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 tool
6 :
7 : import (
8 : "bytes"
9 : "cmp"
10 : "fmt"
11 : "io"
12 : "path/filepath"
13 : "slices"
14 : "sort"
15 :
16 : "github.com/cockroachdb/pebble"
17 : "github.com/cockroachdb/pebble/internal/base"
18 : "github.com/cockroachdb/pebble/internal/keyspan"
19 : "github.com/cockroachdb/pebble/internal/manifest"
20 : "github.com/cockroachdb/pebble/internal/private"
21 : "github.com/cockroachdb/pebble/internal/rangedel"
22 : "github.com/cockroachdb/pebble/record"
23 : "github.com/cockroachdb/pebble/sstable"
24 : "github.com/cockroachdb/pebble/wal"
25 : "github.com/spf13/cobra"
26 : )
27 :
28 : type findRef struct {
29 : key base.InternalKey
30 : value []byte
31 : fileNum base.FileNum
32 : filename string
33 : }
34 :
35 : // findT implements the find tool.
36 : //
37 : // TODO(bananabrick): Add support for virtual sstables in this tool. Currently,
38 : // the tool will work because we're parsing files from disk, so virtual sstables
39 : // will never be added to findT.tables. The manifest could contain information
40 : // about virtual sstables. This is fine because the manifest is only used to
41 : // compute the findT.editRefs, and editRefs is only used if a file in
42 : // findT.tables contains a key. Of course, the tool won't be completely
43 : // accurate without dealing with virtual sstable case.
44 : type findT struct {
45 : Root *cobra.Command
46 :
47 : // Configuration.
48 : opts *pebble.Options
49 : comparers sstable.Comparers
50 : mergers sstable.Mergers
51 :
52 : // Flags.
53 : comparerName string
54 : fmtKey keyFormatter
55 : fmtValue valueFormatter
56 : verbose bool
57 :
58 : // Map from file num to version edit index which references the file num.
59 : editRefs map[base.DiskFileNum][]int
60 : // List of version edits.
61 : edits []manifest.VersionEdit
62 : // List of WAL files sorted by disk file num.
63 : logs []wal.LogicalLog
64 : // List of manifest files sorted by disk file num.
65 : manifests []fileLoc
66 : // List of table files sorted by disk file num.
67 : tables []fileLoc
68 : // Set of tables that contains references to the search key.
69 : tableRefs map[base.FileNum]bool
70 : // Map from file num to table metadata.
71 : tableMeta map[base.FileNum]*manifest.FileMetadata
72 : // List of error messages for SSTables that could not be decoded.
73 : errors []string
74 : }
75 :
76 : type fileLoc struct {
77 : base.DiskFileNum
78 : path string
79 : }
80 :
81 1 : func cmpFileLoc(a, b fileLoc) int {
82 1 : return cmp.Compare(a.DiskFileNum, b.DiskFileNum)
83 1 : }
84 :
85 : func newFind(
86 : opts *pebble.Options,
87 : comparers sstable.Comparers,
88 : defaultComparer string,
89 : mergers sstable.Mergers,
90 1 : ) *findT {
91 1 : f := &findT{
92 1 : opts: opts,
93 1 : comparers: comparers,
94 1 : mergers: mergers,
95 1 : }
96 1 : f.fmtKey.mustSet("quoted")
97 1 : f.fmtValue.mustSet("[%x]")
98 1 :
99 1 : f.Root = &cobra.Command{
100 1 : Use: "find <dir> <key>",
101 1 : Short: "find references to the specified key",
102 1 : Long: `
103 1 : Find references to the specified key and any range tombstones that contain the
104 1 : key. This includes references to the key in WAL files and sstables, and the
105 1 : provenance of the sstables (flushed, ingested, compacted).
