This is mainly for initiator zoning. During the delete/update sequence of removing target wwpns for an initiator zone, if 'activate=True', the code first does a hard delete of the zone, and then a subsequent recreation. During the time in between the deletion and creation of the zone, the VM loses all access to the volumes disks.
brcd_fc_zone_driver.py
@lockutils.synchronized('brcd', 'fcfabric-', True)
def delete_connection(self, fabric, initiator_target_map, host_name=None,
...
try:
# Update zone membership.
if zone_map:
conn.add_zones(
zone_map, zone_activate,
cfgmap_from_fabric)
brcd_fc_zone_client_cli.py
def add_zones(self, zones, activate, active_zone_set=None):
"""Add zone configuration.
for zone in zones.keys():
# If zone exists, its an update. Delete & insert
# TODO(skolathur): This still need to be optimized
# to an update call later. Now we just handled the
# same zone name with same zone members.
if (zone in zone_list):
if set(zones[zone]) == set(zone_list[zone]):
break
try:
self.delete_zones(zone, activate, active_zone_set)
except exception.BrocadeZoningCliException:
with excutils.save_and_reraise_exception():
LOG.error(_LE("Deleting zone failed %s"), zone)
LOG.debug("Deleted Zone before insert : %s", zone)
This can be avoided if the delete_zones is a transient operation, by setting activate=False. And when the zone is recreated-the subsequent cfgenable call will configure everything in one go.
This can be fixed by addressing the TODO item. Instead of doing zonedelete/