Line data Source code
1 : // Copyright 2012 The LevelDB-Go and Pebble Authors. All rights reserved. Use
2 : // of this source code is governed by a BSD-style license that can be found in
3 : // the LICENSE file.
4 :
5 : package pebble
6 :
7 : import (
8 : "bytes"
9 : "context"
10 : "encoding/binary"
11 : "fmt"
12 : "io"
13 : "math"
14 : "os"
15 : "slices"
16 : "sync/atomic"
17 : "time"
18 :
19 : "github.com/cockroachdb/crlib/crtime"
20 : "github.com/cockroachdb/errors"
21 : "github.com/cockroachdb/errors/oserror"
22 : "github.com/cockroachdb/pebble/batchrepr"
23 : "github.com/cockroachdb/pebble/internal/arenaskl"
24 : "github.com/cockroachdb/pebble/internal/base"
25 : "github.com/cockroachdb/pebble/internal/cache"
26 : "github.com/cockroachdb/pebble/internal/constants"
27 : "github.com/cockroachdb/pebble/internal/invariants"
28 : "github.com/cockroachdb/pebble/internal/keyspan"
29 : "github.com/cockroachdb/pebble/internal/manifest"
30 : "github.com/cockroachdb/pebble/internal/manual"
31 : "github.com/cockroachdb/pebble/objstorage"
32 : "github.com/cockroachdb/pebble/objstorage/objstorageprovider"
33 : "github.com/cockroachdb/pebble/objstorage/remote"
34 : "github.com/cockroachdb/pebble/record"
35 : "github.com/cockroachdb/pebble/vfs"
36 : "github.com/cockroachdb/pebble/wal"
37 : "github.com/prometheus/client_golang/prometheus"
38 : )
39 :
40 : const (
41 : initialMemTableSize = 256 << 10 // 256 KB
42 :
43 : // The max batch size is limited by the uint32 offsets stored in
44 : // internal/batchskl.node, DeferredBatchOp, and flushableBatchEntry.
45 : //
46 : // We limit the size to MaxUint32 (just short of 4GB) so that the exclusive
47 : // end of an allocation fits in uint32.
48 : //
49 : // On 32-bit systems, slices are naturally limited to MaxInt (just short of
50 : // 2GB).
51 : maxBatchSize = constants.MaxUint32OrInt
52 :
53 : // The max memtable size is limited by the uint32 offsets stored in
54 : // internal/arenaskl.node, DeferredBatchOp, and flushableBatchEntry.
55 : //
56 : // We limit the size to MaxUint32 (just short of 4GB) so that the exclusive
57 : // end of an allocation fits in uint32.
58 : //
59 : // On 32-bit systems, slices are naturally limited to MaxInt (just short of
60 : // 2GB).
61 : maxMemTableSize = constants.MaxUint32OrInt
62 : )
63 :
64 : // FileCacheSize can be used to determine the file
65 : // cache size for a single db, given the maximum open
66 : // files which can be used by a file cache which is
67 : // only used by a single db.
68 1 : func FileCacheSize(maxOpenFiles int) int {
69 1 : fileCacheSize := maxOpenFiles - numNonFileCacheFiles
70 1 : if fileCacheSize < minFileCacheSize {
71 1 : fileCacheSize = minFileCacheSize
72 1 : }
73 1 : return fileCacheSize
74 : }
75 :
76 : // Open opens a DB whose files live in the given directory.
77 : //
78 : // IsCorruptionError() can be use to determine if the error is caused by on-disk
79 : // corruption.
80 1 : func Open(dirname string, opts *Options) (db *DB, err error) {
81 1 : // Make a copy of the options so that we don't mutate the passed in options.
82 1 : opts = opts.Clone()
83 1 : opts.EnsureDefaults()
84 1 : if opts.Experimental.CompactionScheduler == nil {
85 0 : opts.Experimental.CompactionScheduler = newConcurrencyLimitScheduler(defaultTimeSource{})
86 0 : }
87 1 : if err := opts.Validate(); err != nil {
88 0 : return nil, err
89 0 : }
90 1 : if opts.LoggerAndTracer == nil {
91 1 : opts.LoggerAndTracer = &base.LoggerWithNoopTracer{Logger: opts.Logger}
92 1 : } else {
93 0 : opts.Logger = opts.LoggerAndTracer
94 0 : }
95 :
96 1 : if invariants.Sometimes(5) {
97 1 : assertComparer := base.MakeAssertComparer(*opts.Comparer)
98 1 : opts.Comparer = &assertComparer
99 1 : }
100 :
101 : // In all error cases, we return db = nil; this is used by various
102 : // deferred cleanups.
103 :
104 : // Open the database and WAL directories first.
105 1 : walDirname, dataDir, err := prepareAndOpenDirs(dirname, opts)
106 1 : if err != nil {
107 0 : return nil, errors.Wrapf(err, "error opening database at %q", dirname)
108 0 : }
109 1 : defer func() {
110 1 : if db == nil {
111 0 : dataDir.Close()
112 0 : }
113 : }()
114 :
115 : // Lock the database directory.
116 1 : var fileLock *Lock
117 1 : if opts.Lock != nil {
118 0 : // The caller already acquired the database lock. Ensure that the
119 0 : // directory matches.
120 0 : if err := opts.Lock.pathMatches(dirname); err != nil {
121 0 : return nil, err
122 0 : }
123 0 : if err := opts.Lock.refForOpen(); err != nil {
124 0 : return nil, err
125 0 : }
126 0 : fileLock = opts.Lock
127 1 : } else {
128 1 : fileLock, err = LockDirectory(dirname, opts.FS)
129 1 : if err != nil {
130 0 : return nil, err
131 0 : }
132 : }
133 1 : defer func() {
134 1 : if db == nil {
135 0 : _ = fileLock.Close()
136 0 : }
137 : }()
138 :
139 : // List the directory contents. This also happens to include WAL log files, if
140 : // they are in the same dir, but we will ignore those below. The provider is
141 : // also given this list, but it ignores non sstable files.
142 1 : ls, err := opts.FS.List(dirname)
143 1 : if err != nil {
144 0 : return nil, err
145 0 : }
146 :
147 : // Establish the format major version.
148 1 : formatVersion, formatVersionMarker, err := lookupFormatMajorVersion(opts.FS, dirname, ls)
149 1 : if err != nil {
150 0 : return nil, err
151 0 : }
152 1 : defer func() {
153 1 : if db == nil {
154 0 : _ = formatVersionMarker.Close()
155 0 : }
156 : }()
157 :
158 1 : noFormatVersionMarker := formatVersion == FormatDefault
159 1 : if noFormatVersionMarker {
160 1 : // We will initialize the store at the minimum possible format, then upgrade
161 1 : // the format to the desired one. This helps test the format upgrade code.
162 1 : formatVersion = FormatMinSupported
163 1 : if opts.Experimental.CreateOnShared != remote.CreateOnSharedNone {
164 1 : formatVersion = FormatMinForSharedObjects
165 1 : }
166 : // There is no format version marker file. There are three cases:
167 : // - we are trying to open an existing store that was created at
168 : // FormatMostCompatible (the only one without a version marker file)
169 : // - we are creating a new store;
170 : // - we are retrying a failed creation.
171 : //
172 : // To error in the first case, we set ErrorIfNotPristine.
173 1 : opts.ErrorIfNotPristine = true
174 1 : defer func() {
175 1 : if err != nil && errors.Is(err, ErrDBNotPristine) {
176 0 : // We must be trying to open an existing store at FormatMostCompatible.
