-
Notifications
You must be signed in to change notification settings - Fork 39.2k
/
eventhandlers.go
532 lines (486 loc) · 19.1 KB
/
eventhandlers.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
/*
Copyright 2019 The Kubernetes Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package scheduler
import (
"fmt"
"reflect"
"strings"
"k8s.io/apimachinery/pkg/runtime/schema"
"k8s.io/client-go/dynamic/dynamicinformer"
"k8s.io/klog/v2"
v1 "k8s.io/api/core/v1"
storagev1 "k8s.io/api/storage/v1"
utilruntime "k8s.io/apimachinery/pkg/util/runtime"
"k8s.io/apiserver/pkg/util/feature"
"k8s.io/client-go/informers"
"k8s.io/client-go/tools/cache"
v1helper "k8s.io/component-helpers/scheduling/corev1"
"k8s.io/component-helpers/scheduling/corev1/nodeaffinity"
"k8s.io/kubernetes/pkg/features"
"k8s.io/kubernetes/pkg/scheduler/framework"
"k8s.io/kubernetes/pkg/scheduler/framework/plugins/nodename"
"k8s.io/kubernetes/pkg/scheduler/framework/plugins/nodeports"
"k8s.io/kubernetes/pkg/scheduler/framework/plugins/noderesources"
"k8s.io/kubernetes/pkg/scheduler/internal/queue"
"k8s.io/kubernetes/pkg/scheduler/profile"
)
func (sched *Scheduler) onStorageClassAdd(obj interface{}) {
sc, ok := obj.(*storagev1.StorageClass)
if !ok {
klog.ErrorS(nil, "Cannot convert to *storagev1.StorageClass", "obj", obj)
return
}
// CheckVolumeBindingPred fails if pod has unbound immediate PVCs. If these
// PVCs have specified StorageClass name, creating StorageClass objects
// with late binding will cause predicates to pass, so we need to move pods
// to active queue.
// We don't need to invalidate cached results because results will not be
// cached for pod that has unbound immediate PVCs.
if sc.VolumeBindingMode != nil && *sc.VolumeBindingMode == storagev1.VolumeBindingWaitForFirstConsumer {
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(queue.StorageClassAdd, nil)
}
}
func (sched *Scheduler) addNodeToCache(obj interface{}) {
node, ok := obj.(*v1.Node)
if !ok {
klog.ErrorS(nil, "Cannot convert to *v1.Node", "obj", obj)
return
}
nodeInfo := sched.SchedulerCache.AddNode(node)
klog.V(3).InfoS("Add event for node", "node", klog.KObj(node))
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(queue.NodeAdd, preCheckForNode(nodeInfo))
}
func (sched *Scheduler) updateNodeInCache(oldObj, newObj interface{}) {
oldNode, ok := oldObj.(*v1.Node)
if !ok {
klog.ErrorS(nil, "Cannot convert oldObj to *v1.Node", "oldObj", oldObj)
return
}
newNode, ok := newObj.(*v1.Node)
if !ok {
klog.ErrorS(nil, "Cannot convert newObj to *v1.Node", "newObj", newObj)
return
}
nodeInfo := sched.SchedulerCache.UpdateNode(oldNode, newNode)
// Only requeue unschedulable pods if the node became more schedulable.
if event := nodeSchedulingPropertiesChange(newNode, oldNode); event != nil {
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(*event, preCheckForNode(nodeInfo))
}
}
func (sched *Scheduler) deleteNodeFromCache(obj interface{}) {
var node *v1.Node
switch t := obj.(type) {
case *v1.Node:
node = t
case cache.DeletedFinalStateUnknown:
var ok bool
node, ok = t.Obj.(*v1.Node)
if !ok {
klog.ErrorS(nil, "Cannot convert to *v1.Node", "obj", t.Obj)
return
}
default:
klog.ErrorS(nil, "Cannot convert to *v1.Node", "obj", t)
return
}
klog.V(3).InfoS("Delete event for node", "node", klog.KObj(node))
// NOTE: Updates must be written to scheduler cache before invalidating
// equivalence cache, because we could snapshot equivalence cache after the
// invalidation and then snapshot the cache itself. If the cache is
// snapshotted before updates are written, we would update equivalence
// cache with stale information which is based on snapshot of old cache.
