Ceph » Orchestrator

OpenStack manila wants to use cephadm managed nfs-ganesha service to export CephFS subvolumes to OpenStack clients. Currently, OpenStack deployments use TripleO ansible deployed nfs-ganesha daemon setup in a HA active-passive configuration managed by pacemaker-corosync. If the ganesha's host goes down, the pacemaker-corosync redeploy ganesha in an available host with the same virtual IP. Since we want the cephadm managed ganesha service to be a drop-in replacement for existing ganesha service in OpenStack, we expect the cephadm managed ganesha service to be redeployed with same virtual IP in another host if its current host goes offline. Considering ganesha service is stateless it might not be complex to reschedule on another host, but cephadm might also need to fence the offline host.