The example in the docs and in the architecture repo (and CI Job) has a problem. This does not require a code fix to ship but an example file fix.

Alan Bishop wrote:

I can identify at least one configuration issue. There's a volume service running in az0, az1 and az2 (I assume az0 is the "central" site) but the backup service is using the default storage_availability_zone value of "nova". Unless the tempest tests are actively aware of how to do cross-AZ backups, the backup service must be running in the same AZ as the one where the volume was created. I'm looking at the first failure (test_backup_crossproject_admin_negative[...]) and see a volume created in az0, but there's no backup service in that AZ (it's running in the "nova" AZ).

To fix this problem, the cinderBackup CR needs to be updated to include the following:

[DEFAULT]
backup_driver = cinder.backup.drivers.ceph.CephBackupDriver
backup_ceph_conf = /etc/ceph/az0.conf
backup_ceph_pool = backups
backup_ceph_user = openstack
storage_availability_zone = az0      #             <====================== Add this line