-
Notifications
You must be signed in to change notification settings - Fork 9.7k
/
kvstore.go
710 lines (580 loc) · 16.2 KB
/
kvstore.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
// Copyright 2015 The etcd Authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package mvcc
import (
"encoding/binary"
"errors"
"math"
"math/rand"
"sync"
"time"
"github.com/coreos/etcd/lease"
"github.com/coreos/etcd/mvcc/backend"
"github.com/coreos/etcd/mvcc/mvccpb"
"github.com/coreos/etcd/pkg/schedule"
"github.com/coreos/pkg/capnslog"
"golang.org/x/net/context"
)
var (
keyBucketName = []byte("key")
metaBucketName = []byte("meta")
// markedRevBytesLen is the byte length of marked revision.
// The first `revBytesLen` bytes represents a normal revision. The last
// one byte is the mark.
markedRevBytesLen = revBytesLen + 1
markBytePosition = markedRevBytesLen - 1
markTombstone byte = 't'
consistentIndexKeyName = []byte("consistent_index")
scheduledCompactKeyName = []byte("scheduledCompactRev")
finishedCompactKeyName = []byte("finishedCompactRev")
ErrTxnIDMismatch = errors.New("mvcc: txn id mismatch")
ErrCompacted = errors.New("mvcc: required revision has been compacted")
ErrFutureRev = errors.New("mvcc: required revision is a future revision")
ErrCanceled = errors.New("mvcc: watcher is canceled")
plog = capnslog.NewPackageLogger("github.com/coreos/etcd", "mvcc")
)
// ConsistentIndexGetter is an interface that wraps the Get method.
// Consistent index is the offset of an entry in a consistent replicated log.
type ConsistentIndexGetter interface {
// ConsistentIndex returns the consistent index of current executing entry.
ConsistentIndex() uint64
}
type store struct {
mu sync.Mutex // guards the following
ig ConsistentIndexGetter
b backend.Backend
kvindex index
le lease.Lessor
currentRev revision
// the main revision of the last compaction
compactMainRev int64
tx backend.BatchTx
txnID int64 // tracks the current txnID to verify txn operations
txnModify bool
// bytesBuf8 is a byte slice of length 8
// to avoid a repetitive allocation in saveIndex.
bytesBuf8 []byte
changes []mvccpb.KeyValue
fifoSched schedule.Scheduler
stopc chan struct{}
}
// NewStore returns a new store. It is useful to create a store inside
// mvcc pkg. It should only be used for testing externally.
func NewStore(b backend.Backend, le lease.Lessor, ig ConsistentIndexGetter) *store {
s := &store{
b: b,
ig: ig,
kvindex: newTreeIndex(),
le: le,
currentRev: revision{main: 1},
compactMainRev: -1,
bytesBuf8: make([]byte, 8, 8),
fifoSched: schedule.NewFIFOScheduler(),
stopc: make(chan struct{}),
}
if s.le != nil {
s.le.SetRangeDeleter(s)
}
tx := s.b.BatchTx()
tx.Lock()
tx.UnsafeCreateBucket(keyBucketName)
tx.UnsafeCreateBucket(metaBucketName)
tx.Unlock()
s.b.ForceCommit()
if err := s.restore(); err != nil {
// TODO: return the error instead of panic here?
