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 : "fmt"
9 : "io"
10 : "sync"
11 : "sync/atomic"
12 :
13 : "github.com/cockroachdb/errors"
14 : "github.com/cockroachdb/pebble/internal/base"
15 : "github.com/cockroachdb/pebble/internal/invariants"
16 : "github.com/cockroachdb/pebble/internal/manifest"
17 : "github.com/cockroachdb/pebble/objstorage"
18 : "github.com/cockroachdb/pebble/record"
19 : "github.com/cockroachdb/pebble/vfs"
20 : "github.com/cockroachdb/pebble/vfs/atomicfs"
21 : )
22 :
23 : const numLevels = manifest.NumLevels
24 :
25 : const manifestMarkerName = `manifest`
26 :
27 : // Provide type aliases for the various manifest structs.
28 : type bulkVersionEdit = manifest.BulkVersionEdit
29 : type deletedFileEntry = manifest.DeletedFileEntry
30 : type fileMetadata = manifest.FileMetadata
31 : type physicalMeta = manifest.PhysicalFileMeta
32 : type virtualMeta = manifest.VirtualFileMeta
33 : type fileBacking = manifest.FileBacking
34 : type newFileEntry = manifest.NewFileEntry
35 : type version = manifest.Version
36 : type versionEdit = manifest.VersionEdit
37 : type versionList = manifest.VersionList
38 :
39 : // versionSet manages a collection of immutable versions, and manages the
40 : // creation of a new version from the most recent version. A new version is
41 : // created from an existing version by applying a version edit which is just
42 : // like it sounds: a delta from the previous version. Version edits are logged
43 : // to the MANIFEST file, which is replayed at startup.
44 : type versionSet struct {
45 : // Next seqNum to use for WAL writes.
46 : logSeqNum atomic.Uint64
47 :
48 : // The upper bound on sequence numbers that have been assigned so far. A
49 : // suffix of these sequence numbers may not have been written to a WAL. Both
50 : // logSeqNum and visibleSeqNum are atomically updated by the commitPipeline.
51 : // visibleSeqNum is <= logSeqNum.
52 : visibleSeqNum atomic.Uint64
53 :
54 : // Number of bytes present in sstables being written by in-progress
55 : // compactions. This value will be zero if there are no in-progress
56 : // compactions. Updated and read atomically.
57 : atomicInProgressBytes atomic.Int64
58 :
59 : // Immutable fields.
60 : dirname string
61 : provider objstorage.Provider
62 : // Set to DB.mu.
63 : mu *sync.Mutex
64 : opts *Options
65 : fs vfs.FS
66 : cmp *base.Comparer
67 : // Dynamic base level allows the dynamic base level computation to be
68 : // disabled. Used by tests which want to create specific LSM structures.
69 : dynamicBaseLevel bool
70 :
71 : // Mutable fields.
72 : versions versionList
73 : picker compactionPicker
74 :
75 : // Not all metrics are kept here. See DB.Metrics().
76 : metrics Metrics
77 :
78 : // A pointer to versionSet.addObsoleteLocked. Avoids allocating a new closure
79 : // on the creation of every version.
80 : obsoleteFn func(obsolete []*fileBacking)
81 : obsoleteTables []tableInfo
82 : obsoleteManifests []fileInfo
83 : obsoleteOptions []fileInfo
84 :
85 : // Zombie tables which have been removed from the current version but are
86 : // still referenced by an inuse iterator.
87 : zombieTables map[base.DiskFileNum]tableInfo
88 :
89 : // virtualBackings contains information about the FileBackings which support
90 : // virtual sstables in the latest version. It is mainly used to determine when
91 : // a backing is no longer in use by the tables in the latest version; this is
92 : // not a trivial problem because compactions and other operations can happen
93 : // in parallel (and they can finish in unpredictable order).
94 : //
95 : // This mechanism is complementary to the backing Ref/Unref mechanism, which
96 : // determines when a backing is no longer used by *any* live version and can
97 : // be removed.
98 : //
99 : // In principle this should have been a copy-on-write structure, with each
100 : // Version having its own record of its virtual backings (similar to the
101 : // B-tree that holds the tables). However, in practice we only need it for the
102 : // latest version, so we use a simpler structure and store it in the
103 : // versionSet instead.
104 : //
105 : // virtualBackings is modified under DB.mu and the log lock. If it is accessed
106 : // under DB.mu and a version update is in progress, it reflects the state of
107 : // the next version.
108 : virtualBackings manifest.VirtualBackings
109 :
110 : // minUnflushedLogNum is the smallest WAL log file number corresponding to
111 : // mutations that have not been flushed to an sstable.
112 : minUnflushedLogNum base.DiskFileNum
113 :
114 : // The next file number. A single counter is used to assign file
115 : // numbers for the WAL, MANIFEST, sstable, and OPTIONS files.
116 : nextFileNum uint64
117 :
118 : // The current manifest file number.
119 : manifestFileNum base.DiskFileNum
120 : manifestMarker *atomicfs.Marker
121 :
122 : manifestFile vfs.File
123 : manifest *record.Writer
124 : getFormatMajorVersion func() FormatMajorVersion
125 :
126 : writing bool
127 : writerCond sync.Cond
128 : // State for deciding when to write a snapshot. Protected by mu.
129 : rotationHelper record.RotationHelper
130 : }
131 :
132 : type tableInfo struct {
133 : fileInfo
134 : isLocal bool
135 : }
136 :
137 : func (vs *versionSet) init(
138 : dirname string,
139 : provider objstorage.Provider,
140 : opts *Options,
141 : marker *atomicfs.Marker,
142 : getFMV func() FormatMajorVersion,
143 : mu *sync.Mutex,
144 1 : ) {
145 1 : vs.dirname = dirname
146 1 : vs.provider = provider
147 1 : vs.mu = mu
148 1 : vs.writerCond.L = mu
149 1 : vs.opts = opts
150 1 : vs.fs = opts.FS
151 1 : vs.cmp = opts.Comparer
152 1 : vs.dynamicBaseLevel = true
153 1 : vs.versions.Init(mu)
154 1 : vs.obsoleteFn = vs.addObsoleteLocked
155 1 : vs.zombieTables = make(map[base.DiskFileNum]tableInfo)
156 1 : vs.virtualBackings = manifest.MakeVirtualBackings()
157 1 : vs.nextFileNum = 1
158 1 : vs.manifestMarker = marker
159 1 : vs.getFormatMajorVersion = getFMV
160 1 : }
161 :
162 : // create creates a version set for a fresh DB.
