forked from influxdata/influxdb
-
Notifications
You must be signed in to change notification settings - Fork 0
/
file_store.go
1150 lines (961 loc) · 27.3 KB
/
file_store.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
package tsm1
import (
"fmt"
"io"
"io/ioutil"
"log"
"math"
"os"
"path/filepath"
"sort"
"strconv"
"strings"
"sync"
"sync/atomic"
"time"
"github.com/influxdata/influxdb/models"
)
type TSMFile interface {
// Path returns the underlying file path for the TSMFile. If the file
// has not be written or loaded from disk, the zero value is returned.
Path() string
// Read returns all the values in the block where time t resides
Read(key string, t int64) ([]Value, error)
// ReadAt returns all the values in the block identified by entry.
ReadAt(entry *IndexEntry, values []Value) ([]Value, error)
ReadFloatBlockAt(entry *IndexEntry, tdec *TimeDecoder, vdec *FloatDecoder, values *[]FloatValue) ([]FloatValue, error)
ReadIntegerBlockAt(entry *IndexEntry, tdec *TimeDecoder, vdec *IntegerDecoder, values *[]IntegerValue) ([]IntegerValue, error)
ReadStringBlockAt(entry *IndexEntry, tdec *TimeDecoder, vdec *StringDecoder, values *[]StringValue) ([]StringValue, error)
ReadBooleanBlockAt(entry *IndexEntry, tdec *TimeDecoder, vdec *BooleanDecoder, values *[]BooleanValue) ([]BooleanValue, error)
// Entries returns the index entries for all blocks for the given key.
Entries(key string) []IndexEntry
ReadEntries(key string, entries *[]IndexEntry)
// Returns true if the TSMFile may contain a value with the specified
// key and time
ContainsValue(key string, t int64) bool
// Contains returns true if the file contains any values for the given
// key.
Contains(key string) bool
// TimeRange returns the min and max time across all keys in the file.
TimeRange() (int64, int64)
// TombstoneRange returns ranges of time that are deleted for the given key.
TombstoneRange(key string) []TimeRange
// KeyRange returns the min and max keys in the file.
KeyRange() (string, string)
// KeyCount returns the number of distict keys in the file.
KeyCount() int
// KeyAt returns the key located at index position idx
KeyAt(idx int) (string, byte)
// Type returns the block type of the values stored for the key. Returns one of
// BlockFloat64, BlockInt64, BlockBoolean, BlockString. If key does not exist,
// an error is returned.
Type(key string) (byte, error)
// Delete removes the keys from the set of keys available in this file.
Delete(keys []string) error
// DeleteRange removes the values for keys between min and max.
DeleteRange(keys []string, min, max int64) error
// HasTombstones returns true if file contains values that have been deleted.
HasTombstones() bool
// TombstoneFiles returns the tombstone filestats if there are any tombstones
// written for this file.
TombstoneFiles() []FileStat
// Close the underlying file resources
Close() error
// Size returns the size of the file on disk in bytes.
Size() uint32
// Rename renames the existing TSM file to a new name and replaces the mmap backing slice using the new
// file name. Index and Reader state are not re-initialized.
Rename(path string) error
// Remove deletes the file from the filesystem
Remove() error
// Returns true if the file is currently in use by queries
InUse() bool
// Ref records that this file is actively in use
Ref()
// Unref records that this file is no longer in user
Unref()
// Stats returns summary information about the TSM file.
Stats() FileStat
// BlockIterator returns an iterator pointing to the first block in the file and
// allows sequential iteration to each every block.
BlockIterator() *BlockIterator
}
// Statistics gathered by the FileStore.
const (
statFileStoreBytes = "diskBytes"
statFileStoreCount = "numFiles"
)
type FileStore struct {
mu sync.RWMutex
lastModified time.Time
currentGeneration int
dir string
files []TSMFile
logger *log.Logger // Logger to be used for important messages
traceLogger *log.Logger // Logger to be used when trace-logging is on.
logOutput io.Writer // Writer to be logger and traceLogger if active.
