I have two Kubernetes deployments that each creates a pod. I need to make sure that the pods run in different nodes in the cluster.
I don’t want to use node labelling because I need to make sure that if the cluster is restarted the pods will be deployed (as far as I have tested, when I restart the cluster, at least in AKS, the nodes loose the labels that I create to solve this problem using nodeSelector.
Until I figure out how to have persistent node labeling, is there a way to be 100% sure that the pods from the two deployments will be scheduled to different nodes in the cluster?