Am 24.11.19 um 11:58 schrieb Benny Zlotnik:
The current plan to integrate ceph is via cinderlib integration[1]
(currently in tech preview mode) because we still have no packaging
ready, there are some manual installation steps required, but there is
no need to install and configure openstack/cinder
> 1. Does this require you to install OpenStack, or will a vanilla Ceph installation
work?
a vanilla installation will work with cinderlib
> 2. Is it possible to deploy Ceph on the same nodes that run oVirt? (i.e. is a 3-node
oVirt + Ceph cluster possible?)
I haven't tried it, but should be possible
> 3. Is there any monitoring/management of Ceph from within oVirt? (Guessing no?)
No, cinderlib is storage agnostic
> 4. Are all the normal VM features working yet, or is this planned?
Most features (starting/stopping/snapshots/live migration) are
working, but not all are fully tested (specifically snapshots)
> 5. Is making Ceph a first-class citizen (like Gluster) on oVirt on the roadmap?
Not at the moment, maybe once cinderlib integration matures and we
have more feedback and users for the feature
[1]
https://www.ovirt.org/develop/release-management/features/storage/cinderl...
We are using Ceph with oVirt (via standalone Cinder) extensively in a
production environment.
I tested oVirt cinderlib integration in our dev environment, gave some
feedback here on the list and am currently waiting for the future
development. IMHO cinderlib in oVirt is currently not fit for production
use, I think this matches your assessment.
What can we do to help advance Ceph integration in oVirt?
What are the plans for oVirt 4.4?
Will standalone Cinder storage domains still be supported in oVirt 4.4?
Will there be a migration scenario from standalone Cinder to cinderlib?
Accidentally just yesterday I had an incident in our test environment
where migration of a VM with MANAGED_BLOCK_STORAGE (=cinderlib) disks
failed (causes are known and unrelated to cinderlib). Restarting the VM
failed because of leftover rbdmapped devices. This is similar to the
case I reported in
https://bugzilla.redhat.com/show_bug.cgi?id=1697496.
I don't clearly see if this fixed or not. Shall I report my recent problem?
thanks
Matthias