forked from openark/orchestrator
/
orchestrator.go
363 lines (314 loc) · 11 KB
/
orchestrator.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
/*
Copyright 2014 Outbrain Inc.
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 logic
import (
"os"
"os/signal"
"sync/atomic"
"syscall"
"time"
"github.com/github/orchestrator/go/agent"
"github.com/github/orchestrator/go/config"
"github.com/github/orchestrator/go/discovery"
"github.com/github/orchestrator/go/inst"
ometrics "github.com/github/orchestrator/go/metrics"
"github.com/github/orchestrator/go/process"
"github.com/openark/golib/log"
"github.com/patrickmn/go-cache"
"github.com/rcrowley/go-metrics"
)
// discoveryQueue is a channel of deduplicated instanceKey-s
// that were requested for discovery. It can be continuously updated
// as discovery process progresses.
var discoveryQueue *discovery.Queue
var discoveriesCounter = metrics.NewCounter()
var failedDiscoveriesCounter = metrics.NewCounter()
var discoveryQueueLengthGauge = metrics.NewGauge()
var discoveryRecentCountGauge = metrics.NewGauge()
var isElectedGauge = metrics.NewGauge()
var isElectedNode int64 = 0
var recentDiscoveryOperationKeys *cache.Cache
func init() {
metrics.Register("discoveries.attempt", discoveriesCounter)
metrics.Register("discoveries.fail", failedDiscoveriesCounter)
metrics.Register("discoveries.queue_length", discoveryQueueLengthGauge)
metrics.Register("discoveries.recent_count", discoveryRecentCountGauge)
metrics.Register("elect.is_elected", isElectedGauge)
ometrics.OnGraphiteTick(func() { discoveryQueueLengthGauge.Update(int64(discoveryQueue.QueueLen())) })
ometrics.OnGraphiteTick(func() {
if recentDiscoveryOperationKeys == nil {
return
}
discoveryRecentCountGauge.Update(int64(recentDiscoveryOperationKeys.ItemCount()))
})
ometrics.OnGraphiteTick(func() { isElectedGauge.Update(int64(atomic.LoadInt64(&isElectedNode))) })
}
// acceptSignals registers for OS signals
func acceptSignals() {
c := make(chan os.Signal, 1)
signal.Notify(c, syscall.SIGHUP)
go func() {
for sig := range c {
switch sig {
case syscall.SIGHUP:
log.Debugf("Received SIGHUP. Reloading configuration")
config.Reload()
inst.AuditOperation("reload-configuration", nil, "Triggered via SIGHUP")
}
}
}()
}
// handleDiscoveryRequests iterates the discoveryQueue channel and calls upon
// instance discovery per entry.
func handleDiscoveryRequests() {
discoveryQueue = discovery.CreateOrReturnQueue("DEFAULT")
// create a pool of discovery workers
for i := uint(0); i < config.Config.DiscoveryMaxConcurrency; i++ {
go func() {
for {
instanceKey := discoveryQueue.Consume()
// Possibly this used to be the elected node, but has
// been demoted, while still the queue is full.
if atomic.LoadInt64(&isElectedNode) != 1 {
log.Debugf("Node apparently demoted. Skipping discovery of %+v. "+
"Remaining queue size: %+v", instanceKey, discoveryQueue.QueueLen())
discoveryQueue.Release(instanceKey)
continue
}
discoverInstance(instanceKey)
discoveryQueue.Release(instanceKey)
}
}()
}
}
// discoverInstance will attempt discovering an instance (unless it is already up to date) and will
// list down its master and replicas (if any) for further discovery.
func discoverInstance(instanceKey inst.InstanceKey) {
start := time.Now()
defer func() {
discoveryTime := time.Since(start)
if discoveryTime > time.Duration(config.Config.InstancePollSeconds)*time.Second {
log.Warningf("discoverInstance for key %v took %.4fs", instanceKey, discoveryTime.Seconds())
}
}()
instanceKey.Formalize()
if !instanceKey.IsValid() {
return
}
if existsInCacheError := recentDiscoveryOperationKeys.Add(instanceKey.DisplayString(), true, cache.DefaultExpiration); existsInCacheError != nil {
// Just recently attempted
return
}
instance, found, err := inst.ReadInstance(&instanceKey)
if found && instance.IsUpToDate && instance.IsLastCheckValid {
// we've already discovered this one. Skip!
