-
Notifications
You must be signed in to change notification settings - Fork 5.9k
/
Copy pathrestore.go
1513 lines (1370 loc) · 53.9 KB
/
restore.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
// Copyright 2020 PingCAP, Inc. Licensed under Apache-2.0.
package task
import (
"cmp"
"context"
"fmt"
"slices"
"strings"
"time"
"github.com/google/uuid"
"github.com/opentracing/opentracing-go"
"github.com/pingcap/errors"
"github.com/pingcap/failpoint"
backuppb "github.com/pingcap/kvproto/pkg/brpb"
"github.com/pingcap/log"
"github.com/pingcap/tidb/br/pkg/checkpoint"
pconfig "github.com/pingcap/tidb/br/pkg/config"
"github.com/pingcap/tidb/br/pkg/conn"
connutil "github.com/pingcap/tidb/br/pkg/conn/util"
berrors "github.com/pingcap/tidb/br/pkg/errors"
"github.com/pingcap/tidb/br/pkg/glue"
"github.com/pingcap/tidb/br/pkg/httputil"
"github.com/pingcap/tidb/br/pkg/logutil"
"github.com/pingcap/tidb/br/pkg/metautil"
"github.com/pingcap/tidb/br/pkg/restore"
snapclient "github.com/pingcap/tidb/br/pkg/restore/snap_client"
"github.com/pingcap/tidb/br/pkg/restore/tiflashrec"
"github.com/pingcap/tidb/br/pkg/summary"
"github.com/pingcap/tidb/br/pkg/utils"
"github.com/pingcap/tidb/br/pkg/version"
"github.com/pingcap/tidb/pkg/config"
"github.com/pingcap/tidb/pkg/domain"
"github.com/pingcap/tidb/pkg/kv"
"github.com/pingcap/tidb/pkg/parser/model"
"github.com/pingcap/tidb/pkg/tablecodec"
"github.com/pingcap/tidb/pkg/util"
"github.com/pingcap/tidb/pkg/util/collate"
"github.com/pingcap/tidb/pkg/util/mathutil"
"github.com/spf13/cobra"
"github.com/spf13/pflag"
"github.com/tikv/client-go/v2/tikv"
pd "github.com/tikv/pd/client"
clientv3 "go.etcd.io/etcd/client/v3"
"go.uber.org/multierr"
"go.uber.org/zap"
)
const (
flagOnline = "online"
flagNoSchema = "no-schema"
flagLoadStats = "load-stats"
flagGranularity = "granularity"
flagConcurrencyPerStore = "tikv-max-restore-concurrency"
// FlagMergeRegionSizeBytes is the flag name of merge small regions by size
FlagMergeRegionSizeBytes = "merge-region-size-bytes"
// FlagMergeRegionKeyCount is the flag name of merge small regions by key count
FlagMergeRegionKeyCount = "merge-region-key-count"
// FlagPDConcurrency controls concurrency pd-relative operations like split & scatter.
FlagPDConcurrency = "pd-concurrency"
// FlagStatsConcurrency controls concurrency to restore statistic.
FlagStatsConcurrency = "stats-concurrency"
// FlagBatchFlushInterval controls after how long the restore batch would be auto sended.
FlagBatchFlushInterval = "batch-flush-interval"
// FlagDdlBatchSize controls batch ddl size to create a batch of tables
FlagDdlBatchSize = "ddl-batch-size"
// FlagWithPlacementPolicy corresponds to tidb config with-tidb-placement-mode
// current only support STRICT or IGNORE, the default is STRICT according to tidb.
FlagWithPlacementPolicy = "with-tidb-placement-mode"
// FlagKeyspaceName corresponds to tidb config keyspace-name
FlagKeyspaceName = "keyspace-name"
// FlagWaitTiFlashReady represents whether wait tiflash replica ready after table restored and checksumed.
FlagWaitTiFlashReady = "wait-tiflash-ready"
// FlagStreamStartTS and FlagStreamRestoreTS is used for log restore timestamp range.
FlagStreamStartTS = "start-ts"
FlagStreamRestoreTS = "restored-ts"
// FlagStreamFullBackupStorage is used for log restore, represents the full backup storage.
FlagStreamFullBackupStorage = "full-backup-storage"
// FlagPiTRBatchCount and FlagPiTRBatchSize are used for restore log with batch method.
FlagPiTRBatchCount = "pitr-batch-count"
FlagPiTRBatchSize = "pitr-batch-size"
FlagPiTRConcurrency = "pitr-concurrency"
FlagResetSysUsers = "reset-sys-users"
defaultPiTRBatchCount = 8
defaultPiTRBatchSize = 16 * 1024 * 1024
defaultRestoreConcurrency = 128
defaultPiTRConcurrency = 16
defaultPDConcurrency = 1
defaultStatsConcurrency = 12
defaultBatchFlushInterval = 16 * time.Second
defaultFlagDdlBatchSize = 128
resetSpeedLimitRetryTimes = 3
maxRestoreBatchSizeLimit = 10240
)
const (
FullRestoreCmd = "Full Restore"
DBRestoreCmd = "DataBase Restore"
TableRestoreCmd = "Table Restore"
PointRestoreCmd = "Point Restore"
RawRestoreCmd = "Raw Restore"
TxnRestoreCmd = "Txn Restore"
)
// RestoreCommonConfig is the common configuration for all BR restore tasks.
type RestoreCommonConfig struct {
Online bool `json:"online" toml:"online"`
Granularity string `json:"granularity" toml:"granularity"`
ConcurrencyPerStore pconfig.ConfigTerm[uint] `json:"tikv-max-restore-concurrency" toml:"tikv-max-restore-concurrency"`
// MergeSmallRegionSizeBytes is the threshold of merging small regions (Default 96MB, region split size).
