-
Notifications
You must be signed in to change notification settings - Fork 298
/
reconciler.go
1141 lines (998 loc) · 35.4 KB
/
reconciler.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 liveupdate
import (
"context"
"errors"
"fmt"
"sync"
"time"
v1 "k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/runtime"
"k8s.io/apimachinery/pkg/types"
"sigs.k8s.io/controller-runtime/pkg/builder"
"sigs.k8s.io/controller-runtime/pkg/client"
"sigs.k8s.io/controller-runtime/pkg/handler"
apierrors "k8s.io/apimachinery/pkg/api/errors"
ctrl "sigs.k8s.io/controller-runtime"
ctrlclient "sigs.k8s.io/controller-runtime/pkg/client"
"sigs.k8s.io/controller-runtime/pkg/reconcile"
"github.com/tilt-dev/tilt/internal/build"
"github.com/tilt-dev/tilt/internal/container"
"github.com/tilt-dev/tilt/internal/containerupdate"
"github.com/tilt-dev/tilt/internal/controllers/apicmp"
"github.com/tilt-dev/tilt/internal/controllers/apis/configmap"
"github.com/tilt-dev/tilt/internal/controllers/apis/liveupdate"
"github.com/tilt-dev/tilt/internal/controllers/indexer"
"github.com/tilt-dev/tilt/internal/k8s"
"github.com/tilt-dev/tilt/internal/ospath"
"github.com/tilt-dev/tilt/internal/sliceutils"
"github.com/tilt-dev/tilt/internal/store"
"github.com/tilt-dev/tilt/internal/store/buildcontrols"
"github.com/tilt-dev/tilt/internal/store/k8sconv"
"github.com/tilt-dev/tilt/internal/store/liveupdates"
"github.com/tilt-dev/tilt/pkg/apis"
"github.com/tilt-dev/tilt/pkg/apis/core/v1alpha1"
"github.com/tilt-dev/tilt/pkg/logger"
"github.com/tilt-dev/tilt/pkg/model"
"github.com/tilt-dev/tilt/pkg/model/logstore"
)
const LiveUpdateSource = "liveupdate"
var discoveryGVK = v1alpha1.SchemeGroupVersion.WithKind("KubernetesDiscovery")
var dcsGVK = v1alpha1.SchemeGroupVersion.WithKind("DockerComposeService")
var applyGVK = v1alpha1.SchemeGroupVersion.WithKind("KubernetesApply")
var fwGVK = v1alpha1.SchemeGroupVersion.WithKind("FileWatch")
var imageMapGVK = v1alpha1.SchemeGroupVersion.WithKind("ImageMap")
var reasonObjectNotFound = "ObjectNotFound"
// Manages the LiveUpdate API object.
type Reconciler struct {
client ctrlclient.Client
indexer *indexer.Indexer
store store.RStore
ExecUpdater containerupdate.ContainerUpdater
DockerUpdater containerupdate.ContainerUpdater
updateMode liveupdates.UpdateMode
kubeContext k8s.KubeContext
startedTime metav1.MicroTime
monitors map[string]*monitor
// We need to be able to map trigger events to known resources while
// Reconcile() is running.
mu sync.Mutex
}
var _ reconcile.Reconciler = &Reconciler{}
// Dependency-inject a live update reconciler.
func NewReconciler(
st store.RStore,
dcu *containerupdate.DockerUpdater,
ecu *containerupdate.ExecUpdater,
updateMode liveupdates.UpdateMode,
kubeContext k8s.KubeContext,
client ctrlclient.Client,
scheme *runtime.Scheme) *Reconciler {
return &Reconciler{
DockerUpdater: dcu,
ExecUpdater: ecu,
updateMode: updateMode,
kubeContext: kubeContext,
client: client,
indexer: indexer.NewIndexer(scheme, indexLiveUpdate),
store: st,
startedTime: apis.NowMicro(),
monitors: make(map[string]*monitor),
}
}
// Create a reconciler baked by a fake ContainerUpdater and Client.