177 0 : // Correct the error in this case -we
178 0 : err = errors.Newf(
179 0 : "pebble: database %q written in format major version 1 which is no longer supported",
180 0 : dirname)
181 0 : }
182 : }()
183 : }
184 :
185 : // Find the currently active manifest, if there is one.
186 1 : manifestMarker, manifestFileNum, manifestExists, err := findCurrentManifest(opts.FS, dirname, ls)
187 1 : if err != nil {
188 0 : return nil, errors.Wrapf(err, "pebble: database %q", dirname)
189 0 : }
190 1 : defer func() {
191 1 : if db == nil {
192 0 : _ = manifestMarker.Close()
193 0 : }
194 : }()
195 :
196 : // Atomic markers may leave behind obsolete files if there's a crash
197 : // mid-update. Clean these up if we're not in read-only mode.
198 1 : if !opts.ReadOnly {
199 1 : if err := formatVersionMarker.RemoveObsolete(); err != nil {
200 0 : return nil, err
201 0 : }
202 1 : if err := manifestMarker.RemoveObsolete(); err != nil {
203 0 : return nil, err
204 0 : }
205 : }
206 :
207 1 : if opts.Cache == nil {
208 1 : opts.Cache = cache.New(opts.CacheSize)
209 1 : defer opts.Cache.Unref()
210 1 : }
211 :
212 1 : d := &DB{
213 1 : cacheHandle: opts.Cache.NewHandle(),
214 1 : dirname: dirname,
215 1 : opts: opts,
216 1 : cmp: opts.Comparer.Compare,
217 1 : equal: opts.Comparer.Equal,
218 1 : merge: opts.Merger.Merge,
219 1 : split: opts.Comparer.Split,
220 1 : abbreviatedKey: opts.Comparer.AbbreviatedKey,
221 1 : largeBatchThreshold: (opts.MemTableSize - uint64(memTableEmptySize)) / 2,
222 1 : fileLock: fileLock,
223 1 : dataDir: dataDir,
224 1 : closed: new(atomic.Value),
225 1 : closedCh: make(chan struct{}),
226 1 : }
227 1 : d.mu.versions = &versionSet{}
228 1 : d.diskAvailBytes.Store(math.MaxUint64)
229 1 : d.problemSpans.Init(manifest.NumLevels, opts.Comparer.Compare)
230 1 :
231 1 : defer func() {
232 1 : // If an error or panic occurs during open, attempt to release the manually
233 1 : // allocated memory resources. Note that rather than look for an error, we
234 1 : // look for the return of a nil DB pointer.
235 1 : if r := recover(); db == nil {
236 0 : // If there's an unused, recycled memtable, we need to release its memory.
237 0 : if obsoleteMemTable := d.memTableRecycle.Swap(nil); obsoleteMemTable != nil {
238 0 : d.freeMemTable(obsoleteMemTable)
239 0 : }
240 :
241 0 : if d.fileCache != nil {
242 0 : _ = d.fileCache.Close()
243 0 : }
244 0 : d.cacheHandle.Close()
245 0 :
246 0 : for _, mem := range d.mu.mem.queue {
247 0 : switch t := mem.flushable.(type) {
248 0 : case *memTable:
249 0 : manual.Free(manual.MemTable, t.arenaBuf)
250 0 : t.arenaBuf = manual.Buf{}
251 : }
252 : }
253 0 : if d.cleanupManager != nil {
254 0 : d.cleanupManager.Close()
255 0 : }
256 0 : if d.objProvider != nil {
257 0 : _ = d.objProvider.Close()
258 0 : }
259 0 : if r != nil {
260 0 : panic(r)
261 : }
262 : }
263 : }()
264 :
265 1 : d.commit = newCommitPipeline(commitEnv{
266 1 : logSeqNum: &d.mu.versions.logSeqNum,
267 1 : visibleSeqNum: &d.mu.versions.visibleSeqNum,
268 1 : apply: d.commitApply,
269 1 : write: d.commitWrite,
270 1 : })
271 1 : d.mu.nextJobID = 1
272 1 : d.mu.mem.nextSize = opts.MemTableSize
273 1 : if d.mu.mem.nextSize > initialMemTableSize {
274 1 : d.mu.mem.nextSize = initialMemTableSize
275 1 : }
276 1 : d.mu.compact.cond.L = &d.mu.Mutex
277 1 : d.mu.compact.inProgress = make(map[*compaction]struct{})
278 1 : d.mu.compact.noOngoingFlushStartTime = crtime.NowMono()
279 1 : d.mu.snapshots.init()
280 1 : // logSeqNum is the next sequence number that will be assigned.
281 1 : // Start assigning sequence numbers from base.SeqNumStart to leave
282 1 : // room for reserved sequence numbers (see comments around
283 1 : // SeqNumStart).
284 1 : d.mu.versions.logSeqNum.Store(base.SeqNumStart)
285 1 : d.mu.formatVers.vers.Store(uint64(formatVersion))
286 1 : d.mu.formatVers.marker = formatVersionMarker
287 1 :
288 1 : d.timeNow = time.Now
289 1 : d.openedAt = d.timeNow()
290 1 :
291 1 : d.mu.Lock()
292 1 : defer d.mu.Unlock()
293 1 :
294 1 : jobID := d.newJobIDLocked()
295 1 :
296 1 : providerSettings := opts.MakeObjStorageProviderSettings(dirname)
297 1 : providerSettings.FSDirInitialListing = ls
298 1 : d.objProvider, err = objstorageprovider.Open(providerSettings)
299 1 : if err != nil {
300 0 : return nil, err
301 0 : }
302 :
303 1 : if !manifestExists {
304 1 : // DB does not exist.
305 1 : if d.opts.ErrorIfNotExists || d.opts.ReadOnly {
306 0 : return nil, errors.Wrapf(ErrDBDoesNotExist, "dirname=%q", dirname)
307 0 : }
308 :
309 : // Create the DB.
310 1 : if err := d.mu.versions.create(
311 1 : jobID, dirname, d.objProvider, opts, manifestMarker, d.FormatMajorVersion, &d.mu.Mutex); err != nil {
312 0 : return nil, err
313 0 : }
314 1 : } else {
315 1 : if opts.ErrorIfExists {
316 0 : return nil, errors.Wrapf(ErrDBAlreadyExists, "dirname=%q", dirname)
317 0 : }
318 : // Load the version set.
319 1 : if err := d.mu.versions.load(
320 1 : dirname, d.objProvider, opts, manifestFileNum, manifestMarker, d.FormatMajorVersion, &d.mu.Mutex); err != nil {
321 0 : return nil, err
322 0 : }
323 1 : if opts.ErrorIfNotPristine {
324 0 : liveFileNums := make(map[base.DiskFileNum]struct{})
325 0 : d.mu.versions.addLiveFileNums(liveFileNums)
326 0 : if len(liveFileNums) != 0 {
327 0 : return nil, errors.Wrapf(ErrDBNotPristine, "dirname=%q", dirname)
328 0 : }
329 : }
330 : }
331 :
332 : // In read-only mode, we replay directly into the mutable memtable but never
333 : // flush it. We need to delay creation of the memtable until we know the
334 : // sequence number of the first batch that will be inserted.
