-
Notifications
You must be signed in to change notification settings - Fork 0
/
monitor.go
559 lines (457 loc) · 20.5 KB
/
monitor.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
package master
import (
"database/sql"
"fmt"
"github.com/KIT-MAMID/mamid/model"
"github.com/KIT-MAMID/mamid/msp"
"github.com/Sirupsen/logrus"
"github.com/jinzhu/gorm"
"sync"
"time"
)
var monitorLog = logrus.WithField("module", "monitor")
type Monitor struct {
DB *model.DB
BusWriteChannel chan<- interface{}
MSPClient msp.MSPClient
Interval time.Duration
}
func (m *Monitor) Run() {
ticker := time.NewTicker(m.Interval)
quit := make(chan struct{})
go func() {
for {
select {
case <-ticker.C:
monitorLog.Info("Monitor running")
//Get all slaves from database
tx := m.DB.Begin()
var slaves []model.Slave
err := tx.Find(&slaves).Error
if err != nil {
monitorLog.WithError(err).Error("Could not get slaves")
}
tx.Rollback()
wg := sync.WaitGroup{}
type observation struct {
result []msp.Mongod
err *msp.Error
theSlave model.Slave
}
observationChan := make(chan observation)
//Observe active slaves
for _, slave := range slaves {
wg.Add(1)
go func(s model.Slave) {
//Request mongod states from slave
observedMongods, mspError := m.MSPClient.RequestStatus(msp.HostPort{s.Hostname, msp.PortNumber(s.Port)})
observationChan <- observation{
result: observedMongods,
err: mspError,
theSlave: s, //Do not call this slave or vet will fail
}
wg.Done()
}(slave)
}
//Wait for all slaves to be observed and close channel to make consumer loop break
go func() {
wg.Wait()
close(observationChan)
}()
//Consumer loop that saves result to database
//We do this so that all transactions happen after eachother == prevent concurrent database access
for observationRes := range observationChan {
m.handleObservation(observationRes.result, observationRes.err, observationRes.theSlave)
}
//Check degradation of replica sets
m.observeReplicaSets()
case <-quit:
ticker.Stop()
return
}
}
}()
}
func mongodTuple(s model.Slave, m msp.Mongod) string {
return fmt.Sprintf("(%s(id=%d),%d,%s)", s.Hostname, s.ID, m.Port, m.ReplicaSetConfig.ReplicaSetName)
}
func (m *Monitor) handleObservation(observedMongods []msp.Mongod, mspError *msp.Error, slave model.Slave) {
// Notify about reachablility
comErr := msp.Error{}
if mspError != nil {
//TODO Handle other slave errors => check identifiers != CommunicationError
monitorLog.Errorf("monitor: error observing slave: %#v", mspError)
comErr = *mspError
}
m.BusWriteChannel <- model.ConnectionStatus{
Slave: slave,
Unreachable: mspError != nil && mspError.Identifier == msp.CommunicationError,
CommunicationError: comErr,
}
tx := m.DB.Begin()
if err := m.updateSlaveObservationError(tx, slave, mspError); err != nil {
monitorLog.WithError(err).Error("error persisting slave observation error")
tx.Rollback()
return
}
// Return early if there were observation errors, since Mongods' ObservedState cannot be updated
if mspError != nil {
if err := tx.Commit().Error; err != nil {
monitorLog.WithError(err).Error("could not commit monitor run")
}
return
}
// NOTE: we assume observedMonogds to be valid from this point on (caught by early return in case of slave observation error)
if err := m.updateMongodStatesNotObservedOnSlave(tx, slave, observedMongods); err != nil {
monitorLog.Errorf("error deleting MongodStates that are present in the database but not reported by slave `%s`: %s", slave.Hostname, err)
tx.Rollback()
return
}
modelToObservedMap, err := m.updateOrCreateObservedMongodStates(tx, slave, observedMongods)
if err != nil {
monitorLog.Errorf("error updating or creating ObservedState of Mongods reported by slave `%s`: %s", slave.Hostname, err)
tx.Rollback()
