-
Notifications
You must be signed in to change notification settings - Fork 1.9k
/
task_runner.go
1955 lines (1672 loc) · 60.3 KB
/
task_runner.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
package client
import (
"bytes"
"crypto/md5"
"encoding/hex"
"fmt"
"io"
"io/ioutil"
"log"
"os"
"path/filepath"
"strings"
"sync"
"time"
metrics "github.com/armon/go-metrics"
"github.com/boltdb/bolt"
"github.com/golang/snappy"
"github.com/hashicorp/consul-template/signals"
"github.com/hashicorp/go-multierror"
version "github.com/hashicorp/go-version"
"github.com/hashicorp/nomad/client/allocdir"
"github.com/hashicorp/nomad/client/config"
"github.com/hashicorp/nomad/client/driver"
"github.com/hashicorp/nomad/client/getter"
"github.com/hashicorp/nomad/client/vaultclient"
"github.com/hashicorp/nomad/command/agent/consul"
"github.com/hashicorp/nomad/nomad/structs"
"github.com/ugorji/go/codec"
"github.com/hashicorp/nomad/client/driver/env"
dstructs "github.com/hashicorp/nomad/client/driver/structs"
cstructs "github.com/hashicorp/nomad/client/structs"
)
const (
// killBackoffBaseline is the baseline time for exponential backoff while
// killing a task.
killBackoffBaseline = 5 * time.Second
// killBackoffLimit is the limit of the exponential backoff for killing
// the task.
killBackoffLimit = 2 * time.Minute
// killFailureLimit is how many times we will attempt to kill a task before
// giving up and potentially leaking resources.
killFailureLimit = 5
// vaultBackoffBaseline is the baseline time for exponential backoff when
// attempting to retrieve a Vault token
vaultBackoffBaseline = 5 * time.Second
// vaultBackoffLimit is the limit of the exponential backoff when attempting
// to retrieve a Vault token
vaultBackoffLimit = 3 * time.Minute
// vaultTokenFile is the name of the file holding the Vault token inside the
// task's secret directory
vaultTokenFile = "vault_token"
)
var (
// taskRunnerStateAllKey holds all the task runners state. At the moment
// there is no need to split it
taskRunnerStateAllKey = []byte("simple-all")
)
// taskRestartEvent wraps a TaskEvent with additional metadata to control
// restart behavior.
type taskRestartEvent struct {
// taskEvent to report
taskEvent *structs.TaskEvent
// if false, don't count against restart count
failure bool
}
func newTaskRestartEvent(reason string, failure bool) *taskRestartEvent {
return &taskRestartEvent{
taskEvent: structs.NewTaskEvent(structs.TaskRestartSignal).SetRestartReason(reason),
failure: failure,
}
}
// TaskRunner is used to wrap a task within an allocation and provide the execution context.
type TaskRunner struct {
stateDB *bolt.DB
config *config.Config
updater TaskStateUpdater
logger *log.Logger
restartTracker *RestartTracker
consul ConsulServiceAPI
// running marks whether the task is running
running bool
runningLock sync.Mutex
resourceUsage *cstructs.TaskResourceUsage
resourceUsageLock sync.RWMutex
alloc *structs.Allocation
task *structs.Task
taskDir *allocdir.TaskDir
// envBuilder is used to build the task's environment
envBuilder *env.Builder
// driverNet is the network information returned by the driver
driverNet *cstructs.DriverNetwork
driverNetLock sync.Mutex
// updateCh is used to receive updated versions of the allocation
updateCh chan *structs.Allocation
handle driver.DriverHandle
handleLock sync.Mutex
// artifactsDownloaded tracks whether the tasks artifacts have been
// downloaded
//
// Must acquire persistLock when accessing
artifactsDownloaded bool
// taskDirBuilt tracks whether the task has built its directory.
//
// Must acquire persistLock when accessing
taskDirBuilt bool
// createdResources are all the resources created by the task driver
// across all attempts to start the task.