// MergeSmallRegionKeyCount is the threshold of merging smalle regions (Default 960_000, region split key count).
// See https://github.com/tikv/tikv/blob/v4.0.8/components/raftstore/src/coprocessor/config.rs#L35-L38
MergeSmallRegionSizeBytes pconfig.ConfigTerm[uint64] `json:"merge-region-size-bytes" toml:"merge-region-size-bytes"`
MergeSmallRegionKeyCount pconfig.ConfigTerm[uint64] `json:"merge-region-key-count" toml:"merge-region-key-count"`
// determines whether enable restore sys table on default, see fullClusterRestore in restore/client.go
WithSysTable bool `json:"with-sys-table" toml:"with-sys-table"`
ResetSysUsers []string `json:"reset-sys-users" toml:"reset-sys-users"`
}
// adjust adjusts the abnormal config value in the current config.
// useful when not starting BR from CLI (e.g. from BRIE in SQL).
func (cfg *RestoreCommonConfig) adjust() {
if !cfg.MergeSmallRegionKeyCount.Modified {
cfg.MergeSmallRegionKeyCount.Value = conn.DefaultMergeRegionKeyCount
}
if !cfg.MergeSmallRegionSizeBytes.Modified {
cfg.MergeSmallRegionSizeBytes.Value = conn.DefaultMergeRegionSizeBytes
}
if len(cfg.Granularity) == 0 {
cfg.Granularity = string(restore.CoarseGrained)
}
if !cfg.ConcurrencyPerStore.Modified {
cfg.ConcurrencyPerStore.Value = conn.DefaultImportNumGoroutines
}
}
// DefineRestoreCommonFlags defines common flags for the restore command.
func DefineRestoreCommonFlags(flags *pflag.FlagSet) {
// TODO remove experimental tag if it's stable
flags.Bool(flagOnline, false, "(experimental) Whether online when restore")
flags.String(flagGranularity, string(restore.CoarseGrained), "(deprecated) Whether split & scatter regions using fine-grained way during restore")
flags.Uint(flagConcurrencyPerStore, 128, "The size of thread pool on each store that executes tasks")
flags.Uint32(flagConcurrency, 128, "(deprecated) The size of thread pool on BR that executes tasks, "+
"where each task restores one SST file to TiKV")
flags.Uint64(FlagMergeRegionSizeBytes, conn.DefaultMergeRegionSizeBytes,
"the threshold of merging small regions (Default 96MB, region split size)")
flags.Uint64(FlagMergeRegionKeyCount, conn.DefaultMergeRegionKeyCount,
"the threshold of merging small regions (Default 960_000, region split key count)")
flags.Uint(FlagPDConcurrency, defaultPDConcurrency,
"concurrency pd-relative operations like split & scatter.")
flags.Uint(FlagStatsConcurrency, defaultStatsConcurrency,
"concurrency to restore statistic")
flags.Duration(FlagBatchFlushInterval, defaultBatchFlushInterval,
"after how long a restore batch would be auto sent.")
flags.Uint(FlagDdlBatchSize, defaultFlagDdlBatchSize,
"batch size for ddl to create a batch of tables once.")
flags.Bool(flagWithSysTable, true, "whether restore system privilege tables on default setting")
flags.StringArrayP(FlagResetSysUsers, "", []string{"cloud_admin", "root"}, "whether reset these users after restoration")
flags.Bool(flagUseFSR, false, "whether enable FSR for AWS snapshots")
_ = flags.MarkHidden(FlagResetSysUsers)
_ = flags.MarkHidden(FlagMergeRegionSizeBytes)
_ = flags.MarkHidden(FlagMergeRegionKeyCount)
_ = flags.MarkHidden(FlagPDConcurrency)
_ = flags.MarkHidden(FlagStatsConcurrency)
_ = flags.MarkHidden(FlagBatchFlushInterval)
_ = flags.MarkHidden(FlagDdlBatchSize)
}
// ParseFromFlags parses the config from the flag set.
func (cfg *RestoreCommonConfig) ParseFromFlags(flags *pflag.FlagSet) error {
var err error
cfg.Online, err = flags.GetBool(flagOnline)
if err != nil {
return errors.Trace(err)
}
cfg.Granularity, err = flags.GetString(flagGranularity)
if err != nil {
return errors.Trace(err)
}
cfg.ConcurrencyPerStore.Value, err = flags.GetUint(flagConcurrencyPerStore)
if err != nil {
return errors.Trace(err)
}
cfg.ConcurrencyPerStore.Modified = flags.Changed(flagConcurrencyPerStore)
cfg.MergeSmallRegionKeyCount.Value, err = flags.GetUint64(FlagMergeRegionKeyCount)
if err != nil {
return errors.Trace(err)
}
cfg.MergeSmallRegionKeyCount.Modified = flags.Changed(FlagMergeRegionKeyCount)
cfg.MergeSmallRegionSizeBytes.Value, err = flags.GetUint64(FlagMergeRegionSizeBytes)
if err != nil {
return errors.Trace(err)
}
cfg.MergeSmallRegionSizeBytes.Modified = flags.Changed(FlagMergeRegionSizeBytes)
if flags.Lookup(flagWithSysTable) != nil {
cfg.WithSysTable, err = flags.GetBool(flagWithSysTable)
if err != nil {
return errors.Trace(err)
}
}
cfg.ResetSysUsers, err = flags.GetStringArray(FlagResetSysUsers)
if err != nil {
return errors.Trace(err)
}
return errors.Trace(err)
}
// RestoreConfig is the configuration specific for restore tasks.
