-
Notifications
You must be signed in to change notification settings - Fork 1.9k
/
util.go
862 lines (751 loc) · 25.6 KB
/
util.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
package scheduler
import (
"fmt"
"math/rand"
"reflect"
log "github.com/hashicorp/go-hclog"
memdb "github.com/hashicorp/go-memdb"
"github.com/hashicorp/nomad/nomad/structs"
)
// allocTuple is a tuple of the allocation name and potential alloc ID
type allocTuple struct {
Name string
TaskGroup *structs.TaskGroup
Alloc *structs.Allocation
}
// materializeTaskGroups is used to materialize all the task groups
// a job requires. This is used to do the count expansion.
func materializeTaskGroups(job *structs.Job) map[string]*structs.TaskGroup {
out := make(map[string]*structs.TaskGroup)
if job.Stopped() {
return out
}
for _, tg := range job.TaskGroups {
for i := 0; i < tg.Count; i++ {
name := fmt.Sprintf("%s.%s[%d]", job.Name, tg.Name, i)
out[name] = tg
}
}
return out
}
// diffResult is used to return the sets that result from the diff
type diffResult struct {
place, update, migrate, stop, ignore, lost []allocTuple
}
func (d *diffResult) GoString() string {
return fmt.Sprintf("allocs: (place %d) (update %d) (migrate %d) (stop %d) (ignore %d) (lost %d)",
len(d.place), len(d.update), len(d.migrate), len(d.stop), len(d.ignore), len(d.lost))
}
func (d *diffResult) Append(other *diffResult) {
d.place = append(d.place, other.place...)
d.update = append(d.update, other.update...)
d.migrate = append(d.migrate, other.migrate...)
d.stop = append(d.stop, other.stop...)
d.ignore = append(d.ignore, other.ignore...)
d.lost = append(d.lost, other.lost...)
}
// diffAllocs is used to do a set difference between the target allocations
// and the existing allocations. This returns 6 sets of results, the list of
// named task groups that need to be placed (no existing allocation), the
// allocations that need to be updated (job definition is newer), allocs that
// need to be migrated (node is draining), the allocs that need to be evicted
// (no longer required), those that should be ignored and those that are lost
// that need to be replaced (running on a lost node).
//
// job is the job whose allocs is going to be diff-ed.
// taintedNodes is an index of the nodes which are either down or in drain mode
// by name.
// required is a set of allocations that must exist.
// allocs is a list of non terminal allocations.
// terminalAllocs is an index of the latest terminal allocations by name.
func diffAllocs(job *structs.Job, taintedNodes map[string]*structs.Node,
required map[string]*structs.TaskGroup, allocs []*structs.Allocation,
terminalAllocs map[string]*structs.Allocation) *diffResult {
result := &diffResult{}
// Scan the existing updates
existing := make(map[string]struct{})
for _, exist := range allocs {
// Index the existing node
name := exist.Name
existing[name] = struct{}{}
// Check for the definition in the required set
tg, ok := required[name]
// If not required, we stop the alloc
if !ok {
result.stop = append(result.stop, allocTuple{
Name: name,
TaskGroup: tg,
Alloc: exist,
})
continue
}
// If we have been marked for migration and aren't terminal, migrate
if !exist.TerminalStatus() && exist.DesiredTransition.ShouldMigrate() {
result.migrate = append(result.migrate, allocTuple{
Name: name,
TaskGroup: tg,
Alloc: exist,
})
continue
}
// If we are on a tainted node, we must migrate if we are a service or
// if the batch allocation did not finish
if node, ok := taintedNodes[exist.NodeID]; ok {
// If the job is batch and finished successfully, the fact that the
// node is tainted does not mean it should be migrated or marked as
// lost as the work was already successfully finished. However for
// service/system jobs, tasks should never complete. The check of
// batch type, defends against client bugs.
if exist.Job.Type == structs.JobTypeBatch && exist.RanSuccessfully() {
goto IGNORE
}
if !exist.TerminalStatus() && (node == nil || node.TerminalStatus()) {
result.lost = append(result.lost, allocTuple{
Name: name,
TaskGroup: tg,
Alloc: exist,
})
} else {
goto IGNORE
}
continue
}
// If the definition is updated we need to update
if job.JobModifyIndex != exist.Job.JobModifyIndex {
result.update = append(result.update, allocTuple{
Name: name,
TaskGroup: tg,
Alloc: exist,
})
continue
}
// Everything is up-to-date
IGNORE:
result.ignore = append(result.ignore, allocTuple{
Name: name,
TaskGroup: tg,
Alloc: exist,
})
}
// Scan the required groups
for name, tg := range required {
// Check for an existing allocation
_, ok := existing[name]
// Require a placement if no existing allocation. If there
// is an existing allocation, we would have checked for a potential
// update or ignore above.