163 : func (vs *versionSet) create(
164 : jobID JobID,
165 : dirname string,
166 : provider objstorage.Provider,
167 : opts *Options,
168 : marker *atomicfs.Marker,
169 : getFormatMajorVersion func() FormatMajorVersion,
170 : mu *sync.Mutex,
171 1 : ) error {
172 1 : vs.init(dirname, provider, opts, marker, getFormatMajorVersion, mu)
173 1 : newVersion := &version{}
174 1 : vs.append(newVersion)
175 1 : var err error
176 1 :
177 1 : vs.picker = newCompactionPickerByScore(newVersion, &vs.virtualBackings, vs.opts, nil)
178 1 : // Note that a "snapshot" version edit is written to the manifest when it is
179 1 : // created.
180 1 : vs.manifestFileNum = vs.getNextDiskFileNum()
181 1 : err = vs.createManifest(vs.dirname, vs.manifestFileNum, vs.minUnflushedLogNum, vs.nextFileNum, nil /* virtualBackings */)
182 1 : if err == nil {
183 1 : if err = vs.manifest.Flush(); err != nil {
184 0 : vs.opts.Logger.Fatalf("MANIFEST flush failed: %v", err)
185 0 : }
186 : }
187 1 : if err == nil {
188 1 : if err = vs.manifestFile.Sync(); err != nil {
189 0 : vs.opts.Logger.Fatalf("MANIFEST sync failed: %v", err)
190 0 : }
191 : }
192 1 : if err == nil {
193 1 : // NB: Move() is responsible for syncing the data directory.
194 1 : if err = vs.manifestMarker.Move(base.MakeFilename(fileTypeManifest, vs.manifestFileNum)); err != nil {
195 0 : vs.opts.Logger.Fatalf("MANIFEST set current failed: %v", err)
196 0 : }
197 : }
198 :
199 1 : vs.opts.EventListener.ManifestCreated(ManifestCreateInfo{
200 1 : JobID: int(jobID),
201 1 : Path: base.MakeFilepath(vs.fs, vs.dirname, fileTypeManifest, vs.manifestFileNum),
202 1 : FileNum: vs.manifestFileNum,
203 1 : Err: err,
204 1 : })
205 1 : if err != nil {
206 0 : return err
207 0 : }
208 1 : return nil
209 : }
210 :
211 : // load loads the version set from the manifest file.
212 : func (vs *versionSet) load(
213 : dirname string,
214 : provider objstorage.Provider,
215 : opts *Options,
216 : manifestFileNum base.DiskFileNum,
217 : marker *atomicfs.Marker,
218 : getFormatMajorVersion func() FormatMajorVersion,
219 : mu *sync.Mutex,
220 1 : ) error {
221 1 : vs.init(dirname, provider, opts, marker, getFormatMajorVersion, mu)
222 1 :
223 1 : vs.manifestFileNum = manifestFileNum
224 1 : manifestPath := base.MakeFilepath(opts.FS, dirname, fileTypeManifest, vs.manifestFileNum)
225 1 : manifestFilename := opts.FS.PathBase(manifestPath)
226 1 :
227 1 : // Read the versionEdits in the manifest file.
228 1 : var bve bulkVersionEdit
229 1 : bve.AddedByFileNum = make(map[base.FileNum]*fileMetadata)
230 1 : manifest, err := vs.fs.Open(manifestPath)
231 1 : if err != nil {
232 0 : return errors.Wrapf(err, "pebble: could not open manifest file %q for DB %q",
233 0 : errors.Safe(manifestFilename), dirname)
234 0 : }
235 1 : defer manifest.Close()
236 1 : rr := record.NewReader(manifest, 0 /* logNum */)
237 1 : for {
238 1 : r, err := rr.Next()
239 1 : if err == io.EOF || record.IsInvalidRecord(err) {
240 1 : break
241 : }
242 1 : if err != nil {
243 0 : return errors.Wrapf(err, "pebble: error when loading manifest file %q",
244 0 : errors.Safe(manifestFilename))
245 0 : }
246 1 : var ve versionEdit
247 1 : err = ve.Decode(r)
248 1 : if err != nil {
249 0 : // Break instead of returning an error if the record is corrupted
250 0 : // or invalid.
251 0 : if err == io.EOF || record.IsInvalidRecord(err) {
252 0 : break
253 : }
254 0 : return err
255 : }
256 1 : if ve.ComparerName != "" {
257 1 : if ve.ComparerName != vs.cmp.Name {
258 0 : return errors.Errorf("pebble: manifest file %q for DB %q: "+
259 0 : "comparer name from file %q != comparer name from Options %q",
260 0 : errors.Safe(manifestFilename), dirname, errors.Safe(ve.ComparerName), errors.Safe(vs.cmp.Name))
261 0 : }
262 : }
263 1 : if err := bve.Accumulate(&ve); err != nil {
264 0 : return err
265 0 : }
266 1 : if ve.MinUnflushedLogNum != 0 {
267 1 : vs.minUnflushedLogNum = ve.MinUnflushedLogNum
268 1 : }
269 1 : if ve.NextFileNum != 0 {
270 1 : vs.nextFileNum = ve.NextFileNum
271 1 : }
272 1 : if ve.LastSeqNum != 0 {
273 1 : // logSeqNum is the _next_ sequence number that will be assigned,
274 1 : // while LastSeqNum is the last assigned sequence number. Note that
275 1 : // this behaviour mimics that in RocksDB; the first sequence number
276 1 : // assigned is one greater than the one present in the manifest
277 1 : // (assuming no WALs contain higher sequence numbers than the
278 1 : // manifest's LastSeqNum). Increment LastSeqNum by 1 to get the
279 1 : // next sequence number that will be assigned.
280 1 : //
281 1 : // If LastSeqNum is less than SeqNumStart, increase it to at least
282 1 : // SeqNumStart to leave ample room for reserved sequence numbers.
283 1 : if ve.LastSeqNum+1 < base.SeqNumStart {
284 0 : vs.logSeqNum.Store(base.SeqNumStart)
285 1 : } else {
286 1 : vs.logSeqNum.Store(ve.LastSeqNum + 1)
287 1 : }
288 : }
289 : }
290 : // We have already set vs.nextFileNum = 2 at the beginning of the
291 : // function and could have only updated it to some other non-zero value,
292 : // so it cannot be 0 here.
