Repository navigation
Expand file tree
/
Copy pathprovisionerdserver.go
More file actions
3898 lines (3538 loc) · 140 KB
/
Copy pathprovisionerdserver.go
File metadata and controls
3898 lines (3538 loc) · 140 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
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 provisionerdserver
import (
"context"
"crypto/sha256"
"database/sql"
"encoding/hex"
"encoding/json"
"errors"
"fmt"
"net/http"
"net/url"
"reflect"
"slices"
"sort"
"strconv"
"strings"
"sync"
"sync/atomic"
"time"
"github.com/google/uuid"
"github.com/sqlc-dev/pqtype"
semconv "go.opentelemetry.io/otel/semconv/v1.14.0"
"go.opentelemetry.io/otel/trace"
"golang.org/x/exp/maps"
"golang.org/x/oauth2"
"golang.org/x/xerrors"
protobuf "google.golang.org/protobuf/proto"
"cdr.dev/slog/v3"
"github.com/coder/coder/v2/coderd/aiseats"
"github.com/coder/coder/v2/coderd/apikey"
"github.com/coder/coder/v2/coderd/audit"
"github.com/coder/coder/v2/coderd/database"
"github.com/coder/coder/v2/coderd/database/dbauthz"
"github.com/coder/coder/v2/coderd/database/dbtime"
"github.com/coder/coder/v2/coderd/database/pubsub"
"github.com/coder/coder/v2/coderd/externalauth"
"github.com/coder/coder/v2/coderd/notifications"
"github.com/coder/coder/v2/coderd/prebuilds"
"github.com/coder/coder/v2/coderd/promoauth"
"github.com/coder/coder/v2/coderd/schedule"
"github.com/coder/coder/v2/coderd/telemetry"
"github.com/coder/coder/v2/coderd/tracing"
"github.com/coder/coder/v2/coderd/usage"
"github.com/coder/coder/v2/coderd/util/slice"
"github.com/coder/coder/v2/coderd/wspubsub"
"github.com/coder/coder/v2/codersdk"
"github.com/coder/coder/v2/codersdk/agentsdk"
"github.com/coder/coder/v2/codersdk/drpcsdk"
"github.com/coder/coder/v2/provisioner"
"github.com/coder/coder/v2/provisionerd/proto"
"github.com/coder/coder/v2/provisionersdk"
sdkproto "github.com/coder/coder/v2/provisionersdk/proto"
"github.com/coder/quartz"
)
const (
tarMimeType = "application/x-tar"
)
const (
// DefaultAcquireJobLongPollDur is the time the (deprecated) AcquireJob rpc waits to try to obtain a job before
// canceling and returning an empty job.
DefaultAcquireJobLongPollDur = time.Second * 5
// DefaultHeartbeatInterval is the interval at which the provisioner daemon
// will update its last seen at timestamp in the database.
DefaultHeartbeatInterval = time.Minute
// StaleInterval is the amount of time after the last heartbeat for which
// the provisioner will be reported as 'stale'.
StaleInterval = 90 * time.Second
)
type Options struct {
OIDCConfig promoauth.OAuth2Config
ExternalAuthConfigs []*externalauth.Config
AISeatTracker aiseats.SeatTracker
// KeyID is the provisioner key the daemon authenticated with, or the
// zero value if it did not authenticate with a key.
KeyID uuid.UUID
// SessionCancel terminates the daemon's session. Required when KeyID is a
// deletable provisioner key; optional otherwise.
SessionCancel context.CancelFunc
// Clock for testing
Clock quartz.Clock
// AcquireJobLongPollDur is used in tests
AcquireJobLongPollDur time.Duration
// HeartbeatInterval is the interval at which the provisioner daemon
// will update its last seen at timestamp in the database.
HeartbeatInterval time.Duration
// HeartbeatFn is the function that will be called at the interval
// specified by HeartbeatInterval.
// The default function just calls UpdateProvisionerDaemonLastSeenAt.
// This is mainly used for testing.
HeartbeatFn func(context.Context) error
}
type server struct {
apiVersion string
// lifecycleCtx must be tied to the API server's lifecycle
// as when the API server shuts down, we want to cancel any
// long-running operations.
lifecycleCtx context.Context
AccessURL *url.URL
ID uuid.UUID
KeyID uuid.UUID
sessionCancel context.CancelFunc
OrganizationID uuid.UUID
Logger slog.Logger
Provisioners []database.ProvisionerType
ExternalAuthConfigs []*externalauth.Config
Tags Tags
Database database.Store
Pubsub pubsub.Pubsub
Acquirer *Acquirer
Telemetry telemetry.Reporter
Tracer trace.Tracer
QuotaCommitter *atomic.Pointer[proto.QuotaCommitter]
Auditor *atomic.Pointer[audit.Auditor]
TemplateScheduleStore *atomic.Pointer[schedule.TemplateScheduleStore]
UserQuietHoursScheduleStore *atomic.Pointer[schedule.UserQuietHoursScheduleStore]
DeploymentValues *codersdk.DeploymentValues
NotificationsEnqueuer notifications.Enqueuer
PrebuildsOrchestrator *atomic.Pointer[prebuilds.ReconciliationOrchestrator]
UsageInserter *atomic.Pointer[usage.Inserter]
AISeatTracker aiseats.SeatTracker
Experiments codersdk.Experiments
OIDCConfig promoauth.OAuth2Config
Clock quartz.Clock
acquireJobLongPollDur time.Duration
heartbeatInterval time.Duration
heartbeatFn func(ctx context.Context) error
// jobMu guards activeJobs and terminationPending.