if err := sched.SchedulerCache.RemoveNode(node); err != nil {
klog.ErrorS(err, "Scheduler cache RemoveNode failed")
}
}
func (sched *Scheduler) addPodToSchedulingQueue(obj interface{}) {
pod := obj.(*v1.Pod)
klog.V(3).InfoS("Add event for unscheduled pod", "pod", klog.KObj(pod))
if err := sched.SchedulingQueue.Add(pod); err != nil {
utilruntime.HandleError(fmt.Errorf("unable to queue %T: %v", obj, err))
}
}
func (sched *Scheduler) updatePodInSchedulingQueue(oldObj, newObj interface{}) {
oldPod, newPod := oldObj.(*v1.Pod), newObj.(*v1.Pod)
// Bypass update event that carries identical objects; otherwise, a duplicated
// Pod may go through scheduling and cause unexpected behavior (see #96071).
if oldPod.ResourceVersion == newPod.ResourceVersion {
return
}
isAssumed, err := sched.SchedulerCache.IsAssumedPod(newPod)
if err != nil {
utilruntime.HandleError(fmt.Errorf("failed to check whether pod %s/%s is assumed: %v", newPod.Namespace, newPod.Name, err))
}
if isAssumed {
return
}
if err := sched.SchedulingQueue.Update(oldPod, newPod); err != nil {
utilruntime.HandleError(fmt.Errorf("unable to update %T: %v", newObj, err))
}
}
func (sched *Scheduler) deletePodFromSchedulingQueue(obj interface{}) {
var pod *v1.Pod
switch t := obj.(type) {
case *v1.Pod:
pod = obj.(*v1.Pod)
case cache.DeletedFinalStateUnknown:
var ok bool
pod, ok = t.Obj.(*v1.Pod)
if !ok {
utilruntime.HandleError(fmt.Errorf("unable to convert object %T to *v1.Pod in %T", obj, sched))
return
}
default:
utilruntime.HandleError(fmt.Errorf("unable to handle object in %T: %T", sched, obj))
return
}
klog.V(3).InfoS("Delete event for unscheduled pod", "pod", klog.KObj(pod))
if err := sched.SchedulingQueue.Delete(pod); err != nil {
utilruntime.HandleError(fmt.Errorf("unable to dequeue %T: %v", obj, err))
}
fwk, err := sched.frameworkForPod(pod)
if err != nil {
// This shouldn't happen, because we only accept for scheduling the pods
// which specify a scheduler name that matches one of the profiles.
klog.ErrorS(err, "Unable to get profile", "pod", klog.KObj(pod))
return
}
// If a waiting pod is rejected, it indicates it's previously assumed and we're
// removing it from the scheduler cache. In this case, signal a AssignedPodDelete
// event to immediately retry some unscheduled Pods.
if fwk.RejectWaitingPod(pod.UID) {
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(queue.AssignedPodDelete, nil)
}
}
func (sched *Scheduler) addPodToCache(obj interface{}) {
pod, ok := obj.(*v1.Pod)
if !ok {
klog.ErrorS(nil, "Cannot convert to *v1.Pod", "obj", obj)
return
}
klog.V(3).InfoS("Add event for scheduled pod", "pod", klog.KObj(pod))
if err := sched.SchedulerCache.AddPod(pod); err != nil {
klog.ErrorS(err, "Scheduler cache AddPod failed", "pod", klog.KObj(pod))
}
sched.SchedulingQueue.AssignedPodAdded(pod)
}
func (sched *Scheduler) updatePodInCache(oldObj, newObj interface{}) {
oldPod, ok := oldObj.(*v1.Pod)
if !ok {
klog.ErrorS(nil, "Cannot convert oldObj to *v1.Pod", "oldObj", oldObj)
return
}
newPod, ok := newObj.(*v1.Pod)
if !ok {
klog.ErrorS(nil, "Cannot convert newObj to *v1.Pod", "newObj", newObj)
return
}
// A Pod delete event followed by an immediate Pod add event may be merged
// into a Pod update event. In this case, we should invalidate the old Pod, and
// then add the new Pod.
if oldPod.UID != newPod.UID {
sched.deletePodFromCache(oldObj)
sched.addPodToCache(newObj)
return
}
// NOTE: Updates must be written to scheduler cache before invalidating
// equivalence cache, because we could snapshot equivalence cache after the
// invalidation and then snapshot the cache itself. If the cache is
// snapshotted before updates are written, we would update equivalence
// cache with stale information which is based on snapshot of old cache.
