[feature request] When CloneSet volumeClaimTemplates changed, Pod must ReCreate update

[zmberg]

What would you like to be added:
Question: When CloneSet volumeClaimTemplates and Image changed, Pod will in-place update, then volumeClaimTemplates not updated to Pod volumes, because the volumes field is only recalculated when the Pod is created.

What happened:

1. Create CloneSet with volumeClaimTemplates, as follows:

apiVersion: apps.kruise.io/v1alpha1
kind: CloneSet
metadata:
  name: web
spec:
  updateStrategy:
    type: InPlaceIfPossible
  replicas: 2
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
        - name: nginx
          image: nginx:stable-alpine3.17
          env:
          - name: version
            value: v1
          volumeMounts:
            - name: www-data
              mountPath: /usr/share/nginx/html
  volumeClaimTemplates:
    - metadata:
        name: www-data
      spec:
        accessModes: ["ReadWriteOnce"]
        resources:
          requests:
            storage: 1Gi

2. Update image tag from stable-alpine3.17 to stable-alpine, and volumeClaimTemplates storage from 1Gi to 2Gi.

apiVersion: apps.kruise.io/v1alpha1
kind: CloneSet
spec:
  template:
    spec:
      containers:
        - name: nginx
          image: nginx:stable-alpine
  volumeClaimTemplates:
    - metadata:
        name: www-data
      spec:
        accessModes: ["ReadWriteOnce"]
        resources:
          requests:
            storage: 2Gi

3. Update cloneSet, but pods only in-place update and pvc capacity don't change.

% kubectl get clonesets
NAME   DESIRED   UPDATED   UPDATED_READY   READY   TOTAL   AGE
web    2         2         2               2       2       2m27s

% kubectl get pods
NAME        READY   STATUS    RESTARTS      AGE
web-c48gr   1/1     Running   1 (16s ago)   2m33s
web-mjmvl   1/1     Running   1 (23s ago)   2m33s

% kubectl get pvc
NAME                 STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
www-data-web-c48gr   Bound    pvc-4100c7c4-0fb0-41b6-90e1-696cfdd970c8   1Gi        RWO            standard       2m38s
www-data-web-mjmvl   Bound    pvc-256c0b60-3cee-4da5-ba66-5530ce153b1b   1Gi        RWO            standard       2m38s

Goals

• Image and volumeClaimTemplates changed, ReCreate update Pod.

Non-Goals

• Only volumeClaimTemplates changed, will not trigger Pod update

Solution

1. When CloneSet contains volumeClaimTemplates, NewRevision with volumeClaimTemplates hash. refer code: https://github.com/openkruise/kruise/blob/release-1.5/pkg/controller/cloneset/revision/cloneset_revision.go#L51

apiVersion: apps/v1
kind: ControllerRevision
metadata:
  annotations:
    kruise.io/cloneset-volumeclaimtemplate-hash: {volumeClaimTemplates.spec hash}

2. CanUpdateInPlace determines if the hash value of ControllerRevision has changed, then returns false. refer code: https://github.com/openkruise/kruise/blob/v1.6.1/pkg/util/inplaceupdate/inplace_update.go#L270

[hzliangbin]

@zmberg I'd like to take this shot.If nobody doing this job, can I summit a PR?

[furykerry]

@hzliangbin thanks, however we have some internal user problem related to this issue in Alibaba, so we'd like to work on this. You can start with other issue

[ABNER-1]

Please consider assigning this issue to me.

[zmberg]

/assign @ABNER-1