first_level_iter, arena, false));
875 876 877
  }
}

878
void Version::AddRangeDelIteratorsForLevel(
879 880 881
    const ReadOptions& read_options, const EnvOptions& soptions, int level,
    std::vector<InternalIterator*>* range_del_iters) {
  range_del_iters->clear();
882 883
  for (size_t i = 0; i < storage_info_.LevelFilesBrief(level).num_files; i++) {
    const auto& file = storage_info_.LevelFilesBrief(level).files[i];
884 885 886 887 888 889 890
    auto* range_del_iter = cfd_->table_cache()->NewRangeTombstoneIterator(
        read_options, soptions, cfd_->internal_comparator(), file.fd,
        cfd_->internal_stats()->GetFileReadHist(level),
        false /* skip_filters */, level);
    if (range_del_iter != nullptr) {
      range_del_iters->push_back(range_del_iter);
    }
891 892 893
  }
}

S
sdong 已提交
894 895
VersionStorageInfo::VersionStorageInfo(
    const InternalKeyComparator* internal_comparator,
I
Igor Canadi 已提交
896
    const Comparator* user_comparator, int levels,
897 898
    CompactionStyle compaction_style, VersionStorageInfo* ref_vstorage,
    bool _force_consistency_checks)
S
sdong 已提交
899 900
    : internal_comparator_(internal_comparator),
      user_comparator_(user_comparator),
901
      // cfd is nullptr if Version is dummy
I
Igor Canadi 已提交
902
      num_levels_(levels),
S
sdong 已提交
903
      num_non_empty_levels_(0),
S
sdong 已提交
904 905
      file_indexer_(user_comparator),
      compaction_style_(compaction_style),
906
      files_(new std::vector<FileMetaData*>[num_levels_]),
S
sdong 已提交
907
      base_level_(num_levels_ == 1 ? -1 : 1),
908
      files_by_compaction_pri_(num_levels_),
909
      level0_non_overlapping_(false),
910 911 912
      next_file_to_compact_by_size_(num_levels_),
      compaction_score_(num_levels_),
      compaction_level_(num_levels_),
S
sdong 已提交
913
      l0_delay_trigger_count_(0),
914 915 916 917 918
      accumulated_file_size_(0),
      accumulated_raw_key_size_(0),
      accumulated_raw_value_size_(0),
      accumulated_num_non_deletions_(0),
      accumulated_num_deletions_(0),
919 920 921
      current_num_non_deletions_(0),
      current_num_deletions_(0),
      current_num_samples_(0),
922
      estimated_compaction_needed_bytes_(0),
923 924
      finalized_(false),
      force_consistency_checks_(_force_consistency_checks) {
S
sdong 已提交
925 926 927 928 929 930 931
  if (ref_vstorage != nullptr) {
    accumulated_file_size_ = ref_vstorage->accumulated_file_size_;
    accumulated_raw_key_size_ = ref_vstorage->accumulated_raw_key_size_;
    accumulated_raw_value_size_ = ref_vstorage->accumulated_raw_value_size_;
    accumulated_num_non_deletions_ =
        ref_vstorage->accumulated_num_non_deletions_;
    accumulated_num_deletions_ = ref_vstorage->accumulated_num_deletions_;
932 933 934
    current_num_non_deletions_ = ref_vstorage->current_num_non_deletions_;
    current_num_deletions_ = ref_vstorage->current_num_deletions_;
    current_num_samples_ = ref_vstorage->current_num_samples_;
935
  }
936
}
937

I
Igor Canadi 已提交
938
Version::Version(ColumnFamilyData* column_family_data, VersionSet* vset,
S
sdong 已提交
939
                 uint64_t version_number)
940 941
    : env_(vset->env_),
      cfd_(column_family_data),
I
Igor Canadi 已提交
942 943 944 945 946 947
      info_log_((cfd_ == nullptr) ? nullptr : cfd_->ioptions()->info_log),
      db_statistics_((cfd_ == nullptr) ? nullptr
                                       : cfd_->ioptions()->statistics),
      table_cache_((cfd_ == nullptr) ? nullptr : cfd_->table_cache()),
      merge_operator_((cfd_ == nullptr) ? nullptr
                                        : cfd_->ioptions()->merge_operator),
948 949 950 951 952 953 954 955 956 957
      storage_info_(
          (cfd_ == nullptr) ? nullptr : &cfd_->internal_comparator(),
          (cfd_ == nullptr) ? nullptr : cfd_->user_comparator(),
          cfd_ == nullptr ? 0 : cfd_->NumberLevels(),
          cfd_ == nullptr ? kCompactionStyleLevel
                          : cfd_->ioptions()->compaction_style,
          (cfd_ == nullptr || cfd_->current() == nullptr)
              ? nullptr
              : cfd_->current()->storage_info(),
          cfd_ == nullptr ? false : cfd_->ioptions()->force_consistency_checks),
S
sdong 已提交
958 959 960 961 962 963
      vset_(vset),
      next_(this),
      prev_(this),
      refs_(0),
      version_number_(version_number) {}

964
void Version::Get(const ReadOptions& read_options, const LookupKey& k,
M
Maysam Yabandeh 已提交
965
                  PinnableSlice* value, Status* status,
A
Andrew Kryczka 已提交
966 967
                  MergeContext* merge_context,
                  RangeDelAggregator* range_del_agg, bool* value_found,
968
                  bool* key_exists, SequenceNumber* seq) {
969 970
  Slice ikey = k.internal_key();
  Slice user_key = k.user_key();
971 972

  assert(status->ok() || status->IsMergeInProgress());
973

974 975 976 977 978
  if (key_exists != nullptr) {
    // will falsify below if not found
    *key_exists = true;
  }

979
  PinnedIteratorsManager pinned_iters_mgr;
S
sdong 已提交
980
  GetContext get_context(
S
sdong 已提交
981
      user_comparator(), merge_operator_, info_log_, db_statistics_,
S
sdong 已提交
982
      status->ok() ? GetContext::kNotFound : GetContext::kMerge, user_key,
M
Maysam Yabandeh 已提交
983
      value, value_found, merge_context, range_del_agg, this->env_, seq,
984 985 986 987 988 989
      merge_operator_ ? &pinned_iters_mgr : nullptr);

  // Pin blocks that we read to hold merge operands
  if (merge_operator_) {
    pinned_iters_mgr.StartPinning();
  }
990

S
sdong 已提交
991 992 993 994
  FilePicker fp(
      storage_info_.files_, user_key, ikey, &storage_info_.level_files_brief_,
      storage_info_.num_non_empty_levels_, &storage_info_.file_indexer_,
      user_comparator(), internal_comparator());
995 996
  FdWithKeyRange* f = fp.GetNextFile();
  while (f != nullptr) {
997 998 999
    if (get_context.sample()) {
      sample_file_read_inc(f->file_metadata);
    }
1000 1001
    *status = table_cache_->Get(
        read_options, *internal_comparator(), f->fd, ikey, &get_context,
1002
        cfd_->internal_stats()->GetFileReadHist(fp.GetHitFileLevel()),
1003
        IsFilterSkipped(static_cast<int>(fp.GetHitFileLevel()),
1004 1005
                        fp.IsHitFileLastInLevel()),
        fp.GetCurrentLevel());
1006 1007 1008
    // TODO: examine the behavior for corrupted key
    if (!status->ok()) {
      return;
1009
    }
1010

1011 1012 1013 1014 1015
    switch (get_context.State()) {
      case GetContext::kNotFound:
        // Keep searching in other files
        break;
      case GetContext::kFound:
1016 1017 1018 1019 1020 1021 1022
        if (fp.GetHitFileLevel() == 0) {
          RecordTick(db_statistics_, GET_HIT_L0);
        } else if (fp.GetHitFileLevel() == 1) {
          RecordTick(db_statistics_, GET_HIT_L1);
        } else if (fp.GetHitFileLevel() >= 2) {
          RecordTick(db_statistics_, GET_HIT_L2_AND_UP);
        }
1023
        return;
1024 1025 1026
      case GetContext::kDeleted:
        // Use empty error message for speed
        *status = Status::NotFound();
1027
        return;
1028
      case GetContext::kCorrupt:
1029 1030
        *status = Status::Corruption("corrupted key for ", user_key);
        return;
1031
      case GetContext::kMerge:
1032
        break;
1033
    }
1034
    f = fp.GetNextFile();
1035 1036
  }

1037
  if (GetContext::kMerge == get_context.State()) {
1038 1039 1040 1041 1042
    if (!merge_operator_) {
      *status =  Status::InvalidArgument(
          "merge_operator is not properly initialized.");
      return;
    }
1043 1044
    // merge_operands are in saver and we hit the beginning of the key history
    // do a final merge of nullptr and operands;
M
Maysam Yabandeh 已提交
1045 1046 1047
    std::string* str_value = value != nullptr ? value->GetSelf() : nullptr;
    *status = MergeHelper::TimedFullMerge(
        merge_operator_, user_key, nullptr, merge_context->GetOperands(),
1048 1049
        str_value, info_log_, db_statistics_, env_,
        nullptr /* result_operand */, true);
M
Maysam Yabandeh 已提交
1050 1051 1052
    if (LIKELY(value != nullptr)) {
      value->PinSelf();
    }
1053
  } else {
1054 1055 1056
    if (key_exists != nullptr) {
      *key_exists = false;
    }
1057
    *status = Status::NotFound(); // Use an empty error message for speed
1058
  }
1059 1060
}

1061
bool Version::IsFilterSkipped(int level, bool is_file_last_in_level) {
1062 1063
  // Reaching the bottom level implies misses at all upper levels, so we'll
  // skip checking the filters when we predict a hit.
1064 1065
  return cfd_->ioptions()->optimize_filters_for_hits &&
         (level > 0 || is_file_last_in_level) &&
1066 1067 1068
         level == storage_info_.num_non_empty_levels() - 1;
}

S
sdong 已提交
1069
void VersionStorageInfo::GenerateLevelFilesBrief() {
1070
  level_files_brief_.resize(num_non_empty_levels_);
1071
  for (int level = 0; level < num_non_empty_levels_; level++) {
1072 1073
    DoGenerateLevelFilesBrief(
        &level_files_brief_[level], files_[level], &arena_);
1074 1075 1076
  }
}

1077 1078 1079 1080
void Version::PrepareApply(
    const MutableCFOptions& mutable_cf_options,
    bool update_stats) {
  UpdateAccumulatedStats(update_stats);
S
sdong 已提交
1081
  storage_info_.UpdateNumNonEmptyLevels();
1082
  storage_info_.CalculateBaseBytes(*cfd_->ioptions(), mutable_cf_options);
Y
Yi Wu 已提交
1083
  storage_info_.UpdateFilesByCompactionPri(cfd_->ioptions()->compaction_pri);
S
sdong 已提交
1084 1085
  storage_info_.GenerateFileIndexer();
  storage_info_.GenerateLevelFilesBrief();
1086
  storage_info_.GenerateLevel0NonOverlapping();
1087 1088
}

1089
bool Version::MaybeInitializeFileMetaData(FileMetaData* file_meta) {
1090 1091
  if (file_meta->init_stats_from_file ||
      file_meta->compensated_file_size > 0) {
1092 1093 1094 1095
    return false;
  }
  std::shared_ptr<const TableProperties> tp;
  Status s = GetTableProperties(&tp, file_meta);
1096
  file_meta->init_stats_from_file = true;
1097
  if (!s.ok()) {
1098 1099 1100 1101
    ROCKS_LOG_ERROR(vset_->db_options_->info_log,
                    "Unable to load table properties for file %" PRIu64
                    " --- %s\n",
                    file_meta->fd.GetNumber(), s.ToString().c_str());
1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112
    return false;
  }
  if (tp.get() == nullptr) return false;
  file_meta->num_entries = tp->num_entries;
  file_meta->num_deletions = GetDeletedKeys(tp->user_collected_properties);
  file_meta->raw_value_size = tp->raw_value_size;
  file_meta->raw_key_size = tp->raw_key_size;

  return true;
}

S
sdong 已提交
1113
void VersionStorageInfo::UpdateAccumulatedStats(FileMetaData* file_meta) {
1114 1115 1116 1117 1118 1119 1120
  assert(file_meta->init_stats_from_file);
  accumulated_file_size_ += file_meta->fd.GetFileSize();
  accumulated_raw_key_size_ += file_meta->raw_key_size;
  accumulated_raw_value_size_ += file_meta->raw_value_size;
  accumulated_num_non_deletions_ +=
      file_meta->num_entries - file_meta->num_deletions;
  accumulated_num_deletions_ += file_meta->num_deletions;
1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134

  current_num_non_deletions_ +=
      file_meta->num_entries - file_meta->num_deletions;
  current_num_deletions_ += file_meta->num_deletions;
  current_num_samples_++;
}

void VersionStorageInfo::RemoveCurrentStats(FileMetaData* file_meta) {
  if (file_meta->init_stats_from_file) {
    current_num_non_deletions_ -=
        file_meta->num_entries - file_meta->num_deletions;
    current_num_deletions_ -= file_meta->num_deletions;
    current_num_samples_--;
  }
1135 1136
}

1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
void Version::UpdateAccumulatedStats(bool update_stats) {
  if (update_stats) {
    // maximum number of table properties loaded from files.
    const int kMaxInitCount = 20;
    int init_count = 0;
    // here only the first kMaxInitCount files which haven't been
    // initialized from file will be updated with num_deletions.
    // The motivation here is to cap the maximum I/O per Version creation.
    // The reason for choosing files from lower-level instead of higher-level
    // is that such design is able to propagate the initialization from
    // lower-level to higher-level:  When the num_deletions of lower-level
    // files are updated, it will make the lower-level files have accurate
    // compensated_file_size, making lower-level to higher-level compaction
    // will be triggered, which creates higher-level files whose num_deletions
    // will be updated here.
    for (int level = 0;
         level < storage_info_.num_levels_ && init_count < kMaxInitCount;
         ++level) {
      for (auto* file_meta : storage_info_.files_[level]) {
        if (MaybeInitializeFileMetaData(file_meta)) {
          // each FileMeta will be initialized only once.
          storage_info_.UpdateAccumulatedStats(file_meta);
1159 1160
          // when option "max_open_files" is -1, all the file metadata has
          // already been read, so MaybeInitializeFileMetaData() won't incur
L
Leonidas Galanis 已提交
1161 1162
          // any I/O cost. "max_open_files=-1" means that the table cache passed
          // to the VersionSet and then to the ColumnFamilySet has a size of
1163
          // TableCache::kInfiniteCapacity
L
Leonidas Galanis 已提交
1164
          if (vset_->GetColumnFamilySet()->get_table_cache()->GetCapacity() ==
1165
              TableCache::kInfiniteCapacity) {
1166 1167
            continue;
          }
1168 1169 1170
          if (++init_count >= kMaxInitCount) {
            break;
          }
1171 1172 1173
        }
      }
    }
1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
    // In case all sampled-files contain only deletion entries, then we
    // load the table-property of a file in higher-level to initialize
    // that value.
    for (int level = storage_info_.num_levels_ - 1;
         storage_info_.accumulated_raw_value_size_ == 0 && level >= 0;
         --level) {
      for (int i = static_cast<int>(storage_info_.files_[level].size()) - 1;
           storage_info_.accumulated_raw_value_size_ == 0 && i >= 0; --i) {
        if (MaybeInitializeFileMetaData(storage_info_.files_[level][i])) {
          storage_info_.UpdateAccumulatedStats(storage_info_.files_[level][i]);
        }
1185 1186 1187 1188
      }
    }
  }

S
sdong 已提交
1189
  storage_info_.ComputeCompensatedSizes();
S
sdong 已提交
1190 1191 1192 1193
}

void VersionStorageInfo::ComputeCompensatedSizes() {
  static const int kDeletionWeightOnCompaction = 2;
1194 1195 1196 1197 1198
  uint64_t average_value_size = GetAverageValueSize();

  // compute the compensated size
  for (int level = 0; level < num_levels_; level++) {
    for (auto* file_meta : files_[level]) {
1199
      // Here we only compute compensated_file_size for those file_meta
I
Igor Canadi 已提交
1200 1201 1202
      // which compensated_file_size is uninitialized (== 0). This is true only
      // for files that have been created right now and no other thread has
      // access to them. That's why we can safely mutate compensated_file_size.
1203
      if (file_meta->compensated_file_size == 0) {
1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214
        file_meta->compensated_file_size = file_meta->fd.GetFileSize();
        // Here we only boost the size of deletion entries of a file only
        // when the number of deletion entries is greater than the number of
        // non-deletion entries in the file.  The motivation here is that in
        // a stable workload, the number of deletion entries should be roughly
        // equal to the number of non-deletion entries.  If we compensate the
        // size of deletion entries in a stable workload, the deletion
        // compensation logic might introduce unwanted effet which changes the
        // shape of LSM tree.
        if (file_meta->num_deletions * 2 >= file_meta->num_entries) {
          file_meta->compensated_file_size +=
1215 1216
              (file_meta->num_deletions * 2 - file_meta->num_entries) *
              average_value_size * kDeletionWeightOnCompaction;
1217
        }
1218
      }
1219 1220 1221 1222
    }
  }
}

S
sdong 已提交
1223 1224
int VersionStorageInfo::MaxInputLevel() const {
  if (compaction_style_ == kCompactionStyleLevel) {
1225
    return num_levels() - 2;
S
sdong 已提交
1226 1227 1228 1229
  }
  return 0;
}

1230 1231 1232 1233
void VersionStorageInfo::EstimateCompactionBytesNeeded(
    const MutableCFOptions& mutable_cf_options) {
  // Only implemented for level-based compaction
  if (compaction_style_ != kCompactionStyleLevel) {
1234
    estimated_compaction_needed_bytes_ = 0;
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247
    return;
  }

  // Start from Level 0, if level 0 qualifies compaction to level 1,
  // we estimate the size of compaction.
  // Then we move on to the next level and see whether it qualifies compaction
  // to the next level. The size of the level is estimated as the actual size
  // on the level plus the input bytes from the previous level if there is any.
  // If it exceeds, take the exceeded bytes as compaction input and add the size
  // of the compaction size to tatal size.
  // We keep doing it to Level 2, 3, etc, until the last level and return the
  // accumulated bytes.