panic("failed to recover store from backend")
}
return s
}
func (s *store) Rev() int64 {
s.mu.Lock()
defer s.mu.Unlock()
return s.currentRev.main
}
func (s *store) FirstRev() int64 {
s.mu.Lock()
defer s.mu.Unlock()
return s.compactMainRev
}
func (s *store) Put(key, value []byte, lease lease.LeaseID) int64 {
id := s.TxnBegin()
s.put(key, value, lease)
s.txnEnd(id)
putCounter.Inc()
return int64(s.currentRev.main)
}
func (s *store) Range(key, end []byte, ro RangeOptions) (r *RangeResult, err error) {
id := s.TxnBegin()
kvs, count, rev, err := s.rangeKeys(key, end, ro.Limit, ro.Rev, ro.Count)
s.txnEnd(id)
rangeCounter.Inc()
r = &RangeResult{
KVs: kvs,
Count: count,
Rev: rev,
}
return r, err
}
func (s *store) DeleteRange(key, end []byte) (n, rev int64) {
id := s.TxnBegin()
n = s.deleteRange(key, end)
s.txnEnd(id)
deleteCounter.Inc()
return n, int64(s.currentRev.main)
}
func (s *store) TxnBegin() int64 {
s.mu.Lock()
s.currentRev.sub = 0
s.tx = s.b.BatchTx()
s.tx.Lock()
s.txnID = rand.Int63()
return s.txnID
}
func (s *store) TxnEnd(txnID int64) error {
err := s.txnEnd(txnID)
if err != nil {
return err
}
txnCounter.Inc()
return nil
}
// txnEnd is used for unlocking an internal txn. It does
// not increase the txnCounter.
func (s *store) txnEnd(txnID int64) error {
if txnID != s.txnID {
return ErrTxnIDMismatch
}
// only update index if the txn modifies the mvcc state.
// read only txn might execute with one write txn concurrently,
// it should not write its index to mvcc.
if s.txnModify {
s.saveIndex()
}
s.txnModify = false
s.tx.Unlock()
if s.currentRev.sub != 0 {
s.currentRev.main += 1
}
s.currentRev.sub = 0
dbTotalSize.Set(float64(s.b.Size()))
s.mu.Unlock()
return nil
}
func (s *store) TxnRange(txnID int64, key, end []byte, ro RangeOptions) (r *RangeResult, err error) {
if txnID != s.txnID {
return nil, ErrTxnIDMismatch
}
kvs, count, rev, err := s.rangeKeys(key, end, ro.Limit, ro.Rev, ro.Count)
r = &RangeResult{
KVs: kvs,
Count: count,
Rev: rev,
}
return r, err
}
func (s *store) TxnPut(txnID int64, key, value []byte, lease lease.LeaseID) (rev int64, err error) {
if txnID != s.txnID {
return 0, ErrTxnIDMismatch
}
s.put(key, value, lease)
return int64(s.currentRev.main + 1), nil
}
func (s *store) TxnDeleteRange(txnID int64, key, end []byte) (n, rev int64, err error) {
if txnID != s.txnID {
return 0, 0, ErrTxnIDMismatch
}
n = s.deleteRange(key, end)
if n != 0 || s.currentRev.sub != 0 {
rev = int64(s.currentRev.main + 1)
} else {
rev = int64(s.currentRev.main)
}
return n, rev, nil
}
func (s *store) compactBarrier(ctx context.Context, ch chan struct{}) {
if ctx == nil || ctx.Err() != nil {
s.mu.Lock()
select {
case <-s.stopc:
default:
f := func(ctx context.Context) { s.compactBarrier(ctx, ch) }
s.fifoSched.Schedule(f)
}
s.mu.Unlock()
return
}
close(ch)
}
func (s *store) Compact(rev int64) (<-chan struct{}, error) {
s.mu.Lock()
defer s.mu.Unlock()
if rev <= s.compactMainRev {
ch := make(chan struct{})
f := func(ctx context.Context) { s.compactBarrier(ctx, ch) }
s.fifoSched.Schedule(f)
return ch, ErrCompacted
}
if rev > s.currentRev.main {
return nil, ErrFutureRev
}
start := time.Now()
s.compactMainRev = rev
rbytes := newRevBytes()
revToBytes(revision{main: rev}, rbytes)
tx := s.b.BatchTx()
tx.Lock()
tx.UnsafePut(metaBucketName, scheduledCompactKeyName, rbytes)
tx.Unlock()
// ensure that desired compaction is persisted
s.b.ForceCommit()
keep := s.kvindex.Compact(rev)
ch := make(chan struct{})
var j = func(ctx context.Context) {
if ctx.Err() != nil {
s.compactBarrier(ctx, ch)
return
}
if !s.scheduleCompaction(rev, keep) {
s.compactBarrier(nil, ch)
return
}
close(ch)
}
s.fifoSched.Schedule(j)
indexCompactionPauseDurations.Observe(float64(time.Since(start) / time.Millisecond))
return ch, nil
}
// DefaultIgnores is a map of keys to ignore in hash checking.
