Line data Source code
1 : // Copyright 2020 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 : "context"
9 : "fmt"
10 : "math"
11 : "slices"
12 : "time"
13 :
14 : "github.com/cockroachdb/crlib/crtime"
15 : "github.com/cockroachdb/errors"
16 : "github.com/cockroachdb/pebble/internal/base"
17 : "github.com/cockroachdb/pebble/internal/invariants"
18 : "github.com/cockroachdb/pebble/internal/keyspan"
19 : "github.com/cockroachdb/pebble/internal/keyspan/keyspanimpl"
20 : "github.com/cockroachdb/pebble/internal/manifest"
21 : "github.com/cockroachdb/pebble/sstable"
22 : "github.com/cockroachdb/pebble/sstable/block"
23 : "github.com/cockroachdb/redact"
24 : )
25 :
26 : // In-memory statistics about tables help inform compaction picking, but may
27 : // be expensive to calculate or load from disk. Every time a database is
28 : // opened, these statistics must be reloaded or recalculated. To minimize
29 : // impact on user activity and compactions, we load these statistics
30 : // asynchronously in the background and store loaded statistics in each
31 : // table's *TableMetadata.
32 : //
33 : // This file implements the asynchronous loading of statistics by maintaining
34 : // a list of files that require statistics, alongside their LSM levels.
35 : // Whenever new files are added to the LSM, the files are appended to
36 : // d.mu.tableStats.pending. If a stats collection job is not currently
37 : // running, one is started in a separate goroutine.
38 : //
39 : // The stats collection job grabs and clears the pending list, computes table
40 : // statistics relative to the current readState and updates the tables' file
41 : // metadata. New pending files may accumulate during a stats collection job,
42 : // so a completing job triggers a new job if necessary. Only one job runs at a
43 : // time.
44 : //
45 : // When an existing database is opened, all files lack in-memory statistics.
46 : // These files' stats are loaded incrementally whenever the pending list is
47 : // empty by scanning a current readState for files missing statistics. Once a
48 : // job completes a scan without finding any remaining files without
49 : // statistics, it flips a `loadedInitial` flag. From then on, the stats
50 : // collection job only needs to load statistics for new files appended to the
51 : // pending list.
52 :
53 1 : func (d *DB) maybeCollectTableStatsLocked() {
54 1 : if d.shouldCollectTableStatsLocked() {
55 1 : go d.collectTableStats()
56 1 : }
57 : }
58 :
59 : // updateTableStatsLocked is called when new files are introduced, after the
60 : // read state has been updated. It may trigger a new stat collection.
61 : // DB.mu must be locked when calling.
62 1 : func (d *DB) updateTableStatsLocked(newTables []manifest.NewTableEntry) {
63 1 : var needStats bool
64 1 : for _, nf := range newTables {
65 1 : if !nf.Meta.StatsValid() {
66 1 : needStats = true
67 1 : break
68 : }
69 : }
70 1 : if !needStats {
71 1 : return
72 1 : }
73 :
74 1 : d.mu.tableStats.pending = append(d.mu.tableStats.pending, newTables...)
75 1 : d.maybeCollectTableStatsLocked()
76 : }
77 :
78 1 : func (d *DB) shouldCollectTableStatsLocked() bool {
79 1 : return !d.mu.tableStats.loading &&
80 1 : d.closed.Load() == nil &&
81 1 : !d.opts.DisableTableStats &&
82 1 : (len(d.mu.tableStats.pending) > 0 || !d.mu.tableStats.loadedInitial)
83 1 : }
84 :
85 : // collectTableStats runs a table stats collection job, returning true if the
86 : // invocation did the collection work, false otherwise (e.g. if another job was
87 : // already running).
88 1 : func (d *DB) collectTableStats() bool {
89 1 : const maxTableStatsPerScan = 50
90 1 :
91 1 : d.mu.Lock()
92 1 : if !d.shouldCollectTableStatsLocked() {
93 1 : d.mu.Unlock()
94 1 : return false
95 1 : }
96 1 : ctx := context.Background()
97 1 :
98 1 : pending := d.mu.tableStats.pending
99 1 : d.mu.tableStats.pending = nil
100 1 : d.mu.tableStats.loading = true
101 1 : jobID := d.newJobIDLocked()
102 1 : loadedInitial := d.mu.tableStats.loadedInitial
103 1 : // Drop DB.mu before performing IO.
104 1 : d.mu.Unlock()
105 1 :
106 1 : // Every run of collectTableStats either collects stats from the pending
107 1 : // list (if non-empty) or from scanning the version (loadedInitial is
108 1 : // false). This job only runs if at least one of those conditions holds.
109 1 :
110 1 : // Grab a read state to scan for tables.
111 1 : rs := d.loadReadState()
112 1 : var collected []collectedStats
113 1 : var hints []deleteCompactionHint
114 1 : if len(pending) > 0 {
115 1 : collected, hints = d.loadNewFileStats(ctx, rs, pending)
116 1 : } else {
117 1 : var moreRemain bool
118 1 : var buf [maxTableStatsPerScan]collectedStats
119 1 : collected, hints, moreRemain = d.scanReadStateTableStats(ctx, rs, buf[:0])
120 1 : loadedInitial = !moreRemain
121 1 : }
122 1 : rs.unref()
123 1 :
124 1 : // Update the TableMetadata with the loaded stats while holding d.mu.
125 1 : d.mu.Lock()
126 1 : defer d.mu.Unlock()
127 1 : d.mu.tableStats.loading = false
128 1 : if loadedInitial && !d.mu.tableStats.loadedInitial {
129 1 : d.mu.tableStats.loadedInitial = loadedInitial
130 1 : d.opts.EventListener.TableStatsLoaded(TableStatsInfo{
131 1 : JobID: int(jobID),
132 1 : })
133 1 : }
134 :
135 1 : maybeCompact := false
136 1 : for _, c := range collected {
137 1 : c.TableMetadata.Stats = c.TableStats
138 1 : maybeCompact = maybeCompact || tableTombstoneCompensation(c.TableMetadata) > 0
139 1 : sanityCheckStats(c.TableMetadata, d.opts.Logger, "collected stats")
140 1 : c.TableMetadata.StatsMarkValid()
141 1 : }
142 :
143 1 : d.mu.tableStats.cond.Broadcast()
144 1 : d.maybeCollectTableStatsLocked()
145 1 : if len(hints) > 0 && !d.opts.private.disableDeleteOnlyCompactions {
146 1 : // Verify that all of the hint tombstones' files still exist in the
147 1 : // current version. Otherwise, the tombstone itself may have been
148 1 : // compacted into L6 and more recent keys may have had their sequence
149 1 : // numbers zeroed.
150 1 : //
151 1 : // Note that it's possible that the tombstone file is being compacted
152 1 : // presently. In that case, the file will be present in v. When the
153 1 : // compaction finishes compacting the tombstone file, it will detect
154 1 : // and clear the hint.
155 1 : //
156 1 : // See DB.maybeUpdateDeleteCompactionHints.