func NewFakeReconciler(
st store.RStore,
cu containerupdate.ContainerUpdater,
client ctrlclient.Client) *Reconciler {
scheme := v1alpha1.NewScheme()
return &Reconciler{
DockerUpdater: cu,
ExecUpdater: cu,
updateMode: liveupdates.UpdateModeAuto,
kubeContext: k8s.KubeContext("fake-context"),
client: client,
indexer: indexer.NewIndexer(scheme, indexLiveUpdate),
store: st,
startedTime: apis.NowMicro(),
monitors: make(map[string]*monitor),
}
}
func (r *Reconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.Result, error) {
r.mu.Lock()
defer r.mu.Unlock()
lu := &v1alpha1.LiveUpdate{}
err := r.client.Get(ctx, req.NamespacedName, lu)
r.indexer.OnReconcile(req.NamespacedName, lu)
if err != nil && !apierrors.IsNotFound(err) {
return ctrl.Result{}, fmt.Errorf("liveupdate reconcile: %v", err)
}
if apierrors.IsNotFound(err) || lu.ObjectMeta.DeletionTimestamp != nil {
r.store.Dispatch(liveupdates.NewLiveUpdateDeleteAction(req.Name))
delete(r.monitors, req.Name)
return ctrl.Result{}, nil
}
// The apiserver is the source of truth, and will ensure the engine state is up to date.
r.store.Dispatch(liveupdates.NewLiveUpdateUpsertAction(lu))
ctx = store.MustObjectLogHandler(ctx, r.store, lu)
if lu.Annotations[v1alpha1.AnnotationManagedBy] != "" {
// A LiveUpdate can't be managed by the reconciler until all the objects
// it depends on are managed by the reconciler. The Tiltfile controller
// is responsible for marking objects that we want to manage with ForceApply().
return ctrl.Result{}, nil
}
invalidSelectorFailedState := r.ensureSelectorValid(lu)
if invalidSelectorFailedState != nil {
return r.handleFailure(ctx, lu, invalidSelectorFailedState)
}
monitor := r.ensureMonitorExists(lu.Name, lu)
hasFileChanges, err := r.reconcileSources(ctx, monitor)
if err != nil {
if apierrors.IsNotFound(err) {
return r.handleFailure(ctx, lu, createFailedState(lu, reasonObjectNotFound, err.Error()))
}
return ctrl.Result{}, err
}
hasKubernetesChanges, err := r.reconcileKubernetesResource(ctx, monitor)
if err != nil {
if apierrors.IsNotFound(err) {
return r.handleFailure(ctx, lu, createFailedState(lu, reasonObjectNotFound, err.Error()))
}
return ctrl.Result{}, err
}
hasDockerComposeChanges, err := r.reconcileDockerComposeService(ctx, monitor)
if err != nil {
if apierrors.IsNotFound(err) {
return r.handleFailure(ctx, lu, createFailedState(lu, reasonObjectNotFound, err.Error()))
}
return ctrl.Result{}, err
}
hasTriggerQueueChanges, err := r.reconcileTriggerQueue(ctx, monitor)
if err != nil {
return ctrl.Result{}, err
}
if hasFileChanges || hasKubernetesChanges || hasDockerComposeChanges || hasTriggerQueueChanges {
monitor.hasChangesToSync = true
}
if monitor.hasChangesToSync {
status := r.maybeSync(ctx, lu, monitor)
if status.Failed != nil {
// Log any new failures.
isNew := lu.Status.Failed == nil || !apicmp.DeepEqual(lu.Status.Failed, status.Failed)
if isNew && r.shouldLogFailureReason(status.Failed) {
logger.Get(ctx).Infof("LiveUpdate %q %s: %v", lu.Name, status.Failed.Reason, status.Failed.Message)
}
}
if !apicmp.DeepEqual(lu.Status, status) {
update := lu.DeepCopy()
update.Status = status
err := r.client.Status().Update(ctx, update)
if err != nil {
return ctrl.Result{}, err
}
}
}
monitor.hasChangesToSync = false
return ctrl.Result{}, nil
}
func (r *Reconciler) shouldLogFailureReason(obj *v1alpha1.LiveUpdateStateFailed) bool {
// ObjectNotFound errors are normal before the Apply has created the KubernetesDiscovery object.
return obj.Reason != reasonObjectNotFound
}
// Check for some invalid states.
func (r *Reconciler) ensureSelectorValid(lu *v1alpha1.LiveUpdate) *v1alpha1.LiveUpdateStateFailed {
selector := lu.Spec.Selector
if selector.Kubernetes != nil {
if selector.Kubernetes.DiscoveryName == "" {
return createFailedState(lu, "Invalid", "Kubernetes selector requires DiscoveryName")
}
return nil
}
if selector.DockerCompose != nil {
if selector.DockerCompose.Service == "" {
return createFailedState(lu, "Invalid", "DockerCompose selector requires Service")
}
return nil
}
return createFailedState(lu, "Invalid", "No valid selector")
}
// If the failure state has changed, log it and write it to the apiserver.