335 1 : if !d.opts.ReadOnly {
336 1 : var entry *flushableEntry
337 1 : d.mu.mem.mutable, entry = d.newMemTable(0 /* logNum */, d.mu.versions.logSeqNum.Load(), 0 /* minSize */)
338 1 : d.mu.mem.queue = append(d.mu.mem.queue, entry)
339 1 : }
340 :
341 1 : d.mu.log.metrics.fsyncLatency = prometheus.NewHistogram(prometheus.HistogramOpts{
342 1 : Buckets: FsyncLatencyBuckets,
343 1 : })
344 1 : walOpts := wal.Options{
345 1 : Primary: wal.Dir{FS: opts.FS, Dirname: walDirname},
346 1 : Secondary: wal.Dir{},
347 1 : MinUnflushedWALNum: wal.NumWAL(d.mu.versions.minUnflushedLogNum),
348 1 : MaxNumRecyclableLogs: opts.MemTableStopWritesThreshold + 1,
349 1 : NoSyncOnClose: opts.NoSyncOnClose,
350 1 : BytesPerSync: opts.WALBytesPerSync,
351 1 : PreallocateSize: d.walPreallocateSize,
352 1 : MinSyncInterval: opts.WALMinSyncInterval,
353 1 : FsyncLatency: d.mu.log.metrics.fsyncLatency,
354 1 : QueueSemChan: d.commit.logSyncQSem,
355 1 : Logger: opts.Logger,
356 1 : EventListener: walEventListenerAdaptor{l: opts.EventListener},
357 1 : WriteWALSyncOffsets: func() bool { return d.FormatMajorVersion() >= FormatWALSyncChunks },
358 : }
359 1 : if opts.WALFailover != nil {
360 1 : walOpts.Secondary = opts.WALFailover.Secondary
361 1 : walOpts.Secondary.Dirname = resolveStorePath(dirname, walOpts.Secondary.Dirname)
362 1 : walOpts.FailoverOptions = opts.WALFailover.FailoverOptions
363 1 : walOpts.FailoverWriteAndSyncLatency = prometheus.NewHistogram(prometheus.HistogramOpts{
364 1 : Buckets: FsyncLatencyBuckets,
365 1 : })
366 1 : }
367 1 : walDirs := walOpts.Dirs()
368 1 : for _, dir := range opts.WALRecoveryDirs {
369 0 : dir.Dirname = resolveStorePath(dirname, dir.Dirname)
370 0 : walDirs = append(walDirs, dir)
371 0 : }
372 1 : wals, err := wal.Scan(walDirs...)
373 1 : if err != nil {
374 0 : return nil, err
375 0 : }
376 1 : walManager, err := wal.Init(walOpts, wals)
377 1 : if err != nil {
378 0 : return nil, err
379 0 : }
380 1 : defer func() {
381 1 : if db == nil {
382 0 : _ = walManager.Close()
383 0 : }
384 : }()
385 :
386 1 : d.mu.log.manager = walManager
387 1 :
388 1 : d.cleanupManager = openCleanupManager(opts, d.objProvider, d.getDeletionPacerInfo)
389 1 :
390 1 : if manifestExists && !opts.DisableConsistencyCheck {
391 1 : curVersion := d.mu.versions.currentVersion()
392 1 : if err := checkConsistency(curVersion, d.objProvider); err != nil {
393 0 : return nil, err
394 0 : }
395 : }
396 :
397 1 : fileCacheSize := FileCacheSize(opts.MaxOpenFiles)
398 1 : if opts.FileCache == nil {
399 1 : opts.FileCache = NewFileCache(opts.Experimental.FileCacheShards, fileCacheSize)
400 1 : defer opts.FileCache.Unref()
401 1 : }
402 1 : d.fileCache = opts.FileCache.newHandle(d.cacheHandle, d.objProvider, d.opts.LoggerAndTracer, d.opts.MakeReaderOptions(), d.reportCorruption)
403 1 : d.newIters = d.fileCache.newIters
404 1 : d.tableNewRangeKeyIter = tableNewRangeKeyIter(d.newIters)
405 1 :
406 1 : d.mu.annotators.totalFileSize = d.makeFileSizeAnnotator(func(f *manifest.TableMetadata) bool {
407 0 : return true
408 0 : })
409 1 : d.mu.annotators.remoteSize = d.makeFileSizeAnnotator(func(f *manifest.TableMetadata) bool {
410 0 : meta, err := d.objProvider.Lookup(base.FileTypeTable, f.TableBacking.DiskFileNum)
411 0 : if err != nil {
412 0 : return false
413 0 : }
414 0 : return meta.IsRemote()
415 : })
416 1 : d.mu.annotators.externalSize = d.makeFileSizeAnnotator(func(f *manifest.TableMetadata) bool {
417 0 : meta, err := d.objProvider.Lookup(base.FileTypeTable, f.TableBacking.DiskFileNum)
418 0 : if err != nil {
419 0 : return false
420 0 : }
421 0 : return meta.IsRemote() && meta.Remote.CleanupMethod == objstorage.SharedNoCleanup
422 : })
423 :
424 1 : var previousOptionsFileNum base.DiskFileNum
425 1 : var previousOptionsFilename string
426 1 : for _, filename := range ls {
427 1 : ft, fn, ok := base.ParseFilename(opts.FS, filename)
428 1 : if !ok {
429 1 : continue
430 : }
431 :
432 : // Don't reuse any obsolete file numbers to avoid modifying an
433 : // ingested sstable's original external file.
434 1 : d.mu.versions.markFileNumUsed(fn)
435 1 :
436 1 : switch ft {
437 0 : case base.FileTypeLog:
438 : // Ignore.
439 1 : case base.FileTypeOptions:
440 1 : if previousOptionsFileNum < fn {
441 1 : previousOptionsFileNum = fn
442 1 : previousOptionsFilename = filename
443 1 : }
444 0 : case base.FileTypeTemp, base.FileTypeOldTemp:
445 0 : if !d.opts.ReadOnly {
446 0 : // Some codepaths write to a temporary file and then
447 0 : // rename it to its final location when complete. A
448 0 : // temp file is leftover if a process exits before the
449 0 : // rename. Remove it.
450 0 : err := opts.FS.Remove(opts.FS.PathJoin(dirname, filename))
451 0 : if err != nil {
452 0 : return nil, err
453 0 : }
454 : }
455 : }
456 : }
457 1 : if n := len(wals); n > 0 {
458 1 : // Don't reuse any obsolete file numbers to avoid modifying an
459 1 : // ingested sstable's original external file.
460 1 : d.mu.versions.markFileNumUsed(base.DiskFileNum(wals[n-1].Num))
461 1 : }
462 :
463 : // Ratchet d.mu.versions.nextFileNum ahead of all known objects in the
464 : // objProvider. This avoids FileNum collisions with obsolete sstables.
465 1 : objects := d.objProvider.List()
466 1 : for _, obj := range objects {
467 1 : d.mu.versions.markFileNumUsed(obj.DiskFileNum)
468 1 : }
469 :
470 : // Validate the most-recent OPTIONS file, if there is one.
471 1 : if previousOptionsFilename != "" {
472 1 : path := opts.FS.PathJoin(dirname, previousOptionsFilename)
473 1 : previousOptions, err := readOptionsFile(opts, path)
474 1 : if err != nil {
475 0 : return nil, err
476 0 : }
477 1 : if err := opts.CheckCompatibility(dirname, previousOptions); err != nil {
478 0 : return nil, err
479 0 : }
480 : }
481 :
482 : // Replay any newer log files than the ones named in the manifest.
483 1 : var replayWALs wal.Logs
484 1 : for i, w := range wals {
485 1 : if base.DiskFileNum(w.Num) >= d.mu.versions.minUnflushedLogNum {
486 1 : replayWALs = wals[i:]
487 1 : break
488 : }
489 : }
490 1 : var flushableIngests []*ingestedFlushable
491 1 : for i, lf := range replayWALs {
492 1 : // WALs other than the last one would have been closed cleanly.