106 1 : `,
107 1 : Args: cobra.ExactArgs(2),
108 1 : Run: f.run,
109 1 : }
110 1 :
111 1 : f.Root.Flags().BoolVarP(
112 1 : &f.verbose, "verbose", "v", false, "verbose output")
113 1 : f.Root.Flags().StringVar(
114 1 : &f.comparerName, "comparer", defaultComparer, "comparer name")
115 1 : f.Root.Flags().Var(
116 1 : &f.fmtKey, "key", "key formatter")
117 1 : f.Root.Flags().Var(
118 1 : &f.fmtValue, "value", "value formatter")
119 1 : return f
120 1 : }
121 :
122 1 : func (f *findT) run(cmd *cobra.Command, args []string) {
123 1 : stdout, stderr := cmd.OutOrStdout(), cmd.OutOrStderr()
124 1 : var key key
125 1 : if err := key.Set(args[1]); err != nil {
126 0 : fmt.Fprintf(stdout, "%s\n", err)
127 0 : return
128 0 : }
129 :
130 1 : if err := f.findFiles(stdout, stderr, args[0]); err != nil {
131 1 : fmt.Fprintf(stdout, "%s\n", err)
132 1 : return
133 1 : }
134 1 : f.readManifests(stdout)
135 1 :
136 1 : f.opts.Comparer = f.comparers[f.comparerName]
137 1 : if f.opts.Comparer == nil {
138 0 : fmt.Fprintf(stderr, "unknown comparer %q", f.comparerName)
139 0 : return
140 0 : }
141 1 : f.fmtKey.setForComparer(f.opts.Comparer.Name, f.comparers)
142 1 : f.fmtValue.setForComparer(f.opts.Comparer.Name, f.comparers)
143 1 :
144 1 : refs := f.search(stdout, key)
145 1 : var lastFilename string
146 1 : for i := range refs {
147 1 : r := &refs[i]
148 1 : if lastFilename != r.filename {
149 1 : lastFilename = r.filename
150 1 : fmt.Fprintf(stdout, "%s", r.filename)
151 1 : if m := f.tableMeta[r.fileNum]; m != nil {
152 1 : fmt.Fprintf(stdout, " ")
153 1 : formatKeyRange(stdout, f.fmtKey, &m.Smallest, &m.Largest)
154 1 : }
155 1 : fmt.Fprintf(stdout, "\n")
156 1 : if p := f.tableProvenance(r.fileNum); p != "" {
157 1 : fmt.Fprintf(stdout, " (%s)\n", p)
158 1 : }
159 : }
160 1 : fmt.Fprintf(stdout, " ")
161 1 : formatKeyValue(stdout, f.fmtKey, f.fmtValue, &r.key, r.value)
162 : }
163 :
164 1 : for _, errorMsg := range f.errors {
165 1 : fmt.Fprint(stdout, errorMsg)
166 1 : }
167 : }
168 :
169 : // Find all of the manifests, logs, and tables in the specified directory.
170 1 : func (f *findT) findFiles(stdout, stderr io.Writer, dir string) error {
171 1 : f.editRefs = make(map[base.DiskFileNum][]int)
172 1 : f.logs = nil
173 1 : f.manifests = nil
174 1 : f.tables = nil
175 1 : f.tableMeta = make(map[base.FileNum]*manifest.FileMetadata)
176 1 :
177 1 : if _, err := f.opts.FS.Stat(dir); err != nil {
178 1 : return err
179 1 : }
180 :
181 1 : var walAccumulator wal.FileAccumulator
182 1 : walk(stderr, f.opts.FS, dir, func(path string) {
183 1 : if isLogFile, err := walAccumulator.MaybeAccumulate(f.opts.FS, path); err != nil {
184 0 : fmt.Fprintf(stderr, "%s: %v\n", path, err)
185 0 : return
186 1 : } else if isLogFile {
187 1 : return
188 1 : }
189 :
190 1 : ft, fileNum, ok := base.ParseFilename(f.opts.FS, path)
191 1 : if !ok {
192 1 : return
193 1 : }
194 1 : fl := fileLoc{DiskFileNum: fileNum, path: path}
195 1 : switch ft {
196 1 : case base.FileTypeManifest:
197 1 : f.manifests = append(f.manifests, fl)
198 1 : case base.FileTypeTable:
199 1 : f.tables = append(f.tables, fl)
200 1 : default:
201 1 : return
202 : }
203 : })
204 :
205 : // TODO(jackson): Provide a means of scanning the secondary WAL directory
206 : // too.