return
}
discoveriesCounter.Inc(1)
// First we've ever heard of this instance. Continue investigation:
instance, err = inst.ReadTopologyInstanceBufferable(&instanceKey, config.Config.BufferInstanceWrites)
// panic can occur (IO stuff). Therefore it may happen
// that instance is nil. Check it.
if instance == nil {
failedDiscoveriesCounter.Inc(1)
log.Warningf("discoverInstance(%+v) instance is nil in %.3fs, error=%+v", instanceKey, time.Since(start).Seconds(), err)
return
}
log.Debugf("Discovered host: %+v, master: %+v, version: %+v in %.3fs", instance.Key, instance.MasterKey, instance.Version, time.Since(start).Seconds())
if atomic.LoadInt64(&isElectedNode) == 0 {
// Maybe this node was elected before, but isn't elected anymore.
// If not elected, stop drilling up/down the topology
return
}
// Investigate replicas:
for _, replicaKey := range instance.SlaveHosts.GetInstanceKeys() {
replicaKey := replicaKey
if replicaKey.IsValid() {
discoveryQueue.Push(replicaKey)
}
}
// Investigate master:
if instance.MasterKey.IsValid() {
discoveryQueue.Push(instance.MasterKey)
}
}
func onDiscoveryTick() {
wasAlreadyElected := atomic.LoadInt64(&isElectedNode)
myIsElectedNode, err := process.AttemptElection()
if err != nil {
log.Errore(err)
}
if myIsElectedNode {
atomic.StoreInt64(&isElectedNode, 1)
} else {
atomic.StoreInt64(&isElectedNode, 0)
}
if !myIsElectedNode {
if electedNode, _, err := process.ElectedNode(); err == nil {
log.Debugf("Not elected as active node; active node: %v; polling", electedNode.Hostname)
} else {
log.Debugf("Not elected as active node; active node: Unable to determine: %v; polling", err)
}
return
}
// I'm elected!
instanceKeys, err := inst.ReadOutdatedInstanceKeys()
if err != nil {
log.Errore(err)
}
if wasAlreadyElected == 0 {
// Just turned to be leader!
go process.RegisterNode("", "", false)
}
if len(instanceKeys) > 0 {
if len(instanceKeys) > config.Config.MaxOutdatedKeysToShow {
log.Debugf("polling %d outdated keys", len(instanceKeys))
} else {
log.Debugf("outdated keys: %+v", instanceKeys)
}
for _, instanceKey := range instanceKeys {
instanceKey := instanceKey
if instanceKey.IsValid() {
discoveryQueue.Push(instanceKey)
}
}
}
}
// ContinuousDiscovery starts an asynchronuous infinite discovery process where instances are
// periodically investigated and their status captured, and long since unseen instances are
// purged and forgotten.
