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.DeletedTableEntry
30 : type tableMetadata = manifest.TableMetadata
31 : type physicalMeta = manifest.PhysicalTableMeta
32 : type virtualMeta = manifest.VirtualTableMeta
33 : type fileBacking = manifest.FileBacking
34 : type newTableEntry = manifest.NewTableEntry
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 base.AtomicSeqNum
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 base.AtomicSeqNum
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(manifest.ObsoleteFiles)
81 : obsoleteTables []objectInfo
82 : obsoleteBlobs []objectInfo
83 : obsoleteManifests []fileInfo
84 : obsoleteOptions []fileInfo
85 :
86 : // Zombie tables which have been removed from the current version but are
87 : // still referenced by an inuse iterator.
88 : zombieTables map[base.DiskFileNum]objectInfo
89 :
90 : // virtualBackings contains information about the FileBackings which support
91 : // virtual sstables in the latest version. It is mainly used to determine when
92 : // a backing is no longer in use by the tables in the latest version; this is
93 : // not a trivial problem because compactions and other operations can happen
94 : // in parallel (and they can finish in unpredictable order).
95 : //
96 : // This mechanism is complementary to the backing Ref/Unref mechanism, which
97 : // determines when a backing is no longer used by *any* live version and can
98 : // be removed.
99 : //
100 : // In principle this should have been a copy-on-write structure, with each
101 : // Version having its own record of its virtual backings (similar to the
102 : // B-tree that holds the tables). However, in practice we only need it for the
103 : // latest version, so we use a simpler structure and store it in the
104 : // versionSet instead.
105 : //
106 : // virtualBackings is modified under DB.mu and the log lock. If it is accessed
107 : // under DB.mu and a version update is in progress, it reflects the state of
108 : // the next version.
109 : virtualBackings manifest.VirtualBackings
110 :
111 : // minUnflushedLogNum is the smallest WAL log file number corresponding to
112 : // mutations that have not been flushed to an sstable.
113 : minUnflushedLogNum base.DiskFileNum
114 :
115 : // The next file number. A single counter is used to assign file
116 : // numbers for the WAL, MANIFEST, sstable, and OPTIONS files.
117 : nextFileNum atomic.Uint64
118 :
119 : // The current manifest file number.
120 : manifestFileNum base.DiskFileNum
121 : manifestMarker *atomicfs.Marker
122 :
123 : manifestFile vfs.File
124 : manifest *record.Writer
125 : getFormatMajorVersion func() FormatMajorVersion
126 :
127 : writing bool
128 : writerCond sync.Cond
129 : // State for deciding when to write a snapshot. Protected by mu.
130 : rotationHelper record.RotationHelper
131 : }
132 :
133 : // objectInfo describes an object in object storage (either a sstable or a blob
134 : // file).
135 : type objectInfo struct {
136 : fileInfo
137 : isLocal bool
138 : }
139 :
140 : func (vs *versionSet) init(
141 : dirname string,
142 : provider objstorage.Provider,
143 : opts *Options,
144 : marker *atomicfs.Marker,
145 : getFMV func() FormatMajorVersion,
146 : mu *sync.Mutex,
147 1 : ) {
148 1 : vs.dirname = dirname
149 1 : vs.provider = provider
150 1 : vs.mu = mu
151 1 : vs.writerCond.L = mu
152 1 : vs.opts = opts
153 1 : vs.fs = opts.FS
154 1 : vs.cmp = opts.Comparer
155 1 : vs.dynamicBaseLevel = true
156 1 : vs.versions.Init(mu)
157 1 : vs.obsoleteFn = vs.addObsoleteLocked
158 1 : vs.zombieTables = make(map[base.DiskFileNum]objectInfo)
159 1 : vs.virtualBackings = manifest.MakeVirtualBackings()
160 1 : vs.nextFileNum.Store(1)
161 1 : vs.manifestMarker = marker
162 1 : vs.getFormatMajorVersion = getFMV
163 1 : }
164 :
165 : // create creates a version set for a fresh DB.
166 : func (vs *versionSet) create(
167 : jobID JobID,
168 : dirname string,
169 : provider objstorage.Provider,
170 : opts *Options,
171 : marker *atomicfs.Marker,
172 : getFormatMajorVersion func() FormatMajorVersion,
173 : mu *sync.Mutex,
174 1 : ) error {
175 1 : vs.init(dirname, provider, opts, marker, getFormatMajorVersion, mu)
176 1 : var bve bulkVersionEdit
177 1 : newVersion, err := bve.Apply(nil /* curr */, opts.Comparer, opts.FlushSplitBytes, opts.Experimental.ReadCompactionRate)
178 1 : if err != nil {
179 0 : return err
180 0 : }
181 1 : vs.append(newVersion)
182 1 :
183 1 : vs.picker = newCompactionPickerByScore(newVersion, &vs.virtualBackings, vs.opts, nil)
184 1 : // Note that a "snapshot" version edit is written to the manifest when it is
185 1 : // created.
186 1 : vs.manifestFileNum = vs.getNextDiskFileNum()
187 1 : err = vs.createManifest(vs.dirname, vs.manifestFileNum, vs.minUnflushedLogNum, vs.nextFileNum.Load(), nil /* virtualBackings */)
188 1 : if err == nil {
189 1 : if err = vs.manifest.Flush(); err != nil {
190 0 : vs.opts.Logger.Fatalf("MANIFEST flush failed: %v", err)
191 0 : }
192 : }
193 1 : if err == nil {
194 1 : if err = vs.manifestFile.Sync(); err != nil {
195 0 : vs.opts.Logger.Fatalf("MANIFEST sync failed: %v", err)
196 0 : }
197 : }
198 1 : if err == nil {
199 1 : // NB: Move() is responsible for syncing the data directory.
200 1 : if err = vs.manifestMarker.Move(base.MakeFilename(base.FileTypeManifest, vs.manifestFileNum)); err != nil {
201 0 : vs.opts.Logger.Fatalf("MANIFEST set current failed: %v", err)
202 0 : }
203 : }
204 :
205 1 : vs.opts.EventListener.ManifestCreated(ManifestCreateInfo{
206 1 : JobID: int(jobID),
207 1 : Path: base.MakeFilepath(vs.fs, vs.dirname, base.FileTypeManifest, vs.manifestFileNum),
208 1 : FileNum: vs.manifestFileNum,
209 1 : Err: err,
210 1 : })
211 1 : if err != nil {
212 0 : return err
213 0 : }
214 1 : return nil
215 : }
216 :
217 : // load loads the version set from the manifest file.
