-
Notifications
You must be signed in to change notification settings - Fork 877
/
workloadrebalancer_controller.go
300 lines (257 loc) · 12.5 KB
/
workloadrebalancer_controller.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
/*
Copyright 2024 The Karmada 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 workloadrebalancer
import (
"context"
"fmt"
"reflect"
"sort"
"time"
apierrors "k8s.io/apimachinery/pkg/api/errors"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/client-go/util/retry"
"k8s.io/klog/v2"
controllerruntime "sigs.k8s.io/controller-runtime"
"sigs.k8s.io/controller-runtime/pkg/builder"
"sigs.k8s.io/controller-runtime/pkg/client"
"sigs.k8s.io/controller-runtime/pkg/event"
"sigs.k8s.io/controller-runtime/pkg/predicate"
appsv1alpha1 "github.com/karmada-io/karmada/pkg/apis/apps/v1alpha1"
workv1alpha2 "github.com/karmada-io/karmada/pkg/apis/work/v1alpha2"
"github.com/karmada-io/karmada/pkg/util/names"
)
const (
// ControllerName is the controller name that will be used when reporting events.
ControllerName = "workload-rebalancer"
)
// RebalancerController is to handle a rebalance to workloads selected by WorkloadRebalancer object.
type RebalancerController struct {
Client client.Client
}
// SetupWithManager creates a controller and register to controller manager.
func (c *RebalancerController) SetupWithManager(mgr controllerruntime.Manager) error {
var predicateFunc = predicate.Funcs{
CreateFunc: func(_ event.CreateEvent) bool { return true },
UpdateFunc: func(e event.UpdateEvent) bool {
oldObj := e.ObjectOld.(*appsv1alpha1.WorkloadRebalancer)
newObj := e.ObjectNew.(*appsv1alpha1.WorkloadRebalancer)
return !reflect.DeepEqual(oldObj.Spec, newObj.Spec)
},
DeleteFunc: func(event.DeleteEvent) bool { return false },
GenericFunc: func(event.GenericEvent) bool { return false },
}
return controllerruntime.NewControllerManagedBy(mgr).
Named(ControllerName).
For(&appsv1alpha1.WorkloadRebalancer{}, builder.WithPredicates(predicateFunc)).
Complete(c)
}
// Reconcile performs a full reconciliation for the object referred to by the Request.
// The Controller will requeue the Request to be processed again if an error is non-nil or
// Result.Requeue is true, otherwise upon completion it will remove the work from the queue.
func (c *RebalancerController) Reconcile(ctx context.Context, req controllerruntime.Request) (controllerruntime.Result, error) {
klog.V(4).Infof("Reconciling for WorkloadRebalancer %s", req.Name)
// 1. get latest WorkloadRebalancer
rebalancer := &appsv1alpha1.WorkloadRebalancer{}
if err := c.Client.Get(ctx, req.NamespacedName, rebalancer); err != nil {
if apierrors.IsNotFound(err) {
klog.Infof("no need to reconcile WorkloadRebalancer for it not found")
return controllerruntime.Result{}, nil
}
return controllerruntime.Result{}, err
}
// 2. list latest target workloads and trigger its rescheduling by updating its referenced binding.
newStatus, err := c.handleWorkloadRebalance(ctx, rebalancer)
if err != nil {
return controllerruntime.Result{}, err
}
if rebalancer.Status.FinishTime == nil {
// should never reach here.
klog.Errorf("finishTime shouldn't be nil, current status: %+v", rebalancer.Status)
return controllerruntime.Result{}, nil
}
// 3. when all workloads finished, judge whether the rebalancer needs cleanup.
rebalancer.Status = *newStatus
if rebalancer.Spec.TTLSecondsAfterFinished != nil {
remainingTTL := timeLeft(rebalancer)
if remainingTTL > 0 {
return controllerruntime.Result{RequeueAfter: remainingTTL}, nil
}
if err := c.deleteWorkloadRebalancer(ctx, rebalancer); err != nil {
return controllerruntime.Result{}, err
}
}
return controllerruntime.Result{}, nil
}
// When spec filed of WorkloadRebalancer updated, we shall refresh the workload list in status.observedWorkloads:
// 1. a new workload added to spec list, just add it into status list and do the rebalance.
// 2. a workload deleted from previous spec list, keep it in status list if already success, and remove it if not.
// 3. a workload is modified, just regard it as deleted an old one and inserted a new one.