// Simple gets and sets should use {get,set}CreatedResources
createdResources *driver.CreatedResources
createdResourcesLock sync.Mutex
// payloadRendered tracks whether the payload has been rendered to disk
payloadRendered bool
// vaultFuture is the means to wait for and get a Vault token
vaultFuture *tokenFuture
// recoveredVaultToken is the token that was recovered through a restore
recoveredVaultToken string
// vaultClient is used to retrieve and renew any needed Vault token
vaultClient vaultclient.VaultClient
// templateManager is used to manage any consul-templates this task may have
templateManager *TaskTemplateManager
// startCh is used to trigger the start of the task
startCh chan struct{}
// unblockCh is used to unblock the starting of the task
unblockCh chan struct{}
unblocked bool
unblockLock sync.Mutex
// restartCh is used to restart a task
restartCh chan *taskRestartEvent
// signalCh is used to send a signal to a task
signalCh chan SignalEvent
destroy bool
destroyCh chan struct{}
destroyLock sync.Mutex
destroyEvent *structs.TaskEvent
// waitCh closing marks the run loop as having exited
waitCh chan struct{}
// persistLock must be acquired when accessing fields stored by
// SaveState. SaveState is called asynchronously to TaskRunner.Run by
// AllocRunner, so all state fields must be synchronized using this
// lock.
persistLock sync.Mutex
// persistedHash is the hash of the last persisted snapshot. It is used to
// detect if a new snapshot has to be written to disk.
persistedHash []byte
// baseLabels are used when emitting tagged metrics. All task runner metrics
// will have these tags, and optionally more.
baseLabels []metrics.Label
}
// taskRunnerState is used to snapshot the state of the task runner
type taskRunnerState struct {
Version string
HandleID string
ArtifactDownloaded bool
TaskDirBuilt bool
PayloadRendered bool
CreatedResources *driver.CreatedResources
DriverNetwork *cstructs.DriverNetwork
}
func (s *taskRunnerState) Hash() []byte {
h := md5.New()
io.WriteString(h, s.Version)
io.WriteString(h, s.HandleID)
io.WriteString(h, fmt.Sprintf("%v", s.ArtifactDownloaded))
io.WriteString(h, fmt.Sprintf("%v", s.TaskDirBuilt))
io.WriteString(h, fmt.Sprintf("%v", s.PayloadRendered))
h.Write(s.CreatedResources.Hash())
h.Write(s.DriverNetwork.Hash())
return h.Sum(nil)
}
// TaskStateUpdater is used to signal that tasks state has changed. If lazySync
// is set the event won't be immediately pushed to the server.
type TaskStateUpdater func(taskName, state string, event *structs.TaskEvent, lazySync bool)
// SignalEvent is a tuple of the signal and the event generating it
type SignalEvent struct {
// s is the signal to be sent
s os.Signal
// e is the task event generating the signal
e *structs.TaskEvent
// result should be used to send back the result of the signal
result chan<- error
}
// NewTaskRunner is used to create a new task context
func NewTaskRunner(logger *log.Logger, config *config.Config,
stateDB *bolt.DB, updater TaskStateUpdater, taskDir *allocdir.TaskDir,
alloc *structs.Allocation, task *structs.Task,
vaultClient vaultclient.VaultClient, consulClient ConsulServiceAPI) *TaskRunner {
// Merge in the task resources
task.Resources = alloc.TaskResources[task.Name]
// Build the restart tracker.