218 : func (vs *versionSet) load(
219 : dirname string,
220 : provider objstorage.Provider,
221 : opts *Options,
222 : manifestFileNum base.DiskFileNum,
223 : marker *atomicfs.Marker,
224 : getFormatMajorVersion func() FormatMajorVersion,
225 : mu *sync.Mutex,
226 1 : ) error {
227 1 : vs.init(dirname, provider, opts, marker, getFormatMajorVersion, mu)
228 1 :
229 1 : vs.manifestFileNum = manifestFileNum
230 1 : manifestPath := base.MakeFilepath(opts.FS, dirname, base.FileTypeManifest, vs.manifestFileNum)
231 1 : manifestFilename := opts.FS.PathBase(manifestPath)
232 1 :
233 1 : // Read the versionEdits in the manifest file.
234 1 : var bve bulkVersionEdit
235 1 : bve.AddedTablesByFileNum = make(map[base.FileNum]*tableMetadata)
236 1 : manifest, err := vs.fs.Open(manifestPath)
237 1 : if err != nil {
238 0 : return errors.Wrapf(err, "pebble: could not open manifest file %q for DB %q",
239 0 : errors.Safe(manifestFilename), dirname)
240 0 : }
241 1 : defer manifest.Close()
242 1 : rr := record.NewReader(manifest, 0 /* logNum */)
243 1 : for {
244 1 : r, err := rr.Next()
245 1 : if err == io.EOF || record.IsInvalidRecord(err) {
246 1 : break
247 : }
248 1 : if err != nil {
249 0 : return errors.Wrapf(err, "pebble: error when loading manifest file %q",
250 0 : errors.Safe(manifestFilename))
251 0 : }
252 1 : var ve versionEdit
253 1 : err = ve.Decode(r)
254 1 : if err != nil {
255 0 : // Break instead of returning an error if the record is corrupted
256 0 : // or invalid.
257 0 : if err == io.EOF || record.IsInvalidRecord(err) {
258 0 : break
259 : }
260 0 : return err
261 : }
262 1 : if ve.ComparerName != "" {
263 1 : if ve.ComparerName != vs.cmp.Name {
264 0 : return errors.Errorf("pebble: manifest file %q for DB %q: "+
265 0 : "comparer name from file %q != comparer name from Options %q",
266 0 : errors.Safe(manifestFilename), dirname, errors.Safe(ve.ComparerName), errors.Safe(vs.cmp.Name))
267 0 : }
268 : }
269 1 : if err := bve.Accumulate(&ve); err != nil {
270 0 : return err
271 0 : }
272 1 : if ve.MinUnflushedLogNum != 0 {
273 1 : vs.minUnflushedLogNum = ve.MinUnflushedLogNum
274 1 : }
275 1 : if ve.NextFileNum != 0 {
276 1 : vs.nextFileNum.Store(ve.NextFileNum)
277 1 : }
278 1 : if ve.LastSeqNum != 0 {
279 1 : // logSeqNum is the _next_ sequence number that will be assigned,
280 1 : // while LastSeqNum is the last assigned sequence number. Note that
281 1 : // this behaviour mimics that in RocksDB; the first sequence number
282 1 : // assigned is one greater than the one present in the manifest
283 1 : // (assuming no WALs contain higher sequence numbers than the
284 1 : // manifest's LastSeqNum). Increment LastSeqNum by 1 to get the
285 1 : // next sequence number that will be assigned.
286 1 : //
287 1 : // If LastSeqNum is less than SeqNumStart, increase it to at least
288 1 : // SeqNumStart to leave ample room for reserved sequence numbers.
289 1 : if ve.LastSeqNum+1 < base.SeqNumStart {
290 0 : vs.logSeqNum.Store(base.SeqNumStart)
291 1 : } else {
292 1 : vs.logSeqNum.Store(ve.LastSeqNum + 1)
293 1 : }
294 : }
295 : }
296 : // We have already set vs.nextFileNum = 2 at the beginning of the
297 : // function and could have only updated it to some other non-zero value,
298 : // so it cannot be 0 here.
299 1 : if vs.minUnflushedLogNum == 0 {
300 1 : if vs.nextFileNum.Load() >= 2 {
301 1 : // We either have a freshly created DB, or a DB created by RocksDB
302 1 : // that has not had a single flushed SSTable yet. This is because
303 1 : // RocksDB bumps up nextFileNum in this case without bumping up
304 1 : // minUnflushedLogNum, even if WALs with non-zero file numbers are
305 1 : // present in the directory.
306 1 : } else {
307 0 : return base.CorruptionErrorf("pebble: malformed manifest file %q for DB %q",
308 0 : errors.Safe(manifestFilename), dirname)
309 0 : }
310 : }
311 1 : vs.markFileNumUsed(vs.minUnflushedLogNum)
312 1 :
313 1 : // Populate the fileBackingMap and the FileBacking for virtual sstables since
314 1 : // we have finished version edit accumulation.
315 1 : for _, b := range bve.AddedFileBacking {
316 1 : vs.virtualBackings.AddAndRef(b)
317 1 : }
318 :
319 1 : for _, addedLevel := range bve.AddedTables {
320 1 : for _, m := range addedLevel {
321 1 : if m.Virtual {
322 1 : vs.virtualBackings.AddTable(m)
323 1 : }
324 : }
325 : }
326 :
327 1 : if invariants.Enabled {
328 1 : // There should be no deleted tables or backings, since we're starting from
329 1 : // an empty state.