// 4. just list order is disrupted, no additional action.
func (c *RebalancerController) syncWorkloadsFromSpecToStatus(rebalancer *appsv1alpha1.WorkloadRebalancer) *appsv1alpha1.WorkloadRebalancerStatus {
observedWorkloads := make([]appsv1alpha1.ObservedWorkload, 0)
specWorkloadSet := sets.New[appsv1alpha1.ObjectReference]()
for _, workload := range rebalancer.Spec.Workloads {
specWorkloadSet.Insert(workload)
}
for _, item := range rebalancer.Status.ObservedWorkloads {
// if item still exist in `spec`, keep it in `status` and remove it from `specWorkloadSet`.
// if item no longer exist in `spec`, keep it in `status` if it already success, otherwise remove it from `status`.
if specWorkloadSet.Has(item.Workload) {
observedWorkloads = append(observedWorkloads, item)
specWorkloadSet.Delete(item.Workload)
} else if item.Result == appsv1alpha1.RebalanceSuccessful {
observedWorkloads = append(observedWorkloads, item)
}
}
// since item exist in both `spec` and `status` has been removed, the left means the newly added workload,
// add them into `status`.
for workload := range specWorkloadSet {
observedWorkloads = append(observedWorkloads, appsv1alpha1.ObservedWorkload{Workload: workload})
}
sort.Slice(observedWorkloads, func(i, j int) bool {
wi := observedWorkloads[i].Workload
wj := observedWorkloads[j].Workload
stri := fmt.Sprintf("%s %s %s %s", wi.APIVersion, wi.Kind, wi.Namespace, wi.Name)
strj := fmt.Sprintf("%s %s %s %s", wj.APIVersion, wj.Kind, wj.Namespace, wj.Name)
return stri < strj
})
return &appsv1alpha1.WorkloadRebalancerStatus{ObservedWorkloads: observedWorkloads, ObservedGeneration: rebalancer.Generation}
}
func (c *RebalancerController) handleWorkloadRebalance(ctx context.Context, rebalancer *appsv1alpha1.WorkloadRebalancer) (*appsv1alpha1.WorkloadRebalancerStatus, error) {
// 1. get previous status and update basing on it
newStatus := c.syncWorkloadsFromSpecToStatus(rebalancer)
// 2. trigger the rescheduling of target workloads listed in newStatus
newStatus, retryNum := c.triggerReschedule(ctx, rebalancer.ObjectMeta, newStatus)
// 3. judge whether status changed and update finishTime field if so
statusChanged := false
newStatus.FinishTime = rebalancer.Status.FinishTime
// compare whether newStatus equals to old status except finishTime field.
if !reflect.DeepEqual(rebalancer.Status, *newStatus) {
statusChanged = true
// retryNum is 0 means the latest rebalancer has finished, which should update finishTime.
if retryNum == 0 {
finishTime := metav1.Now()
newStatus.FinishTime = &finishTime
}
}
// 4. update status of WorkloadRebalancer if statusChanged
if statusChanged {
if err := c.updateWorkloadRebalancerStatus(ctx, rebalancer, newStatus); err != nil {
return newStatus, fmt.Errorf("update status failed: %w", err)
}
}
if retryNum > 0 {
return newStatus, fmt.Errorf("%d resource reschedule triggered failed and need retry", retryNum)
}
return newStatus, nil
}
func (c *RebalancerController) triggerReschedule(ctx context.Context, metadata metav1.ObjectMeta, newStatus *appsv1alpha1.WorkloadRebalancerStatus) (
*appsv1alpha1.WorkloadRebalancerStatus, int64) {
successNum, retryNum := int64(0), int64(0)
for i, resource := range newStatus.ObservedWorkloads {
if resource.Result == appsv1alpha1.RebalanceSuccessful {
successNum++
continue
}
if resource.Result == appsv1alpha1.RebalanceFailed {
continue
}
bindingName := names.GenerateBindingName(resource.Workload.Kind, resource.Workload.Name)
// resource with empty namespace represents it is a cluster wide resource.