if err := sched.SchedulerCache.UpdatePod(oldPod, newPod); err != nil {
klog.ErrorS(err, "Scheduler cache UpdatePod failed", "oldPod", klog.KObj(oldPod), "newPod", klog.KObj(newPod))
}
sched.SchedulingQueue.AssignedPodUpdated(newPod)
}
func (sched *Scheduler) deletePodFromCache(obj interface{}) {
var pod *v1.Pod
switch t := obj.(type) {
case *v1.Pod:
pod = t
case cache.DeletedFinalStateUnknown:
var ok bool
pod, ok = t.Obj.(*v1.Pod)
if !ok {
klog.ErrorS(nil, "Cannot convert to *v1.Pod", "obj", t.Obj)
return
}
default:
klog.ErrorS(nil, "Cannot convert to *v1.Pod", "obj", t)
return
}
klog.V(3).InfoS("Delete event for scheduled pod", "pod", klog.KObj(pod))
// NOTE: Updates must be written to scheduler cache before invalidating
// equivalence cache, because we could snapshot equivalence cache after the
// invalidation and then snapshot the cache itself. If the cache is
// snapshotted before updates are written, we would update equivalence
// cache with stale information which is based on snapshot of old cache.
if err := sched.SchedulerCache.RemovePod(pod); err != nil {
klog.ErrorS(err, "Scheduler cache RemovePod failed", "pod", klog.KObj(pod))
}
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(queue.AssignedPodDelete, nil)
}
// assignedPod selects pods that are assigned (scheduled and running).
func assignedPod(pod *v1.Pod) bool {
return len(pod.Spec.NodeName) != 0
}
// responsibleForPod returns true if the pod has asked to be scheduled by the given scheduler.
func responsibleForPod(pod *v1.Pod, profiles profile.Map) bool {
return profiles.HandlesSchedulerName(pod.Spec.SchedulerName)
}
// addAllEventHandlers is a helper function used in tests and in Scheduler
// to add event handlers for various informers.
func addAllEventHandlers(
sched *Scheduler,
informerFactory informers.SharedInformerFactory,
dynInformerFactory dynamicinformer.DynamicSharedInformerFactory,
gvkMap map[framework.GVK]framework.ActionType,
) {
// scheduled pod cache
informerFactory.Core().V1().Pods().Informer().AddEventHandler(
cache.FilteringResourceEventHandler{
FilterFunc: func(obj interface{}) bool {
switch t := obj.(type) {
case *v1.Pod:
return assignedPod(t)
case cache.DeletedFinalStateUnknown:
if pod, ok := t.Obj.(*v1.Pod); ok {
return assignedPod(pod)
}
utilruntime.HandleError(fmt.Errorf("unable to convert object %T to *v1.Pod in %T", obj, sched))
return false
default:
utilruntime.HandleError(fmt.Errorf("unable to handle object in %T: %T", sched, obj))
return false
}
},
Handler: cache.ResourceEventHandlerFuncs{
AddFunc: sched.addPodToCache,
UpdateFunc: sched.updatePodInCache,
DeleteFunc: sched.deletePodFromCache,
},
},
)
// unscheduled pod queue
informerFactory.Core().V1().Pods().Informer().AddEventHandler(
cache.FilteringResourceEventHandler{
FilterFunc: func(obj interface{}) bool {
switch t := obj.(type) {
case *v1.Pod:
return !assignedPod(t) && responsibleForPod(t, sched.Profiles)
case cache.DeletedFinalStateUnknown:
if pod, ok := t.Obj.(*v1.Pod); ok {
return !assignedPod(pod) && responsibleForPod(pod, sched.Profiles)
}
utilruntime.HandleError(fmt.Errorf("unable to convert object %T to *v1.Pod in %T", obj, sched))
return false
default:
utilruntime.HandleError(fmt.Errorf("unable to handle object in %T: %T", sched, obj))
return false
}
},
Handler: cache.ResourceEventHandlerFuncs{
AddFunc: sched.addPodToSchedulingQueue,
UpdateFunc: sched.updatePodInSchedulingQueue,
DeleteFunc: sched.deletePodFromSchedulingQueue,
},
},
)
informerFactory.Core().V1().Nodes().Informer().AddEventHandler(
cache.ResourceEventHandlerFuncs{
AddFunc: sched.addNodeToCache,
UpdateFunc: sched.updateNodeInCache,
DeleteFunc: sched.deleteNodeFromCache,
},
)
