-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmaster.go
863 lines (789 loc) · 22.3 KB
/
master.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
package sentinel
import (
"context"
"fmt"
"math/rand"
"strings"
"sync"
"time"
"golang.org/x/sync/errgroup"
"golang.org/x/sync/semaphore"
)
func (m *masterInstance) keepSendingPeriodRequest() bool {
m.mu.Lock()
defer m.mu.Unlock()
return m.stopSendingRequest
}
func (m *masterInstance) killed() bool {
m.mu.Lock()
defer m.mu.Unlock()
return m.isKilled
}
func (s *Sentinel) masterPingRoutine(m *masterInstance) {
ticker := time.NewTicker(time.Second)
defer ticker.Stop()
for !m.keepSendingPeriodRequest() {
<-ticker.C
_, err := m.client.Ping()
if err != nil {
if time.Since(m.lastSuccessfulPing) > m.sentinelConf.DownAfter {
//notify if it is down
if m.getState() == masterStateUp {
m.mu.Lock()
m.state = masterStateSubjDown
m.downSince = time.Now()
s.logger.Warnf("master %s is subjectively down", m.name)
m.mu.Unlock()
select {
case m.subjDownNotify <- struct{}{}:
default:
}
}
}
continue
} else {
m.mu.Lock()
state := m.state
m.lastSuccessfulPing = time.Now()
m.mu.Unlock()
if state == masterStateSubjDown || state == masterStateObjDown {
m.mu.Lock()
m.state = masterStateUp
m.mu.Unlock()
//select {
//case m.reviveNotify <- struct{}{}:
//default:
//}
//revive
}
}
}
}
func (m *masterInstance) getFailOverStartTime() time.Time {
m.mu.Lock()
defer m.mu.Unlock()
return m.failOverStartTime
}
func (m *masterInstance) getFailOverEpoch() int {
m.mu.Lock()
defer m.mu.Unlock()
return m.failOverEpoch
}
func (m *masterInstance) getFailOverState() failOverState {
m.mu.Lock()
defer m.mu.Unlock()
return m.failOverState
}
func (m *masterInstance) getAddr() string {
m.mu.Lock()
defer m.mu.Unlock()
return fmt.Sprintf("%s:%s", m.host, m.port)
}
func (m *masterInstance) getState() masterInstanceState {
m.mu.Lock()
defer m.mu.Unlock()
return m.state
}
// for now, only say hello through master, redis also say hello through slaves
func (s *Sentinel) sayHelloRoutineToMaster(m *masterInstance, helloChan HelloChan) {
for !m.keepSendingPeriodRequest() {
time.Sleep(2 * time.Second)
m.mu.Lock()
masterName := m.name
masterIP := m.host
masterPort := m.port
masterConfigEpoch := m.configEpoch
m.mu.Unlock()
info := fmt.Sprintf("%s,%s,%s,%d,%s,%s,%s,%d",
s.conf.Binds[0], s.conf.Port, s.selfID(),
s.getCurrentEpoch(), masterName, masterIP, masterPort, masterConfigEpoch,
)
err := helloChan.Publish(info)
if err != nil {
s.logger.Errorf("SayHelloRoutineToMaster helloChan.Publish: %s", err)
}
}
}
/* Format is composed of 8 tokens:
* 0=ip,1=port,2=runid,3=current_epoch,4=master_name,
* 5=master_ip,6=master_port,7=master_config_epoch.