I
Igor Canadi 已提交
1248
  uint64_t bytes_compact_to_next_level = 0;
1249 1250 1251 1252
  uint64_t level_size = 0;
  for (auto* f : files_[0]) {
    level_size += f->fd.GetFileSize();
  }
1253 1254
  // Level 0
  bool level0_compact_triggered = false;
1255 1256 1257
  if (static_cast<int>(files_[0].size()) >=
          mutable_cf_options.level0_file_num_compaction_trigger ||
      level_size >= mutable_cf_options.max_bytes_for_level_base) {
1258
    level0_compact_triggered = true;
1259 1260
    estimated_compaction_needed_bytes_ = level_size;
    bytes_compact_to_next_level = level_size;
1261 1262 1263 1264 1265
  } else {
    estimated_compaction_needed_bytes_ = 0;
  }

  // Level 1 and up.
1266
  uint64_t bytes_next_level = 0;
1267
  for (int level = base_level(); level <= MaxInputLevel(); level++) {
1268
    level_size = 0;
1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282
    if (bytes_next_level > 0) {
#ifndef NDEBUG
      uint64_t level_size2 = 0;
      for (auto* f : files_[level]) {
        level_size2 += f->fd.GetFileSize();
      }
      assert(level_size2 == bytes_next_level);
#endif
      level_size = bytes_next_level;
      bytes_next_level = 0;
    } else {
      for (auto* f : files_[level]) {
        level_size += f->fd.GetFileSize();
      }
1283 1284 1285 1286 1287 1288 1289 1290
    }
    if (level == base_level() && level0_compact_triggered) {
      // Add base level size to compaction if level0 compaction triggered.
      estimated_compaction_needed_bytes_ += level_size;
    }
    // Add size added by previous compaction
    level_size += bytes_compact_to_next_level;
    bytes_compact_to_next_level = 0;
I
Igor Canadi 已提交
1291
    uint64_t level_target = MaxBytesForLevel(level);
1292 1293
    if (level_size > level_target) {
      bytes_compact_to_next_level = level_size - level_target;
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310
      // Estimate the actual compaction fan-out ratio as size ratio between
      // the two levels.

      assert(bytes_next_level == 0);
      if (level + 1 < num_levels_) {
        for (auto* f : files_[level + 1]) {
          bytes_next_level += f->fd.GetFileSize();
        }
      }
      if (bytes_next_level > 0) {
        assert(level_size > 0);
        estimated_compaction_needed_bytes_ += static_cast<uint64_t>(
            static_cast<double>(bytes_compact_to_next_level) *
            (static_cast<double>(bytes_next_level) /
                 static_cast<double>(level_size) +
             1));
      }
1311 1312 1313 1314
    }
  }
}

S
Sagar Vemuri 已提交
1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340
namespace {
uint32_t GetExpiredTtlFilesCount(const ImmutableCFOptions& ioptions,
                                 const std::vector<FileMetaData*>& files) {
  uint32_t ttl_expired_files_count = 0;

  int64_t _current_time;
  auto status = ioptions.env->GetCurrentTime(&_current_time);
  if (status.ok()) {
    const uint64_t current_time = static_cast<uint64_t>(_current_time);
    for (auto f : files) {
      if (!f->being_compacted && f->fd.table_reader != nullptr &&
          f->fd.table_reader->GetTableProperties() != nullptr) {
        auto creation_time =
            f->fd.table_reader->GetTableProperties()->creation_time;
        if (creation_time > 0 &&
            creation_time <
                (current_time - ioptions.compaction_options_fifo.ttl)) {
          ttl_expired_files_count++;
        }
      }
    }
  }
  return ttl_expired_files_count;
}
}  // anonymous namespace

S
sdong 已提交
1341
void VersionStorageInfo::ComputeCompactionScore(
Y
Yi Wu 已提交
1342
    const ImmutableCFOptions& immutable_cf_options,
1343
    const MutableCFOptions& mutable_cf_options) {
S
sdong 已提交
1344
  for (int level = 0; level <= MaxInputLevel(); level++) {
1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
    double score;
    if (level == 0) {
      // We treat level-0 specially by bounding the number of files
      // instead of number of bytes for two reasons:
      //
      // (1) With larger write-buffer sizes, it is nice not to do too
      // many level-0 compactions.
      //
      // (2) The files in level-0 are merged on every read and
      // therefore we wish to avoid too many files when the individual
      // file size is small (perhaps because of a small write-buffer
      // setting, or very high compression ratios, or lots of
      // overwrites/deletions).
S
sdong 已提交
1358
      int num_sorted_runs = 0;
I
Igor Canadi 已提交
1359
      uint64_t total_size = 0;
1360 1361 1362
      for (auto* f : files_[level]) {
        if (!f->being_compacted) {
          total_size += f->compensated_file_size;
S
sdong 已提交
1363
          num_sorted_runs++;
1364 1365
        }
      }
S
sdong 已提交
1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376
      if (compaction_style_ == kCompactionStyleUniversal) {
        // For universal compaction, we use level0 score to indicate
        // compaction score for the whole DB. Adding other levels as if
        // they are L0 files.
        for (int i = 1; i < num_levels(); i++) {
          if (!files_[i].empty() && !files_[i][0]->being_compacted) {
            num_sorted_runs++;
          }
        }
      }

S
sdong 已提交
1377
      if (compaction_style_ == kCompactionStyleFIFO) {
Y
Yi Wu 已提交
1378 1379 1380
        score =
            static_cast<double>(total_size) /
            immutable_cf_options.compaction_options_fifo.max_table_files_size;
1381 1382 1383 1384 1385 1386
        if (immutable_cf_options.compaction_options_fifo.allow_compaction) {
          score = std::max(
              static_cast<double>(num_sorted_runs) /
                  mutable_cf_options.level0_file_num_compaction_trigger,
              score);
        }
S
Sagar Vemuri 已提交
1387 1388 1389 1390 1391
        if (immutable_cf_options.compaction_options_fifo.ttl > 0) {
          score = std::max(static_cast<double>(GetExpiredTtlFilesCount(
                               immutable_cf_options, files_[level])),
                           score);
        }
1392

1393
      } else {
S
sdong 已提交
1394
        score = static_cast<double>(num_sorted_runs) /
1395
                mutable_cf_options.level0_file_num_compaction_trigger;
A
Andrew Kryczka 已提交
1396 1397 1398 1399 1400
        if (compaction_style_ == kCompactionStyleLevel && num_levels() > 1) {
          // Level-based involves L0->L0 compactions that can lead to oversized
          // L0 files. Take into account size as well to avoid later giant
          // compactions to the base level.
          score = std::max(
1401 1402
              score, static_cast<double>(total_size) /
                     mutable_cf_options.max_bytes_for_level_base);
A
Andrew Kryczka 已提交
1403
        }
1404 1405 1406
      }
    } else {
      // Compute the ratio of current size to size limit.
I
Igor Canadi 已提交
1407 1408
      uint64_t level_bytes_no_compacting = 0;
      for (auto f : files_[level]) {
1409
        if (!f->being_compacted) {
I
Igor Canadi 已提交
1410 1411 1412 1413
          level_bytes_no_compacting += f->compensated_file_size;
        }
      }
      score = static_cast<double>(level_bytes_no_compacting) /
1414
              MaxBytesForLevel(level);
1415 1416 1417 1418 1419 1420 1421
    }
    compaction_level_[level] = level;
    compaction_score_[level] = score;
  }

  // sort all the levels based on their score. Higher scores get listed
  // first. Use bubble sort because the number of entries are small.
1422 1423
  for (int i = 0; i < num_levels() - 2; i++) {
    for (int j = i + 1; j < num_levels() - 1; j++) {
1424 1425 1426 1427 1428 1429 1430 1431 1432 1433
      if (compaction_score_[i] < compaction_score_[j]) {
        double score = compaction_score_[i];
        int level = compaction_level_[i];
        compaction_score_[i] = compaction_score_[j];
        compaction_level_[i] = compaction_level_[j];
        compaction_score_[j] = score;
        compaction_level_[j] = level;
      }
    }
  }
1434
  ComputeFilesMarkedForCompaction();
1435
  EstimateCompactionBytesNeeded(mutable_cf_options);
1436 1437 1438 1439
}

void VersionStorageInfo::ComputeFilesMarkedForCompaction() {
  files_marked_for_compaction_.clear();
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
  int last_qualify_level = 0;

  // Do not include files from the last level with data
  // If table properties collector suggests a file on the last level,
  // we should not move it to a new level.
  for (int level = num_levels() - 1; level >= 1; level--) {
    if (!files_[level].empty()) {
      last_qualify_level = level - 1;
      break;
    }
  }

  for (int level = 0; level <= last_qualify_level; level++) {
1453 1454 1455 1456 1457 1458
    for (auto* f : files_[level]) {
      if (!f->being_compacted && f->marked_for_compaction) {
        files_marked_for_compaction_.emplace_back(level, f);
      }
    }
  }
1459 1460 1461
}

namespace {
1462 1463 1464

// used to sort files by size
struct Fsize {
1465
  size_t index;
1466 1467 1468
  FileMetaData* file;
};

1469 1470
// Compator that is used to sort files based on their size
// In normal mode: descending size
1471
bool CompareCompensatedSizeDescending(const Fsize& first, const Fsize& second) {
1472 1473
  return (first.file->compensated_file_size >
      second.file->compensated_file_size);
1474
}
K
kailiu 已提交
1475
} // anonymous namespace
1476

1477
void VersionStorageInfo::AddFile(int level, FileMetaData* f, Logger* info_log) {
S
sdong 已提交
1478 1479
  auto* level_files = &files_[level];
  // Must not overlap
1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497
#ifndef NDEBUG
  if (level > 0 && !level_files->empty() &&
      internal_comparator_->Compare(
          (*level_files)[level_files->size() - 1]->largest, f->smallest) >= 0) {
    auto* f2 = (*level_files)[level_files->size() - 1];
    if (info_log != nullptr) {
      Error(info_log, "Adding new file %" PRIu64
                      " range (%s, %s) to level %d but overlapping "
                      "with existing file %" PRIu64 " %s %s",
            f->fd.GetNumber(), f->smallest.DebugString(true).c_str(),
            f->largest.DebugString(true).c_str(), level, f2->fd.GetNumber(),
            f2->smallest.DebugString(true).c_str(),
            f2->largest.DebugString(true).c_str());
      LogFlush(info_log);
    }
    assert(false);
  }
#endif
S
sdong 已提交
1498 1499 1500 1501
  f->refs++;
  level_files->push_back(f);
}

1502 1503 1504 1505
// Version::PrepareApply() need to be called before calling the function, or
// following functions called:
// 1. UpdateNumNonEmptyLevels();
// 2. CalculateBaseBytes();
1506
// 3. UpdateFilesByCompactionPri();
1507 1508
// 4. GenerateFileIndexer();
// 5. GenerateLevelFilesBrief();
1509
// 6. GenerateLevel0NonOverlapping();
1510 1511 1512
void VersionStorageInfo::SetFinalized() {
  finalized_ = true;
#ifndef NDEBUG
S
sdong 已提交
1513 1514 1515 1516 1517 1518
  if (compaction_style_ != kCompactionStyleLevel) {
    // Not level based compaction.
    return;
  }
  assert(base_level_ < 0 || num_levels() == 1 ||
         (base_level_ >= 1 && base_level_ < num_levels()));
1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
  // Verify all levels newer than base_level are empty except L0
  for (int level = 1; level < base_level(); level++) {
    assert(NumLevelBytes(level) == 0);
  }
  uint64_t max_bytes_prev_level = 0;
  for (int level = base_level(); level < num_levels() - 1; level++) {
    if (LevelFiles(level).size() == 0) {
      continue;
    }
    assert(MaxBytesForLevel(level) >= max_bytes_prev_level);
    max_bytes_prev_level = MaxBytesForLevel(level);
  }
  int num_empty_non_l0_level = 0;
  for (int level = 0; level < num_levels(); level++) {
    assert(LevelFiles(level).size() == 0 ||
           LevelFiles(level).size() == LevelFilesBrief(level).num_files);
    if (level > 0 && NumLevelBytes(level) > 0) {
      num_empty_non_l0_level++;
    }
    if (LevelFiles(level).size() > 0) {
      assert(level < num_non_empty_levels());
    }
  }
  assert(compaction_level_.size() > 0);
  assert(compaction_level_.size() == compaction_score_.size());
#endif
}

S
sdong 已提交
1547
void VersionStorageInfo::UpdateNumNonEmptyLevels() {
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557
  num_non_empty_levels_ = num_levels_;
  for (int i = num_levels_ - 1; i >= 0; i--) {
    if (files_[i].size() != 0) {
      return;
    } else {
      num_non_empty_levels_ = i;
    }
  }
}

1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
namespace {
// Sort `temp` based on ratio of overlapping size over file size
void SortFileByOverlappingRatio(
    const InternalKeyComparator& icmp, const std::vector<FileMetaData*>& files,
    const std::vector<FileMetaData*>& next_level_files,
    std::vector<Fsize>* temp) {
  std::unordered_map<uint64_t, uint64_t> file_to_order;
  auto next_level_it = next_level_files.begin();

  for (auto& file : files) {
    uint64_t overlapping_bytes = 0;
    // Skip files in next level that is smaller than current file
    while (next_level_it != next_level_files.end() &&
           icmp.Compare((*next_level_it)->largest, file->smallest) < 0) {
      next_level_it++;
    }

    while (next_level_it != next_level_files.end() &&
           icmp.Compare((*next_level_it)->smallest, file->largest) < 0) {
      overlapping_bytes += (*next_level_it)->fd.file_size;

      if (icmp.Compare((*next_level_it)->largest, file->largest) > 0) {
        // next level file cross large boundary of current file.
        break;
      }
      next_level_it++;
    }

    assert(file->fd.file_size != 0);
    file_to_order[file->fd.GetNumber()] =
        overlapping_bytes * 1024u / file->fd.file_size;
  }

  std::sort(temp->begin(), temp->end(),
            [&](const Fsize& f1, const Fsize& f2) -> bool {
              return file_to_order[f1.file->fd.GetNumber()] <
                     file_to_order[f2.file->fd.GetNumber()];
            });
}
}  // namespace

1599
void VersionStorageInfo::UpdateFilesByCompactionPri(
Y
Yi Wu 已提交
1600
    CompactionPri compaction_pri) {
S
sdong 已提交
1601 1602
  if (compaction_style_ == kCompactionStyleFIFO ||
      compaction_style_ == kCompactionStyleUniversal) {
I
Igor Canadi 已提交
1603 1604 1605
    // don't need this
    return;
  }
1606
  // No need to sort the highest level because it is never compacted.
1607
  for (int level = 0; level < num_levels() - 1; level++) {
1608
    const std::vector<FileMetaData*>& files = files_[level];
1609 1610
    auto& files_by_compaction_pri = files_by_compaction_pri_[level];
    assert(files_by_compaction_pri.size() == 0);
1611 1612 1613

    // populate a temp vector for sorting based on size
    std::vector<Fsize> temp(files.size());
1614
    for (size_t i = 0; i < files.size(); i++) {
1615 1616 1617 1618
      temp[i].index = i;
      temp[i].file = files[i];
    }