traceLogging bool
stats *FileStoreStatistics
purger *purger
currentTempDirID int
}
type FileStat struct {
Path string
HasTombstone bool
Size uint32
LastModified int64
MinTime, MaxTime int64
MinKey, MaxKey string
}
func (f FileStat) OverlapsTimeRange(min, max int64) bool {
return f.MinTime <= max && f.MaxTime >= min
}
func (f FileStat) OverlapsKeyRange(min, max string) bool {
return min != "" && max != "" && f.MinKey <= max && f.MaxKey >= min
}
func (f FileStat) ContainsKey(key string) bool {
return f.MinKey >= key || key <= f.MaxKey
}
func NewFileStore(dir string) *FileStore {
logger := log.New(os.Stderr, "[filestore] ", log.LstdFlags)
return &FileStore{
dir: dir,
lastModified: time.Now(),
logger: logger,
traceLogger: log.New(ioutil.Discard, "[filestore] ", log.LstdFlags),
logOutput: os.Stderr,
stats: &FileStoreStatistics{},
purger: &purger{
files: map[string]TSMFile{},
logger: logger,
},
}
}
// enableTraceLogging must be called before the FileStore is opened.
func (f *FileStore) enableTraceLogging(enabled bool) {
f.traceLogging = enabled
if enabled {
f.traceLogger.SetOutput(f.logOutput)
}
}
// SetLogOutput sets the logger used for all messages. It is safe for concurrent
// use.
func (f *FileStore) SetLogOutput(w io.Writer) {
f.logger.SetOutput(w)
// Set the trace logger's output only if trace logging is enabled.
if f.traceLogging {
f.traceLogger.SetOutput(w)
}
f.mu.Lock()
f.logOutput = w
f.mu.Unlock()
}
// FileStoreStatistics keeps statistics about the file store.
type FileStoreStatistics struct {
DiskBytes int64
FileCount int64
}
// Statistics returns statistics for periodic monitoring.
func (f *FileStore) Statistics(tags map[string]string) []models.Statistic {
return []models.Statistic{{
Name: "tsm1_filestore",
Tags: tags,
Values: map[string]interface{}{
statFileStoreBytes: atomic.LoadInt64(&f.stats.DiskBytes),
statFileStoreCount: atomic.LoadInt64(&f.stats.FileCount),
},
}}
}
// Returns the number of TSM files currently loaded
func (f *FileStore) Count() int {
f.mu.RLock()
defer f.mu.RUnlock()
return len(f.files)
}
// Files returns TSM files currently loaded.
func (f *FileStore) Files() []TSMFile {
f.mu.RLock()
defer f.mu.RUnlock()
return f.files
}
// CurrentGeneration returns the current generation of the TSM files
func (f *FileStore) CurrentGeneration() int {
f.mu.RLock()
defer f.mu.RUnlock()
return f.currentGeneration
}
// NextGeneration returns the max file ID + 1
func (f *FileStore) NextGeneration() int {
f.mu.Lock()
defer f.mu.Unlock()
f.currentGeneration++
return f.currentGeneration
}
func (f *FileStore) Add(files ...TSMFile) {
f.mu.Lock()
defer f.mu.Unlock()
for _, file := range files {
atomic.AddInt64(&f.stats.DiskBytes, int64(file.Size()))
}
f.files = append(f.files, files...)
sort.Sort(tsmReaders(f.files))
atomic.StoreInt64(&f.stats.FileCount, int64(len(f.files)))
}
// Remove removes the files with matching paths from the set of active files. It does
// not remove the paths from disk.
func (f *FileStore) Remove(paths ...string) {
f.mu.Lock()
defer f.mu.Unlock()
var active []TSMFile
for _, file := range f.files {
keep := true
for _, remove := range paths {
if remove == file.Path() {
keep = false
break
}
}
if keep {
active = append(active, file)
} else {
// Removing the file, remove the file size from the total file store bytes
atomic.AddInt64(&f.stats.DiskBytes, -int64(file.Size()))
}
}
f.files = active
sort.Sort(tsmReaders(f.files))
atomic.StoreInt64(&f.stats.FileCount, int64(len(f.files)))
}
// WalkKeys calls fn for every key in every TSM file known to the FileStore. If the key
// exists in multiple files, it will be invoked for each file.