jobMu sync.Mutex
// activeJobs tracks jobs claimed by this session that have not yet been
// completed or failed. The in-tree provisioner daemon runs jobs serially,
// so at most one entry is expected; the protocol does not enforce this.
activeJobs map[uuid.UUID]struct{}
// terminationPending records a termination request that arrived while a
// job was active; it is performed when the last active job finishes.
terminationPending bool
metrics *Metrics
}
// We use the null byte (0x00) in generating a canonical map key for tags, so
// it cannot be used in the tag keys or values.
var ErrTagsContainNullByte = xerrors.New("tags cannot contain the null byte (0x00)")
// ErrProvisionerKeyDeleted is returned from job acquisition when the
// provisioner key the daemon authenticated with no longer exists.
var ErrProvisionerKeyDeleted = xerrors.New("provisioner key was deleted")
type Tags map[string]string
func (t Tags) ToJSON() (json.RawMessage, error) {
r, err := json.Marshal(t)
if err != nil {
return nil, err
}
return r, err
}
func (t Tags) Valid() error {
for k, v := range t {
if slices.Contains([]byte(k), 0x00) || slices.Contains([]byte(v), 0x00) {
return ErrTagsContainNullByte
}
}
return nil
}
// Server is the provisioner daemon DRPC server plus session-lifecycle hooks
// used by the serve handlers.
type Server interface {
proto.DRPCProvisionerDaemonServer
// TerminateSession cancels the session once no acquired job is
// active. Safe to call from any goroutine.
TerminateSession()
}
func NewServer(
lifecycleCtx context.Context,
apiVersion string,
accessURL *url.URL,
id uuid.UUID,
organizationID uuid.UUID,
logger slog.Logger,
provisioners []database.ProvisionerType,
tags Tags,
db database.Store,
ps pubsub.Pubsub,
acquirer *Acquirer,
tel telemetry.Reporter,
tracer trace.Tracer,
quotaCommitter *atomic.Pointer[proto.QuotaCommitter],
auditor *atomic.Pointer[audit.Auditor],
templateScheduleStore *atomic.Pointer[schedule.TemplateScheduleStore],
userQuietHoursScheduleStore *atomic.Pointer[schedule.UserQuietHoursScheduleStore],
usageInserter *atomic.Pointer[usage.Inserter],
deploymentValues *codersdk.DeploymentValues,
options Options,
enqueuer notifications.Enqueuer,
prebuildsOrchestrator *atomic.Pointer[prebuilds.ReconciliationOrchestrator],
metrics *Metrics,
experiments codersdk.Experiments,
) (Server, error) {
// Fail-fast if pointers are nil
if lifecycleCtx == nil {
return nil, xerrors.New("ctx is nil")
}
// A deletable key's session must be cancelable, otherwise key deletion
// cannot terminate it.
if codersdk.IsDeletableProvisionerKey(options.KeyID) && options.SessionCancel == nil {
return nil, xerrors.New("SessionCancel is required when KeyID is a deletable provisioner key")
}
if quotaCommitter == nil {
return nil, xerrors.New("quotaCommitter is nil")
}
if auditor == nil {
return nil, xerrors.New("auditor is nil")
}
if templateScheduleStore == nil {
return nil, xerrors.New("templateScheduleStore is nil")
}
if userQuietHoursScheduleStore == nil {
return nil, xerrors.New("userQuietHoursScheduleStore is nil")
}
if usageInserter == nil {
return nil, xerrors.New("usageCollector is nil")
}
if deploymentValues == nil {
return nil, xerrors.New("deploymentValues is nil")
}
if acquirer == nil {
return nil, xerrors.New("acquirer is nil")
}
if tags == nil {
return nil, xerrors.Errorf("tags is nil")
}
if err := tags.Valid(); err != nil {
return nil, xerrors.Errorf("invalid tags: %w", err)
}
if options.AISeatTracker == nil {
options.AISeatTracker = aiseats.Noop{}
}
if options.AcquireJobLongPollDur == 0 {
options.AcquireJobLongPollDur = DefaultAcquireJobLongPollDur
}
if options.HeartbeatInterval == 0 {
options.HeartbeatInterval = DefaultHeartbeatInterval
}
if options.Clock == nil {
options.Clock = quartz.NewReal()
}
s := &server{
lifecycleCtx: lifecycleCtx,
apiVersion: apiVersion,
AccessURL: accessURL,
ID: id,
KeyID: options.KeyID,
sessionCancel: options.SessionCancel,
OrganizationID: organizationID,
Logger: logger,
Provisioners: provisioners,
ExternalAuthConfigs: options.ExternalAuthConfigs,
Tags: tags,
Database: db,
Pubsub: ps,
Acquirer: acquirer,
NotificationsEnqueuer: enqueuer,
Telemetry: tel,
Tracer: tracer,
QuotaCommitter: quotaCommitter,
Auditor: auditor,
TemplateScheduleStore: templateScheduleStore,
UserQuietHoursScheduleStore: userQuietHoursScheduleStore,
DeploymentValues: deploymentValues,
OIDCConfig: options.OIDCConfig,
Clock: options.Clock,
acquireJobLongPollDur: options.AcquireJobLongPollDur,
heartbeatInterval: options.HeartbeatInterval,
heartbeatFn: options.HeartbeatFn,
PrebuildsOrchestrator: prebuildsOrchestrator,
UsageInserter: usageInserter,
AISeatTracker: options.AISeatTracker,
activeJobs: map[uuid.UUID]struct{}{},
metrics: metrics,
Experiments: experiments,
}
if s.heartbeatFn == nil {
s.heartbeatFn = s.defaultHeartbeat
}
go s.heartbeatLoop()
return s, nil
}
// timeNow should be used when trying to get the current time for math
// calculations regarding workspace start and stop time.