tg := alloc.Job.LookupTaskGroup(alloc.TaskGroup)
if tg == nil {
logger.Printf("[ERR] client: alloc %q for missing task group %q", alloc.ID, alloc.TaskGroup)
return nil
}
restartTracker := newRestartTracker(tg.RestartPolicy, alloc.Job.Type)
// Initialize the environment builder
envBuilder := env.NewBuilder(config.Node, alloc, task, config.Region)
tc := &TaskRunner{
config: config,
stateDB: stateDB,
updater: updater,
logger: logger,
restartTracker: restartTracker,
alloc: alloc,
task: task,
taskDir: taskDir,
envBuilder: envBuilder,
createdResources: driver.NewCreatedResources(),
consul: consulClient,
vaultClient: vaultClient,
vaultFuture: NewTokenFuture().Set(""),
updateCh: make(chan *structs.Allocation, 64),
destroyCh: make(chan struct{}),
waitCh: make(chan struct{}),
startCh: make(chan struct{}, 1),
unblockCh: make(chan struct{}),
restartCh: make(chan *taskRestartEvent),
signalCh: make(chan SignalEvent),
}
tc.baseLabels = []metrics.Label{
{
Name: "job",
Value: tc.alloc.Job.Name,
},
{
Name: "task_group",
Value: tc.alloc.TaskGroup,
},
{
Name: "alloc_id",
Value: tc.alloc.ID,
},
{
Name: "task",
Value: tc.task.Name,
},
}
return tc
}
// MarkReceived marks the task as received.
func (r *TaskRunner) MarkReceived() {
// We lazy sync this since there will be a follow up message almost
// immediately.
r.updater(r.task.Name, structs.TaskStatePending, structs.NewTaskEvent(structs.TaskReceived), true)
}
// WaitCh returns a channel to wait for termination
func (r *TaskRunner) WaitCh() <-chan struct{} {
return r.waitCh
}
// getHandle returns the task's handle or nil
func (r *TaskRunner) getHandle() driver.DriverHandle {
r.handleLock.Lock()
h := r.handle
r.handleLock.Unlock()
return h
}
// pre060StateFilePath returns the path to our state file that would have been
// written pre v0.6.0
// COMPAT: Remove in 0.7.0
func (r *TaskRunner) pre060StateFilePath() string {
// Get the MD5 of the task name
hashVal := md5.Sum([]byte(r.task.Name))
hashHex := hex.EncodeToString(hashVal[:])
dirName := fmt.Sprintf("task-%s", hashHex)
// Generate the path
return filepath.Join(r.config.StateDir, "alloc", r.alloc.ID, dirName, "state.json")
}
// RestoreState is used to restore our state. If a non-empty string is returned
// the task is restarted with the string as the reason. This is useful for
// backwards incompatible upgrades that need to restart tasks with a new
// executor.
func (r *TaskRunner) RestoreState() (string, error) {
// COMPAT: Remove in 0.7.0
// 0.6.0 transitioned from individual state files to a single bolt-db.
// The upgrade path is to:
// Check if old state exists
// If so, restore from that and delete old state
// Restore using state database
var snap taskRunnerState
// Check if the old snapshot is there
oldPath := r.pre060StateFilePath()
if err := pre060RestoreState(oldPath, &snap); err == nil {
// Delete the old state
os.RemoveAll(oldPath)
} else if !os.IsNotExist(err) {
// Something corrupt in the old state file
return "", err
} else {
// We are doing a normal restore
err := r.stateDB.View(func(tx *bolt.Tx) error {
bkt, err := getTaskBucket(tx, r.alloc.ID, r.task.Name)
if err != nil {
return fmt.Errorf("failed to get task bucket: %v", err)
}
if err := getObject(bkt, taskRunnerStateAllKey, &snap); err != nil {
return fmt.Errorf("failed to read task runner state: %v", err)
}
return nil
})
if err != nil {
return "", err
}
}
// Restore fields from the snapshot
r.artifactsDownloaded = snap.ArtifactDownloaded
r.taskDirBuilt = snap.TaskDirBuilt
r.payloadRendered = snap.PayloadRendered
r.setCreatedResources(snap.CreatedResources)
r.driverNet = snap.DriverNetwork
if r.task.Vault != nil {
// Read the token from the secret directory
tokenPath := filepath.Join(r.taskDir.SecretsDir, vaultTokenFile)
data, err := ioutil.ReadFile(tokenPath)
if err != nil {
if !os.IsNotExist(err) {
return "", fmt.Errorf("failed to read token for task %q in alloc %q: %v", r.task.Name, r.alloc.ID, err)
}
// Token file doesn't exist
} else {
// Store the recovered token
r.recoveredVaultToken = string(data)
}
}
// Restore the driver
restartReason := ""
if snap.HandleID != "" {
d, err := r.createDriver()
if err != nil {
return "", err
}
// Add the restored network driver to the environment
r.envBuilder.SetDriverNetwork(r.driverNet)
// Open a connection to the driver handle
ctx := driver.NewExecContext(r.taskDir, r.envBuilder.Build())
handle, err := d.Open(ctx, snap.HandleID)
// In the case it fails, we relaunch the task in the Run() method.