func (r *Reconciler) handleFailure(ctx context.Context, lu *v1alpha1.LiveUpdate, failed *v1alpha1.LiveUpdateStateFailed) (ctrl.Result, error) {
isNew := lu.Status.Failed == nil || !apicmp.DeepEqual(lu.Status.Failed, failed)
if !isNew {
return ctrl.Result{}, nil
}
if r.shouldLogFailureReason(failed) {
logger.Get(ctx).Infof("LiveUpdate %q %s: %v", lu.Name, failed.Reason, failed.Message)
}
update := lu.DeepCopy()
update.Status.Failed = failed
err := r.client.Status().Update(ctx, update)
return ctrl.Result{}, err
}
// Create the monitor that tracks a live update. If the live update
// spec changes, wipe out all accumulated state.
func (r *Reconciler) ensureMonitorExists(name string, obj *v1alpha1.LiveUpdate) *monitor {
spec := obj.Spec
m, ok := r.monitors[name]
if ok && apicmp.DeepEqual(obj.Spec, m.spec) {
return m
}
m = &monitor{
manifestName: obj.Annotations[v1alpha1.AnnotationManifest],
spec: spec,
sources: make(map[string]*monitorSource),
containers: make(map[monitorContainerKey]monitorContainerStatus),
}
r.monitors[name] = m
return m
}
// Consume all FileEvents off the FileWatch objects.
// Returns true if we saw new file events.
//
// TODO(nick): Currently, it's entirely possible to miss file events. This has
// always been true (since operating systems themselves put limits on the event
// queue.) But it gets worse in a world where we read FileEvents from the API,
// since the FileWatch API itself adds lower limits.
//
// Long-term, we ought to have some way to reconnect/resync like other
// sync systems do (syncthing/rsync). e.g., diff the two file systems
// and update based on changes. But it also might make more sense to switch to a
// different library for syncing (e.g., Mutagen) now that live updates
// are decoupled from other file event-triggered tasks.
//
// In the meantime, Milas+Nick should figure out a way to handle this
// better in the short term.
func (r *Reconciler) reconcileSources(ctx context.Context, monitor *monitor) (bool, error) {
if len(monitor.spec.Sources) == 0 {
return false, nil
}
hasChange := false
for _, s := range monitor.spec.Sources {
oneChange, err := r.reconcileOneSource(ctx, monitor, s)
if err != nil {
return false, err
}
if oneChange {
hasChange = true
}
}
return hasChange, nil
}
// Consume one Source object.
func (r *Reconciler) reconcileOneSource(ctx context.Context, monitor *monitor, source v1alpha1.LiveUpdateSource) (bool, error) {
fwn := source.FileWatch
imn := source.ImageMap
var fw v1alpha1.FileWatch
if fwn != "" {
err := r.client.Get(ctx, types.NamespacedName{Name: fwn}, &fw)
if err != nil {
return false, err
}
}
var im v1alpha1.ImageMap
if imn != "" {
err := r.client.Get(ctx, types.NamespacedName{Name: imn}, &im)
if err != nil {
return false, err
}
}
events := fw.Status.FileEvents
if len(events) == 0 || fwn == "" {
return false, nil
}
mSource, ok := monitor.sources[fwn]
if !ok {
mSource = &monitorSource{
modTimeByPath: make(map[string]metav1.MicroTime),
}
monitor.sources[fwn] = mSource
}
newImageStatus := im.Status
imageChanged := false
if imn != "" {
imageChanged = mSource.lastImageStatus == nil ||
!apicmp.DeepEqual(&newImageStatus, mSource.lastImageStatus)
mSource.lastImageStatus = &im.Status
}
newLastFileEvent := events[len(events)-1]
event := mSource.lastFileEvent
fileWatchChanged := event == nil || !apicmp.DeepEqual(&newLastFileEvent, event)
mSource.lastFileEvent = &newLastFileEvent
if fileWatchChanged {
// Consume all the file events.
for _, event := range events {
eventTime := event.Time.Time
if newImageStatus.BuildStartTime != nil && newImageStatus.BuildStartTime.After(eventTime) {
continue
}
for _, f := range event.SeenFiles {
existing, ok := mSource.modTimeByPath[f]
if !ok || existing.Time.Before(event.Time.Time) {
mSource.modTimeByPath[f] = event.Time
}
}
}
}
return fileWatchChanged || imageChanged, nil
}
// Consume the TriggerQueue.
// This isn't formally represented in the API right now, it's just
// a ConfigMap to pull attributes off of.
