So proper solution would be: • stop the replica • replace data-dir with backup • run the replica • wait for health status • repeat for other replicas Do we need to do it in a proper order (starting from leader node, for example)?