Improvements
- Cluster deletion: now after we’ve deleted the load balancer—which must be removed before the masters—we switch to the
master1's kubeconfig context. This allows us to continue identifying (via kubectl) and deleting autoscaled nodes as well static nodes. Previously, autoscaled nodes were left behind because the load balancer was no longer available at that point, so we couldn't reach the Kubernetes API to detect autoscaled nodes. - When removing unused placement groups, we now focus only on those whose names begin with the cluster name. This helps prevent accidentally deleting placement groups that were created by the user or other tools within the same project.