// Returns true if we saw any changes.
func (r *Reconciler) reconcileTriggerQueue(ctx context.Context, monitor *monitor) (bool, error) {
queue, err := configmap.TriggerQueue(ctx, r.client)
if err != nil {
return false, client.IgnoreNotFound(err)
}
if monitor.lastTriggerQueue != nil && apicmp.DeepEqual(queue.Data, monitor.lastTriggerQueue.Data) {
return false, nil
}
monitor.lastTriggerQueue = queue
return true, nil
}
// Consume all objects off the KubernetesSelector.
// Returns true if we saw any changes to the objects we're watching.
func (r *Reconciler) reconcileKubernetesResource(ctx context.Context, monitor *monitor) (bool, error) {
selector := monitor.spec.Selector.Kubernetes
if selector == nil {
return false, nil
}
var kd *v1alpha1.KubernetesDiscovery
var ka *v1alpha1.KubernetesApply
var im *v1alpha1.ImageMap
changed := false
if selector.ApplyName != "" {
ka = &v1alpha1.KubernetesApply{}
err := r.client.Get(ctx, types.NamespacedName{Name: selector.ApplyName}, ka)
if err != nil {
return false, err
}
if monitor.lastKubernetesApplyStatus == nil ||
!apicmp.DeepEqual(monitor.lastKubernetesApplyStatus, &(ka.Status)) {
changed = true
}
}
kd = &v1alpha1.KubernetesDiscovery{}
err := r.client.Get(ctx, types.NamespacedName{Name: selector.DiscoveryName}, kd)
if err != nil {
return false, err
}
if selector.ImageMapName != "" {
im = &v1alpha1.ImageMap{}
if err := r.client.Get(ctx, types.NamespacedName{Name: selector.ImageMapName}, im); err != nil {
return false, err
}
if monitor.lastImageMap == nil || !apicmp.DeepEqual(monitor.lastImageMap, im) {
changed = true
}
}
if monitor.lastKubernetesDiscovery == nil ||
!apicmp.DeepEqual(monitor.lastKubernetesDiscovery.Status, kd.Status) {
changed = true
}
if ka == nil {
monitor.lastKubernetesApplyStatus = nil
} else {
monitor.lastKubernetesApplyStatus = &(ka.Status)
}
monitor.lastKubernetesDiscovery = kd
monitor.lastImageMap = im
return changed, nil
}
// Consume all objects off the DockerComposeSelector.
// Returns true if we saw any changes to the objects we're watching.
func (r *Reconciler) reconcileDockerComposeService(ctx context.Context, monitor *monitor) (bool, error) {
selector := monitor.spec.Selector.DockerCompose
if selector == nil {
return false, nil
}
var dcs v1alpha1.DockerComposeService
err := r.client.Get(ctx, types.NamespacedName{Name: selector.Service}, &dcs)
if err != nil {
return false, err
}
changed := false
if monitor.lastDockerComposeService == nil ||
!apicmp.DeepEqual(monitor.lastDockerComposeService.Status, dcs.Status) {
changed = true
}
monitor.lastDockerComposeService = &dcs
return changed, nil
}
// Go through all the file changes, and delete files that aren't relevant
// to the current build.
//
// Determining the current build is a bit tricky, but our
// order of preference is:
// 1. If we have an ImageMap.BuildStartedAt, this is the gold standard.
// 2. If there's no ImageMap, we prefer the KubernetesApply.LastApplyStartTime.
// 3. If there's no KubernetesApply, we prefer the oldest pod
// in the filtered pod list.
func (r *Reconciler) garbageCollectFileChanges(res luResource, monitor *monitor) {
for _, source := range monitor.spec.Sources {
fwn := source.FileWatch
mSource, ok := monitor.sources[fwn]
if !ok {
continue
}
lastImageStatus := mSource.lastImageStatus
var gcTime time.Time
if lastImageStatus != nil && lastImageStatus.BuildStartTime != nil {
gcTime = lastImageStatus.BuildStartTime.Time
} else {
gcTime = res.bestStartTime()
}
if !gcTime.IsZero() {
// Delete all file events that happened before the
// latest build started.
for p, t := range mSource.modTimeByPath {
if gcTime.After(t.Time) {
delete(mSource.modTimeByPath, p)
}
}
// Delete all failures that happened before the
// latest build started.