func (f *FileStore) WalkKeys(fn func(key string, typ byte) error) error {
f.mu.RLock()
defer f.mu.RUnlock()
for _, f := range f.files {
for i := 0; i < f.KeyCount(); i++ {
key, typ := f.KeyAt(i)
if err := fn(key, typ); err != nil {
return err
}
}
}
return nil
}
// Keys returns all keys and types for all files
func (f *FileStore) Keys() map[string]byte {
f.mu.RLock()
defer f.mu.RUnlock()
uniqueKeys := map[string]byte{}
for _, f := range f.files {
for i := 0; i < f.KeyCount(); i++ {
key, typ := f.KeyAt(i)
uniqueKeys[key] = typ
}
}
return uniqueKeys
}
func (f *FileStore) Type(key string) (byte, error) {
f.mu.RLock()
defer f.mu.RUnlock()
for _, f := range f.files {
if f.Contains(key) {
return f.Type(key)
}
}
return 0, fmt.Errorf("unknown type for %v", key)
}
func (f *FileStore) Delete(keys []string) error {
return f.DeleteRange(keys, math.MinInt64, math.MaxInt64)
}
// DeleteRange removes the values for keys between min and max.
func (f *FileStore) DeleteRange(keys []string, min, max int64) error {
f.mu.Lock()
f.lastModified = time.Now()
f.mu.Unlock()
return f.walkFiles(func(tsm TSMFile) error {
return tsm.DeleteRange(keys, min, max)
})
}
func (f *FileStore) Open() error {
f.mu.Lock()
defer f.mu.Unlock()
// Not loading files from disk so nothing to do
if f.dir == "" {
return nil
}
// find the current max ID for temp directories
tmpfiles, err := ioutil.ReadDir(f.dir)
if err != nil {
return err
}
for _, fi := range tmpfiles {
if fi.IsDir() && strings.HasSuffix(fi.Name(), ".tmp") {
ss := strings.Split(filepath.Base(fi.Name()), ".")
if len(ss) == 2 {
if i, err := strconv.Atoi(ss[0]); err != nil {
if i > f.currentTempDirID {
f.currentTempDirID = i
}
}
}
}
}
files, err := filepath.Glob(filepath.Join(f.dir, fmt.Sprintf("*.%s", TSMFileExtension)))
if err != nil {
return err
}
// struct to hold the result of opening each reader in a goroutine
type res struct {
r *TSMReader
err error
}
readerC := make(chan *res)
for i, fn := range files {
// Keep track of the latest ID
generation, _, err := ParseTSMFileName(fn)
if err != nil {
return err
}
if generation >= f.currentGeneration {
f.currentGeneration = generation + 1
}
file, err := os.OpenFile(fn, os.O_RDONLY, 0666)
if err != nil {
return fmt.Errorf("error opening file %s: %v", fn, err)
}
// Accumulate file store size stat
if fi, err := file.Stat(); err == nil {
atomic.AddInt64(&f.stats.DiskBytes, fi.Size())
}
go func(idx int, file *os.File) {
start := time.Now()
df, err := NewTSMReader(file)
f.logger.Printf("%s (#%d) opened in %v", file.Name(), idx, time.Now().Sub(start))
if err != nil {
readerC <- &res{r: df, err: fmt.Errorf("error opening memory map for file %s: %v", file.Name(), err)}
return
}
readerC <- &res{r: df}
}(i, file)
}
for range files {
res := <-readerC
if res.err != nil {
return res.err
}
f.files = append(f.files, res.r)
}
close(readerC)
sort.Sort(tsmReaders(f.files))
atomic.StoreInt64(&f.stats.FileCount, int64(len(f.files)))
return nil
}
func (f *FileStore) Close() error {
f.mu.Lock()
defer f.mu.Unlock()
for _, f := range f.files {
f.Close()
}
f.files = nil
atomic.StoreInt64(&f.stats.FileCount, 0)
return nil
}
func (f *FileStore) Read(key string, t int64) ([]Value, error) {
f.mu.RLock()
defer f.mu.RUnlock()
for _, f := range f.files {
// Can this file possibly contain this key and timestamp?
if !f.Contains(key) {
continue
}
// May have the key and time we are looking for so try to find
v, err := f.Read(key, t)
if err != nil {
return nil, err
}
if len(v) > 0 {
return v, nil
}
}
return nil, nil
}
func (f *FileStore) KeyCursor(key string, t int64, ascending bool) *KeyCursor {
f.mu.RLock()
defer f.mu.RUnlock()
return newKeyCursor(f, key, t, ascending)
}
func (f *FileStore) Stats() []FileStat {
f.mu.RLock()
defer f.mu.RUnlock()
stats := make([]FileStat, len(f.files))
for i, fd := range f.files {
stats[i] = fd.Stats()
}
return stats
}
func (f *FileStore) Replace(oldFiles, newFiles []string) error {
f.mu.Lock()
defer f.mu.Unlock()
f.lastModified = time.Now()
// Copy the current set of active files while we rename
// and load the new files. We copy the pointers here to minimize
// the time that locks are held as well as to ensure that the replacement
// is atomic.©
var updated []TSMFile
for _, t := range f.files {
updated = append(updated, t)
}
// Rename all the new files to make them live on restart
for _, file := range newFiles {
var newName = file
if strings.HasSuffix(file, ".tmp") {
// The new TSM files have a tmp extension. First rename them.