*/
// Todo: this function is per instance, not per master
func (s *Sentinel) subscribeHello(m *masterInstance) {
helloChan := m.client.SubscribeHelloChan()
defer helloChan.Close()
selfID := s.selfID()
go s.sayHelloRoutineToMaster(m, helloChan)
for !m.keepSendingPeriodRequest() {
newmsg, err := helloChan.Receive()
if err != nil {
s.logger.Errorf("helloChan.Receive: %s", err)
continue
//skip for now
}
parts := strings.Split(newmsg, ",")
if len(parts) != 8 {
s.logger.Errorf("helloChan.Receive: invalid format for newmsg: %s", newmsg)
continue
}
runid := parts[2]
if runid == selfID {
continue
}
m.mu.Lock()
_, ok := m.sentinels[runid]
if !ok {
client, err := NewRedisProtoInternalClient(parts[0], parts[1])
if err != nil {
s.logger.Errorf("NewRedisProtoInternalClient: cannot create new client to other sentinel with info: %s: %s", newmsg, err)
m.mu.Unlock()
continue
}
si := &sentinelInstance{
mu: sync.Mutex{},
sdown: false,
client: client,
runID: runid,
}
m.sentinels[runid] = si
s.logger.Infof("subscribeHello: connected to new sentinel: %s", newmsg)
}
m.mu.Unlock()
//ignore if exist for now
}
}
func (s *Sentinel) masterRoutine(m *masterInstance) {
s.logger.Debugw(logEventMasterInstanceCreated,
"master_name", m.name,
"run_id", m.runID,
"sentinel_run_id", s.runID,
"epoch", m.configEpoch,
"addr", m.getAddr(),
)
go s.masterPingRoutine(m)
go s.subscribeHello(m)
infoTicker := time.NewTicker(10 * time.Second)
for !m.killed() {
switch m.getState() {
case masterStateUp:
select {
case <-m.shutdownChan:
return
case <-m.subjDownNotify:
//notify allowing break select loop early, rather than waiting for 10s
infoTicker.Stop()
case <-infoTicker.C:
info, err := m.client.Info()
if err != nil {
//TODO continue for now
continue
}
roleSwitched, err := s.parseInfoMaster(m.getAddr(), info)
if err != nil {
s.logger.Errorf("parseInfoMaster: %v", err)
continue
// continue for now
}
if roleSwitched {
// if master has been switched to slave, this loop should
// should have been killed and started as a loop for slave
panic("unimlemented")
//TODO
// s.parseInfoSlave()
}
}
case masterStateSubjDown:
SdownLoop:
for {
switch m.getState() {
case masterStateSubjDown:
// check if quorum as met
s.askSentinelsIfMasterIsDown(m)
s.checkObjDown(m)
if m.getState() == masterStateObjDown {
s.notifyMasterDownToSlaveRoutines(m)
break SdownLoop
}
case masterStateObjDown:
panic("no other process should set masterStateObjDown")
case masterStateUp:
break SdownLoop
}
time.Sleep(1 * time.Second)
}
case masterStateObjDown:
// stopAskingOtherSentinels: used when the fsm state wnats to turn back to this current line
// canceling the on going goroutine, or when the failover is successful
ctx, stopAskingOtherSentinels := context.WithCancel(context.Background())
defer stopAskingOtherSentinels()
go s.askOtherSentinelsEach1Sec(ctx, m)
sleepToNextFailover := s.nextFailoverAllowed(m)
timer := time.NewTimer(sleepToNextFailover)
defer timer.Stop()
select {
case <-timer.C:
case <-m.followerNewMasterNotify:
s.resetMasterInstance(m, true)
// this is a follower sentinel and new master recognize
// from leader sentinel through hello message
return
}
// If any logic changing state of master to something != masterStateObjDown, this code below will be broken
failOverFSM:
for {
switch m.getFailOverState() {
// non-leader can not reach next step
case failOverWaitLeaderElection:
aborted, isleader := s.checkElectionStatus(m)
if isleader {
stopAskingOtherSentinels()
// to next state of fsm