157 1 : v := d.mu.versions.currentVersion()
158 1 : keepHints := hints[:0]
159 1 : for _, h := range hints {
160 1 : if v.Contains(h.tombstoneLevel, h.tombstoneFile) {
161 1 : keepHints = append(keepHints, h)
162 1 : }
163 : }
164 1 : d.mu.compact.deletionHints = append(d.mu.compact.deletionHints, keepHints...)
165 : }
166 1 : if maybeCompact {
167 1 : d.maybeScheduleCompaction()
168 1 : }
169 1 : return true
170 : }
171 :
172 : type collectedStats struct {
173 : *manifest.TableMetadata
174 : manifest.TableStats
175 : }
176 :
177 : func (d *DB) loadNewFileStats(
178 : ctx context.Context, rs *readState, pending []manifest.NewTableEntry,
179 1 : ) ([]collectedStats, []deleteCompactionHint) {
180 1 : var hints []deleteCompactionHint
181 1 : collected := make([]collectedStats, 0, len(pending))
182 1 : for _, nf := range pending {
183 1 : // A file's stats might have been populated by an earlier call to
184 1 : // loadNewFileStats if the file was moved.
185 1 : // NB: We're not holding d.mu which protects f.Stats, but only
186 1 : // collectTableStats updates f.Stats for active files, and we
187 1 : // ensure only one goroutine runs it at a time through
188 1 : // d.mu.tableStats.loading.
189 1 : if nf.Meta.StatsValid() {
190 1 : continue
191 : }
192 :
193 : // The file isn't guaranteed to still be live in the readState's
194 : // version. It may have been deleted or moved. Skip it if it's not in
195 : // the expected level.
196 1 : if !rs.current.Contains(nf.Level, nf.Meta) {
197 1 : continue
198 : }
199 :
200 1 : stats, newHints, err := d.loadTableStats(ctx, rs.current, nf.Level, nf.Meta)
201 1 : if err != nil {
202 0 : d.opts.EventListener.BackgroundError(err)
203 0 : continue
204 : }
205 : // NB: We don't update the TableMetadata yet, because we aren't holding
206 : // DB.mu. We'll copy it to the TableMetadata after we're finished with
207 : // IO.
208 1 : collected = append(collected, collectedStats{
209 1 : TableMetadata: nf.Meta,
210 1 : TableStats: stats,
211 1 : })
212 1 : hints = append(hints, newHints...)
213 : }
214 1 : return collected, hints
215 : }
216 :
217 : // scanReadStateTableStats is run by an active stat collection job when there
218 : // are no pending new files, but there might be files that existed at Open for
219 : // which we haven't loaded table stats.
220 : func (d *DB) scanReadStateTableStats(
221 : ctx context.Context, rs *readState, fill []collectedStats,
222 1 : ) ([]collectedStats, []deleteCompactionHint, bool) {
223 1 : moreRemain := false
224 1 : var hints []deleteCompactionHint
225 1 : sizesChecked := make(map[base.DiskFileNum]struct{})
226 1 : for l, levelMetadata := range rs.current.Levels {
227 1 : for f := range levelMetadata.All() {
228 1 : // NB: We're not holding d.mu which protects f.Stats, but only the
229 1 : // active stats collection job updates f.Stats for active files,
230 1 : // and we ensure only one goroutine runs it at a time through
231 1 : // d.mu.tableStats.loading. This makes it safe to read validity
232 1 : // through f.Stats.ValidLocked despite not holding d.mu.
233 1 : if f.StatsValid() {
234 1 : continue
235 : }
236 :
237 : // Limit how much work we do per read state. The older the read
238 : // state is, the higher the likelihood files are no longer being
239 : // used in the current version. If we've exhausted our allowance,
240 : // return true for the last return value to signal there's more
241 : // work to do.
242 1 : if len(fill) == cap(fill) {
243 1 : moreRemain = true
244 1 : return fill, hints, moreRemain
245 1 : }
246 :
247 : // If the file is remote and not SharedForeign, we should check if its size
248 : // matches. This is because checkConsistency skips over remote files.
249 : //
250 : // SharedForeign and External files are skipped as their sizes are allowed
251 : // to have a mismatch; the size stored in the TableBacking is just the part
252 : // of the file that is referenced by this Pebble instance, not the size of
253 : // the whole object.
254 1 : objMeta, err := d.objProvider.Lookup(base.FileTypeTable, f.TableBacking.DiskFileNum)
255 1 : if err != nil {
256 0 : // Set `moreRemain` so we'll try again.
257 0 : moreRemain = true
258 0 : d.opts.EventListener.BackgroundError(err)
259 0 : continue
260 : }
261 :
262 1 : shouldCheckSize := objMeta.IsRemote() &&
263 1 : !d.objProvider.IsSharedForeign(objMeta) &&
264 1 : !objMeta.IsExternal()
265 1 : if _, ok := sizesChecked[f.TableBacking.DiskFileNum]; !ok && shouldCheckSize {
266 1 : size, err := d.objProvider.Size(objMeta)
267 1 : fileSize := f.TableBacking.Size
268 1 : if err != nil {
269 0 : moreRemain = true
270 0 : d.opts.EventListener.BackgroundError(err)
271 0 : continue
272 : }
273 1 : if size != int64(fileSize) {
274 0 : err := errors.Errorf(
275 0 : "during consistency check in loadTableStats: L%d: %s: object size mismatch (%s): %d (provider) != %d (MANIFEST)",
276 0 : errors.Safe(l), f.TableNum, d.objProvider.Path(objMeta),
277 0 : errors.Safe(size), errors.Safe(fileSize))
278 0 : d.opts.EventListener.BackgroundError(err)
279 0 : d.opts.Logger.Fatalf("%s", err)
280 0 : }
281 :
282 1 : sizesChecked[f.TableBacking.DiskFileNum] = struct{}{}
283 : }
284 :
285 1 : stats, newHints, err := d.loadTableStats(ctx, rs.current, l, f)
286 1 : if err != nil {
287 0 : // Set `moreRemain` so we'll try again.
288 0 : moreRemain = true
289 0 : d.opts.EventListener.BackgroundError(err)
290 0 : continue
291 : }
292 1 : fill = append(fill, collectedStats{
293 1 : TableMetadata: f,
294 1 : TableStats: stats,
295 1 : })
296 1 : hints = append(hints, newHints...)