493 1 : //
494 1 : // Note: we used to never require strict WAL tails when reading from older
495 1 : // versions: RocksDB 6.2.1 and the version of Pebble included in CockroachDB
496 1 : // 20.1 do not guarantee that closed WALs end cleanly. But the earliest
497 1 : // compatible Pebble format is newer and guarantees a clean EOF.
498 1 : strictWALTail := i < len(replayWALs)-1
499 1 : fi, maxSeqNum, err := d.replayWAL(jobID, lf, strictWALTail)
500 1 : if err != nil {
501 0 : return nil, err
502 0 : }
503 1 : if len(fi) > 0 {
504 1 : flushableIngests = append(flushableIngests, fi...)
505 1 : }
506 1 : if d.mu.versions.logSeqNum.Load() < maxSeqNum {
507 1 : d.mu.versions.logSeqNum.Store(maxSeqNum)
508 1 : }
509 : }
510 1 : if d.mu.mem.mutable == nil {
511 1 : // Recreate the mutable memtable if replayWAL got rid of it.
512 1 : var entry *flushableEntry
513 1 : d.mu.mem.mutable, entry = d.newMemTable(d.mu.versions.getNextDiskFileNum(), d.mu.versions.logSeqNum.Load(), 0 /* minSize */)
514 1 : d.mu.mem.queue = append(d.mu.mem.queue, entry)
515 1 : }
516 1 : d.mu.versions.visibleSeqNum.Store(d.mu.versions.logSeqNum.Load())
517 1 :
518 1 : // Register with the CompactionScheduler before calling
519 1 : // d.maybeScheduleFlush, since completion of the flush can trigger
520 1 : // compactions.
521 1 : d.opts.Experimental.CompactionScheduler.Register(2, d)
522 1 : if !d.opts.ReadOnly {
523 1 : d.maybeScheduleFlush()
524 1 : for d.mu.compact.flushing {
525 1 : d.mu.compact.cond.Wait()
526 1 : }
527 :
528 : // Create an empty .log file for the mutable memtable.
529 1 : newLogNum := d.mu.versions.getNextDiskFileNum()
530 1 : d.mu.log.writer, err = d.mu.log.manager.Create(wal.NumWAL(newLogNum), int(jobID))
531 1 : if err != nil {
532 0 : return nil, err
533 0 : }
534 :
535 : // This isn't strictly necessary as we don't use the log number for
536 : // memtables being flushed, only for the next unflushed memtable.
537 1 : d.mu.mem.queue[len(d.mu.mem.queue)-1].logNum = newLogNum
538 : }
539 1 : d.updateReadStateLocked(d.opts.DebugCheck)
540 1 :
541 1 : if !d.opts.ReadOnly {
542 1 : // If the Options specify a format major version higher than the
543 1 : // loaded database's, upgrade it. If this is a new database, this
544 1 : // code path also performs an initial upgrade from the starting
545 1 : // implicit MinSupported version.
546 1 : //
547 1 : // We ratchet the version this far into Open so that migrations have a read
548 1 : // state available. Note that this also results in creating/updating the
549 1 : // format version marker file.
550 1 : if opts.FormatMajorVersion > d.FormatMajorVersion() {
551 1 : if err := d.ratchetFormatMajorVersionLocked(opts.FormatMajorVersion); err != nil {
552 0 : return nil, err
553 0 : }
554 1 : } else if noFormatVersionMarker {
555 1 : // We are creating a new store. Create the format version marker file.
556 1 : if err := d.writeFormatVersionMarker(d.FormatMajorVersion()); err != nil {
557 0 : return nil, err
558 0 : }
559 : }
560 :
561 : // Write the current options to disk.
562 1 : d.optionsFileNum = d.mu.versions.getNextDiskFileNum()
563 1 : tmpPath := base.MakeFilepath(opts.FS, dirname, base.FileTypeTemp, d.optionsFileNum)
564 1 : optionsPath := base.MakeFilepath(opts.FS, dirname, base.FileTypeOptions, d.optionsFileNum)
565 1 :
566 1 : // Write them to a temporary file first, in case we crash before
567 1 : // we're done. A corrupt options file prevents opening the
568 1 : // database.
569 1 : optionsFile, err := opts.FS.Create(tmpPath, vfs.WriteCategoryUnspecified)
570 1 : if err != nil {
571 0 : return nil, err
572 0 : }
573 1 : serializedOpts := []byte(opts.String())
574 1 : if _, err := optionsFile.Write(serializedOpts); err != nil {
575 0 : return nil, errors.CombineErrors(err, optionsFile.Close())
576 0 : }
577 1 : d.optionsFileSize = uint64(len(serializedOpts))
578 1 : if err := optionsFile.Sync(); err != nil {
579 0 : return nil, errors.CombineErrors(err, optionsFile.Close())
580 0 : }
581 1 : if err := optionsFile.Close(); err != nil {
582 0 : return nil, err
583 0 : }
584 : // Atomically rename to the OPTIONS-XXXXXX path. This rename is
585 : // guaranteed to be atomic because the destination path does not
586 : // exist.
587 1 : if err := opts.FS.Rename(tmpPath, optionsPath); err != nil {
588 0 : return nil, err
589 0 : }
590 1 : if err := d.dataDir.Sync(); err != nil {
591 0 : return nil, err
592 0 : }
593 : }
594 :
595 1 : if !d.opts.ReadOnly {
596 1 : // Get a fresh list of files, in case some of the earlier flushes/compactions
597 1 : // have deleted some files.
598 1 : ls, err := opts.FS.List(dirname)
599 1 : if err != nil {
600 0 : return nil, err
601 0 : }
602 1 : d.scanObsoleteFiles(ls, flushableIngests)
603 1 : d.deleteObsoleteFiles(jobID)
604 : }
605 : // Else, nothing is obsolete.
606 :
607 1 : d.mu.tableStats.cond.L = &d.mu.Mutex
608 1 : d.mu.tableValidation.cond.L = &d.mu.Mutex
609 1 : if !d.opts.ReadOnly {
610 1 : d.maybeCollectTableStatsLocked()
611 1 : }
612 1 : d.calculateDiskAvailableBytes()
613 1 :
614 1 : d.maybeScheduleFlush()
615 1 : d.maybeScheduleCompaction()
616 1 :
617 1 : // Note: this is a no-op if invariants are disabled or race is enabled.
618 1 : //
619 1 : // Setting a finalizer on *DB causes *DB to never be reclaimed and the
620 1 : // finalizer to never be run. The problem is due to this limitation of
621 1 : // finalizers mention in the SetFinalizer docs:
622 1 : //
623 1 : // If a cyclic structure includes a block with a finalizer, that cycle is
624 1 : // not guaranteed to be garbage collected and the finalizer is not
625 1 : // guaranteed to run, because there is no ordering that respects the
626 1 : // dependencies.
627 1 : //
628 1 : // DB has cycles with several of its internal structures: readState,
629 1 : // newIters, fileCache, versions, etc. Each of this individually cause a
630 1 : // cycle and prevent the finalizer from being run. But we can workaround this
631 1 : // finializer limitation by setting a finalizer on another object that is
632 1 : // tied to the lifetime of DB: the DB.closed atomic.Value.
633 1 : dPtr := fmt.Sprintf("%p", d)
634 1 : invariants.SetFinalizer(d.closed, func(obj interface{}) {
635 0 : v := obj.(*atomic.Value)
636 0 : if err := v.Load(); err == nil {
637 0 : fmt.Fprintf(os.Stderr, "%s: unreferenced DB not closed\n", dPtr)
638 0 : os.Exit(1)
639 0 : }
640 : })
641 :
642 1 : return d, nil
643 : }
644 :
645 : // prepareAndOpenDirs opens the directories for the store (and creates them if
646 : // necessary).