207 :
208 1 : f.logs = walAccumulator.Finish()
209 1 : slices.SortFunc(f.manifests, cmpFileLoc)
210 1 : slices.SortFunc(f.tables, cmpFileLoc)
211 1 :
212 1 : if f.verbose {
213 1 : fmt.Fprintf(stdout, "%s\n", dir)
214 1 : fmt.Fprintf(stdout, "%5d %s\n", len(f.manifests), makePlural("manifest", int64(len(f.manifests))))
215 1 : fmt.Fprintf(stdout, "%5d %s\n", len(f.logs), makePlural("log", int64(len(f.logs))))
216 1 : fmt.Fprintf(stdout, "%5d %s\n", len(f.tables), makePlural("sstable", int64(len(f.tables))))
217 1 : }
218 1 : return nil
219 : }
220 :
221 : // Read the manifests and populate the editRefs map which is used to determine
222 : // the provenance and metadata of tables.
223 1 : func (f *findT) readManifests(stdout io.Writer) {
224 1 : for _, fl := range f.manifests {
225 1 : func() {
226 1 : mf, err := f.opts.FS.Open(fl.path)
227 1 : if err != nil {
228 0 : fmt.Fprintf(stdout, "%s\n", err)
229 0 : return
230 0 : }
231 1 : defer mf.Close()
232 1 :
233 1 : if f.verbose {
234 1 : fmt.Fprintf(stdout, "%s\n", fl.path)
235 1 : }
236 :
237 1 : rr := record.NewReader(mf, 0 /* logNum */)
238 1 : for {
239 1 : r, err := rr.Next()
240 1 : if err != nil {
241 1 : if err != io.EOF {
242 0 : fmt.Fprintf(stdout, "%s: %s\n", fl.path, err)
243 0 : }
244 1 : break
245 : }
246 :
247 1 : var ve manifest.VersionEdit
248 1 : if err := ve.Decode(r); err != nil {
249 0 : fmt.Fprintf(stdout, "%s: %s\n", fl.path, err)
250 0 : break
251 : }
252 1 : i := len(f.edits)
253 1 : f.edits = append(f.edits, ve)
254 1 :
255 1 : if ve.ComparerName != "" {
256 1 : f.comparerName = ve.ComparerName
257 1 : }
258 1 : if num := ve.MinUnflushedLogNum; num != 0 {
259 1 : f.editRefs[num] = append(f.editRefs[num], i)
260 1 : }
261 1 : for df := range ve.DeletedFiles {
262 1 : diskFileNum := base.PhysicalTableDiskFileNum(df.FileNum)
263 1 : f.editRefs[diskFileNum] = append(f.editRefs[diskFileNum], i)
264 1 : }
265 1 : for _, nf := range ve.NewFiles {
266 1 : // The same file can be deleted and added in a single version edit
267 1 : // which indicates a "move" compaction. Only add the edit to the list
268 1 : // once.
269 1 : diskFileNum := base.PhysicalTableDiskFileNum(nf.Meta.FileNum)
270 1 : refs := f.editRefs[diskFileNum]
271 1 : if n := len(refs); n == 0 || refs[n-1] != i {
272 1 : f.editRefs[diskFileNum] = append(refs, i)
273 1 : }
274 1 : if _, ok := f.tableMeta[nf.Meta.FileNum]; !ok {
275 1 : f.tableMeta[nf.Meta.FileNum] = nf.Meta
276 1 : }
277 : }
278 : }
279 : }()
280 : }
281 :
282 1 : if f.verbose {
283 1 : fmt.Fprintf(stdout, "%5d %s\n", len(f.edits), makePlural("edit", int64(len(f.edits))))
284 1 : }
285 : }
286 :
287 : // Search the logs and sstables for references to the specified key.