297 : }
298 : }
299 1 : return fill, hints, moreRemain
300 : }
301 :
302 : func (d *DB) loadTableStats(
303 : ctx context.Context, v *manifest.Version, level int, meta *manifest.TableMetadata,
304 1 : ) (manifest.TableStats, []deleteCompactionHint, error) {
305 1 : var stats manifest.TableStats
306 1 : var compactionHints []deleteCompactionHint
307 1 :
308 1 : err := d.fileCache.withReader(
309 1 : ctx, block.NoReadEnv, meta, func(r *sstable.Reader, env sstable.ReadEnv) (err error) {
310 1 : loadedProps, err := r.ReadPropertiesBlock(ctx, nil /* buffer pool */)
311 1 : if err != nil {
312 0 : return err
313 0 : }
314 1 : props := loadedProps.CommonProperties
315 1 : if meta.Virtual {
316 1 : props = loadedProps.GetScaledProperties(meta.TableBacking.Size, meta.Size)
317 1 : }
318 1 : stats.NumEntries = props.NumEntries
319 1 : stats.NumDeletions = props.NumDeletions
320 1 : stats.NumRangeKeySets = props.NumRangeKeySets
321 1 : stats.ValueBlocksSize = props.ValueBlocksSize
322 1 : stats.RawKeySize = props.RawKeySize
323 1 : stats.RawValueSize = props.RawValueSize
324 1 : stats.CompressionType = block.CompressionProfileByName(props.CompressionName)
325 1 : if props.NumDataBlocks > 0 {
326 1 : stats.TombstoneDenseBlocksRatio = float64(props.NumTombstoneDenseBlocks) / float64(props.NumDataBlocks)
327 1 : }
328 :
329 1 : if props.NumPointDeletions() > 0 {
330 1 : if err = d.loadTablePointKeyStats(ctx, &props, v, level, meta, &stats); err != nil {
331 0 : return
332 0 : }
333 : }
334 1 : if r.Attributes.Intersects(sstable.AttributeRangeDels | sstable.AttributeRangeKeyDels) {
335 1 : compactionHints, err = d.loadTableRangeDelStats(ctx, r, v, level, meta, &stats, env)
336 1 : if err != nil {
337 0 : return
338 0 : }
339 : }
340 1 : return
341 : })
342 1 : if err != nil {
343 0 : return stats, nil, err
344 0 : }
345 1 : return stats, compactionHints, nil
346 : }
347 :
348 : // loadTablePointKeyStats calculates the point key statistics for the given
349 : // table. The provided manifest.TableStats are updated.
350 : func (d *DB) loadTablePointKeyStats(
351 : ctx context.Context,
352 : props *sstable.CommonProperties,
353 : v *manifest.Version,
354 : level int,
355 : meta *manifest.TableMetadata,
356 : stats *manifest.TableStats,
357 1 : ) error {
358 1 : // TODO(jackson): If the file has a wide keyspace, the average
359 1 : // value size beneath the entire file might not be representative
360 1 : // of the size of the keys beneath the point tombstones.
361 1 : // We could write the ranges of 'clusters' of point tombstones to
362 1 : // a sstable property and call averageValueSizeBeneath for each of
363 1 : // these narrower ranges to improve the estimate.
364 1 : avgValLogicalSize, compressionRatio, err := d.estimateSizesBeneath(ctx, v, level, meta, props)
365 1 : if err != nil {
366 0 : return err
367 0 : }
368 1 : stats.PointDeletionsBytesEstimate =
369 1 : pointDeletionsBytesEstimate(props, avgValLogicalSize, compressionRatio)
370 1 : return nil
371 : }
372 :
373 : // loadTableRangeDelStats calculates the range deletion and range key deletion
374 : // statistics for the given table.
375 : func (d *DB) loadTableRangeDelStats(
376 : ctx context.Context,
377 : r *sstable.Reader,
378 : v *manifest.Version,
379 : level int,
380 : meta *manifest.TableMetadata,
381 : stats *manifest.TableStats,
382 : env sstable.ReadEnv,
383 1 : ) ([]deleteCompactionHint, error) {
384 1 : iter, err := newCombinedDeletionKeyspanIter(ctx, d.opts.Comparer, r, meta, env)
385 1 : if err != nil {
386 0 : return nil, err
387 0 : }
388 1 : defer iter.Close()
389 1 : var compactionHints []deleteCompactionHint
390 1 : // We iterate over the defragmented range tombstones and range key deletions,
391 1 : // which ensures we don't double count ranges deleted at different sequence
392 1 : // numbers. Also, merging abutting tombstones reduces the number of calls to
393 1 : // estimateReclaimedSizeBeneath which is costly, and improves the accuracy of
394 1 : // our overall estimate.
395 1 : s, err := iter.First()
396 1 : for ; s != nil; s, err = iter.Next() {
397 1 : start, end := s.Start, s.End
398 1 : // We only need to consider deletion size estimates for tables that contain
399 1 : // RANGEDELs.
400 1 : var maxRangeDeleteSeqNum base.SeqNum
401 1 : for _, k := range s.Keys {
402 1 : if k.Kind() == base.InternalKeyKindRangeDelete && maxRangeDeleteSeqNum < k.SeqNum() {
403 1 : maxRangeDeleteSeqNum = k.SeqNum()
404 1 : break
405 : }
406 : }
407 :
408 : // If the file is in the last level of the LSM, there is no data beneath
409 : // it. The fact that there is still a range tombstone in a bottommost file
410 : // indicates two possibilites:
411 : // 1. an open snapshot kept the tombstone around, and the data the
412 : // tombstone deletes is contained within the file itself.
413 : // 2. the file was ingested.
414 : // In the first case, we'd like to estimate disk usage within the file
415 : // itself since compacting the file will drop that covered data. In the
416 : // second case, we expect that compacting the file will NOT drop any
417 : // data and rewriting the file is a waste of write bandwidth. We can
418 : // distinguish these cases by looking at the table metadata's sequence
419 : // numbers. A file's range deletions can only delete data within the
420 : // file at lower sequence numbers. All keys in an ingested sstable adopt
421 : // the same sequence number, preventing tombstones from deleting keys
422 : // within the same file. We check here if the largest RANGEDEL sequence
423 : // number is greater than the file's smallest sequence number. If it is,
424 : // the RANGEDEL could conceivably (although inconclusively) delete data
425 : // within the same file.
426 : //
427 : // Note that this heuristic is imperfect. If a table containing a range
428 : // deletion is ingested into L5 and subsequently compacted into L6 but
429 : // an open snapshot prevents elision of covered keys in L6, the
430 : // resulting RangeDeletionsBytesEstimate will incorrectly include all
431 : // covered keys.
432 : //
433 : // TODO(jackson): We could prevent the above error in the heuristic by
434 : // computing the file's RangeDeletionsBytesEstimate during the
435 : // compaction itself. It's unclear how common this is.
436 : //
437 : // NOTE: If the span `s` wholly contains a table containing range keys,
438 : // the returned size estimate will be slightly inflated by the range key
439 : // block. However, in practice, range keys are expected to be rare, and
440 : // the size of the range key block relative to the overall size of the
441 : // table is expected to be small.
442 1 : if level == numLevels-1 && meta.SmallestSeqNum < maxRangeDeleteSeqNum {
443 1 : size, err := r.EstimateDiskUsage(start, end, env)
444 1 : if err != nil {
445 0 : return nil, err
446 0 : }
447 1 : stats.RangeDeletionsBytesEstimate += size
448 1 :
449 1 : // As the file is in the bottommost level, there is no need to collect a
450 1 : // deletion hint.
451 1 : continue
452 : }
453 :
454 : // While the size estimates for point keys should only be updated if this
455 : // span contains a range del, the sequence numbers are required for the
456 : // hint. Unconditionally descend, but conditionally update the estimates.
457 1 : hintType := compactionHintFromKeys(s.Keys)
458 1 : estimate, hintSeqNum, err := d.estimateReclaimedSizeBeneath(ctx, v, level, start, end, hintType)
459 1 : if err != nil {
460 0 : return nil, err
461 0 : }
462 1 : stats.RangeDeletionsBytesEstimate += estimate
463 1 :
464 1 : // hintSeqNum is the smallest sequence number contained in any
465 1 : // file overlapping with the hint and in a level below it.