type RestoreConfig struct {
Config
RestoreCommonConfig
NoSchema bool `json:"no-schema" toml:"no-schema"`
LoadStats bool `json:"load-stats" toml:"load-stats"`
PDConcurrency uint `json:"pd-concurrency" toml:"pd-concurrency"`
StatsConcurrency uint `json:"stats-concurrency" toml:"stats-concurrency"`
BatchFlushInterval time.Duration `json:"batch-flush-interval" toml:"batch-flush-interval"`
// DdlBatchSize use to define the size of batch ddl to create tables
DdlBatchSize uint `json:"ddl-batch-size" toml:"ddl-batch-size"`
WithPlacementPolicy string `json:"with-tidb-placement-mode" toml:"with-tidb-placement-mode"`
// FullBackupStorage is used to run `restore full` before `restore log`.
// if it is empty, directly take restoring log justly.
FullBackupStorage string `json:"full-backup-storage" toml:"full-backup-storage"`
// [startTs, RestoreTS] is used to `restore log` from StartTS to RestoreTS.
StartTS uint64 `json:"start-ts" toml:"start-ts"`
RestoreTS uint64 `json:"restore-ts" toml:"restore-ts"`
tiflashRecorder *tiflashrec.TiFlashRecorder `json:"-" toml:"-"`
PitrBatchCount uint32 `json:"pitr-batch-count" toml:"pitr-batch-count"`
PitrBatchSize uint32 `json:"pitr-batch-size" toml:"pitr-batch-size"`
PitrConcurrency uint32 `json:"-" toml:"-"`
UseCheckpoint bool `json:"use-checkpoint" toml:"use-checkpoint"`
checkpointSnapshotRestoreTaskName string `json:"-" toml:"-"`
checkpointLogRestoreTaskName string `json:"-" toml:"-"`
checkpointTaskInfoClusterID uint64 `json:"-" toml:"-"`
WaitTiflashReady bool `json:"wait-tiflash-ready" toml:"wait-tiflash-ready"`
// for ebs-based restore
FullBackupType FullBackupType `json:"full-backup-type" toml:"full-backup-type"`
Prepare bool `json:"prepare" toml:"prepare"`
OutputFile string `json:"output-file" toml:"output-file"`
SkipAWS bool `json:"skip-aws" toml:"skip-aws"`
CloudAPIConcurrency uint `json:"cloud-api-concurrency" toml:"cloud-api-concurrency"`
VolumeType pconfig.EBSVolumeType `json:"volume-type" toml:"volume-type"`
VolumeIOPS int64 `json:"volume-iops" toml:"volume-iops"`
VolumeThroughput int64 `json:"volume-throughput" toml:"volume-throughput"`
VolumeEncrypted bool `json:"volume-encrypted" toml:"volume-encrypted"`
ProgressFile string `json:"progress-file" toml:"progress-file"`
TargetAZ string `json:"target-az" toml:"target-az"`
UseFSR bool `json:"use-fsr" toml:"use-fsr"`
}
// DefineRestoreFlags defines common flags for the restore tidb command.
func DefineRestoreFlags(flags *pflag.FlagSet) {
flags.Bool(flagNoSchema, false, "skip creating schemas and tables, reuse existing empty ones")
flags.Bool(flagLoadStats, true, "Run load stats at end of snapshot restore task")
// Do not expose this flag
_ = flags.MarkHidden(flagNoSchema)
flags.String(FlagWithPlacementPolicy, "STRICT", "correspond to tidb global/session variable with-tidb-placement-mode")
flags.String(FlagKeyspaceName, "", "correspond to tidb config keyspace-name")
flags.Bool(flagUseCheckpoint, true, "use checkpoint mode")
_ = flags.MarkHidden(flagUseCheckpoint)
flags.Bool(FlagWaitTiFlashReady, false, "whether wait tiflash replica ready if tiflash exists")
DefineRestoreCommonFlags(flags)
}
// DefineStreamRestoreFlags defines for the restore log command.
func DefineStreamRestoreFlags(command *cobra.Command) {
command.Flags().String(FlagStreamStartTS, "", "the start timestamp which log restore from.\n"+
"support TSO or datetime, e.g. '400036290571534337' or '2018-05-11 01:42:23+0800'")
command.Flags().String(FlagStreamRestoreTS, "", "the point of restore, used for log restore.\n"+
"support TSO or datetime, e.g. '400036290571534337' or '2018-05-11 01:42:23+0800'")
command.Flags().String(FlagStreamFullBackupStorage, "", "specify the backup full storage. "+
"fill it if want restore full backup before restore log.")
command.Flags().Uint32(FlagPiTRBatchCount, defaultPiTRBatchCount, "specify the batch count to restore log.")
command.Flags().Uint32(FlagPiTRBatchSize, defaultPiTRBatchSize, "specify the batch size to retore log.")
command.Flags().Uint32(FlagPiTRConcurrency, defaultPiTRConcurrency, "specify the concurrency to restore log.")