//
// This mechanism isn't perfect - for example, it will start resyncing
// again to a container that's going to be replaced by the current
// build. But we also can't determine if a container is going to be
// replaced or not (particularly if the image didn't change).
for key, c := range monitor.containers {
if !c.failedLowWaterMark.IsZero() && gcTime.After(c.failedLowWaterMark.Time) {
c.failedLowWaterMark = metav1.MicroTime{}
c.failedReason = ""
c.failedMessage = ""
monitor.containers[key] = c
}
}
}
}
}
// Go through all the container monitors, and delete any that are no longer
// being selected. We don't care why they're not being selected.
func (r *Reconciler) garbageCollectMonitorContainers(res luResource, monitor *monitor) {
// All containers are guaranteed to have container IDs if they're still active.
containerIDs := map[string]bool{}
res.visitSelectedContainers(func(pod v1alpha1.Pod, c v1alpha1.Container) bool {
if c.ID != "" {
containerIDs[c.ID] = true
}
return false
})
for key := range monitor.containers {
if !containerIDs[key.containerID] {
delete(monitor.containers, key)
}
}
}
func (r *Reconciler) dispatchStartBuildAction(ctx context.Context, lu *v1alpha1.LiveUpdate, filesChanged []string) {
manifestName := lu.Annotations[v1alpha1.AnnotationManifest]
spanID := lu.Annotations[v1alpha1.AnnotationSpanID]
r.store.Dispatch(buildcontrols.BuildStartedAction{
ManifestName: model.ManifestName(manifestName),
StartTime: time.Now(),
FilesChanged: filesChanged,
Reason: model.BuildReasonFlagChangedFiles,
SpanID: logstore.SpanID(spanID),
FullBuildTriggered: false,
Source: LiveUpdateSource,
})
buildcontrols.LogBuildEntry(ctx, buildcontrols.BuildEntry{
Name: model.ManifestName(manifestName),
BuildReason: model.BuildReasonFlagChangedFiles,
FilesChanged: filesChanged,
})
}
func (r *Reconciler) dispatchCompleteBuildAction(lu *v1alpha1.LiveUpdate, newStatus v1alpha1.LiveUpdateStatus) {
manifestName := model.ManifestName(lu.Annotations[v1alpha1.AnnotationManifest])
spanID := logstore.SpanID(lu.Annotations[v1alpha1.AnnotationSpanID])
var err error
if newStatus.Failed != nil {
err = errors.New(newStatus.Failed.Message)
} else {
for _, c := range newStatus.Containers {
if c.LastExecError != "" {
err = errors.New(c.LastExecError)
break
}
}
}
resultSet := store.BuildResultSet{}
r.store.Dispatch(buildcontrols.NewBuildCompleteAction(manifestName, LiveUpdateSource, spanID, resultSet, err))
}
func (r *Reconciler) resource(lu *v1alpha1.LiveUpdate, monitor *monitor) (luResource, error) {
k := lu.Spec.Selector.Kubernetes
if k != nil {
r, err := k8sconv.NewKubernetesResource(monitor.lastKubernetesDiscovery, monitor.lastKubernetesApplyStatus)
if err != nil || r == nil {
return nil, fmt.Errorf("creating kube resource: %v", err)
}
return &luK8sResource{
selector: k,
res: r,
im: monitor.lastImageMap,
}, nil
}
dc := lu.Spec.Selector.DockerCompose
if dc != nil {
if monitor.lastDockerComposeService == nil {
return nil, fmt.Errorf("no docker compose status")
}
return &luDCResource{
selector: dc,
res: monitor.lastDockerComposeService,
}, nil
}
return nil, fmt.Errorf("No valid selector")
}
// Convert the currently tracked state into a set of inputs
// to the updater, then apply them.
func (r *Reconciler) maybeSync(ctx context.Context, lu *v1alpha1.LiveUpdate, monitor *monitor) v1alpha1.LiveUpdateStatus {
var status v1alpha1.LiveUpdateStatus
resource, err := r.resource(lu, monitor)
if err != nil {
status.Failed = createFailedState(lu, "Invalid", err.Error())
return status
}
manifestName := lu.Annotations[v1alpha1.AnnotationManifest]
updateMode := lu.Annotations[liveupdate.AnnotationUpdateMode]
inTriggerQueue := monitor.lastTriggerQueue != nil && manifestName != "" &&
configmap.InTriggerQueue(monitor.lastTriggerQueue, types.NamespacedName{Name: manifestName})
isUpdateModeManual := updateMode == liveupdate.UpdateModeManual
isWaitingOnTrigger := false
if isUpdateModeManual && !inTriggerQueue {
// In manual mode, we should always wait for a trigger before live updating anything.
isWaitingOnTrigger = true
}
r.garbageCollectFileChanges(resource, monitor)
r.garbageCollectMonitorContainers(resource, monitor)
// Go through all the container monitors, and check if any of them are unrecoverable.