buildEvtResHandler := func(at framework.ActionType, gvk framework.GVK, shortGVK string) cache.ResourceEventHandlerFuncs {
funcs := cache.ResourceEventHandlerFuncs{}
if at&framework.Add != 0 {
evt := framework.ClusterEvent{Resource: gvk, ActionType: framework.Add, Label: fmt.Sprintf("%vAdd", shortGVK)}
funcs.AddFunc = func(_ interface{}) {
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(evt, nil)
}
}
if at&framework.Update != 0 {
evt := framework.ClusterEvent{Resource: gvk, ActionType: framework.Update, Label: fmt.Sprintf("%vUpdate", shortGVK)}
funcs.UpdateFunc = func(_, _ interface{}) {
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(evt, nil)
}
}
if at&framework.Delete != 0 {
evt := framework.ClusterEvent{Resource: gvk, ActionType: framework.Delete, Label: fmt.Sprintf("%vDelete", shortGVK)}
funcs.DeleteFunc = func(_ interface{}) {
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(evt, nil)
}
}
return funcs
}
for gvk, at := range gvkMap {
switch gvk {
case framework.Node, framework.Pod:
// Do nothing.
case framework.CSINode:
informerFactory.Storage().V1().CSINodes().Informer().AddEventHandler(
buildEvtResHandler(at, framework.CSINode, "CSINode"),
)
case framework.CSIDriver:
informerFactory.Storage().V1().CSIDrivers().Informer().AddEventHandler(
buildEvtResHandler(at, framework.CSIDriver, "CSIDriver"),
)
case framework.CSIStorageCapacity:
informerFactory.Storage().V1beta1().CSIStorageCapacities().Informer().AddEventHandler(
buildEvtResHandler(at, framework.CSIStorageCapacity, "CSIStorageCapacity"),
)
case framework.PersistentVolume:
// MaxPDVolumeCountPredicate: since it relies on the counts of PV.
//
// PvAdd: Pods created when there are no PVs available will be stuck in
// unschedulable queue. But unbound PVs created for static provisioning and
// delay binding storage class are skipped in PV controller dynamic
// provisioning and binding process, will not trigger events to schedule pod
// again. So we need to move pods to active queue on PV add for this
// scenario.
//
// PvUpdate: Scheduler.bindVolumesWorker may fail to update assumed pod volume
// bindings due to conflicts if PVs are updated by PV controller or other
// parties, then scheduler will add pod back to unschedulable queue. We
// need to move pods to active queue on PV update for this scenario.
informerFactory.Core().V1().PersistentVolumes().Informer().AddEventHandler(
buildEvtResHandler(at, framework.PersistentVolume, "Pv"),
)
case framework.PersistentVolumeClaim:
// MaxPDVolumeCountPredicate: add/update PVC will affect counts of PV when it is bound.
informerFactory.Core().V1().PersistentVolumeClaims().Informer().AddEventHandler(
buildEvtResHandler(at, framework.PersistentVolumeClaim, "Pvc"),
)
case framework.StorageClass:
if at&framework.Add != 0 {
informerFactory.Storage().V1().StorageClasses().Informer().AddEventHandler(
cache.ResourceEventHandlerFuncs{
AddFunc: sched.onStorageClassAdd,
},
)
}
if at&framework.Update != 0 {
informerFactory.Storage().V1().StorageClasses().Informer().AddEventHandler(
cache.ResourceEventHandlerFuncs{
UpdateFunc: func(_, _ interface{}) {
sched.SchedulingQueue.MoveAllToActiveOrBackoffQueue(queue.StorageClassUpdate, nil)
},
},
)
}
case framework.Service:
// ServiceAffinity: affected by the selector of the service is updated.
// Also, if new service is added, equivalence cache will also become invalid since
// existing pods may be "captured" by this service and change this predicate result.
informerFactory.Core().V1().Services().Informer().AddEventHandler(
buildEvtResHandler(at, framework.Service, "Service"),
)
default:
// Tests may not instantiate dynInformerFactory.
if dynInformerFactory == nil {
continue
}
// GVK is expected to be at least 3-folded, separated by dots.