return
}
if err := tx.Commit().Error; err != nil {
monitorLog.WithError(err).Error("Could not commit monitor run")
}
// Read-only transaction
tx = m.DB.Begin()
defer tx.Rollback()
if err := m.sendMongodMismatchStatusToBus(tx, slave, modelToObservedMap); err != nil {
monitorLog.WithError(err).Error()
}
}
func (m *Monitor) updateSlaveObservationError(tx *gorm.DB, slave model.Slave, slaveObservationError *msp.Error) (criticalError error) {
if slaveObservationError != nil { // update observation error field
monitorLog.Debugf("monitor: persisting observation error for slave `%s:%d` in database", slave.Hostname, slave.Port)
var updateError error
modelObservationErr := mspErrorToModelMSPError(slaveObservationError)
if slave.ObservationErrorID.Valid {
// Replace existing entry
monitorLog.Debugf("monitor: replacing existing observation error for slave `%s:%d` in database", slave.Hostname, slave.Port)
modelObservationErr.ID = slave.ObservationErrorID.Int64
updateError = tx.Save(&modelObservationErr).Error
} else {
monitorLog.Debugf("monitor: creating observation error for slave `%s:%d` in database", slave.Hostname, slave.Port)
updateError = tx.Create(&modelObservationErr).Error
if updateError == nil {
updateError = tx.Model(&slave).Update("ObservationErrorID", modelObservationErr.ID).Error
}
}
if updateError != nil {
return fmt.Errorf("monitor: database error when updating slave `%s:%d` ObservationErrorID field: %s", slave.Hostname, slave.Port, updateError)
}
// return early as there should not be observedMongods in case of slaveObservationError
return nil
} else if slave.ObservationErrorID.Valid { // clear observation error field
monitorLog.Debugf("monitor: clearing observation error for slave `%s:%d` in database", slave.Hostname, slave.Port)
res := tx.Exec(`DELETE FROM msp_errors WHERE id=?`, slave.ObservationErrorID.Int64)
switch {
case res.Error != nil:
return fmt.Errorf("monitor: database error when clearing observation error for slave `%s:%d`: %s", slave.Hostname, slave.Port, res.Error)
case res.RowsAffected == 0:
return fmt.Errorf("monitor: clearing observation error for slave `%s:%d` affected 0 rows", slave.Hostname, slave.Port)
case res.RowsAffected == 1:
monitorLog.Debugf("monitor: cleared observation error for slave `%s:%d`", slave.Hostname, slave.Port)
case res.RowsAffected > 1:
monitorLog.Warnf("monitor: clearing observation error for slave `%s:%d` affected %d != 1 rows", slave.Hostname, slave.Port, res.RowsAffected)
}
}
return nil
}
// Update the monitoring information about Mongods kept in the database
// Returns:
// modelObservationMap map from model.Mongod.ID => observed msp.Mongod from observedMongods
func (m *Monitor) updateOrCreateObservedMongodStates(tx *gorm.DB, slave model.Slave, observedMongods []msp.Mongod) (modelToObservedMap map[int64]msp.Mongod, err error) {
modelToObservedMap = make(map[int64]msp.Mongod)
for _, observedMongod := range observedMongods {
monitorLog.Debugf("monitor: updating observed state for mongod `%s` in database`", mongodTuple(slave, observedMongod))
var dbMongod model.Mongod
dbMongodRes := tx.First(&dbMongod, &model.Mongod{
ParentSlaveID: slave.ID,
Port: model.PortNumber(observedMongod.Port),
ReplSetName: observedMongod.ReplicaSetConfig.ReplicaSetName,
})
if dbMongodRes.Error != nil && !dbMongodRes.RecordNotFound() {
// Early exit
return modelToObservedMap, fmt.Errorf("monitor: database error when querying for Mongod corresponding to observed Mongod `%s`: %s",
mongodTuple(slave, observedMongod), dbMongodRes.Error)
} else if dbMongodRes.RecordNotFound() {
// The slave is running a Mongod which is not in the database.
// This means MAMID does not expect it to be there.
// => model this situation in the database by settings desiredState = destroyed
// => then continue as if the Mongod had always been in the database
dbMongod = model.Mongod{
ParentSlaveID: slave.ID,
Port: model.PortNumber(observedMongod.Port),
ReplSetName: observedMongod.ReplicaSetConfig.ReplicaSetName,
ReplicaSetID: sql.NullInt64{Valid: false},
}
if err := tx.Create(&dbMongod).Error; err != nil {
return modelToObservedMap, fmt.Errorf("monitor: could not create database representation for unknown observed Mongod `%s`: %s",
mongodTuple(slave, observedMongod), err)
}
shardingRole, err := ProjectMSPShardingRoleToModelShardingRole(observedMongod.ReplicaSetConfig.ShardingRole)
if err != nil {
shardingRole = model.ShardingRoleNone // We don't really care about the sharding role as we want do destroy the mongod anyway
}
desiredState := model.MongodState{
ParentMongodID: dbMongod.ID,
ExecutionState: model.MongodExecutionStateForceDestroyed,
ShardingRole: shardingRole,
}
if err := tx.Create(&desiredState).Error; err != nil {
return modelToObservedMap, fmt.Errorf("monitor: could not create desired MongodState for unknown observed Mongod `%s`: %s",
mongodTuple(slave, observedMongod), err)
}
if err := tx.Model(&dbMongod).Update("DesiredStateID", desiredState.ID).Error; err != nil {
return modelToObservedMap, fmt.Errorf("monitor: could not update DesiredStateID column for unknown observed Mongod `%s`: %s",
mongodTuple(slave, observedMongod), err)
}
}
modelToObservedMap[dbMongod.ID] = observedMongod
//Get desired state if it exists
relatedResult := tx.Model(&dbMongod).Related(&dbMongod.DesiredState, "DesiredState")
if !relatedResult.RecordNotFound() && relatedResult.Error != nil {
return modelToObservedMap, fmt.Errorf("monitor: internal inconsistency: could not get desired state for Mongod `%s`: %s",
mongodTuple(slave, observedMongod), relatedResult.Error.Error())
}
//Get observed state if it exists
relatedResult = tx.Model(&dbMongod).Related(&dbMongod.ObservedState, "ObservedState")
if !relatedResult.RecordNotFound() && relatedResult.Error != nil {
return modelToObservedMap, fmt.Errorf("monitor: database error when querying for observed state of Mongod `%s`: %s",
mongodTuple(slave, observedMongod), relatedResult.Error)
}
// Update database representation of observation
if observedMongod.StatusError == nil {
//TODO Finish this
//Put observations into model
dbMongod.ObservedState.ParentMongodID = dbMongod.ID // we could be creating the ObservedState of Mongod on first observation
if err := m.updateObservedState(tx, observedMongod, &dbMongod.ObservedState); err != nil {
return modelToObservedMap, fmt.Errorf("error updating observed state of Mongod `%s`: %s", mongodTuple(slave, observedMongod), err)
}
dbMongod.ObservationError = model.MSPError{}
} else {
dbMongod.ObservationError = mspErrorToModelMSPError(observedMongod.StatusError)
}
// Persist updated database representation
//TODO Only update observed state and errors to prevent collisions with cluster allocator
saveErr := tx.Save(&dbMongod).Error
if saveErr != nil {
return modelToObservedMap, fmt.Errorf("monitor: error persisting updated observed state for mongod `%s`: %s",
mongodTuple(slave, observedMongod), saveErr.Error())
}
monitorLog.Debugf("monitor: finished updating observed state for mongod `%s` in database`", mongodTuple(slave, observedMongod))
}
return modelToObservedMap, nil
}
// Update the ObservedState of a Mongod, including ReplicaSetMembers
func (m *Monitor) updateObservedState(tx *gorm.DB, observedMongod msp.Mongod, observedState *model.MongodState) (err error) {
observedState.ExecutionState = MspMongodStateToModelExecutionState(observedMongod.State)
observedState.ShardingRole, err = ProjectMSPShardingRoleToModelShardingRole(observedMongod.ReplicaSetConfig.ShardingRole)
if err != nil {
observedState.ShardingRole = model.ShardingRoleNone //TODO Make nullable and remove this but does not really matter
}
return nil
}
// Remove observed state of mongods the slave does not report
// Errors returned by this method should be handled by aborting the transaction tx
func (m *Monitor) updateMongodStatesNotObservedOnSlave(tx *gorm.DB, slave model.Slave, observedMongods []msp.Mongod) (err error) {