293 1 : if vs.minUnflushedLogNum == 0 {
294 0 : if vs.nextFileNum >= 2 {
295 0 : // We either have a freshly created DB, or a DB created by RocksDB
296 0 : // that has not had a single flushed SSTable yet. This is because
297 0 : // RocksDB bumps up nextFileNum in this case without bumping up
298 0 : // minUnflushedLogNum, even if WALs with non-zero file numbers are
299 0 : // present in the directory.
300 0 : } else {
301 0 : return base.CorruptionErrorf("pebble: malformed manifest file %q for DB %q",
302 0 : errors.Safe(manifestFilename), dirname)
303 0 : }
304 : }
305 1 : vs.markFileNumUsed(vs.minUnflushedLogNum)
306 1 :
307 1 : // Populate the fileBackingMap and the FileBacking for virtual sstables since
308 1 : // we have finished version edit accumulation.
309 1 : for _, b := range bve.AddedFileBacking {
310 1 : vs.virtualBackings.AddAndRef(b)
311 1 : }
312 :
313 1 : for _, addedLevel := range bve.Added {
314 1 : for _, m := range addedLevel {
315 1 : if m.Virtual {
316 1 : vs.virtualBackings.AddTable(m)
317 1 : }
318 : }
319 : }
320 :
321 1 : if invariants.Enabled {
322 1 : // There should be no deleted tables or backings, since we're starting from
323 1 : // an empty state.
324 1 : for _, deletedLevel := range bve.Deleted {
325 1 : if len(deletedLevel) != 0 {
326 0 : panic("deleted files after manifest replay")
327 : }
328 : }
329 1 : if len(bve.RemovedFileBacking) > 0 {
330 0 : panic("deleted backings after manifest replay")
331 : }
332 : }
333 :
334 1 : newVersion, err := bve.Apply(nil, opts.Comparer, opts.FlushSplitBytes, opts.Experimental.ReadCompactionRate)
335 1 : if err != nil {
336 0 : return err
337 0 : }
338 1 : newVersion.L0Sublevels.InitCompactingFileInfo(nil /* in-progress compactions */)
339 1 : vs.append(newVersion)
340 1 :
341 1 : for i := range vs.metrics.Levels {
342 1 : l := &vs.metrics.Levels[i]
343 1 : l.NumFiles = int64(newVersion.Levels[i].Len())
344 1 : files := newVersion.Levels[i].Slice()
345 1 : l.Size = int64(files.SizeSum())
346 1 : }
347 1 : for _, l := range newVersion.Levels {
348 1 : iter := l.Iter()
349 1 : for f := iter.First(); f != nil; f = iter.Next() {
350 1 : if !f.Virtual {
351 1 : _, localSize := sizeIfLocal(f.FileBacking, vs.provider)
352 1 : vs.metrics.Table.Local.LiveSize = uint64(int64(vs.metrics.Table.Local.LiveSize) + localSize)
353 1 : }
354 : }
355 : }
356 1 : vs.virtualBackings.ForEach(func(backing *fileBacking) {
357 1 : _, localSize := sizeIfLocal(backing, vs.provider)
358 1 : vs.metrics.Table.Local.LiveSize = uint64(int64(vs.metrics.Table.Local.LiveSize) + localSize)
359 1 : })
360 :
361 1 : vs.picker = newCompactionPickerByScore(newVersion, &vs.virtualBackings, vs.opts, nil)
362 1 : return nil
363 : }
364 :
365 1 : func (vs *versionSet) close() error {
366 1 : if vs.manifestFile != nil {
367 1 : if err := vs.manifestFile.Close(); err != nil {
368 0 : return err
369 0 : }
370 : }
371 1 : if vs.manifestMarker != nil {
372 1 : if err := vs.manifestMarker.Close(); err != nil {
373 0 : return err
374 0 : }
375 : }
376 1 : return nil
377 : }
378 :
379 : // logLock locks the manifest for writing. The lock must be released by either
380 : // a call to logUnlock or logAndApply.
381 : //
382 : // DB.mu must be held when calling this method, but the mutex may be dropped and
383 : // re-acquired during the course of this method.
384 1 : func (vs *versionSet) logLock() {
385 1 : // Wait for any existing writing to the manifest to complete, then mark the
386 1 : // manifest as busy.
387 1 : for vs.writing {
388 1 : // Note: writerCond.L is DB.mu, so we unlock it while we wait.
389 1 : vs.writerCond.Wait()
390 1 : }
391 1 : vs.writing = true
392 : }
393 :
394 : // logUnlock releases the lock for manifest writing.
395 : //
396 : // DB.mu must be held when calling this method.
397 1 : func (vs *versionSet) logUnlock() {
398 1 : if !vs.writing {
399 0 : vs.opts.Logger.Fatalf("MANIFEST not locked for writing")
400 0 : }
401 1 : vs.writing = false
402 1 : vs.writerCond.Signal()
403 : }
404 :
405 : // logAndApply logs the version edit to the manifest, applies the version edit
406 : // to the current version, and installs the new version.
407 : //
408 : // logAndApply fills in the following fields of the VersionEdit: NextFileNum,
409 : // LastSeqNum, RemovedBackingTables. The removed backing tables are those
410 : // backings that are no longer used (in the new version) after applying the edit
411 : // (as per vs.virtualBackings). Other than these fields, the VersionEdit must be
412 : // complete.
413 : //
414 : // New table backing references (FileBacking.Ref) are taken as part of applying
415 : // the version edit. The state of the virtual backings (vs.virtualBackings) is
416 : // updated before logging to the manifest and installing the latest version;
417 : // this is ok because any failure in those steps is fatal.
418 : // TODO(radu): remove the error return.
419 : //
420 : // DB.mu must be held when calling this method and will be released temporarily
421 : // while performing file I/O. Requires that the manifest is locked for writing
422 : // (see logLock). Will unconditionally release the manifest lock (via
423 : // logUnlock) even if an error occurs.
424 : //
425 : // inProgressCompactions is called while DB.mu is held, to get the list of
426 : // in-progress compactions.
427 : func (vs *versionSet) logAndApply(
428 : jobID JobID,
429 : ve *versionEdit,
430 : metrics map[int]*LevelMetrics,
431 : forceRotation bool,
432 : inProgressCompactions func() []compactionInfo,
433 1 : ) error {
434 1 : if !vs.writing {
435 0 : vs.opts.Logger.Fatalf("MANIFEST not locked for writing")
436 0 : }
437 1 : defer vs.logUnlock()
438 1 :
439 1 : if ve.MinUnflushedLogNum != 0 {
440 1 : if ve.MinUnflushedLogNum < vs.minUnflushedLogNum ||
441 1 : vs.nextFileNum <= uint64(ve.MinUnflushedLogNum) {
442 0 : panic(fmt.Sprintf("pebble: inconsistent versionEdit minUnflushedLogNum %d",
443 0 : ve.MinUnflushedLogNum))
444 : }
445 : }
446 :
447 : // This is the next manifest filenum, but if the current file is too big we
448 : // will write this ve to the next file which means what ve encodes is the
449 : // current filenum and not the next one.
