The test won't pass on serverstack due to the snapd taking a copy of the
/mnt/vdb info at boot time and thus not being able to umount it is root
without also running through all of the procs and checking if they also
have it mounted. Thus, this just disables the test on focal and newer.
When testing with TLS there is a chance the deployment will appear
done and idle prior to ceph-radosgw and Keystone have updated the
service catalog. Retry the test in this circumstance.
Increase wait attempts for availability of volume created from image
in CephRBDMirrorTest. This is taking longer as of Nautilus due to
switch to using juju storage backed by undercloud cinder taking
longer than prior method of directory-backed OSD devices.