func (s *server) timeNow(tags ...string) time.Time {
return dbtime.Time(s.Clock.Now(tags...))
}
// heartbeatLoop runs heartbeatOnce at the interval specified by HeartbeatInterval
// until the lifecycle context is canceled.
func (s *server) heartbeatLoop() {
tick := time.NewTicker(time.Nanosecond)
defer tick.Stop()
for {
select {
case <-s.lifecycleCtx.Done():
s.Logger.Debug(s.lifecycleCtx, "heartbeat loop canceled")
return
case <-tick.C:
if s.lifecycleCtx.Err() != nil {
return
}
start := s.timeNow()
hbCtx, hbCancel := context.WithTimeout(s.lifecycleCtx, s.heartbeatInterval)
if err := s.heartbeat(hbCtx); err != nil && !database.IsQueryCanceledError(err) {
s.Logger.Warn(hbCtx, "heartbeat failed", slog.Error(err))
}
// The key check rides the heartbeat tick so a session whose deletable
// key is gone terminates within one interval. Transient errors are
// logged and the session is left running.
if deleted, err := s.keyDeleted(hbCtx); err != nil && !database.IsQueryCanceledError(err) {
s.Logger.Warn(hbCtx, "check provisioner key on heartbeat", slog.Error(err))
} else if deleted {
s.Logger.Warn(hbCtx, "provisioner key deleted, canceling session",
slog.F("provisioner_key_id", s.KeyID))
s.TerminateSession()
}
hbCancel()
elapsed := s.timeNow().Sub(start)
nextBeat := s.heartbeatInterval - elapsed
// avoid negative interval
if nextBeat <= 0 {
nextBeat = time.Nanosecond
}
tick.Reset(nextBeat)
}
}
}
// heartbeat updates the last seen at timestamp in the database.
// If HeartbeatFn is set, it will be called instead.
func (s *server) heartbeat(ctx context.Context) error {
select {
case <-ctx.Done():
return nil
default:
return s.heartbeatFn(ctx)
}
}
func (s *server) defaultHeartbeat(ctx context.Context) error {
//nolint:gocritic // This is specifically for updating the last seen at timestamp.
return s.Database.UpdateProvisionerDaemonLastSeenAt(dbauthz.AsSystemRestricted(ctx), database.UpdateProvisionerDaemonLastSeenAtParams{
ID: s.ID,
LastSeenAt: sql.NullTime{Time: s.timeNow(), Valid: true},
})
}
// keyDeleted reports whether the provisioner key no longer exists.
func (s *server) keyDeleted(ctx context.Context) (bool, error) {
if !codersdk.IsDeletableProvisionerKey(s.KeyID) {
return false, nil
}
_, err := s.Database.GetProvisionerKeyByID(
//nolint:gocritic // Callers' contexts cannot read provisioner keys
// (provisionerd actor or no actor at all), so scope the read to this
// narrow subject.
dbauthz.AsSystemReadProvisionerDaemons(ctx), s.KeyID)
if errors.Is(err, sql.ErrNoRows) {
return true, nil
}
if err != nil {
return false, xerrors.Errorf("get provisioner key: %w", err)
}
return false, nil
}
// TerminateSession cancels the session. Cancellation is deferred while a job
// claimed by this session is active so the daemon can report the job's
// result; the last active job's completion performs it. Requires a configured
// sessionCancel.