continue
}
if aborted {
stopAskingOtherSentinels()
break failOverFSM
}
ticker := time.NewTicker(1 * time.Second)
select {
case <-ticker.C:
case <-m.followerNewMasterNotify:
s.resetMasterInstance(m, true)
// this is a follower sentinel and new master recognize
// from leader sentinel through hello message
return
}
case failOverSelectSlave:
slave := s.selectSlave(m)
// abort failover, start voting for new epoch
if slave == nil {
s.abortFailover(m)
} else {
s.promoteSlave(m, slave)
}
case failOverPromoteSlave:
m.mu.Lock()
promotedSlave := m.promotedSlave
m.mu.Unlock()
timeout := time.NewTimer(m.sentinelConf.FailoverTimeout)
defer timeout.Stop()
select {
case <-timeout.C:
s.abortFailover(m)
// when new slave recognized to have switched role, it will notify this channel
// or time out
case <-promotedSlave.masterRoleSwitchChan:
var (
epoch int
)
locked(&m.mu, func() {
m.failOverState = failOverReconfSlave
m.failOverStateChangeTime = time.Now()
epoch = m.failOverEpoch
m.configEpoch = epoch
})
s.logger.Debugw(logEventSlavePromoted,
"run_id", promotedSlave.runID,
"epoch", epoch)
s.logger.Debugw(logEventFailoverStateChanged,
"epoch", m.getFailOverEpoch(),
"new_state", failOverReconfSlave)
}
case failOverReconfSlave:
s.reconfigRemoteSlaves(m)
s.logger.Debugw(logEventFailoverStateChanged,
"epoch", m.getFailOverEpoch(),
"new_state", failOverResetInstance)
s.resetMasterInstance(m, false)
return
}
}
}
}
}
func (s *Sentinel) resetMasterInstance(m *masterInstance, isfollower bool) {
var (
promoted *slaveInstance
oldAddr string
epoch int
name string
promotedRunID, promotedAddr, promotedHost, promotedPort string
sentinels map[string]*sentinelInstance
)
locked(&m.mu, func() {
if isfollower {
epoch = m.leaderEpoch
} else {
epoch = m.failOverEpoch
}
promoted = m.promotedSlave
name = m.name
m.isKilled = true
oldAddr = fmt.Sprintf("%s:%s", m.host, m.port)
sentinels = m.sentinels
})
locked(&promoted.mu, func() {
promotedAddr = promoted.addr
promotedHost, promotedPort = promoted.host, promoted.port
promotedRunID = m.promotedSlave.runID
})
cl, err := s.clientFactory(promotedAddr)
if err != nil {
panic("todo")
}
newSlaves := map[string]*slaveInstance{}
newMaster := &masterInstance{
runID: promotedRunID,
sentinelConf: m.sentinelConf,
name: name,
host: promotedHost,
port: promotedPort,
configEpoch: m.configEpoch,
mu: sync.Mutex{},
client: cl,
slaves: map[string]*slaveInstance{},
sentinels: sentinels,
state: masterStateUp,
lastSuccessfulPing: time.Now(),
subjDownNotify: make(chan struct{}),
followerNewMasterNotify: make(chan struct{}, 1),
}
locked(&m.mu, func() {
for _, sl := range m.slaves {
locked(&sl.mu, func() {
// kill all running slaves routine, for simplicity
sl.killed = true
if sl.addr != promotedAddr {
newslave, err := s.newSlaveInstance(promotedHost, promotedPort, sl.host, sl.port, sl.replOffset, newMaster)
if err == nil {
newSlaves[sl.addr] = newslave
}
}
})
}
// turn dead master into slave as well
newslave, err := s.newSlaveInstance(promotedHost, promotedPort, m.host, m.port, 0, newMaster)
if err != nil {
} else {
newSlaves[oldAddr] = newslave
}
m.slaves = newSlaves
})
locked(s.mu, func() {
delete(s.masterInstances, oldAddr)
s.masterInstances[newMaster.getAddr()] = newMaster
})
for idx := range newSlaves {
go s.slaveRoutine(newSlaves[idx])
}
s.logger.Debugw(logEventFailoverStateChanged,
"new_state", failOverDone,
"epoch", epoch,
)
go s.masterRoutine(newMaster)
}
// TODO: write test for this func