// If they are, it's not important to figure out why.
resource.visitSelectedContainers(func(pod v1alpha1.Pod, c v1alpha1.Container) bool {
cKey := monitorContainerKey{
containerID: c.ID,
podName: pod.Name,
namespace: pod.Namespace,
}
cStatus, ok := monitor.containers[cKey]
if ok && cStatus.failedReason != "" {
status.Failed = createFailedState(lu, cStatus.failedReason, cStatus.failedMessage)
return true
}
return false
})
if status.Failed != nil {
return status
}
updateEventDispatched := false
// Visit all containers, apply changes, and return their statuses.
terminatedContainerPodName := ""
hasAnyFilesToSync := false
resource.visitSelectedContainers(func(pod v1alpha1.Pod, cInfo v1alpha1.Container) bool {
c := liveupdates.Container{
ContainerID: container.ID(cInfo.ID),
ContainerName: container.Name(cInfo.Name),
PodID: k8s.PodID(pod.Name),
Namespace: k8s.Namespace(pod.Namespace),
}
cKey := monitorContainerKey{
containerID: cInfo.ID,
podName: pod.Name,
namespace: pod.Namespace,
}
highWaterMark := r.startedTime
cStatus, ok := monitor.containers[cKey]
if ok && !cStatus.lastFileTimeSynced.IsZero() {
highWaterMark = cStatus.lastFileTimeSynced
}
// Determine the changed files.
filesChanged := []string{}
newHighWaterMark := highWaterMark
newLowWaterMark := metav1.MicroTime{}
for _, source := range monitor.sources {
for f, t := range source.modTimeByPath {
if t.After(highWaterMark.Time) {
filesChanged = append(filesChanged, f)
if newLowWaterMark.IsZero() || t.Before(&newLowWaterMark) {
newLowWaterMark = t
}
if t.After(newHighWaterMark.Time) {
newHighWaterMark = t
}
}
}
}
// Sort the files so that they're deterministic.
filesChanged = sliceutils.DedupedAndSorted(filesChanged)
if len(filesChanged) > 0 {
hasAnyFilesToSync = true
}
// Ignore completed pods/containers.
// This is a bit tricky to handle correctly, but is handled at
// the end of this function.
if pod.Phase == string(v1.PodSucceeded) || pod.Phase == string(v1.PodFailed) || cInfo.State.Terminated != nil {
if terminatedContainerPodName == "" {
terminatedContainerPodName = pod.Name
}
return false
}
var waiting *v1alpha1.LiveUpdateContainerStateWaiting
// We interpret "no container id" as a waiting state
// (terminated states should have been caught above).
if cInfo.State.Running == nil || cInfo.ID == "" {
waiting = &v1alpha1.LiveUpdateContainerStateWaiting{
Reason: "ContainerWaiting",
Message: "Waiting for container to start",
}
} else if isWaitingOnTrigger {
waiting = &v1alpha1.LiveUpdateContainerStateWaiting{
Reason: "Trigger",
Message: "Only updates on manual trigger",
}
}
// Create a plan to update the container.
filesApplied := false
var oneUpdateStatus v1alpha1.LiveUpdateStatus
plan, failed := r.createLiveUpdatePlan(lu.Spec, filesChanged)
if failed != nil {
// The plan told us to stop updating - this container is unrecoverable.
oneUpdateStatus.Failed = failed
} else if len(plan.SyncPaths) == 0 {
// The plan told us that there are no updates to do.
oneUpdateStatus.Containers = []v1alpha1.LiveUpdateContainerStatus{{
ContainerName: cInfo.Name,
ContainerID: cInfo.ID,
PodName: pod.Name,
Namespace: pod.Namespace,
LastFileTimeSynced: cStatus.lastFileTimeSynced,
Waiting: waiting,
}}
} else if cInfo.State.Waiting != nil && cInfo.State.Waiting.Reason == "CrashLoopBackOff" {
// At this point, the plan told us that we have some files to sync.
// Check if the container is in a state to receive those updates.