S
sdong 已提交
1619 1620
    // sort the top number_of_files_to_sort_ based on file size
    size_t num = VersionStorageInfo::kNumberFilesToSort;
1621 1622
    if (num > temp.size()) {
      num = temp.size();
1623
    }
Y
Yi Wu 已提交
1624
    switch (compaction_pri) {
1625
      case kByCompensatedSize:
1626 1627 1628
        std::partial_sort(temp.begin(), temp.begin() + num, temp.end(),
                          CompareCompensatedSizeDescending);
        break;
1629
      case kOldestLargestSeqFirst:
1630
        std::sort(temp.begin(), temp.end(),
1631
                  [](const Fsize& f1, const Fsize& f2) -> bool {
1632 1633 1634
                    return f1.file->largest_seqno < f2.file->largest_seqno;
                  });
        break;
1635 1636
      case kOldestSmallestSeqFirst:
        std::sort(temp.begin(), temp.end(),
1637
                  [](const Fsize& f1, const Fsize& f2) -> bool {
1638 1639 1640
                    return f1.file->smallest_seqno < f2.file->smallest_seqno;
                  });
        break;
1641 1642 1643 1644
      case kMinOverlappingRatio:
        SortFileByOverlappingRatio(*internal_comparator_, files_[level],
                                   files_[level + 1], &temp);
        break;
1645 1646 1647
      default:
        assert(false);
    }
1648 1649
    assert(temp.size() == files.size());

1650
    // initialize files_by_compaction_pri_
1651 1652
    for (size_t i = 0; i < temp.size(); i++) {
      files_by_compaction_pri.push_back(static_cast<int>(temp[i].index));
1653 1654
    }
    next_file_to_compact_by_size_[level] = 0;
1655
    assert(files_[level].size() == files_by_compaction_pri_[level].size());
1656 1657 1658
  }
}

1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
void VersionStorageInfo::GenerateLevel0NonOverlapping() {
  assert(!finalized_);
  level0_non_overlapping_ = true;
  if (level_files_brief_.size() == 0) {
    return;
  }

  // A copy of L0 files sorted by smallest key
  std::vector<FdWithKeyRange> level0_sorted_file(
      level_files_brief_[0].files,
      level_files_brief_[0].files + level_files_brief_[0].num_files);
1670 1671 1672 1673 1674
  std::sort(level0_sorted_file.begin(), level0_sorted_file.end(),
            [this](const FdWithKeyRange& f1, const FdWithKeyRange& f2) -> bool {
              return (internal_comparator_->Compare(f1.smallest_key,
                                                    f2.smallest_key) < 0);
            });
1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685

  for (size_t i = 1; i < level0_sorted_file.size(); ++i) {
    FdWithKeyRange& f = level0_sorted_file[i];
    FdWithKeyRange& prev = level0_sorted_file[i - 1];
    if (internal_comparator_->Compare(prev.largest_key, f.smallest_key) >= 0) {
      level0_non_overlapping_ = false;
      break;
    }
  }
}

J
jorlow@chromium.org 已提交
1686 1687 1688 1689
void Version::Ref() {
  ++refs_;
}

1690
bool Version::Unref() {
J
jorlow@chromium.org 已提交
1691 1692 1693
  assert(refs_ >= 1);
  --refs_;
  if (refs_ == 0) {
1694
    delete this;
1695
    return true;
J
jorlow@chromium.org 已提交
1696
  }
1697
  return false;
J
jorlow@chromium.org 已提交
1698 1699
}

S
sdong 已提交
1700 1701 1702
bool VersionStorageInfo::OverlapInLevel(int level,
                                        const Slice* smallest_user_key,
                                        const Slice* largest_user_key) {
1703 1704 1705 1706
  if (level >= num_non_empty_levels_) {
    // empty level, no overlap
    return false;
  }
S
sdong 已提交
1707
  return SomeFileOverlapsRange(*internal_comparator_, (level > 0),
1708
                               level_files_brief_[level], smallest_user_key,
1709
                               largest_user_key);
G
Gabor Cselle 已提交
1710 1711 1712
}

// Store in "*inputs" all files in "level" that overlap [begin,end]
A
Abhishek Kona 已提交
1713
// If hint_index is specified, then it points to a file in the
1714 1715
// overlapping range.
// The file_index returns a pointer to any file in an overlapping range.
S
sdong 已提交
1716 1717
void VersionStorageInfo::GetOverlappingInputs(
    int level, const InternalKey* begin, const InternalKey* end,
1718 1719
    std::vector<FileMetaData*>* inputs, int hint_index, int* file_index,
    bool expand_range) const {
1720 1721 1722 1723 1724
  if (level >= num_non_empty_levels_) {
    // this level is empty, no overlapping inputs
    return;
  }

G
Gabor Cselle 已提交
1725 1726
  inputs->clear();
  Slice user_begin, user_end;
A
Abhishek Kona 已提交
1727
  if (begin != nullptr) {
G
Gabor Cselle 已提交
1728 1729
    user_begin = begin->user_key();
  }
A
Abhishek Kona 已提交
1730
  if (end != nullptr) {
G
Gabor Cselle 已提交
1731 1732
    user_end = end->user_key();
  }
1733 1734 1735
  if (file_index) {
    *file_index = -1;
  }
S
sdong 已提交
1736
  const Comparator* user_cmp = user_comparator_;
A
Abhishek Kona 已提交
1737
  if (begin != nullptr && end != nullptr && level > 0) {
A
Aaron Gao 已提交
1738 1739
    GetOverlappingInputsRangeBinarySearch(level, user_begin, user_end, inputs,
                                          hint_index, file_index);
1740 1741
    return;
  }
A
Aaron Gao 已提交
1742

1743 1744
  for (size_t i = 0; i < level_files_brief_[level].num_files; ) {
    FdWithKeyRange* f = &(level_files_brief_[level].files[i++]);
1745 1746
    const Slice file_start = ExtractUserKey(f->smallest_key);
    const Slice file_limit = ExtractUserKey(f->largest_key);
A
Abhishek Kona 已提交
1747
    if (begin != nullptr && user_cmp->Compare(file_limit, user_begin) < 0) {
G
Gabor Cselle 已提交
1748
      // "f" is completely before specified range; skip it
A
Abhishek Kona 已提交
1749
    } else if (end != nullptr && user_cmp->Compare(file_start, user_end) > 0) {
G
Gabor Cselle 已提交
1750 1751
      // "f" is completely after specified range; skip it
    } else {
1752
      inputs->push_back(files_[level][i-1]);
1753
      if (level == 0 && expand_range) {
H
Hans Wennborg 已提交
1754 1755
        // Level-0 files may overlap each other.  So check if the newly
        // added file has expanded the range.  If so, restart search.
A
Abhishek Kona 已提交
1756
        if (begin != nullptr && user_cmp->Compare(file_start, user_begin) < 0) {
H
Hans Wennborg 已提交
1757 1758 1759
          user_begin = file_start;
          inputs->clear();
          i = 0;
A
Abhishek Kona 已提交
1760 1761
        } else if (end != nullptr
            && user_cmp->Compare(file_limit, user_end) > 0) {
H
Hans Wennborg 已提交
1762 1763 1764 1765
          user_end = file_limit;
          inputs->clear();
          i = 0;
        }
1766
      } else if (file_index) {
1767
        *file_index = static_cast<int>(i) - 1;
H
Hans Wennborg 已提交
1768
      }
G
Gabor Cselle 已提交
1769 1770
    }
  }
1771 1772
}

A
Aaron Gao 已提交
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804
// Store in "*inputs" files in "level" that within range [begin,end]
// Guarantee a "clean cut" boundary between the files in inputs
// and the surrounding files and the maxinum number of files.
// This will ensure that no parts of a key are lost during compaction.
// If hint_index is specified, then it points to a file in the range.
// The file_index returns a pointer to any file in an overlapping range.
void VersionStorageInfo::GetCleanInputsWithinInterval(
    int level, const InternalKey* begin, const InternalKey* end,
    std::vector<FileMetaData*>* inputs, int hint_index, int* file_index) const {
  if (level >= num_non_empty_levels_) {
    // this level is empty, no inputs within range
    return;
  }

  inputs->clear();
  Slice user_begin, user_end;
  if (begin != nullptr) {
    user_begin = begin->user_key();
  }
  if (end != nullptr) {
    user_end = end->user_key();
  }
  if (file_index) {
    *file_index = -1;
  }
  if (begin != nullptr && end != nullptr && level > 0) {
    GetOverlappingInputsRangeBinarySearch(level, user_begin, user_end, inputs,
                                          hint_index, file_index,
                                          true /* within_interval */);
  }
}

1805 1806 1807 1808
// Store in "*inputs" all files in "level" that overlap [begin,end]
// Employ binary search to find at least one file that overlaps the
// specified range. From that file, iterate backwards and
// forwards to find all overlapping files.
A
Aaron Gao 已提交
1809 1810 1811 1812
// if within_range is set, then only store the maximum clean inputs
// within range [begin, end]. "clean" means there is a boudnary
// between the files in "*inputs" and the surrounding files
void VersionStorageInfo::GetOverlappingInputsRangeBinarySearch(
S
sdong 已提交
1813
    int level, const Slice& user_begin, const Slice& user_end,
A
Aaron Gao 已提交
1814 1815
    std::vector<FileMetaData*>* inputs, int hint_index, int* file_index,
    bool within_interval) const {
1816 1817 1818
  assert(level > 0);
  int min = 0;
  int mid = 0;
1819
  int max = static_cast<int>(files_[level].size()) - 1;
1820
  bool foundOverlap = false;
S
sdong 已提交
1821
  const Comparator* user_cmp = user_comparator_;
1822 1823 1824 1825 1826 1827 1828 1829 1830

  // if the caller already knows the index of a file that has overlap,
  // then we can skip the binary search.
  if (hint_index != -1) {
    mid = hint_index;
    foundOverlap = true;
  }

  while (!foundOverlap && min <= max) {
1831
    mid = (min + max)/2;
1832
    FdWithKeyRange* f = &(level_files_brief_[level].files[mid]);
1833 1834
    const Slice file_start = ExtractUserKey(f->smallest_key);
    const Slice file_limit = ExtractUserKey(f->largest_key);
A
Aaron Gao 已提交
1835 1836
    if ((!within_interval && user_cmp->Compare(file_limit, user_begin) < 0) ||
        (within_interval && user_cmp->Compare(file_start, user_begin) < 0)) {
1837
      min = mid + 1;
A
Aaron Gao 已提交
1838 1839 1840 1841
    } else if ((!within_interval &&
                user_cmp->Compare(user_end, file_start) < 0) ||
               (within_interval &&
                user_cmp->Compare(user_end, file_limit) < 0)) {
1842 1843 1844 1845 1846 1847
      max = mid - 1;
    } else {
      foundOverlap = true;
      break;
    }
  }
A
Abhishek Kona 已提交
1848

1849 1850 1851 1852
  // If there were no overlapping files, return immediately.
  if (!foundOverlap) {
    return;
  }
1853 1854 1855 1856
  // returns the index where an overlap is found
  if (file_index) {
    *file_index = mid;
  }
A
Aaron Gao 已提交
1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870

  int start_index, end_index;
  if (within_interval) {
    ExtendFileRangeWithinInterval(level, user_begin, user_end, mid, &start_index,
                                  &end_index);
  } else {
    ExtendFileRangeOverlappingInterval(level, user_begin, user_end, mid,
                                       &start_index, &end_index);
  }
  assert(end_index >= start_index);
  // insert overlapping files into vector
  for (int i = start_index; i <= end_index; i++) {
    inputs->push_back(files_[level][i]);
  }
1871
}
A
Abhishek Kona 已提交
1872

A
Aaron Gao 已提交
1873 1874 1875
// Store in *start_index and *end_index the range of all files in
// "level" that overlap [begin,end]
// The mid_index specifies the index of at least one file that
1876 1877
// overlaps the specified range. From that file, iterate backward
// and forward to find all overlapping files.
S
sdong 已提交
1878
// Use FileLevel in searching, make it faster
A
Aaron Gao 已提交
1879
void VersionStorageInfo::ExtendFileRangeOverlappingInterval(
S
sdong 已提交
1880
    int level, const Slice& user_begin, const Slice& user_end,
A
Aaron Gao 已提交
1881
    unsigned int mid_index, int* start_index, int* end_index) const {
S
sdong 已提交
1882
  const Comparator* user_cmp = user_comparator_;
1883
  const FdWithKeyRange* files = level_files_brief_[level].files;
1884 1885
#ifndef NDEBUG
  {
A
Aaron Gao 已提交
1886 1887 1888
    // assert that the file at mid_index overlaps with the range
    assert(mid_index < level_files_brief_[level].num_files);
    const FdWithKeyRange* f = &files[mid_index];
1889 1890
    const Slice fstart = ExtractUserKey(f->smallest_key);
    const Slice flimit = ExtractUserKey(f->largest_key);
1891 1892 1893 1894 1895 1896 1897
    if (user_cmp->Compare(fstart, user_begin) >= 0) {
      assert(user_cmp->Compare(fstart, user_end) <= 0);
    } else {
      assert(user_cmp->Compare(flimit, user_begin) >= 0);
    }
  }
#endif
A
Aaron Gao 已提交
1898 1899
  *start_index = mid_index + 1;
  *end_index = mid_index;
1900
  int count __attribute__((unused)) = 0;
1901 1902

  // check backwards from 'mid' to lower indices
A
Aaron Gao 已提交
1903
  for (int i = mid_index; i >= 0 ; i--) {
1904 1905
    const FdWithKeyRange* f = &files[i];
    const Slice file_limit = ExtractUserKey(f->largest_key);
1906
    if (user_cmp->Compare(file_limit, user_begin) >= 0) {
A
Aaron Gao 已提交
1907
      *start_index = i;
1908
      assert((count++, true));
1909 1910 1911 1912 1913
    } else {
      break;
    }
  }
  // check forward from 'mid+1' to higher indices
A
Aaron Gao 已提交
1914
  for (unsigned int i = mid_index+1;
1915
       i < level_files_brief_[level].num_files; i++) {
1916 1917
    const FdWithKeyRange* f = &files[i];
    const Slice file_start = ExtractUserKey(f->smallest_key);
1918
    if (user_cmp->Compare(file_start, user_end) <= 0) {
1919
      assert((count++, true));
A
Aaron Gao 已提交
1920
      *end_index = i;
1921 1922 1923 1924
    } else {
      break;
    }
  }
A
Aaron Gao 已提交
1925
  assert(count == *end_index - *start_index + 1);
1926 1927
}

A
Aaron Gao 已提交
1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938
// Store in *start_index and *end_index the clean range of all files in
// "level" within [begin,end]
// The mid_index specifies the index of at least one file within
// the specified range. From that file, iterate backward
// and forward to find all overlapping files and then "shrink" to
// the clean range required.
// Use FileLevel in searching, make it faster
void VersionStorageInfo::ExtendFileRangeWithinInterval(
    int level, const Slice& user_begin, const Slice& user_end,
    unsigned int mid_index, int* start_index, int* end_index) const {
  assert(level != 0);
S
sdong 已提交
1939
  const Comparator* user_cmp = user_comparator_;
1940
  const FdWithKeyRange* files = level_files_brief_[level].files;
A
Aaron Gao 已提交
1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971
#ifndef NDEBUG
  {
    // assert that the file at mid_index is within the range
    assert(mid_index < level_files_brief_[level].num_files);
    const FdWithKeyRange* f = &files[mid_index];
    const Slice fstart = ExtractUserKey(f->smallest_key);
    const Slice flimit = ExtractUserKey(f->largest_key);
    assert(user_cmp->Compare(fstart, user_begin) >= 0 &&
           user_cmp->Compare(flimit, user_end) <= 0);
  }
#endif
  ExtendFileRangeOverlappingInterval(level, user_begin, user_end, mid_index,
                                     start_index, end_index);
  int left = *start_index;
  int right = *end_index;
  // shrink from left to right
  while (left <= right) {
    const Slice& first_key_in_range = ExtractUserKey(files[left].smallest_key);
    if (user_cmp->Compare(first_key_in_range, user_begin) < 0) {
      left++;
      continue;
    }
    if (left > 0) {  // If not first file
      const Slice& last_key_before =
          ExtractUserKey(files[left - 1].largest_key);
      if (user_cmp->Equal(first_key_in_range, last_key_before)) {
        // The first user key in range overlaps with the previous file's last
        // key
        left++;
        continue;
      }
1972
    }
A
Aaron Gao 已提交
1973
    break;
1974
  }
A
Aaron Gao 已提交
1975 1976 1977 1978 1979 1980
  // shrink from right to left
  while (left <= right) {
    const Slice last_key_in_range = ExtractUserKey(files[right].largest_key);
    if (user_cmp->Compare(last_key_in_range, user_end) > 0) {
      right--;
      continue;
1981
    }
A
Aaron Gao 已提交
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992
    if (right < static_cast<int>(level_files_brief_[level].num_files) -
                    1) {  // If not the last file
      const Slice first_key_after =
          ExtractUserKey(files[right + 1].smallest_key);
      if (user_cmp->Equal(last_key_in_range, first_key_after)) {
        // The last user key in range overlaps with the next file's first key
        right--;
        continue;
      }
    }
    break;
1993 1994
  }