newName = file[:len(file)-4]
if err := os.Rename(file, newName); err != nil {
return err
}
}
fd, err := os.Open(newName)
if err != nil {
return err
}
tsm, err := NewTSMReader(fd)
if err != nil {
return err
}
updated = append(updated, tsm)
}
// We need to prune our set of active files now
var active, inuse []TSMFile
for _, file := range updated {
keep := true
for _, remove := range oldFiles {
if remove == file.Path() {
keep = false
// If queries are running against this file, then we need to move it out of the
// way and let them complete. We'll then delete the original file to avoid
// blocking callers upstream. If the process crashes, the temp file is
// cleaned up at startup automatically.
if file.InUse() {
// Copy all the tombstones related to this TSM file
var deletes []string
for _, t := range file.TombstoneFiles() {
deletes = append(deletes, t.Path)
}
deletes = append(deletes, file.Path())
// Rename the TSM file used by this reader
tempPath := file.Path() + ".tmp"
if err := file.Rename(tempPath); err != nil {
return err
}
// Remove the old file and tombstones. We can't use the normal TSMReader.Remove()
// because it now refers to our temp file which we can't remove.
for _, f := range deletes {
if err := os.RemoveAll(f); err != nil {
return err
}
}
inuse = append(inuse, file)
continue
}
if err := file.Close(); err != nil {
return err
}
if err := file.Remove(); err != nil {
return err
}
break
}
}
if keep {
active = append(active, file)
}
}
if err := syncDir(f.dir); err != nil {
return err
}
// Tell the purger about our in-use files we need to remove
f.purger.add(inuse)
f.files = active
sort.Sort(tsmReaders(f.files))
atomic.StoreInt64(&f.stats.FileCount, int64(len(f.files)))
// Recalculate the disk size stat
var totalSize int64
for _, file := range f.files {
totalSize += int64(file.Size())
}
atomic.StoreInt64(&f.stats.DiskBytes, totalSize)
return nil
}
// LastModified returns the last time the file store was updated with new
// TSM files or a delete
func (f *FileStore) LastModified() time.Time {
f.mu.RLock()
defer f.mu.RUnlock()
return f.lastModified
}
// BlockCount returns number of values stored in the block at location idx
// in the file at path. If path does not match any file in the store, 0 is
// returned. If idx is out of range for the number of blocks in the file,
// 0 is returned.
func (f *FileStore) BlockCount(path string, idx int) int {
f.mu.RLock()
defer f.mu.RUnlock()
if idx < 0 {
return 0
}
for _, fd := range f.files {
if fd.Path() == path {
iter := fd.BlockIterator()
for i := 0; i < idx; i++ {
if !iter.Next() {
return 0
}
}
_, _, _, _, block, _ := iter.Read()
return BlockCount(block)
}
}
return 0
}
// walkFiles calls fn for every files in filestore in parallel
func (f *FileStore) walkFiles(fn func(f TSMFile) error) error {
f.mu.RLock()
defer f.mu.RUnlock()
// struct to hold the result of opening each reader in a goroutine
errC := make(chan error, len(f.files))
for _, f := range f.files {
go func(tsm TSMFile) {
if err := fn(tsm); err != nil {
errC <- fmt.Errorf("file %s: %s", tsm.Path(), err)
return
}
errC <- nil
}(f)
}
for i := 0; i < cap(errC); i++ {
res := <-errC
if res != nil {
return res
}
}
return nil
}
// locations returns the files and index blocks for a key and time. ascending indicates
// whether the key will be scan in ascending time order or descenging time order.
