2014-05-09 15:55 GMT+02:00 Nicolas Ecarnot <nico...@ecarnot.net>:
> Hi,
> On our second oVirt setup in 3.4.0-1.el6 (that was running fine), I did a
> yum upgrade on the engine (...sigh...).
> Then rebooted the engine.
> This machine is hosting the NFS export domain.
> Though the VM are still running, the storage domain is in invalid status.
> You'll find below the engine.log.
> At first sight, I thought it was the same issue as :
> http://lists.ovirt.org/pipermail/users/2014-March/022161.html
> because it looked very similar.
> But the NFS export domain connection seemed OK (tested).
> I did try every trick I could thought of, restarting, checking anything...
> Our cluster stayed in a broken state.
> On second sight, I saw that when rebooting the engine, then NFS export
> domain was not mounted correctly (I wrote a static /dev/sd-something in
> fstab, and the iscsi manager changed the letter. Next time, I'll use LVM or
> a label).
> So the NFS served was void/empty/black hole.
> I just realized all the above, and spent my afternoon in cold sweat.
> Correcting the NFS mounting and restarting the engine did the trick.
> What still disturbs me is that the unavailability of the NFS export domain
> should NOT be a reason for the MASTER storage domain to break!
> Following the URL above and the BZ opened by the user
> (https://bugzilla.redhat.com/show_bug.cgi?id=1072900), I see this has been
> corrected in 3.4.1. What gives a perfectly connected NFS export domain, but
> empty?

sorry for jumping late on an old thread, I'm the one reporting that bug.
I have two things to say:
- taking advantage of a rare opportunity to turn off my production
cluster I put it back in that critical situation and I can confirm
that with oVirt 3.4.1 the problem has been solved.

> PS : I see no 3.4.1 update on CentOS repo.

- me too, until I installed ovirt-release34.rpm (see
http://www.ovirt.org/OVirt_3.4.1_release_notes ). All went smooth
after that.

Best Regards,
Users mailing list

Reply via email to