330 1 : for _, deletedLevel := range bve.DeletedTables {
331 1 : if len(deletedLevel) != 0 {
332 0 : panic("deleted files after manifest replay")
333 : }
334 : }
335 1 : if len(bve.RemovedFileBacking) > 0 {
336 0 : panic("deleted backings after manifest replay")
337 : }
338 : }
339 :
340 1 : newVersion, err := bve.Apply(nil, opts.Comparer, opts.FlushSplitBytes, opts.Experimental.ReadCompactionRate)
341 1 : if err != nil {
342 0 : return err
343 0 : }
344 1 : newVersion.L0Sublevels.InitCompactingFileInfo(nil /* in-progress compactions */)
345 1 : vs.append(newVersion)
346 1 :
347 1 : for i := range vs.metrics.Levels {
348 1 : l := &vs.metrics.Levels[i]
349 1 : l.NumFiles = int64(newVersion.Levels[i].Len())
350 1 : files := newVersion.Levels[i].Slice()
351 1 : l.Size = int64(files.SizeSum())
352 1 : }
353 1 : for _, l := range newVersion.Levels {
354 1 : iter := l.Iter()
355 1 : for f := iter.First(); f != nil; f = iter.Next() {
356 1 : if !f.Virtual {
357 1 : _, localSize := sizeIfLocal(f.FileBacking, vs.provider)
358 1 : vs.metrics.Table.Local.LiveSize = uint64(int64(vs.metrics.Table.Local.LiveSize) + localSize)
359 1 : }
360 : }
361 : }
362 1 : vs.virtualBackings.ForEach(func(backing *fileBacking) {
363 1 : _, localSize := sizeIfLocal(backing, vs.provider)
364 1 : vs.metrics.Table.Local.LiveSize = uint64(int64(vs.metrics.Table.Local.LiveSize) + localSize)
365 1 : })
366 :
367 1 : vs.picker = newCompactionPickerByScore(newVersion, &vs.virtualBackings, vs.opts, nil)
368 1 : return nil
369 : }
370 :
371 1 : func (vs *versionSet) close() error {
372 1 : if vs.manifestFile != nil {
373 1 : if err := vs.manifestFile.Close(); err != nil {
374 0 : return err
375 0 : }
376 : }
377 1 : if vs.manifestMarker != nil {
378 1 : if err := vs.manifestMarker.Close(); err != nil {
379 0 : return err
380 0 : }
381 : }
382 1 : return nil
383 : }
384 :
385 : // logLock locks the manifest for writing. The lock must be released by either
386 : // a call to logUnlock or logAndApply.
387 : //
388 : // DB.mu must be held when calling this method, but the mutex may be dropped and
389 : // re-acquired during the course of this method.
390 1 : func (vs *versionSet) logLock() {
391 1 : // Wait for any existing writing to the manifest to complete, then mark the
392 1 : // manifest as busy.
393 1 : for vs.writing {
394 1 : // Note: writerCond.L is DB.mu, so we unlock it while we wait.
395 1 : vs.writerCond.Wait()
396 1 : }
397 1 : vs.writing = true
398 : }
399 :
400 : // logUnlock releases the lock for manifest writing.
401 : //
402 : // DB.mu must be held when calling this method.
403 1 : func (vs *versionSet) logUnlock() {
404 1 : if !vs.writing {
405 0 : vs.opts.Logger.Fatalf("MANIFEST not locked for writing")
406 0 : }
407 1 : vs.writing = false
408 1 : vs.writerCond.Signal()
409 : }
410 :
411 : // logAndApply logs the version edit to the manifest, applies the version edit
412 : // to the current version, and installs the new version.
413 : //
414 : // logAndApply fills in the following fields of the VersionEdit: NextFileNum,
415 : // LastSeqNum, RemovedBackingTables. The removed backing tables are those
416 : // backings that are no longer used (in the new version) after applying the edit
417 : // (as per vs.virtualBackings). Other than these fields, the VersionEdit must be
418 : // complete.
419 : //
420 : // New table backing references (FileBacking.Ref) are taken as part of applying
421 : // the version edit. The state of the virtual backings (vs.virtualBackings) is
422 : // updated before logging to the manifest and installing the latest version;
423 : // this is ok because any failure in those steps is fatal.
424 : // TODO(radu): remove the error return.
425 : //
426 : // DB.mu must be held when calling this method and will be released temporarily
427 : // while performing file I/O. Requires that the manifest is locked for writing
428 : // (see logLock). Will unconditionally release the manifest lock (via
429 : // logUnlock) even if an error occurs.
430 : //
431 : // inProgressCompactions is called while DB.mu is held, to get the list of
432 : // in-progress compactions.
433 : func (vs *versionSet) logAndApply(
434 : jobID JobID,
435 : ve *versionEdit,
436 : metrics map[int]*LevelMetrics,
437 : forceRotation bool,
438 : inProgressCompactions func() []compactionInfo,
439 1 : ) error {
440 1 : if !vs.writing {
441 0 : vs.opts.Logger.Fatalf("MANIFEST not locked for writing")
442 0 : }
443 1 : defer vs.logUnlock()
444 1 :
445 1 : if ve.MinUnflushedLogNum != 0 {
446 1 : if ve.MinUnflushedLogNum < vs.minUnflushedLogNum ||
447 1 : vs.nextFileNum.Load() <= uint64(ve.MinUnflushedLogNum) {
448 0 : panic(fmt.Sprintf("pebble: inconsistent versionEdit minUnflushedLogNum %d",
449 0 : ve.MinUnflushedLogNum))
450 : }
451 : }
452 :
453 : // This is the next manifest filenum, but if the current file is too big we
454 : // will write this ve to the next file which means what ve encodes is the
455 : // current filenum and not the next one.
456 : //
457 : // TODO(sbhola): figure out why this is correct and update comment.
458 1 : ve.NextFileNum = vs.nextFileNum.Load()
459 1 :
460 1 : // LastSeqNum is set to the current upper bound on the assigned sequence
461 1 : // numbers. Note that this is exactly the behavior of RocksDB. LastSeqNum is
462 1 : // used to initialize versionSet.logSeqNum and versionSet.visibleSeqNum on
463 1 : // replay. It must be higher than or equal to any than any sequence number
464 1 : // written to an sstable, including sequence numbers in ingested files.
465 1 : // Note that LastSeqNum is not (and cannot be) the minimum unflushed sequence
466 1 : // number. This is fallout from ingestion which allows a sequence number X to
467 1 : // be assigned to an ingested sstable even though sequence number X-1 resides
468 1 : // in an unflushed memtable. logSeqNum is the _next_ sequence number that
469 1 : // will be assigned, so subtract that by 1 to get the upper bound on the
470 1 : // last assigned sequence number.
471 1 : logSeqNum := vs.logSeqNum.Load()
472 1 : ve.LastSeqNum = logSeqNum - 1
473 1 : if logSeqNum == 0 {
474 0 : // logSeqNum is initialized to 1 in Open() if there are no previous WAL
475 0 : // or manifest records, so this case should never happen.