// <kind in plural>.<version>.<group>
// Valid examples:
// - foos.v1.example.com
// - bars.v1beta1.a.b.c
// Invalid examples:
// - foos.v1 (2 sections)
// - foo.v1.example.com (the first section should be plural)
if strings.Count(string(gvk), ".") < 2 {
klog.ErrorS(nil, "incorrect event registration", "gvk", gvk)
continue
}
// Fall back to try dynamic informers.
gvr, _ := schema.ParseResourceArg(string(gvk))
dynInformer := dynInformerFactory.ForResource(*gvr).Informer()
dynInformer.AddEventHandler(
buildEvtResHandler(at, gvk, strings.Title(gvr.Resource)),
)
}
}
}
func nodeSchedulingPropertiesChange(newNode *v1.Node, oldNode *v1.Node) *framework.ClusterEvent {
if nodeSpecUnschedulableChanged(newNode, oldNode) {
return &queue.NodeSpecUnschedulableChange
}
if nodeAllocatableChanged(newNode, oldNode) {
return &queue.NodeAllocatableChange
}
if nodeLabelsChanged(newNode, oldNode) {
return &queue.NodeLabelChange
}
if nodeTaintsChanged(newNode, oldNode) {
return &queue.NodeTaintChange
}
if nodeConditionsChanged(newNode, oldNode) {
return &queue.NodeConditionChange
}
return nil
}
func nodeAllocatableChanged(newNode *v1.Node, oldNode *v1.Node) bool {
return !reflect.DeepEqual(oldNode.Status.Allocatable, newNode.Status.Allocatable)
}
func nodeLabelsChanged(newNode *v1.Node, oldNode *v1.Node) bool {
return !reflect.DeepEqual(oldNode.GetLabels(), newNode.GetLabels())
}
func nodeTaintsChanged(newNode *v1.Node, oldNode *v1.Node) bool {
return !reflect.DeepEqual(newNode.Spec.Taints, oldNode.Spec.Taints)
}
func nodeConditionsChanged(newNode *v1.Node, oldNode *v1.Node) bool {
strip := func(conditions []v1.NodeCondition) map[v1.NodeConditionType]v1.ConditionStatus {
conditionStatuses := make(map[v1.NodeConditionType]v1.ConditionStatus, len(conditions))
for i := range conditions {
conditionStatuses[conditions[i].Type] = conditions[i].Status
}
return conditionStatuses
}
return !reflect.DeepEqual(strip(oldNode.Status.Conditions), strip(newNode.Status.Conditions))
}
func nodeSpecUnschedulableChanged(newNode *v1.Node, oldNode *v1.Node) bool {
return newNode.Spec.Unschedulable != oldNode.Spec.Unschedulable && !newNode.Spec.Unschedulable
}
func preCheckForNode(nodeInfo *framework.NodeInfo) queue.PreEnqueueCheck {
// In addition to the checks in kubelet (pkg/kubelet/lifecycle/predicate.go#GeneralPredicates),
// the following logic appends a taint/toleration check.
// TODO: verify if kubelet should also apply the taint/toleration check, and then unify the
// logic with kubelet and move to a shared place.
//
// Note: the following checks doesn't take preemption into considerations, in very rare
// cases (e.g., node resizing), "pod" may still fail a check but preemption helps. We deliberately
// chose to ignore those cases as unschedulable pods will be re-queued eventually.
return func(pod *v1.Pod) bool {
if len(noderesources.Fits(pod, nodeInfo, feature.DefaultFeatureGate.Enabled(features.PodOverhead))) != 0 {
return false
}
// Ignore parsing errors for backwards compatibility.
matches, _ := nodeaffinity.GetRequiredNodeAffinity(pod).Match(nodeInfo.Node())
if !matches {
return false
}
if !nodename.Fits(pod, nodeInfo) {
return false
}
if !nodeports.Fits(pod, nodeInfo) {
return false
}
_, isUntolerated := v1helper.FindMatchingUntoleratedTaint(nodeInfo.Node().Spec.Taints, pod.Spec.Tolerations, func(t *v1.Taint) bool {
// PodToleratesNodeTaints is only interested in NoSchedule and NoExecute taints.
return t.Effect == v1.TaintEffectNoSchedule || t.Effect == v1.TaintEffectNoExecute
})
return !isUntolerated
}
}