647 : //
648 : // Returns an error if ReadOnly is set and the directories don't exist.
649 : func prepareAndOpenDirs(
650 : dirname string, opts *Options,
651 1 : ) (walDirname string, dataDir vfs.File, err error) {
652 1 : walDirname = dirname
653 1 : if opts.WALDir != "" {
654 1 : walDirname = resolveStorePath(dirname, opts.WALDir)
655 1 : }
656 :
657 : // Create directories if needed.
658 1 : if !opts.ReadOnly {
659 1 : f, err := mkdirAllAndSyncParents(opts.FS, dirname)
660 1 : if err != nil {
661 0 : return "", nil, err
662 0 : }
663 1 : f.Close()
664 1 : if walDirname != dirname {
665 1 : f, err := mkdirAllAndSyncParents(opts.FS, walDirname)
666 1 : if err != nil {
667 0 : return "", nil, err
668 0 : }
669 1 : f.Close()
670 : }
671 1 : if opts.WALFailover != nil {
672 1 : secondary := opts.WALFailover.Secondary
673 1 : f, err := mkdirAllAndSyncParents(secondary.FS, resolveStorePath(dirname, secondary.Dirname))
674 1 : if err != nil {
675 0 : return "", nil, err
676 0 : }
677 1 : f.Close()
678 : }
679 : }
680 :
681 1 : dataDir, err = opts.FS.OpenDir(dirname)
682 1 : if err != nil {
683 0 : if opts.ReadOnly && oserror.IsNotExist(err) {
684 0 : return "", nil, errors.Errorf("pebble: database %q does not exist", dirname)
685 0 : }
686 0 : return "", nil, err
687 : }
688 1 : if opts.ReadOnly && walDirname != dirname {
689 0 : // Check that the wal dir exists.
690 0 : walDir, err := opts.FS.OpenDir(walDirname)
691 0 : if err != nil {
692 0 : dataDir.Close()
693 0 : return "", nil, err
694 0 : }
695 0 : walDir.Close()
696 : }
697 :
698 1 : return walDirname, dataDir, nil
699 : }
700 :
701 : // GetVersion returns the engine version string from the latest options
702 : // file present in dir. Used to check what Pebble or RocksDB version was last
703 : // used to write to the database stored in this directory. An empty string is
704 : // returned if no valid OPTIONS file with a version key was found.
705 0 : func GetVersion(dir string, fs vfs.FS) (string, error) {
706 0 : ls, err := fs.List(dir)
707 0 : if err != nil {
708 0 : return "", err
709 0 : }
710 0 : var version string
711 0 : lastOptionsSeen := base.DiskFileNum(0)
712 0 : for _, filename := range ls {
713 0 : ft, fn, ok := base.ParseFilename(fs, filename)
714 0 : if !ok {
715 0 : continue
716 : }
717 0 : switch ft {
718 0 : case base.FileTypeOptions:
719 0 : // If this file has a higher number than the last options file
720 0 : // processed, reset version. This is because rocksdb often
721 0 : // writes multiple options files without deleting previous ones.
722 0 : // Otherwise, skip parsing this options file.
723 0 : if fn > lastOptionsSeen {
724 0 : version = ""
725 0 : lastOptionsSeen = fn
726 0 : } else {
727 0 : continue
728 : }
729 0 : f, err := fs.Open(fs.PathJoin(dir, filename))
730 0 : if err != nil {
731 0 : return "", err
732 0 : }
733 0 : data, err := io.ReadAll(f)
734 0 : f.Close()
735 0 :
736 0 : if err != nil {
737 0 : return "", err
738 0 : }
739 0 : err = parseOptions(string(data), parseOptionsFuncs{
740 0 : visitKeyValue: func(i, j int, section, key, value string) error {
741 0 : switch {
742 0 : case section == "Version":
743 0 : switch key {
744 0 : case "pebble_version":
745 0 : version = value
746 0 : case "rocksdb_version":
747 0 : version = fmt.Sprintf("rocksdb v%s", value)
748 : }
749 : }
750 0 : return nil
751 : },
752 : })
753 0 : if err != nil {
754 0 : return "", err
755 0 : }
756 : }
757 : }
758 0 : return version, nil
759 : }
760 :
761 : func (d *DB) replayIngestedFlushable(
762 : b *Batch, logNum base.DiskFileNum,
763 1 : ) (entry *flushableEntry, err error) {
764 1 : br := b.Reader()
765 1 : seqNum := b.SeqNum()
766 1 :
767 1 : fileNums := make([]base.DiskFileNum, 0, b.Count())
768 1 : var exciseSpan KeyRange
769 1 : addFileNum := func(encodedFileNum []byte) {
770 1 : fileNum, n := binary.Uvarint(encodedFileNum)
771 1 : if n <= 0 {
772 0 : panic("pebble: ingest sstable file num is invalid")
773 : }
774 1 : fileNums = append(fileNums, base.DiskFileNum(fileNum))
775 : }
776 :
777 1 : for i := 0; i < int(b.Count()); i++ {
778 1 : kind, key, val, ok, err := br.Next()
779 1 : if err != nil {
780 0 : return nil, err
781 0 : }
782 1 : if kind != InternalKeyKindIngestSST && kind != InternalKeyKindExcise {
783 0 : panic("pebble: invalid batch key kind")
784 : }
785 1 : if !ok {
786 0 : panic("pebble: invalid batch count")
787 : }
788 1 : if kind == base.InternalKeyKindExcise {
789 1 : if exciseSpan.Valid() {
790 0 : panic("pebble: multiple excise spans in a single batch")
791 : }
792 1 : exciseSpan.Start = slices.Clone(key)
793 1 : exciseSpan.End = slices.Clone(val)
794 1 : continue
795 : }
796 1 : addFileNum(key)
797 : }
798 :
799 1 : if _, _, _, ok, err := br.Next(); err != nil {
800 0 : return nil, err
801 1 : } else if ok {
802 0 : panic("pebble: invalid number of entries in batch")
803 : }
804 :
805 1 : meta := make([]*manifest.TableMetadata, len(fileNums))
806 1 : var lastRangeKey keyspan.Span
807 1 : for i, n := range fileNums {
808 1 : readable, err := d.objProvider.OpenForReading(context.TODO(), base.FileTypeTable, n,
809 1 : objstorage.OpenOptions{MustExist: true})
810 1 : if err != nil {
811 0 : return nil, errors.Wrap(err, "pebble: error when opening flushable ingest files")
812 0 : }
813 : // NB: ingestLoad1 will close readable.
814 1 : meta[i], lastRangeKey, err = ingestLoad1(context.TODO(), d.opts, d.FormatMajorVersion(),
815 1 : readable, d.cacheHandle, base.PhysicalTableFileNum(n), disableRangeKeyChecks())
816 1 : if err != nil {
817 0 : return nil, errors.Wrap(err, "pebble: error when loading flushable ingest files")
818 0 : }
819 : }
820 1 : if lastRangeKey.Valid() && d.opts.Comparer.Split.HasSuffix(lastRangeKey.End) {
821 0 : return nil, errors.AssertionFailedf("pebble: last ingest sstable has suffixed range key end %s",
822 0 : d.opts.Comparer.FormatKey(lastRangeKey.End))
823 0 : }
824 :
825 1 : numFiles := len(meta)
826 1 : if exciseSpan.Valid() {
827 1 : numFiles++
828 1 : }
829 1 : if uint32(numFiles) != b.Count() {
830 0 : panic("pebble: couldn't load all files in WAL entry")
831 : }
832 :
833 1 : return d.newIngestedFlushableEntry(meta, seqNum, logNum, exciseSpan)
834 : }
835 :
836 : // replayWAL replays the edits in the specified WAL. If the DB is in read
837 : // only mode, then the WALs are replayed into memtables and not flushed. If
838 : // the DB is not in read only mode, then the contents of the WAL are
839 : // guaranteed to be flushed when a flush is scheduled after this method is run.