476 0 : vs.opts.Logger.Fatalf("logSeqNum must be a positive integer: %d", logSeqNum)
477 0 : }
478 :
479 1 : currentVersion := vs.currentVersion()
480 1 : var newVersion *version
481 1 :
482 1 : // Generate a new manifest if we don't currently have one, or forceRotation
483 1 : // is true, or the current one is too large.
484 1 : //
485 1 : // For largeness, we do not exclusively use MaxManifestFileSize size
486 1 : // threshold since we have had incidents where due to either large keys or
487 1 : // large numbers of files, each edit results in a snapshot + write of the
488 1 : // edit. This slows the system down since each flush or compaction is
489 1 : // writing a new manifest snapshot. The primary goal of the size-based
490 1 : // rollover logic is to ensure that when reopening a DB, the number of edits
491 1 : // that need to be replayed on top of the snapshot is "sane". Rolling over
492 1 : // to a new manifest after each edit is not relevant to that goal.
493 1 : //
494 1 : // Consider the following cases:
495 1 : // - The number of live files F in the DB is roughly stable: after writing
496 1 : // the snapshot (with F files), say we require that there be enough edits
497 1 : // such that the cumulative number of files in those edits, E, be greater
498 1 : // than F. This will ensure that the total amount of time in logAndApply
499 1 : // that is spent in snapshot writing is ~50%.
500 1 : //
501 1 : // - The number of live files F in the DB is shrinking drastically, say from
502 1 : // F to F/10: This can happen for various reasons, like wide range
503 1 : // tombstones, or large numbers of smaller than usual files that are being
504 1 : // merged together into larger files. And say the new files generated
505 1 : // during this shrinkage is insignificant compared to F/10, and so for
506 1 : // this example we will assume it is effectively 0. After this shrinking,
507 1 : // E = 0.9F, and so if we used the previous snapshot file count, F, as the
508 1 : // threshold that needs to be exceeded, we will further delay the snapshot
509 1 : // writing. Which means on DB reopen we will need to replay 0.9F edits to
510 1 : // get to a version with 0.1F files. It would be better to create a new
511 1 : // snapshot when E exceeds the number of files in the current version.
512 1 : //
513 1 : // - The number of live files F in the DB is growing via perfect ingests
514 1 : // into L6: Say we wrote the snapshot when there were F files and now we
515 1 : // have 10F files, so E = 9F. We will further delay writing a new
516 1 : // snapshot. This case can be critiqued as contrived, but we consider it
517 1 : // nonetheless.
518 1 : //
519 1 : // The logic below uses the min of the last snapshot file count and the file
520 1 : // count in the current version.
521 1 : vs.rotationHelper.AddRecord(int64(len(ve.DeletedTables) + len(ve.NewTables)))
522 1 : sizeExceeded := vs.manifest.Size() >= vs.opts.MaxManifestFileSize
523 1 : requireRotation := forceRotation || vs.manifest == nil
524 1 :
525 1 : var nextSnapshotFilecount int64
526 1 : for i := range vs.metrics.Levels {
527 1 : nextSnapshotFilecount += vs.metrics.Levels[i].NumFiles
528 1 : }
529 1 : if sizeExceeded && !requireRotation {
530 1 : requireRotation = vs.rotationHelper.ShouldRotate(nextSnapshotFilecount)
531 1 : }
532 1 : var newManifestFileNum base.DiskFileNum
533 1 : var prevManifestFileSize uint64
534 1 : var newManifestVirtualBackings []*fileBacking
535 1 : if requireRotation {
536 1 : newManifestFileNum = vs.getNextDiskFileNum()
537 1 : prevManifestFileSize = uint64(vs.manifest.Size())
538 1 :
539 1 : // We want the virtual backings *before* applying the version edit, because
540 1 : // the new manifest will contain the pre-apply version plus the last version
541 1 : // edit.
542 1 : newManifestVirtualBackings = vs.virtualBackings.Backings()
543 1 : }
544 :
545 : // Grab certain values before releasing vs.mu, in case createManifest() needs
546 : // to be called.
547 1 : minUnflushedLogNum := vs.minUnflushedLogNum
548 1 : nextFileNum := vs.nextFileNum.Load()
549 1 :
550 1 : // Note: this call populates ve.RemovedBackingTables.
551 1 : zombieBackings, removedVirtualBackings, localLiveSizeDelta :=
552 1 : getZombiesAndUpdateVirtualBackings(ve, &vs.virtualBackings, vs.provider)
553 1 :
554 1 : if err := func() error {
555 1 : vs.mu.Unlock()
556 1 : defer vs.mu.Lock()
557 1 :
558 1 : if vs.getFormatMajorVersion() < FormatVirtualSSTables && len(ve.CreatedBackingTables) > 0 {
559 0 : return base.AssertionFailedf("MANIFEST cannot contain virtual sstable records due to format major version")
560 0 : }
561 1 : var b bulkVersionEdit
562 1 : err := b.Accumulate(ve)
563 1 : if err != nil {
564 0 : return errors.Wrap(err, "MANIFEST accumulate failed")
565 0 : }
566 1 : newVersion, err = b.Apply(
567 1 : currentVersion, vs.cmp, vs.opts.FlushSplitBytes, vs.opts.Experimental.ReadCompactionRate,
568 1 : )
569 1 : if err != nil {
570 0 : return errors.Wrap(err, "MANIFEST apply failed")
571 0 : }
572 :
573 1 : if newManifestFileNum != 0 {
574 1 : if err := vs.createManifest(vs.dirname, newManifestFileNum, minUnflushedLogNum, nextFileNum, newManifestVirtualBackings); err != nil {
575 0 : vs.opts.EventListener.ManifestCreated(ManifestCreateInfo{
576 0 : JobID: int(jobID),
577 0 : Path: base.MakeFilepath(vs.fs, vs.dirname, base.FileTypeManifest, newManifestFileNum),
578 0 : FileNum: newManifestFileNum,
579 0 : Err: err,
580 0 : })
581 0 : return errors.Wrap(err, "MANIFEST create failed")
582 0 : }
583 : }
584 :
585 1 : w, err := vs.manifest.Next()
586 1 : if err != nil {
587 0 : return errors.Wrap(err, "MANIFEST next record write failed")
588 0 : }
589 :
590 : // NB: Any error from this point on is considered fatal as we don't know if
591 : // the MANIFEST write occurred or not. Trying to determine that is
592 : // fraught. Instead we rely on the standard recovery mechanism run when a
593 : // database is open. In particular, that mechanism generates a new MANIFEST
594 : // and ensures it is synced.