if err != nil {
r.logger.Printf("[ERR] client: failed to open handle to task %q for alloc %q: %v",
r.task.Name, r.alloc.ID, err)
return "", nil
}
if pre06ScriptCheck(snap.Version, r.task.Driver, r.task.Services) {
restartReason = pre06ScriptCheckReason
}
if err := r.registerServices(d, handle, r.driverNet); err != nil {
// Don't hard fail here as there's a chance this task
// registered with Consul properly when it initial
// started.
r.logger.Printf("[WARN] client: failed to register services and checks with consul for task %q in alloc %q: %v",
r.task.Name, r.alloc.ID, err)
}
r.handleLock.Lock()
r.handle = handle
r.handleLock.Unlock()
r.runningLock.Lock()
r.running = true
r.runningLock.Unlock()
}
return restartReason, nil
}
// ver06 is used for checking for pre-0.6 script checks
var ver06 = version.Must(version.NewVersion("0.6.0dev"))
// pre06ScriptCheckReason is the restart reason given when a pre-0.6 script
// check is found on an exec/java task.
const pre06ScriptCheckReason = "upgrading pre-0.6 script checks"
// pre06ScriptCheck returns true if version is prior to 0.6.0dev, has a script
// check, and uses exec or java drivers.
func pre06ScriptCheck(ver, driver string, services []*structs.Service) bool {
if driver != "exec" && driver != "java" && driver != "mock_driver" {
// Only exec and java are affected
return false
}
v, err := version.NewVersion(ver)
if err != nil {
// Treat it as old
return true
}
if !v.LessThan(ver06) {
// >= 0.6.0dev
return false
}
for _, service := range services {
for _, check := range service.Checks {
if check.Type == "script" {
return true
}
}
}
return false
}
// SaveState is used to snapshot our state
func (r *TaskRunner) SaveState() error {
r.destroyLock.Lock()
defer r.destroyLock.Unlock()
if r.destroy {
// Don't save state if already destroyed
return nil
}
r.persistLock.Lock()
defer r.persistLock.Unlock()
snap := taskRunnerState{
Version: r.config.Version.VersionNumber(),
ArtifactDownloaded: r.artifactsDownloaded,
TaskDirBuilt: r.taskDirBuilt,
PayloadRendered: r.payloadRendered,
CreatedResources: r.getCreatedResources(),
}
r.handleLock.Lock()
if r.handle != nil {
snap.HandleID = r.handle.ID()
}
r.handleLock.Unlock()
r.driverNetLock.Lock()
snap.DriverNetwork = r.driverNet.Copy()
r.driverNetLock.Unlock()
// If nothing has changed avoid the write
h := snap.Hash()
if bytes.Equal(h, r.persistedHash) {
return nil
}
// Serialize the object
var buf bytes.Buffer
if err := codec.NewEncoder(&buf, structs.MsgpackHandle).Encode(&snap); err != nil {
return fmt.Errorf("failed to serialize snapshot: %v", err)
}
// Start the transaction.
