/v2/clusters?action=stop shutdowns down all caches and makes the servers exit. This does not work well with a k8s StatefulSet as the controller will always attempt to have `.replica` pods ready, resulting in the pods being recreated after stop is called.

To avoid this, we should expose a /v2/clusters?action=stop-containers action that only shutdowns the container content and leaves the server process running. The StatefulSet controller can then terminate the server process when spec.replica = 0 and the pods are scaled down.