595 1 : if err := ve.Encode(w); err != nil {
596 0 : return errors.Wrap(err, "MANIFEST write failed")
597 0 : }
598 1 : if err := vs.manifest.Flush(); err != nil {
599 0 : return errors.Wrap(err, "MANIFEST flush failed")
600 0 : }
601 1 : if err := vs.manifestFile.Sync(); err != nil {
602 0 : return errors.Wrap(err, "MANIFEST sync failed")
603 0 : }
604 1 : if newManifestFileNum != 0 {
605 1 : // NB: Move() is responsible for syncing the data directory.
606 1 : if err := vs.manifestMarker.Move(base.MakeFilename(base.FileTypeManifest, newManifestFileNum)); err != nil {
607 0 : return errors.Wrap(err, "MANIFEST set current failed")
608 0 : }
609 1 : vs.opts.EventListener.ManifestCreated(ManifestCreateInfo{
610 1 : JobID: int(jobID),
611 1 : Path: base.MakeFilepath(vs.fs, vs.dirname, base.FileTypeManifest, newManifestFileNum),
612 1 : FileNum: newManifestFileNum,
613 1 : })
614 : }
615 1 : return nil
616 0 : }(); err != nil {
617 0 : // Any error encountered during any of the operations in the previous
618 0 : // closure are considered fatal. Treating such errors as fatal is preferred
619 0 : // to attempting to unwind various file and b-tree reference counts, and
620 0 : // re-generating L0 sublevel metadata. This may change in the future, if
621 0 : // certain manifest / WAL operations become retryable. For more context, see
622 0 : // #1159 and #1792.
623 0 : vs.opts.Logger.Fatalf("%s", err)
624 0 : return err
625 0 : }
626 :
627 1 : if requireRotation {
628 1 : // Successfully rotated.
629 1 : vs.rotationHelper.Rotate(nextSnapshotFilecount)
630 1 : }
631 : // Now that DB.mu is held again, initialize compacting file info in
632 : // L0Sublevels.
633 1 : inProgress := inProgressCompactions()
634 1 :
635 1 : newVersion.L0Sublevels.InitCompactingFileInfo(inProgressL0Compactions(inProgress))
636 1 :
637 1 : // Update the zombie tables set first, as installation of the new version
638 1 : // will unref the previous version which could result in addObsoleteLocked
639 1 : // being called.
640 1 : for _, b := range zombieBackings {
641 1 : vs.zombieTables[b.backing.DiskFileNum] = objectInfo{
642 1 : fileInfo: fileInfo{
643 1 : FileNum: b.backing.DiskFileNum,
644 1 : FileSize: b.backing.Size,
645 1 : },
646 1 : isLocal: b.isLocal,
647 1 : }
648 1 : }
649 :
650 : // Unref the removed backings and report those that already became obsolete.
651 : // Note that the only case where we report obsolete tables here is when
652 : // VirtualBackings.Protect/Unprotect was used to keep a backing alive without
653 : // it being used in the current version.
654 1 : var obsoleteVirtualBackings manifest.ObsoleteFiles
655 1 : for _, b := range removedVirtualBackings {
656 1 : if b.backing.Unref() == 0 {
657 1 : obsoleteVirtualBackings.FileBackings = append(obsoleteVirtualBackings.FileBackings, b.backing)
658 1 : }
659 : }
660 1 : vs.addObsoleteLocked(obsoleteVirtualBackings)
661 1 :
662 1 : // Install the new version.
663 1 : vs.append(newVersion)
664 1 :
665 1 : if ve.MinUnflushedLogNum != 0 {
666 1 : vs.minUnflushedLogNum = ve.MinUnflushedLogNum
667 1 : }
668 1 : if newManifestFileNum != 0 {
669 1 : if vs.manifestFileNum != 0 {
670 1 : vs.obsoleteManifests = append(vs.obsoleteManifests, fileInfo{
671 1 : FileNum: vs.manifestFileNum,
672 1 : FileSize: prevManifestFileSize,
673 1 : })
674 1 : }
675 1 : vs.manifestFileNum = newManifestFileNum
676 : }
677 :
678 1 : for level, update := range metrics {
679 1 : vs.metrics.Levels[level].Add(update)
680 1 : }
681 1 : for i := range vs.metrics.Levels {
682 1 : l := &vs.metrics.Levels[i]
683 1 : l.NumFiles = int64(newVersion.Levels[i].Len())
684 1 : l.NumVirtualFiles = newVersion.Levels[i].NumVirtual
685 1 : l.VirtualSize = newVersion.Levels[i].VirtualSize
686 1 : l.Size = int64(newVersion.Levels[i].Size())
687 1 :
688 1 : l.Sublevels = 0
689 1 : if l.NumFiles > 0 {
690 1 : l.Sublevels = 1
691 1 : }
692 1 : if invariants.Enabled {
693 1 : levelFiles := newVersion.Levels[i].Slice()
694 1 : if size := int64(levelFiles.SizeSum()); l.Size != size {
695 0 : vs.opts.Logger.Fatalf("versionSet metrics L%d Size = %d, actual size = %d", i, l.Size, size)
696 0 : }
697 1 : if nVirtual := levelFiles.NumVirtual(); nVirtual != l.NumVirtualFiles {
698 0 : vs.opts.Logger.Fatalf(
699 0 : "versionSet metrics L%d NumVirtual = %d, actual NumVirtual = %d",
700 0 : i, l.NumVirtualFiles, nVirtual,
701 0 : )
702 0 : }
703 1 : if vSize := levelFiles.VirtualSizeSum(); vSize != l.VirtualSize {
704 0 : vs.opts.Logger.Fatalf(
705 0 : "versionSet metrics L%d Virtual size = %d, actual size = %d",
706 0 : i, l.VirtualSize, vSize,
707 0 : )
708 0 : }
709 : }
710 : }
711 1 : vs.metrics.Levels[0].Sublevels = int32(len(newVersion.L0SublevelFiles))
712 1 : vs.metrics.Table.Local.LiveSize = uint64(int64(vs.metrics.Table.Local.LiveSize) + localLiveSizeDelta)
713 1 :
714 1 : vs.picker = newCompactionPickerByScore(newVersion, &vs.virtualBackings, vs.opts, inProgress)
715 1 : if !vs.dynamicBaseLevel {
716 0 : vs.picker.forceBaseLevel1()
717 0 : }
718 1 : return nil
719 : }
720 :
721 : type fileBackingInfo struct {
722 : backing *fileBacking
723 : isLocal bool
724 : }
725 :
726 : // getZombiesAndUpdateVirtualBackings updates the virtual backings with the
727 : // changes in the versionEdit and populates ve.RemovedBackingTables.
