Deleting IBM Storage Scale instances from the cluster

Read the following information carefully before you delete a IBM Storage Scale server instance.

CAUTION:

You must remove the IBM Storage Scale nodes and disks from the cluster before deleting the instances. Before deleting a Primary, Mirror or Tiebreak instance, you must remove it from the active-active configuration first, if it belongs to one. Failure to do so might result in the cluster losing quorum, and the IBM Storage Scale shared file system being unavailable, requiring manual intervention to clean up the cluster and reestablish quorum.

The IBM Storage Scale pattern deployment guards against this issue when the Mirror or Tiebreak (or Primary ) instances are being deleted while they are part of a IBM Storage Scale cluster. When an instance is deleted, the pattern sends a notification to the IBM Storage Scale cluster to remove the instance nodes from the active-active configuration. The cluster can then clean up deleted nodes, but only if the cluster maintains quorum while those nodes and NSD disks are being removed.

For example, if you delete both the Mirror and Tiebreak instances at the same time, even if the pattern sends the notification to the cluster to remove the Mirror and Tiebreak nodes and disks, because the remaining nodes cannot maintain quorum, the cluster fails to remove the nodes and the cluster goes down.

If the Mirror or Tiebreak instances are deleted and the cluster fails to clean up those nodes, the cluster is in an invalid state and you must remove the nodes and disks from the cluster manually.

If the Mirror or Tiebreak IP addresses are reused by a new deployment before the IBM Storage Scale cluster has a chance to clean up the Mirror and Tiebreak nodes and disks, the IBM Storage Scale cluster cannot recover, because it assumes the Mirror and Tiebreak nodes are active but unreachable (the nodes do not have the cluster keys set up). In this case you must stop the deployment that is using the Mirror or Tiebreaker IP addresses, repair the IBM Storage Scale server instance (by removing Mirror and Tiebreak nodes and disks), and then restart those instances.