Hi Strahil, 

Thanks for your reply!

Since we are using 4.3.9 from the installation ISO, issues regarding CentOS 8 rolling releases do not apply here.
We will do that upgrade sooner or later, but then i would try to do local yum repository mirrors that are always snapshots. More or less how you suggest…

So my success rate of 4 in 10 attempts has nothing to do with changing package or Ovirt versions. 

It happens when applying exactly the same procedure, scripts, installation media, on the exact same machine. 
So the input is always the same, but the result is at least 2 different error states in 6 cases, and only 4 times a working system.

I would like to understand if that is normal and everybody needs their handful of attempts to deploy the engine until it succeeds or if this is a sign that we still do something wrong that we need to track down and fix.


On Sat 30. Oct 2021 at 14:31, Strahil Nikolov <hunter86_bg@yahoo.com> wrote:
If you want to increase your deployment success, you will need to use repository management and freeze your OS & oVirt repos to a working level .
For example if you use RHEL 8.4 and current level of oVirt - you will have dependency issues until RHEL 8.5 is released.

Once this happens and your build succeeds, you will lock your repos and deploy from them again.Once you test each 'batch' of repos and confirm they work for you -> you will create a new set of repos ...

oVirt is dynamic project with new features constantly comming up (and sometimes go away).

Another approach is to use the ovirt Node image which is based on CentOS Stream and is validated in the dev infrastructure .

Best Regards,
Strahil Nikolov

On Wed, Oct 27, 2021 at 21:39, Henning Sprang

I've just inherited a project where we need to bring a prototype of a
small Ovirt system (single node or 3 node hyperconverged, with
glusterFS on the same machine, a bunch of different VM's ) running in
an industrial machine into serial production.

This means, we want to build a new 1 or 3 node Ovirt system each day
up until 3 times a day.

In my tests so far, the failure rate of the Ovirt engine deployment
(via the included scripts as well as the web UI) turns out to be
pretty high - it's between 40-60%, meaning until we have a running
system, we would have to try the installation and/or final engine
deployment about 2-4 times until we are successful.

So far I could not identify clear error messages that let me tell how
to fix the problem.

Before going into details of the errors I would like to ask if people
deeper into Ovirt would consider this a somewhat normal success rate,
or if this indicates we are doing something generally wrong and we
should definitely spend a few more hours or maybe days into finding
sources of problems.

More info about the system and errors

* OVirt 4.3.9 (because the prototype was made and verified with that
version - would be interesting to know, too, if it's strongly
considered to upgrade for more stable installation/deployment)
* The errors that appear are changing between the deployment process
seeming not to be able to transfer the "LocalHostedEngine" VM to the
glusterFS storage to become a "HostedEngine", and the other seems to
be when the engine is already up and running, but never being really
connected to the Ovirt system, continuously restarting, and also
showing XFS filesystem errors in it's dmesg output.

Any hints on our chances on getting this solved or requests for more
information about the error are welcome - thanks in advance.

Henning Sprang