}
// ParseStreamRestoreFlags parses the `restore stream` flags from the flag set.
func (cfg *RestoreConfig) ParseStreamRestoreFlags(flags *pflag.FlagSet) error {
tsString, err := flags.GetString(FlagStreamStartTS)
if err != nil {
return errors.Trace(err)
}
if cfg.StartTS, err = ParseTSString(tsString, true); err != nil {
return errors.Trace(err)
}
tsString, err = flags.GetString(FlagStreamRestoreTS)
if err != nil {
return errors.Trace(err)
}
if cfg.RestoreTS, err = ParseTSString(tsString, true); err != nil {
return errors.Trace(err)
}
if cfg.FullBackupStorage, err = flags.GetString(FlagStreamFullBackupStorage); err != nil {
return errors.Trace(err)
}
if cfg.StartTS > 0 && len(cfg.FullBackupStorage) > 0 {
return errors.Annotatef(berrors.ErrInvalidArgument, "%v and %v are mutually exclusive",
FlagStreamStartTS, FlagStreamFullBackupStorage)
}
if cfg.PitrBatchCount, err = flags.GetUint32(FlagPiTRBatchCount); err != nil {
return errors.Trace(err)
}
if cfg.PitrBatchSize, err = flags.GetUint32(FlagPiTRBatchSize); err != nil {
return errors.Trace(err)
}
if cfg.PitrConcurrency, err = flags.GetUint32(FlagPiTRConcurrency); err != nil {
return errors.Trace(err)
}
return nil
}
// ParseFromFlags parses the restore-related flags from the flag set.
func (cfg *RestoreConfig) ParseFromFlags(flags *pflag.FlagSet) error {
var err error
cfg.NoSchema, err = flags.GetBool(flagNoSchema)
if err != nil {
return errors.Trace(err)
}
cfg.LoadStats, err = flags.GetBool(flagLoadStats)
if err != nil {
return errors.Trace(err)
}
err = cfg.Config.ParseFromFlags(flags)
if err != nil {
return errors.Trace(err)
}
err = cfg.RestoreCommonConfig.ParseFromFlags(flags)
if err != nil {
return errors.Trace(err)
}
cfg.Concurrency, err = flags.GetUint32(flagConcurrency)
if err != nil {
return errors.Trace(err)
}
if cfg.Config.Concurrency == 0 {
cfg.Config.Concurrency = defaultRestoreConcurrency
}
cfg.PDConcurrency, err = flags.GetUint(FlagPDConcurrency)
if err != nil {
return errors.Annotatef(err, "failed to get flag %s", FlagPDConcurrency)
}
cfg.StatsConcurrency, err = flags.GetUint(FlagStatsConcurrency)
if err != nil {
return errors.Annotatef(err, "failed to get flag %s", FlagStatsConcurrency)
}
cfg.BatchFlushInterval, err = flags.GetDuration(FlagBatchFlushInterval)
if err != nil {
return errors.Annotatef(err, "failed to get flag %s", FlagBatchFlushInterval)
}
cfg.DdlBatchSize, err = flags.GetUint(FlagDdlBatchSize)
if err != nil {
return errors.Annotatef(err, "failed to get flag %s", FlagDdlBatchSize)
}
cfg.WithPlacementPolicy, err = flags.GetString(FlagWithPlacementPolicy)
if err != nil {
return errors.Annotatef(err, "failed to get flag %s", FlagWithPlacementPolicy)
}
cfg.KeyspaceName, err = flags.GetString(FlagKeyspaceName)
if err != nil {
return errors.Annotatef(err, "failed to get flag %s", FlagKeyspaceName)
}
cfg.UseCheckpoint, err = flags.GetBool(flagUseCheckpoint)
if err != nil {
return errors.Annotatef(err, "failed to get flag %s", flagUseCheckpoint)
}
cfg.WaitTiflashReady, err = flags.GetBool(FlagWaitTiFlashReady)
if err != nil {
return errors.Annotatef(err, "failed to get flag %s", FlagWaitTiFlashReady)
}
if flags.Lookup(flagFullBackupType) != nil {
// for restore full only
fullBackupType, err := flags.GetString(flagFullBackupType)
if err != nil {
return errors.Trace(err)
}
if !FullBackupType(fullBackupType).Valid() {
return errors.New("invalid full backup type")
}
cfg.FullBackupType = FullBackupType(fullBackupType)
cfg.Prepare, err = flags.GetBool(flagPrepare)
if err != nil {
return errors.Trace(err)
}
cfg.SkipAWS, err = flags.GetBool(flagSkipAWS)
if err != nil {
return errors.Trace(err)
}
cfg.CloudAPIConcurrency, err = flags.GetUint(flagCloudAPIConcurrency)
if err != nil {
return errors.Trace(err)
}
cfg.OutputFile, err = flags.GetString(flagOutputMetaFile)
if err != nil {
return errors.Trace(err)
}
volumeType, err := flags.GetString(flagVolumeType)
if err != nil {
return errors.Trace(err)
}
cfg.VolumeType = pconfig.EBSVolumeType(volumeType)
if !cfg.VolumeType.Valid() {
return errors.New("invalid volume type: " + volumeType)
}
if cfg.VolumeIOPS, err = flags.GetInt64(flagVolumeIOPS); err != nil {
return errors.Trace(err)
}
if cfg.VolumeThroughput, err = flags.GetInt64(flagVolumeThroughput); err != nil {
return errors.Trace(err)
}
if cfg.VolumeEncrypted, err = flags.GetBool(flagVolumeEncrypted); err != nil {
return errors.Trace(err)
}
cfg.ProgressFile, err = flags.GetString(flagProgressFile)
if err != nil {
return errors.Trace(err)
}
cfg.TargetAZ, err = flags.GetString(flagTargetAZ)
if err != nil {
return errors.Trace(err)
}
cfg.UseFSR, err = flags.GetBool(flagUseFSR)
if err != nil {
return errors.Trace(err)
}
// iops: gp3 [3,000-16,000]; io1/io2 [100-32,000]
// throughput: gp3 [125, 1000]; io1/io2 cannot set throughput
// io1 and io2 volumes support up to 64,000 IOPS only on Instances built on the Nitro System.