466 1 : if hintSeqNum == math.MaxUint64 {
467 1 : continue
468 : }
469 1 : compactionHints = append(compactionHints, deleteCompactionHint{
470 1 : hintType: hintType,
471 1 : start: slices.Clone(start),
472 1 : end: slices.Clone(end),
473 1 : tombstoneFile: meta,
474 1 : tombstoneLevel: level,
475 1 : tombstoneLargestSeqNum: s.LargestSeqNum(),
476 1 : tombstoneSmallestSeqNum: s.SmallestSeqNum(),
477 1 : fileSmallestSeqNum: hintSeqNum,
478 1 : })
479 : }
480 1 : if err != nil {
481 0 : return nil, err
482 0 : }
483 1 : return compactionHints, nil
484 : }
485 :
486 : func (d *DB) estimateSizesBeneath(
487 : ctx context.Context,
488 : v *manifest.Version,
489 : level int,
490 : meta *manifest.TableMetadata,
491 : fileProps *sstable.CommonProperties,
492 1 : ) (avgValueLogicalSize, compressionRatio float64, err error) {
493 1 : // Find all files in lower levels that overlap with meta,
494 1 : // summing their value sizes and entry counts.
495 1 :
496 1 : // Include the file itself. This is important because in some instances, the
497 1 : // computed compression ratio is applied to the tombstones contained within
498 1 : // `meta` itself. If there are no files beneath `meta` in the LSM, we would
499 1 : // calculate a compression ratio of 0 which is not accurate for the file's
500 1 : // own tombstones.
501 1 : var (
502 1 : // TODO(sumeer): The entryCount includes the tombstones, which can be small,
503 1 : // resulting in a lower than expected avgValueLogicalSize. For an example of
504 1 : // this effect see the estimate in testdata/compaction_picker_scores (search
505 1 : // for "point-deletions-bytes-estimate: 163850").
506 1 : fileSum = meta.Size + meta.EstimatedReferenceSize()
507 1 : entryCount = fileProps.NumEntries
508 1 : keySum = fileProps.RawKeySize
509 1 : valSum = fileProps.RawValueSize
510 1 : )
511 1 :
512 1 : for l := level + 1; l < numLevels; l++ {
513 1 : for tableBeneath := range v.Overlaps(l, meta.UserKeyBounds()).All() {
514 1 : fileSum += tableBeneath.Size + tableBeneath.EstimatedReferenceSize()
515 1 : if tableBeneath.StatsValid() {
516 1 : entryCount += tableBeneath.Stats.NumEntries
517 1 : keySum += tableBeneath.Stats.RawKeySize
518 1 : valSum += tableBeneath.Stats.RawValueSize
519 1 : continue
520 : }
521 : // If stats aren't available, we need to read the properties block.
522 1 : err := d.fileCache.withReader(ctx, block.NoReadEnv, tableBeneath, func(v *sstable.Reader, _ sstable.ReadEnv) (err error) {
523 1 : loadedProps, err := v.ReadPropertiesBlock(ctx, nil /* buffer pool */)
524 1 : if err != nil {
525 0 : return err
526 0 : }
527 1 : props := loadedProps.CommonProperties
528 1 : if tableBeneath.Virtual {
529 1 : props = loadedProps.GetScaledProperties(tableBeneath.TableBacking.Size, tableBeneath.Size)
530 1 : }
531 :
532 1 : entryCount += props.NumEntries
533 1 : keySum += props.RawKeySize
534 1 : valSum += props.RawValueSize
535 1 : return nil
536 : })
537 1 : if err != nil {
538 0 : return 0, 0, err
539 0 : }
540 : }
541 : }
542 1 : if entryCount == 0 {
543 0 : return 0, 0, nil
544 0 : }
545 : // RawKeySize and RawValueSize are uncompressed totals. We'll need to scale
546 : // the value sum according to the data size to account for compression,
547 : // index blocks and metadata overhead. Eg:
548 : //
549 : // Compression rate × Average uncompressed value size
550 : //
551 : // ↓
552 : //
553 : // FileSize RawValueSize
554 : // ----------------------- × ------------
555 : // RawKeySize+RawValueSize NumEntries
556 : //
557 : // We return the average logical value size plus the compression ratio,
558 : // leaving the scaling to the caller. This allows the caller to perform
559 : // additional compression ratio scaling if necessary.
560 1 : uncompressedSum := float64(keySum + valSum)
561 1 : compressionRatio = float64(fileSum) / uncompressedSum
562 1 : if compressionRatio > 1 {
563 1 : // We can get huge compression ratios due to the fixed overhead of files
564 1 : // containing a tiny amount of data. By setting this to 1, we are ignoring
565 1 : // that overhead, but we accept that tradeoff since the total bytes in
566 1 : // such overhead is not large.
567 1 : compressionRatio = 1
568 1 : }
569 1 : avgValueLogicalSize = (float64(valSum) / float64(entryCount))
570 1 : return avgValueLogicalSize, compressionRatio, nil
571 : }
572 :
573 : func (d *DB) estimateReclaimedSizeBeneath(
574 : ctx context.Context,
575 : v *manifest.Version,
576 : level int,
577 : start, end []byte,
578 : hintType deleteCompactionHintType,
579 1 : ) (estimate uint64, hintSeqNum base.SeqNum, err error) {
580 1 : // Find all files in lower levels that overlap with the deleted range
581 1 : // [start, end).
582 1 : //
583 1 : // An overlapping file might be completely contained by the range
584 1 : // tombstone, in which case we can count the entire file size in
585 1 : // our estimate without doing any additional I/O.
586 1 : //
587 1 : // Otherwise, estimating the range for the file requires
588 1 : // additional I/O to read the file's index blocks.
589 1 : hintSeqNum = math.MaxUint64
590 1 : // TODO(jbowens): When there are multiple sub-levels in L0 and the RANGEDEL
591 1 : // is from a higher sub-level, we incorrectly skip the files in the lower
592 1 : // sub-levels when estimating this overlap.
593 1 : for l := level + 1; l < numLevels; l++ {
594 1 : for file := range v.Overlaps(l, base.UserKeyBoundsEndExclusive(start, end)).All() {
595 1 : // Determine whether we need to update size estimates and hint seqnums
596 1 : // based on the type of hint and the type of keys in this file.
597 1 : var updateEstimates, updateHints bool
598 1 : switch hintType {
599 1 : case deleteCompactionHintTypePointKeyOnly:
600 1 : // The range deletion byte estimates should only be updated if this
601 1 : // table contains point keys. This ends up being an overestimate in
602 1 : // the case that table also has range keys, but such keys are expected
603 1 : // to contribute a negligible amount of the table's overall size,
604 1 : // relative to point keys.
605 1 : if file.HasPointKeys {
606 1 : updateEstimates = true
607 1 : }
608 : // As the initiating span contained only range dels, hints can only be
609 : // updated if this table does _not_ contain range keys.
610 1 : if !file.HasRangeKeys {
611 1 : updateHints = true
612 1 : }
613 1 : case deleteCompactionHintTypeRangeKeyOnly:
614 1 : // The initiating span contained only range key dels. The estimates
615 1 : // apply only to point keys, and are therefore not updated.