288 1 : func (f *findT) search(stdout io.Writer, key []byte) []findRef {
289 1 : refs := f.searchLogs(stdout, key, nil)
290 1 : refs = f.searchTables(stdout, key, refs)
291 1 :
292 1 : // For a given file (log or table) the references are already in the correct
293 1 : // order. We simply want to order the references by fileNum using a stable
294 1 : // sort.
295 1 : //
296 1 : // TODO(peter): I'm not sure if this is perfectly correct with regards to log
297 1 : // files and ingested sstables, but it is close enough and doing something
298 1 : // better is onerous. Revisit if this ever becomes problematic (e.g. if we
299 1 : // allow finding more than one key at a time).
300 1 : //
301 1 : // An example of the problem with logs and ingestion (which can only occur
302 1 : // with distinct keys). If I write key "a" to a log, I can then ingested key
303 1 : // "b" without causing "a" to be flushed. Then I can write key "c" to the
304 1 : // log. Ideally, we'd show the key "a" from the log, then the key "b" from
305 1 : // the ingested sstable, then key "c" from the log.
306 1 : slices.SortStableFunc(refs, func(a, b findRef) int {
307 1 : if v := cmp.Compare(a.fileNum, b.fileNum); v != 0 {
308 1 : return v
309 1 : }
310 1 : return cmp.Compare(a.filename, b.filename)
311 : })
312 1 : return refs
313 : }
314 :
315 : // Search the logs for references to the specified key.
316 1 : func (f *findT) searchLogs(stdout io.Writer, searchKey []byte, refs []findRef) []findRef {
317 1 : cmp := f.opts.Comparer.Compare
318 1 : for _, ll := range f.logs {
319 1 : _ = func() (err error) {
320 1 : rr := ll.OpenForRead()
321 1 : if f.verbose {
322 1 : fmt.Fprintf(stdout, "%s", ll)
323 1 : defer fmt.Fprintf(stdout, "\n")
324 1 : }
325 1 : defer func() {
326 1 : switch err {
327 0 : case record.ErrZeroedChunk:
328 0 : if f.verbose {
329 0 : fmt.Fprintf(stdout, ": EOF [%s] (may be due to WAL preallocation)", err)
330 0 : }
331 0 : case record.ErrInvalidChunk:
332 0 : if f.verbose {
333 0 : fmt.Fprintf(stdout, ": EOF [%s] (may be due to WAL recycling)", err)
334 0 : }
335 1 : default:
336 1 : if err != io.EOF {
337 0 : if f.verbose {
338 0 : fmt.Fprintf(stdout, ": %s", err)
339 0 : } else {
340 0 : fmt.Fprintf(stdout, "%s: %s\n", ll, err)
341 0 : }
342 : }
343 : }
344 : }()
345 :
346 1 : var b pebble.Batch
347 1 : var buf bytes.Buffer
348 1 : for {
349 1 : r, off, err := rr.NextRecord()
350 1 : if err == nil {
351 1 : buf.Reset()
352 1 : _, err = io.Copy(&buf, r)
353 1 : }
354 1 : if err != nil {
355 1 : return err
356 1 : }
357 :
358 1 : b = pebble.Batch{}
359 1 : if err := b.SetRepr(buf.Bytes()); err != nil {
360 0 : fmt.Fprintf(stdout, "%s: corrupt log file: %v", ll, err)
361 0 : continue
362 : }
363 1 : seqNum := b.SeqNum()
364 1 : for r := b.Reader(); ; seqNum++ {
365 1 : kind, ukey, value, ok, err := r.Next()
366 1 : if !ok {
367 1 : if err != nil {
368 0 : fmt.Fprintf(stdout, "%s: corrupt log file: %v", ll, err)
369 0 : break
370 : }
371 1 : break
372 : }
373 1 : ikey := base.MakeInternalKey(ukey, seqNum, kind)
374 1 : switch kind {
375 : case base.InternalKeyKindDelete,
376 : base.InternalKeyKindDeleteSized,
377 : base.InternalKeyKindSet,
378 : base.InternalKeyKindMerge,
379 : base.InternalKeyKindSingleDelete,
380 1 : base.InternalKeyKindSetWithDelete:
381 1 : if cmp(searchKey, ikey.UserKey) != 0 {
382 1 : continue
383 : }
384 1 : case base.InternalKeyKindRangeDelete:
385 1 : // Output tombstones that contain or end with the search key.