monitorLog.Debugf("monitor: handling unobserved Mongods of slave `%s`", slave.Hostname)
var modelMongods []model.Mongod
if err := tx.Model(&slave).Related(&modelMongods, "Mongods").Error; err != nil {
return err
}
outer:
for _, modelMongod := range modelMongods {
//Check if slave reported this mongod
for _, observedMongod := range observedMongods {
if modelMongod.Port == model.PortNumber(observedMongod.Port) &&
modelMongod.ReplSetName == observedMongod.ReplicaSetConfig.ReplicaSetName {
continue outer
}
}
// We didn't find the expected modelMongod on the slave
// => delete observation
monitorLog.Infof("setting observed state of Mongod `%s:%d` to destroyed as it was not reported by slave `%s`", slave.Hostname, modelMongod.Port, slave.Hostname)
if modelMongod.ObservedStateID.Valid {
tx.Model(&model.MongodState{ID: modelMongod.ObservedStateID.Int64}).Update("ExecutionState", model.MongodExecutionStateDestroyed)
} else {
observedState := model.MongodState{
ParentMongodID: modelMongod.ID,
ExecutionState: model.MongodExecutionStateDestroyed,
ShardingRole: model.ShardingRoleNone, //Does not matter but we have to set one
}
if err := tx.Create(&observedState).Error; err != nil {
return fmt.Errorf("monitor: could not create observed MongodState for observed destroyed Mongod `%s`: %s", modelMongod.ID, err)
}
if err := tx.Model(&modelMongod).Update("ObservedStateID", observedState.ID).Error; err != nil {
return fmt.Errorf("monitor: could not create observed MongodState backreference for observed destroyed Mongod `%s`: %s", modelMongod.ID, err)
}
}
if modelMongod.ObservationErrorID.Valid {
deleteErr := tx.Model(&modelMongod).Update("ObservationErrorID", sql.NullInt64{Valid: false}).Error
if deleteErr != nil {
monitorLog.Errorf("error removing observation error of Mongod `%s:%d`: %s", slave.Hostname, modelMongod.Port, deleteErr)
return deleteErr
}
}
}
monitorLog.Debugf("monitor: finished handling unobserved Mongods of slave `%s`", slave.Hostname)
return nil
}
// Check every Mongod of the Slave for mismatches between DesiredState and ObservedState
// and send an appropriate MongodMismatchStatus to the Bus
// The desired/observed msp.ReplicaSetMembers are computed ad-hoc.
func (m *Monitor) sendMongodMismatchStatusToBus(tx *gorm.DB, slave model.Slave, modelToObservedMap map[int64]msp.Mongod) (err error) {
monitorLog.Debugf("monitor: preparing Mongod Mismatch Status messages for slave `%s`", slave.Hostname)
var modelMongods []model.Mongod
if err := tx.Model(&slave).Related(&modelMongods, "Mongods").Error; err != nil {
return err
}
for _, modelMongod := range modelMongods {
var busMessage interface{}
busMessage, err = m.compareStates(tx, modelMongod, modelToObservedMap[modelMongod.ID])
if err != nil {
monitorLog.Errorf("error comparing Mongod `%d on %s` Desired and Observed state: %s", modelMongod.ID, slave.Hostname, err)
continue
}
monitorLog.Debugf("monitor: sending bus message for slave `%s`", slave.Hostname, busMessage)
m.BusWriteChannel <- busMessage
monitorLog.Debugf("monitor: sent bus message for slave `%s`", slave.Hostname, busMessage)
}
monitorLog.Debugf("monitor: finished sending Mongod Mismatch Status messages for slave `%s`", slave.Hostname)
return nil
}
func (m *Monitor) compareStates(tx *gorm.DB, mongod model.Mongod, observedMongod msp.Mongod) (s model.MongodMatchStatus, err error) {
if !mongod.ObservedStateID.Valid {
// If we have no observations, it is a mismatch (since we can't know what the actual state is)
// Example: a new Mongod that has never been observed. Will be deployed by the Deployer when informed about Mismatch
return model.MongodMatchStatus{
Mismatch: true,
Mongod: mongod,
}, nil
}
if err = tx.Model(mongod).Related(&mongod.DesiredState, "DesiredState").Error; err != nil {
return s, fmt.Errorf("error fetching DesiredState for mongod `%v`: %s", mongod, err)
}
observedStateRes := tx.Model(mongod).Related(&mongod.ObservedState, "ObservedState")
if observedStateRes.Error != nil {
return s, fmt.Errorf("error fetching ObservedState for mongod `%v`: %s", mongod, observedStateRes.Error)
}
mongodStatesEquivalent := MongodStatesEquivalent(mongod.DesiredState, mongod.ObservedState)