A
Aaron Gao 已提交
1995 1996
  *start_index = left;
  *end_index = right;
1997 1998
}

S
sdong 已提交
1999
uint64_t VersionStorageInfo::NumLevelBytes(int level) const {
2000
  assert(level >= 0);
2001
  assert(level < num_levels());
2002 2003 2004
  return TotalFileSize(files_[level]);
}

S
sdong 已提交
2005 2006
const char* VersionStorageInfo::LevelSummary(
    LevelSummaryStorage* scratch) const {
2007
  int len = 0;
S
sdong 已提交
2008
  if (compaction_style_ == kCompactionStyleLevel && num_levels() > 1) {
2009 2010
    assert(base_level_ < static_cast<int>(level_max_bytes_.size()));
    len = snprintf(scratch->buffer, sizeof(scratch->buffer),
S
sdong 已提交
2011
                   "base level %d max bytes base %" PRIu64 " ", base_level_,
2012 2013 2014
                   level_max_bytes_[base_level_]);
  }
  len +=
S
sdong 已提交
2015
      snprintf(scratch->buffer + len, sizeof(scratch->buffer) - len, "files[");
2016
  for (int i = 0; i < num_levels(); i++) {
2017 2018 2019 2020 2021
    int sz = sizeof(scratch->buffer) - len;
    int ret = snprintf(scratch->buffer + len, sz, "%d ", int(files_[i].size()));
    if (ret < 0 || ret >= sz) break;
    len += ret;
  }
I
Igor Canadi 已提交
2022 2023 2024 2025
  if (len > 0) {
    // overwrite the last space
    --len;
  }
2026 2027 2028 2029 2030
  len += snprintf(scratch->buffer + len, sizeof(scratch->buffer) - len,
                  "] max score %.2f", compaction_score_[0]);

  if (!files_marked_for_compaction_.empty()) {
    snprintf(scratch->buffer + len, sizeof(scratch->buffer) - len,
2031
             " (%" ROCKSDB_PRIszt " files need compaction)",
2032 2033 2034
             files_marked_for_compaction_.size());
  }

2035 2036 2037
  return scratch->buffer;
}

S
sdong 已提交
2038 2039
const char* VersionStorageInfo::LevelFileSummary(FileSummaryStorage* scratch,
                                                 int level) const {
2040 2041 2042
  int len = snprintf(scratch->buffer, sizeof(scratch->buffer), "files_size[");
  for (const auto& f : files_[level]) {
    int sz = sizeof(scratch->buffer) - len;
I
Igor Canadi 已提交
2043
    char sztxt[16];
2044
    AppendHumanBytes(f->fd.GetFileSize(), sztxt, sizeof(sztxt));
2045
    int ret = snprintf(scratch->buffer + len, sz,
2046 2047
                       "#%" PRIu64 "(seq=%" PRIu64 ",sz=%s,%d) ",
                       f->fd.GetNumber(), f->smallest_seqno, sztxt,
I
Igor Canadi 已提交
2048
                       static_cast<int>(f->being_compacted));
2049 2050 2051 2052
    if (ret < 0 || ret >= sz)
      break;
    len += ret;
  }
I
Igor Canadi 已提交
2053 2054 2055 2056
  // overwrite the last space (only if files_[level].size() is non-zero)
  if (files_[level].size() && len > 0) {
    --len;
  }
2057 2058 2059 2060
  snprintf(scratch->buffer + len, sizeof(scratch->buffer) - len, "]");
  return scratch->buffer;
}

S
sdong 已提交
2061
int64_t VersionStorageInfo::MaxNextLevelOverlappingBytes() {
2062 2063
  uint64_t result = 0;
  std::vector<FileMetaData*> overlaps;
2064
  for (int level = 1; level < num_levels() - 1; level++) {
2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
    for (const auto& f : files_[level]) {
      GetOverlappingInputs(level + 1, &f->smallest, &f->largest, &overlaps);
      const uint64_t sum = TotalFileSize(overlaps);
      if (sum > result) {
        result = sum;
      }
    }
  }
  return result;
}

2076 2077 2078 2079 2080 2081 2082 2083 2084 2085
uint64_t VersionStorageInfo::MaxBytesForLevel(int level) const {
  // Note: the result for level zero is not really used since we set
  // the level-0 compaction threshold based on number of files.
  assert(level >= 0);
  assert(level < static_cast<int>(level_max_bytes_.size()));
  return level_max_bytes_[level];
}

void VersionStorageInfo::CalculateBaseBytes(const ImmutableCFOptions& ioptions,
                                            const MutableCFOptions& options) {
S
sdong 已提交
2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100
  // Special logic to set number of sorted runs.
  // It is to match the previous behavior when all files are in L0.
  int num_l0_count = static_cast<int>(files_[0].size());
  if (compaction_style_ == kCompactionStyleUniversal) {
    // For universal compaction, we use level0 score to indicate
    // compaction score for the whole DB. Adding other levels as if
    // they are L0 files.
    for (int i = 1; i < num_levels(); i++) {
      if (!files_[i].empty()) {
        num_l0_count++;
      }
    }
  }
  set_l0_delay_trigger_count(num_l0_count);

2101 2102
  level_max_bytes_.resize(ioptions.num_levels);
  if (!ioptions.level_compaction_dynamic_level_bytes) {
S
sdong 已提交
2103
    base_level_ = (ioptions.compaction_style == kCompactionStyleLevel) ? 1 : -1;
2104 2105 2106 2107 2108 2109 2110 2111 2112

    // Calculate for static bytes base case
    for (int i = 0; i < ioptions.num_levels; ++i) {
      if (i == 0 && ioptions.compaction_style == kCompactionStyleUniversal) {
        level_max_bytes_[i] = options.max_bytes_for_level_base;
      } else if (i > 1) {
        level_max_bytes_[i] = MultiplyCheckOverflow(
            MultiplyCheckOverflow(level_max_bytes_[i - 1],
                                  options.max_bytes_for_level_multiplier),
S
sdong 已提交
2113
            options.MaxBytesMultiplerAdditional(i - 1));
2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148
      } else {
        level_max_bytes_[i] = options.max_bytes_for_level_base;
      }
    }
  } else {
    uint64_t max_level_size = 0;

    int first_non_empty_level = -1;
    // Find size of non-L0 level of most data.
    // Cannot use the size of the last level because it can be empty or less
    // than previous levels after compaction.
    for (int i = 1; i < num_levels_; i++) {
      uint64_t total_size = 0;
      for (const auto& f : files_[i]) {
        total_size += f->fd.GetFileSize();
      }
      if (total_size > 0 && first_non_empty_level == -1) {
        first_non_empty_level = i;
      }
      if (total_size > max_level_size) {
        max_level_size = total_size;
      }
    }

    // Prefill every level's max bytes to disallow compaction from there.
    for (int i = 0; i < num_levels_; i++) {
      level_max_bytes_[i] = std::numeric_limits<uint64_t>::max();
    }

    if (max_level_size == 0) {
      // No data for L1 and up. L0 compacts to last level directly.
      // No compaction from L1+ needs to be scheduled.
      base_level_ = num_levels_ - 1;
    } else {
      uint64_t base_bytes_max = options.max_bytes_for_level_base;
2149 2150
      uint64_t base_bytes_min = static_cast<uint64_t>(
          base_bytes_max / options.max_bytes_for_level_multiplier);
2151 2152 2153 2154 2155

      // Try whether we can make last level's target size to be max_level_size
      uint64_t cur_level_size = max_level_size;
      for (int i = num_levels_ - 2; i >= first_non_empty_level; i--) {
        // Round up after dividing
2156 2157
        cur_level_size = static_cast<uint64_t>(
            cur_level_size / options.max_bytes_for_level_multiplier);
2158 2159 2160
      }

      // Calculate base level and its size.
2161
      uint64_t base_level_size;
2162 2163 2164 2165
      if (cur_level_size <= base_bytes_min) {
        // Case 1. If we make target size of last level to be max_level_size,
        // target size of the first non-empty level would be smaller than
        // base_bytes_min. We set it be base_bytes_min.
2166
        base_level_size = base_bytes_min + 1U;
2167
        base_level_ = first_non_empty_level;
2168 2169 2170
        ROCKS_LOG_WARN(ioptions.info_log,
                       "More existing levels in DB than needed. "
                       "max_bytes_for_level_multiplier may not be guaranteed.");
2171 2172 2173 2174 2175
      } else {
        // Find base level (where L0 data is compacted to).
        base_level_ = first_non_empty_level;
        while (base_level_ > 1 && cur_level_size > base_bytes_max) {
          --base_level_;
2176 2177
          cur_level_size = static_cast<uint64_t>(
              cur_level_size / options.max_bytes_for_level_multiplier);
2178 2179 2180 2181
        }
        if (cur_level_size > base_bytes_max) {
          // Even L1 will be too large
          assert(base_level_ == 1);
2182
          base_level_size = base_bytes_max;
2183
        } else {
2184
          base_level_size = cur_level_size;
2185 2186 2187
        }
      }

2188
      uint64_t level_size = base_level_size;
2189 2190
      for (int i = base_level_; i < num_levels_; i++) {
        if (i > base_level_) {
2191 2192
          level_size = MultiplyCheckOverflow(
              level_size, options.max_bytes_for_level_multiplier);
2193
        }
2194 2195 2196 2197 2198
        // Don't set any level below base_bytes_max. Otherwise, the LSM can
        // assume an hourglass shape where L1+ sizes are smaller than L0. This
        // causes compaction scoring, which depends on level sizes, to favor L1+
        // at the expense of L0, which may fill up and stall.
        level_max_bytes_[i] = std::max(level_size, base_bytes_max);
2199 2200 2201 2202 2203
      }
    }
  }
}

A
Andres Notzli 已提交
2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238
uint64_t VersionStorageInfo::EstimateLiveDataSize() const {
  // Estimate the live data size by adding up the size of the last level for all
  // key ranges. Note: Estimate depends on the ordering of files in level 0
  // because files in level 0 can be overlapping.
  uint64_t size = 0;

  auto ikey_lt = [this](InternalKey* x, InternalKey* y) {
    return internal_comparator_->Compare(*x, *y) < 0;
  };
  // (Ordered) map of largest keys in non-overlapping files
  std::map<InternalKey*, FileMetaData*, decltype(ikey_lt)> ranges(ikey_lt);

  for (int l = num_levels_ - 1; l >= 0; l--) {
    bool found_end = false;
    for (auto file : files_[l]) {
      // Find the first file where the largest key is larger than the smallest
      // key of the current file. If this file does not overlap with the
      // current file, none of the files in the map does. If there is
      // no potential overlap, we can safely insert the rest of this level
      // (if the level is not 0) into the map without checking again because
      // the elements in the level are sorted and non-overlapping.
      auto lb = (found_end && l != 0) ?
        ranges.end() : ranges.lower_bound(&file->smallest);
      found_end = (lb == ranges.end());
      if (found_end || internal_comparator_->Compare(
            file->largest, (*lb).second->smallest) < 0) {
          ranges.emplace_hint(lb, &file->largest, file);
          size += file->fd.file_size;
      }
    }
  }
  return size;
}


2239
void Version::AddLiveFiles(std::vector<FileDescriptor>* live) {
2240
  for (int level = 0; level < storage_info_.num_levels(); level++) {
S
sdong 已提交
2241
    const std::vector<FileMetaData*>& files = storage_info_.files_[level];
2242
    for (const auto& file : files) {
2243
      live->push_back(file->fd);
2244 2245 2246 2247
    }
  }
}

2248
std::string Version::DebugString(bool hex, bool print_stats) const {
J
jorlow@chromium.org 已提交
2249
  std::string r;
S
sdong 已提交
2250
  for (int level = 0; level < storage_info_.num_levels_; level++) {
2251 2252 2253 2254
    // E.g.,
    //   --- level 1 ---
    //   17:123['a' .. 'd']
    //   20:43['e' .. 'g']
2255 2256 2257
    //
    // if print_stats=true:
    //   17:123['a' .. 'd'](4096)
2258
    r.append("--- level ");
J
jorlow@chromium.org 已提交
2259
    AppendNumberTo(&r, level);
2260 2261
    r.append(" --- version# ");
    AppendNumberTo(&r, version_number_);
2262
    r.append(" ---\n");
S
sdong 已提交
2263
    const std::vector<FileMetaData*>& files = storage_info_.files_[level];
D
dgrogan@chromium.org 已提交
2264
    for (size_t i = 0; i < files.size(); i++) {
J
jorlow@chromium.org 已提交
2265
      r.push_back(' ');
2266
      AppendNumberTo(&r, files[i]->fd.GetNumber());
J
jorlow@chromium.org 已提交
2267
      r.push_back(':');
2268
      AppendNumberTo(&r, files[i]->fd.GetFileSize());
G
Gabor Cselle 已提交
2269
      r.append("[");
Z
Zheng Shao 已提交
2270
      r.append(files[i]->smallest.DebugString(hex));
G
Gabor Cselle 已提交
2271
      r.append(" .. ");
Z
Zheng Shao 已提交
2272
      r.append(files[i]->largest.DebugString(hex));
2273 2274 2275 2276 2277 2278 2279 2280
      r.append("]");
      if (print_stats) {
        r.append("(");
        r.append(ToString(
            files[i]->stats.num_reads_sampled.load(std::memory_order_relaxed)));
        r.append(")");
      }
      r.append("\n");
J
jorlow@chromium.org 已提交
2281 2282 2283 2284 2285
    }
  }
  return r;
}

2286 2287 2288 2289
// this is used to batch writes to the manifest file
struct VersionSet::ManifestWriter {
  Status status;
  bool done;
2290
  InstrumentedCondVar cv;
2291
  ColumnFamilyData* cfd;
2292
  const autovector<VersionEdit*>& edit_list;
A
Abhishek Kona 已提交
2293

2294
  explicit ManifestWriter(InstrumentedMutex* mu, ColumnFamilyData* _cfd,
2295 2296
                          const autovector<VersionEdit*>& e)
      : done(false), cv(mu), cfd(_cfd), edit_list(e) {}
2297 2298
};

2299 2300
VersionSet::VersionSet(const std::string& dbname,
                       const ImmutableDBOptions* db_options,
I
Igor Canadi 已提交
2301
                       const EnvOptions& storage_options, Cache* table_cache,
2302
                       WriteBufferManager* write_buffer_manager,
2303
                       WriteController* write_controller)
2304 2305 2306
    : column_family_set_(
          new ColumnFamilySet(dbname, db_options, storage_options, table_cache,
                              write_buffer_manager, write_controller)),
2307
      env_(db_options->env),
J
jorlow@chromium.org 已提交
2308
      dbname_(dbname),
2309
      db_options_(db_options),
J
jorlow@chromium.org 已提交
2310 2311
      next_file_number_(2),
      manifest_file_number_(0),  // Filled by Recover()
2312
      pending_manifest_file_number_(0),
2313
      last_sequence_(0),
2314
      last_to_be_written_sequence_(0),
2315
      prev_log_number_(0),
A
Abhishek Kona 已提交
2316
      current_version_number_(0),
2317
      manifest_file_size_(0),
I
Igor Canadi 已提交
2318
      env_options_(storage_options),
2319 2320
      env_options_compactions_(
          env_->OptimizeForCompactionTableRead(env_options_, *db_options_)) {}
J
jorlow@chromium.org 已提交
2321

2322 2323 2324 2325 2326
void CloseTables(void* ptr, size_t) {
  TableReader* table_reader = reinterpret_cast<TableReader*>(ptr);
  table_reader->Close();
}

J
jorlow@chromium.org 已提交
2327
VersionSet::~VersionSet() {
2328 2329
  // we need to delete column_family_set_ because its destructor depends on
  // VersionSet
2330 2331
  column_family_set_->get_table_cache()->ApplyToAllCacheEntries(&CloseTables,
                                                                false);
2332
  column_family_set_.reset();
2333 2334 2335 2336
  for (auto file : obsolete_files_) {
    delete file;
  }
  obsolete_files_.clear();
2337 2338
}