if !ok {
result.place = append(result.place, allocTuple{
Name: name,
TaskGroup: tg,
Alloc: terminalAllocs[name],
})
}
}
return result
}
// diffSystemAllocs is like diffAllocs however, the allocations in the
// diffResult contain the specific nodeID they should be allocated on.
//
// job is the job whose allocs is going to be diff-ed.
// nodes is a list of nodes in ready state.
// taintedNodes is an index of the nodes which are either down or in drain mode
// by name.
// allocs is a list of non terminal allocations.
// terminalAllocs is an index of the latest terminal allocations by name.
func diffSystemAllocs(job *structs.Job, nodes []*structs.Node, taintedNodes map[string]*structs.Node,
allocs []*structs.Allocation, terminalAllocs map[string]*structs.Allocation) *diffResult {
// Build a mapping of nodes to all their allocs.
nodeAllocs := make(map[string][]*structs.Allocation, len(allocs))
for _, alloc := range allocs {
nallocs := append(nodeAllocs[alloc.NodeID], alloc)
nodeAllocs[alloc.NodeID] = nallocs
}
for _, node := range nodes {
if _, ok := nodeAllocs[node.ID]; !ok {
nodeAllocs[node.ID] = nil
}
}
// Create the required task groups.
required := materializeTaskGroups(job)
result := &diffResult{}
for nodeID, allocs := range nodeAllocs {
diff := diffAllocs(job, taintedNodes, required, allocs, terminalAllocs)
// If the node is tainted there should be no placements made
if _, ok := taintedNodes[nodeID]; ok {
diff.place = nil
} else {
// Mark the alloc as being for a specific node.
for i := range diff.place {
alloc := &diff.place[i]
// If the new allocation isn't annotated with a previous allocation
// or if the previous allocation isn't from the same node then we
// annotate the allocTuple with a new Allocation
if alloc.Alloc == nil || alloc.Alloc.NodeID != nodeID {
alloc.Alloc = &structs.Allocation{NodeID: nodeID}
}
}
}
result.Append(diff)
}
return result
}
// readyNodesInDCs returns all the ready nodes in the given datacenters and a
// mapping of each data center to the count of ready nodes.
func readyNodesInDCs(state State, dcs []string) ([]*structs.Node, map[string]int, error) {
// Index the DCs
dcMap := make(map[string]int, len(dcs))
for _, dc := range dcs {
dcMap[dc] = 0
}
// Scan the nodes
ws := memdb.NewWatchSet()
var out []*structs.Node
iter, err := state.Nodes(ws)
if err != nil {
return nil, nil, err
}
for {
raw := iter.Next()
if raw == nil {
break
}
// Filter on datacenter and status
node := raw.(*structs.Node)
if node.Status != structs.NodeStatusReady {
continue
}
if node.Drain {
continue
}
if node.SchedulingEligibility != structs.NodeSchedulingEligible {
continue
}
if _, ok := dcMap[node.Datacenter]; !ok {
continue
}
out = append(out, node)
dcMap[node.Datacenter]++
}
return out, dcMap, nil
}
// retryMax is used to retry a callback until it returns success or
// a maximum number of attempts is reached. An optional reset function may be
// passed which is called after each failed iteration. If the reset function is
// set and returns true, the number of attempts is reset back to max.
func retryMax(max int, cb func() (bool, error), reset func() bool) error {
attempts := 0
for attempts < max {
done, err := cb()
if err != nil {
return err
}
if done {
return nil
}
// Check if we should reset the number attempts
if reset != nil && reset() {
attempts = 0
} else {
attempts++
}
}
return &SetStatusError{
Err: fmt.Errorf("maximum attempts reached (%d)", max),
EvalStatus: structs.EvalStatusFailed,
}
}
// progressMade checks to see if the plan result made allocations or updates.
// If the result is nil, false is returned.
func progressMade(result *structs.PlanResult) bool {
return result != nil && (len(result.NodeUpdate) != 0 ||
len(result.NodeAllocation) != 0 || result.Deployment != nil ||
len(result.DeploymentUpdates) != 0)
}
// taintedNodes is used to scan the allocations and then check if the
// underlying nodes are tainted, and should force a migration of the allocation.