386 1 : t := rangedel.Decode(ikey, value, nil)
387 1 : if !t.Contains(cmp, searchKey) && cmp(t.End, searchKey) != 0 {
388 1 : continue
389 : }
390 0 : default:
391 0 : continue
392 : }
393 :
394 1 : refs = append(refs, findRef{
395 1 : key: ikey.Clone(),
396 1 : value: append([]byte(nil), value...),
397 1 : fileNum: base.FileNum(ll.Num),
398 1 : filename: filepath.Base(off.PhysicalFile),
399 1 : })
400 : }
401 : }
402 : }()
403 : }
404 1 : return refs
405 : }
406 :
407 : // Search the tables for references to the specified key.
408 1 : func (f *findT) searchTables(stdout io.Writer, searchKey []byte, refs []findRef) []findRef {
409 1 : cache := pebble.NewCache(128 << 20 /* 128 MB */)
410 1 : defer cache.Unref()
411 1 :
412 1 : f.tableRefs = make(map[base.FileNum]bool)
413 1 : for _, fl := range f.tables {
414 1 : _ = func() (err error) {
415 1 : tf, err := f.opts.FS.Open(fl.path)
416 1 : if err != nil {
417 0 : fmt.Fprintf(stdout, "%s\n", err)
418 0 : return
419 0 : }
420 :
421 1 : m := f.tableMeta[base.PhysicalTableFileNum(fl.DiskFileNum)]
422 1 : if f.verbose {
423 1 : fmt.Fprintf(stdout, "%s", fl.path)
424 1 : if m != nil && m.SmallestSeqNum == m.LargestSeqNum {
425 1 : fmt.Fprintf(stdout, ": global seqnum: %d", m.LargestSeqNum)
426 1 : }
427 1 : defer fmt.Fprintf(stdout, "\n")
428 : }
429 1 : defer func() {
430 1 : switch {
431 0 : case err != nil:
432 0 : if f.verbose {
433 0 : fmt.Fprintf(stdout, ": %v", err)
434 0 : } else {
435 0 : fmt.Fprintf(stdout, "%s: %v\n", fl.path, err)
436 0 : }
437 : }
438 : }()
439 :
440 1 : opts := sstable.ReaderOptions{
441 1 : Cache: cache,
442 1 : Comparer: f.opts.Comparer,
443 1 : Filters: f.opts.Filters,
444 1 : }
445 1 : readable, err := sstable.NewSimpleReadable(tf)
446 1 : if err != nil {
447 0 : return err
448 0 : }
449 1 : r, err := sstable.NewReader(readable, opts, f.comparers, f.mergers,
450 1 : private.SSTableRawTombstonesOpt.(sstable.ReaderOption))
451 1 : if err != nil {
452 1 : f.errors = append(f.errors, fmt.Sprintf("Unable to decode sstable %s, %s", fl.path, err.Error()))
453 1 : // Ensure the error only gets printed once.
454 1 : err = nil
455 1 : return
456 1 : }
457 1 : defer r.Close()
458 1 :
459 1 : var transforms sstable.IterTransforms
460 1 : if m != nil {
461 1 : transforms = m.IterTransforms()
462 1 : }
463 :
464 1 : iter, err := r.NewIter(transforms, nil, nil)
465 1 : if err != nil {
466 0 : return err
467 0 : }
468 1 : defer iter.Close()
469 1 : kv := iter.SeekGE(searchKey, base.SeekGEFlagsNone)
470 1 :
471 1 : // We configured sstable.Reader to return raw tombstones which requires a
472 1 : // bit more work here to put them in a form that can be iterated in
473 1 : // parallel with the point records.
