Troubleshooting Guide: Rolling Update gets stuck

[ghost]

We have experienced difficulties with cluster rolling updates that cannot be completed for some reason.
For example because the OpenStack instance cannot be created or the kubelet service on the newly created instance cannot be started.

• One way would be to eliminate the issues to complete the rolling update.
• Another way would be to delete the cluster, recreate it via create_cluster.sh and re-deploy the workloads from GitOps-repo/ backup including recovering Persistent Volumes. But that would require downtime.

But what should be done if eliminating the issues is not possible and downtime is not an option? How can the rolling update be canceled and rolled back?

Maybe some troubleshooting information and lessons learned regarding this could be shared in the Maintenance and Troubleshooting Guide.