return r.stateDB.Batch(func(tx *bolt.Tx) error {
// Grab the task bucket
taskBkt, err := getTaskBucket(tx, r.alloc.ID, r.task.Name)
if err != nil {
return fmt.Errorf("failed to retrieve allocation bucket: %v", err)
}
if err := putData(taskBkt, taskRunnerStateAllKey, buf.Bytes()); err != nil {
return fmt.Errorf("failed to write task_runner state: %v", err)
}
// Store the hash that was persisted
tx.OnCommit(func() {
r.persistedHash = h
})
return nil
})
}
// DestroyState is used to cleanup after ourselves
func (r *TaskRunner) DestroyState() error {
r.persistLock.Lock()
defer r.persistLock.Unlock()
return r.stateDB.Update(func(tx *bolt.Tx) error {
if err := deleteTaskBucket(tx, r.alloc.ID, r.task.Name); err != nil {
return fmt.Errorf("failed to delete task bucket: %v", err)
}
return nil
})
}
// setState is used to update the state of the task runner
func (r *TaskRunner) setState(state string, event *structs.TaskEvent, lazySync bool) {
event.PopulateEventDisplayMessage()
// Persist our state to disk.
if err := r.SaveState(); err != nil {
r.logger.Printf("[ERR] client: failed to save state of Task Runner for task %q: %v", r.task.Name, err)
}
// Indicate the task has been updated.
r.updater(r.task.Name, state, event, lazySync)
}
// createDriver makes a driver for the task
func (r *TaskRunner) createDriver() (driver.Driver, error) {
// Create a task-specific event emitter callback to expose minimal
// state to drivers
eventEmitter := func(m string, args ...interface{}) {
msg := fmt.Sprintf(m, args...)
r.logger.Printf("[DEBUG] client: driver event for alloc %q: %s", r.alloc.ID, msg)
r.setState(structs.TaskStatePending, structs.NewTaskEvent(structs.TaskDriverMessage).SetDriverMessage(msg), false)
}
driverCtx := driver.NewDriverContext(r.alloc.Job.Name, r.alloc.TaskGroup, r.task.Name, r.alloc.ID, r.config, r.config.Node, r.logger, eventEmitter)
d, err := driver.NewDriver(r.task.Driver, driverCtx)
if err != nil {
return nil, fmt.Errorf("failed to create driver '%s' for alloc %s: %v",
r.task.Driver, r.alloc.ID, err)
}
return d, err
}
// Run is a long running routine used to manage the task
func (r *TaskRunner) Run() {
defer close(r.waitCh)
r.logger.Printf("[DEBUG] client: starting task context for '%s' (alloc '%s')",
r.task.Name, r.alloc.ID)
if err := r.validateTask(); err != nil {
r.setState(
structs.TaskStateDead,
structs.NewTaskEvent(structs.TaskFailedValidation).SetValidationError(err).SetFailsTask(),
false)
return
}
// Create a temporary driver so that we can determine the FSIsolation
// required. run->startTask will create a new driver after environment
// has been setup (env vars, templates, artifacts, secrets, etc).
tmpDrv, err := r.createDriver()
if err != nil {
e := fmt.Errorf("failed to create driver of task %q for alloc %q: %v", r.task.Name, r.alloc.ID, err)
r.setState(
structs.TaskStateDead,
structs.NewTaskEvent(structs.TaskSetupFailure).SetSetupError(e).SetFailsTask(),
false)
return
}
// Build base task directory structure regardless of FS isolation abilities.
// This needs to happen before we start the Vault manager and call prestart
// as both those can write to the task directories
if err := r.buildTaskDir(tmpDrv.FSIsolation()); err != nil {
e := fmt.Errorf("failed to build task directory for %q: %v", r.task.Name, err)
r.setState(
structs.TaskStateDead,
structs.NewTaskEvent(structs.TaskSetupFailure).SetSetupError(e).SetFailsTask(),
false)
return
}
// If there is no Vault policy leave the static future created in
// NewTaskRunner
if r.task.Vault != nil {
// Start the go-routine to get a Vault token
r.vaultFuture.Clear()
go r.vaultManager(r.recoveredVaultToken)
}
// Start the run loop
r.run()
// Do any cleanup necessary
r.postrun()
return
}
// validateTask validates the fields of the task and returns an error if the
// task is invalid.