var replicaSetMembersEquivalent = false
if mongod.ReplicaSetID.Valid {
desiredMembers, _, err := DesiredMSPReplicaSetMembersForReplicaSetID(tx, mongod.ReplicaSetID.Int64)
if err != nil {
return s, fmt.Errorf("error computing ReplicaSetMembers: %s", err)
}
replicaSetMembersEquivalent = MSPReplicaSetMembersDeepEqualsIgnoringOrder(desiredMembers, observedMongod.ReplicaSetConfig.ReplicaSetMembers)
}
return model.MongodMatchStatus{
Mismatch: !(mongodStatesEquivalent && replicaSetMembersEquivalent),
Mongod: mongod,
}, nil
}
// check if MongodStates are equivalent with regards to monitored attributes
func MongodStatesEquivalent(a, b model.MongodState) (e bool) {
e = a.ExecutionState == b.ExecutionState
e = e && a.ShardingRole == b.ShardingRole
return e
}
// Deep-equals of two _sets_ (=> lists without order) of msp.ReplicaSetMembers
func MSPReplicaSetMembersDeepEqualsIgnoringOrder(a, b []msp.ReplicaSetMember) (equal bool) {
if len(a) != len(b) {
return false
}
seen := make(map[msp.ReplicaSetMember]interface{})
for _, m := range a {
seen[m] = struct{}{}
}
for _, m := range b {
_, exists := seen[m]
if !exists {
return false
}
}
return true
}
func ReplicaSetMembersEquivalent(a, b msp.ReplicaSetMember) bool {
return a.HostPort == b.HostPort && a.Priority == b.Priority && a.Votes == b.Votes
}
func MspMongodStateToModelExecutionState(e msp.MongodState) model.MongodExecutionState {
switch e {
case msp.MongodStateDestroyed:
return model.MongodExecutionStateDestroyed
case msp.MongodStateNotRunning:
return model.MongodExecutionStateNotRunning
case msp.MongodStateRecovering:
return model.MongodExecutionStateRecovering
case msp.MongodStateRunning:
return model.MongodExecutionStateRunning
case msp.MongodStateForceDestroyed:
return model.MongodExecutionStateForceDestroyed
case msp.MongodStateUninitialized:
return model.MongodExecutionStateUninitiated
default:
return 0 // Invalid
//TODO New states
}
}
func mspErrorToModelMSPError(mspError *msp.Error) model.MSPError {
return model.MSPError{
Identifier: mspError.Identifier,
Description: mspError.Description,
LongDescription: mspError.LongDescription,
}
}
func (m *Monitor) observeReplicaSets() {
tx := m.DB.Begin()
defer tx.Rollback()
// Get replica sets and the count of their actually configured members from the database
replicaSetsWithMemberCounts, err := tx.Raw(`SELECT
r.*,
(SELECT COUNT(*) FROM replica_set_configured_members WHERE replica_set_id = r.id AND persistent_storage = ?)
AS configured_persistent_members,
(SELECT COUNT(*) FROM replica_set_configured_members WHERE replica_set_id = r.id AND persistent_storage = ?)
AS configured_volatile_members,
(SELECT COUNT(*) FROM replica_set_effective_members_for_monitoring WHERE replica_set_id = r.id AND persistent_storage = ?)
AS actual_persistent_members,
(SELECT COUNT(*) FROM replica_set_effective_members_for_monitoring WHERE replica_set_id = r.id AND persistent_storage = ?)
AS actual_volatile_members
FROM replica_sets r
`, true, false, true, false).Rows()
if err != nil {
monitorLog.WithError(err).Error("Error getting configured and actual member counts of replica sets")
return
}
defer replicaSetsWithMemberCounts.Close()
for replicaSetsWithMemberCounts.Next() {
var replicaSet model.ReplicaSet
tx.ScanRows(replicaSetsWithMemberCounts, &replicaSet)
memberCounts := struct {
ConfiguredPersistentMembers uint
ConfiguredVolatileMembers uint
ActualPersistentMembers uint
ActualVolatileMembers uint
}{}
tx.ScanRows(replicaSetsWithMemberCounts, &memberCounts)
unsatisfied := memberCounts.ConfiguredVolatileMembers > memberCounts.ActualVolatileMembers ||
memberCounts.ConfiguredPersistentMembers > memberCounts.ActualPersistentMembers
m.BusWriteChannel <- model.ObservedReplicaSetConstraintStatus{
Unsatisfied: unsatisfied,
ReplicaSet: replicaSet,
ConfiguredPersistentCount: memberCounts.ConfiguredPersistentMembers,
ConfiguredVolatileCount: memberCounts.ConfiguredVolatileMembers,
ActualPersistentCount: memberCounts.ActualPersistentMembers,
ActualVolatileCount: memberCounts.ActualVolatileMembers,
}
m.BusWriteChannel <- model.ReplicaSetInitiationStatus{
Initiated: replicaSet.Initiated,
ReplicaSet: replicaSet,
}
}
replicaSetsWithMemberCounts.Close()
}