450 : //
451 : // TODO(sbhola): figure out why this is correct and update comment.
452 1 : ve.NextFileNum = vs.nextFileNum
453 1 :
454 1 : // LastSeqNum is set to the current upper bound on the assigned sequence
455 1 : // numbers. Note that this is exactly the behavior of RocksDB. LastSeqNum is
456 1 : // used to initialize versionSet.logSeqNum and versionSet.visibleSeqNum on
457 1 : // replay. It must be higher than or equal to any than any sequence number
458 1 : // written to an sstable, including sequence numbers in ingested files.
459 1 : // Note that LastSeqNum is not (and cannot be) the minimum unflushed sequence
460 1 : // number. This is fallout from ingestion which allows a sequence number X to
461 1 : // be assigned to an ingested sstable even though sequence number X-1 resides
462 1 : // in an unflushed memtable. logSeqNum is the _next_ sequence number that
463 1 : // will be assigned, so subtract that by 1 to get the upper bound on the
464 1 : // last assigned sequence number.
465 1 : logSeqNum := vs.logSeqNum.Load()
466 1 : ve.LastSeqNum = logSeqNum - 1
467 1 : if logSeqNum == 0 {
468 0 : // logSeqNum is initialized to 1 in Open() if there are no previous WAL
469 0 : // or manifest records, so this case should never happen.
470 0 : vs.opts.Logger.Fatalf("logSeqNum must be a positive integer: %d", logSeqNum)
471 0 : }
472 :
473 1 : currentVersion := vs.currentVersion()
474 1 : var newVersion *version
475 1 :
476 1 : // Generate a new manifest if we don't currently have one, or forceRotation
477 1 : // is true, or the current one is too large.
478 1 : //
479 1 : // For largeness, we do not exclusively use MaxManifestFileSize size
480 1 : // threshold since we have had incidents where due to either large keys or
481 1 : // large numbers of files, each edit results in a snapshot + write of the
482 1 : // edit. This slows the system down since each flush or compaction is
483 1 : // writing a new manifest snapshot. The primary goal of the size-based
484 1 : // rollover logic is to ensure that when reopening a DB, the number of edits
485 1 : // that need to be replayed on top of the snapshot is "sane". Rolling over
486 1 : // to a new manifest after each edit is not relevant to that goal.
487 1 : //
488 1 : // Consider the following cases:
489 1 : // - The number of live files F in the DB is roughly stable: after writing
490 1 : // the snapshot (with F files), say we require that there be enough edits
491 1 : // such that the cumulative number of files in those edits, E, be greater
492 1 : // than F. This will ensure that the total amount of time in logAndApply
493 1 : // that is spent in snapshot writing is ~50%.
494 1 : //
495 1 : // - The number of live files F in the DB is shrinking drastically, say from
496 1 : // F to F/10: This can happen for various reasons, like wide range
497 1 : // tombstones, or large numbers of smaller than usual files that are being
498 1 : // merged together into larger files. And say the new files generated
499 1 : // during this shrinkage is insignificant compared to F/10, and so for
500 1 : // this example we will assume it is effectively 0. After this shrinking,
501 1 : // E = 0.9F, and so if we used the previous snapshot file count, F, as the
502 1 : // threshold that needs to be exceeded, we will further delay the snapshot
503 1 : // writing. Which means on DB reopen we will need to replay 0.9F edits to
504 1 : // get to a version with 0.1F files. It would be better to create a new
505 1 : // snapshot when E exceeds the number of files in the current version.
506 1 : //
507 1 : // - The number of live files F in the DB is growing via perfect ingests
508 1 : // into L6: Say we wrote the snapshot when there were F files and now we
509 1 : // have 10F files, so E = 9F. We will further delay writing a new
510 1 : // snapshot. This case can be critiqued as contrived, but we consider it
511 1 : // nonetheless.
512 1 : //
513 1 : // The logic below uses the min of the last snapshot file count and the file
514 1 : // count in the current version.
515 1 : vs.rotationHelper.AddRecord(int64(len(ve.DeletedFiles) + len(ve.NewFiles)))
516 1 : sizeExceeded := vs.manifest.Size() >= vs.opts.MaxManifestFileSize
517 1 : requireRotation := forceRotation || vs.manifest == nil
518 1 :
519 1 : var nextSnapshotFilecount int64
520 1 : for i := range vs.metrics.Levels {
521 1 : nextSnapshotFilecount += vs.metrics.Levels[i].NumFiles
522 1 : }
523 1 : if sizeExceeded && !requireRotation {
524 1 : requireRotation = vs.rotationHelper.ShouldRotate(nextSnapshotFilecount)
525 1 : }
526 1 : var newManifestFileNum base.DiskFileNum
527 1 : var prevManifestFileSize uint64
528 1 : var newManifestVirtualBackings []*fileBacking
529 1 : if requireRotation {
530 1 : newManifestFileNum = vs.getNextDiskFileNum()
531 1 : prevManifestFileSize = uint64(vs.manifest.Size())
532 1 :
533 1 : // We want the virtual backings *before* applying the version edit, because
534 1 : // the new manifest will contain the pre-apply version plus the last version
535 1 : // edit.
536 1 : newManifestVirtualBackings = vs.virtualBackings.Backings()
537 1 : }
538 :
539 : // Grab certain values before releasing vs.mu, in case createManifest() needs
540 : // to be called.
541 1 : minUnflushedLogNum := vs.minUnflushedLogNum
542 1 : nextFileNum := vs.nextFileNum
543 1 :
544 1 : // Note: this call populates ve.RemovedBackingTables.