840 : // Note that this flushing is very important for guaranteeing durability:
841 : // the application may have had a number of pending
842 : // fsyncs to the WAL before the process crashed, and those fsyncs may not have
843 : // happened but the corresponding data may now be readable from the WAL (while
844 : // sitting in write-back caches in the kernel or the storage device). By
845 : // reading the WAL (including the non-fsynced data) and then flushing all
846 : // these changes (flush does fsyncs), we are able to guarantee that the
847 : // initial state of the DB is durable.
848 : //
849 : // This method mutates d.mu.mem.queue and possibly d.mu.mem.mutable and replays
850 : // WALs into the flushable queue. Flushing of the queue is expected to be handled
851 : // by callers. A list of flushable ingests (but not memtables) replayed is returned.
852 : //
853 : // d.mu must be held when calling this, but the mutex may be dropped and
854 : // re-acquired during the course of this method.
855 : func (d *DB) replayWAL(
856 : jobID JobID, ll wal.LogicalLog, strictWALTail bool,
857 1 : ) (flushableIngests []*ingestedFlushable, maxSeqNum base.SeqNum, err error) {
858 1 : rr := ll.OpenForRead()
859 1 : defer func() { _ = rr.Close() }()
860 1 : var (
861 1 : b Batch
862 1 : buf bytes.Buffer
863 1 : mem *memTable
864 1 : entry *flushableEntry
865 1 : offset wal.Offset
866 1 : lastFlushOffset int64
867 1 : keysReplayed int64 // number of keys replayed
868 1 : batchesReplayed int64 // number of batches replayed
869 1 : )
870 1 :
871 1 : // TODO(jackson): This function is interspersed with panics, in addition to
872 1 : // corruption error propagation. Audit them to ensure we're truly only
873 1 : // panicking where the error points to Pebble bug and not user or
874 1 : // hardware-induced corruption.
875 1 :
876 1 : // "Flushes" (ie. closes off) the current memtable, if not nil.
877 1 : flushMem := func() {
878 1 : if mem == nil {
879 1 : return
880 1 : }
881 1 : mem.writerUnref()
882 1 : if d.mu.mem.mutable == mem {
883 1 : d.mu.mem.mutable = nil
884 1 : }
885 1 : entry.flushForced = !d.opts.ReadOnly
886 1 : var logSize uint64
887 1 : mergedOffset := offset.Physical + offset.PreviousFilesBytes
888 1 : if mergedOffset >= lastFlushOffset {
889 1 : logSize = uint64(mergedOffset - lastFlushOffset)
890 1 : }
891 : // Else, this was the initial memtable in the read-only case which must have
892 : // been empty, but we need to flush it since we don't want to add to it later.
893 1 : lastFlushOffset = mergedOffset
894 1 : entry.logSize = logSize
895 1 : mem, entry = nil, nil
896 : }
897 :
898 1 : mem = d.mu.mem.mutable
899 1 : if mem != nil {
900 1 : entry = d.mu.mem.queue[len(d.mu.mem.queue)-1]
901 1 : if !d.opts.ReadOnly {
902 1 : flushMem()
903 1 : }
904 : }
905 :
906 : // Creates a new memtable if there is no current memtable.
907 1 : ensureMem := func(seqNum base.SeqNum) {
908 1 : if mem != nil {
909 1 : return
910 1 : }
911 1 : mem, entry = d.newMemTable(base.DiskFileNum(ll.Num), seqNum, 0 /* minSize */)
912 1 : d.mu.mem.mutable = mem
913 1 : d.mu.mem.queue = append(d.mu.mem.queue, entry)
914 : }
915 :
916 1 : defer func() {
917 1 : if err != nil {
918 0 : err = errors.WithDetailf(err, "replaying wal %d, offset %s", ll.Num, offset)
919 0 : }
920 : }()
921 :
922 1 : for {
923 1 : var r io.Reader
924 1 : var err error
925 1 : r, offset, err = rr.NextRecord()
926 1 : if err == nil {
927 1 : _, err = io.Copy(&buf, r)
928 1 : }
929 1 : if err != nil {
930 1 : // It is common to encounter a zeroed or invalid chunk due to WAL
931 1 : // preallocation and WAL recycling. However zeroed or invalid chunks
932 1 : // can also be a consequence of corruption / disk rot. When the log
933 1 : // reader encounters one of these cases, it attempts to disambiguate
934 1 : // by reading ahead looking for a future record. If a future chunk
935 1 : // indicates the chunk at the original offset should've been valid, it
936 1 : // surfaces record.ErrInvalidChunk or record.ErrZeroedChunk. These
937 1 : // errors are always indicative of corruption and data loss.
938 1 : //
939 1 : // Otherwise, the reader surfaces record.ErrUnexpectedEOF indicating
940 1 : // that the WAL terminated uncleanly and ambiguously. If the WAL is
941 1 : // the most recent logical WAL, the caller passes in
942 1 : // (strictWALTail=false), indicating we should tolerate the unclean
943 1 : // ending. If the WAL is an older WAL, the caller passes in
944 1 : // (strictWALTail=true), indicating that the WAL should have been
945 1 : // closed cleanly, and we should interpret the
946 1 : // `record.ErrUnexpectedEOF` as corruption and stop recovery.
947 1 : if errors.Is(err, io.EOF) {
948 1 : break
949 0 : } else if errors.Is(err, record.ErrUnexpectedEOF) && !strictWALTail {
950 0 : break
951 0 : } else if (errors.Is(err, record.ErrUnexpectedEOF) && strictWALTail) ||
952 0 : errors.Is(err, record.ErrInvalidChunk) || errors.Is(err, record.ErrZeroedChunk) {
953 0 : // If a read-ahead returns record.ErrInvalidChunk or
954 0 : // record.ErrZeroedChunk, then there's definitively corruption.
955 0 : //
956 0 : // If strictWALTail=true, then record.ErrUnexpectedEOF should
957 0 : // also be considered corruption because the strictWALTail
958 0 : // indicates we expect a clean end to the WAL.
959 0 : //
960 0 : // Other I/O related errors should not be marked with corruption
961 0 : // and simply returned.
962 0 : err = errors.Mark(err, ErrCorruption)
963 0 : }
964 :
965 0 : return nil, 0, errors.Wrap(err, "pebble: error when replaying WAL")
966 : }
967 :
968 1 : if buf.Len() < batchrepr.HeaderLen {
969 0 : return nil, 0, base.CorruptionErrorf("pebble: corrupt wal %s (offset %s)",
970 0 : errors.Safe(base.DiskFileNum(ll.Num)), offset)
971 0 : }
972 :
973 1 : if d.opts.ErrorIfNotPristine {
974 0 : return nil, 0, errors.WithDetailf(ErrDBNotPristine, "location: %q", d.dirname)
975 0 : }
976 :
977 : // Specify Batch.db so that Batch.SetRepr will compute Batch.memTableSize
978 : // which is used below.