// Other instance families support performance up to 32,000 IOPS.
// https://docs.aws.amazon.com/AWSEC2/latest/APIReference/API_CreateVolume.html
// todo: check lower/upper bound
}
return nil
}
// Adjust is use for BR(binary) and BR in TiDB.
// When new config was added and not included in parser.
// we should set proper value in this function.
// so that both binary and TiDB will use same default value.
func (cfg *RestoreConfig) Adjust() {
cfg.Config.adjust()
cfg.RestoreCommonConfig.adjust()
if cfg.Config.Concurrency == 0 {
cfg.Config.Concurrency = defaultRestoreConcurrency
}
if cfg.Config.SwitchModeInterval == 0 {
cfg.Config.SwitchModeInterval = defaultSwitchInterval
}
if cfg.PDConcurrency == 0 {
cfg.PDConcurrency = defaultPDConcurrency
}
if cfg.StatsConcurrency == 0 {
cfg.StatsConcurrency = defaultStatsConcurrency
}
if cfg.BatchFlushInterval == 0 {
cfg.BatchFlushInterval = defaultBatchFlushInterval
}
if cfg.DdlBatchSize == 0 {
cfg.DdlBatchSize = defaultFlagDdlBatchSize
}
if cfg.CloudAPIConcurrency == 0 {
cfg.CloudAPIConcurrency = defaultCloudAPIConcurrency
}
}
func (cfg *RestoreConfig) adjustRestoreConfigForStreamRestore() {
if cfg.PitrConcurrency == 0 {
cfg.PitrConcurrency = defaultPiTRConcurrency
}
if cfg.PitrBatchCount == 0 {
cfg.PitrBatchCount = defaultPiTRBatchCount
}
if cfg.PitrBatchSize == 0 {
cfg.PitrBatchSize = defaultPiTRBatchSize
}
// another goroutine is used to iterate the backup file
cfg.PitrConcurrency += 1
log.Info("set restore kv files concurrency", zap.Int("concurrency", int(cfg.PitrConcurrency)))
cfg.Config.Concurrency = cfg.PitrConcurrency
}
// generateLogRestoreTaskName generates the log restore taskName for checkpoint
func (cfg *RestoreConfig) generateLogRestoreTaskName(clusterID, startTS, restoreTs uint64) string {
cfg.checkpointTaskInfoClusterID = clusterID
cfg.checkpointLogRestoreTaskName = fmt.Sprintf("%d/%d.%d", clusterID, startTS, restoreTs)
return cfg.checkpointLogRestoreTaskName
}
// generateSnapshotRestoreTaskName generates the snapshot restore taskName for checkpoint
func (cfg *RestoreConfig) generateSnapshotRestoreTaskName(clusterID uint64) string {
cfg.checkpointSnapshotRestoreTaskName = fmt.Sprint(clusterID)
return cfg.checkpointSnapshotRestoreTaskName
}
func configureRestoreClient(ctx context.Context, client *snapclient.SnapClient, cfg *RestoreConfig) error {
client.SetRateLimit(cfg.RateLimit)
client.SetCrypter(&cfg.CipherInfo)
if cfg.NoSchema {
client.EnableSkipCreateSQL()
}
client.SetBatchDdlSize(cfg.DdlBatchSize)
client.SetPlacementPolicyMode(cfg.WithPlacementPolicy)
client.SetWithSysTable(cfg.WithSysTable)
client.SetRewriteMode(ctx)
return nil
}
func CheckNewCollationEnable(
backupNewCollationEnable string,
g glue.Glue,
storage kv.Storage,
CheckRequirements bool,
) (bool, error) {
se, err := g.CreateSession(storage)
if err != nil {
return false, errors.Trace(err)
}
newCollationEnable, err := se.GetGlobalVariable(utils.GetTidbNewCollationEnabled())
if err != nil {
return false, errors.Trace(err)
}
// collate.newCollationEnabled is set to 1 when the collate package is initialized,
// so we need to modify this value according to the config of the cluster
// before using the collate package.
enabled := newCollationEnable == "True"
// modify collate.newCollationEnabled according to the config of the cluster
collate.SetNewCollationEnabledForTest(enabled)
log.Info(fmt.Sprintf("set %s", utils.TidbNewCollationEnabled), zap.Bool("new_collation_enabled", enabled))
if backupNewCollationEnable == "" {
if CheckRequirements {
return enabled, errors.Annotatef(berrors.ErrUnknown,
"the value '%s' not found in backupmeta. "+
"you can use \"SELECT VARIABLE_VALUE FROM mysql.tidb WHERE VARIABLE_NAME='%s';\" to manually check the config. "+
"if you ensure the value '%s' in backup cluster is as same as restore cluster, use --check-requirements=false to skip this check",
utils.TidbNewCollationEnabled, utils.TidbNewCollationEnabled, utils.TidbNewCollationEnabled)
}
log.Warn(fmt.Sprintf("the config '%s' is not in backupmeta", utils.TidbNewCollationEnabled))
return enabled, nil
}
if !strings.EqualFold(backupNewCollationEnable, newCollationEnable) {
return enabled, errors.Annotatef(berrors.ErrUnknown,
"the config '%s' not match, upstream:%v, downstream: %v",
utils.TidbNewCollationEnabled, backupNewCollationEnable, newCollationEnable)
}
return enabled, nil
}
// CheckRestoreDBAndTable is used to check whether the restore dbs or tables have been backup
func CheckRestoreDBAndTable(schemas []*metautil.Database, cfg *RestoreConfig) error {
if len(cfg.Schemas) == 0 && len(cfg.Tables) == 0 {
return nil
}
schemasMap := make(map[string]struct{})
tablesMap := make(map[string]struct{})
for _, db := range schemas {
dbName := db.Info.Name.L
if dbCIStrName, ok := utils.GetSysDBCIStrName(db.Info.Name); utils.IsSysDB(dbCIStrName.O) && ok {
dbName = dbCIStrName.L
}
schemasMap[utils.EncloseName(dbName)] = struct{}{}
for _, table := range db.Tables {
if table.Info == nil {
// we may back up empty database.
