ceph-csi/docs/ceph-csi-upgrade.md

Ceph-csi Upgrade

• Ceph-csi Upgrade
  • Pre-upgrade considerations
    • RBD CSI Snapshot Incompatibility
  • Upgrading from v1.2 to v2.0
  • Upgrading from v2.0 to v2.1
  • Upgrading from v2.1 to v3.0
    • Upgrading CephFS
      • 1. Upgrade CephFS Provisioner resources
        • 1.1 Update the CephFS Provisioner RBAC
        • 1.2 Update the CephFS Provisioner deployment
      • 2. Upgrade CephFS Nodeplugin resources
        • 2.1 Update the CephFS Nodeplugin RBAC
        • 2.2 Update the CephFS Nodeplugin daemonset
        • 2.3 Manual deletion of CephFS Nodeplugin daemonset pods
    • Upgrading RBD
      • 3. Upgrade RBD Provisioner resources
        • 3.1 Update the RBD Provisioner RBAC
        • 3.2 Update the RBD Provisioner deployment
        • 3.3 Update snapshot CRD from Alpha to Beta
      • 4. Upgrade RBD Nodeplugin resources
        • 4.1 Update the RBD Nodeplugin RBAC
        • 4.2 Update the RBD Nodeplugin daemonset
        • 4.3 Manual deletion of RBD Nodeplugin daemonset pods
    • Handling node reboot hangs due to existing network mounts

Pre-upgrade considerations

In some scenarios there is an issue in the CSI driver that will cause application pods to be disconnected from their mounts when the CSI driver is restarted. Since the upgrade would cause the CSI driver to restart if it is updated, you need to be aware of whether this affects your applications. This issue will happen when using the Ceph fuse client or rbd-nbd:

CephFS: If you are provision volumes for CephFS and have a kernel less than version 4.17, The CSI driver will fall back to use the FUSE client.

RBD: If you have set the mounter: rbd-nbd option in the RBD storage class, the NBD mounter will have this issue.

If you are affected by this issue, you will need to proceed carefully during the upgrade to restart your application pods. The recommended step is to modify the update strategy of the CSI nodeplugin daemonsets to OnDelete so that you can control when the CSI driver pods are restarted on each node.

To avoid this issue in future upgrades, we recommend that you do not use the fuse client or rbd-nbd as of now.

This guide will walk you through the steps to upgrade the software in a cluster from v2.1 to v3.0

RBD CSI Snapshot Incompatibility

CSI snapshot is moved from Alpha to Beta and is not backward compatible. The snapshots created with the Alpha version must be deleted before the upgrade.

Check if we have any v1alpha1 CRD created in our kubernetes cluster, If there is no v1alpha1 CRD created you can skip this step.

[$]kubectl get crd volumesnapshotclasses.snapshot.storage.k8s.io -o yaml |grep v1alpha1
  - name: v1alpha1
  - v1alpha1
[$]kubectl get crd volumesnapshotcontents.snapshot.storage.k8s.io -o yaml |grep v1alpha1
  - name: v1alpha1
  - v1alpha1
[$]kubectl get crd volumesnapshots.snapshot.storage.k8s.io -o yaml |grep v1alpha1
  - name: v1alpha1
  - v1alpha1

• List all the volumesnapshots created

[$] kubectl get volumesnapshot
NAME               AGE
rbd-pvc-snapshot   22s

• Delete all volumesnapshots

[$] kubectl delete volumesnapshot rbd-pvc-snapshot
volumesnapshot.snapshot.storage.k8s.io "rbd-pvc-snapshot" deleted

• List all volumesnapshotclasses created

[$] kubectl get volumesnapshotclass
NAME                      AGE
csi-rbdplugin-snapclass   86s

• Delete all volumesnapshotclasses

[$] kubectl delete volumesnapshotclass csi-rbdplugin-snapclass
volumesnapshotclass.snapshot.storage.k8s.io "csi-rbdplugin-snapclass" deleted

Note: The underlying snapshots on the storage system will be deleted by ceph-csi

Upgrading from v1.2 to v2.0

Refer upgrade-from-1.2-v2.0 to upgrade from cephcsi 1.2 to v2.0

Upgrading from v2.0 to v2.1

Refer upgrade-from-2.0-v2.1 to upgrade from cephcsi v2.0 to v2.1

Upgrading from v2.1 to v3.0

Ceph-csi releases from master are expressly unsupported. It is strongly recommended that you use official releases of Ceph-csi. Unreleased versions from the master branch are subject to changes and incompatibilities that will not be supported in the official releases. Builds from the master branch can have functionality changed and even removed at any time without compatibility support and without prior notice.

git checkout v3.0.0 tag

[$] git clone https://github.com/ceph/ceph-csi.git
[$] cd ./ceph-csi
[$] git checkout v3.0.0

Note: While upgrading please Ignore warning messages from kubectl output

Warning: kubectl apply should be used on resource created by either kubectl create --save-config or kubectl apply

Upgrading CephFS

Upgrading cephfs csi includes upgrade of cephfs driver and as well as kubernetes sidecar containers and also the permissions required for the kubernetes sidecar containers, lets upgrade the things one by one

1. Upgrade CephFS Provisioner resources

Upgrade provisioner resources include updating the provisioner RBAC and Provisioner deployment

1.1 Update the CephFS Provisioner RBAC

[$] kubectl apply -f deploy/cephfs/kubernetes/csi-provisioner-rbac.yaml
serviceaccount/cephfs-csi-provisioner configured
clusterrole.rbac.authorization.k8s.io/cephfs-external-provisioner-runner configured
clusterrole.rbac.authorization.k8s.io/cephfs-external-provisioner-runner-rules configured
clusterrolebinding.rbac.authorization.k8s.io/cephfs-csi-provisioner-role configured
role.rbac.authorization.k8s.io/cephfs-external-provisioner-cfg configured
rolebinding.rbac.authorization.k8s.io/cephfs-csi-provisioner-role-cfg configured

1.2 Update the CephFS Provisioner deployment

[$]kubectl apply -f deploy/cephfs/kubernetes/csi-cephfsplugin-provisioner.yaml
service/csi-cephfsplugin-provisioner configured
deployment.apps/csi-cephfsplugin-provisioner configured

wait for the deployment to complete

[$]kubectl get deployment
NAME                           READY   UP-TO-DATE   AVAILABLE   AGE
csi-cephfsplugin-provisioner   3/3     1            3           104m

deployment UP-TO-DATE value must be same as READY

2. Upgrade CephFS Nodeplugin resources

Upgrading nodeplugin resources include updating the nodeplugin RBAC and nodeplugin daemonset

2.1 Update the CephFS Nodeplugin RBAC

[$]kubectl apply -f deploy/cephfs/kubernetes/csi-nodeplugin-rbac.yaml
serviceaccount/cephfs-csi-nodeplugin configured
clusterrole.rbac.authorization.k8s.io/cephfs-csi-nodeplugin configured
clusterrole.rbac.authorization.k8s.io/cephfs-csi-nodeplugin-rules configured
clusterrolebinding.rbac.authorization.k8s.io/cephfs-csi-nodeplugin configured

If you determined in Pre-upgrade considerations that you were affected by the CSI driver restart issue that disconnects the application pods from their mounts, continue with this section. Otherwise, you can skip to step 2.2

vi deploy/cephfs/kubernetes/csi-cephfsplugin.yaml

kind: DaemonSet
apiVersion: apps/v1
metadata:
  name: csi-cephfsplugin
spec:
  selector:
    matchLabels:
      app: csi-cephfsplugin
  updateStrategy:
    type: OnDelete
  template:
    metadata:
      labels:
        app: csi-cephfsplugin
    spec:
      serviceAccount: cephfs-csi-nodeplugin

in the above template we have added updateStrategy and its type to the daemonset spec

2.2 Update the CephFS Nodeplugin daemonset

[$]kubectl apply -f deploy/cephfs/kubernetes/csi-cephfsplugin.yaml
daemonset.apps/csi-cephfsplugin configured
service/csi-metrics-cephfsplugin configured

2.3 Manual deletion of CephFS Nodeplugin daemonset pods

If you determined in Pre-upgrade considerations that you were affected by the CSI driver restart issue that disconnects the application pods from their mounts, continue with this section. Otherwise, you can skip this section.

As we have set the updateStrategy to OnDelete the CSI driver pods will not be updated until you delete them manually. This allows you to control when your application pods will be affected by the CSI driver restart.

For each node:

• Drain your application pods from the node
• Delete the CSI driver pods on the node
  • The pods to delete will be named with a csi-cephfsplugin prefix and have a random suffix on each node. However, no need to delete the provisioner pods: csi-cephfsplugin-provisioner-* .
  • The pod deletion causes the pods to be restarted and updated automatically on the node.

we have successfully upgraded cephfs csi from v2.1 to v3.0

Upgrading RBD

Upgrading rbd csi includes upgrade of rbd driver and as well as kubernetes sidecar containers and also the permissions required for the kubernetes sidecar containers, lets upgrade the things one by one

3. Upgrade RBD Provisioner resources

Upgrading provisioner resources include updating the provisioner RBAC and Provisioner deployment

3.1 Update the RBD Provisioner RBAC

[$]kubectl apply -f deploy/rbd/kubernetes/csi-provisioner-rbac.yaml
serviceaccount/rbd-csi-provisioner configured
clusterrole.rbac.authorization.k8s.io/rbd-external-provisioner-runner configured
clusterrole.rbac.authorization.k8s.io/rbd-external-provisioner-runner-rules configured
clusterrolebinding.rbac.authorization.k8s.io/rbd-csi-provisioner-role configured
role.rbac.authorization.k8s.io/rbd-external-provisioner-cfg configured
rolebinding.rbac.authorization.k8s.io/rbd-csi-provisioner-role-cfg configured