979 1 : b = Batch{}
980 1 : b.db = d
981 1 : if err := b.SetRepr(buf.Bytes()); err != nil {
982 0 : return nil, 0, err
983 0 : }
984 1 : seqNum := b.SeqNum()
985 1 : maxSeqNum = seqNum + base.SeqNum(b.Count())
986 1 : keysReplayed += int64(b.Count())
987 1 : batchesReplayed++
988 1 : {
989 1 : br := b.Reader()
990 1 : if kind, _, _, ok, err := br.Next(); err != nil {
991 0 : return nil, 0, err
992 1 : } else if ok && (kind == InternalKeyKindIngestSST || kind == InternalKeyKindExcise) {
993 1 : // We're in the flushable ingests (+ possibly excises) case.
994 1 : //
995 1 : // Ingests require an up-to-date view of the LSM to determine the target
996 1 : // level of ingested sstables, and to accurately compute excises. Instead of
997 1 : // doing an ingest in this function, we just enqueue a flushable ingest
998 1 : // in the flushables queue and run a regular flush.
999 1 : flushMem()
1000 1 : // mem is nil here.
1001 1 : entry, err = d.replayIngestedFlushable(&b, base.DiskFileNum(ll.Num))
1002 1 : if err != nil {
1003 0 : return nil, 0, err
1004 0 : }
1005 1 : fi := entry.flushable.(*ingestedFlushable)
1006 1 : flushableIngests = append(flushableIngests, fi)
1007 1 : d.mu.mem.queue = append(d.mu.mem.queue, entry)
1008 1 : // A flushable ingest is always followed by a WAL rotation.
1009 1 : break
1010 : }
1011 : }
1012 :
1013 1 : if b.memTableSize >= uint64(d.largeBatchThreshold) {
1014 0 : flushMem()
1015 0 : // Make a copy of the data slice since it is currently owned by buf and will
1016 0 : // be reused in the next iteration.
1017 0 : b.data = slices.Clone(b.data)
1018 0 : b.flushable, err = newFlushableBatch(&b, d.opts.Comparer)
1019 0 : if err != nil {
1020 0 : return nil, 0, err
1021 0 : }
1022 0 : entry := d.newFlushableEntry(b.flushable, base.DiskFileNum(ll.Num), b.SeqNum())
1023 0 : // Disable memory accounting by adding a reader ref that will never be
1024 0 : // removed.
1025 0 : entry.readerRefs.Add(1)
1026 0 : d.mu.mem.queue = append(d.mu.mem.queue, entry)
1027 1 : } else {
1028 1 : ensureMem(seqNum)
1029 1 : if err = mem.prepare(&b); err != nil && err != arenaskl.ErrArenaFull {
1030 0 : return nil, 0, err
1031 0 : }
1032 : // We loop since DB.newMemTable() slowly grows the size of allocated memtables, so the
1033 : // batch may not initially fit, but will eventually fit (since it is smaller than
1034 : // largeBatchThreshold).
1035 1 : for err == arenaskl.ErrArenaFull {
1036 1 : flushMem()
1037 1 : ensureMem(seqNum)
1038 1 : err = mem.prepare(&b)
1039 1 : if err != nil && err != arenaskl.ErrArenaFull {
1040 0 : return nil, 0, err
1041 0 : }
1042 : }
1043 1 : if err = mem.apply(&b, seqNum); err != nil {
1044 0 : return nil, 0, err
1045 0 : }
1046 1 : mem.writerUnref()
1047 : }
1048 1 : buf.Reset()
1049 : }
1050 :
1051 1 : d.opts.Logger.Infof("[JOB %d] WAL %s stopped reading at offset: %s; replayed %d keys in %d batches",
1052 1 : jobID, ll.String(), offset, keysReplayed, batchesReplayed)
1053 1 : if !d.opts.ReadOnly {
1054 1 : flushMem()
1055 1 : }
1056 :
1057 : // mem is nil here, if !ReadOnly.
1058 1 : return flushableIngests, maxSeqNum, err
1059 : }
1060 :
1061 1 : func readOptionsFile(opts *Options, path string) (string, error) {
1062 1 : f, err := opts.FS.Open(path)
1063 1 : if err != nil {
1064 0 : return "", err
1065 0 : }
1066 1 : defer f.Close()
1067 1 :
1068 1 : data, err := io.ReadAll(f)
1069 1 : if err != nil {
1070 0 : return "", err
1071 0 : }
1072 1 : return string(data), nil
1073 : }
1074 :
1075 : // DBDesc briefly describes high-level state about a database.
1076 : type DBDesc struct {
1077 : // Exists is true if an existing database was found.
1078 : Exists bool
1079 : // FormatMajorVersion indicates the database's current format
1080 : // version.
1081 : FormatMajorVersion FormatMajorVersion
1082 : // ManifestFilename is the filename of the current active manifest,
1083 : // if the database exists.
1084 : ManifestFilename string
1085 : // OptionsFilename is the filename of the most recent OPTIONS file, if it
1086 : // exists.
1087 : OptionsFilename string
1088 : }
1089 :
1090 : // String implements fmt.Stringer.
1091 0 : func (d *DBDesc) String() string {
1092 0 : if !d.Exists {
1093 0 : return "uninitialized"
1094 0 : }
1095 0 : var buf bytes.Buffer
1096 0 : fmt.Fprintf(&buf, "initialized at format major version %s\n", d.FormatMajorVersion)
1097 0 : fmt.Fprintf(&buf, "manifest: %s\n", d.ManifestFilename)
1098 0 : fmt.Fprintf(&buf, "options: %s", d.OptionsFilename)
1099 0 : return buf.String()
1100 : }
1101 :
1102 : // Peek looks for an existing database in dirname on the provided FS. It
1103 : // returns a brief description of the database. Peek is read-only and
1104 : // does not open the database
1105 0 : func Peek(dirname string, fs vfs.FS) (*DBDesc, error) {
1106 0 : ls, err := fs.List(dirname)
1107 0 : if err != nil {
1108 0 : return nil, err
1109 0 : }
1110 :
1111 0 : vers, versMarker, err := lookupFormatMajorVersion(fs, dirname, ls)
1112 0 : if err != nil {
1113 0 : return nil, err
1114 0 : }
1115 : // TODO(jackson): Immediately closing the marker is clunky. Add a
1116 : // PeekMarker variant that avoids opening the directory.
1117 0 : if err := versMarker.Close(); err != nil {
1118 0 : return nil, err
1119 0 : }
1120 :
1121 : // Find the currently active manifest, if there is one.
1122 0 : manifestMarker, manifestFileNum, exists, err := findCurrentManifest(fs, dirname, ls)
1123 0 : if err != nil {
1124 0 : return nil, err
1125 0 : }
1126 : // TODO(jackson): Immediately closing the marker is clunky. Add a
1127 : // PeekMarker variant that avoids opening the directory.
1128 0 : if err := manifestMarker.Close(); err != nil {
1129 0 : return nil, err
1130 0 : }
1131 :
1132 0 : desc := &DBDesc{
1133 0 : Exists: exists,
1134 0 : FormatMajorVersion: vers,
1135 0 : }
1136 0 :
1137 0 : // Find the OPTIONS file with the highest file number within the list of
1138 0 : // directory entries.