func (s *server) TerminateSession() {
s.jobMu.Lock()
if len(s.activeJobs) > 0 {
s.terminationPending = true
s.jobMu.Unlock()
s.Logger.Info(s.lifecycleCtx, "deferring session cancellation until active jobs finish",
slog.F("provisioner_key_id", s.KeyID))
return
}
s.jobMu.Unlock()
s.sessionCancel()
}
// jobStarted records a job claimed by this session as active.
func (s *server) jobStarted(id uuid.UUID) {
s.jobMu.Lock()
defer s.jobMu.Unlock()
s.activeJobs[id] = struct{}{}
}
// jobFinished removes an active job and performs a termination deferred while
// jobs were active. The daemon may not receive the final RPC response when
// this cancels the session; the job's outcome is already persisted.
func (s *server) jobFinished(id uuid.UUID) {
s.jobMu.Lock()
delete(s.activeJobs, id)
terminate := s.terminationPending && len(s.activeJobs) == 0
s.jobMu.Unlock()
if !terminate {
return
}
s.Logger.Warn(s.lifecycleCtx, "canceling session after job completion",
slog.F("provisioner_key_id", s.KeyID))
s.sessionCancel()
}
// AcquireJob queries the database to lock a job.
//
// Deprecated: This method is only available for back-level provisioner daemons.
func (s *server) AcquireJob(ctx context.Context, _ *proto.Empty) (*proto.AcquiredJob, error) {
//nolint:gocritic // Provisionerd has specific authz rules.
ctx = dbauthz.AsProvisionerd(ctx)
if deleted, err := s.keyDeleted(ctx); err != nil {
return nil, xerrors.Errorf("acquire job: check provisioner key: %w", err)
} else if deleted {
s.Logger.Warn(ctx, "provisioner key deleted, rejecting job acquisition",
slog.F("provisioner_key_id", s.KeyID))
s.TerminateSession()
return nil, xerrors.Errorf("acquire job: %w", ErrProvisionerKeyDeleted)
}
// Since AcquireJob blocks until a job is available, we set a long (5s by default) timeout. This allows back-level
// provisioner daemons to gracefully shut down within a few seconds, but keeps them from rapidly polling the
// database.
acqCtx, acqCancel := context.WithTimeout(ctx, s.acquireJobLongPollDur)
defer acqCancel()
job, err := s.Acquirer.AcquireJob(acqCtx, s.OrganizationID, s.ID, s.Provisioners, s.Tags, s.KeyID)
if database.IsQueryCanceledError(err) {
s.Logger.Debug(ctx, "successful cancel")
return &proto.AcquiredJob{}, nil
}
if errors.Is(err, ErrProvisionerKeyDeleted) {
s.Logger.Warn(ctx, "provisioner key deleted, rejecting job acquisition",
slog.F("provisioner_key_id", s.KeyID))
s.TerminateSession()
}
if err != nil {
return nil, xerrors.Errorf("acquire job: %w", err)
}
s.Logger.Debug(ctx, "locked job from database", slog.F("job_id", job.ID))
s.jobStarted(job.ID)
pj, err := s.acquireProtoJob(ctx, job)
if err != nil {
s.jobFinished(job.ID)
return nil, err
}
return pj, nil
}
type jobAndErr struct {
job database.ProvisionerJob
err error
}
// AcquireJobWithCancel queries the database to lock a job.
func (s *server) AcquireJobWithCancel(stream proto.DRPCProvisionerDaemon_AcquireJobWithCancelStream) (retErr error) {
//nolint:gocritic // Provisionerd has specific authz rules.
streamCtx := dbauthz.AsProvisionerd(stream.Context())
defer func() {
closeErr := stream.Close()
s.Logger.Debug(streamCtx, "closed stream", slog.Error(closeErr))
if retErr == nil {
retErr = closeErr
}
}()
if deleted, err := s.keyDeleted(streamCtx); err != nil {
return xerrors.Errorf("acquire job: check provisioner key: %w", err)
} else if deleted {
s.Logger.Warn(streamCtx, "provisioner key deleted, rejecting job acquisition",
slog.F("provisioner_key_id", s.KeyID))
s.TerminateSession()
return xerrors.Errorf("acquire job: %w", ErrProvisionerKeyDeleted)
}
acqCtx, acqCancel := context.WithCancel(streamCtx)
defer acqCancel()
recvCh := make(chan error, 1)
go func() {
_, err := stream.Recv() // cancel is the only message
recvCh <- err
}()
jec := make(chan jobAndErr, 1)
go func() {
job, err := s.Acquirer.AcquireJob(acqCtx, s.OrganizationID, s.ID, s.Provisioners, s.Tags, s.KeyID)
jec <- jobAndErr{job: job, err: err}
}()
var recvErr error
var je jobAndErr
select {
case recvErr = <-recvCh:
acqCancel()
je = <-jec
case je = <-jec:
}
if database.IsQueryCanceledError(je.err) {
s.Logger.Debug(streamCtx, "successful cancel")
err := stream.Send(&proto.AcquiredJob{})
if err != nil {
// often this is just because the other side hangs up and doesn't wait for the cancel, so log at INFO
s.Logger.Info(streamCtx, "failed to send empty job", slog.Error(err))
return err
}
return nil
}
if errors.Is(je.err, ErrProvisionerKeyDeleted) {
s.Logger.Warn(streamCtx, "provisioner key deleted, rejecting job acquisition",
slog.F("provisioner_key_id", s.KeyID))
s.TerminateSession()
}
if je.err != nil {
return xerrors.Errorf("acquire job: %w", je.err)
}
logger := s.Logger.With(slog.F("job_id", je.job.ID))
logger.Debug(streamCtx, "locked job from database")
s.jobStarted(je.job.ID)
if recvErr != nil {
logger.Error(streamCtx, "recv error and failed to cancel acquire job", slog.Error(recvErr))
// Well, this is awkward. We hit an error receiving from the stream, but didn't cancel before we locked a job
// in the database. We need to mark this job as failed so the end user can retry if they want to.
