Readd RBD lock detection and clearing on startup
This is still needed due to the nature of the locks and freeing them on startup, and to preserve lock=fail behaviour on VM startup. Also fixes the fencing lock flush to directly use the client library outside of Celery. I don't like this hack but it seems prudent until we move fencing to the workers as well.
This commit is contained in:
@ -23,7 +23,7 @@ import time
|
||||
|
||||
import daemon_lib.common as common
|
||||
|
||||
from pvcnoded.objects.VMInstance import VMInstance
|
||||
from daemon_lib.vm import vm_worker_flush_locks
|
||||
|
||||
|
||||
#
|
||||
@ -121,7 +121,7 @@ def migrateFromFencedNode(zkhandler, node_name, config, logger):
|
||||
|
||||
# Migrate a VM after a flush
|
||||
def fence_migrate_vm(dom_uuid):
|
||||
VMInstance.flush_locks(zkhandler, logger, dom_uuid)
|
||||
vm_worker_flush_locks(zkhandler, None, dom_uuid, force_unlock=True)
|
||||
|
||||
target_node = common.findTargetNode(zkhandler, dom_uuid)
|
||||
|
||||
|
Reference in New Issue
Block a user