// If the container is crashlooping, that means it might not be up long enough
// to be able to receive a live-update. Treat this as an unrecoverable failure case.
oneUpdateStatus.Failed = createFailedState(lu, "CrashLoopBackOff",
fmt.Sprintf("Cannot live update because container crashing. Pod: %s", pod.Name))
} else if waiting != nil {
// Mark the container as waiting, so we have a record of it. No need to sync any files.
oneUpdateStatus.Containers = []v1alpha1.LiveUpdateContainerStatus{{
ContainerName: cInfo.Name,
ContainerID: cInfo.ID,
PodName: pod.Name,
Namespace: pod.Namespace,
LastFileTimeSynced: cStatus.lastFileTimeSynced,
Waiting: waiting,
}}
} else {
// Log progress and treat this as an update in the engine state.
if !updateEventDispatched {
updateEventDispatched = true
r.dispatchStartBuildAction(ctx, lu, filesChanged)
}
// Apply the change to the container.
oneUpdateStatus = r.applyInternal(ctx, lu.Spec, Input{
IsDC: lu.Spec.Selector.DockerCompose != nil,
ChangedFiles: plan.SyncPaths,
Containers: []liveupdates.Container{c},
LastFileTimeSynced: newHighWaterMark,
})
filesApplied = true
}
// Merge the status from the single update into the overall liveupdate status.
adjustFailedStateTimestamps(lu, &oneUpdateStatus)
// Update the monitor based on the result of the applied changes.
if oneUpdateStatus.Failed != nil {
cStatus.failedReason = oneUpdateStatus.Failed.Reason
cStatus.failedMessage = oneUpdateStatus.Failed.Message
cStatus.failedLowWaterMark = newLowWaterMark
} else if filesApplied {
cStatus.lastFileTimeSynced = newHighWaterMark
}
monitor.containers[cKey] = cStatus
// Update the status based on the result of the applied changes.
if oneUpdateStatus.Failed != nil {
status.Failed = oneUpdateStatus.Failed
status.Containers = nil
return true
}
status.Containers = append(status.Containers, oneUpdateStatus.Containers...)
return false
})
// If the only containers we're connected to are terminated containers,
// there are two cases we need to worry about:
//
// 1) The pod has completed, and will never run again (like a Job).
// 2) This is an old pod, and we're waiting for the new pod to rollout.
//
// We don't really have a great way to distinguish between these two cases.
//
// If we get to the end of this loop and haven't found any "live" pods,
// we assume we're in state (1) (to prevent waiting forever).
if status.Failed == nil && terminatedContainerPodName != "" &&
hasAnyFilesToSync && len(status.Containers) == 0 {
status.Failed = createFailedState(lu, "Terminated",
fmt.Sprintf("Container for live update is stopped. Pod name: %s", terminatedContainerPodName))
}
if updateEventDispatched {
r.dispatchCompleteBuildAction(lu, status)
}
return status
}
func (r *Reconciler) createLiveUpdatePlan(spec v1alpha1.LiveUpdateSpec, filesChanged []string) (liveupdates.LiveUpdatePlan, *v1alpha1.LiveUpdateStateFailed) {
plan, err := liveupdates.NewLiveUpdatePlan(spec, filesChanged)
if err != nil {
return plan, &v1alpha1.LiveUpdateStateFailed{
Reason: "UpdateStopped",
Message: fmt.Sprintf("No update plan: %v", err),
}
}
if len(plan.NoMatchPaths) > 0 {
return plan, &v1alpha1.LiveUpdateStateFailed{
Reason: "UpdateStopped",
Message: fmt.Sprintf("Found file(s) not matching any sync (files: %s)",
ospath.FormatFileChangeList(plan.NoMatchPaths)),
}
}
// If any changed files match a FallBackOn file, fall back to next BuildAndDeployer
if len(plan.StopPaths) != 0 {
return plan, &v1alpha1.LiveUpdateStateFailed{
Reason: "UpdateStopped",
Message: fmt.Sprintf("Detected change to stop file %q", plan.StopPaths[0]),
}
}
return plan, nil
}
// Generate the correct transition time on the Failed state.
func adjustFailedStateTimestamps(obj *v1alpha1.LiveUpdate, newStatus *v1alpha1.LiveUpdateStatus) {
if newStatus.Failed == nil {
return
}
newStatus.Failed = createFailedState(obj, newStatus.Failed.Reason, newStatus.Failed.Message)
}
// Create a new failed state and update the transition timestamp if appropriate.
func createFailedState(obj *v1alpha1.LiveUpdate, reason, msg string) *v1alpha1.LiveUpdateStateFailed {
failed := &v1alpha1.LiveUpdateStateFailed{Reason: reason, Message: msg}
transitionTime := apis.NowMicro()
if obj.Status.Failed != nil && obj.Status.Failed.Reason == failed.Reason {
// If the reason hasn't changed, don't treat this as a transition.