728 : // Returns:
729 : // - zombieBackings: all backings (physical and virtual) that will no longer be
730 : // needed when we apply ve.
731 : // - removedVirtualBackings: the virtual backings that will be removed by the
732 : // VersionEdit and which must be Unref()ed by the caller. These backings
733 : // match ve.RemovedBackingTables.
734 : // - localLiveSizeDelta: the delta in local live bytes.
735 : func getZombiesAndUpdateVirtualBackings(
736 : ve *versionEdit, virtualBackings *manifest.VirtualBackings, provider objstorage.Provider,
737 1 : ) (zombieBackings, removedVirtualBackings []fileBackingInfo, localLiveSizeDelta int64) {
738 1 : // First, deal with the physical tables.
739 1 : //
740 1 : // A physical backing has become unused if it is in DeletedFiles but not in
741 1 : // NewFiles or CreatedBackingTables.
742 1 : //
743 1 : // Note that for the common case where there are very few elements, the map
744 1 : // will stay on the stack.
745 1 : stillUsed := make(map[base.DiskFileNum]struct{})
746 1 : for _, nf := range ve.NewTables {
747 1 : if !nf.Meta.Virtual {
748 1 : stillUsed[nf.Meta.FileBacking.DiskFileNum] = struct{}{}
749 1 : _, localFileDelta := sizeIfLocal(nf.Meta.FileBacking, provider)
750 1 : localLiveSizeDelta += localFileDelta
751 1 : }
752 : }
753 1 : for _, b := range ve.CreatedBackingTables {
754 1 : stillUsed[b.DiskFileNum] = struct{}{}
755 1 : }
756 1 : for _, m := range ve.DeletedTables {
757 1 : if !m.Virtual {
758 1 : // NB: this deleted file may also be in NewFiles or
759 1 : // CreatedBackingTables, due to a file moving between levels, or
760 1 : // becoming virtualized. In which case there is no change due to this
761 1 : // file in the localLiveSizeDelta -- the subtraction below compensates
762 1 : // for the addition.
763 1 : isLocal, localFileDelta := sizeIfLocal(m.FileBacking, provider)
764 1 : localLiveSizeDelta -= localFileDelta
765 1 : if _, ok := stillUsed[m.FileBacking.DiskFileNum]; !ok {
766 1 : zombieBackings = append(zombieBackings, fileBackingInfo{
767 1 : backing: m.FileBacking,
768 1 : isLocal: isLocal,
769 1 : })
770 1 : }
771 : }
772 : }
773 :
774 : // Now deal with virtual tables.
775 : //
776 : // When a virtual table moves between levels we AddTable() then RemoveTable(),
777 : // which works out.
778 1 : for _, b := range ve.CreatedBackingTables {
779 1 : virtualBackings.AddAndRef(b)
780 1 : _, localFileDelta := sizeIfLocal(b, provider)
781 1 : localLiveSizeDelta += localFileDelta
782 1 : }
783 1 : for _, nf := range ve.NewTables {
784 1 : if nf.Meta.Virtual {
785 1 : virtualBackings.AddTable(nf.Meta)
786 1 : }
787 : }
788 1 : for _, m := range ve.DeletedTables {
789 1 : if m.Virtual {
790 1 : virtualBackings.RemoveTable(m)
791 1 : }
792 : }
793 :
794 1 : if unused := virtualBackings.Unused(); len(unused) > 0 {
795 1 : // Virtual backings that are no longer used are zombies and are also added
796 1 : // to RemovedBackingTables (before the version edit is written to disk).
797 1 : ve.RemovedBackingTables = make([]base.DiskFileNum, len(unused))
798 1 : for i, b := range unused {
799 1 : isLocal, localFileDelta := sizeIfLocal(b, provider)
800 1 : localLiveSizeDelta -= localFileDelta
801 1 : ve.RemovedBackingTables[i] = b.DiskFileNum
802 1 : zombieBackings = append(zombieBackings, fileBackingInfo{
803 1 : backing: b,
804 1 : isLocal: isLocal,
805 1 : })
806 1 : virtualBackings.Remove(b.DiskFileNum)
807 1 : }
808 1 : removedVirtualBackings = zombieBackings[len(zombieBackings)-len(unused):]
809 : }
810 1 : return zombieBackings, removedVirtualBackings, localLiveSizeDelta
811 : }
812 :
813 : // sizeIfLocal returns backing.Size if the backing is a local file, else 0.
814 : func sizeIfLocal(
815 : backing *fileBacking, provider objstorage.Provider,
816 1 : ) (isLocal bool, localSize int64) {
817 1 : isLocal = objstorage.IsLocalTable(provider, backing.DiskFileNum)
818 1 : if isLocal {
819 1 : return true, int64(backing.Size)
820 1 : }
821 1 : return false, 0
822 : }
823 :
824 : func (vs *versionSet) incrementCompactions(
825 : kind compactionKind, extraLevels []*compactionLevel, pickerMetrics compactionPickerMetrics,
826 1 : ) {
827 1 : switch kind {
828 1 : case compactionKindDefault:
829 1 : vs.metrics.Compact.Count++
830 1 : vs.metrics.Compact.DefaultCount++
831 :
832 1 : case compactionKindFlush, compactionKindIngestedFlushable:
833 1 : vs.metrics.Flush.Count++
834 :
835 1 : case compactionKindMove:
836 1 : vs.metrics.Compact.Count++
837 1 : vs.metrics.Compact.MoveCount++
838 :
839 1 : case compactionKindDeleteOnly:
840 1 : vs.metrics.Compact.Count++
841 1 : vs.metrics.Compact.DeleteOnlyCount++
842 :
843 1 : case compactionKindElisionOnly:
844 1 : vs.metrics.Compact.Count++
845 1 : vs.metrics.Compact.ElisionOnlyCount++
846 :
847 0 : case compactionKindRead:
848 0 : vs.metrics.Compact.Count++
849 0 : vs.metrics.Compact.ReadCount++
850 :
851 1 : case compactionKindTombstoneDensity:
852 1 : vs.metrics.Compact.Count++
853 1 : vs.metrics.Compact.TombstoneDensityCount++
854 :
855 1 : case compactionKindRewrite:
856 1 : vs.metrics.Compact.Count++
857 1 : vs.metrics.Compact.RewriteCount++
858 :
859 1 : case compactionKindCopy:
860 1 : vs.metrics.Compact.Count++
861 1 : vs.metrics.Compact.CopyCount++
862 :
863 0 : default:
864 0 : if invariants.Enabled {
865 0 : panic("unhandled compaction kind")
866 : }
867 : }
868 1 : if len(extraLevels) > 0 {
869 1 : vs.metrics.Compact.MultiLevelCount++
870 1 : }
871 : }
872 :
873 1 : func (vs *versionSet) incrementCompactionBytes(numBytes int64) {
874 1 : vs.atomicInProgressBytes.Add(numBytes)
875 1 : }
876 :
877 : // createManifest creates a manifest file that contains a snapshot of vs.
