On 31.10.2014 12:02, Markus Stockhausen wrote:
Hi,
maybe a stupid one. Just want to make sure that nothing goes wrong. We
plan to make a rolling upgrade of our landscape going to 3.5.1 in december.
So we always need some hypervisor nodes up and running during the
process.
This went fairly well with me. I did not loose a single machine and the
process was transparent for them (I even had a script running lots of
queries against a web server in one vm).
Looking at older posts I assume that upgrading the engine should be
sufficient, while hypervisor nodes with vdsm 4.14.17 (from 3.4 repos) and
the active VMs keep on running. The start of the engine after the upgrade
should catch up and resync all states.
The Engine and VDSM are fully backward
compatible to each other [1]. So
you should not see any problems there; only missing some features maybe.
Afterwards I clean hypervisor hosts one by one and clean/reinstall it into
the cluster?
This what I have done; remember you need to add the 3.5 repos
otherwise
VDSM will be (re)installed from the latest 3.4 repos and not upgraded to
the '3.5' version. Also, this process went like a charm without any
issues (let alone loosing guest)
Thanks in advance.
Please keep in mind, upgrading your datacenter(s) to 3.5 might
lead to
some issues, some of them already fixed for 3.5.1 (I had some more
severe ones). I would be careful with this setting in a production env.
Also, some featues are available on cluster versions only; not requiring
the more invasive datacenter upgrade [2].
[1]
https://bugzilla.redhat.com/show_bug.cgi?id=1154620#c3
[2]
http://www.mail-archive.com/users@ovirt.org/msg22249.html
--
Daniel Helgenberger
m box bewegtbild GmbH
P: +49/30/2408781-22
F: +49/30/2408781-10
ACKERSTR. 19
D-10115 BERLIN
www.m-box.de www.monkeymen.tv
Geschäftsführer: Martin Retschitzegger / Michaela Göllner
Handeslregister: Amtsgericht Charlottenburg / HRB 112767