var DefaultIgnores map[backend.IgnoreKey]struct{}
func init() {
DefaultIgnores = map[backend.IgnoreKey]struct{}{
// consistent index might be changed due to v2 internal sync, which
// is not controllable by the user.
{Bucket: string(metaBucketName), Key: string(consistentIndexKeyName)}: {},
}
}
func (s *store) Hash() (uint32, int64, error) {
s.mu.Lock()
defer s.mu.Unlock()
s.b.ForceCommit()
h, err := s.b.Hash(DefaultIgnores)
rev := s.currentRev.main
return h, rev, err
}
func (s *store) Commit() {
s.mu.Lock()
defer s.mu.Unlock()
s.tx = s.b.BatchTx()
s.tx.Lock()
s.saveIndex()
s.tx.Unlock()
s.b.ForceCommit()
}
func (s *store) Restore(b backend.Backend) error {
s.mu.Lock()
defer s.mu.Unlock()
close(s.stopc)
s.fifoSched.Stop()
s.b = b
s.kvindex = newTreeIndex()
s.currentRev = revision{main: 1}
s.compactMainRev = -1
s.tx = b.BatchTx()
s.txnID = -1
s.fifoSched = schedule.NewFIFOScheduler()
s.stopc = make(chan struct{})
return s.restore()
}
func (s *store) restore() error {
min, max := newRevBytes(), newRevBytes()
revToBytes(revision{main: 1}, min)
revToBytes(revision{main: math.MaxInt64, sub: math.MaxInt64}, max)
keyToLease := make(map[string]lease.LeaseID)
// use an unordered map to hold the temp index data to speed up
// the initial key index recovery.
// we will convert this unordered map into the tree index later.
unordered := make(map[string]*keyIndex, 100000)
// restore index
tx := s.b.BatchTx()
tx.Lock()
_, finishedCompactBytes := tx.UnsafeRange(metaBucketName, finishedCompactKeyName, nil, 0)
if len(finishedCompactBytes) != 0 {
s.compactMainRev = bytesToRev(finishedCompactBytes[0]).main
plog.Printf("restore compact to %d", s.compactMainRev)
}
// TODO: limit N to reduce max memory usage
keys, vals := tx.UnsafeRange(keyBucketName, min, max, 0)
for i, key := range keys {
var kv mvccpb.KeyValue
if err := kv.Unmarshal(vals[i]); err != nil {
plog.Fatalf("cannot unmarshal event: %v", err)
}
rev := bytesToRev(key[:revBytesLen])
// restore index
switch {
case isTombstone(key):
if ki, ok := unordered[string(kv.Key)]; ok {
ki.tombstone(rev.main, rev.sub)
}
delete(keyToLease, string(kv.Key))
default:
ki, ok := unordered[string(kv.Key)]
if ok {
ki.put(rev.main, rev.sub)
} else {
ki = &keyIndex{key: kv.Key}
ki.restore(revision{kv.CreateRevision, 0}, rev, kv.Version)
unordered[string(kv.Key)] = ki
}
if lid := lease.LeaseID(kv.Lease); lid != lease.NoLease {
keyToLease[string(kv.Key)] = lid
} else {
delete(keyToLease, string(kv.Key))
}
}
// update revision
s.currentRev = rev
}
// restore the tree index from the unordered index.
for _, v := range unordered {
s.kvindex.Insert(v)
}
// keys in the range [compacted revision -N, compaction] might all be deleted due to compaction.