s.failAcquiredJob(logger, je.job.ID, "connection to provisioner daemon broken")
s.jobFinished(je.job.ID)
return recvErr
}
pj, err := s.acquireProtoJob(streamCtx, je.job)
if err != nil {
// acquireProtoJob marks the job failed itself.
s.jobFinished(je.job.ID)
return err
}
err = stream.Send(pj)
if err != nil {
s.Logger.Error(streamCtx, "failed to send job", slog.Error(err))
// The job was locked but never delivered, so mark it failed instead of
// leaving it assigned to a worker that does not have it.
s.failAcquiredJob(logger, je.job.ID, "connection to provisioner daemon broken")
s.jobFinished(je.job.ID)
return err
}
return nil
}
// failAcquiredJob marks a job that was claimed but never delivered to the
// daemon as failed. It uses a fresh context so the update succeeds even when
// the session context is canceled.
func (s *server) failAcquiredJob(logger slog.Logger, jobID uuid.UUID, message string) {
now := s.timeNow()
err := s.Database.UpdateProvisionerJobWithCompleteByID(
//nolint:gocritic // Provisionerd has specific authz rules.
dbauthz.AsProvisionerd(context.Background()),
database.UpdateProvisionerJobWithCompleteByIDParams{
ID: jobID,
CompletedAt: sql.NullTime{
Time: now,
Valid: true,
},
UpdatedAt: now,
Error: sql.NullString{
String: message,
Valid: true,
},
ErrorCode: sql.NullString{},
})
if err != nil {
logger.Error(s.lifecycleCtx, "error updating failed job", slog.Error(err))
}
}
func (s *server) acquireProtoJob(ctx context.Context, job database.ProvisionerJob) (*proto.AcquiredJob, error) {
// Marks the acquired job as failed with the error message provided. The
// update runs on a fresh context so it succeeds even when the session
// context is canceled; otherwise the claimed job would stay assigned to
// this worker until the job reaper.
failJob := func(errorMessage string) error {
err := s.Database.UpdateProvisionerJobWithCompleteByID(
//nolint:gocritic // Provisionerd has specific authz rules.
dbauthz.AsProvisionerd(context.Background()),
database.UpdateProvisionerJobWithCompleteByIDParams{
ID: job.ID,
CompletedAt: sql.NullTime{
Time: s.timeNow(),
Valid: true,
},
Error: sql.NullString{
String: errorMessage,
Valid: true,
},
ErrorCode: job.ErrorCode,
UpdatedAt: s.timeNow(),
})
if err != nil {
return xerrors.Errorf("update provisioner job: %w", err)
}
return xerrors.Errorf("request job was invalidated: %s", errorMessage)
}
user, err := s.Database.GetUserByID(ctx, job.InitiatorID)
if err != nil {
return nil, failJob(fmt.Sprintf("get user: %s", err))
}
jobTraceMetadata := map[string]string{}
if job.TraceMetadata.Valid {
err := json.Unmarshal(job.TraceMetadata.RawMessage, &jobTraceMetadata)
if err != nil {
return nil, failJob(fmt.Sprintf("unmarshal metadata: %s", err))
}
}
protoJob := &proto.AcquiredJob{
JobId: job.ID.String(),
CreatedAt: job.CreatedAt.UnixMilli(),
Provisioner: string(job.Provisioner),
UserName: user.Username,
TraceMetadata: jobTraceMetadata,
}
// jobTransition and jobBuildReason are used for metrics; only set for workspace builds.