transitionTime = obj.Status.Failed.LastTransitionTime
}
failed.LastTransitionTime = transitionTime
return failed
}
// Like apply, but doesn't write the status to the apiserver.
func (r *Reconciler) applyInternal(
ctx context.Context,
spec v1alpha1.LiveUpdateSpec,
input Input) v1alpha1.LiveUpdateStatus {
var result v1alpha1.LiveUpdateStatus
cu := r.containerUpdater(input)
l := logger.Get(ctx)
containers := input.Containers
names := liveupdates.ContainerDisplayNames(containers)
suffix := ""
if len(containers) != 1 {
suffix = "(s)"
}
runSteps := liveupdate.RunSteps(spec)
changedFiles := input.ChangedFiles
hotReload := !liveupdate.ShouldRestart(spec)
boiledSteps, err := build.BoilRuns(runSteps, changedFiles)
if err != nil {
result.Failed = &v1alpha1.LiveUpdateStateFailed{
Reason: "Invalid",
Message: fmt.Sprintf("Building exec: %v", err),
}
return result
}
// rm files from container
toRemove, toArchive, err := build.MissingLocalPaths(ctx, changedFiles)
if err != nil {
result.Failed = &v1alpha1.LiveUpdateStateFailed{
Reason: "Invalid",
Message: fmt.Sprintf("Mapping paths: %v", err),
}
return result
}
if len(toRemove) > 0 {
l.Infof("Will delete %d file(s) from container%s: %s", len(toRemove), suffix, names)
for _, pm := range toRemove {
l.Infof("- '%s' (matched local path: '%s')", pm.ContainerPath, pm.LocalPath)
}
}
if len(toArchive) > 0 {
l.Infof("Will copy %d file(s) to container%s: %s", len(toArchive), suffix, names)
for _, pm := range toArchive {
l.Infof("- %s", pm.PrettyStr())
}
}
var lastExecErrorStatus *v1alpha1.LiveUpdateContainerStatus
for _, cInfo := range containers {
// TODO(nick): We should try to distinguish between cases where the tar writer
// fails (which is recoverable) vs when the server-side unpacking
// fails (which may not be recoverable).
archive := build.TarArchiveForPaths(ctx, toArchive, nil)
err = cu.UpdateContainer(ctx, cInfo, archive,
build.PathMappingsToContainerPaths(toRemove), boiledSteps, hotReload)
_ = archive.Close()
lastFileTimeSynced := input.LastFileTimeSynced
if lastFileTimeSynced.IsZero() {
lastFileTimeSynced = apis.NowMicro()
}
cStatus := v1alpha1.LiveUpdateContainerStatus{
ContainerName: cInfo.ContainerName.String(),
ContainerID: cInfo.ContainerID.String(),
PodName: cInfo.PodID.String(),
Namespace: string(cInfo.Namespace),
LastFileTimeSynced: lastFileTimeSynced,
}
if err != nil {
if build.IsRunStepFailure(err) {
// Keep running updates -- we want all containers to have the same files on them
// even if the Runs don't succeed
logger.Get(ctx).Infof(" → Failed to update container %s: %v",
cInfo.DisplayName(), err)
cStatus.LastExecError = err.Error()
lastExecErrorStatus = &cStatus
} else {
// Something went wrong with this update and it's NOT the user's fault--
// likely a infrastructure error. Bail, and fall back to full build.
msg := ""
if cStatus.PodName != "" {
msg = fmt.Sprintf("Updating pod %s: %v", cStatus.PodName, err)
} else {
msg = fmt.Sprintf("Updating container %s: %v", cInfo.DisplayName(), err)
}
result.Failed = &v1alpha1.LiveUpdateStateFailed{
Reason: "UpdateFailed",
Message: msg,
}
return result
}
} else {
logger.Get(ctx).Infof(" → Container %s updated!", cInfo.DisplayName())
if lastExecErrorStatus != nil {
// This build succeeded, but previously at least one failed due to user error.
// We may have inconsistent state--bail, and fall back to full build.
result.Failed = &v1alpha1.LiveUpdateStateFailed{
Reason: "PodsInconsistent",
Message: fmt.Sprintf("Pods in inconsistent state. Success: pod %s. Failure: pod %s. Error: %v",
cStatus.PodName, lastExecErrorStatus.PodName, lastExecErrorStatus.LastExecError),
}
return result
}
}
result.Containers = append(result.Containers, cStatus)
}
return result
}
func (r *Reconciler) containerUpdater(input Input) containerupdate.ContainerUpdater {
isDC := input.IsDC
if isDC || r.updateMode == liveupdates.UpdateModeContainer {