616 1 : updateEstimates = false
617 1 : // As the initiating span contained only range key dels, hints can
618 1 : // only be updated if this table does _not_ contain point keys.
619 1 : if !file.HasPointKeys {
620 1 : updateHints = true
621 1 : }
622 1 : case deleteCompactionHintTypePointAndRangeKey:
623 1 : // Always update the estimates and hints, as this hint type can drop a
624 1 : // file, irrespective of the mixture of keys. Similar to above, the
625 1 : // range del bytes estimates is an overestimate.
626 1 : updateEstimates, updateHints = true, true
627 0 : default:
628 0 : panic(fmt.Sprintf("pebble: unknown hint type %s", hintType))
629 : }
630 1 : startCmp := d.cmp(start, file.Smallest().UserKey)
631 1 : endCmp := d.cmp(file.Largest().UserKey, end)
632 1 : if startCmp <= 0 && (endCmp < 0 || endCmp == 0 && file.Largest().IsExclusiveSentinel()) {
633 1 : // The range fully contains the file, so skip looking it up in table
634 1 : // cache/looking at its indexes and add the full file size.
635 1 : if updateEstimates {
636 1 : estimate += file.Size
637 1 : }
638 1 : if updateHints && hintSeqNum > file.SmallestSeqNum {
639 1 : hintSeqNum = file.SmallestSeqNum
640 1 : }
641 1 : } else if d.cmp(file.Smallest().UserKey, end) <= 0 && d.cmp(start, file.Largest().UserKey) <= 0 {
642 1 : // Partial overlap.
643 1 : if hintType == deleteCompactionHintTypeRangeKeyOnly {
644 1 : // If the hint that generated this overlap contains only range keys,
645 1 : // there is no need to calculate disk usage, as the reclaimable space
646 1 : // is expected to be minimal relative to point keys.
647 1 : continue
648 : }
649 1 : var size uint64
650 1 : err := d.fileCache.withReader(ctx, block.NoReadEnv, file,
651 1 : func(r *sstable.Reader, env sstable.ReadEnv) (err error) {
652 1 : size, err = r.EstimateDiskUsage(start, end, env)
653 1 : return err
654 1 : })
655 1 : if err != nil {
656 0 : return 0, hintSeqNum, err
657 0 : }
658 1 : estimate += size
659 1 : if updateHints && hintSeqNum > file.SmallestSeqNum && d.FormatMajorVersion() >= FormatVirtualSSTables {
660 1 : // If the format major version is past Virtual SSTables, deletion only
661 1 : // hints can also apply to partial overlaps with sstables.
662 1 : hintSeqNum = file.SmallestSeqNum
663 1 : }
664 : }
665 : }
666 : }
667 1 : return estimate, hintSeqNum, nil
668 : }
669 :
670 : var lastSanityCheckStatsLog crtime.AtomicMono
671 :
672 1 : func sanityCheckStats(meta *manifest.TableMetadata, logger Logger, info string) {
673 1 : // Values for PointDeletionsBytesEstimate and RangeDeletionsBytesEstimate that
674 1 : // exceed this value are likely indicative of a bug (eg, underflow).
675 1 : const maxDeletionBytesEstimate = 1 << 50 // 1 PiB
676 1 :
677 1 : if meta.Stats.PointDeletionsBytesEstimate > maxDeletionBytesEstimate ||
678 1 : meta.Stats.RangeDeletionsBytesEstimate > maxDeletionBytesEstimate {
679 0 : if invariants.Enabled {
680 0 : panic(fmt.Sprintf("%s: table %s has extreme deletion bytes estimates: point=%d range=%d",
681 0 : info, meta.TableNum,
682 0 : redact.Safe(meta.Stats.PointDeletionsBytesEstimate),
683 0 : redact.Safe(meta.Stats.RangeDeletionsBytesEstimate),
684 0 : ))
685 : }
686 0 : if v := lastSanityCheckStatsLog.Load(); v == 0 || v.Elapsed() > 30*time.Second {
687 0 : logger.Errorf("%s: table %s has extreme deletion bytes estimates: point=%d range=%d",
688 0 : info, meta.TableNum,
689 0 : redact.Safe(meta.Stats.PointDeletionsBytesEstimate),
690 0 : redact.Safe(meta.Stats.RangeDeletionsBytesEstimate),
691 0 : )
692 0 : lastSanityCheckStatsLog.Store(crtime.NowMono())
693 0 : }
694 : }
695 : }
696 :
697 : func maybeSetStatsFromProperties(
698 : meta *manifest.TableMetadata, props *sstable.CommonProperties, logger Logger,
699 1 : ) bool {
700 1 : // If a table contains range deletions or range key deletions, we defer the
701 1 : // stats collection. There are two main reasons for this:
702 1 : //
703 1 : // 1. Estimating the potential for reclaimed space due to a range deletion
704 1 : // tombstone requires scanning the LSM - a potentially expensive operation
705 1 : // that should be deferred.
706 1 : // 2. Range deletions and / or range key deletions present an opportunity to
707 1 : // compute "deletion hints", which also requires a scan of the LSM to
708 1 : // compute tables that would be eligible for deletion.
709 1 : //
710 1 : // These two tasks are deferred to the table stats collector goroutine.
711 1 : if props.NumRangeDeletions != 0 || props.NumRangeKeyDels != 0 {
712 1 : return false
713 1 : }
714 :
715 : // If a table is more than 10% point deletions without user-provided size
716 : // estimates, don't calculate the PointDeletionsBytesEstimate statistic
717 : // using our limited knowledge. The table stats collector can populate the
718 : // stats and calculate an average of value size of all the tables beneath
719 : // the table in the LSM, which will be more accurate.
720 1 : if unsizedDels := (props.NumDeletions - props.NumSizedDeletions); unsizedDels > props.NumEntries/10 {
721 1 : return false
722 1 : }
723 :
724 1 : var pointEstimate uint64
725 1 : if props.NumEntries > 0 {
726 1 : // Use the file's own average key and value sizes as an estimate. This
727 1 : // doesn't require any additional IO and since the number of point
728 1 : // deletions in the file is low, the error introduced by this crude
729 1 : // estimate is expected to be small.
730 1 : avgValSize, compressionRatio := estimatePhysicalSizes(meta, props)
731 1 : pointEstimate = pointDeletionsBytesEstimate(props, avgValSize, compressionRatio)
732 1 : }
733 :
734 1 : meta.Stats.NumEntries = props.NumEntries
735 1 : meta.Stats.NumDeletions = props.NumDeletions
736 1 : meta.Stats.NumRangeKeySets = props.NumRangeKeySets
737 1 : meta.Stats.PointDeletionsBytesEstimate = pointEstimate
738 1 : meta.Stats.RangeDeletionsBytesEstimate = 0
739 1 : meta.Stats.ValueBlocksSize = props.ValueBlocksSize
740 1 : meta.Stats.RawKeySize = props.RawKeySize
741 1 : meta.Stats.RawValueSize = props.RawValueSize
742 1 : meta.Stats.CompressionType = block.CompressionProfileByName(props.CompressionName)
743 1 : meta.StatsMarkValid()
744 1 : sanityCheckStats(meta, logger, "stats from properties")
745 1 : return true
746 : }
747 :
748 : func pointDeletionsBytesEstimate(
749 : props *sstable.CommonProperties, avgValLogicalSize, compressionRatio float64,
750 1 : ) (estimate uint64) {
751 1 : if props.NumEntries == 0 {
752 0 : return 0
753 0 : }
754 1 : numPointDels := props.NumPointDeletions()
755 1 : if numPointDels == 0 {
756 1 : return 0
757 1 : }
758 : // Estimate the potential space to reclaim using the table's own properties.