// All the nodes returned in the map are tainted.
func taintedNodes(state State, allocs []*structs.Allocation) (map[string]*structs.Node, error) {
out := make(map[string]*structs.Node)
for _, alloc := range allocs {
if _, ok := out[alloc.NodeID]; ok {
continue
}
ws := memdb.NewWatchSet()
node, err := state.NodeByID(ws, alloc.NodeID)
if err != nil {
return nil, err
}
// If the node does not exist, we should migrate
if node == nil {
out[alloc.NodeID] = nil
continue
}
if structs.ShouldDrainNode(node.Status) || node.Drain {
out[alloc.NodeID] = node
}
}
return out, nil
}
// shuffleNodes randomizes the slice order with the Fisher-Yates algorithm
func shuffleNodes(nodes []*structs.Node) {
n := len(nodes)
for i := n - 1; i > 0; i-- {
j := rand.Intn(i + 1)
nodes[i], nodes[j] = nodes[j], nodes[i]
}
}
// tasksUpdated does a diff between task groups to see if the
// tasks, their drivers, environment variables or config have updated. The
// inputs are the task group name to diff and two jobs to diff.
func tasksUpdated(jobA, jobB *structs.Job, taskGroup string) bool {
a := jobA.LookupTaskGroup(taskGroup)
b := jobB.LookupTaskGroup(taskGroup)
// If the number of tasks do not match, clearly there is an update
if len(a.Tasks) != len(b.Tasks) {
return true
}
// Check ephemeral disk
if !reflect.DeepEqual(a.EphemeralDisk, b.EphemeralDisk) {
return true
}
// Check that the network resources haven't changed
if networkUpdated(a.Networks, b.Networks) {
return true
}
// Check each task
for _, at := range a.Tasks {
bt := b.LookupTask(at.Name)
if bt == nil {
return true
}
if at.Driver != bt.Driver {
return true
}
if at.User != bt.User {
return true
}
if !reflect.DeepEqual(at.Config, bt.Config) {
return true
}
if !reflect.DeepEqual(at.Env, bt.Env) {
return true
}
if !reflect.DeepEqual(at.Artifacts, bt.Artifacts) {
return true
}
if !reflect.DeepEqual(at.Vault, bt.Vault) {
return true
}
if !reflect.DeepEqual(at.Templates, bt.Templates) {
return true
}
// Check the metadata
if !reflect.DeepEqual(
jobA.CombinedTaskMeta(taskGroup, at.Name),
jobB.CombinedTaskMeta(taskGroup, bt.Name)) {
return true
}
// Inspect the network to see if the dynamic ports are different
if networkUpdated(at.Resources.Networks, bt.Resources.Networks) {
return true
}
// Inspect the non-network resources
if ar, br := at.Resources, bt.Resources; ar.CPU != br.CPU {
return true
} else if ar.MemoryMB != br.MemoryMB {
return true
}
}
return false
}
func networkUpdated(netA, netB []*structs.NetworkResource) bool {
if len(netA) != len(netB) {
return true
}
for idx := range netA {
an := netA[idx]
bn := netB[idx]
if an.MBits != bn.MBits {
return true
}
aPorts, bPorts := networkPortMap(an), networkPortMap(bn)
if !reflect.DeepEqual(aPorts, bPorts) {
return true
}
}
return false
}
// networkPortMap takes a network resource and returns a map of port labels to
// values. The value for dynamic ports is disregarded even if it is set. This
// makes this function suitable for comparing two network resources for changes.
func networkPortMap(n *structs.NetworkResource) map[string]int {
m := make(map[string]int, len(n.DynamicPorts)+len(n.ReservedPorts))
for _, p := range n.ReservedPorts {
m[p.Label] = p.Value
}
for _, p := range n.DynamicPorts {
m[p.Label] = -1
}
return m
}
// setStatus is used to update the status of the evaluation
func setStatus(logger log.Logger, planner Planner,
eval, nextEval, spawnedBlocked *structs.Evaluation,
tgMetrics map[string]*structs.AllocMetric, status, desc string,
queuedAllocs map[string]int, deploymentID string) error {
logger.Debug("setting eval status", "status", status)
newEval := eval.Copy()
newEval.Status = status
newEval.StatusDescription = desc
newEval.DeploymentID = deploymentID
newEval.FailedTGAllocs = tgMetrics
if nextEval != nil {
newEval.NextEval = nextEval.ID
}
if spawnedBlocked != nil {
newEval.BlockedEval = spawnedBlocked.ID
}
if queuedAllocs != nil {
newEval.QueuedAllocations = queuedAllocs
}
return planner.UpdateEval(newEval)
}
// inplaceUpdate attempts to update allocations in-place where possible. It
// returns the allocs that couldn't be done inplace and then those that could.