545 1 : zombieBackings, removedVirtualBackings, localLiveSizeDelta :=
546 1 : getZombiesAndUpdateVirtualBackings(ve, &vs.virtualBackings, vs.provider)
547 1 :
548 1 : if err := func() error {
549 1 : vs.mu.Unlock()
550 1 : defer vs.mu.Lock()
551 1 :
552 1 : if vs.getFormatMajorVersion() < FormatVirtualSSTables && len(ve.CreatedBackingTables) > 0 {
553 0 : return base.AssertionFailedf("MANIFEST cannot contain virtual sstable records due to format major version")
554 0 : }
555 1 : var b bulkVersionEdit
556 1 : err := b.Accumulate(ve)
557 1 : if err != nil {
558 0 : return errors.Wrap(err, "MANIFEST accumulate failed")
559 0 : }
560 1 : newVersion, err = b.Apply(
561 1 : currentVersion, vs.cmp, vs.opts.FlushSplitBytes, vs.opts.Experimental.ReadCompactionRate,
562 1 : )
563 1 : if err != nil {
564 0 : return errors.Wrap(err, "MANIFEST apply failed")
565 0 : }
566 :
567 1 : if newManifestFileNum != 0 {
568 1 : if err := vs.createManifest(vs.dirname, newManifestFileNum, minUnflushedLogNum, nextFileNum, newManifestVirtualBackings); err != nil {
569 0 : vs.opts.EventListener.ManifestCreated(ManifestCreateInfo{
570 0 : JobID: int(jobID),
571 0 : Path: base.MakeFilepath(vs.fs, vs.dirname, fileTypeManifest, newManifestFileNum),
572 0 : FileNum: newManifestFileNum,
573 0 : Err: err,
574 0 : })
575 0 : return errors.Wrap(err, "MANIFEST create failed")
576 0 : }
577 : }
578 :
579 1 : w, err := vs.manifest.Next()
580 1 : if err != nil {
581 0 : return errors.Wrap(err, "MANIFEST next record write failed")
582 0 : }
583 :
584 : // NB: Any error from this point on is considered fatal as we don't know if
585 : // the MANIFEST write occurred or not. Trying to determine that is
586 : // fraught. Instead we rely on the standard recovery mechanism run when a
587 : // database is open. In particular, that mechanism generates a new MANIFEST
588 : // and ensures it is synced.
589 1 : if err := ve.Encode(w); err != nil {
590 0 : return errors.Wrap(err, "MANIFEST write failed")
591 0 : }
592 1 : if err := vs.manifest.Flush(); err != nil {
593 0 : return errors.Wrap(err, "MANIFEST flush failed")
594 0 : }
595 1 : if err := vs.manifestFile.Sync(); err != nil {
596 0 : return errors.Wrap(err, "MANIFEST sync failed")
597 0 : }
598 1 : if newManifestFileNum != 0 {
599 1 : // NB: Move() is responsible for syncing the data directory.
600 1 : if err := vs.manifestMarker.Move(base.MakeFilename(fileTypeManifest, newManifestFileNum)); err != nil {
601 0 : return errors.Wrap(err, "MANIFEST set current failed")
602 0 : }
603 1 : vs.opts.EventListener.ManifestCreated(ManifestCreateInfo{
604 1 : JobID: int(jobID),
605 1 : Path: base.MakeFilepath(vs.fs, vs.dirname, fileTypeManifest, newManifestFileNum),
606 1 : FileNum: newManifestFileNum,
607 1 : })
608 : }
609 1 : return nil
610 0 : }(); err != nil {
611 0 : // Any error encountered during any of the operations in the previous
612 0 : // closure are considered fatal. Treating such errors as fatal is preferred
613 0 : // to attempting to unwind various file and b-tree reference counts, and
614 0 : // re-generating L0 sublevel metadata. This may change in the future, if
615 0 : // certain manifest / WAL operations become retryable. For more context, see
616 0 : // #1159 and #1792.
617 0 : vs.opts.Logger.Fatalf("%s", err)
618 0 : return err
619 0 : }
620 :
621 1 : if requireRotation {
622 1 : // Successfully rotated.
623 1 : vs.rotationHelper.Rotate(nextSnapshotFilecount)
624 1 : }
625 : // Now that DB.mu is held again, initialize compacting file info in
626 : // L0Sublevels.
627 1 : inProgress := inProgressCompactions()
628 1 :
629 1 : newVersion.L0Sublevels.InitCompactingFileInfo(inProgressL0Compactions(inProgress))
630 1 :
631 1 : // Update the zombie tables set first, as installation of the new version
632 1 : // will unref the previous version which could result in addObsoleteLocked
633 1 : // being called.
634 1 : for _, b := range zombieBackings {
635 1 : vs.zombieTables[b.backing.DiskFileNum] = tableInfo{
636 1 : fileInfo: fileInfo{
637 1 : FileNum: b.backing.DiskFileNum,
638 1 : FileSize: b.backing.Size,
639 1 : },
640 1 : isLocal: b.isLocal,
641 1 : }
642 1 : }
643 :
644 : // Unref the removed backings and report those that already became obsolete.
645 : // Note that the only case where we report obsolete tables here is when
646 : // VirtualBackings.Protect/Unprotect was used to keep a backing alive without
647 : // it being used in the current version.
648 1 : var obsoleteVirtualBackings []*fileBacking
649 1 : for _, b := range removedVirtualBackings {
650 1 : if b.backing.Unref() == 0 {
651 1 : obsoleteVirtualBackings = append(obsoleteVirtualBackings, b.backing)
652 1 : }
653 : }
654 1 : vs.addObsoleteLocked(obsoleteVirtualBackings)
655 1 :
656 1 : // Install the new version.