759 : // There may or may not be keys covered by any individual point tombstone.
760 : // If not, compacting the point tombstone into L6 will at least allow us to
761 : // drop the point deletion key and will reclaim the tombstone's key bytes.
762 : // If there are covered key(s), we also get to drop key and value bytes for
763 : // each covered key.
764 : //
765 : // Some point tombstones (DELSIZEDs) carry a user-provided estimate of the
766 : // uncompressed size of entries that will be elided by fully compacting the
767 : // tombstone. For these tombstones, there's no guesswork—we use the
768 : // RawPointTombstoneValueSizeHint property which is the sum of all these
769 : // tombstones' encoded values.
770 : //
771 : // For un-sized point tombstones (DELs), we estimate assuming that each
772 : // point tombstone on average covers 1 key and using average value sizes.
773 : // This is almost certainly an overestimate, but that's probably okay
774 : // because point tombstones can slow range iterations even when they don't
775 : // cover a key.
776 : //
777 : // TODO(jackson): This logic doesn't directly incorporate fixed per-key
778 : // overhead (8-byte trailer, plus at least 1 byte encoding the length of the
779 : // key and 1 byte encoding the length of the value). This overhead is
780 : // indirectly incorporated through the compression ratios, but that results
781 : // in the overhead being smeared per key-byte and value-byte, rather than
782 : // per-entry. This per-key fixed overhead can be nontrivial, especially for
783 : // dense swaths of point tombstones. Give some thought as to whether we
784 : // should directly include fixed per-key overhead in the calculations.
785 :
786 : // Below, we calculate the tombstone contributions and the shadowed keys'
787 : // contributions separately.
788 1 : var tombstonesLogicalSize float64
789 1 : var shadowedLogicalSize float64
790 1 :
791 1 : // 1. Calculate the contribution of the tombstone keys themselves.
792 1 : if props.RawPointTombstoneKeySize > 0 {
793 1 : tombstonesLogicalSize += float64(props.RawPointTombstoneKeySize)
794 1 : } else {
795 0 : // This sstable predates the existence of the RawPointTombstoneKeySize
796 0 : // property. We can use the average key size within the file itself and
797 0 : // the count of point deletions to estimate the size.
798 0 : tombstonesLogicalSize += float64(numPointDels * props.RawKeySize / props.NumEntries)
799 0 : }
800 :
801 : // 2. Calculate the contribution of the keys shadowed by tombstones.
802 : //
803 : // 2a. First account for keys shadowed by DELSIZED tombstones. THE DELSIZED
804 : // tombstones encode the size of both the key and value of the shadowed KV
805 : // entries. These sizes are aggregated into a sstable property.
806 1 : shadowedLogicalSize += float64(props.RawPointTombstoneValueSize)
807 1 :
808 1 : // 2b. Calculate the contribution of the KV entries shadowed by ordinary DEL
809 1 : // keys.
810 1 : numUnsizedDels := invariants.SafeSub(numPointDels, props.NumSizedDeletions)
811 1 : {
812 1 : // The shadowed keys have the same exact user keys as the tombstones
813 1 : // themselves, so we can use the `tombstonesLogicalSize` we computed
814 1 : // earlier as an estimate. There's a complication that
815 1 : // `tombstonesLogicalSize` may include DELSIZED keys we already
816 1 : // accounted for.
817 1 : shadowedLogicalSize += float64(tombstonesLogicalSize) / float64(numPointDels) * float64(numUnsizedDels)
818 1 :
819 1 : // Calculate the contribution of the deleted values. The caller has
820 1 : // already computed an average logical size (possibly computed across
821 1 : // many sstables).
822 1 : shadowedLogicalSize += float64(numUnsizedDels) * avgValLogicalSize
823 1 : }
824 :
825 : // Scale both tombstone and shadowed totals by logical:physical ratios to
826 : // account for compression, metadata overhead, etc.
827 : //
828 : // Physical FileSize
829 : // ----------- = -----------------------
830 : // Logical RawKeySize+RawValueSize
831 : //
832 1 : return uint64((tombstonesLogicalSize + shadowedLogicalSize) * compressionRatio)
833 : }
834 :
835 : func estimatePhysicalSizes(
836 : tableMeta *manifest.TableMetadata, props *sstable.CommonProperties,
837 1 : ) (avgValLogicalSize, compressionRatio float64) {
838 1 : // RawKeySize and RawValueSize are uncompressed totals. Scale according to
839 1 : // the data size to account for compression, index blocks and metadata
840 1 : // overhead. Eg:
841 1 : //
842 1 : // Compression rate × Average uncompressed value size
843 1 : //
844 1 : // ↓
845 1 : //
846 1 : // FileSize RawValSize
847 1 : // ----------------------- × ----------
848 1 : // RawKeySize+RawValueSize NumEntries
849 1 : //
850 1 : physicalSize := tableMeta.Size + tableMeta.EstimatedReferenceSize()
851 1 : uncompressedSum := props.RawKeySize + props.RawValueSize
852 1 : compressionRatio = float64(physicalSize) / float64(uncompressedSum)
853 1 : if compressionRatio > 1 {
854 1 : // We can get huge compression ratios due to the fixed overhead of files
855 1 : // containing a tiny amount of data. By setting this to 1, we are ignoring
856 1 : // that overhead, but we accept that tradeoff since the total bytes in
857 1 : // such overhead is not large.
858 1 : compressionRatio = 1
859 1 : }
860 1 : avgValLogicalSize = (float64(props.RawValueSize) / float64(props.NumEntries))
861 1 : return avgValLogicalSize, compressionRatio
862 : }
863 :
864 : // newCombinedDeletionKeyspanIter returns a keyspan.FragmentIterator that
865 : // returns "ranged deletion" spans for a single table, providing a combined view
866 : // of both range deletion and range key deletion spans. The
867 : // tableRangedDeletionIter is intended for use in the specific case of computing
868 : // the statistics and deleteCompactionHints for a single table.
869 : //
870 : // As an example, consider the following set of spans from the range deletion
871 : // and range key blocks of a table:
872 : //
873 : // |---------| |---------| |-------| RANGEKEYDELs
874 : // |-----------|-------------| |-----| RANGEDELs
875 : // __________________________________________________________
876 : // a b c d e f g h i j k l m n o p q r s t u v w x y z
877 : //
878 : // The tableRangedDeletionIter produces the following set of output spans, where
879 : // '1' indicates a span containing only range deletions, '2' is a span
880 : // containing only range key deletions, and '3' is a span containing a mixture
881 : // of both range deletions and range key deletions.
