Currently the operator allows custom server images to be specified via the spec.image field, resulting in existing clusters being restarted via a GracefulShutdown and all in-memory entries being lost. This is not ideal for integrating patched server images, as the majority of patches will not break cluster communication compatibility and so a rolling upgrade of the cluster should be possible.

To ensure that the necessity for a Graceful shutdown vs Rolling Upgrade can be determined on a per patched image basis, with no changes to the Operator code required, we can add image inspection capabilities to the Operator. The Operator can retrieve the metadata associated with the spec.image and utilises container Labels to determine what type of upgrade is possible. Skopeo allows containers to be inspected remotely without having to pull the containers content to the local FS.

Adding such capabilities to the Operator means that it's possible for DG patches to be deployed on customer sites without the need to also patch the Operator.