var jobTransition string
var jobBuildReason string
switch job.Type {
case database.ProvisionerJobTypeWorkspaceBuild:
var input WorkspaceProvisionJob
err = json.Unmarshal(job.Input, &input)
if err != nil {
return nil, failJob(fmt.Sprintf("unmarshal job input %q: %s", job.Input, err))
}
workspaceBuild, err := s.Database.GetWorkspaceBuildByID(ctx, input.WorkspaceBuildID)
if err != nil {
return nil, failJob(fmt.Sprintf("get workspace build: %s", err))
}
workspace, err := s.Database.GetWorkspaceByID(ctx, workspaceBuild.WorkspaceID)
if err != nil {
return nil, failJob(fmt.Sprintf("get workspace: %s", err))
}
templateVersion, err := s.Database.GetTemplateVersionByID(ctx, workspaceBuild.TemplateVersionID)
if err != nil {
return nil, failJob(fmt.Sprintf("get template version: %s", err))
}
templateVariables, err := s.Database.GetTemplateVersionVariables(ctx, templateVersion.ID)
if err != nil && !xerrors.Is(err, sql.ErrNoRows) {
return nil, failJob(fmt.Sprintf("get template version variables: %s", err))
}
template, err := s.Database.GetTemplateByID(ctx, templateVersion.TemplateID.UUID)
if err != nil {
return nil, failJob(fmt.Sprintf("get template: %s", err))
}
owner, err := s.Database.GetUserByID(ctx, workspace.OwnerID)
if err != nil {
return nil, failJob(fmt.Sprintf("get owner: %s", err))
}
// Fetch the file id of the cached module files if it exists. Modules
// stay cached for parameter rendering even when the cache is
// disabled; they are only withheld from the build.
versionModulesFile := ""
if !codersdk.ModuleCacheDisabled(s.DeploymentValues, template.DisableModuleCache) {
tfvals, err := s.Database.GetTemplateVersionTerraformValues(ctx, templateVersion.ID)
if err != nil && !xerrors.Is(err, sql.ErrNoRows) {
// Older templates (before dynamic parameters) will not have cached module files.
return nil, failJob(fmt.Sprintf("get template version terraform values: %s", err))
}
if err == nil && tfvals.CachedModuleFiles.Valid {
versionModulesFile = tfvals.CachedModuleFiles.UUID.String()
}
}
var ownerSSHPublicKey, ownerSSHPrivateKey string
if ownerSSHKey, err := s.Database.GetGitSSHKey(ctx, owner.ID); err != nil {
if !xerrors.Is(err, sql.ErrNoRows) {
return nil, failJob(fmt.Sprintf("get owner ssh key: %s", err))
}
} else {
ownerSSHPublicKey = ownerSSHKey.PublicKey
ownerSSHPrivateKey = ownerSSHKey.PrivateKey
}
ownerGroups, err := s.Database.GetGroups(ctx, database.GetGroupsParams{
HasMemberID: owner.ID,
OrganizationID: s.OrganizationID,
})
if err != nil {
return nil, failJob(fmt.Sprintf("get owner group names: %s", err))
}
ownerGroupNames := []string{}
for _, group := range ownerGroups {
ownerGroupNames = append(ownerGroupNames, group.Group.Name)
}
msg, err := json.Marshal(wspubsub.WorkspaceEvent{
Kind: wspubsub.WorkspaceEventKindStateChange,
WorkspaceID: workspace.ID,
})
if err != nil {
return nil, failJob(fmt.Sprintf("marshal workspace update event: %s", err))
}
err = s.Pubsub.Publish(wspubsub.WorkspaceEventChannel(workspace.OwnerID), msg)
if err != nil {
return nil, failJob(fmt.Sprintf("publish workspace update: %s", err))
}
var workspaceOwnerOIDCAccessToken string
// The check `s.OIDCConfig != nil` is not as strict, since it can be an interface
// pointing to a typed nil.
if !reflect.ValueOf(s.OIDCConfig).IsNil() {
workspaceOwnerOIDCAccessToken, err = ObtainOIDCAccessToken(ctx, s.Logger, s.Database, s.OIDCConfig, owner.ID)
if err != nil {
return nil, failJob(fmt.Sprintf("obtain OIDC access token: %s", err))
}
}
var sessionToken string
switch workspaceBuild.Transition {
case database.WorkspaceTransitionStart:
sessionToken, err = s.regenerateSessionToken(ctx, owner, workspace)
if err != nil {
return nil, failJob(fmt.Sprintf("regenerate session token: %s", err))
}
case database.WorkspaceTransitionStop, database.WorkspaceTransitionDelete:
err = deleteSessionToken(ctx, s.Database, workspace)
if err != nil {
return nil, failJob(fmt.Sprintf("delete session token: %s", err))
}
}
transition, err := convertWorkspaceTransition(workspaceBuild.Transition)
if err != nil {
return nil, failJob(fmt.Sprintf("convert workspace transition: %s", err))
}
jobTransition = string(workspaceBuild.Transition)
// Prebuilds use BuildReasonInitiator in the database but we want to
// track them separately in metrics. Check the initiator ID to detect
// prebuild jobs.
if job.InitiatorID == database.PrebuildsSystemUserID {
jobBuildReason = BuildReasonPrebuild
} else {
jobBuildReason = string(workspaceBuild.Reason)
}
// A previous workspace build exists
var lastWorkspaceBuildParameters []database.WorkspaceBuildParameter
if workspaceBuild.BuildNumber > 1 {
// TODO: Should we fetch the last build that succeeded? This fetches the
// previous build regardless of the status of the build.