474 1 : rangeDelIter, err := func() (keyspan.FragmentIterator, error) {
475 1 : iter, err := r.NewRawRangeDelIter(transforms)
476 1 : if err != nil {
477 0 : return nil, err
478 0 : }
479 1 : if iter == nil {
480 1 : return keyspan.NewIter(r.Compare, nil), nil
481 1 : }
482 1 : defer iter.Close()
483 1 :
484 1 : var tombstones []keyspan.Span
485 1 : t, err := iter.First()
486 1 : for ; t != nil; t, err = iter.Next() {
487 1 : if !t.Contains(r.Compare, searchKey) {
488 1 : continue
489 : }
490 1 : tombstones = append(tombstones, t.ShallowClone())
491 : }
492 1 : if err != nil {
493 0 : return nil, err
494 0 : }
495 :
496 1 : slices.SortFunc(tombstones, func(a, b keyspan.Span) int {
497 0 : return r.Compare(a.Start, b.Start)
498 0 : })
499 1 : return keyspan.NewIter(r.Compare, tombstones), nil
500 : }()
501 1 : if err != nil {
502 0 : return err
503 0 : }
504 :
505 1 : defer rangeDelIter.Close()
506 1 : rangeDel, err := rangeDelIter.First()
507 1 : if err != nil {
508 0 : return err
509 0 : }
510 :
511 1 : foundRef := false
512 1 : for kv != nil || rangeDel != nil {
513 1 : if kv != nil &&
514 1 : (rangeDel == nil || r.Compare(kv.K.UserKey, rangeDel.Start) < 0) {
515 1 : if r.Compare(searchKey, kv.K.UserKey) != 0 {
516 1 : kv = nil
517 1 : continue
518 : }
519 1 : v, _, err := kv.Value(nil)
520 1 : if err != nil {
521 0 : return err
522 0 : }
523 1 : refs = append(refs, findRef{
524 1 : key: kv.K.Clone(),
525 1 : value: slices.Clone(v),
526 1 : fileNum: base.PhysicalTableFileNum(fl.DiskFileNum),
527 1 : filename: filepath.Base(fl.path),
528 1 : })
529 1 : kv = iter.Next()
530 1 : } else {
531 1 : // Use rangedel.Encode to add a reference for each key
532 1 : // within the span.
533 1 : err := rangedel.Encode(rangeDel, func(k base.InternalKey, v []byte) error {
534 1 : refs = append(refs, findRef{
535 1 : key: k.Clone(),
536 1 : value: slices.Clone(v),
537 1 : fileNum: base.PhysicalTableFileNum(fl.DiskFileNum),
538 1 : filename: filepath.Base(fl.path),
539 1 : })
540 1 : return nil
541 1 : })
542 1 : if err != nil {
543 0 : return err
544 0 : }
545 1 : rangeDel, err = rangeDelIter.Next()
546 1 : if err != nil {
547 0 : return err
548 0 : }
549 : }
550 1 : foundRef = true
551 : }
552 :
553 1 : if foundRef {
554 1 : f.tableRefs[base.PhysicalTableFileNum(fl.DiskFileNum)] = true
555 1 : }
556 1 : return nil
557 : }()
558 : }
559 1 : return refs
560 : }
561 :
562 : // Determine the provenance of the specified table. We search the version edits
563 : // for the first edit which created the table, and then analyze the edit to
564 : // determine if it was a compaction, flush, or ingestion. Returns an empty
565 : // string if the provenance of a table cannot be determined.