continue
}
tablesMap[utils.EncloseDBAndTable(dbName, table.Info.Name.L)] = struct{}{}
}
}
restoreSchemas := cfg.Schemas
restoreTables := cfg.Tables
for schema := range restoreSchemas {
schemaLName := strings.ToLower(schema)
if _, ok := schemasMap[schemaLName]; !ok {
return errors.Annotatef(berrors.ErrUndefinedRestoreDbOrTable,
"[database: %v] has not been backup, please ensure you has input a correct database name", schema)
}
}
for table := range restoreTables {
tableLName := strings.ToLower(table)
if _, ok := tablesMap[tableLName]; !ok {
return errors.Annotatef(berrors.ErrUndefinedRestoreDbOrTable,
"[table: %v] has not been backup, please ensure you has input a correct table name", table)
}
}
return nil
}
func isFullRestore(cmdName string) bool {
return cmdName == FullRestoreCmd
}
// IsStreamRestore checks the command is `restore point`
func IsStreamRestore(cmdName string) bool {
return cmdName == PointRestoreCmd
}
func registerTaskToPD(ctx context.Context, etcdCLI *clientv3.Client) (closeF func(context.Context) error, err error) {
register := utils.NewTaskRegister(etcdCLI, utils.RegisterRestore, fmt.Sprintf("restore-%s", uuid.New()))
err = register.RegisterTask(ctx)
return register.Close, errors.Trace(err)
}
func removeCheckpointDataForSnapshotRestore(ctx context.Context, storageName string, taskName string, config *Config) error {
_, s, err := GetStorage(ctx, storageName, config)
if err != nil {
return errors.Trace(err)
}
return errors.Trace(checkpoint.RemoveCheckpointDataForRestore(ctx, s, taskName))
}
func removeCheckpointDataForLogRestore(ctx context.Context, storageName string, taskName string, clusterID uint64, config *Config) error {
_, s, err := GetStorage(ctx, storageName, config)
if err != nil {
return errors.Trace(err)
}
return errors.Trace(checkpoint.RemoveCheckpointDataForLogRestore(ctx, s, taskName, clusterID))
}
func DefaultRestoreConfig() RestoreConfig {
fs := pflag.NewFlagSet("dummy", pflag.ContinueOnError)
DefineCommonFlags(fs)
DefineRestoreFlags(fs)
cfg := RestoreConfig{}
err := multierr.Combine(
cfg.ParseFromFlags(fs),
cfg.RestoreCommonConfig.ParseFromFlags(fs),
cfg.Config.ParseFromFlags(fs),
)
if err != nil {
log.Panic("infallible failed.", zap.Error(err))
}
return cfg
}
// RunRestore starts a restore task inside the current goroutine.
func RunRestore(c context.Context, g glue.Glue, cmdName string, cfg *RestoreConfig) error {
etcdCLI, err := dialEtcdWithCfg(c, cfg.Config)
if err != nil {
return err
}
defer func() {
if err := etcdCLI.Close(); err != nil {
log.Error("failed to close the etcd client", zap.Error(err))
}
}()
if err := checkTaskExists(c, cfg, etcdCLI); err != nil {
return errors.Annotate(err, "failed to check task exists")
}
closeF, err := registerTaskToPD(c, etcdCLI)
if err != nil {
return errors.Annotate(err, "failed to register task to pd")
}
defer func() {
_ = closeF(c)
}()
config.UpdateGlobal(func(conf *config.Config) {
conf.KeyspaceName = cfg.KeyspaceName
})
var restoreError error
if IsStreamRestore(cmdName) {
restoreError = RunStreamRestore(c, g, cmdName, cfg)
} else {
restoreError = runRestore(c, g, cmdName, cfg)
}
if restoreError != nil {
return errors.Trace(restoreError)
}
// Clear the checkpoint data
if cfg.UseCheckpoint {
if len(cfg.checkpointLogRestoreTaskName) > 0 {
log.Info("start to remove checkpoint data for log restore")
err = removeCheckpointDataForLogRestore(c, cfg.Config.Storage, cfg.checkpointLogRestoreTaskName, cfg.checkpointTaskInfoClusterID, &cfg.Config)
if err != nil {
log.Warn("failed to remove checkpoint data for log restore", zap.Error(err))
}
}
if len(cfg.checkpointSnapshotRestoreTaskName) > 0 {
log.Info("start to remove checkpoint data for snapshot restore.")
var storage string
if IsStreamRestore(cmdName) {
storage = cfg.FullBackupStorage
} else {
storage = cfg.Config.Storage
}
err = removeCheckpointDataForSnapshotRestore(c, storage, cfg.checkpointSnapshotRestoreTaskName, &cfg.Config)
if err != nil {
log.Warn("failed to remove checkpoint data for snapshot restore", zap.Error(err))
}
}
log.Info("all the checkpoint data is removed.")
