https://github.com/rhcs-dashboard/ceph-dev/blob/master/docker/ceph/restart-dashboard.sh actually all you need to do is to run `ceph mgr fail <pidof mgr>`. this will restart the MGR once, instead of twice. https://github.com/ricardoasmarques/ceph-dev-docker/issues/117