func (r *TaskRunner) validateTask() error {
var mErr multierror.Error
// Validate the user.
unallowedUsers := r.config.ReadStringListToMapDefault("user.blacklist", config.DefaultUserBlacklist)
checkDrivers := r.config.ReadStringListToMapDefault("user.checked_drivers", config.DefaultUserCheckedDrivers)
if _, driverMatch := checkDrivers[r.task.Driver]; driverMatch {
if _, unallowed := unallowedUsers[r.task.User]; unallowed {
mErr.Errors = append(mErr.Errors, fmt.Errorf("running as user %q is disallowed", r.task.User))
}
}
// Validate the artifacts
for i, artifact := range r.task.Artifacts {
// Verify the artifact doesn't escape the task directory.
if err := artifact.Validate(); err != nil {
// If this error occurs there is potentially a server bug or
// malicious, server spoofing.
r.logger.Printf("[ERR] client: allocation %q, task %v, artifact %#v (%v) fails validation: %v",
r.alloc.ID, r.task.Name, artifact, i, err)
mErr.Errors = append(mErr.Errors, fmt.Errorf("artifact (%d) failed validation: %v", i, err))
}
}
// Validate the Service names
taskEnv := r.envBuilder.Build()
for i, service := range r.task.Services {
name := taskEnv.ReplaceEnv(service.Name)
if err := service.ValidateName(name); err != nil {
mErr.Errors = append(mErr.Errors, fmt.Errorf("service (%d) failed validation: %v", i, err))
}
}
if len(mErr.Errors) == 1 {
return mErr.Errors[0]
}
return mErr.ErrorOrNil()
}
// tokenFuture stores the Vault token and allows consumers to block till a valid
// token exists
type tokenFuture struct {
waiting []chan struct{}
token string
set bool
m sync.Mutex
}
// NewTokenFuture returns a new token future without any token set
func NewTokenFuture() *tokenFuture {
return &tokenFuture{}
}
// Wait returns a channel that can be waited on. When this channel unblocks, a
// valid token will be available via the Get method
func (f *tokenFuture) Wait() <-chan struct{} {
f.m.Lock()
defer f.m.Unlock()
c := make(chan struct{})
if f.set {
close(c)
return c
}
f.waiting = append(f.waiting, c)
return c
}
// Set sets the token value and unblocks any caller of Wait
func (f *tokenFuture) Set(token string) *tokenFuture {
f.m.Lock()
defer f.m.Unlock()
f.set = true
f.token = token
for _, w := range f.waiting {
close(w)
}
f.waiting = nil
return f
}
// Clear clears the set vault token.
func (f *tokenFuture) Clear() *tokenFuture {
f.m.Lock()
defer f.m.Unlock()
f.token = ""
f.set = false
return f
}
// Get returns the set Vault token
func (f *tokenFuture) Get() string {
f.m.Lock()
defer f.m.Unlock()
return f.token
}
// vaultManager should be called in a go-routine and manages the derivation,
// renewal and handling of errors with the Vault token. The optional parameter
// allows setting the initial Vault token. This is useful when the Vault token
// is recovered off disk.