func (s *Sentinel) reconfigRemoteSlaves(m *masterInstance) error {
m.mu.Lock()
slaveList := make([]*slaveInstance, 0, len(m.slaves))
var promotedHost, promotedPort string
for runID := range m.slaves {
slaveList = append(slaveList, m.slaves[runID])
}
m.mu.Unlock()
sema := semaphore.NewWeighted(int64(m.sentinelConf.ParallelSync))
errgrp := &errgroup.Group{}
ctx, cancel := context.WithTimeout(context.Background(), m.sentinelConf.ReconfigSlaveTimeout)
defer cancel()
locked(&m.promotedSlave.mu, func() {
promotedHost, promotedPort = m.promotedSlave.host, m.promotedSlave.port
})
for idx := range slaveList {
slave := slaveList[idx]
if slave.runID == m.promotedSlave.runID {
continue
}
errgrp.Go(func() error {
var done bool
for {
err := sema.Acquire(ctx, 1)
if err != nil {
return err
}
// goroutine exit if ctx is canceled
select {
case <-ctx.Done():
return ctx.Err()
default:
}
err = slave.client.SlaveOf(promotedHost, promotedPort)
if err != nil {
sema.Release(1)
time.Sleep(100 * time.Millisecond)
continue
}
slave.mu.Lock()
slave.reconfigFlag |= reconfigSent
slave.mu.Unlock()
break
}
for {
slave.mu.Lock()
if slave.reconfigFlag&reconfigDone != 0 {
slave.reconfigFlag = 0
done = true
}
slave.mu.Unlock()
if done {
sema.Release(1)
return nil
}
select {
case <-ctx.Done():
return ctx.Err()
default:
}
// TODO: may use sync.Cond instead of ticking
time.Sleep(1 * time.Second)
}
})
}
return errgrp.Wait()
// update state no matter what
}
func (s *Sentinel) nextFailoverAllowed(m *masterInstance) time.Duration {
// this code only has to wait in case failover timeout reached and block until the next failover is allowed to
// continue (2* failover timeout)
retry := false
startTime := m.getFailOverStartTime()
secondsLeft := 2*m.sentinelConf.FailoverTimeout - time.Since(startTime)
if secondsLeft <= 0 {
locked(s.mu, func() {
locked(&m.mu, func() {
if m.failOverStartTime.Equal(startTime) {
// if some how, this sentinel receive new epoch from hello message
// before this code block runs,
// it will uncessarily increment the epoch one more time, causing already
// voting candidates to fail, need logic to check this
// TODO
s.currentEpoch += 1
if m.failOverState != failOverWaitLeaderElection {
s.logger.Debugw(logEventFailoverStateChanged,
"new_state", failOverWaitLeaderElection,
"epoch", s.currentEpoch,
)
s.logger.Debugw(logEventRequestingElection,
"epoch", s.currentEpoch,
"sentinel_run_id", s.runID,
)
}
m.failOverState = failOverWaitLeaderElection
m.failOverStartTime = time.Now()
m.failOverEpoch = s.currentEpoch
} else {
// some other thread has modify this start time already
retry = true
}
})
})
// mostly, when obj down is met, multiples sentinels will try to send request for vote to be leader
// to prevent split vote, sleep for a random small duration
secondsLeft = 0
}
if retry {
return s.nextFailoverAllowed(m)
}
return secondsLeft + time.Duration(rand.Intn(SENTINEL_MAX_DESYNC)*int(time.Millisecond))
}
func (s *Sentinel) checkElectionStatus(m *masterInstance) (aborted bool, isLeader bool) {
leader, epoch := s.checkWhoIsLeader(m)
isLeader = leader != "" && leader == s.selfID()
if !isLeader {
//abort if failover take too long
if time.Since(m.getFailOverStartTime()) > m.sentinelConf.FailoverTimeout {
s.abortFailover(m)
return true, false
}
return false, false
}
// do not call cancel(), keep receiving update from other sentinel
s.logger.Debugw(logEventBecameTermLeader,
"run_id", leader,
"epoch", epoch)
m.mu.Lock()
m.failOverState = failOverSelectSlave
