On 15 Apr 2020, at 22:14, Dominik Holler <dholler@redhat.com> wrote:

On Wed, Apr 15, 2020 at 8:54 PM Michal Skrivanek <michal.skrivanek@redhat.com> wrote:

On 14 Apr 2020, at 11:05, