Upgrading to Component Version v8.0.0

The component was updated to use HostNetwork for the Ceph-CSI driver pods and to disable holder pods.

Using HostNetwork for the Ceph-CSI driver pods avoids running into issues with existing mounts after the Ceph-CSI driver pods are restarted. Holder pods will be deprecated in one of the next Rook-Ceph releases.

This how-to is intended for users which are upgrading an existing Rook-Ceph setup to component version v8.0.0 from a previous component version.

Prerequisites

  • cluster-admin access to the cluster

  • kubectl

  • jq

Steps

  1. Upgrade the component to version v8.0.0 and wait for the upgrade to finish. Ensure the ArgoCD application is synced, in a healthy state, and the OSDs were restarted.

  2. Delete the Ceph-CSI driver configmap:

    kubectl --as=cluster-admin delete cm -n syn-rook-ceph-operator rook-ceph-csi-config

  3. Restart the Rook-Ceph operator:

    kubectl --as=cluster-admin delete pod -n syn-rook-ceph-operator -l app=rook-ceph-operator

  4. Delete the Ceph-CSI driver pods:

    kubectl --as=cluster-admin -n syn-rook-ceph-operator delete pod -l app=csi-cephfsplugin
kubectl --as=cluster-admin -n syn-rook-ceph-operator delete pod -l app=csi-rbdplugin

  5. Drain each node in the cluster to ensure no volume mounts still rely on the holder pods

    Refer to the relevant documentation for your Kubernetes distribution for commands to safely drain each node in your cluster. See the APPUiO Managed OpenShift 4 documentation for a possible approach to take for draining all nodes of an OpenShift 4 cluster.

  6. Delete holder pods

    Make sure that all pods that mount volumes have been restarted before deleting the holder daemonsets.

    		 
    kubectl --as=cluster-admin delete -n syn-rook-ceph-operator daemonsets.apps csi-cephfsplugin-holder-cluster
kubectl --as=cluster-admin delete -n syn-rook-ceph-operator daemonsets.apps csi-rbdplugin-holder-cluster

Sources