// This function assumes the read-lock has been taken.
func (f *FileStore) locations(key string, t int64, ascending bool) []*location {
var locations []*location
filesSnapshot := make([]TSMFile, len(f.files))
for i := range f.files {
filesSnapshot[i] = f.files[i]
}
var entries []IndexEntry
for _, fd := range filesSnapshot {
minTime, maxTime := fd.TimeRange()
tombstones := fd.TombstoneRange(key)
// If we ascending and the max time of the file is before where we want to start
// skip it.
if ascending && maxTime < t {
continue
// If we are descending and the min time of the file is after where we want to start,
// then skip it.
} else if !ascending && minTime > t {
continue
}
// This file could potential contain points we are looking for so find the blocks for
// the given key.
fd.ReadEntries(key, &entries)
for _, ie := range entries {
// Skip any blocks only contain values that are tombstoned.
var skip bool
for _, t := range tombstones {
if t.Min <= ie.MinTime && t.Max >= ie.MaxTime {
skip = true
break
}
}
if skip {
continue
}
// If we ascending and the max time of a block is before where we are looking, skip
// it since the data is out of our range
if ascending && ie.MaxTime < t {
continue
// If we descending and the min time of a block is after where we are looking, skip
// it since the data is out of our range
} else if !ascending && ie.MinTime > t {
continue
}
location := &location{
r: fd,
entry: ie,
}
if ascending {
// For an ascending cursor, mark everything before the seek time as read
// so we can filter it out at query time
location.readMin = math.MinInt64
location.readMax = t - 1
} else {
// For an ascending cursort, mark everything after the seek time as read
// so we can filter it out at query time
location.readMin = t + 1
location.readMax = math.MaxInt64
}
// Otherwise, add this file and block location
locations = append(locations, location)
}
}
return locations
}
// CreateSnapshot will create hardlinks for all tsm and tombstone files
// in the path provided
func (f *FileStore) CreateSnapshot() (string, error) {
f.traceLogger.Printf("Creating snapshot in %s", f.dir)
files := f.Files()
f.mu.Lock()
f.currentTempDirID += 1
f.mu.Unlock()
f.mu.RLock()
defer f.mu.RUnlock()
// get a tmp directory name
tmpPath := fmt.Sprintf("%s/%d.tmp", f.dir, f.currentTempDirID)
err := os.Mkdir(tmpPath, 0777)
if err != nil {
return "", err
}
for _, tsmf := range files {
newpath := filepath.Join(tmpPath, filepath.Base(tsmf.Path()))
if err := os.Link(tsmf.Path(), newpath); err != nil {
return "", fmt.Errorf("error creating tsm hard link: %q", err)
}
// Check for tombstones and link those as well
for _, tf := range tsmf.TombstoneFiles() {
newpath := filepath.Join(tmpPath, filepath.Base(tf.Path))
if err := os.Link(tf.Path, newpath); err != nil {
return "", fmt.Errorf("error creating tombstone hard link: %q", err)
}
}
}
return tmpPath, nil
}
// ParseTSMFileName parses the generation and sequence from a TSM file name.
func ParseTSMFileName(name string) (int, int, error) {
base := filepath.Base(name)
idx := strings.Index(base, ".")
if idx == -1 {
return 0, 0, fmt.Errorf("file %s is named incorrectly", name)
}
id := base[:idx]
parts := strings.Split(id, "-")
if len(parts) != 2 {
return 0, 0, fmt.Errorf("file %s is named incorrectly", name)
}
generation, err := strconv.ParseUint(parts[0], 10, 32)
sequence, err := strconv.ParseUint(parts[1], 10, 32)
return int(generation), int(sequence), err
}
type KeyCursor struct {
key string
fs *FileStore
// seeks is all the file locations that we need to return during iteration.
seeks []*location
// current is the set of blocks possibly containing the next set of points.
// Normally this is just one entry, but there may be multiple if points have
// been overwritten.
current []*location
buf []Value
// pos is the index within seeks. Based on ascending, it will increment or
// decrement through the size of seeks slice.
pos int
ascending bool
// duplicates is a hint that there are overlapping blocks for this key in
// multiple files (e.g. points have been overwritten but not fully compacted)
// If this is true, we need to scan the duplicate blocks and dedup the points
// as query time until they are compacted.
