Incompatible Upgrade Guide

For the case when the desired operator version is not compatible with an existing cluster created by a previous operator version, the cluster will need to be recreated from a backup under the new operator.

This is a more general way to upgrade the operator when an in-place update is not possible.

First make a backup of the current cluster. See the backup service doc to see how to do this from inside the kubernetes cluster.

Next delete the current cluster and operator. Make sure that the cleanupBackupsOnClusterDelete field was not set to true for the cluster configuration since that would clean the backup upon deleting the cluster.

$ kubectl delete cluster <cluster-name>
$ kubectl delete deployment etcd-operator

Now modify the yaml/json spec file used to create the original operator deployment by setting the value of the spec.template.spec.containers.image field to the desired image version like .

Then simply create operator using the updated deployment spec:

$ kubectl create -f deployment.yaml

To restore from the backup, the new cluster spec will look similar to the template below depending on your type of storage used for the backup:

apiVersion: ""
kind: "EtcdCluster"
  name: <new-cluster-name> # can be the same as the old cluster name
  size: <cluster-size>
  version: <etcd-cluster-version>
    backupIntervalInSecond: <interval>
    maxBackups: <num-backups>
    storageType: <type> # S3 or PV
    backupClusterName: <previous-cluster-name>
    storageType: <type> # S3 or PV

Finally restore the new cluster from the backup of the previous cluster:

$ kubectl create -f cluster-restore-from-backup.yaml