// the correct revision should be set to compaction revision in the case, not the largest revision
// we have seen.
if s.currentRev.main < s.compactMainRev {
s.currentRev.main = s.compactMainRev
}
for key, lid := range keyToLease {
if s.le == nil {
panic("no lessor to attach lease")
}
err := s.le.Attach(lid, []lease.LeaseItem{{Key: key}})
if err != nil {
plog.Errorf("unexpected Attach error: %v", err)
}
}
_, scheduledCompactBytes := tx.UnsafeRange(metaBucketName, scheduledCompactKeyName, nil, 0)
scheduledCompact := int64(0)
if len(scheduledCompactBytes) != 0 {
scheduledCompact = bytesToRev(scheduledCompactBytes[0]).main
if scheduledCompact <= s.compactMainRev {
scheduledCompact = 0
}
}
tx.Unlock()
if scheduledCompact != 0 {
s.Compact(scheduledCompact)
plog.Printf("resume scheduled compaction at %d", scheduledCompact)
}
return nil
}
func (s *store) Close() error {
close(s.stopc)
s.fifoSched.Stop()
return nil
}
func (a *store) Equal(b *store) bool {
if a.currentRev != b.currentRev {
return false
}
if a.compactMainRev != b.compactMainRev {
return false
}
return a.kvindex.Equal(b.kvindex)
}
// range is a keyword in Go, add Keys suffix.
func (s *store) rangeKeys(key, end []byte, limit, rangeRev int64, countOnly bool) (kvs []mvccpb.KeyValue, count int, curRev int64, err error) {
curRev = int64(s.currentRev.main)
if s.currentRev.sub > 0 {
curRev += 1
}
if rangeRev > curRev {
return nil, -1, s.currentRev.main, ErrFutureRev
}
var rev int64
if rangeRev <= 0 {
rev = curRev
} else {
rev = rangeRev
}
if rev < s.compactMainRev {
return nil, -1, 0, ErrCompacted
}
_, revpairs := s.kvindex.Range(key, end, int64(rev))
if len(revpairs) == 0 {
return nil, 0, curRev, nil
}
if countOnly {
return nil, len(revpairs), curRev, nil
}
for _, revpair := range revpairs {
start, end := revBytesRange(revpair)
_, vs := s.tx.UnsafeRange(keyBucketName, start, end, 0)
if len(vs) != 1 {
plog.Fatalf("range cannot find rev (%d,%d)", revpair.main, revpair.sub)
}
var kv mvccpb.KeyValue
if err := kv.Unmarshal(vs[0]); err != nil {
plog.Fatalf("cannot unmarshal event: %v", err)
}
kvs = append(kvs, kv)
if limit > 0 && len(kvs) >= int(limit) {
break
}
}
return kvs, len(revpairs), curRev, nil
}
func (s *store) put(key, value []byte, leaseID lease.LeaseID) {
s.txnModify = true
rev := s.currentRev.main + 1
c := rev
oldLease := lease.NoLease
// if the key exists before, use its previous created and
// get its previous leaseID
_, created, ver, err := s.kvindex.Get(key, rev)
if err == nil {
c = created.main
oldLease = s.le.GetLease(lease.LeaseItem{Key: string(key)})
}
ibytes := newRevBytes()
revToBytes(revision{main: rev, sub: s.currentRev.sub}, ibytes)
ver = ver + 1
kv := mvccpb.KeyValue{
Key: key,
Value: value,
CreateRevision: c,
ModRevision: rev,
Version: ver,
Lease: int64(leaseID),
}
d, err := kv.Marshal()
if err != nil {
plog.Fatalf("cannot marshal event: %v", err)
}
s.tx.UnsafeSeqPut(keyBucketName, ibytes, d)
s.kvindex.Put(key, revision{main: rev, sub: s.currentRev.sub})
s.changes = append(s.changes, kv)