duplicates bool
// The distinct set of TSM files references by the cursor
refs map[string]TSMFile
}
type location struct {
r TSMFile
entry IndexEntry
readMin, readMax int64
}
func (l *location) read() bool {
return l.readMin <= l.entry.MinTime && l.readMax >= l.entry.MaxTime
}
func (l *location) markRead(min, max int64) {
if min < l.readMin {
l.readMin = min
}
if max > l.readMax {
l.readMax = max
}
}
type descLocations []*location
// Sort methods
func (a descLocations) Len() int { return len(a) }
func (a descLocations) Swap(i, j int) { a[i], a[j] = a[j], a[i] }
func (a descLocations) Less(i, j int) bool {
if a[i].entry.OverlapsTimeRange(a[j].entry.MinTime, a[j].entry.MaxTime) {
return a[i].r.Path() < a[j].r.Path()
}
return a[i].entry.MaxTime < a[j].entry.MaxTime
}
type ascLocations []*location
// Sort methods
func (a ascLocations) Len() int { return len(a) }
func (a ascLocations) Swap(i, j int) { a[i], a[j] = a[j], a[i] }
func (a ascLocations) Less(i, j int) bool {
if a[i].entry.OverlapsTimeRange(a[j].entry.MinTime, a[j].entry.MaxTime) {
return a[i].r.Path() < a[j].r.Path()
}
return a[i].entry.MinTime < a[j].entry.MinTime
}
// newKeyCursor returns a new instance of KeyCursor.
// This function assumes the read-lock has been taken.
func newKeyCursor(fs *FileStore, key string, t int64, ascending bool) *KeyCursor {
c := &KeyCursor{
key: key,
fs: fs,
seeks: fs.locations(key, t, ascending),
ascending: ascending,
refs: map[string]TSMFile{},
}
c.duplicates = c.hasOverlappingBlocks()
if ascending {
sort.Sort(ascLocations(c.seeks))
} else {
sort.Sort(descLocations(c.seeks))
}
// Determine the distinct set of TSM files in use and mark then as in-use
for _, f := range c.seeks {
if _, ok := c.refs[f.r.Path()]; !ok {
f.r.Ref()
c.refs[f.r.Path()] = f.r
}
}
c.seek(t)
return c
}
// Close removes all references on the cursor.
func (c *KeyCursor) Close() {
// Remove all of our in-use references since we're done
for _, f := range c.refs {
f.Unref()
}
c.buf = nil
c.seeks = nil
c.fs = nil
c.current = nil
}
// hasOverlappingBlocks returns true if blocks have overlapping time ranges.
// This result is computed once and stored as the "duplicates" field.
func (c *KeyCursor) hasOverlappingBlocks() bool {
if len(c.seeks) == 0 {
return false
}
for i := 1; i < len(c.seeks); i++ {
prev := c.seeks[i-1]
cur := c.seeks[i]
if prev.entry.MaxTime >= cur.entry.MinTime {
return true
}
}
return false
}
// seek positions the cursor at the given time.
func (c *KeyCursor) seek(t int64) {
if len(c.seeks) == 0 {
return
}
c.current = nil
if c.ascending {
c.seekAscending(t)
} else {
c.seekDescending(t)
}
}
func (c *KeyCursor) seekAscending(t int64) {
for i, e := range c.seeks {
if t < e.entry.MinTime || e.entry.Contains(t) {
// Record the position of the first block matching our seek time
if len(c.current) == 0 {
c.pos = i
}
c.current = append(c.current, e)
// Exit if we don't have duplicates.
// Otherwise, keep looking for additional blocks containing this point.
if !c.duplicates {
return
}
}
}
}
func (c *KeyCursor) seekDescending(t int64) {
for i := len(c.seeks) - 1; i >= 0; i-- {
e := c.seeks[i]
if t > e.entry.MaxTime || e.entry.Contains(t) {
// Record the position of the first block matching our seek time
if len(c.current) == 0 {
c.pos = i
}
c.current = append(c.current, e)
// Exit if we don't have duplicates.
// Otherwise, keep looking for additional blocks containing this point.
if !c.duplicates {
return
}
}
}
}
// Next moves the cursor to the next position.
// Data should be read by the ReadBlock functions.
func (c *KeyCursor) Next() {
if len(c.current) == 0 {
return
}
// Do we still have unread values in the current block
if !c.current[0].read() {
return
}
c.current = c.current[:0]
if c.ascending {
c.nextAscending()
} else {
c.nextDescending()
}
}
func (c *KeyCursor) nextAscending() {
for {
c.pos++
if c.pos >= len(c.seeks) {