SVC storwize regression after fixing 1913363

[Wojciech]

A lot of errors
 Stderr:
'CMMVC6578E The command has failed because the iSCSI name is already assigned or is not valid.\n': oslo_concurrency.processutils.ProcessExecutionError: Unexpected error while running command.
Stderr: 'CMMVC6578E The command has failed because the iSCSI name is already assigned or is not valid.\n': oslo_concurrency.processutils.ProcessExecutionError: Unexpected error while running command.
 stderr: CMMVC6578E The command has failed because the iSCSI name is already assigned or is not valid.
 stderr: CMMVC6578E The command has failed because the iSCSI name is already assigned or is not valid.
 stderr: CMMVC6578E The command has failed because the iSCSI name is already assigned or is not valid.
,
probably after https://bugs.launchpad.net/cinder/+bug/1913363
even on small env
this race condition causes volumes to not be able to attach to vm, ending in error state
happy to provide more logs/tests if needed.

[Sofia Enriquez (lsofia-enriquez)]

Greetings,
Would you mind sharing the steps to reproduce?
Thanks in advance

[Wojciech (suzumushi)]

Deploy openstack (ussuri in this case) with Availability zones in separate geo locations
each geo location has it's own ceph and svc storage, managed by 3 cinder instnaces

and i think this could be reason for that error, and could be related to

https://bugs.launchpad.net/cinder/+bug/1740021/

but need confirmation that there is no possibility to use more than one cinder instance to manage one svc.
in our case we have 50+ nodes per AZ, and one cinder instance basically disqualifies usage of svc as storage backend.
If it's what i think, is there any plan to implement A/A for svc driver ?

regards

w