s.currentRev.sub += 1
if oldLease != lease.NoLease {
if s.le == nil {
panic("no lessor to detach lease")
}
err = s.le.Detach(oldLease, []lease.LeaseItem{{Key: string(key)}})
if err != nil {
plog.Errorf("unexpected error from lease detach: %v", err)
}
}
if leaseID != lease.NoLease {
if s.le == nil {
panic("no lessor to attach lease")
}
err = s.le.Attach(leaseID, []lease.LeaseItem{{Key: string(key)}})
if err != nil {
panic("unexpected error from lease Attach")
}
}
}
func (s *store) deleteRange(key, end []byte) int64 {
s.txnModify = true
rrev := s.currentRev.main
if s.currentRev.sub > 0 {
rrev += 1
}
keys, revs := s.kvindex.Range(key, end, rrev)
if len(keys) == 0 {
return 0
}
for i, key := range keys {
s.delete(key, revs[i])
}
return int64(len(keys))
}
func (s *store) delete(key []byte, rev revision) {
mainrev := s.currentRev.main + 1
ibytes := newRevBytes()
revToBytes(revision{main: mainrev, sub: s.currentRev.sub}, ibytes)
ibytes = appendMarkTombstone(ibytes)
kv := mvccpb.KeyValue{
Key: key,
}
d, err := kv.Marshal()
if err != nil {
plog.Fatalf("cannot marshal event: %v", err)
}
s.tx.UnsafeSeqPut(keyBucketName, ibytes, d)
err = s.kvindex.Tombstone(key, revision{main: mainrev, sub: s.currentRev.sub})
if err != nil {
plog.Fatalf("cannot tombstone an existing key (%s): %v", string(key), err)
}
s.changes = append(s.changes, kv)
s.currentRev.sub += 1
item := lease.LeaseItem{Key: string(key)}
leaseID := s.le.GetLease(item)
if leaseID != lease.NoLease {
err = s.le.Detach(leaseID, []lease.LeaseItem{item})
if err != nil {
plog.Errorf("cannot detach %v", err)
}
}
}
func (s *store) getChanges() []mvccpb.KeyValue {
changes := s.changes
s.changes = make([]mvccpb.KeyValue, 0, 4)
return changes
}
func (s *store) saveIndex() {
if s.ig == nil {
return
}
tx := s.tx
bs := s.bytesBuf8
binary.BigEndian.PutUint64(bs, s.ig.ConsistentIndex())
// put the index into the underlying backend
// tx has been locked in TxnBegin, so there is no need to lock it again
tx.UnsafePut(metaBucketName, consistentIndexKeyName, bs)
}
func (s *store) ConsistentIndex() uint64 {
// TODO: cache index in a uint64 field?
tx := s.b.BatchTx()
tx.Lock()
defer tx.Unlock()
_, vs := tx.UnsafeRange(metaBucketName, consistentIndexKeyName, nil, 0)
if len(vs) == 0 {
return 0
}
return binary.BigEndian.Uint64(vs[0])
}
// appendMarkTombstone appends tombstone mark to normal revision bytes.
func appendMarkTombstone(b []byte) []byte {
if len(b) != revBytesLen {
plog.Panicf("cannot append mark to non normal revision bytes")
}
return append(b, markTombstone)
}
// isTombstone checks whether the revision bytes is a tombstone.
func isTombstone(b []byte) bool {
return len(b) == markedRevBytesLen && b[markBytePosition] == markTombstone
}
// revBytesRange returns the range of revision bytes at
// the given revision.
func revBytesRange(rev revision) (start, end []byte) {
start = newRevBytes()
revToBytes(rev, start)
end = newRevBytes()
endRev := revision{main: rev.main, sub: rev.sub + 1}
revToBytes(endRev, end)
return start, end
}