Gitops - all kubernetes configuration is represented in the form of manifests (yaml files) so instead of making stateful backups using etcd snapshot - put your kubernetes manifests in git and use a tool like fleet to deploy the required manifests to the correct clusters. That way if the cluster is destroyed you can just redeploy the entire configuration of the cluster and let kubernetes work for you - treat your infrastructure like code by versioning and treat your cluster as cattle not pets rancher employee