566 1 : func (f *findT) tableProvenance(fileNum base.FileNum) string {
567 1 : editRefs := f.editRefs[base.PhysicalTableDiskFileNum(fileNum)]
568 1 : for len(editRefs) > 0 {
569 1 : ve := f.edits[editRefs[0]]
570 1 : editRefs = editRefs[1:]
571 1 : for _, nf := range ve.NewFiles {
572 1 : if fileNum != nf.Meta.FileNum {
573 1 : continue
574 : }
575 :
576 1 : var buf bytes.Buffer
577 1 : switch {
578 1 : case len(ve.DeletedFiles) > 0:
579 1 : // A version edit with deleted files is a compaction. The deleted
580 1 : // files are the inputs to the compaction. We're going to
581 1 : // reconstruct the input files and display those inputs that
582 1 : // contain the search key (i.e. are list in refs) and use an
583 1 : // ellipsis to indicate when there were other inputs that have
584 1 : // been elided.
585 1 : var sourceLevels []int
586 1 : levels := make(map[int][]base.FileNum)
587 1 : for df := range ve.DeletedFiles {
588 1 : files := levels[df.Level]
589 1 : if len(files) == 0 {
590 1 : sourceLevels = append(sourceLevels, df.Level)
591 1 : }
592 1 : levels[df.Level] = append(files, df.FileNum)
593 : }
594 :
595 1 : sort.Ints(sourceLevels)
596 1 : if sourceLevels[len(sourceLevels)-1] != nf.Level {
597 0 : sourceLevels = append(sourceLevels, nf.Level)
598 0 : }
599 :
600 1 : sep := " "
601 1 : fmt.Fprintf(&buf, "compacted")
602 1 : for _, level := range sourceLevels {
603 1 : files := levels[level]
604 1 : slices.Sort(files)
605 1 :
606 1 : fmt.Fprintf(&buf, "%sL%d [", sep, level)
607 1 : sep = ""
608 1 : elided := false
609 1 : for _, fileNum := range files {
610 1 : if f.tableRefs[fileNum] {
611 1 : fmt.Fprintf(&buf, "%s%s", sep, fileNum)
612 1 : sep = " "
613 1 : } else {
614 1 : elided = true
615 1 : }
616 : }
617 1 : if elided {
618 1 : fmt.Fprintf(&buf, "%s...", sep)
619 1 : }
620 1 : fmt.Fprintf(&buf, "]")
621 1 : sep = " + "
622 : }
623 :
624 1 : case ve.MinUnflushedLogNum != 0:
625 1 : // A version edit with a min-unflushed log indicates a flush
626 1 : // operation.
627 1 : fmt.Fprintf(&buf, "flushed to L%d", nf.Level)
628 :
629 1 : case nf.Meta.SmallestSeqNum == nf.Meta.LargestSeqNum:
630 1 : // If the smallest and largest seqnum are the same, the file was
631 1 : // ingested. Note that this can also occur for a flushed sstable
632 1 : // that contains only a single key, though that would have
633 1 : // already been captured above.
634 1 : fmt.Fprintf(&buf, "ingested to L%d", nf.Level)
635 :
636 0 : default:
637 0 : // The provenance of the table is unclear. This is usually due to
638 0 : // the MANIFEST rolling over and taking a snapshot of the LSM
639 0 : // state.
640 0 : fmt.Fprintf(&buf, "added to L%d", nf.Level)
641 : }
642 :
643 : // After a table is created, it can be moved to a different level via a
644 : // move compaction. This is indicated by a version edit that deletes the
645 : // table from one level and adds the same table to a different
646 : // level. Loop over the remaining version edits for the table looking for
647 : // such moves.
648 1 : for len(editRefs) > 0 {
649 1 : ve := f.edits[editRefs[0]]
650 1 : editRefs = editRefs[1:]
651 1 : for _, nf := range ve.NewFiles {
652 1 : if fileNum == nf.Meta.FileNum {
653 1 : for df := range ve.DeletedFiles {
654 1 : if fileNum == df.FileNum {
655 1 : fmt.Fprintf(&buf, ", moved to L%d", nf.Level)
656 1 : break
657 : }
658 : }
659 1 : break
660 : }
661 : }
662 : }
663 :
664 1 : return buf.String()
665 : }
666 : }
667 1 : return ""
668 : }
|