2339 2340
void VersionSet::AppendVersion(ColumnFamilyData* column_family_data,
                               Version* v) {
I
Igor Canadi 已提交
2341 2342
  // compute new compaction score
  v->storage_info()->ComputeCompactionScore(
Y
Yi Wu 已提交
2343
      *column_family_data->ioptions(),
2344
      *column_family_data->GetLatestMutableCFOptions());
I
Igor Canadi 已提交
2345

2346
  // Mark v finalized
S
sdong 已提交
2347
  v->storage_info_.SetFinalized();
2348

2349 2350
  // Make "v" current
  assert(v->refs_ == 0);
2351 2352 2353 2354 2355
  Version* current = column_family_data->current();
  assert(v != current);
  if (current != nullptr) {
    assert(current->refs_ > 0);
    current->Unref();
2356
  }
2357
  column_family_data->SetCurrent(v);
2358 2359 2360
  v->Ref();

  // Append to linked list
2361 2362
  v->prev_ = column_family_data->dummy_versions()->prev_;
  v->next_ = column_family_data->dummy_versions();
2363 2364 2365 2366
  v->prev_->next_ = v;
  v->next_->prev_ = v;
}

2367
Status VersionSet::LogAndApply(ColumnFamilyData* column_family_data,
2368
                               const MutableCFOptions& mutable_cf_options,
2369 2370 2371
                               const autovector<VersionEdit*>& edit_list,
                               InstrumentedMutex* mu, Directory* db_directory,
                               bool new_descriptor_log,
2372
                               const ColumnFamilyOptions* new_cf_options) {
2373
  mu->AssertHeld();
2374 2375 2376 2377 2378
  // num of edits
  auto num_edits = edit_list.size();
  if (num_edits == 0) {
    return Status::OK();
  } else if (num_edits > 1) {
2379
#ifndef NDEBUG
2380 2381 2382 2383
    // no group commits for column family add or drop
    for (auto& edit : edit_list) {
      assert(!edit->IsColumnFamilyManipulation());
    }
2384
#endif
2385
  }
2386

I
Igor Canadi 已提交
2387 2388 2389
  // column_family_data can be nullptr only if this is column_family_add.
  // in that case, we also need to specify ColumnFamilyOptions
  if (column_family_data == nullptr) {
2390 2391
    assert(num_edits == 1);
    assert(edit_list[0]->is_column_family_add_);
2392
    assert(new_cf_options != nullptr);
2393
  }
2394

2395
  // queue our request
2396
  ManifestWriter w(mu, column_family_data, edit_list);
2397 2398 2399
  manifest_writers_.push_back(&w);
  while (!w.done && &w != manifest_writers_.front()) {
    w.cv.Wait();
2400
  }
2401 2402 2403
  if (w.done) {
    return w.status;
  }
I
Igor Canadi 已提交
2404 2405 2406 2407 2408 2409 2410 2411
  if (column_family_data != nullptr && column_family_data->IsDropped()) {
    // if column family is dropped by the time we get here, no need to write
    // anything to the manifest
    manifest_writers_.pop_front();
    // Notify new head of write queue
    if (!manifest_writers_.empty()) {
      manifest_writers_.front()->cv.Signal();
    }
2412 2413
    // we steal this code to also inform about cf-drop
    return Status::ShutdownInProgress();
I
Igor Canadi 已提交
2414
  }
A
Abhishek Kona 已提交
2415

2416
  autovector<VersionEdit*> batch_edits;
2417
  Version* v = nullptr;
S
sdong 已提交
2418
  std::unique_ptr<BaseReferencedVersionBuilder> builder_guard(nullptr);
2419

2420 2421 2422
  // process all requests in the queue
  ManifestWriter* last_writer = &w;
  assert(!manifest_writers_.empty());
2423
  assert(manifest_writers_.front() == &w);
2424
  if (w.edit_list.front()->IsColumnFamilyManipulation()) {
2425
    // no group commits for column family add or drop
2426 2427
    LogAndApplyCFHelper(w.edit_list.front());
    batch_edits.push_back(w.edit_list.front());
2428 2429
  } else {
    v = new Version(column_family_data, this, current_version_number_++);
S
sdong 已提交
2430
    builder_guard.reset(new BaseReferencedVersionBuilder(column_family_data));
2431
    auto* builder = builder_guard->version_builder();
2432
    for (const auto& writer : manifest_writers_) {
2433
      if (writer->edit_list.front()->IsColumnFamilyManipulation() ||
2434 2435 2436 2437 2438 2439
          writer->cfd->GetID() != column_family_data->GetID()) {
        // no group commits for column family add or drop
        // also, group commits across column families are not supported
        break;
      }
      last_writer = writer;
2440 2441 2442 2443
      for (const auto& edit : writer->edit_list) {
        LogAndApplyHelper(column_family_data, builder, v, edit, mu);
        batch_edits.push_back(edit);
      }
2444
    }
S
sdong 已提交
2445
    builder->SaveTo(v->storage_info());
J
jorlow@chromium.org 已提交
2446 2447 2448 2449
  }

  // Initialize new descriptor log file if necessary by creating
  // a temporary file that contains a snapshot of the current version.
2450
  uint64_t new_manifest_file_size = 0;
2451
  Status s;
A
Abhishek Kona 已提交
2452

2453
  assert(pending_manifest_file_number_ == 0);
2454
  if (!descriptor_log_ ||
2455
      manifest_file_size_ > db_options_->max_manifest_file_size) {
2456
    pending_manifest_file_number_ = NewFileNumber();
2457
    batch_edits.back()->SetNextFile(next_file_number_.load());
A
Abhishek Kona 已提交
2458
    new_descriptor_log = true;
2459 2460
  } else {
    pending_manifest_file_number_ = manifest_file_number_;
A
Abhishek Kona 已提交
2461 2462
  }

I
Igor Canadi 已提交
2463
  if (new_descriptor_log) {
2464
    // if we're writing out new snapshot make sure to persist max column family
I
Igor Canadi 已提交
2465
    if (column_family_set_->GetMaxColumnFamily() > 0) {
2466 2467
      w.edit_list.front()->SetMaxColumnFamily(
          column_family_set_->GetMaxColumnFamily());
I
Igor Canadi 已提交
2468
    }
J
jorlow@chromium.org 已提交
2469 2470
  }

2471
  // Unlock during expensive operations. New writes cannot get here
2472
  // because &w is ensuring that all new writes get queued.
2473
  {
2474

2475
    mu->Unlock();
2476

S
sdong 已提交
2477
    TEST_SYNC_POINT("VersionSet::LogAndApply:WriteManifest");
2478
    if (!w.edit_list.front()->IsColumnFamilyManipulation() &&
L
Leonidas Galanis 已提交
2479
        this->GetColumnFamilySet()->get_table_cache()->GetCapacity() ==
2480
            TableCache::kInfiniteCapacity) {
2481 2482
      // unlimited table cache. Pre-load table handle now.
      // Need to do it out of the mutex.
2483
      builder_guard->version_builder()->LoadTableHandlers(
2484
          column_family_data->internal_stats(),
2485 2486
          column_family_data->ioptions()->optimize_filters_for_hits,
          true /* prefetch_index_and_filter_in_cache */);
2487 2488
    }

I
Igor Canadi 已提交
2489 2490
    // This is fine because everything inside of this block is serialized --
    // only one thread can be here at the same time
I
Igor Canadi 已提交
2491
    if (new_descriptor_log) {
2492
      // create manifest file
2493 2494
      ROCKS_LOG_INFO(db_options_->info_log, "Creating manifest %" PRIu64 "\n",
                     pending_manifest_file_number_);
I
Igor Canadi 已提交
2495
      unique_ptr<WritableFile> descriptor_file;
2496
      EnvOptions opt_env_opts = env_->OptimizeForManifestWrite(env_options_);
S
sdong 已提交
2497 2498
      s = NewWritableFile(
          env_, DescriptorFileName(dbname_, pending_manifest_file_number_),
2499
          &descriptor_file, opt_env_opts);
I
Igor Canadi 已提交
2500
      if (s.ok()) {
2501
        descriptor_file->SetPreallocationBlockSize(
2502
            db_options_->manifest_preallocation_size);
2503 2504 2505

        unique_ptr<WritableFileWriter> file_writer(
            new WritableFileWriter(std::move(descriptor_file), opt_env_opts));
2506 2507
        descriptor_log_.reset(
            new log::Writer(std::move(file_writer), 0, false));
I
Igor Canadi 已提交
2508 2509 2510 2511
        s = WriteSnapshot(descriptor_log_.get());
      }
    }

2512
    if (!w.edit_list.front()->IsColumnFamilyManipulation()) {
2513
      // This is cpu-heavy operations, which should be called outside mutex.
2514
      v->PrepareApply(mutable_cf_options, true);
2515
    }
2516 2517

    // Write new record to MANIFEST log
J
jorlow@chromium.org 已提交
2518
    if (s.ok()) {
I
Igor Canadi 已提交
2519 2520
      for (auto& e : batch_edits) {
        std::string record;
2521 2522 2523 2524 2525
        if (!e->EncodeTo(&record)) {
          s = Status::Corruption(
              "Unable to Encode VersionEdit:" + e->DebugString(true));
          break;
        }
S
sdong 已提交
2526 2527
        TEST_KILL_RANDOM("VersionSet::LogAndApply:BeforeAddRecord",
                         rocksdb_kill_odds * REDUCE_ODDS2);
2528 2529 2530 2531 2532
        s = descriptor_log_->AddRecord(record);
        if (!s.ok()) {
          break;
        }
      }
S
sdong 已提交
2533 2534
      if (s.ok()) {
        s = SyncManifest(env_, db_options_, descriptor_log_->file());
2535
      }
2536
      if (!s.ok()) {
2537 2538
        ROCKS_LOG_ERROR(db_options_->info_log, "MANIFEST write: %s\n",
                        s.ToString().c_str());
2539
      }
J
jorlow@chromium.org 已提交
2540 2541
    }

2542 2543
    // If we just created a new descriptor file, install it by writing a
    // new CURRENT file that points to it.
2544
    if (s.ok() && new_descriptor_log) {
A
Aaron Gao 已提交
2545 2546
      s = SetCurrentFile(env_, dbname_, pending_manifest_file_number_,
                         db_directory);
2547 2548
    }

2549 2550 2551 2552
    if (s.ok()) {
      // find offset in manifest file where this version is stored.
      new_manifest_file_size = descriptor_log_->file()->GetFileSize();
    }
A
Abhishek Kona 已提交
2553

2554
    if (w.edit_list.front()->is_column_family_drop_) {
2555
      TEST_SYNC_POINT("VersionSet::LogAndApply::ColumnFamilyDrop:0");
2556 2557 2558 2559
      TEST_SYNC_POINT("VersionSet::LogAndApply::ColumnFamilyDrop:1");
      TEST_SYNC_POINT("VersionSet::LogAndApply::ColumnFamilyDrop:2");
    }

2560
    LogFlush(db_options_->info_log);
2561
    TEST_SYNC_POINT("VersionSet::LogAndApply:WriteManifestDone");
2562
    mu->Lock();
J
jorlow@chromium.org 已提交
2563 2564
  }

2565 2566 2567 2568 2569 2570 2571
  // Append the old mainfest file to the obsolete_manifests_ list to be deleted
  // by PurgeObsoleteFiles later.
  if (s.ok() && new_descriptor_log) {
    obsolete_manifests_.emplace_back(
        DescriptorFileName("", manifest_file_number_));
  }

J
jorlow@chromium.org 已提交
2572 2573
  // Install the new version
  if (s.ok()) {
2574
    if (w.edit_list.front()->is_column_family_add_) {
2575 2576
      // no group commit on column family add
      assert(batch_edits.size() == 1);
2577
      assert(new_cf_options != nullptr);
2578 2579
      CreateColumnFamily(*new_cf_options, w.edit_list.front());
    } else if (w.edit_list.front()->is_column_family_drop_) {
2580
      assert(batch_edits.size() == 1);
I
Igor Canadi 已提交
2581
      column_family_data->SetDropped();
2582 2583 2584 2585
      if (column_family_data->Unref()) {
        delete column_family_data;
      }
    } else {
I
Igor Canadi 已提交
2586 2587 2588 2589 2590 2591 2592 2593
      uint64_t max_log_number_in_batch  = 0;
      for (auto& e : batch_edits) {
        if (e->has_log_number_) {
          max_log_number_in_batch =
              std::max(max_log_number_in_batch, e->log_number_);
        }
      }
      if (max_log_number_in_batch != 0) {
I
Igor Canadi 已提交
2594
        assert(column_family_data->GetLogNumber() <= max_log_number_in_batch);
I
Igor Canadi 已提交
2595 2596
        column_family_data->SetLogNumber(max_log_number_in_batch);
      }
2597 2598 2599
      AppendVersion(column_family_data, v);
    }

2600
    manifest_file_number_ = pending_manifest_file_number_;
2601
    manifest_file_size_ = new_manifest_file_size;
2602
    prev_log_number_ = w.edit_list.front()->prev_log_number_;
J
jorlow@chromium.org 已提交
2603
  } else {
2604 2605 2606 2607
    std::string version_edits;
    for (auto& e : batch_edits) {
      version_edits = version_edits + "\n" + e->DebugString(true);
    }
2608 2609
    ROCKS_LOG_ERROR(
        db_options_->info_log,
2610 2611 2612
        "[%s] Error in committing version edit to MANIFEST: %s",
        column_family_data ? column_family_data->GetName().c_str() : "<null>",
        version_edits.c_str());
J
jorlow@chromium.org 已提交
2613
    delete v;
2614
    if (new_descriptor_log) {
2615 2616 2617
      ROCKS_LOG_INFO(db_options_->info_log, "Deleting manifest %" PRIu64
                                            " current manifest %" PRIu64 "\n",
                     manifest_file_number_, pending_manifest_file_number_);
2618
      descriptor_log_.reset();
2619 2620
      env_->DeleteFile(
          DescriptorFileName(dbname_, pending_manifest_file_number_));
J
jorlow@chromium.org 已提交
2621 2622
    }
  }
2623
  pending_manifest_file_number_ = 0;
J
jorlow@chromium.org 已提交
2624

2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639
  // wake up all the waiting writers
  while (true) {
    ManifestWriter* ready = manifest_writers_.front();
    manifest_writers_.pop_front();
    if (ready != &w) {
      ready->status = s;
      ready->done = true;
      ready->cv.Signal();
    }
    if (ready == last_writer) break;
  }
  // Notify new head of write queue
  if (!manifest_writers_.empty()) {
    manifest_writers_.front()->cv.Signal();
  }
J
jorlow@chromium.org 已提交
2640 2641 2642
  return s;
}

I
Igor Canadi 已提交
2643 2644
void VersionSet::LogAndApplyCFHelper(VersionEdit* edit) {
  assert(edit->IsColumnFamilyManipulation());
2645
  edit->SetNextFile(next_file_number_.load());
I
Igor Canadi 已提交
2646 2647 2648 2649 2650 2651 2652 2653
  edit->SetLastSequence(last_sequence_);
  if (edit->is_column_family_drop_) {
    // if we drop column family, we have to make sure to save max column family,
    // so that we don't reuse existing ID
    edit->SetMaxColumnFamily(column_family_set_->GetMaxColumnFamily());
  }
}

S
sdong 已提交
2654
void VersionSet::LogAndApplyHelper(ColumnFamilyData* cfd,
2655
                                   VersionBuilder* builder, Version* v,
2656
                                   VersionEdit* edit, InstrumentedMutex* mu) {
2657
  mu->AssertHeld();
I
Igor Canadi 已提交
2658
  assert(!edit->IsColumnFamilyManipulation());
2659

2660 2661
  if (edit->has_log_number_) {
    assert(edit->log_number_ >= cfd->GetLogNumber());
2662
    assert(edit->log_number_ < next_file_number_.load());
I
Igor Canadi 已提交
2663
  }
2664

I
Igor Canadi 已提交
2665 2666 2667
  if (!edit->has_prev_log_number_) {
    edit->SetPrevLogNumber(prev_log_number_);
  }
2668
  edit->SetNextFile(next_file_number_.load());
I
Igor Canadi 已提交
2669 2670
  edit->SetLastSequence(last_sequence_);

2671
  builder->Apply(edit);
2672 2673
}

I
Igor Canadi 已提交
2674
Status VersionSet::Recover(
2675 2676
    const std::vector<ColumnFamilyDescriptor>& column_families,
    bool read_only) {
I
Igor Canadi 已提交
2677 2678 2679 2680 2681 2682 2683
  std::unordered_map<std::string, ColumnFamilyOptions> cf_name_to_options;
  for (auto cf : column_families) {
    cf_name_to_options.insert({cf.name, cf.options});
  }
  // keeps track of column families in manifest that were not found in
  // column families parameters. if those column families are not dropped
  // by subsequent manifest records, Recover() will return failure status
I
Igor Canadi 已提交
2684
  std::unordered_map<int, std::string> column_families_not_found;
J
jorlow@chromium.org 已提交
2685 2686