}
return nil
}
func runRestore(c context.Context, g glue.Glue, cmdName string, cfg *RestoreConfig) error {
cfg.Adjust()
defer summary.Summary(cmdName)
ctx, cancel := context.WithCancel(c)
defer cancel()
if span := opentracing.SpanFromContext(ctx); span != nil && span.Tracer() != nil {
span1 := span.Tracer().StartSpan("task.RunRestore", opentracing.ChildOf(span.Context()))
defer span1.Finish()
ctx = opentracing.ContextWithSpan(ctx, span1)
}
// Restore needs domain to do DDL.
needDomain := true
keepaliveCfg := GetKeepalive(&cfg.Config)
mgr, err := NewMgr(ctx, g, cfg.PD, cfg.TLS, keepaliveCfg, cfg.CheckRequirements, needDomain, conn.NormalVersionChecker)
if err != nil {
return errors.Trace(err)
}
defer mgr.Close()
codec := mgr.GetStorage().GetCodec()
// need retrieve these configs from tikv if not set in command.
kvConfigs := &pconfig.KVConfig{
ImportGoroutines: cfg.ConcurrencyPerStore,
MergeRegionSize: cfg.MergeSmallRegionSizeBytes,
MergeRegionKeyCount: cfg.MergeSmallRegionKeyCount,
}
// according to https://github.com/pingcap/tidb/issues/34167.
// we should get the real config from tikv to adapt the dynamic region.
httpCli := httputil.NewClient(mgr.GetTLSConfig())
mgr.ProcessTiKVConfigs(ctx, kvConfigs, httpCli)
keepaliveCfg.PermitWithoutStream = true
client := snapclient.NewRestoreClient(mgr.GetPDClient(), mgr.GetPDHTTPClient(), mgr.GetTLSConfig(), keepaliveCfg)
// using tikv config to set the concurrency-per-store for client.
client.SetConcurrencyPerStore(kvConfigs.ImportGoroutines.Value)
err = configureRestoreClient(ctx, client, cfg)
if err != nil {
return errors.Trace(err)
}
// Init DB connection sessions
err = client.Init(g, mgr.GetStorage())
defer client.Close()
if err != nil {
return errors.Trace(err)
}
u, s, backupMeta, err := ReadBackupMeta(ctx, metautil.MetaFile, &cfg.Config)
if err != nil {
return errors.Trace(err)
}
if cfg.CheckRequirements {
err := checkIncompatibleChangefeed(ctx, backupMeta.EndVersion, mgr.GetDomain().GetEtcdClient())
log.Info("Checking incompatible TiCDC changefeeds before restoring.",
logutil.ShortError(err), zap.Uint64("restore-ts", backupMeta.EndVersion))
if err != nil {
return errors.Trace(err)
}
}
backupVersion := version.NormalizeBackupVersion(backupMeta.ClusterVersion)
if cfg.CheckRequirements && backupVersion != nil {
if versionErr := version.CheckClusterVersion(ctx, mgr.GetPDClient(), version.CheckVersionForBackup(backupVersion)); versionErr != nil {
return errors.Trace(versionErr)
}
}
if _, err = CheckNewCollationEnable(backupMeta.GetNewCollationsEnabled(), g, mgr.GetStorage(), cfg.CheckRequirements); err != nil {
return errors.Trace(err)
}
reader := metautil.NewMetaReader(backupMeta, s, &cfg.CipherInfo)
if err = client.InitBackupMeta(c, backupMeta, u, reader, cfg.LoadStats); err != nil {
return errors.Trace(err)
}
if client.IsRawKvMode() {
return errors.Annotate(berrors.ErrRestoreModeMismatch, "cannot do transactional restore from raw kv data")
}
if err = CheckRestoreDBAndTable(client.GetDatabases(), cfg); err != nil {
return err
}
files, tables, dbs := filterRestoreFiles(client, cfg)
if len(dbs) == 0 && len(tables) != 0 {
return errors.Annotate(berrors.ErrRestoreInvalidBackup, "contain tables but no databases")
}
archiveSize := reader.ArchiveSize(ctx, files)
g.Record(summary.RestoreDataSize, archiveSize)
//restore from tidb will fetch a general Size issue https://github.com/pingcap/tidb/issues/27247
g.Record("Size", archiveSize)
restoreTS, err := restore.GetTSWithRetry(ctx, mgr.GetPDClient())
if err != nil {
return errors.Trace(err)
}
if client.IsIncremental() {
// don't support checkpoint for the ddl restore
log.Info("the incremental snapshot restore doesn't support checkpoint mode, so unuse checkpoint.")
cfg.UseCheckpoint = false
}
importModeSwitcher := restore.NewImportModeSwitcher(mgr.GetPDClient(), cfg.Config.SwitchModeInterval, mgr.GetTLSConfig())
restoreSchedulers, schedulersConfig, err := restore.RestorePreWork(ctx, mgr, importModeSwitcher, cfg.Online, true)
if err != nil {
return errors.Trace(err)
}
schedulersRemovable := false
defer func() {
// don't reset pd scheduler if checkpoint mode is used and restored is not finished
if cfg.UseCheckpoint && !schedulersRemovable {
log.Info("skip removing pd schehduler for next retry")
return
}
log.Info("start to remove the pd scheduler")
// run the post-work to avoid being stuck in the import
// mode or emptied schedulers.