m.mu.Unlock()
s.logger.Debugw(logEventFailoverStateChanged,
"new_state", failOverSelectSlave,
"epoch", epoch, // epoch = failover epoch = sentinel current epoch
)
return false, true
}
func (s *Sentinel) promoteSlave(m *masterInstance, slave *slaveInstance) {
m.mu.Lock()
m.promotedSlave = slave
m.failOverState = failOverPromoteSlave
m.failOverStateChangeTime = time.Now()
epoch := m.failOverEpoch
m.mu.Unlock()
slave.mu.Lock()
slaveAddr := fmt.Sprintf("%s:%s", slave.host, slave.port)
slaveID := slave.runID
client := slave.client
slave.mu.Unlock()
s.logger.Debugw(logEventSelectedSlave,
"slave_addr", slaveAddr,
"slave_id", slaveID,
"epoch", epoch,
)
s.logger.Debugw(logEventFailoverStateChanged,
"new_state", failOverPromoteSlave,
"epoch", epoch,
)
err := client.SlaveOfNoOne()
if err != nil {
s.logger.Errorf("send slave of no one to slave %s return error %s", slaveAddr, err.Error())
}
}
func (s *Sentinel) abortFailover(m *masterInstance) {
m.mu.Lock()
m.failOverState = failOverNone
m.failOverStateChangeTime = time.Now()
epoch := m.failOverEpoch
m.mu.Unlock()
s.logger.Debugw(logEventFailoverStateChanged,
"new_state", failOverNone,
"epoch", epoch,
)
}
func (s *Sentinel) notifyMasterDownToSlaveRoutines(m *masterInstance) {
m.mu.Lock()
for idx := range m.slaves {
m.slaves[idx].masterDownNotify <- struct{}{}
}
m.mu.Unlock()
}
func (s *Sentinel) checkWhoIsLeader(m *masterInstance) (string, int) {
instancesVotes := map[string]int{}
currentEpoch := s.getCurrentEpoch()
// instancesVoters := map[string][]string{}
m.mu.Lock()
for _, sen := range m.sentinels {
sen.mu.Lock()
if sen.leaderID != "" && sen.leaderEpoch == currentEpoch {
instancesVotes[sen.leaderID] = instancesVotes[sen.leaderID] + 1
// instancesVoters[sen.leaderID] = append(instancesVoters[sen.leaderID], sen.runID)
}
sen.mu.Unlock()
}
totalSentinels := len(m.sentinels) + 1
m.mu.Unlock()
maxVote := 0
var winner string
for runID, vote := range instancesVotes {
if vote > maxVote {
maxVote = vote
winner = runID
}
}
epoch := m.getFailOverEpoch()
var leaderEpoch int
var votedByMe string
if winner != "" {
// vote for the most winning candidate
leaderEpoch, votedByMe = s.voteLeader(m, epoch, winner)
} else {
leaderEpoch, votedByMe = s.voteLeader(m, epoch, s.selfID())
}
if votedByMe != "" && leaderEpoch == epoch {
instancesVotes[votedByMe] = instancesVotes[votedByMe] + 1
if instancesVotes[votedByMe] > maxVote {
maxVote = instancesVotes[votedByMe]
winner = votedByMe
}
// instancesVoters[votedByMe] = append(instancesVoters[votedByMe], s.selfID())
}
quorum := totalSentinels/2 + 1
if winner != "" && (maxVote < quorum || maxVote < m.sentinelConf.Quorum) {
winner = ""
}
return winner, currentEpoch
}
func (s *Sentinel) checkObjDown(m *masterInstance) {
m.mu.Lock()
defer m.mu.Unlock()
quorum := m.sentinelConf.Quorum
if len(m.sentinels)+1 < quorum {
panic(fmt.Sprintf("quorum too large (%d) or too few sentinel instances (%d)", quorum, len(m.sentinels)))
// TODO: log warning only
}
down := 1
for _, sentinel := range m.sentinels {
sentinel.mu.Lock()
if sentinel.sdown {
down++
}
sentinel.mu.Unlock()
}
if down >= quorum {
m.state = masterStateObjDown
m.stopSendingRequest = true
s.logger.Debugw(logEventODown)
}
}
func (s *Sentinel) askOtherSentinelsEach1Sec(ctx context.Context, m *masterInstance) {
m.mu.Lock()
masterName := m.name
masterIp := m.host
masterPort := m.port
for idx := range m.sentinels {
sentinel := m.sentinels[idx]
go func() {
var loggedForNeighbor bool