  // Read "CURRENT" file, which contains a pointer to the current manifest file
K
kailiu 已提交
2687 2688 2689 2690
  std::string manifest_filename;
  Status s = ReadFileToString(
      env_, CurrentFileName(dbname_), &manifest_filename
  );
J
jorlow@chromium.org 已提交
2691 2692 2693
  if (!s.ok()) {
    return s;
  }
K
kailiu 已提交
2694 2695
  if (manifest_filename.empty() ||
      manifest_filename.back() != '\n') {
J
jorlow@chromium.org 已提交
2696 2697
    return Status::Corruption("CURRENT file does not end with newline");
  }
K
kailiu 已提交
2698 2699
  // remove the trailing '\n'
  manifest_filename.resize(manifest_filename.size() - 1);
2700 2701 2702 2703 2704 2705
  FileType type;
  bool parse_ok =
      ParseFileName(manifest_filename, &manifest_file_number_, &type);
  if (!parse_ok || type != kDescriptorFile) {
    return Status::Corruption("CURRENT file corrupted");
  }
J
jorlow@chromium.org 已提交
2706

2707 2708
  ROCKS_LOG_INFO(db_options_->info_log, "Recovering from manifest file: %s\n",
                 manifest_filename.c_str());
H
heyongqiang 已提交
2709

K
kailiu 已提交
2710
  manifest_filename = dbname_ + "/" + manifest_filename;
2711 2712 2713 2714
  unique_ptr<SequentialFileReader> manifest_file_reader;
  {
    unique_ptr<SequentialFile> manifest_file;
    s = env_->NewSequentialFile(manifest_filename, &manifest_file,
2715
                                env_->OptimizeForManifestRead(env_options_));
2716 2717 2718 2719 2720
    if (!s.ok()) {
      return s;
    }
    manifest_file_reader.reset(
        new SequentialFileReader(std::move(manifest_file)));
J
jorlow@chromium.org 已提交
2721
  }
I
Igor Canadi 已提交
2722 2723
  uint64_t current_manifest_file_size;
  s = env_->GetFileSize(manifest_filename, &current_manifest_file_size);
2724 2725 2726
  if (!s.ok()) {
    return s;
  }
J
jorlow@chromium.org 已提交
2727 2728

  bool have_log_number = false;
2729
  bool have_prev_log_number = false;
J
jorlow@chromium.org 已提交
2730 2731 2732
  bool have_next_file = false;
  bool have_last_sequence = false;
  uint64_t next_file = 0;
2733 2734
  uint64_t last_sequence = 0;
  uint64_t log_number = 0;
I
Igor Canadi 已提交
2735
  uint64_t previous_log_number = 0;
2736
  uint32_t max_column_family = 0;
S
sdong 已提交
2737
  std::unordered_map<uint32_t, BaseReferencedVersionBuilder*> builders;
J
jorlow@chromium.org 已提交
2738

2739
  // add default column family
2740
  auto default_cf_iter = cf_name_to_options.find(kDefaultColumnFamilyName);
I
Igor Canadi 已提交
2741
  if (default_cf_iter == cf_name_to_options.end()) {
I
Igor Canadi 已提交
2742
    return Status::InvalidArgument("Default column family not specified");
I
Igor Canadi 已提交
2743
  }
I
Igor Canadi 已提交
2744
  VersionEdit default_cf_edit;
2745
  default_cf_edit.AddColumnFamily(kDefaultColumnFamilyName);
I
Igor Canadi 已提交
2746 2747 2748
  default_cf_edit.SetColumnFamily(0);
  ColumnFamilyData* default_cfd =
      CreateColumnFamily(default_cf_iter->second, &default_cf_edit);
2749 2750 2751
  // In recovery, nobody else can access it, so it's fine to set it to be
  // initialized earlier.
  default_cfd->set_initialized();
S
sdong 已提交
2752
  builders.insert({0, new BaseReferencedVersionBuilder(default_cfd)});
2753

J
jorlow@chromium.org 已提交
2754
  {
I
Igor Canadi 已提交
2755
    VersionSet::LogReporter reporter;
J
jorlow@chromium.org 已提交
2756
    reporter.status = &s;
2757 2758
    log::Reader reader(NULL, std::move(manifest_file_reader), &reporter,
                       true /*checksum*/, 0 /*initial_offset*/, 0);
J
jorlow@chromium.org 已提交
2759 2760 2761
    Slice record;
    std::string scratch;
    while (reader.ReadRecord(&record, &scratch) && s.ok()) {
2762
      VersionEdit edit;
J
jorlow@chromium.org 已提交
2763
      s = edit.DecodeFrom(record);
2764 2765
      if (!s.ok()) {
        break;
J
jorlow@chromium.org 已提交
2766 2767
      }

I
Igor Canadi 已提交
2768 2769 2770 2771 2772
      // Not found means that user didn't supply that column
      // family option AND we encountered column family add
      // record. Once we encounter column family drop record,
      // we will delete the column family from
      // column_families_not_found.
I
Igor Canadi 已提交
2773 2774 2775
      bool cf_in_not_found =
          column_families_not_found.find(edit.column_family_) !=
          column_families_not_found.end();
I
Igor Canadi 已提交
2776 2777
      // in builders means that user supplied that column family
      // option AND that we encountered column family add record
I
Igor Canadi 已提交
2778 2779 2780 2781 2782 2783
      bool cf_in_builders =
          builders.find(edit.column_family_) != builders.end();

      // they can't both be true
      assert(!(cf_in_not_found && cf_in_builders));

2784 2785
      ColumnFamilyData* cfd = nullptr;

2786
      if (edit.is_column_family_add_) {
I
Igor Canadi 已提交
2787 2788 2789 2790 2791 2792 2793
        if (cf_in_builders || cf_in_not_found) {
          s = Status::Corruption(
              "Manifest adding the same column family twice");
          break;
        }
        auto cf_options = cf_name_to_options.find(edit.column_family_name_);
        if (cf_options == cf_name_to_options.end()) {
I
Igor Canadi 已提交
2794 2795
          column_families_not_found.insert(
              {edit.column_family_, edit.column_family_name_});
I
Igor Canadi 已提交
2796
        } else {
2797
          cfd = CreateColumnFamily(cf_options->second, &edit);
2798
          cfd->set_initialized();
S
sdong 已提交
2799 2800
          builders.insert(
              {edit.column_family_, new BaseReferencedVersionBuilder(cfd)});
I
Igor Canadi 已提交
2801
        }
2802
      } else if (edit.is_column_family_drop_) {
I
Igor Canadi 已提交
2803 2804 2805 2806 2807
        if (cf_in_builders) {
          auto builder = builders.find(edit.column_family_);
          assert(builder != builders.end());
          delete builder->second;
          builders.erase(builder);
2808
          cfd = column_family_set_->GetColumnFamily(edit.column_family_);
2809 2810
          if (cfd->Unref()) {
            delete cfd;
2811
            cfd = nullptr;
2812 2813 2814 2815
          } else {
            // who else can have reference to cfd!?
            assert(false);
          }
I
Igor Canadi 已提交
2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829
        } else if (cf_in_not_found) {
          column_families_not_found.erase(edit.column_family_);
        } else {
          s = Status::Corruption(
              "Manifest - dropping non-existing column family");
          break;
        }
      } else if (!cf_in_not_found) {
        if (!cf_in_builders) {
          s = Status::Corruption(
              "Manifest record referencing unknown column family");
          break;
        }

2830
        cfd = column_family_set_->GetColumnFamily(edit.column_family_);
I
Igor Canadi 已提交
2831 2832
        // this should never happen since cf_in_builders is true
        assert(cfd != nullptr);
2833
        if (edit.max_level_ >= cfd->current()->storage_info()->num_levels()) {
2834 2835 2836 2837 2838
          s = Status::InvalidArgument(
              "db has more levels than options.num_levels");
          break;
        }

2839 2840 2841 2842 2843
        // if it is not column family add or column family drop,
        // then it's a file add/delete, which should be forwarded
        // to builder
        auto builder = builders.find(edit.column_family_);
        assert(builder != builders.end());
2844
        builder->second->version_builder()->Apply(&edit);
2845 2846 2847
      }

      if (cfd != nullptr) {
2848
        if (edit.has_log_number_) {
I
Igor Canadi 已提交
2849
          if (cfd->GetLogNumber() > edit.log_number_) {
2850 2851
            ROCKS_LOG_WARN(
                db_options_->info_log,
2852 2853
                "MANIFEST corruption detected, but ignored - Log numbers in "
                "records NOT monotonically increasing");
2854 2855 2856
          } else {
            cfd->SetLogNumber(edit.log_number_);
            have_log_number = true;
I
Igor Canadi 已提交
2857
          }
2858
        }
I
Igor Canadi 已提交
2859 2860 2861 2862 2863 2864 2865
        if (edit.has_comparator_ &&
            edit.comparator_ != cfd->user_comparator()->Name()) {
          s = Status::InvalidArgument(
              cfd->user_comparator()->Name(),
              "does not match existing comparator " + edit.comparator_);
          break;
        }
2866 2867
      }

2868
      if (edit.has_prev_log_number_) {
I
Igor Canadi 已提交
2869
        previous_log_number = edit.prev_log_number_;
2870 2871 2872
        have_prev_log_number = true;
      }

J
jorlow@chromium.org 已提交
2873 2874 2875 2876 2877
      if (edit.has_next_file_number_) {
        next_file = edit.next_file_number_;
        have_next_file = true;
      }

2878 2879 2880 2881
      if (edit.has_max_column_family_) {
        max_column_family = edit.max_column_family_;
      }

J
jorlow@chromium.org 已提交
2882
      if (edit.has_last_sequence_) {
2883
        last_sequence = edit.last_sequence_;
J
jorlow@chromium.org 已提交
2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896
        have_last_sequence = true;
      }
    }
  }

  if (s.ok()) {
    if (!have_next_file) {
      s = Status::Corruption("no meta-nextfile entry in descriptor");
    } else if (!have_log_number) {
      s = Status::Corruption("no meta-lognumber entry in descriptor");
    } else if (!have_last_sequence) {
      s = Status::Corruption("no last-sequence-number entry in descriptor");
    }
2897 2898

    if (!have_prev_log_number) {
I
Igor Canadi 已提交
2899
      previous_log_number = 0;
2900
    }
2901

2902 2903
    column_family_set_->UpdateMaxColumnFamily(max_column_family);

2904 2905
    MarkFileNumberUsedDuringRecovery(previous_log_number);
    MarkFileNumberUsedDuringRecovery(log_number);
J
jorlow@chromium.org 已提交
2906 2907
  }

I
Igor Canadi 已提交
2908
  // there were some column families in the MANIFEST that weren't specified
2909
  // in the argument. This is OK in read_only mode
2910
  if (read_only == false && !column_families_not_found.empty()) {
2911
    std::string list_of_not_found;
I
Igor Canadi 已提交
2912 2913
    for (const auto& cf : column_families_not_found) {
      list_of_not_found += ", " + cf.second;
2914 2915
    }
    list_of_not_found = list_of_not_found.substr(2);
I
Igor Canadi 已提交
2916
    s = Status::InvalidArgument(
I
Igor Canadi 已提交
2917 2918
        "You have to open all column families. Column families not opened: " +
        list_of_not_found);
I
Igor Canadi 已提交
2919 2920
  }

J
jorlow@chromium.org 已提交
2921
  if (s.ok()) {
I
Igor Canadi 已提交
2922
    for (auto cfd : *column_family_set_) {
2923 2924 2925
      if (cfd->IsDropped()) {
        continue;
      }
2926
      assert(cfd->initialized());
2927 2928
      auto builders_iter = builders.find(cfd->GetID());
      assert(builders_iter != builders.end());
2929
      auto* builder = builders_iter->second->version_builder();
2930

2931 2932
      if (GetColumnFamilySet()->get_table_cache()->GetCapacity() ==
          TableCache::kInfiniteCapacity) {
2933 2934
        // unlimited table cache. Pre-load table handle now.
        // Need to do it out of the mutex.
2935 2936 2937
        builder->LoadTableHandlers(
            cfd->internal_stats(), db_options_->max_file_opening_threads,
            false /* prefetch_index_and_filter_in_cache */);
2938
      }
2939

2940
      Version* v = new Version(cfd, this, current_version_number_++);
S
sdong 已提交
2941
      builder->SaveTo(v->storage_info());
2942

2943
      // Install recovered version
2944 2945
      v->PrepareApply(*cfd->GetLatestMutableCFOptions(),
          !(db_options_->skip_stats_update_on_db_open));
I
Igor Canadi 已提交
2946
      AppendVersion(cfd, v);
2947
    }
2948

I
Igor Canadi 已提交
2949
    manifest_file_size_ = current_manifest_file_size;
2950
    next_file_number_.store(next_file + 1);
2951
    last_to_be_written_sequence_ = last_sequence;
2952
    last_sequence_ = last_sequence;
I
Igor Canadi 已提交
2953
    prev_log_number_ = previous_log_number;
H
heyongqiang 已提交
2954

2955 2956
    ROCKS_LOG_INFO(
        db_options_->info_log,
2957
        "Recovered from manifest file:%s succeeded,"
K
Kai Liu 已提交
2958 2959
        "manifest_file_number is %lu, next_file_number is %lu, "
        "last_sequence is %lu, log_number is %lu,"
2960 2961
        "prev_log_number is %lu,"
        "max_column_family is %u\n",
2962
        manifest_filename.c_str(), (unsigned long)manifest_file_number_,
2963
        (unsigned long)next_file_number_.load(), (unsigned long)last_sequence_,
2964
        (unsigned long)log_number, (unsigned long)prev_log_number_,
2965
        column_family_set_->GetMaxColumnFamily());
2966 2967

    for (auto cfd : *column_family_set_) {
2968 2969 2970
      if (cfd->IsDropped()) {
        continue;
      }
2971 2972 2973
      ROCKS_LOG_INFO(db_options_->info_log,
                     "Column family [%s] (ID %u), log number is %" PRIu64 "\n",
                     cfd->GetName().c_str(), cfd->GetID(), cfd->GetLogNumber());
2974
    }
J
jorlow@chromium.org 已提交
2975 2976
  }

Y
Yi Wu 已提交
2977
  for (auto& builder : builders) {
2978 2979 2980
    delete builder.second;
  }

J
jorlow@chromium.org 已提交
2981 2982 2983
  return s;
}

I
Igor Canadi 已提交
2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000
Status VersionSet::ListColumnFamilies(std::vector<std::string>* column_families,
                                      const std::string& dbname, Env* env) {
  // these are just for performance reasons, not correcntes,
  // so we're fine using the defaults
  EnvOptions soptions;
  // Read "CURRENT" file, which contains a pointer to the current manifest file
  std::string current;
  Status s = ReadFileToString(env, CurrentFileName(dbname), &current);
  if (!s.ok()) {
    return s;
  }
  if (current.empty() || current[current.size()-1] != '\n') {
    return Status::Corruption("CURRENT file does not end with newline");
  }
  current.resize(current.size() - 1);

  std::string dscname = dbname + "/" + current;
3001 3002 3003

  unique_ptr<SequentialFileReader> file_reader;
  {
I
Igor Canadi 已提交
3004 3005 3006 3007 3008
  unique_ptr<SequentialFile> file;
  s = env->NewSequentialFile(dscname, &file, soptions);
  if (!s.ok()) {
    return s;
  }
3009 3010
  file_reader.reset(new SequentialFileReader(std::move(file)));
  }
I
Igor Canadi 已提交
3011 3012 3013

  std::map<uint32_t, std::string> column_family_names;
  // default column family is always implicitly there
3014
  column_family_names.insert({0, kDefaultColumnFamilyName});
I
Igor Canadi 已提交
3015 3016
  VersionSet::LogReporter reporter;
  reporter.status = &s;
3017 3018
  log::Reader reader(NULL, std::move(file_reader), &reporter, true /*checksum*/,
                     0 /*initial_offset*/, 0);
I
Igor Canadi 已提交
3019 3020 3021
  Slice record;
  std::string scratch;
  while (reader.ReadRecord(&record, &scratch) && s.ok()) {
3022 3023 3024 3025 3026 3027
    VersionEdit edit;
    s = edit.DecodeFrom(record);
    if (!s.ok()) {
      break;
    }
    if (edit.is_column_family_add_) {
3028 3029 3030 3031 3032
      if (column_family_names.find(edit.column_family_) !=
          column_family_names.end()) {
        s = Status::Corruption("Manifest adding the same column family twice");
        break;
      }
3033 3034 3035
      column_family_names.insert(
          {edit.column_family_, edit.column_family_name_});
    } else if (edit.is_column_family_drop_) {
3036 3037 3038 3039 3040 3041
      if (column_family_names.find(edit.column_family_) ==
          column_family_names.end()) {
        s = Status::Corruption(
            "Manifest - dropping non-existing column family");
        break;
      }
3042 3043
      column_family_names.erase(edit.column_family_);
    }
I
Igor Canadi 已提交
3044 3045 3046 3047 3048 3049
  }

  column_families->clear();
  if (s.ok()) {
    for (const auto& iter : column_family_names) {
      column_families->push_back(iter.second);
3050
    }
I
Igor Canadi 已提交
3051 3052 3053 3054
  }

  return s;
}
3055

I
Igor Canadi 已提交
3056
#ifndef ROCKSDB_LITE
3057 3058
Status VersionSet::ReduceNumberOfLevels(const std::string& dbname,
                                        const Options* options,
3059
                                        const EnvOptions& env_options,
3060 3061 3062 3063 3064 3065
                                        int new_levels) {
  if (new_levels <= 1) {
    return Status::InvalidArgument(
        "Number of levels needs to be bigger than 1");
  }