657 1 : vs.append(newVersion)
658 1 :
659 1 : if ve.MinUnflushedLogNum != 0 {
660 1 : vs.minUnflushedLogNum = ve.MinUnflushedLogNum
661 1 : }
662 1 : if newManifestFileNum != 0 {
663 1 : if vs.manifestFileNum != 0 {
664 1 : vs.obsoleteManifests = append(vs.obsoleteManifests, fileInfo{
665 1 : FileNum: vs.manifestFileNum,
666 1 : FileSize: prevManifestFileSize,
667 1 : })
668 1 : }
669 1 : vs.manifestFileNum = newManifestFileNum
670 : }
671 :
672 1 : for level, update := range metrics {
673 1 : vs.metrics.Levels[level].Add(update)
674 1 : }
675 1 : for i := range vs.metrics.Levels {
676 1 : l := &vs.metrics.Levels[i]
677 1 : l.NumFiles = int64(newVersion.Levels[i].Len())
678 1 : l.NumVirtualFiles = newVersion.Levels[i].NumVirtual
679 1 : l.VirtualSize = newVersion.Levels[i].VirtualSize
680 1 : l.Size = int64(newVersion.Levels[i].Size())
681 1 :
682 1 : l.Sublevels = 0
683 1 : if l.NumFiles > 0 {
684 1 : l.Sublevels = 1
685 1 : }
686 1 : if invariants.Enabled {
687 1 : levelFiles := newVersion.Levels[i].Slice()
688 1 : if size := int64(levelFiles.SizeSum()); l.Size != size {
689 0 : vs.opts.Logger.Fatalf("versionSet metrics L%d Size = %d, actual size = %d", i, l.Size, size)
690 0 : }
691 1 : if nVirtual := levelFiles.NumVirtual(); nVirtual != l.NumVirtualFiles {
692 0 : vs.opts.Logger.Fatalf(
693 0 : "versionSet metrics L%d NumVirtual = %d, actual NumVirtual = %d",
694 0 : i, l.NumVirtualFiles, nVirtual,
695 0 : )
696 0 : }
697 1 : if vSize := levelFiles.VirtualSizeSum(); vSize != l.VirtualSize {
698 0 : vs.opts.Logger.Fatalf(
699 0 : "versionSet metrics L%d Virtual size = %d, actual size = %d",
700 0 : i, l.VirtualSize, vSize,
701 0 : )
702 0 : }
703 : }
704 : }
705 1 : vs.metrics.Levels[0].Sublevels = int32(len(newVersion.L0SublevelFiles))
706 1 : vs.metrics.Table.Local.LiveSize = uint64(int64(vs.metrics.Table.Local.LiveSize) + localLiveSizeDelta)
707 1 :
708 1 : vs.picker = newCompactionPickerByScore(newVersion, &vs.virtualBackings, vs.opts, inProgress)
709 1 : if !vs.dynamicBaseLevel {
710 0 : vs.picker.forceBaseLevel1()
711 0 : }
712 1 : return nil
713 : }
714 :
715 : type fileBackingInfo struct {
716 : backing *fileBacking
717 : isLocal bool
718 : }
719 :
720 : // getZombiesAndUpdateVirtualBackings updates the virtual backings with the
721 : // changes in the versionEdit and populates ve.RemovedBackingTables.
722 : // Returns:
723 : // - zombieBackings: all backings (physical and virtual) that will no longer be
724 : // needed when we apply ve.
725 : // - removedVirtualBackings: the virtual backings that will be removed by the
726 : // VersionEdit and which must be Unref()ed by the caller. These backings
727 : // match ve.RemovedBackingTables.
728 : // - localLiveSizeDelta: the delta in local live bytes.
729 : func getZombiesAndUpdateVirtualBackings(
730 : ve *versionEdit, virtualBackings *manifest.VirtualBackings, provider objstorage.Provider,
731 1 : ) (zombieBackings, removedVirtualBackings []fileBackingInfo, localLiveSizeDelta int64) {
732 1 : // First, deal with the physical tables.
733 1 : //
734 1 : // A physical backing has become unused if it is in DeletedFiles but not in
735 1 : // NewFiles or CreatedBackingTables.
736 1 : //
737 1 : // Note that for the common case where there are very few elements, the map
738 1 : // will stay on the stack.
739 1 : stillUsed := make(map[base.DiskFileNum]struct{})
740 1 : for _, nf := range ve.NewFiles {
741 1 : if !nf.Meta.Virtual {
742 1 : stillUsed[nf.Meta.FileBacking.DiskFileNum] = struct{}{}
743 1 : _, localFileDelta := sizeIfLocal(nf.Meta.FileBacking, provider)
744 1 : localLiveSizeDelta += localFileDelta
745 1 : }
746 : }
747 1 : for _, b := range ve.CreatedBackingTables {
748 1 : stillUsed[b.DiskFileNum] = struct{}{}
749 1 : }
750 1 : for _, m := range ve.DeletedFiles {
751 1 : if !m.Virtual {
752 1 : // NB: this deleted file may also be in NewFiles or
753 1 : // CreatedBackingTables, due to a file moving between levels, or
754 1 : // becoming virtualized. In which case there is no change due to this
755 1 : // file in the localLiveSizeDelta -- the subtraction below compensates
756 1 : // for the addition.
757 1 : isLocal, localFileDelta := sizeIfLocal(m.FileBacking, provider)
758 1 : localLiveSizeDelta -= localFileDelta
759 1 : if _, ok := stillUsed[m.FileBacking.DiskFileNum]; !ok {
760 1 : zombieBackings = append(zombieBackings, fileBackingInfo{
761 1 : backing: m.FileBacking,
762 1 : isLocal: isLocal,
763 1 : })
764 1 : }
765 : }
766 : }
767 :
768 : // Now deal with virtual tables.
769 : //
770 : // When a virtual table moves between levels we AddTable() then RemoveTable(),
771 : // which works out.
772 1 : for _, b := range ve.CreatedBackingTables {
773 1 : virtualBackings.AddAndRef(b)
774 1 : _, localFileDelta := sizeIfLocal(b, provider)
775 1 : localLiveSizeDelta += localFileDelta
776 1 : }
777 1 : for _, nf := range ve.NewFiles {
778 1 : if nf.Meta.Virtual {
779 1 : virtualBackings.AddTable(nf.Meta)
780 1 : }
781 : }
782 1 : for _, m := range ve.DeletedFiles {
783 1 : if m.Virtual {
784 1 : virtualBackings.RemoveTable(m)
785 1 : }
786 : }
787 :
788 1 : if unused := virtualBackings.Unused(); len(unused) > 0 {
789 1 : // Virtual backings that are no longer used are zombies and are also added
790 1 : // to RemovedBackingTables (before the version edit is written to disk).
791 1 : ve.RemovedBackingTables = make([]base.DiskFileNum, len(unused))
792 1 : for i, b := range unused {
793 1 : isLocal, localFileDelta := sizeIfLocal(b, provider)
794 1 : localLiveSizeDelta -= localFileDelta
795 1 : ve.RemovedBackingTables[i] = b.DiskFileNum
796 1 : zombieBackings = append(zombieBackings, fileBackingInfo{
797 1 : backing: b,
798 1 : isLocal: isLocal,
799 1 : })
800 1 : virtualBackings.Remove(b.DiskFileNum)
801 1 : }
802 1 : removedVirtualBackings = zombieBackings[len(zombieBackings)-len(unused):]
803 : }
804 1 : return zombieBackings, removedVirtualBackings, localLiveSizeDelta
805 : }
806 :
807 : // sizeIfLocal returns backing.Size if the backing is a local file, else 0.
