Repository navigation
Expand file tree
/
Copy pathchecker.go
More file actions
348 lines (310 loc) · 13.9 KB
/
Copy pathchecker.go
File metadata and controls
348 lines (310 loc) · 13.9 KB
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
package redisfailover
import (
"errors"
"strconv"
"time"
redisfailoverv1 "github.com/freshworks/redis-operator/api/redisfailover/v1"
"github.com/freshworks/redis-operator/metrics"
)
// UpdateRedisesPods deletes Redis pods with a stale StatefulSet revision (OnDelete rollout).
// DisableMasterRollout when true, only slave pods are rolled out on spec change (OnDelete); master is not deleted until the flag is removed.
func (r *RedisFailoverHandler) UpdateRedisesPods(rf *redisfailoverv1.RedisFailover) error {
redises, err := r.rfChecker.GetRedisesIPs(rf)
if err != nil {
return err
}
masterIP := ""
if !rf.Bootstrapping() {
masterIP, _ = r.rfChecker.GetMasterIP(rf)
}
// No perform updates when nodes are syncing, still not connected, etc.
for _, rip := range redises {
if rip != masterIP {
ready, err := r.rfChecker.CheckRedisSlavesReady(rip, rf)
if err != nil {
return err
}
if !ready {
return nil
}
}
}
ssUR, err := r.rfChecker.GetStatefulSetUpdateRevision(rf)
if err != nil {
return err
}
redisesPods, err := r.rfChecker.GetRedisesSlavesPods(rf)
if err != nil {
return err
}
// Update stale pods with slave role
for _, pod := range redisesPods {
revision, err := r.rfChecker.GetRedisRevisionHash(pod, rf)
if err != nil {
return err
}
if revision != ssUR {
//Delete pod and wait next round to check if the new one is synced
err = r.rfHealer.DeletePod(pod, rf)
if err != nil {
return err
}
return nil
}
}
if !rf.Bootstrapping() && !rf.Spec.Redis.DisableMasterRollout {
master, err := r.rfChecker.GetRedisesMasterPod(rf)
if err != nil {
return err
}
masterRevision, err := r.rfChecker.GetRedisRevisionHash(master, rf)
if err != nil {
return err
}
if masterRevision != ssUR {
err = r.rfHealer.DeletePod(master, rf)
if err != nil {
return err
}
return nil
}
}
return nil
}
// CheckAndHeal runs verifcation checks to ensure the RedisFailover is in an expected and healthy state.
// If the checks do not match up to expectations, an attempt will be made to "heal" the RedisFailover into a healthy state.
func (r *RedisFailoverHandler) CheckAndHeal(rf *redisfailoverv1.RedisFailover) error {
if rf.Bootstrapping() {
return r.checkAndHealBootstrapMode(rf)
}
// Number of redis is equal as the set on the RF spec
// Number of sentinel is equal as the set on the RF spec
// Check only one master
// Number of redis master is 1
// All redis slaves have the same master
// All sentinels points to the same redis master
// Sentinel has not death nodes
// Sentinel knows the correct slave number
if !r.rfChecker.IsRedisRunning(rf) {
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.REDIS_REPLICA_MISMATCH, metrics.NOT_APPLICABLE, errors.New("not all replicas running"))
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Debugf("Number of redis mismatch, waiting for redis statefulset reconcile")
return nil
}
if !r.rfChecker.IsSentinelRunning(rf) {
setRedisCheckerMetrics(r.mClient, "sentinel", rf.Namespace, rf.Name, metrics.SENTINEL_REPLICA_MISMATCH, metrics.NOT_APPLICABLE, errors.New("not all replicas running"))
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Debugf("Number of sentinel mismatch, waiting for sentinel deployment reconcile")
return nil
}
nMasters, err := r.rfChecker.GetNumberMasters(rf)
if err != nil {
return err
}
switch nMasters {
case 0:
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.NO_MASTER, metrics.NOT_APPLICABLE, errors.New("no masters detected"))
//when number of redis replicas is 1 , the redis is configured for standalone master mode
//Configure to master
if rf.Spec.Redis.Replicas == 1 {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Infof("Resource spec with standalone master - operator will set the master")
err = r.rfHealer.SetOldestAsMaster(rf)
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.NO_MASTER, metrics.NOT_APPLICABLE, err)
if err != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Errorf("Error in Setting oldest Pod as master")
return err
}
return nil
}
//During the First boot(New deployment or all pods of the statefulsets have restarted),
//Sentinesl will not be able to choose the master , so operator should select a master
//Also in scenarios where Sentinels is not in a position to choose a master like , No quorum reached
//Operator can choose a master , These scenarios can be checked by asking the all the sentinels
//if its in a postion to choose a master also check if the redis is configured with local host IP as master.