3066
  ImmutableDBOptions db_options(*options);
I
Igor Canadi 已提交
3067
  ColumnFamilyOptions cf_options(*options);
3068 3069
  std::shared_ptr<Cache> tc(NewLRUCache(options->max_open_files - 10,
                                        options->table_cache_numshardbits));
S
sdong 已提交
3070
  WriteController wc(options->delayed_write_rate);
3071
  WriteBufferManager wb(options->db_write_buffer_size);
3072
  VersionSet versions(dbname, &db_options, env_options, tc.get(), &wb, &wc);
3073 3074
  Status status;

3075
  std::vector<ColumnFamilyDescriptor> dummy;
3076
  ColumnFamilyDescriptor dummy_descriptor(kDefaultColumnFamilyName,
I
Igor Canadi 已提交
3077
                                          ColumnFamilyOptions(*options));
I
Igor Canadi 已提交
3078
  dummy.push_back(dummy_descriptor);
3079
  status = versions.Recover(dummy);
3080 3081 3082 3083
  if (!status.ok()) {
    return status;
  }

3084
  Version* current_version =
3085
      versions.GetColumnFamilySet()->GetDefault()->current();
S
sdong 已提交
3086
  auto* vstorage = current_version->storage_info();
3087
  int current_levels = vstorage->num_levels();
3088 3089 3090 3091 3092 3093 3094 3095 3096 3097

  if (current_levels <= new_levels) {
    return Status::OK();
  }

  // Make sure there are file only on one level from
  // (new_levels-1) to (current_levels-1)
  int first_nonempty_level = -1;
  int first_nonempty_level_filenum = 0;
  for (int i = new_levels - 1; i < current_levels; i++) {
S
sdong 已提交
3098
    int file_num = vstorage->NumLevelFiles(i);
3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114
    if (file_num != 0) {
      if (first_nonempty_level < 0) {
        first_nonempty_level = i;
        first_nonempty_level_filenum = file_num;
      } else {
        char msg[255];
        snprintf(msg, sizeof(msg),
                 "Found at least two levels containing files: "
                 "[%d:%d],[%d:%d].\n",
                 first_nonempty_level, first_nonempty_level_filenum, i,
                 file_num);
        return Status::InvalidArgument(msg);
      }
    }
  }

I
Igor Canadi 已提交
3115 3116 3117
  // we need to allocate an array with the old number of levels size to
  // avoid SIGSEGV in WriteSnapshot()
  // however, all levels bigger or equal to new_levels will be empty
3118
  std::vector<FileMetaData*>* new_files_list =
I
Igor Canadi 已提交
3119
      new std::vector<FileMetaData*>[current_levels];
3120
  for (int i = 0; i < new_levels - 1; i++) {
S
sdong 已提交
3121
    new_files_list[i] = vstorage->LevelFiles(i);
3122 3123 3124
  }

  if (first_nonempty_level > 0) {
S
sdong 已提交
3125
    new_files_list[new_levels - 1] = vstorage->LevelFiles(first_nonempty_level);
3126 3127
  }

S
sdong 已提交
3128 3129 3130
  delete[] vstorage -> files_;
  vstorage->files_ = new_files_list;
  vstorage->num_levels_ = new_levels;
3131

Y
Yi Wu 已提交
3132
  MutableCFOptions mutable_cf_options(*options);
3133
  VersionEdit ve;
3134 3135
  InstrumentedMutex dummy_mutex;
  InstrumentedMutexLock l(&dummy_mutex);
3136 3137 3138
  return versions.LogAndApply(
      versions.GetColumnFamilySet()->GetDefault(),
      mutable_cf_options, &ve, &dummy_mutex, nullptr, true);
3139 3140
}

I
Igor Canadi 已提交
3141
Status VersionSet::DumpManifest(Options& options, std::string& dscname,
3142
                                bool verbose, bool hex, bool json) {
3143
  // Open the specified manifest file.
3144 3145 3146 3147
  unique_ptr<SequentialFileReader> file_reader;
  Status s;
  {
    unique_ptr<SequentialFile> file;
3148 3149
    s = options.env->NewSequentialFile(
        dscname, &file, env_->OptimizeForManifestRead(env_options_));
3150 3151 3152 3153
    if (!s.ok()) {
      return s;
    }
    file_reader.reset(new SequentialFileReader(std::move(file)));
3154 3155 3156 3157 3158 3159 3160
  }

  bool have_prev_log_number = false;
  bool have_next_file = false;
  bool have_last_sequence = false;
  uint64_t next_file = 0;
  uint64_t last_sequence = 0;
I
Igor Canadi 已提交
3161
  uint64_t previous_log_number = 0;
3162
  int count = 0;
3163
  std::unordered_map<uint32_t, std::string> comparators;
S
sdong 已提交
3164
  std::unordered_map<uint32_t, BaseReferencedVersionBuilder*> builders;
3165 3166 3167

  // add default column family
  VersionEdit default_cf_edit;
3168
  default_cf_edit.AddColumnFamily(kDefaultColumnFamilyName);
3169 3170 3171
  default_cf_edit.SetColumnFamily(0);
  ColumnFamilyData* default_cfd =
      CreateColumnFamily(ColumnFamilyOptions(options), &default_cf_edit);
S
sdong 已提交
3172
  builders.insert({0, new BaseReferencedVersionBuilder(default_cfd)});
3173 3174

  {
I
Igor Canadi 已提交
3175
    VersionSet::LogReporter reporter;
3176
    reporter.status = &s;
3177 3178
    log::Reader reader(NULL, std::move(file_reader), &reporter,
                       true /*checksum*/, 0 /*initial_offset*/, 0);
3179 3180 3181
    Slice record;
    std::string scratch;
    while (reader.ReadRecord(&record, &scratch) && s.ok()) {
3182
      VersionEdit edit;
3183
      s = edit.DecodeFrom(record);
3184 3185
      if (!s.ok()) {
        break;
3186 3187
      }

3188
      // Write out each individual edit
3189 3190 3191 3192
      if (verbose && !json) {
        printf("%s\n", edit.DebugString(hex).c_str());
      } else if (json) {
        printf("%s\n", edit.DebugJSON(count, hex).c_str());
3193 3194 3195
      }
      count++;

3196 3197 3198 3199 3200
      bool cf_in_builders =
          builders.find(edit.column_family_) != builders.end();

      if (edit.has_comparator_) {
        comparators.insert({edit.column_family_, edit.comparator_});
3201 3202
      }

3203 3204
      ColumnFamilyData* cfd = nullptr;

3205 3206 3207 3208 3209 3210
      if (edit.is_column_family_add_) {
        if (cf_in_builders) {
          s = Status::Corruption(
              "Manifest adding the same column family twice");
          break;
        }
3211
        cfd = CreateColumnFamily(ColumnFamilyOptions(options), &edit);
3212
        cfd->set_initialized();
S
sdong 已提交
3213 3214
        builders.insert(
            {edit.column_family_, new BaseReferencedVersionBuilder(cfd)});
3215 3216 3217 3218 3219 3220 3221 3222 3223 3224
      } else if (edit.is_column_family_drop_) {
        if (!cf_in_builders) {
          s = Status::Corruption(
              "Manifest - dropping non-existing column family");
          break;
        }
        auto builder_iter = builders.find(edit.column_family_);
        delete builder_iter->second;
        builders.erase(builder_iter);
        comparators.erase(edit.column_family_);
3225
        cfd = column_family_set_->GetColumnFamily(edit.column_family_);
3226 3227 3228
        assert(cfd != nullptr);
        cfd->Unref();
        delete cfd;
3229
        cfd = nullptr;
3230 3231 3232 3233 3234 3235 3236
      } else {
        if (!cf_in_builders) {
          s = Status::Corruption(
              "Manifest record referencing unknown column family");
          break;
        }

3237
        cfd = column_family_set_->GetColumnFamily(edit.column_family_);
3238 3239 3240 3241 3242 3243 3244 3245
        // this should never happen since cf_in_builders is true
        assert(cfd != nullptr);

        // if it is not column family add or column family drop,
        // then it's a file add/delete, which should be forwarded
        // to builder
        auto builder = builders.find(edit.column_family_);
        assert(builder != builders.end());
3246
        builder->second->version_builder()->Apply(&edit);
3247 3248
      }

3249 3250 3251 3252
      if (cfd != nullptr && edit.has_log_number_) {
        cfd->SetLogNumber(edit.log_number_);
      }

3253
      if (edit.has_prev_log_number_) {
I
Igor Canadi 已提交
3254
        previous_log_number = edit.prev_log_number_;
3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266
        have_prev_log_number = true;
      }

      if (edit.has_next_file_number_) {
        next_file = edit.next_file_number_;
        have_next_file = true;
      }

      if (edit.has_last_sequence_) {
        last_sequence = edit.last_sequence_;
        have_last_sequence = true;
      }
3267 3268 3269 3270

      if (edit.has_max_column_family_) {
        column_family_set_->UpdateMaxColumnFamily(edit.max_column_family_);
      }
3271 3272
    }
  }
3273
  file_reader.reset();
3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284

  if (s.ok()) {
    if (!have_next_file) {
      s = Status::Corruption("no meta-nextfile entry in descriptor");
      printf("no meta-nextfile entry in descriptor");
    } else if (!have_last_sequence) {
      printf("no last-sequence-number entry in descriptor");
      s = Status::Corruption("no last-sequence-number entry in descriptor");
    }

    if (!have_prev_log_number) {
I
Igor Canadi 已提交
3285
      previous_log_number = 0;
3286 3287 3288 3289
    }
  }

  if (s.ok()) {
3290
    for (auto cfd : *column_family_set_) {
3291 3292 3293
      if (cfd->IsDropped()) {
        continue;
      }
3294 3295
      auto builders_iter = builders.find(cfd->GetID());
      assert(builders_iter != builders.end());
3296
      auto builder = builders_iter->second->version_builder();
3297 3298

      Version* v = new Version(cfd, this, current_version_number_++);
S
sdong 已提交
3299
      builder->SaveTo(v->storage_info());
3300
      v->PrepareApply(*cfd->GetLatestMutableCFOptions(), false);
3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313

      printf("--------------- Column family \"%s\"  (ID %u) --------------\n",
             cfd->GetName().c_str(), (unsigned int)cfd->GetID());
      printf("log number: %lu\n", (unsigned long)cfd->GetLogNumber());
      auto comparator = comparators.find(cfd->GetID());
      if (comparator != comparators.end()) {
        printf("comparator: %s\n", comparator->second.c_str());
      } else {
        printf("comparator: <NO COMPARATOR>\n");
      }
      printf("%s \n", v->DebugString(hex).c_str());
      delete v;
    }
3314

3315 3316 3317 3318 3319
    // Free builders
    for (auto& builder : builders) {
      delete builder.second;
    }

3320
    next_file_number_.store(next_file + 1);
3321
    last_to_be_written_sequence_ = last_sequence;
3322
    last_sequence_ = last_sequence;
I
Igor Canadi 已提交
3323
    prev_log_number_ = previous_log_number;
3324

3325
    printf(
3326
        "next_file_number %lu last_sequence "
3327
        "%lu  prev_log_number %lu max_column_family %u\n",
3328
        (unsigned long)next_file_number_.load(), (unsigned long)last_sequence,
I
Igor Canadi 已提交
3329
        (unsigned long)previous_log_number,
3330
        column_family_set_->GetMaxColumnFamily());
3331
  }
3332

3333 3334
  return s;
}
I
Igor Canadi 已提交
3335
#endif  // ROCKSDB_LITE
3336

3337 3338 3339 3340 3341
void VersionSet::MarkFileNumberUsedDuringRecovery(uint64_t number) {
  // only called during recovery which is single threaded, so this works because
  // there can't be concurrent calls
  if (next_file_number_.load(std::memory_order_relaxed) <= number) {
    next_file_number_.store(number + 1, std::memory_order_relaxed);
3342 3343 3344
  }
}

J
jorlow@chromium.org 已提交
3345 3346
Status VersionSet::WriteSnapshot(log::Writer* log) {
  // TODO: Break up into multiple records to reduce memory usage on recovery?
3347

I
Igor Canadi 已提交
3348 3349
  // WARNING: This method doesn't hold a mutex!!

I
Igor Canadi 已提交
3350 3351
  // This is done without DB mutex lock held, but only within single-threaded
  // LogAndApply. Column family manipulations can only happen within LogAndApply
I
Igor Canadi 已提交
3352
  // (the same single thread), so we're safe to iterate.
I
Igor Canadi 已提交
3353
  for (auto cfd : *column_family_set_) {
3354 3355 3356
    if (cfd->IsDropped()) {
      continue;
    }
3357
    assert(cfd->initialized());
3358 3359 3360
    {
      // Store column family info
      VersionEdit edit;
3361
      if (cfd->GetID() != 0) {
3362 3363
        // default column family is always there,
        // no need to explicitly write it
3364 3365
        edit.AddColumnFamily(cfd->GetName());
        edit.SetColumnFamily(cfd->GetID());
I
Igor Canadi 已提交
3366 3367 3368 3369
      }
      edit.SetComparatorName(
          cfd->internal_comparator().user_comparator()->Name());
      std::string record;
3370 3371 3372 3373
      if (!edit.EncodeTo(&record)) {
        return Status::Corruption(
            "Unable to Encode VersionEdit:" + edit.DebugString(true));
      }
I
Igor Canadi 已提交
3374 3375 3376
      Status s = log->AddRecord(record);
      if (!s.ok()) {
        return s;
3377
      }
3378
    }
3379

3380 3381 3382
    {
      // Save files
      VersionEdit edit;
3383
      edit.SetColumnFamily(cfd->GetID());
3384

I
Igor Canadi 已提交
3385
      for (int level = 0; level < cfd->NumberLevels(); level++) {
S
sdong 已提交
3386 3387
        for (const auto& f :
             cfd->current()->storage_info()->LevelFiles(level)) {
3388 3389
          edit.AddFile(level, f->fd.GetNumber(), f->fd.GetPathId(),
                       f->fd.GetFileSize(), f->smallest, f->largest,
3390 3391
                       f->smallest_seqno, f->largest_seqno,
                       f->marked_for_compaction);
3392 3393
        }
      }
3394
      edit.SetLogNumber(cfd->GetLogNumber());
3395
      std::string record;
3396 3397 3398 3399
      if (!edit.EncodeTo(&record)) {
        return Status::Corruption(
            "Unable to Encode VersionEdit:" + edit.DebugString(true));
      }
3400 3401 3402
      Status s = log->AddRecord(record);
      if (!s.ok()) {
        return s;
3403
      }
3404 3405 3406
    }
  }

I
Igor Canadi 已提交
3407
  return Status::OK();
J
jorlow@chromium.org 已提交
3408 3409
}

3410 3411 3412 3413 3414 3415
// TODO(aekmekji): in CompactionJob::GenSubcompactionBoundaries(), this
// function is called repeatedly with consecutive pairs of slices. For example
// if the slice list is [a, b, c, d] this function is called with arguments
// (a,b) then (b,c) then (c,d). Knowing this, an optimization is possible where
// we avoid doing binary search for the keys b and c twice and instead somehow
// maintain state of where they first appear in the files.
3416
uint64_t VersionSet::ApproximateSize(Version* v, const Slice& start,
3417 3418
                                     const Slice& end, int start_level,
                                     int end_level) {
3419 3420
  // pre-condition
  assert(v->cfd_->internal_comparator().Compare(start, end) <= 0);
3421