func inplaceUpdate(ctx Context, eval *structs.Evaluation, job *structs.Job,
stack Stack, updates []allocTuple) (destructive, inplace []allocTuple) {
// doInplace manipulates the updates map to make the current allocation
// an inplace update.
doInplace := func(cur, last, inplaceCount *int) {
updates[*cur], updates[*last-1] = updates[*last-1], updates[*cur]
*cur--
*last--
*inplaceCount++
}
ws := memdb.NewWatchSet()
n := len(updates)
inplaceCount := 0
for i := 0; i < n; i++ {
// Get the update
update := updates[i]
// Check if the task drivers or config has changed, requires
// a rolling upgrade since that cannot be done in-place.
existing := update.Alloc.Job
if tasksUpdated(job, existing, update.TaskGroup.Name) {
continue
}
// Terminal batch allocations are not filtered when they are completed
// successfully. We should avoid adding the allocation to the plan in
// the case that it is an in-place update to avoid both additional data
// in the plan and work for the clients.
if update.Alloc.TerminalStatus() {
doInplace(&i, &n, &inplaceCount)
continue
}
// Get the existing node
node, err := ctx.State().NodeByID(ws, update.Alloc.NodeID)
if err != nil {
ctx.Logger().Error("failed to get node", "node_id", update.Alloc.NodeID, "error", err)
continue
}
if node == nil {
continue
}
// Set the existing node as the base set
stack.SetNodes([]*structs.Node{node})
// Stage an eviction of the current allocation. This is done so that
// the current allocation is discounted when checking for feasibility.
// Otherwise we would be trying to fit the tasks current resources and
// updated resources. After select is called we can remove the evict.
ctx.Plan().AppendStoppedAlloc(update.Alloc, allocInPlace, "")
// Attempt to match the task group
option := stack.Select(update.TaskGroup, nil) // This select only looks at one node so we don't pass selectOptions
// Pop the allocation
ctx.Plan().PopUpdate(update.Alloc)
// Skip if we could not do an in-place update
if option == nil {
continue
}
// Restore the network offers from the existing allocation.
// We do not allow network resources (reserved/dynamic ports)
// to be updated. This is guarded in taskUpdated, so we can
// safely restore those here.
for task, resources := range option.TaskResources {
var networks structs.Networks
if update.Alloc.AllocatedResources != nil {
if tr, ok := update.Alloc.AllocatedResources.Tasks[task]; ok {
networks = tr.Networks
}
} else if tr, ok := update.Alloc.TaskResources[task]; ok {
networks = tr.Networks
}
// Add thhe networks back
resources.Networks = networks
}
// Create a shallow copy
newAlloc := new(structs.Allocation)
*newAlloc = *update.Alloc
// Update the allocation
newAlloc.EvalID = eval.ID
newAlloc.Job = nil // Use the Job in the Plan
newAlloc.Resources = nil // Computed in Plan Apply
newAlloc.AllocatedResources = &structs.AllocatedResources{
Tasks: option.TaskResources,
Shared: structs.AllocatedSharedResources{
DiskMB: int64(update.TaskGroup.EphemeralDisk.SizeMB),
},
}
newAlloc.Metrics = ctx.Metrics()
ctx.Plan().AppendAlloc(newAlloc)
// Remove this allocation from the slice
doInplace(&i, &n, &inplaceCount)
}
if len(updates) > 0 {
ctx.Logger().Debug("made in-place updates", "in-place", inplaceCount, "total_updates", len(updates))
}
return updates[:n], updates[n:]
}
// evictAndPlace is used to mark allocations for evicts and add them to the
// placement queue. evictAndPlace modifies both the diffResult and the
// limit. It returns true if the limit has been reached.