3.2 Update the RBD Provisioner deployment

[$]kubectl apply -f deploy/rbd/kubernetes/csi-rbdplugin-provisioner.yaml
service/csi-rbdplugin-provisioner configured
deployment.apps/csi-rbdplugin-provisioner configured

wait for the deployment to complete

[$]kubectl get deployments
NAME                           READY   UP-TO-DATE   AVAILABLE   AGE
csi-rbdplugin-provisioner      3/3     3            3           139m

deployment UP-TO-DATE value must be same as READY

3.3 Update snapshot CRD from Alpha to Beta

As we are updating the snapshot resources from Alpha to Beta we need to delete the old alphav1 snapshot CRD created by external-snapshotter sidecar container

Check if we have any v1alpha1 CRD created in our kubernetes cluster

[$]kubectl get crd volumesnapshotclasses.snapshot.storage.k8s.io -o yaml |grep v1alpha1
  - name: v1alpha1
  - v1alpha1
[$]kubectl get crd volumesnapshotcontents.snapshot.storage.k8s.io -o yaml |grep v1alpha1
  - name: v1alpha1
  - v1alpha1
[$]kubectl get crd volumesnapshots.snapshot.storage.k8s.io -o yaml |grep v1alpha1
  - name: v1alpha1
  - v1alpha1

As we have v1alpha1 CRD created in our kubernetes cluster,we need to delete the Alpha CRD

[$]kubectl delete crd volumesnapshotclasses.snapshot.storage.k8s.io
customresourcedefinition.apiextensions.k8s.io "volumesnapshotclasses.snapshot.storage.k8s.io" deleted
[$]kubectl delete crd volumesnapshotcontents.snapshot.storage.k8s.io
customresourcedefinition.apiextensions.k8s.io "volumesnapshotcontents.snapshot.storage.k8s.io" deleted
[$]kubectl delete crd volumesnapshots.snapshot.storage.k8s.io
customresourcedefinition.apiextensions.k8s.io "volumesnapshots.snapshot.storage.k8s.io" deleted

Note: Its kubernetes distributor responsibility to install new snapshot controller and snapshot beta CRD. more info can be found here

4. Upgrade RBD Nodeplugin resources

Upgrading nodeplugin resources include updating the nodeplugin RBAC and nodeplugin daemonset

4.1 Update the RBD Nodeplugin RBAC

[$]kubectl apply -f deploy/rbd/kubernetes/csi-nodeplugin-rbac.yaml
serviceaccount/rbd-csi-nodeplugin configured
clusterrole.rbac.authorization.k8s.io/rbd-csi-nodeplugin configured
clusterrole.rbac.authorization.k8s.io/rbd-csi-nodeplugin-rules configured
clusterrolebinding.rbac.authorization.k8s.io/rbd-csi-nodeplugin configured

If you determined in Pre-upgrade considerations that you were affected by the CSI driver restart issue that disconnects the application pods from their mounts, continue with this section. Otherwise, you can skip to step 4.2

vi deploy/rbd/kubernetes/csi-rbdplugin.yaml

kind: DaemonSet
apiVersion: apps/v1
metadata:
  name: csi-rbdplugin
spec:
  selector:
    matchLabels:
      app: csi-rbdplugin
  updateStrategy:
    type: OnDelete
  template:
    metadata:
      labels:
        app: csi-rbdplugin
    spec:
      serviceAccount: rbd-csi-nodeplugin

in the above template we have added updateStrategy and its type to the daemonset spec

4.2 Update the RBD Nodeplugin daemonset

[$]kubectl apply -f deploy/rbd/kubernetes/csi-rbdplugin.yaml
daemonset.apps/csi-rbdplugin configured
service/csi-metrics-rbdplugin configured

4.3 Manual deletion of RBD Nodeplugin daemonset pods

If you determined in Pre-upgrade considerations that you were affected by the CSI driver restart issue that disconnects the application pods from their mounts, continue with this section. Otherwise, you can skip this section.

As we have set the updateStrategy to OnDelete the CSI driver pods will not be updated until you delete them manually. This allows you to control when your application pods will be affected by the CSI driver restart.

For each node:

• Drain your application pods from the node
• Delete the CSI driver pods on the node
  • The pods to delete will be named with a csi-rbdplugin prefix and have a random suffix on each node. However, no need to delete the provisioner pods csi-rbdplugin-provisioner-* .
  • The pod deletion causes the pods to be restarted and updated automatically on the node.

we have successfully upgraded RBD csi from v2.1 to v3.0

Handling node reboot hangs due to existing network mounts

With prior versions of ceph-csi a node reboot may hang the reboot process. This is due to existing cephfs or rbd network mounts not carrying the _netdev mount option. The missing mount option is fixed, and future mounts using ceph-csi would automatically carry the required flags needed for a clean reboot.

It is suggested to upgrade to the latest ceph-csi release and drain application pods on all the nodes so that new mount option _netdev can be added to all the mountpoints.