808 : func sizeIfLocal(
809 : backing *fileBacking, provider objstorage.Provider,
810 1 : ) (isLocal bool, localSize int64) {
811 1 : isLocal = objstorage.IsLocalTable(provider, backing.DiskFileNum)
812 1 : if isLocal {
813 1 : return true, int64(backing.Size)
814 1 : }
815 1 : return false, 0
816 : }
817 :
818 : func (vs *versionSet) incrementCompactions(
819 : kind compactionKind, extraLevels []*compactionLevel, pickerMetrics compactionPickerMetrics,
820 1 : ) {
821 1 : switch kind {
822 1 : case compactionKindDefault:
823 1 : vs.metrics.Compact.Count++
824 1 : vs.metrics.Compact.DefaultCount++
825 :
826 1 : case compactionKindFlush, compactionKindIngestedFlushable:
827 1 : vs.metrics.Flush.Count++
828 :
829 1 : case compactionKindMove:
830 1 : vs.metrics.Compact.Count++
831 1 : vs.metrics.Compact.MoveCount++
832 :
833 1 : case compactionKindDeleteOnly:
834 1 : vs.metrics.Compact.Count++
835 1 : vs.metrics.Compact.DeleteOnlyCount++
836 :
837 1 : case compactionKindElisionOnly:
838 1 : vs.metrics.Compact.Count++
839 1 : vs.metrics.Compact.ElisionOnlyCount++
840 :
841 0 : case compactionKindRead:
842 0 : vs.metrics.Compact.Count++
843 0 : vs.metrics.Compact.ReadCount++
844 :
845 1 : case compactionKindRewrite:
846 1 : vs.metrics.Compact.Count++
847 1 : vs.metrics.Compact.RewriteCount++
848 : }
849 1 : if len(extraLevels) > 0 {
850 1 : vs.metrics.Compact.MultiLevelCount++
851 1 : }
852 : }
853 :
854 1 : func (vs *versionSet) incrementCompactionBytes(numBytes int64) {
855 1 : vs.atomicInProgressBytes.Add(numBytes)
856 1 : }
857 :
858 : // createManifest creates a manifest file that contains a snapshot of vs.
859 : func (vs *versionSet) createManifest(
860 : dirname string,
861 : fileNum, minUnflushedLogNum base.DiskFileNum,
862 : nextFileNum uint64,
863 : virtualBackings []*fileBacking,
864 1 : ) (err error) {
865 1 : var (
866 1 : filename = base.MakeFilepath(vs.fs, dirname, fileTypeManifest, fileNum)
867 1 : manifestFile vfs.File
868 1 : manifest *record.Writer
869 1 : )
870 1 : defer func() {
871 1 : if manifest != nil {
872 0 : manifest.Close()
873 0 : }
874 1 : if manifestFile != nil {
875 0 : manifestFile.Close()
876 0 : }
877 1 : if err != nil {
878 0 : vs.fs.Remove(filename)
879 0 : }
880 : }()
881 1 : manifestFile, err = vs.fs.Create(filename, "pebble-manifest")
882 1 : if err != nil {
883 0 : return err
884 0 : }
885 1 : manifest = record.NewWriter(manifestFile)
886 1 :
887 1 : snapshot := versionEdit{
888 1 : ComparerName: vs.cmp.Name,
889 1 : }
890 1 :
891 1 : for level, levelMetadata := range vs.currentVersion().Levels {
892 1 : iter := levelMetadata.Iter()
893 1 : for meta := iter.First(); meta != nil; meta = iter.Next() {
894 1 : snapshot.NewFiles = append(snapshot.NewFiles, newFileEntry{
895 1 : Level: level,
896 1 : Meta: meta,
897 1 : })
898 1 : }
899 : }
900 :
901 1 : snapshot.CreatedBackingTables = virtualBackings
902 1 :
903 1 : // When creating a version snapshot for an existing DB, this snapshot VersionEdit will be
904 1 : // immediately followed by another VersionEdit (being written in logAndApply()). That
905 1 : // VersionEdit always contains a LastSeqNum, so we don't need to include that in the snapshot.
906 1 : // But it does not necessarily include MinUnflushedLogNum, NextFileNum, so we initialize those
907 1 : // using the corresponding fields in the versionSet (which came from the latest preceding
908 1 : // VersionEdit that had those fields).
909 1 : snapshot.MinUnflushedLogNum = minUnflushedLogNum
910 1 : snapshot.NextFileNum = nextFileNum
911 1 :
912 1 : w, err1 := manifest.Next()
913 1 : if err1 != nil {
914 0 : return err1
915 0 : }
916 1 : if err := snapshot.Encode(w); err != nil {
917 0 : return err
918 0 : }
919 :
920 1 : if vs.manifest != nil {
921 1 : vs.manifest.Close()
922 1 : vs.manifest = nil
923 1 : }
924 1 : if vs.manifestFile != nil {
925 1 : if err := vs.manifestFile.Close(); err != nil {
926 0 : return err
927 0 : }
928 1 : vs.manifestFile = nil
929 : }
930 :
931 1 : vs.manifest, manifest = manifest, nil
932 1 : vs.manifestFile, manifestFile = manifestFile, nil
933 1 : return nil
934 : }
935 :
936 1 : func (vs *versionSet) markFileNumUsed(fileNum base.DiskFileNum) {
937 1 : if vs.nextFileNum <= uint64(fileNum) {
938 1 : vs.nextFileNum = uint64(fileNum + 1)
939 1 : }
940 : }
941 :
942 1 : func (vs *versionSet) getNextFileNum() base.FileNum {
943 1 : x := vs.nextFileNum
944 1 : vs.nextFileNum++
945 1 : return base.FileNum(x)
946 1 : }
947 :
948 1 : func (vs *versionSet) getNextDiskFileNum() base.DiskFileNum {
949 1 : x := vs.nextFileNum
950 1 : vs.nextFileNum++
951 1 : return base.DiskFileNum(x)
952 1 : }
953 :
954 1 : func (vs *versionSet) append(v *version) {
955 1 : if v.Refs() != 0 {
956 0 : panic("pebble: version should be unreferenced")
957 : }
958 1 : if !vs.versions.Empty() {
959 1 : vs.versions.Back().UnrefLocked()
960 1 : }
961 1 : v.Deleted = vs.obsoleteFn
962 1 : v.Ref()
963 1 : vs.versions.PushBack(v)
964 1 : if invariants.Enabled {
965 1 : // Verify that the virtualBackings contains all the backings referenced by
966 1 : // the version.