func (r *TaskRunner) vaultManager(token string) {
// Helper for stopping token renewal
stopRenewal := func() {
if err := r.vaultClient.StopRenewToken(r.vaultFuture.Get()); err != nil {
r.logger.Printf("[WARN] client: failed to stop token renewal for task %v in alloc %q: %v", r.task.Name, r.alloc.ID, err)
}
}
// updatedToken lets us store state between loops. If true, a new token
// has been retrieved and we need to apply the Vault change mode
var updatedToken bool
OUTER:
for {
// Check if we should exit
select {
case <-r.waitCh:
stopRenewal()
return
default:
}
// Clear the token
r.vaultFuture.Clear()
// Check if there already is a token which can be the case for
// restoring the TaskRunner
if token == "" {
// Get a token
var exit bool
token, exit = r.deriveVaultToken()
if exit {
// Exit the manager
return
}
// Write the token to disk
if err := r.writeToken(token); err != nil {
e := fmt.Errorf("failed to write Vault token to disk")
r.logger.Printf("[ERR] client: %v for task %v on alloc %q: %v", e, r.task.Name, r.alloc.ID, err)
r.Kill("vault", e.Error(), true)
return
}
}
// Start the renewal process
renewCh, err := r.vaultClient.RenewToken(token, 30)
// An error returned means the token is not being renewed
if err != nil {
r.logger.Printf("[ERR] client: failed to start renewal of Vault token for task %v on alloc %q: %v", r.task.Name, r.alloc.ID, err)
token = ""
goto OUTER
}
// The Vault token is valid now, so set it
r.vaultFuture.Set(token)
if updatedToken {
switch r.task.Vault.ChangeMode {
case structs.VaultChangeModeSignal:
s, err := signals.Parse(r.task.Vault.ChangeSignal)
if err != nil {
e := fmt.Errorf("failed to parse signal: %v", err)
r.logger.Printf("[ERR] client: %v", err)
r.Kill("vault", e.Error(), true)
return
}
if err := r.Signal("vault", "new Vault token acquired", s); err != nil {
r.logger.Printf("[ERR] client: failed to send signal to task %v for alloc %q: %v", r.task.Name, r.alloc.ID, err)
r.Kill("vault", fmt.Sprintf("failed to send signal to task: %v", err), true)
return
}
case structs.VaultChangeModeRestart:
const noFailure = false
r.Restart("vault", "new Vault token acquired", noFailure)
case structs.VaultChangeModeNoop:
fallthrough
default:
r.logger.Printf("[ERR] client: Invalid Vault change mode: %q", r.task.Vault.ChangeMode)
}
// We have handled it
updatedToken = false
// Call the handler
r.updatedTokenHandler()
}
// Start watching for renewal errors
select {
case err := <-renewCh:
// Clear the token
token = ""
r.logger.Printf("[ERR] client: failed to renew Vault token for task %v on alloc %q: %v", r.task.Name, r.alloc.ID, err)
stopRenewal()
// Check if we have to do anything
if r.task.Vault.ChangeMode != structs.VaultChangeModeNoop {
updatedToken = true
}
case <-r.waitCh:
stopRenewal()
return
}
}
}
// deriveVaultToken derives the Vault token using exponential backoffs. It
// returns the Vault token and whether the manager should exit.
func (r *TaskRunner) deriveVaultToken() (token string, exit bool) {
attempts := 0
for {
tokens, err := r.vaultClient.DeriveToken(r.alloc, []string{r.task.Name})
if err == nil {
return tokens[r.task.Name], false
}
// Check if this is a server side error
if structs.IsServerSide(err) {
r.logger.Printf("[ERR] client: failed to derive Vault token for task %v on alloc %q: %v",
r.task.Name, r.alloc.ID, err)
r.Kill("vault", fmt.Sprintf("server error deriving vault token: %v", err), true)
return "", true
}
// Check if we can't recover from the error
if !structs.IsRecoverable(err) {
r.logger.Printf("[ERR] client: failed to derive Vault token for task %v on alloc %q: %v",
r.task.Name, r.alloc.ID, err)
r.Kill("vault", fmt.Sprintf("failed to derive token: %v", err), true)
return "", true
}
// Handle the retry case
backoff := (1 << (2 * uint64(attempts))) * vaultBackoffBaseline
if backoff > vaultBackoffLimit {
backoff = vaultBackoffLimit
}
r.logger.Printf("[ERR] client: failed to derive Vault token for task %v on alloc %q: %v; retrying in %v",
r.task.Name, r.alloc.ID, err, backoff)
attempts++
// Wait till retrying
select {
case <-r.waitCh:
return "", true
case <-time.After(backoff):
}
}
}
// writeToken writes the given token to disk
func (r *TaskRunner) writeToken(token string) error {
tokenPath := filepath.Join(r.taskDir.SecretsDir, vaultTokenFile)
if err := ioutil.WriteFile(tokenPath, []byte(token), 0777); err != nil {
return fmt.Errorf("failed to save Vault tokens to secret dir for task %q in alloc %q: %v", r.task.Name, r.alloc.ID, err)
}
return nil
}
// updatedTokenHandler is called when a new Vault token is retrieved. Things
// that rely on the token should be updated here.
