Cluster topology controller is getting stuck if a MD deletion is stuck

[sbueringer]

When a MD deletion is stuck the Cluster topology controller will also get stuck.

This can be easily seen in the Available condition

      message: |-
        * WorkersAvailable:
          * MachineDeployment clusterclass-changes-zvh1ts-md-md-0-mdh97: Deletion in progress
        * TopologyReconciled: error reading current state of the Cluster topology: MachineDeployment clusterclass-changes-idpol7/clusterclass-changes-zvh1ts-md-md-0-mdh97 Bootstrap reference could not be retrieved: failed to retrieve KubeadmConfigTemplate clusterclass-changes-idpol7/clusterclass-changes-zvh1ts-md-0-ckpjx: failed to retrieve KubeadmConfigTemplate clusterclass-changes-idpol7/clusterclass-changes-zvh1ts-md-0-ckpjx: KubeadmConfigTemplate.bootstrap.cluster.x-k8s.io "clusterclass-changes-zvh1ts-md-0-ckpjx" not found

The following happens:

• The MD topology controller deletes the templates when the MD has a deletionTimestamp
• Then the Cluster topology controller is unable to get the templates of the deleting MD

I think the ideal fix would be if we can stop retrieving the templates of a deleting MachineDeployment in the cluster topology controller.

But this requires some research

[sbueringer]

/triage accepted
/assign