967 1 : for _, l := range v.Levels {
968 1 : iter := l.Iter()
969 1 : for f := iter.First(); f != nil; f = iter.Next() {
970 1 : if f.Virtual {
971 1 : if _, ok := vs.virtualBackings.Get(f.FileBacking.DiskFileNum); !ok {
972 0 : panic(fmt.Sprintf("%s is not in virtualBackings", f.FileBacking.DiskFileNum))
973 : }
974 : }
975 : }
976 : }
977 : }
978 : }
979 :
980 1 : func (vs *versionSet) currentVersion() *version {
981 1 : return vs.versions.Back()
982 1 : }
983 :
984 1 : func (vs *versionSet) addLiveFileNums(m map[base.DiskFileNum]struct{}) {
985 1 : current := vs.currentVersion()
986 1 : for v := vs.versions.Front(); true; v = v.Next() {
987 1 : for _, lm := range v.Levels {
988 1 : iter := lm.Iter()
989 1 : for f := iter.First(); f != nil; f = iter.Next() {
990 1 : m[f.FileBacking.DiskFileNum] = struct{}{}
991 1 : }
992 : }
993 1 : if v == current {
994 1 : break
995 : }
996 : }
997 : // virtualBackings contains backings that are referenced by some virtual
998 : // tables in the latest version (which are handled above), and backings that
999 : // are not but are still alive because of the protection mechanism (see
1000 : // manifset.VirtualBackings). This loop ensures the latter get added to the
1001 : // map.
1002 1 : vs.virtualBackings.ForEach(func(b *fileBacking) {
1003 1 : m[b.DiskFileNum] = struct{}{}
1004 1 : })
1005 : }
1006 :
1007 : // addObsoleteLocked will add the fileInfo associated with obsolete backing
1008 : // sstables to the obsolete tables list.
1009 : //
1010 : // The file backings in the obsolete list must not appear more than once.
1011 : //
1012 : // DB.mu must be held when addObsoleteLocked is called.
1013 1 : func (vs *versionSet) addObsoleteLocked(obsolete []*fileBacking) {
1014 1 : if len(obsolete) == 0 {
1015 1 : return
1016 1 : }
1017 :
1018 1 : obsoleteFileInfo := make([]tableInfo, len(obsolete))
1019 1 : for i, bs := range obsolete {
1020 1 : obsoleteFileInfo[i].FileNum = bs.DiskFileNum
1021 1 : obsoleteFileInfo[i].FileSize = bs.Size
1022 1 : }
1023 :
1024 1 : if invariants.Enabled {
1025 1 : dedup := make(map[base.DiskFileNum]struct{})
1026 1 : for _, fi := range obsoleteFileInfo {
1027 1 : dedup[fi.FileNum] = struct{}{}
1028 1 : }
1029 1 : if len(dedup) != len(obsoleteFileInfo) {
1030 0 : panic("pebble: duplicate FileBacking present in obsolete list")
1031 : }
1032 : }
1033 :
1034 1 : for i, fi := range obsoleteFileInfo {
1035 1 : // Note that the obsolete tables are no longer zombie by the definition of
1036 1 : // zombie, but we leave them in the zombie tables map until they are
1037 1 : // deleted from disk.
1038 1 : //
1039 1 : // TODO(sumeer): this means that the zombie metrics, like ZombieSize,
1040 1 : // computed in DB.Metrics are also being counted in the obsolete metrics.
1041 1 : // Was this intentional?
1042 1 : info, ok := vs.zombieTables[fi.FileNum]
1043 1 : if !ok {
1044 0 : vs.opts.Logger.Fatalf("MANIFEST obsolete table %s not marked as zombie", fi.FileNum)
1045 0 : }
1046 1 : obsoleteFileInfo[i].isLocal = info.isLocal
1047 : }
1048 :
1049 1 : vs.obsoleteTables = append(vs.obsoleteTables, obsoleteFileInfo...)
1050 1 : vs.updateObsoleteTableMetricsLocked()
1051 : }
1052 :
1053 : // addObsolete will acquire DB.mu, so DB.mu must not be held when this is
1054 : // called.
1055 1 : func (vs *versionSet) addObsolete(obsolete []*fileBacking) {
1056 1 : vs.mu.Lock()
1057 1 : defer vs.mu.Unlock()
1058 1 : vs.addObsoleteLocked(obsolete)
1059 1 : }
1060 :
1061 1 : func (vs *versionSet) updateObsoleteTableMetricsLocked() {
1062 1 : vs.metrics.Table.ObsoleteCount = int64(len(vs.obsoleteTables))
1063 1 : vs.metrics.Table.ObsoleteSize = 0
1064 1 : vs.metrics.Table.Local.ObsoleteSize = 0
1065 1 : for _, fi := range vs.obsoleteTables {
1066 1 : vs.metrics.Table.ObsoleteSize += fi.FileSize
1067 1 : if fi.isLocal {
1068 1 : vs.metrics.Table.Local.ObsoleteSize += fi.FileSize
1069 1 : }
1070 : }
1071 : }
1072 :
1073 : func findCurrentManifest(
1074 : fs vfs.FS, dirname string, ls []string,
1075 1 : ) (marker *atomicfs.Marker, manifestNum base.DiskFileNum, exists bool, err error) {
1076 1 : // Locating a marker should succeed even if the marker has never been placed.
1077 1 : var filename string
1078 1 : marker, filename, err = atomicfs.LocateMarkerInListing(fs, dirname, manifestMarkerName, ls)
1079 1 : if err != nil {
1080 0 : return nil, 0, false, err
1081 0 : }
1082 :
1083 1 : if filename == "" {
1084 1 : // The marker hasn't been set yet. This database doesn't exist.
1085 1 : return marker, 0, false, nil
1086 1 : }
1087 :
1088 1 : var ok bool
1089 1 : _, manifestNum, ok = base.ParseFilename(fs, filename)
1090 1 : if !ok {
1091 0 : return marker, 0, false, base.CorruptionErrorf("pebble: MANIFEST name %q is malformed", errors.Safe(filename))
1092 0 : }
1093 1 : return marker, manifestNum, true, nil
1094 : }
1095 :
1096 1 : func newFileMetrics(newFiles []manifest.NewFileEntry) map[int]*LevelMetrics {
1097 1 : m := map[int]*LevelMetrics{}
1098 1 : for _, nf := range newFiles {
1099 1 : lm := m[nf.Level]
1100 1 : if lm == nil {
1101 1 : lm = &LevelMetrics{}
1102 1 : m[nf.Level] = lm
1103 1 : }
1104 1 : lm.NumFiles++
1105 1 : lm.Size += int64(nf.Meta.Size)
1106 : }
1107 1 : return m
1108 : }
|