882 : //
883 : // 1 3 1 3 2 1 3 2
884 : // |-----|---------|-----|---|-----| |---|-|-----|
885 : // __________________________________________________________
886 : // a b c d e f g h i j k l m n o p q r s t u v w x y z
887 : //
888 : // Algorithm.
889 : //
890 : // The iterator first defragments the range deletion and range key blocks
891 : // separately. During this defragmentation, the range key block is also filtered
892 : // so that keys other than range key deletes are ignored. The range delete and
893 : // range key delete keyspaces are then merged.
894 : //
895 : // Note that the only fragmentation introduced by merging is from where a range
896 : // del span overlaps with a range key del span. Within the bounds of any overlap
897 : // there is guaranteed to be no further fragmentation, as the constituent spans
898 : // have already been defragmented. To the left and right of any overlap, the
899 : // same reasoning applies. For example,
900 : //
901 : // |--------| |-------| RANGEKEYDEL
902 : // |---------------------------| RANGEDEL
903 : // |----1---|----3---|----1----|---2---| Merged, fragmented spans.
904 : // __________________________________________________________
905 : // a b c d e f g h i j k l m n o p q r s t u v w x y z
906 : //
907 : // Any fragmented abutting spans produced by the merging iter will be of
908 : // differing types (i.e. a transition from a span with homogenous key kinds to a
909 : // heterogeneous span, or a transition from a span with exclusively range dels
910 : // to a span with exclusively range key dels). Therefore, further
911 : // defragmentation is not required.
912 : //
913 : // Each span returned by the tableRangeDeletionIter will have at most four keys,
914 : // corresponding to the largest and smallest sequence numbers encountered across
915 : // the range deletes and range keys deletes that comprised the merged spans.
916 : func newCombinedDeletionKeyspanIter(
917 : ctx context.Context,
918 : comparer *base.Comparer,
919 : r *sstable.Reader,
920 : m *manifest.TableMetadata,
921 : env sstable.ReadEnv,
922 1 : ) (keyspan.FragmentIterator, error) {
923 1 : // The range del iter and range key iter are each wrapped in their own
924 1 : // defragmenting iter. For each iter, abutting spans can always be merged.
925 1 : var equal = keyspan.DefragmentMethodFunc(func(_ base.CompareRangeSuffixes, a, b *keyspan.Span) bool { return true })
926 : // Reduce keys by maintaining a slice of at most length two, corresponding to
927 : // the largest and smallest keys in the defragmented span. This maintains the
928 : // contract that the emitted slice is sorted by (SeqNum, Kind) descending.
929 1 : reducer := func(current, incoming []keyspan.Key) []keyspan.Key {
930 1 : if len(current) == 0 && len(incoming) == 0 {
931 0 : // While this should never occur in practice, a defensive return is used
932 0 : // here to preserve correctness.
933 0 : return current
934 0 : }
935 1 : var largest, smallest keyspan.Key
936 1 : var set bool
937 1 : for _, keys := range [2][]keyspan.Key{current, incoming} {
938 1 : if len(keys) == 0 {
939 0 : continue
940 : }
941 1 : first, last := keys[0], keys[len(keys)-1]
942 1 : if !set {
943 1 : largest, smallest = first, last
944 1 : set = true
945 1 : continue
946 : }
947 1 : if first.Trailer > largest.Trailer {
948 1 : largest = first
949 1 : }
950 1 : if last.Trailer < smallest.Trailer {
951 1 : smallest = last
952 1 : }
953 : }
954 1 : if largest.Equal(comparer.CompareRangeSuffixes, smallest) {
955 1 : current = append(current[:0], largest)
956 1 : } else {
957 1 : current = append(current[:0], largest, smallest)
958 1 : }
959 1 : return current
960 : }
961 :
962 : // The separate iters for the range dels and range keys are wrapped in a
963 : // merging iter to join the keyspaces into a single keyspace. The separate
964 : // iters are only added if the particular key kind is present.
965 1 : mIter := &keyspanimpl.MergingIter{}
966 1 : var transform = keyspan.TransformerFunc(func(_ base.CompareRangeSuffixes, in keyspan.Span, out *keyspan.Span) error {
967 1 : if in.KeysOrder != keyspan.ByTrailerDesc {
968 0 : return base.AssertionFailedf("combined deletion iter encountered keys in non-trailer descending order")
969 0 : }
970 1 : out.Start, out.End = in.Start, in.End
971 1 : out.Keys = append(out.Keys[:0], in.Keys...)
972 1 : out.KeysOrder = keyspan.ByTrailerDesc
973 1 : // NB: The order of by-trailer descending may have been violated,
974 1 : // because we've layered rangekey and rangedel iterators from the same
975 1 : // sstable into the same keyspanimpl.MergingIter. The MergingIter will
976 1 : // return the keys in the order that the child iterators were provided.
977 1 : // Sort the keys to ensure they're sorted by trailer descending.
978 1 : keyspan.SortKeysByTrailer(out.Keys)
979 1 : return nil
980 : })
981 1 : mIter.Init(comparer, transform, new(keyspanimpl.MergingBuffers))
982 1 : iter, err := r.NewRawRangeDelIter(ctx, m.FragmentIterTransforms(), env)
983 1 : if err != nil {
984 0 : return nil, err
985 0 : }
986 1 : if iter != nil {
987 1 : // Assert expected bounds. In previous versions of Pebble, range
988 1 : // deletions persisted to sstables could exceed the bounds of the
989 1 : // containing files due to "split user keys." This required readers to
990 1 : // constrain the tombstones' bounds to the containing file at read time.
991 1 : // See docs/range_deletions.md for an extended discussion of the design
992 1 : // and invariants at that time.
993 1 : //
994 1 : // We've since compacted away all 'split user-keys' and in the process
995 1 : // eliminated all "untruncated range tombstones" for physical sstables.
996 1 : // We no longer need to perform truncation at read time for these
997 1 : // sstables.
998 1 : //
999 1 : // At the same time, we've also introduced the concept of "virtual
1000 1 : // SSTables" where the table metadata's effective bounds can again be
1001 1 : // reduced to be narrower than the contained tombstones. These virtual
1002 1 : // SSTables handle truncation differently, performing it using
1003 1 : // keyspan.Truncate when the sstable's range deletion iterator is
1004 1 : // opened.
1005 1 : //
1006 1 : // Together, these mean that we should never see untruncated range
1007 1 : // tombstones any more—and the merging iterator no longer accounts for
1008 1 : // their existence. Since there's abundant subtlety that we're relying
1009 1 : // on, we choose to be conservative and assert that these invariants
1010 1 : // hold. We could (and previously did) choose to only validate these
1011 1 : // bounds in invariants builds, but the most likely avenue for these
1012 1 : // tombstones' existence is through a bug in a migration and old data
1013 1 : // sitting around in an old store from long ago.
1014 1 : //
1015 1 : // The table stats collector will read all files' range deletions
1016 1 : // asynchronously after Open, and provides a perfect opportunity to
1017 1 : // validate our invariants without harming user latency. We also
1018 1 : // previously performed truncation here which similarly required key
1019 1 : // comparisons, so replacing those key comparisons with assertions
1020 1 : // should be roughly similar in performance.
1021 1 : //
1022 1 : // TODO(jackson): Only use AssertBounds in invariants builds in the
1023 1 : // following release.