buildNum := workspaceBuild.BuildNumber - 1
previous, err := s.Database.GetWorkspaceBuildByWorkspaceIDAndBuildNumber(ctx, database.GetWorkspaceBuildByWorkspaceIDAndBuildNumberParams{
WorkspaceID: workspaceBuild.WorkspaceID,
BuildNumber: buildNum,
})
// If the error is ErrNoRows, then assume previous values are empty.
if err != nil && !xerrors.Is(err, sql.ErrNoRows) {
return nil, xerrors.Errorf("get last build with number=%d: %w", buildNum, err)
}
if err == nil {
lastWorkspaceBuildParameters, err = s.Database.GetWorkspaceBuildParameters(ctx, previous.ID)
if err != nil {
return nil, xerrors.Errorf("get last build parameters %q: %w", previous.ID, err)
}
}
}
workspaceBuildParameters, err := s.Database.GetWorkspaceBuildParameters(ctx, workspaceBuild.ID)
if err != nil {
return nil, failJob(fmt.Sprintf("get workspace build parameters: %s", err))
}
dbExternalAuthProviders := []database.ExternalAuthProvider{}
err = json.Unmarshal(templateVersion.ExternalAuthProviders, &dbExternalAuthProviders)
if err != nil {
return nil, xerrors.Errorf("failed to deserialize external_auth_providers value: %w", err)
}
externalAuthProviders := make([]*sdkproto.ExternalAuthProvider, 0, len(dbExternalAuthProviders))
for _, p := range dbExternalAuthProviders {
link, err := s.Database.GetExternalAuthLink(ctx, database.GetExternalAuthLinkParams{
ProviderID: p.ID,
UserID: owner.ID,
})
if errors.Is(err, sql.ErrNoRows) {
continue
}
if err != nil {
return nil, failJob(fmt.Sprintf("acquire external auth link: %s", err))
}
var config *externalauth.Config
for _, c := range s.ExternalAuthConfigs {
if c.ID != p.ID {
continue
}
config = c
break
}
// We weren't able to find a matching config for the ID!
if config == nil {
s.Logger.Warn(ctx, "workspace build job is missing external auth provider",
slog.F("provider_id", p.ID),
slog.F("template_version_id", templateVersion.ID),
slog.F("workspace_id", workspaceBuild.WorkspaceID))
continue
}
refreshed, err := config.RefreshToken(ctx, s.Database, link)
if err != nil && !externalauth.IsInvalidTokenError(err) {
return nil, failJob(fmt.Sprintf("refresh external auth link %q: %s", p.ID, err))
}
if err != nil {
// Invalid tokens are skipped
continue
}
externalAuthProviders = append(externalAuthProviders, &sdkproto.ExternalAuthProvider{
Id: p.ID,
AccessToken: refreshed.OAuthAccessToken,
})
}
allUserRoles, err := s.Database.GetAuthorizationUserRoles(ctx, owner.ID)
if err != nil {
return nil, failJob(fmt.Sprintf("get owner authorization roles: %s", err))
}
ownerRbacRoles := []*sdkproto.Role{}
roles, err := allUserRoles.RoleNames()
if err == nil {
for _, role := range roles {
if role.OrganizationID != uuid.Nil && role.OrganizationID != s.OrganizationID {
continue // Only include site wide and org specific roles
}
orgID := role.OrganizationID.String()
if role.OrganizationID == uuid.Nil {
orgID = ""
}
ownerRbacRoles = append(ownerRbacRoles, &sdkproto.Role{Name: role.Name, OrgId: orgID})
}
}
runningAgentAuthTokens := []*sdkproto.RunningAgentAuthToken{}
if input.PrebuiltWorkspaceBuildStage == sdkproto.PrebuiltWorkspaceBuildStage_CLAIM {
// runningAgentAuthTokens are *only* used for prebuilds. We fetch them when we want to rebuild a prebuilt workspace
// but not generate new agent tokens. The provisionerdserver will push them down to
// the provisioner (and ultimately to the `coder_agent` resource in the Terraform provider) where they will be
// reused. Context: the agent token is often used in immutable attributes of workspace resource (e.g. VM/container)
// to initialize the agent, so if that value changes it will necessitate a replacement of that resource, thus
// obviating the whole point of the prebuild.