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Warningf("Number of Masters running is 0")
maxUptime, err := r.rfChecker.GetMaxRedisPodTime(rf)
if err != nil {
return err
}
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Infof("No master avaiable but max pod up time is : %f", maxUptime.Round(time.Second).Seconds())
//Check If Sentinel has quorum to take a failover decision
noqrm_cnt, err := r.rfChecker.CheckSentinelQuorum(rf)
if err != nil {
// Sentinels are not in a situation to choose a master we pick one
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Warningf("Quorum not available for sentinel to choose master,estimated unhealthy sentinels :%d , Operator to step-in", noqrm_cnt)
err2 := r.rfHealer.SetOldestAsMaster(rf)
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.NO_MASTER, metrics.NOT_APPLICABLE, err2)
if err2 != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Errorf("Error in Setting oldest Pod as master")
return err2
}
} else {
//sentinels are having a quorum to make a failover , but check if redis are not having local hostip (first boot) as master
status, err2 := r.rfChecker.CheckIfMasterLocalhost(rf)
if err2 != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Errorf("CheckIfMasterLocalhost failed retry later")
return err2
} else if status {
// all avaialable redis pods have local host ip as master
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Errorf("all available redis is having local loop back as master , operator initiates master selection")
err3 := r.rfHealer.SetOldestAsMaster(rf)
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.NO_MASTER, metrics.NOT_APPLICABLE, err3)
if err3 != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Errorf("Error in Setting oldest Pod as master")
return err3
}
} else {
// We'll wait until failover is done
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Infof("no master found, wait until failover or fix manually")
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.NO_MASTER, metrics.NOT_APPLICABLE, errors.New("no master not fixed, wait until failover or fix manually"))
return nil
}
}
case 1:
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.NUMBER_OF_MASTERS, metrics.NOT_APPLICABLE, nil)
default:
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.NUMBER_OF_MASTERS, metrics.NOT_APPLICABLE, errors.New("multiple masters detected"))
return errors.New("more than one master, fix manually")
}
master, err := r.rfChecker.GetMasterIP(rf)
if err != nil {
return err
}
err = r.rfChecker.CheckAllSlavesFromMaster(master, rf)
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.SLAVE_WRONG_MASTER, metrics.NOT_APPLICABLE, err)
if err != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Warningf("Slave not associated to master: %s", err.Error())
if err = r.rfHealer.SetMasterOnAll(master, rf); err != nil {
return err
}
}
err = r.applyRedisCustomConfig(rf)
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.APPLY_REDIS_CONFIG, metrics.NOT_APPLICABLE, err)
if err != nil {
return err
}
err = r.UpdateRedisesPods(rf)
if err != nil {
return err
}
sentinels, err := r.rfChecker.GetSentinelsIPs(rf)
if err != nil {
return err
}
port := getRedisPort(rf.Spec.Redis.Port)
for _, sip := range sentinels {
err = r.rfChecker.CheckSentinelMonitor(sip, rf.MasterName(), master, port)
setRedisCheckerMetrics(r.mClient, "sentinel", rf.Namespace, rf.Name, metrics.SENTINEL_WRONG_MASTER, sip, err)
if err != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Warningf("Fixing sentinel not monitoring expected master: %s", err.Error())
if err := r.rfHealer.NewSentinelMonitor(sip, master, rf); err != nil {
return err
}
}
}
return r.checkAndHealSentinels(rf, sentinels)
}
func (r *RedisFailoverHandler) checkAndHealBootstrapMode(rf *redisfailoverv1.RedisFailover) error {
if !r.rfChecker.IsRedisRunning(rf) {
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.REDIS_REPLICA_MISMATCH, metrics.NOT_APPLICABLE, errors.New("not all replicas running"))