func evictAndPlace(ctx Context, diff *diffResult, allocs []allocTuple, desc string, limit *int) bool {
n := len(allocs)
for i := 0; i < n && i < *limit; i++ {
a := allocs[i]
ctx.Plan().AppendStoppedAlloc(a.Alloc, desc, "")
diff.place = append(diff.place, a)
}
if n <= *limit {
*limit -= n
return false
}
*limit = 0
return true
}
// tgConstrainTuple is used to store the total constraints of a task group.
type tgConstrainTuple struct {
// Holds the combined constraints of the task group and all it's sub-tasks.
constraints []*structs.Constraint
// The set of required drivers within the task group.
drivers map[string]struct{}
}
// taskGroupConstraints collects the constraints, drivers and resources required by each
// sub-task to aggregate the TaskGroup totals
func taskGroupConstraints(tg *structs.TaskGroup) tgConstrainTuple {
c := tgConstrainTuple{
constraints: make([]*structs.Constraint, 0, len(tg.Constraints)),
drivers: make(map[string]struct{}),
}
c.constraints = append(c.constraints, tg.Constraints...)
for _, task := range tg.Tasks {
c.drivers[task.Driver] = struct{}{}
c.constraints = append(c.constraints, task.Constraints...)
}
return c
}
// desiredUpdates takes the diffResult as well as the set of inplace and
// destructive updates and returns a map of task groups to their set of desired
// updates.
func desiredUpdates(diff *diffResult, inplaceUpdates,
destructiveUpdates []allocTuple) map[string]*structs.DesiredUpdates {
desiredTgs := make(map[string]*structs.DesiredUpdates)
for _, tuple := range diff.place {
name := tuple.TaskGroup.Name
des, ok := desiredTgs[name]
if !ok {
des = &structs.DesiredUpdates{}
desiredTgs[name] = des
}
des.Place++
}
for _, tuple := range diff.stop {
name := tuple.Alloc.TaskGroup
des, ok := desiredTgs[name]
if !ok {
des = &structs.DesiredUpdates{}
desiredTgs[name] = des
}
des.Stop++
}
for _, tuple := range diff.ignore {
name := tuple.TaskGroup.Name
des, ok := desiredTgs[name]
if !ok {
des = &structs.DesiredUpdates{}
desiredTgs[name] = des
}
des.Ignore++
}
for _, tuple := range diff.migrate {
name := tuple.TaskGroup.Name
des, ok := desiredTgs[name]
if !ok {
des = &structs.DesiredUpdates{}
desiredTgs[name] = des
}
des.Migrate++
}
for _, tuple := range inplaceUpdates {
name := tuple.TaskGroup.Name
des, ok := desiredTgs[name]
if !ok {
des = &structs.DesiredUpdates{}
desiredTgs[name] = des
}
des.InPlaceUpdate++
}
for _, tuple := range destructiveUpdates {
name := tuple.TaskGroup.Name
des, ok := desiredTgs[name]
if !ok {
des = &structs.DesiredUpdates{}
desiredTgs[name] = des
}
des.DestructiveUpdate++
}
return desiredTgs
}
// adjustQueuedAllocations decrements the number of allocations pending per task
// group based on the number of allocations successfully placed
func adjustQueuedAllocations(logger log.Logger, result *structs.PlanResult, queuedAllocs map[string]int) {
if result == nil {
return
}
for _, allocations := range result.NodeAllocation {
for _, allocation := range allocations {
// Ensure that the allocation is newly created. We check that
// the CreateIndex is equal to the ModifyIndex in order to check
// that the allocation was just created. We do not check that
// the CreateIndex is equal to the results AllocIndex because
// the allocations we get back have gone through the planner's
// optimistic snapshot and thus their indexes may not be
// correct, but they will be consistent.
if allocation.CreateIndex != allocation.ModifyIndex {
continue
}
if _, ok := queuedAllocs[allocation.TaskGroup]; ok {
queuedAllocs[allocation.TaskGroup]--
} else {
logger.Error("allocation placed but task group is not in list of unplaced allocations", "task_group", allocation.TaskGroup)
}
}
}
}
// updateNonTerminalAllocsToLost updates the allocations which are in pending/running state on tainted node
// to lost
func updateNonTerminalAllocsToLost(plan *structs.Plan, tainted map[string]*structs.Node, allocs []*structs.Allocation) {
for _, alloc := range allocs {
node, ok := tainted[alloc.NodeID]
if !ok {
continue
}
// Only handle down nodes or nodes that are gone (node == nil)
if node != nil && node.Status != structs.NodeStatusDown {
continue
}
// If the scheduler has marked it as stop already but the alloc wasn't
// terminal on the client change the status to lost.