agents, err := s.Database.GetWorkspaceAgentsByWorkspaceAndBuildNumber(ctx, database.GetWorkspaceAgentsByWorkspaceAndBuildNumberParams{
WorkspaceID: workspace.ID,
BuildNumber: 1,
})
if err != nil {
s.Logger.Error(ctx, "failed to retrieve running agents of claimed prebuilt workspace",
slog.F("workspace_id", workspace.ID), slog.Error(err))
}
for _, agent := range agents {
runningAgentAuthTokens = append(runningAgentAuthTokens, &sdkproto.RunningAgentAuthToken{
AgentId: agent.ID.String(),
Token: agent.AuthToken.String(),
})
}
}
provisionerStateRow, err := s.Database.GetWorkspaceBuildProvisionerStateByID(ctx, workspaceBuild.ID)
if err != nil {
return nil, failJob(fmt.Sprintf("get workspace build provisioner state: %s", err))
}
protoJob.Type = &proto.AcquiredJob_WorkspaceBuild_{
WorkspaceBuild: &proto.AcquiredJob_WorkspaceBuild{
WorkspaceBuildId: workspaceBuild.ID.String(),
WorkspaceName: workspace.Name,
State: provisionerStateRow.ProvisionerState,
RichParameterValues: convertRichParameterValues(workspaceBuildParameters),
PreviousParameterValues: convertRichParameterValues(lastWorkspaceBuildParameters),
VariableValues: asVariableValues(templateVariables),
ExternalAuthProviders: externalAuthProviders,
Metadata: &sdkproto.Metadata{
CoderUrl: s.AccessURL.String(),
WorkspaceTransition: transition,
WorkspaceName: workspace.Name,
WorkspaceOwner: owner.Username,
WorkspaceOwnerEmail: owner.Email,
WorkspaceOwnerName: owner.Name,
WorkspaceOwnerGroups: ownerGroupNames,
WorkspaceOwnerOidcAccessToken: workspaceOwnerOIDCAccessToken,
WorkspaceId: workspace.ID.String(),
WorkspaceOwnerId: owner.ID.String(),
TemplateId: template.ID.String(),
TemplateName: template.Name,
TemplateVersionId: templateVersion.ID.String(),
TemplateVersion: templateVersion.Name,
WorkspaceOwnerSessionToken: sessionToken,
WorkspaceOwnerSshPublicKey: ownerSSHPublicKey,
WorkspaceOwnerSshPrivateKey: ownerSSHPrivateKey,
WorkspaceBuildId: workspaceBuild.ID.String(),
WorkspaceOwnerLoginType: string(owner.LoginType),
WorkspaceOwnerRbacRoles: ownerRbacRoles,
RunningAgentAuthTokens: runningAgentAuthTokens,
PrebuiltWorkspaceBuildStage: input.PrebuiltWorkspaceBuildStage,
TemplateVersionModulesFile: versionModulesFile,
},
LogLevel: input.LogLevel,
},
}
case database.ProvisionerJobTypeTemplateVersionDryRun:
var input TemplateVersionDryRunJob
err = json.Unmarshal(job.Input, &input)
if err != nil {
return nil, failJob(fmt.Sprintf("unmarshal job input %q: %s", job.Input, err))
}
templateVersion, err := s.Database.GetTemplateVersionByID(ctx, input.TemplateVersionID)
if err != nil {
return nil, failJob(fmt.Sprintf("get template version: %s", err))
}
templateVariables, err := s.Database.GetTemplateVersionVariables(ctx, templateVersion.ID)
if err != nil && !xerrors.Is(err, sql.ErrNoRows) {
return nil, failJob(fmt.Sprintf("get template version variables: %s", err))
}
protoJob.Type = &proto.AcquiredJob_TemplateDryRun_{
TemplateDryRun: &proto.AcquiredJob_TemplateDryRun{
RichParameterValues: convertRichParameterValues(input.RichParameterValues),
VariableValues: asVariableValues(templateVariables),
Metadata: &sdkproto.Metadata{
CoderUrl: s.AccessURL.String(),
WorkspaceName: input.WorkspaceName,
// There is no owner for a template import, but we can assume
// the "Everyone" group as a placeholder.
WorkspaceOwnerGroups: []string{database.EveryoneGroup},
},
},
}
case database.ProvisionerJobTypeTemplateVersionImport:
var input TemplateVersionImportJob
err = json.Unmarshal(job.Input, &input)
if err != nil {
return nil, failJob(fmt.Sprintf("unmarshal job input %q: %s", job.Input, err))
}
userVariableValues, err := s.includeLastVariableValues(ctx, input.TemplateVersionID, input.UserVariableValues)
if err != nil {
return nil, failJob(err.Error())
}
templateID := ""
if input.TemplateID.Valid {
templateID = input.TemplateID.UUID.String()
}
protoJob.Type = &proto.AcquiredJob_TemplateImport_{
TemplateImport: &proto.AcquiredJob_TemplateImport{
UserVariableValues: convertVariableValues(userVariableValues),
Metadata: &sdkproto.Metadata{
CoderUrl: s.AccessURL.String(),
// There is no owner for a template import, but we can assume
// the "Everyone" group as a placeholder.
WorkspaceOwnerGroups: []string{database.EveryoneGroup},
TemplateId: templateID,
TemplateVersionId: input.TemplateVersionID.String(),
},
},
}
}
switch job.StorageMethod {
case database.ProvisionerStorageMethodFile:
file, err := s.Database.GetFileByID(ctx, job.FileID)
if err != nil {