restore.RestorePostWork(ctx, importModeSwitcher, restoreSchedulers, cfg.Online)
log.Info("finish removing pd scheduler")
}()
var checkpointTaskName string
var checkpointFirstRun bool = true
if cfg.UseCheckpoint {
checkpointTaskName = cfg.generateSnapshotRestoreTaskName(client.GetClusterID(ctx))
// if the checkpoint metadata exists in the external storage, the restore is not
// for the first time.
existsCheckpointMetadata, err := checkpoint.ExistsRestoreCheckpoint(ctx, s, checkpointTaskName)
if err != nil {
return errors.Trace(err)
}
checkpointFirstRun = !existsCheckpointMetadata
}
if isFullRestore(cmdName) {
if client.NeedCheckFreshCluster(cfg.ExplicitFilter, checkpointFirstRun) {
if err = client.CheckTargetClusterFresh(ctx); err != nil {
return errors.Trace(err)
}
}
// todo: move this check into InitFullClusterRestore, we should move restore config into a separate package
// to avoid import cycle problem which we won't do it in this pr, then refactor this
//
// if it's point restore and reached here, then cmdName=FullRestoreCmd and len(cfg.FullBackupStorage) > 0
if cfg.WithSysTable {
client.InitFullClusterRestore(cfg.ExplicitFilter)
}
}
if client.IsFullClusterRestore() && client.HasBackedUpSysDB() {
if err = snapclient.CheckSysTableCompatibility(mgr.GetDomain(), tables); err != nil {
return errors.Trace(err)
}
}
// reload or register the checkpoint
var checkpointSetWithTableID map[int64]map[string]struct{}
if cfg.UseCheckpoint {
sets, restoreSchedulersConfigFromCheckpoint, err := client.InitCheckpoint(ctx, s, checkpointTaskName, schedulersConfig, checkpointFirstRun)
if err != nil {
return errors.Trace(err)
}
if restoreSchedulersConfigFromCheckpoint != nil {
restoreSchedulers = mgr.MakeUndoFunctionByConfig(*restoreSchedulersConfigFromCheckpoint)
}
checkpointSetWithTableID = sets
defer func() {
// need to flush the whole checkpoint data so that br can quickly jump to
// the log kv restore step when the next retry.
log.Info("wait for flush checkpoint...")
client.WaitForFinishCheckpoint(ctx, len(cfg.FullBackupStorage) > 0 || !schedulersRemovable)
}()
}
sp := utils.BRServiceSafePoint{
BackupTS: restoreTS,
TTL: utils.DefaultBRGCSafePointTTL,
ID: utils.MakeSafePointID(),
}
g.Record("BackupTS", backupMeta.EndVersion)
g.Record("RestoreTS", restoreTS)
cctx, gcSafePointKeeperCancel := context.WithCancel(ctx)
defer func() {
log.Info("start to remove gc-safepoint keeper")
// close the gc safe point keeper at first
gcSafePointKeeperCancel()
// set the ttl to 0 to remove the gc-safe-point
sp.TTL = 0
if err := utils.UpdateServiceSafePoint(ctx, mgr.GetPDClient(), sp); err != nil {
log.Warn("failed to update service safe point, backup may fail if gc triggered",
zap.Error(err),
)
}
log.Info("finish removing gc-safepoint keeper")
}()
// restore checksum will check safe point with its start ts, see details at
// https://github.com/pingcap/tidb/blob/180c02127105bed73712050594da6ead4d70a85f/store/tikv/kv.go#L186-L190
// so, we should keep the safe point unchangeable. to avoid GC life time is shorter than transaction duration.
err = utils.StartServiceSafePointKeeper(cctx, mgr.GetPDClient(), sp)
if err != nil {
return errors.Trace(err)
}
var newTS uint64
if client.IsIncremental() {
newTS = restoreTS
}
ddlJobs := FilterDDLJobs(client.GetDDLJobs(), tables)
ddlJobs = FilterDDLJobByRules(ddlJobs, DDLJobBlockListRule)
err = PreCheckTableTiFlashReplica(ctx, mgr.GetPDClient(), tables, cfg.tiflashRecorder)
if err != nil {
return errors.Trace(err)
}
err = PreCheckTableClusterIndex(tables, ddlJobs, mgr.GetDomain())
if err != nil {
return errors.Trace(err)
}
// pre-set TiDB config for restore
restoreDBConfig := enableTiDBConfig()
defer restoreDBConfig()
if client.GetSupportPolicy() {
// create policy if backupMeta has policies.
policies, err := client.GetPlacementPolicies()
if err != nil {
return errors.Trace(err)
}
if isFullRestore(cmdName) {
// we should restore all policies during full restoration.
err = client.CreatePolicies(ctx, policies)
if err != nil {
return errors.Trace(err)
}
} else {
client.SetPolicyMap(policies)
}
}
// preallocate the table id, because any ddl job or database creation also allocates the global ID
err = client.AllocTableIDs(ctx, tables)
if err != nil {
return errors.Trace(err)
}
// execute DDL first
err = client.ExecDDLs(ctx, ddlJobs)
if err != nil {
return errors.Trace(err)
}
// nothing to restore, maybe only ddl changes in incremental restore
if len(dbs) == 0 && len(tables) == 0 {
log.Info("nothing to restore, all databases and tables are filtered out")
// even nothing to restore, we show a success message since there is no failure.
summary.SetSuccessStatus(true)
return nil
}
if err = client.CreateDatabases(ctx, dbs); err != nil {
return errors.Trace(err)