func (r *TaskRunner) updatedTokenHandler() {
// Update the tasks environment
r.envBuilder.SetVaultToken(r.vaultFuture.Get(), r.task.Vault.Env)
if r.templateManager != nil {
r.templateManager.Stop()
// Create a new templateManager
var err error
r.templateManager, err = NewTaskTemplateManager(&TaskTemplateManagerConfig{
Hooks: r,
Templates: r.task.Templates,
ClientConfig: r.config,
VaultToken: r.vaultFuture.Get(),
TaskDir: r.taskDir.Dir,
EnvBuilder: r.envBuilder,
MaxTemplateEventRate: DefaultMaxTemplateEventRate,
})
if err != nil {
err := fmt.Errorf("failed to build task's template manager: %v", err)
r.setState(structs.TaskStateDead,
structs.NewTaskEvent(structs.TaskSetupFailure).SetSetupError(err).SetFailsTask(),
false)
r.logger.Printf("[ERR] client: alloc %q, task %q %v", r.alloc.ID, r.task.Name, err)
r.Kill("vault", err.Error(), true)
return
}
}
}
// prestart handles life-cycle tasks that occur before the task has started.
// Since it's run asynchronously with the main Run() loop the alloc & task are
// passed in to avoid racing with updates.
func (r *TaskRunner) prestart(alloc *structs.Allocation, task *structs.Task, resultCh chan bool) {
if task.Vault != nil {
// Wait for the token
r.logger.Printf("[DEBUG] client: waiting for Vault token for task %v in alloc %q", task.Name, alloc.ID)
tokenCh := r.vaultFuture.Wait()
select {
case <-tokenCh:
case <-r.waitCh:
resultCh <- false
return
}
r.logger.Printf("[DEBUG] client: retrieved Vault token for task %v in alloc %q", task.Name, alloc.ID)
r.envBuilder.SetVaultToken(r.vaultFuture.Get(), task.Vault.Env)
}
// If the job is a dispatch job and there is a payload write it to disk
requirePayload := len(alloc.Job.Payload) != 0 &&
(r.task.DispatchPayload != nil && r.task.DispatchPayload.File != "")
if !r.payloadRendered && requirePayload {
renderTo := filepath.Join(r.taskDir.LocalDir, task.DispatchPayload.File)
decoded, err := snappy.Decode(nil, alloc.Job.Payload)
if err != nil {
r.setState(
structs.TaskStateDead,
structs.NewTaskEvent(structs.TaskSetupFailure).SetSetupError(err).SetFailsTask(),
false)
resultCh <- false
return
}
if err := os.MkdirAll(filepath.Dir(renderTo), 07777); err != nil {
r.setState(
structs.TaskStateDead,
structs.NewTaskEvent(structs.TaskSetupFailure).SetSetupError(err).SetFailsTask(),
false)
resultCh <- false
return
}
if err := ioutil.WriteFile(renderTo, decoded, 0777); err != nil {
r.setState(
structs.TaskStateDead,
structs.NewTaskEvent(structs.TaskSetupFailure).SetSetupError(err).SetFailsTask(),
false)
resultCh <- false
return
}
r.payloadRendered = true
}
for {
r.persistLock.Lock()
downloaded := r.artifactsDownloaded
r.persistLock.Unlock()
// Download the task's artifacts
if !downloaded && len(task.Artifacts) > 0 {
r.setState(structs.TaskStatePending, structs.NewTaskEvent(structs.TaskDownloadingArtifacts), false)
taskEnv := r.envBuilder.Build()
for _, artifact := range task.Artifacts {
if err := getter.GetArtifact(taskEnv, artifact, r.taskDir.Dir); err != nil {