878 : func (vs *versionSet) createManifest(
879 : dirname string,
880 : fileNum, minUnflushedLogNum base.DiskFileNum,
881 : nextFileNum uint64,
882 : virtualBackings []*fileBacking,
883 1 : ) (err error) {
884 1 : var (
885 1 : filename = base.MakeFilepath(vs.fs, dirname, base.FileTypeManifest, fileNum)
886 1 : manifestFile vfs.File
887 1 : manifest *record.Writer
888 1 : )
889 1 : defer func() {
890 1 : if manifest != nil {
891 0 : manifest.Close()
892 0 : }
893 1 : if manifestFile != nil {
894 0 : manifestFile.Close()
895 0 : }
896 1 : if err != nil {
897 0 : vs.fs.Remove(filename)
898 0 : }
899 : }()
900 1 : manifestFile, err = vs.fs.Create(filename, "pebble-manifest")
901 1 : if err != nil {
902 0 : return err
903 0 : }
904 1 : manifest = record.NewWriter(manifestFile)
905 1 :
906 1 : snapshot := versionEdit{
907 1 : ComparerName: vs.cmp.Name,
908 1 : }
909 1 :
910 1 : for level, levelMetadata := range vs.currentVersion().Levels {
911 1 : iter := levelMetadata.Iter()
912 1 : for meta := iter.First(); meta != nil; meta = iter.Next() {
913 1 : snapshot.NewTables = append(snapshot.NewTables, newTableEntry{
914 1 : Level: level,
915 1 : Meta: meta,
916 1 : })
917 1 : }
918 : }
919 :
920 1 : snapshot.CreatedBackingTables = virtualBackings
921 1 :
922 1 : // When creating a version snapshot for an existing DB, this snapshot VersionEdit will be
923 1 : // immediately followed by another VersionEdit (being written in logAndApply()). That
924 1 : // VersionEdit always contains a LastSeqNum, so we don't need to include that in the snapshot.
925 1 : // But it does not necessarily include MinUnflushedLogNum, NextFileNum, so we initialize those
926 1 : // using the corresponding fields in the versionSet (which came from the latest preceding
927 1 : // VersionEdit that had those fields).
928 1 : snapshot.MinUnflushedLogNum = minUnflushedLogNum
929 1 : snapshot.NextFileNum = nextFileNum
930 1 :
931 1 : w, err1 := manifest.Next()
932 1 : if err1 != nil {
933 0 : return err1
934 0 : }
935 1 : if err := snapshot.Encode(w); err != nil {
936 0 : return err
937 0 : }
938 :
939 1 : if vs.manifest != nil {
940 1 : vs.manifest.Close()
941 1 : vs.manifest = nil
942 1 : }
943 1 : if vs.manifestFile != nil {
944 1 : if err := vs.manifestFile.Close(); err != nil {
945 0 : return err
946 0 : }
947 1 : vs.manifestFile = nil
948 : }
949 :
950 1 : vs.manifest, manifest = manifest, nil
951 1 : vs.manifestFile, manifestFile = manifestFile, nil
952 1 : return nil
953 : }
954 :
955 : // NB: This method is not safe for concurrent use. It is only safe
956 : // to be called when concurrent changes to nextFileNum are not expected.
957 1 : func (vs *versionSet) markFileNumUsed(fileNum base.DiskFileNum) {
958 1 : if vs.nextFileNum.Load() <= uint64(fileNum) {
959 1 : vs.nextFileNum.Store(uint64(fileNum + 1))
960 1 : }
961 : }
962 :
963 : // getNextFileNum returns the next file number to be used.
964 : //
965 : // Can be called without the versionSet's mutex being held.
966 1 : func (vs *versionSet) getNextFileNum() base.FileNum {
967 1 : x := vs.nextFileNum.Add(1) - 1
968 1 : return base.FileNum(x)
969 1 : }
970 :
971 : // Can be called without the versionSet's mutex being held.
972 1 : func (vs *versionSet) getNextDiskFileNum() base.DiskFileNum {
973 1 : x := vs.nextFileNum.Add(1) - 1
974 1 : return base.DiskFileNum(x)
975 1 : }
976 :
977 1 : func (vs *versionSet) append(v *version) {
978 1 : if v.Refs() != 0 {
979 0 : panic("pebble: version should be unreferenced")
980 : }
981 1 : if !vs.versions.Empty() {
982 1 : vs.versions.Back().UnrefLocked()
983 1 : }
984 1 : v.Deleted = vs.obsoleteFn
985 1 : v.Ref()
986 1 : vs.versions.PushBack(v)
987 1 : if invariants.Enabled {
988 1 : // Verify that the virtualBackings contains all the backings referenced by
989 1 : // the version.