3422
  uint64_t size = 0;
S
sdong 已提交
3423
  const auto* vstorage = v->storage_info();
3424 3425 3426
  end_level = end_level == -1
                  ? vstorage->num_non_empty_levels()
                  : std::min(end_level, vstorage->num_non_empty_levels());
3427

3428 3429 3430
  assert(start_level <= end_level);

  for (int level = start_level; level < end_level; level++) {
3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446
    const LevelFilesBrief& files_brief = vstorage->LevelFilesBrief(level);
    if (!files_brief.num_files) {
      // empty level, skip exploration
      continue;
    }

    if (!level) {
      // level 0 data is sorted order, handle the use case explicitly
      size += ApproximateSizeLevel0(v, files_brief, start, end);
      continue;
    }

    assert(level > 0);
    assert(files_brief.num_files > 0);

    // identify the file position for starting key
I
Igor Canadi 已提交
3447 3448 3449
    const uint64_t idx_start = FindFileInRange(
        v->cfd_->internal_comparator(), files_brief, start,
        /*start=*/0, static_cast<uint32_t>(files_brief.num_files - 1));
3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469
    assert(idx_start < files_brief.num_files);

    // scan all files from the starting position until the ending position
    // inferred from the sorted order
    for (uint64_t i = idx_start; i < files_brief.num_files; i++) {
      uint64_t val;
      val = ApproximateSize(v, files_brief.files[i], end);
      if (!val) {
        // the files after this will not have the range
        break;
      }

      size += val;

      if (i == idx_start) {
        // subtract the bytes needed to be scanned to get to the starting
        // key
        val = ApproximateSize(v, files_brief.files[i], start);
        assert(size >= val);
        size -= val;
J
jorlow@chromium.org 已提交
3470 3471 3472
      }
    }
  }
3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508

  return size;
}

uint64_t VersionSet::ApproximateSizeLevel0(Version* v,
                                           const LevelFilesBrief& files_brief,
                                           const Slice& key_start,
                                           const Slice& key_end) {
  // level 0 files are not in sorted order, we need to iterate through
  // the list to compute the total bytes that require scanning
  uint64_t size = 0;
  for (size_t i = 0; i < files_brief.num_files; i++) {
    const uint64_t start = ApproximateSize(v, files_brief.files[i], key_start);
    const uint64_t end = ApproximateSize(v, files_brief.files[i], key_end);
    assert(end >= start);
    size += end - start;
  }
  return size;
}

uint64_t VersionSet::ApproximateSize(Version* v, const FdWithKeyRange& f,
                                     const Slice& key) {
  // pre-condition
  assert(v);

  uint64_t result = 0;
  if (v->cfd_->internal_comparator().Compare(f.largest_key, key) <= 0) {
    // Entire file is before "key", so just add the file size
    result = f.fd.GetFileSize();
  } else if (v->cfd_->internal_comparator().Compare(f.smallest_key, key) > 0) {
    // Entire file is after "key", so ignore
    result = 0;
  } else {
    // "key" falls in the range for this table.  Add the
    // approximate offset of "key" within the table.
    TableReader* table_reader_ptr;
S
sdong 已提交
3509
    InternalIterator* iter = v->cfd_->table_cache()->NewIterator(
3510
        ReadOptions(), env_options_, v->cfd_->internal_comparator(), f.fd,
3511
        nullptr /* range_del_agg */, &table_reader_ptr);
3512 3513 3514 3515 3516
    if (table_reader_ptr != nullptr) {
      result = table_reader_ptr->ApproximateOffsetOf(key);
    }
    delete iter;
  }
J
jorlow@chromium.org 已提交
3517 3518 3519
  return result;
}

3520
void VersionSet::AddLiveFiles(std::vector<FileDescriptor>* live_list) {
3521 3522
  // pre-calculate space requirement
  int64_t total_files = 0;
I
Igor Canadi 已提交
3523
  for (auto cfd : *column_family_set_) {
3524 3525 3526
    if (!cfd->initialized()) {
      continue;
    }
3527 3528
    Version* dummy_versions = cfd->dummy_versions();
    for (Version* v = dummy_versions->next_; v != dummy_versions;
I
Igor Canadi 已提交
3529
         v = v->next_) {
S
sdong 已提交
3530
      const auto* vstorage = v->storage_info();
3531
      for (int level = 0; level < vstorage->num_levels(); level++) {
S
sdong 已提交
3532
        total_files += vstorage->LevelFiles(level).size();
3533
      }
3534 3535 3536 3537
    }
  }

  // just one time extension to the right size
3538
  live_list->reserve(live_list->size() + static_cast<size_t>(total_files));
3539

I
Igor Canadi 已提交
3540
  for (auto cfd : *column_family_set_) {
3541 3542 3543
    if (!cfd->initialized()) {
      continue;
    }
3544 3545
    auto* current = cfd->current();
    bool found_current = false;
3546 3547
    Version* dummy_versions = cfd->dummy_versions();
    for (Version* v = dummy_versions->next_; v != dummy_versions;
I
Igor Canadi 已提交
3548
         v = v->next_) {
3549 3550 3551
      v->AddLiveFiles(live_list);
      if (v == current) {
        found_current = true;
J
jorlow@chromium.org 已提交
3552 3553
      }
    }
3554 3555 3556 3557 3558
    if (!found_current && current != nullptr) {
      // Should never happen unless it is a bug.
      assert(false);
      current->AddLiveFiles(live_list);
    }
J
jorlow@chromium.org 已提交
3559 3560 3561
  }
}

3562 3563
InternalIterator* VersionSet::MakeInputIterator(
    const Compaction* c, RangeDelAggregator* range_del_agg) {
L
Lei Jin 已提交
3564 3565
  auto cfd = c->column_family_data();
  ReadOptions read_options;
3566
  read_options.verify_checksums = true;
L
Lei Jin 已提交
3567
  read_options.fill_cache = false;
3568 3569 3570 3571 3572
  // Compaction iterators shouldn't be confined to a single prefix.
  // Compactions use Seek() for
  // (a) concurrent compactions,
  // (b) CompactionFilter::Decision::kRemoveAndSkipUntil.
  read_options.total_order_seek = true;
J
jorlow@chromium.org 已提交
3573 3574 3575 3576

  // Level-0 files have to be merged together.  For other levels,
  // we will make a concatenating iterator per level.
  // TODO(opt): use concatenating iterator for level-0 if there is no overlap
3577 3578 3579
  const size_t space = (c->level() == 0 ? c->input_levels(0)->num_files +
                                              c->num_input_levels() - 1
                                        : c->num_input_levels());
S
sdong 已提交
3580
  InternalIterator** list = new InternalIterator* [space];
3581 3582
  size_t num = 0;
  for (size_t which = 0; which < c->num_input_levels(); which++) {
F
Feng Zhu 已提交
3583
    if (c->input_levels(which)->num_files != 0) {
3584
      if (c->level(which) == 0) {
3585
        const LevelFilesBrief* flevel = c->input_levels(which);
F
Feng Zhu 已提交
3586
        for (size_t i = 0; i < flevel->num_files; i++) {
L
Lei Jin 已提交
3587
          list[num++] = cfd->table_cache()->NewIterator(
3588
              read_options, env_options_compactions_,
3589 3590 3591
              cfd->internal_comparator(), flevel->files[i].fd, range_del_agg,
              nullptr /* table_reader_ptr */,
              nullptr /* no per level latency histogram */,
3592
              true /* for_compaction */, nullptr /* arena */,
3593
              false /* skip_filters */, (int)which /* level */);
J
jorlow@chromium.org 已提交
3594 3595 3596
        }
      } else {
        // Create concatenating iterator for the files from this level
3597 3598
        list[num++] = NewTwoLevelIterator(
            new LevelFileIteratorState(
3599
                cfd->table_cache(), read_options, env_options_compactions_,
3600 3601
                cfd->internal_comparator(),
                nullptr /* no per level latency histogram */,
3602
                true /* for_compaction */, false /* prefix enabled */,
3603 3604
                false /* skip_filters */, (int)which /* level */,
                range_del_agg),
3605
            new LevelFileNumIterator(cfd->internal_comparator(),
3606 3607
                                     c->input_levels(which),
                                     false /* don't sample compaction */));
J
jorlow@chromium.org 已提交
3608 3609 3610 3611
      }
    }
  }
  assert(num <= space);
S
sdong 已提交
3612
  InternalIterator* result =
3613 3614
      NewMergingIterator(&c->column_family_data()->internal_comparator(), list,
                         static_cast<int>(num));
J
jorlow@chromium.org 已提交
3615 3616 3617 3618
  delete[] list;
  return result;
}

A
Abhishek Kona 已提交
3619
// verify that the files listed in this compaction are present
3620 3621
// in the current version
bool VersionSet::VerifyCompactionFileConsistency(Compaction* c) {
3622
#ifndef NDEBUG
I
Igor Canadi 已提交
3623
  Version* version = c->column_family_data()->current();
S
sdong 已提交
3624
  const VersionStorageInfo* vstorage = version->storage_info();
3625
  if (c->input_version() != version) {
3626 3627
    ROCKS_LOG_INFO(
        db_options_->info_log,
3628 3629
        "[%s] compaction output being applied to a different base version from"
        " input version",
I
Igor Canadi 已提交
3630
        c->column_family_data()->GetName().c_str());
S
sdong 已提交
3631 3632 3633

    if (vstorage->compaction_style_ == kCompactionStyleLevel &&
        c->start_level() == 0 && c->num_input_levels() > 2U) {
3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644
      // We are doing a L0->base_level compaction. The assumption is if
      // base level is not L1, levels from L1 to base_level - 1 is empty.
      // This is ensured by having one compaction from L0 going on at the
      // same time in level-based compaction. So that during the time, no
      // compaction/flush can put files to those levels.
      for (int l = c->start_level() + 1; l < c->output_level(); l++) {
        if (vstorage->NumLevelFiles(l) != 0) {
          return false;
        }
      }
    }
3645 3646
  }

3647
  for (size_t input = 0; input < c->num_input_levels(); ++input) {
3648
    int level = c->level(input);
3649
    for (size_t i = 0; i < c->num_input_files(input); ++i) {
3650 3651
      uint64_t number = c->input(input, i)->fd.GetNumber();
      bool found = false;
3652
      for (size_t j = 0; j < vstorage->files_[level].size(); j++) {
3653 3654 3655 3656 3657
        FileMetaData* f = vstorage->files_[level][j];
        if (f->fd.GetNumber() == number) {
          found = true;
          break;
        }
3658
      }
3659 3660
      if (!found) {
        return false;  // input files non existent in current version
3661 3662 3663
      }
    }
  }
3664
#endif
3665 3666 3667
  return true;     // everything good
}

3668
Status VersionSet::GetMetadataForFile(uint64_t number, int* filelevel,
3669
                                      FileMetaData** meta,
3670 3671
                                      ColumnFamilyData** cfd) {
  for (auto cfd_iter : *column_family_set_) {
3672 3673 3674
    if (!cfd_iter->initialized()) {
      continue;
    }
3675
    Version* version = cfd_iter->current();
S
sdong 已提交
3676
    const auto* vstorage = version->storage_info();
3677
    for (int level = 0; level < vstorage->num_levels(); level++) {
S
sdong 已提交
3678
      for (const auto& file : vstorage->LevelFiles(level)) {
3679
        if (file->fd.GetNumber() == number) {
3680
          *meta = file;
3681
          *filelevel = level;
3682
          *cfd = cfd_iter;
3683 3684
          return Status::OK();
        }
3685 3686 3687 3688 3689 3690
      }
    }
  }
  return Status::NotFound("File not present in any level");
}

3691
void VersionSet::GetLiveFilesMetaData(std::vector<LiveFileMetaData>* metadata) {
I
Igor Canadi 已提交
3692
  for (auto cfd : *column_family_set_) {
3693
    if (cfd->IsDropped() || !cfd->initialized()) {
3694 3695
      continue;
    }
I
Igor Canadi 已提交
3696
    for (int level = 0; level < cfd->NumberLevels(); level++) {
S
sdong 已提交
3697 3698
      for (const auto& file :
           cfd->current()->storage_info()->LevelFiles(level)) {
3699
        LiveFileMetaData filemetadata;
3700
        filemetadata.column_family_name = cfd->GetName();
3701
        uint32_t path_id = file->fd.GetPathId();
3702 3703
        if (path_id < db_options_->db_paths.size()) {
          filemetadata.db_path = db_options_->db_paths[path_id].path;
3704
        } else {
3705 3706
          assert(!db_options_->db_paths.empty());
          filemetadata.db_path = db_options_->db_paths.back().path;
3707 3708
        }
        filemetadata.name = MakeTableFileName("", file->fd.GetNumber());
3709
        filemetadata.level = level;
3710
        filemetadata.size = file->fd.GetFileSize();
I
Igor Canadi 已提交
3711 3712 3713 3714
        filemetadata.smallestkey = file->smallest.user_key().ToString();
        filemetadata.largestkey = file->largest.user_key().ToString();
        filemetadata.smallest_seqno = file->smallest_seqno;
        filemetadata.largest_seqno = file->largest_seqno;
3715 3716
        metadata->push_back(filemetadata);
      }
3717 3718 3719 3720
    }
  }
}

I
Igor Canadi 已提交
3721
void VersionSet::GetObsoleteFiles(std::vector<FileMetaData*>* files,
3722
                                  std::vector<std::string>* manifest_filenames,
I
Igor Canadi 已提交
3723
                                  uint64_t min_pending_output) {
3724 3725
  assert(manifest_filenames->empty());
  obsolete_manifests_.swap(*manifest_filenames);
I
Igor Canadi 已提交
3726 3727 3728 3729 3730 3731 3732 3733 3734
  std::vector<FileMetaData*> pending_files;
  for (auto f : obsolete_files_) {
    if (f->fd.GetNumber() < min_pending_output) {
      files->push_back(f);
    } else {
      pending_files.push_back(f);
    }
  }
  obsolete_files_.swap(pending_files);
I
Igor Canadi 已提交
3735 3736
}

3737
ColumnFamilyData* VersionSet::CreateColumnFamily(
3738
    const ColumnFamilyOptions& cf_options, VersionEdit* edit) {
3739 3740
  assert(edit->is_column_family_add_);

3741
  Version* dummy_versions = new Version(nullptr, this);
3742 3743 3744
  // Ref() dummy version once so that later we can call Unref() to delete it
  // by avoiding calling "delete" explicitly (~Version is private)
  dummy_versions->Ref();
I
Igor Canadi 已提交
3745
  auto new_cfd = column_family_set_->CreateColumnFamily(
3746 3747
      edit->column_family_name_, edit->column_family_, dummy_versions,
      cf_options);
I
Igor Canadi 已提交
3748

3749 3750
  Version* v = new Version(new_cfd, this, current_version_number_++);

3751 3752 3753
  // Fill level target base information.
  v->storage_info()->CalculateBaseBytes(*new_cfd->ioptions(),
                                        *new_cfd->GetLatestMutableCFOptions());
3754
  AppendVersion(new_cfd, v);
3755 3756
  // GetLatestMutableCFOptions() is safe here without mutex since the
  // cfd is not available to client
A
agiardullo 已提交
3757 3758
  new_cfd->CreateNewMemtable(*new_cfd->GetLatestMutableCFOptions(),
                             LastSequence());
I
Igor Canadi 已提交
3759
  new_cfd->SetLogNumber(edit->log_number_);
3760 3761 3762
  return new_cfd;
}

3763 3764 3765 3766 3767 3768 3769 3770
uint64_t VersionSet::GetNumLiveVersions(Version* dummy_versions) {
  uint64_t count = 0;
  for (Version* v = dummy_versions->next_; v != dummy_versions; v = v->next_) {
    count++;
  }
  return count;
}

3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788
uint64_t VersionSet::GetTotalSstFilesSize(Version* dummy_versions) {
  std::unordered_set<uint64_t> unique_files;
  uint64_t total_files_size = 0;
  for (Version* v = dummy_versions->next_; v != dummy_versions; v = v->next_) {
    VersionStorageInfo* storage_info = v->storage_info();
    for (int level = 0; level < storage_info->num_levels_; level++) {
      for (const auto& file_meta : storage_info->LevelFiles(level)) {
        if (unique_files.find(file_meta->fd.packed_number_and_path_id) ==
            unique_files.end()) {
          unique_files.insert(file_meta->fd.packed_number_and_path_id);
          total_files_size += file_meta->fd.GetFileSize();
        }
      }
    }
  }
  return total_files_size;
}

3789
}  // namespace rocksdb