Description

When upgrading cluster versions, in-tree driver Volumes will automatically be updated to use the appropriate CSI node drivers by the cluster.
In order for this to happen with no noticeable workload interruption, the CSI node driver must be running on a node before CSI migration is enabled for the kubelet running on the node.

If migration is not possible due to the node driver not running on the node, WMCO should block an operator upgrade until it is, to prevent workload interruption.

Engineering Details

• Upgrade should be blocked by WMCO's OperatorCondition.
• csi-proxy.exe must be running on a node before the CSI driver pod will go to running.
• On a cluster with CSI drivers installed, a CSINode object will be created for each node, who's status can be checked to determine if the node has a running CSI node driver.

Acceptance Criteria

• If a Windows node has a persistent volume claim bound to it, upgrades for that node are blocked until the cluster administrator deploys a CSI driver daemonset to allow the volumes to be migrated.