func ContinuousDiscovery() {
if config.Config.DatabaselessMode__experimental {
log.Fatal("Cannot execute continuous mode in databaseless mode")
}
log.Infof("Starting continuous discovery")
recentDiscoveryOperationKeys = cache.New(time.Duration(config.Config.InstancePollSeconds)*time.Second, time.Second)
inst.LoadHostnameResolveCache()
go handleDiscoveryRequests()
discoveryTick := time.Tick(time.Duration(config.Config.GetDiscoveryPollSeconds()) * time.Second)
instancePollTick := time.Tick(time.Duration(config.Config.InstancePollSeconds) * time.Second)
caretakingTick := time.Tick(time.Minute)
recoveryTick := time.Tick(time.Duration(config.Config.RecoveryPollSeconds) * time.Second)
var snapshotTopologiesTick <-chan time.Time
if config.Config.SnapshotTopologiesIntervalHours > 0 {
snapshotTopologiesTick = time.Tick(time.Duration(config.Config.SnapshotTopologiesIntervalHours) * time.Hour)
}
go ometrics.InitGraphiteMetrics()
go acceptSignals()
if *config.RuntimeCLIFlags.GrabElection {
process.GrabElection()
}
for {
select {
case <-discoveryTick:
go func() {
onDiscoveryTick()
}()
case <-instancePollTick:
go func() {
// This tick does NOT do instance poll (these are handled by the oversmapling discoveryTick)
// But rather should invoke such routinely operations that need to be as (or roughly as) frequent
// as instance poll
if atomic.LoadInt64(&isElectedNode) == 1 {
go inst.UpdateInstanceRecentRelaylogHistory()
go inst.RecordInstanceCoordinatesHistory()
}
}()
case <-caretakingTick:
// Various periodic internal maintenance tasks
go func() {
if atomic.LoadInt64(&isElectedNode) == 1 {
go inst.RecordInstanceBinlogFileHistory()
go inst.ForgetLongUnseenInstances()
go inst.ForgetUnseenInstancesDifferentlyResolved()
go inst.ForgetExpiredHostnameResolves()
go inst.DeleteInvalidHostnameResolves()
go inst.ReviewUnseenInstances()
go inst.InjectUnseenMasters()
go inst.ResolveUnknownMasterHostnameResolves()
go inst.UpdateClusterAliases()
go inst.ExpireMaintenance()
go inst.ExpireDowntime()
go inst.ExpireCandidateInstances()
go inst.ExpireHostnameUnresolve()
go inst.ExpireClusterDomainName()
go inst.ExpireAudit()
go inst.ExpireMasterPositionEquivalence()
go inst.ExpirePoolInstances()
go inst.FlushNontrivialResolveCacheToDatabase()
go process.ExpireNodesHistory()
go process.ExpireAccessTokens()
} else {
// Take this opportunity to refresh yourself
go inst.LoadHostnameResolveCache()
}
}()
case <-recoveryTick:
go func() {
if atomic.LoadInt64(&isElectedNode) == 1 {
go ClearActiveFailureDetections()
go ClearActiveRecoveries()
go ExpireBlockedRecoveries()
go AcknowledgeCrashedRecoveries()
go inst.ExpireInstanceAnalysisChangelog()
go CheckAndRecover(nil, nil, false)
}
}()
case <-snapshotTopologiesTick:
go func() {
go inst.SnapshotTopologies()
}()
}
}
}
func pollAgent(hostname string) error {
polledAgent, err := agent.GetAgent(hostname)
agent.UpdateAgentLastChecked(hostname)
if err != nil {
return log.Errore(err)
}
err = agent.UpdateAgentInfo(hostname, polledAgent)
if err != nil {
return log.Errore(err)
}
return nil
}
// ContinuousAgentsPoll starts an asynchronuous infinite process where agents are
// periodically investigated and their status captured, and long since unseen agents are
// purged and forgotten.
func ContinuousAgentsPoll() {
log.Infof("Starting continuous agents poll")
go discoverSeededAgents()
tick := time.Tick(time.Duration(config.Config.GetDiscoveryPollSeconds()) * time.Second)
caretakingTick := time.Tick(time.Hour)
for range tick {
agentsHosts, _ := agent.ReadOutdatedAgentsHosts()
log.Debugf("outdated agents hosts: %+v", agentsHosts)
for _, hostname := range agentsHosts {
go pollAgent(hostname)
}
// See if we should also forget agents (lower frequency)
select {
case <-caretakingTick:
agent.ForgetLongUnseenAgents()
agent.FailStaleSeeds()
default:
}
}
}
func discoverSeededAgents() {
for seededAgent := range agent.SeededAgents {
instanceKey := &inst.InstanceKey{Hostname: seededAgent.Hostname, Port: int(seededAgent.MySQLPort)}
go inst.ReadTopologyInstance(instanceKey)
}
}