if alloc.DesiredStatus == structs.AllocDesiredStatusStop &&
(alloc.ClientStatus == structs.AllocClientStatusRunning ||
alloc.ClientStatus == structs.AllocClientStatusPending) {
plan.AppendStoppedAlloc(alloc, allocLost, structs.AllocClientStatusLost)
}
}
}
// genericAllocUpdateFn is a factory for the scheduler to create an allocUpdateType
// function to be passed into the reconciler. The factory takes objects that
// exist only in the scheduler context and returns a function that can be used
// by the reconciler to make decisions about how to update an allocation. The
// factory allows the reconciler to be unaware of how to determine the type of
// update necessary and can minimize the set of objects it is exposed to.
func genericAllocUpdateFn(ctx Context, stack Stack, evalID string) allocUpdateType {
return func(existing *structs.Allocation, newJob *structs.Job, newTG *structs.TaskGroup) (ignore, destructive bool, updated *structs.Allocation) {
// Same index, so nothing to do
if existing.Job.JobModifyIndex == newJob.JobModifyIndex {
return true, false, nil
}
// Check if the task drivers or config has changed, requires
// a destructive upgrade since that cannot be done in-place.
if tasksUpdated(newJob, existing.Job, newTG.Name) {
return false, true, nil
}
// Terminal batch allocations are not filtered when they are completed
// successfully. We should avoid adding the allocation to the plan in
// the case that it is an in-place update to avoid both additional data
// in the plan and work for the clients.
if existing.TerminalStatus() {
return true, false, nil
}
// Get the existing node
ws := memdb.NewWatchSet()
node, err := ctx.State().NodeByID(ws, existing.NodeID)
if err != nil {
ctx.Logger().Error("failed to get node", "node_id", existing.NodeID, "error", err)
return true, false, nil
}
if node == nil {
return false, true, nil
}
// Set the existing node as the base set
stack.SetNodes([]*structs.Node{node})
// Stage an eviction of the current allocation. This is done so that
// the current allocation is discounted when checking for feasibility.
// Otherwise we would be trying to fit the tasks current resources and
// updated resources. After select is called we can remove the evict.
ctx.Plan().AppendStoppedAlloc(existing, allocInPlace, "")
// Attempt to match the task group
option := stack.Select(newTG, nil) // This select only looks at one node so we don't pass selectOptions
// Pop the allocation
ctx.Plan().PopUpdate(existing)
// Require destructive if we could not do an in-place update
if option == nil {
return false, true, nil
}
// Restore the network offers from the existing allocation.
// We do not allow network resources (reserved/dynamic ports)
// to be updated. This is guarded in taskUpdated, so we can
// safely restore those here.
for task, resources := range option.TaskResources {
var networks structs.Networks
if existing.AllocatedResources != nil {
if tr, ok := existing.AllocatedResources.Tasks[task]; ok {
networks = tr.Networks
}
} else if tr, ok := existing.TaskResources[task]; ok {
networks = tr.Networks
}
// Add the networks back
resources.Networks = networks
}
// Create a shallow copy
newAlloc := new(structs.Allocation)
*newAlloc = *existing
// Update the allocation
newAlloc.EvalID = evalID
newAlloc.Job = nil // Use the Job in the Plan
newAlloc.Resources = nil // Computed in Plan Apply
newAlloc.AllocatedResources = &structs.AllocatedResources{
Tasks: option.TaskResources,
Shared: structs.AllocatedSharedResources{
DiskMB: int64(newTG.EphemeralDisk.SizeMB),
},
}
// Since this is an inplace update, we should copy network
// information from the original alloc. This is similar to how
// we copy network info for task level networks above.
//
// existing.AllocatedResources is nil on Allocations created by
// Nomad v0.8 or earlier.
if existing.AllocatedResources != nil {
newAlloc.AllocatedResources.Shared.Networks = existing.AllocatedResources.Shared.Networks
}
// Use metrics from existing alloc for in place upgrade
// This is because if the inplace upgrade succeeded, any scoring metadata from
// when it first went through the scheduler should still be preserved. Using scoring
// metadata from the context would incorrectly replace it with metadata only from a single node that the
// allocation is already on.
newAlloc.Metrics = existing.Metrics.Copy()
return false, false, newAlloc
}
}