1024 1 : iter = keyspan.AssertBounds(
1025 1 : iter, m.PointKeyBounds.Smallest(), m.PointKeyBounds.LargestUserKey(), comparer.Compare,
1026 1 : )
1027 1 : dIter := &keyspan.DefragmentingIter{}
1028 1 : dIter.Init(comparer, iter, equal, reducer, new(keyspan.DefragmentingBuffers))
1029 1 : iter = dIter
1030 1 : mIter.AddLevel(iter)
1031 1 : }
1032 :
1033 1 : iter, err = r.NewRawRangeKeyIter(ctx, m.FragmentIterTransforms(), env)
1034 1 : if err != nil {
1035 0 : return nil, err
1036 0 : }
1037 1 : if iter != nil {
1038 1 : // Assert expected bounds in tests.
1039 1 : if invariants.Sometimes(50) {
1040 1 : if m.HasRangeKeys {
1041 1 : iter = keyspan.AssertBounds(
1042 1 : iter, m.RangeKeyBounds.Smallest(), m.RangeKeyBounds.LargestUserKey(), comparer.Compare,
1043 1 : )
1044 1 : }
1045 : }
1046 : // Wrap the range key iterator in a filter that elides keys other than range
1047 : // key deletions.
1048 1 : iter = keyspan.Filter(iter, func(in *keyspan.Span, buf []keyspan.Key) []keyspan.Key {
1049 1 : keys := buf[:0]
1050 1 : for _, k := range in.Keys {
1051 1 : if k.Kind() != base.InternalKeyKindRangeKeyDelete {
1052 1 : continue
1053 : }
1054 1 : keys = append(keys, k)
1055 : }
1056 1 : return keys
1057 : }, comparer.Compare)
1058 1 : dIter := &keyspan.DefragmentingIter{}
1059 1 : dIter.Init(comparer, iter, equal, reducer, new(keyspan.DefragmentingBuffers))
1060 1 : iter = dIter
1061 1 : mIter.AddLevel(iter)
1062 : }
1063 :
1064 1 : return mIter, nil
1065 : }
1066 :
1067 : // rangeKeySetsAnnotator is a manifest.Annotator that annotates B-Tree nodes
1068 : // with the sum of the files' counts of range key fragments. The count of range
1069 : // key sets may change once a table's stats are loaded asynchronously, so its
1070 : // values are marked as cacheable only if a file's stats have been loaded.
1071 1 : var rangeKeySetsAnnotator = manifest.SumAnnotator(func(f *manifest.TableMetadata) (uint64, bool) {
1072 1 : return f.Stats.NumRangeKeySets, f.StatsValid()
1073 1 : })
1074 :
1075 : // tombstonesAnnotator is a manifest.Annotator that annotates B-Tree nodes
1076 : // with the sum of the files' counts of tombstones (DEL, SINGLEDEL and RANGEDEL
1077 : // keys). The count of tombstones may change once a table's stats are loaded
1078 : // asynchronously, so its values are marked as cacheable only if a file's stats
1079 : // have been loaded.
1080 1 : var tombstonesAnnotator = manifest.SumAnnotator(func(f *manifest.TableMetadata) (uint64, bool) {
1081 1 : return f.Stats.NumDeletions, f.StatsValid()
1082 1 : })
1083 :
1084 : // valueBlocksSizeAnnotator is a manifest.Annotator that annotates B-Tree
1085 : // nodes with the sum of the files' Properties.ValueBlocksSize. The value block
1086 : // size may change once a table's stats are loaded asynchronously, so its
1087 : // values are marked as cacheable only if a file's stats have been loaded.
1088 1 : var valueBlockSizeAnnotator = manifest.SumAnnotator(func(f *manifest.TableMetadata) (uint64, bool) {
1089 1 : return f.Stats.ValueBlocksSize, f.StatsValid()
1090 1 : })
1091 :
1092 : // pointDeletionsBytesEstimateAnnotator is a manifest.Annotator that annotates
1093 : // B-Tree nodes with the sum of the files' PointDeletionsBytesEstimate. This
1094 : // value may change once a table's stats are loaded asynchronously, so its
1095 : // values are marked as cacheable only if a file's stats have been loaded.
1096 1 : var pointDeletionsBytesEstimateAnnotator = manifest.SumAnnotator(func(f *manifest.TableMetadata) (uint64, bool) {
1097 1 : return f.Stats.PointDeletionsBytesEstimate, f.StatsValid()
1098 1 : })
1099 :
1100 : // rangeDeletionsBytesEstimateAnnotator is a manifest.Annotator that annotates
1101 : // B-Tree nodes with the sum of the files' RangeDeletionsBytesEstimate. This
1102 : // value may change once a table's stats are loaded asynchronously, so its
1103 : // values are marked as cacheable only if a file's stats have been loaded.
1104 1 : var rangeDeletionsBytesEstimateAnnotator = manifest.SumAnnotator(func(f *manifest.TableMetadata) (uint64, bool) {
1105 1 : return f.Stats.RangeDeletionsBytesEstimate, f.StatsValid()
1106 1 : })
1107 :
1108 : // compressionTypeAnnotator is a manifest.Annotator that annotates B-tree
1109 : // nodes with the compression type of the file. Its annotation type is
1110 : // compressionTypes. The compression type may change once a table's stats are
1111 : // loaded asynchronously, so its values are marked as cacheable only if a file's
1112 : // stats have been loaded.
1113 : var compressionTypeAnnotator = manifest.Annotator[compressionTypes]{
1114 : Aggregator: compressionTypeAggregator{},
1115 : }
1116 :
1117 : type compressionTypeAggregator struct{}
1118 :
1119 : type compressionTypes struct {
1120 : snappy, zstd, minlz, none, unknown uint64
1121 : }
1122 :
1123 1 : func (a compressionTypeAggregator) Zero(dst *compressionTypes) *compressionTypes {
1124 1 : if dst == nil {
1125 1 : return new(compressionTypes)
1126 1 : }
1127 0 : *dst = compressionTypes{}
1128 0 : return dst
1129 : }
1130 :
1131 : func (a compressionTypeAggregator) Accumulate(
1132 : f *manifest.TableMetadata, dst *compressionTypes,
1133 1 : ) (v *compressionTypes, cacheOK bool) {
1134 1 : switch f.Stats.CompressionType {
1135 1 : case sstable.SnappyCompression:
1136 1 : dst.snappy++
1137 1 : case sstable.ZstdCompression:
1138 1 : dst.zstd++
1139 1 : case sstable.MinLZCompression:
1140 1 : dst.minlz++
1141 1 : case sstable.NoCompression:
1142 1 : dst.none++
1143 1 : default:
1144 1 : dst.unknown++
1145 : }
1146 1 : return dst, f.StatsValid()
1147 : }
1148 :
1149 : func (a compressionTypeAggregator) Merge(
1150 : src *compressionTypes, dst *compressionTypes,
1151 1 : ) *compressionTypes {
1152 1 : dst.snappy += src.snappy
1153 1 : dst.zstd += src.zstd
1154 1 : dst.minlz += src.minlz
1155 1 : dst.none += src.none
1156 1 : dst.unknown += src.unknown
1157 1 : return dst
1158 1 : }
|