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Debugf("Number of redis mismatch, waiting for redis statefulset reconcile")
return nil
}
err := r.UpdateRedisesPods(rf)
if err != nil {
return err
}
err = r.applyRedisCustomConfig(rf)
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.APPLY_REDIS_CONFIG, metrics.NOT_APPLICABLE, err)
if err != nil {
return err
}
bootstrapSettings := rf.Spec.BootstrapNode
err = r.rfHealer.SetExternalMasterOnAll(bootstrapSettings.Host, bootstrapSettings.Port, rf)
setRedisCheckerMetrics(r.mClient, "redis", rf.Namespace, rf.Name, metrics.APPLY_EXTERNAL_MASTER, metrics.NOT_APPLICABLE, err)
if err != nil {
return err
}
if rf.SentinelsAllowed() {
if !r.rfChecker.IsSentinelRunning(rf) {
setRedisCheckerMetrics(r.mClient, "sentinel", rf.Namespace, rf.Name, metrics.SENTINEL_REPLICA_MISMATCH, metrics.NOT_APPLICABLE, errors.New("not all replicas running"))
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Debugf("Number of sentinel mismatch, waiting for sentinel deployment reconcile")
return nil
}
sentinels, err := r.rfChecker.GetSentinelsIPs(rf)
if err != nil {
return err
}
for _, sip := range sentinels {
err = r.rfChecker.CheckSentinelMonitor(sip, rf.MasterName(), bootstrapSettings.Host, bootstrapSettings.Port)
setRedisCheckerMetrics(r.mClient, "sentinel", rf.Namespace, rf.Name, metrics.SENTINEL_WRONG_MASTER, sip, err)
if err != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Warningf("Fixing sentinel not monitoring expected master: %s", err.Error())
if err := r.rfHealer.NewSentinelMonitorWithPort(sip, bootstrapSettings.Host, bootstrapSettings.Port, rf); err != nil {
return err
}
}
}
return r.checkAndHealSentinels(rf, sentinels)
}
return nil
}
func (r *RedisFailoverHandler) applyRedisCustomConfig(rf *redisfailoverv1.RedisFailover) error {
redises, err := r.rfChecker.GetRedisesIPs(rf)
if err != nil {
return err
}
for _, rip := range redises {
if err := r.rfHealer.SetRedisCustomConfig(rip, rf); err != nil {
return err
}
}
return nil
}
func (r *RedisFailoverHandler) checkAndHealSentinels(rf *redisfailoverv1.RedisFailover, sentinels []string) error {
for _, sip := range sentinels {
err := r.rfChecker.CheckSentinelNumberInMemory(sip, rf)
setRedisCheckerMetrics(r.mClient, "sentinel", rf.Namespace, rf.Name, metrics.SENTINEL_NUMBER_IN_MEMORY_MISMATCH, sip, err)
if err != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Warningf("Sentinel %s mismatch number of sentinels in memory. resetting", sip)
if err := r.rfHealer.RestoreSentinel(sip); err != nil {
return err
}
}
}
for _, sip := range sentinels {
err := r.rfChecker.CheckSentinelSlavesNumberInMemory(sip, rf)
setRedisCheckerMetrics(r.mClient, "sentinel", rf.Namespace, rf.Name, metrics.REDIS_SLAVES_NUMBER_IN_MEMORY_MISMATCH, sip, err)
if err != nil {
r.logger.WithField("redisfailover", rf.ObjectMeta.Name).WithField("namespace", rf.ObjectMeta.Namespace).Warningf("Sentinel %s mismatch number of expected slaves in memory. resetting", sip)
if err := r.rfHealer.RestoreSentinel(sip); err != nil {
return err
}
}
}
for _, sip := range sentinels {
err := r.rfHealer.SetSentinelCustomConfig(sip, rf)
setRedisCheckerMetrics(r.mClient, "sentinel", rf.Namespace, rf.Name, metrics.APPLY_SENTINEL_CONFIG, sip, err)
if err != nil {
return err
}
}
return nil
}
func getRedisPort(p int32) string {
return strconv.Itoa(int(p))
}
func setRedisCheckerMetrics(metricsClient metrics.Recorder, mode /* redis or sentinel? */ string, rfNamespace string, rfName string, property string, IP string, err error) {
switch mode {
case "sentinel":
if err != nil {
metricsClient.RecordSentinelCheck(rfNamespace, rfName, property, IP, metrics.STATUS_UNHEALTHY)
} else {
metricsClient.RecordSentinelCheck(rfNamespace, rfName, property, IP, metrics.STATUS_HEALTHY)
}
default: // redis
if err != nil {
metricsClient.RecordRedisCheck(rfNamespace, rfName, property, IP, metrics.STATUS_UNHEALTHY)
} else {
metricsClient.RecordRedisCheck(rfNamespace, rfName, property, IP, metrics.STATUS_HEALTHY)
}
}
}