for m.getState() >= masterStateSubjDown {
select {
case <-ctx.Done():
return
default:
}
s.mu.Lock()
currentEpoch := s.currentEpoch //epoch may change during failover
selfID := s.runID // locked as well for sure
s.mu.Unlock()
// do not ask for vote if has not started failover
if m.getFailOverState() == failOverNone {
selfID = ""
}
reply, err := sentinel.client.IsMasterDownByAddr(IsMasterDownByAddrArgs{
Name: masterName,
IP: masterIp,
Port: masterPort,
CurrentEpoch: currentEpoch,
SelfID: selfID,
})
if err != nil {
s.logger.Errorf("sentinel.client.IsMasterDownByAddr: %s", err)
//skip for now
} else {
sentinel.mu.Lock()
sentinel.sdown = reply.MasterDown
if reply.VotedLeaderID != "" {
if !loggedForNeighbor {
s.logger.Debugw(logEventNeighborVotedFor,
"neighbor_id", sentinel.runID,
"voted_for", reply.VotedLeaderID,
"epoch", reply.LeaderEpoch,
)
loggedForNeighbor = true
}
sentinel.leaderEpoch = reply.LeaderEpoch
sentinel.leaderID = reply.VotedLeaderID
}
sentinel.mu.Unlock()
}
time.Sleep(1 * time.Second)
}
}()
}
m.mu.Unlock()
}
func (s *Sentinel) askSentinelsIfMasterIsDown(m *masterInstance) {
s.mu.Lock()
currentEpoch := s.currentEpoch
s.mu.Unlock()
m.mu.Lock()
masterName := m.name
masterIp := m.host
masterPort := m.port
for _, sentinel := range m.sentinels {
go func(sInstance *sentinelInstance) {
sInstance.mu.Lock()
lastReply := sInstance.lastMasterDownReply
sInstance.mu.Unlock()
if time.Since(lastReply) < 1*time.Second {
return
}
if m.getState() == masterStateSubjDown {
reply, err := sInstance.client.IsMasterDownByAddr(IsMasterDownByAddrArgs{
Name: masterName,
IP: masterIp,
Port: masterPort,
CurrentEpoch: currentEpoch,
SelfID: "",
})
if err != nil {
fmt.Println(err)
//skip for now
} else {
sInstance.mu.Lock()
sInstance.lastMasterDownReply = time.Now()
sInstance.sdown = reply.MasterDown
sInstance.mu.Unlock()
}
} else {
return
}
}(sentinel)
}
m.mu.Unlock()
}
type masterInstance struct {
sentinelConf MasterMonitor
isKilled bool
stopSendingRequest bool
name string
mu sync.Mutex
runID string
slaves map[string]*slaveInstance
promotedSlave *slaveInstance
sentinels map[string]*sentinelInstance
host string
port string
shutdownChan chan struct{}
client InternalClient
// infoClient internalClient
state masterInstanceState
subjDownNotify chan struct{}
followerNewMasterNotify chan struct{}
downSince time.Time
lastSuccessfulPing time.Time
failOverState failOverState
failOverEpoch int
failOverStartTime time.Time
failOverStateChangeTime time.Time
leaderEpoch int // epoch of current leader, only set during failover
leaderID string // current leader id
// failOverStartTime time.Time
configEpoch int
}
type failOverState int
var (
failOverNone failOverState = 0
failOverWaitLeaderElection failOverState = 1
failOverSelectSlave failOverState = 2
failOverPromoteSlave failOverState = 3
failOverReconfSlave failOverState = 4
failOverResetInstance failOverState = 5
failOverDone failOverState = 6
failOverStateValueMap = map[string]failOverState{
"none": failOverNone,
"wait_leader_election": failOverWaitLeaderElection,
"select_slave": failOverSelectSlave,
"promote_slave": failOverPromoteSlave,
"reconfig_slave": failOverReconfSlave,
"failover_done": failOverDone,
"reset_instance": failOverResetInstance,
}
failOverStateMap = map[failOverState]string{
0: "none",
1: "wait_leader_election",
2: "select_slave",
3: "promote_slave",
4: "reconfig_slave",
5: "reset_instance",
6: "failover_done",
}
)
func (s failOverState) String() string {
return failOverStateMap[s]
}