990 1 : for _, l := range v.Levels {
991 1 : iter := l.Iter()
992 1 : for f := iter.First(); f != nil; f = iter.Next() {
993 1 : if f.Virtual {
994 1 : if _, ok := vs.virtualBackings.Get(f.FileBacking.DiskFileNum); !ok {
995 0 : panic(fmt.Sprintf("%s is not in virtualBackings", f.FileBacking.DiskFileNum))
996 : }
997 : }
998 : }
999 : }
1000 : }
1001 : }
1002 :
1003 1 : func (vs *versionSet) currentVersion() *version {
1004 1 : return vs.versions.Back()
1005 1 : }
1006 :
1007 1 : func (vs *versionSet) addLiveFileNums(m map[base.DiskFileNum]struct{}) {
1008 1 : current := vs.currentVersion()
1009 1 : for v := vs.versions.Front(); true; v = v.Next() {
1010 1 : for _, lm := range v.Levels {
1011 1 : iter := lm.Iter()
1012 1 : for f := iter.First(); f != nil; f = iter.Next() {
1013 1 : m[f.FileBacking.DiskFileNum] = struct{}{}
1014 1 : }
1015 : }
1016 1 : if v == current {
1017 1 : break
1018 : }
1019 : }
1020 : // virtualBackings contains backings that are referenced by some virtual
1021 : // tables in the latest version (which are handled above), and backings that
1022 : // are not but are still alive because of the protection mechanism (see
1023 : // manifset.VirtualBackings). This loop ensures the latter get added to the
1024 : // map.
1025 1 : vs.virtualBackings.ForEach(func(b *fileBacking) {
1026 1 : m[b.DiskFileNum] = struct{}{}
1027 1 : })
1028 : }
1029 :
1030 : // addObsoleteLocked will add the fileInfo associated with obsolete backing
1031 : // sstables to the obsolete tables list.
1032 : //
1033 : // The file backings in the obsolete list must not appear more than once.
1034 : //
1035 : // DB.mu must be held when addObsoleteLocked is called.
1036 1 : func (vs *versionSet) addObsoleteLocked(obsolete manifest.ObsoleteFiles) {
1037 1 : if obsolete.Count() == 0 {
1038 1 : return
1039 1 : }
1040 :
1041 1 : obsoleteFileInfo := make([]objectInfo, len(obsolete.FileBackings))
1042 1 : for i, bs := range obsolete.FileBackings {
1043 1 : obsoleteFileInfo[i].FileNum = bs.DiskFileNum
1044 1 : obsoleteFileInfo[i].FileSize = bs.Size
1045 1 : }
1046 :
1047 1 : if invariants.Enabled {
1048 1 : dedup := make(map[base.DiskFileNum]struct{})
1049 1 : for _, fi := range obsoleteFileInfo {
1050 1 : dedup[fi.FileNum] = struct{}{}
1051 1 : }
1052 1 : if len(dedup) != len(obsoleteFileInfo) {
1053 0 : panic("pebble: duplicate FileBacking present in obsolete list")
1054 : }
1055 : }
1056 :
1057 1 : for i, fi := range obsoleteFileInfo {
1058 1 : // Note that the obsolete tables are no longer zombie by the definition of
1059 1 : // zombie, but we leave them in the zombie tables map until they are
1060 1 : // deleted from disk.
1061 1 : //
1062 1 : // TODO(sumeer): this means that the zombie metrics, like ZombieSize,
1063 1 : // computed in DB.Metrics are also being counted in the obsolete metrics.
1064 1 : // Was this intentional?
1065 1 : info, ok := vs.zombieTables[fi.FileNum]
1066 1 : if !ok {
1067 0 : vs.opts.Logger.Fatalf("MANIFEST obsolete table %s not marked as zombie", fi.FileNum)
1068 0 : }
1069 1 : obsoleteFileInfo[i].isLocal = info.isLocal
1070 : }
1071 :
1072 1 : vs.obsoleteTables = append(vs.obsoleteTables, obsoleteFileInfo...)
1073 1 : vs.updateObsoleteTableMetricsLocked()
1074 : }
1075 :
1076 : // addObsolete will acquire DB.mu, so DB.mu must not be held when this is
1077 : // called.
1078 1 : func (vs *versionSet) addObsolete(obsolete manifest.ObsoleteFiles) {
1079 1 : vs.mu.Lock()
1080 1 : defer vs.mu.Unlock()
1081 1 : vs.addObsoleteLocked(obsolete)
1082 1 : }
1083 :
1084 1 : func (vs *versionSet) updateObsoleteTableMetricsLocked() {
1085 1 : vs.metrics.Table.ObsoleteCount = int64(len(vs.obsoleteTables))
1086 1 : vs.metrics.Table.ObsoleteSize = 0
1087 1 : vs.metrics.Table.Local.ObsoleteSize = 0
1088 1 : for _, fi := range vs.obsoleteTables {
1089 1 : vs.metrics.Table.ObsoleteSize += fi.FileSize
1090 1 : if fi.isLocal {
1091 1 : vs.metrics.Table.Local.ObsoleteSize += fi.FileSize
1092 1 : }
1093 : }
1094 : }
1095 :
1096 : func findCurrentManifest(
1097 : fs vfs.FS, dirname string, ls []string,
1098 1 : ) (marker *atomicfs.Marker, manifestNum base.DiskFileNum, exists bool, err error) {
1099 1 : // Locating a marker should succeed even if the marker has never been placed.
1100 1 : var filename string
1101 1 : marker, filename, err = atomicfs.LocateMarkerInListing(fs, dirname, manifestMarkerName, ls)
1102 1 : if err != nil {
1103 0 : return nil, 0, false, err
1104 0 : }
1105 :
1106 1 : if filename == "" {
1107 1 : // The marker hasn't been set yet. This database doesn't exist.
1108 1 : return marker, 0, false, nil
1109 1 : }
1110 :
1111 1 : var ok bool
1112 1 : _, manifestNum, ok = base.ParseFilename(fs, filename)
1113 1 : if !ok {
1114 0 : return marker, 0, false, base.CorruptionErrorf("pebble: MANIFEST name %q is malformed", errors.Safe(filename))
1115 0 : }
1116 1 : return marker, manifestNum, true, nil
1117 : }
1118 :
1119 0 : func newFileMetrics(newFiles []manifest.NewTableEntry) map[int]*LevelMetrics {
1120 0 : m := map[int]*LevelMetrics{}
1121 0 : for _, nf := range newFiles {
1122 0 : lm := m[nf.Level]
1123 0 : if lm == nil {
1124 0 : lm = &LevelMetrics{}
1125 0 : m[nf.Level] = lm
1126 0 : }
1127 0 : lm.NumFiles++
1128 0 : lm.Size += int64(nf.Meta.Size)
1129 : }
1130 0 : return m
1131 : }
|