1139 0 : var previousOptionsFileNum base.DiskFileNum
1140 0 : for _, filename := range ls {
1141 0 : ft, fn, ok := base.ParseFilename(fs, filename)
1142 0 : if !ok || ft != base.FileTypeOptions || fn < previousOptionsFileNum {
1143 0 : continue
1144 : }
1145 0 : previousOptionsFileNum = fn
1146 0 : desc.OptionsFilename = fs.PathJoin(dirname, filename)
1147 : }
1148 :
1149 0 : if exists {
1150 0 : desc.ManifestFilename = base.MakeFilepath(fs, dirname, base.FileTypeManifest, manifestFileNum)
1151 0 : }
1152 0 : return desc, nil
1153 : }
1154 :
1155 : // LockDirectory acquires the database directory lock in the named directory,
1156 : // preventing another process from opening the database. LockDirectory returns a
1157 : // handle to the held lock that may be passed to Open through Options.Lock to
1158 : // subsequently open the database, skipping lock acquistion during Open.
1159 : //
1160 : // LockDirectory may be used to expand the critical section protected by the
1161 : // database lock to include setup before the call to Open.
1162 1 : func LockDirectory(dirname string, fs vfs.FS) (*Lock, error) {
1163 1 : fileLock, err := fs.Lock(base.MakeFilepath(fs, dirname, base.FileTypeLock, base.DiskFileNum(0)))
1164 1 : if err != nil {
1165 0 : return nil, err
1166 0 : }
1167 1 : l := &Lock{dirname: dirname, fileLock: fileLock}
1168 1 : l.refs.Store(1)
1169 1 : invariants.SetFinalizer(l, func(obj interface{}) {
1170 1 : if refs := obj.(*Lock).refs.Load(); refs > 0 {
1171 0 : panic(errors.AssertionFailedf("lock for %q finalized with %d refs", dirname, refs))
1172 : }
1173 : })
1174 1 : return l, nil
1175 : }
1176 :
1177 : // Lock represents a file lock on a directory. It may be passed to Open through
1178 : // Options.Lock to elide lock aquisition during Open.
1179 : type Lock struct {
1180 : dirname string
1181 : fileLock io.Closer
1182 : // refs is a count of the number of handles on the lock. refs must be 0, 1
1183 : // or 2.
1184 : //
1185 : // When acquired by the client and passed to Open, refs = 1 and the Open
1186 : // call increments it to 2. When the database is closed, it's decremented to
1187 : // 1. Finally when the original caller, calls Close on the Lock, it's
1188 : // drecemented to zero and the underlying file lock is released.
1189 : //
1190 : // When Open acquires the file lock, refs remains at 1 until the database is
1191 : // closed.
1192 : refs atomic.Int32
1193 : }
1194 :
1195 0 : func (l *Lock) refForOpen() error {
1196 0 : // During Open, when a user passed in a lock, the reference count must be
1197 0 : // exactly 1. If it's zero, the lock is no longer held and is invalid. If
1198 0 : // it's 2, the lock is already in use by another database within the
1199 0 : // process.
1200 0 : if !l.refs.CompareAndSwap(1, 2) {
1201 0 : return errors.Errorf("pebble: unexpected Lock reference count; is the lock already in use?")
1202 0 : }
1203 0 : return nil
1204 : }
1205 :
1206 : // Close releases the lock, permitting another process to lock and open the
1207 : // database. Close must not be called until after a database using the Lock has
1208 : // been closed.
1209 1 : func (l *Lock) Close() error {
1210 1 : if l.refs.Add(-1) > 0 {
1211 0 : return nil
1212 0 : }
1213 1 : defer func() { l.fileLock = nil }()
1214 1 : return l.fileLock.Close()
1215 : }
1216 :
1217 0 : func (l *Lock) pathMatches(dirname string) error {
1218 0 : if dirname == l.dirname {
1219 0 : return nil
1220 0 : }
1221 : // Check for relative paths, symlinks, etc. This isn't ideal because we're
1222 : // circumventing the vfs.FS interface here.
1223 : //
1224 : // TODO(jackson): We could add support for retrieving file inodes through Stat
1225 : // calls in the VFS interface on platforms where it's available and use that
1226 : // to differentiate.
1227 0 : dirStat, err1 := os.Stat(dirname)
1228 0 : lockDirStat, err2 := os.Stat(l.dirname)
1229 0 : if err1 == nil && err2 == nil && os.SameFile(dirStat, lockDirStat) {
1230 0 : return nil
1231 0 : }
1232 0 : return errors.Join(
1233 0 : errors.Newf("pebble: opts.Lock acquired in %q not %q", l.dirname, dirname),
1234 0 : err1, err2)
1235 : }
1236 :
1237 : // ErrDBDoesNotExist is generated when ErrorIfNotExists is set and the database
1238 : // does not exist.
1239 : //
1240 : // Note that errors can be wrapped with more details; use errors.Is().
1241 : var ErrDBDoesNotExist = errors.New("pebble: database does not exist")
1242 :
1243 : // ErrDBAlreadyExists is generated when ErrorIfExists is set and the database
1244 : // already exists.
1245 : //
1246 : // Note that errors can be wrapped with more details; use errors.Is().
1247 : var ErrDBAlreadyExists = errors.New("pebble: database already exists")
1248 :
1249 : // ErrDBNotPristine is generated when ErrorIfNotPristine is set and the database
1250 : // already exists and is not pristine.
1251 : //
1252 : // Note that errors can be wrapped with more details; use errors.Is().
1253 : var ErrDBNotPristine = errors.New("pebble: database already exists and is not pristine")
1254 :
1255 1 : func checkConsistency(v *manifest.Version, objProvider objstorage.Provider) error {
1256 1 : var errs []error
1257 1 : dedup := make(map[base.DiskFileNum]struct{})
1258 1 : for level, files := range v.Levels {
1259 1 : for f := range files.All() {
1260 1 : backingState := f.TableBacking
1261 1 : if _, ok := dedup[backingState.DiskFileNum]; ok {
1262 1 : continue
1263 : }
1264 1 : dedup[backingState.DiskFileNum] = struct{}{}
1265 1 : fileNum := backingState.DiskFileNum
1266 1 : fileSize := backingState.Size
1267 1 : // We skip over remote objects; those are instead checked asynchronously
1268 1 : // by the table stats loading job.
1269 1 : meta, err := objProvider.Lookup(base.FileTypeTable, fileNum)
1270 1 : var size int64
1271 1 : if err == nil {
1272 1 : if meta.IsRemote() {
1273 1 : continue
1274 : }
1275 1 : size, err = objProvider.Size(meta)
1276 : }
1277 1 : if err != nil {
1278 0 : errs = append(errs, errors.Wrapf(err, "L%d: %s", errors.Safe(level), fileNum))
1279 0 : continue
1280 : }
1281 :
1282 1 : if size != int64(fileSize) {
1283 0 : errs = append(errs, errors.Errorf(
1284 0 : "L%d: %s: object size mismatch (%s): %d (disk) != %d (MANIFEST)",
1285 0 : errors.Safe(level), fileNum, objProvider.Path(meta),
1286 0 : errors.Safe(size), errors.Safe(fileSize)))
1287 0 : continue
1288 : }
1289 : }
1290 : }
1291 1 : return errors.Join(errs...)
1292 : }
1293 :
1294 : type walEventListenerAdaptor struct {
1295 : l *EventListener
1296 : }
1297 :
1298 1 : func (l walEventListenerAdaptor) LogCreated(ci wal.CreateInfo) {
1299 1 : // TODO(sumeer): extend WALCreateInfo for the failover case in case the path
1300 1 : // is insufficient to infer whether primary or secondary.
1301 1 : wci := WALCreateInfo{
1302 1 : JobID: ci.JobID,
1303 1 : Path: ci.Path,
1304 1 : FileNum: base.DiskFileNum(ci.Num),
1305 1 : RecycledFileNum: ci.RecycledFileNum,
1306 1 : Err: ci.Err,
1307 1 : }
1308 1 : l.l.WALCreated(wci)
1309 1 : }
|