if resource.Workload.Namespace != "" {
binding := &workv1alpha2.ResourceBinding{}
if err := c.Client.Get(ctx, client.ObjectKey{Namespace: resource.Workload.Namespace, Name: bindingName}, binding); err != nil {
klog.Errorf("get binding for resource %+v failed: %+v", resource.Workload, err)
c.recordAndCountRebalancerFailed(&newStatus.ObservedWorkloads[i], &retryNum, err)
continue
}
// update spec.rescheduleTriggeredAt of referenced fetchTargetRefBindings to trigger a rescheduling
if c.needTriggerReschedule(metadata.CreationTimestamp, binding.Spec.RescheduleTriggeredAt) {
binding.Spec.RescheduleTriggeredAt = &metadata.CreationTimestamp
if err := c.Client.Update(ctx, binding); err != nil {
klog.Errorf("update binding for resource %+v failed: %+v", resource.Workload, err)
c.recordAndCountRebalancerFailed(&newStatus.ObservedWorkloads[i], &retryNum, err)
continue
}
}
c.recordAndCountRebalancerSuccess(&newStatus.ObservedWorkloads[i], &successNum)
} else {
clusterbinding := &workv1alpha2.ClusterResourceBinding{}
if err := c.Client.Get(ctx, client.ObjectKey{Name: bindingName}, clusterbinding); err != nil {
klog.Errorf("get cluster binding for resource %+v failed: %+v", resource.Workload, err)
c.recordAndCountRebalancerFailed(&newStatus.ObservedWorkloads[i], &retryNum, err)
continue
}
// update spec.rescheduleTriggeredAt of referenced clusterbinding to trigger a rescheduling
if c.needTriggerReschedule(metadata.CreationTimestamp, clusterbinding.Spec.RescheduleTriggeredAt) {
clusterbinding.Spec.RescheduleTriggeredAt = &metadata.CreationTimestamp
if err := c.Client.Update(ctx, clusterbinding); err != nil {
klog.Errorf("update cluster binding for resource %+v failed: %+v", resource.Workload, err)
c.recordAndCountRebalancerFailed(&newStatus.ObservedWorkloads[i], &retryNum, err)
continue
}
}
c.recordAndCountRebalancerSuccess(&newStatus.ObservedWorkloads[i], &successNum)
}
}
klog.V(4).Infof("Finish handling WorkloadRebalancer (%s), %d/%d resource success in all, while %d resource need retry",
metadata.Name, successNum, len(newStatus.ObservedWorkloads), retryNum)
return newStatus, retryNum
}
func (c *RebalancerController) needTriggerReschedule(creationTimestamp metav1.Time, rescheduleTriggeredAt *metav1.Time) bool {
return rescheduleTriggeredAt == nil || creationTimestamp.After(rescheduleTriggeredAt.Time)
}
func (c *RebalancerController) recordAndCountRebalancerSuccess(resource *appsv1alpha1.ObservedWorkload, successNum *int64) {
resource.Result = appsv1alpha1.RebalanceSuccessful
*successNum++
}
func (c *RebalancerController) recordAndCountRebalancerFailed(resource *appsv1alpha1.ObservedWorkload, retryNum *int64, err error) {
reason := apierrors.ReasonForError(err)
if reason == metav1.StatusReasonNotFound {
resource.Result = appsv1alpha1.RebalanceFailed
resource.Reason = appsv1alpha1.RebalanceObjectNotFound
} else {
*retryNum++
}
}
func (c *RebalancerController) updateWorkloadRebalancerStatus(ctx context.Context, rebalancer *appsv1alpha1.WorkloadRebalancer,
newStatus *appsv1alpha1.WorkloadRebalancerStatus) error {
modifiedRebalancer := rebalancer.DeepCopy()
modifiedRebalancer.Status = *newStatus
return retry.RetryOnConflict(retry.DefaultRetry, func() (err error) {
klog.V(4).Infof("Start to patch WorkloadRebalancer(%s) status", rebalancer.Name)
if err = c.Client.Status().Patch(ctx, modifiedRebalancer, client.MergeFrom(rebalancer)); err != nil {
klog.Errorf("Failed to patch WorkloadRebalancer (%s) status, err: %+v", rebalancer.Name, err)
return err
}
klog.V(4).Infof("Patch WorkloadRebalancer(%s) successful", rebalancer.Name)
return nil
})
}
func (c *RebalancerController) deleteWorkloadRebalancer(ctx context.Context, rebalancer *appsv1alpha1.WorkloadRebalancer) error {
klog.V(4).Infof("Start to clean up WorkloadRebalancer(%s)", rebalancer.Name)
options := &client.DeleteOptions{Preconditions: &metav1.Preconditions{ResourceVersion: &rebalancer.ResourceVersion}}
if err := c.Client.Delete(ctx, rebalancer, options); err != nil {
klog.Errorf("Cleaning up WorkloadRebalancer(%s) failed: %+v.", rebalancer.Name, err)
return err
}
klog.V(4).Infof("Cleaning up WorkloadRebalancer(%s) successful", rebalancer.Name)
return nil
}
func timeLeft(r *appsv1alpha1.WorkloadRebalancer) time.Duration {
expireAt := r.Status.FinishTime.Add(time.Duration(*r.Spec.TTLSecondsAfterFinished) * time.Second)
remainingTTL := time.Until(expireAt)
klog.V(4).Infof("Found Rebalancer(%s) finished at: %+v, remainingTTL: %+v", r.Name, r.Status.FinishTime.UTC(), remainingTTL)
return remainingTTL
}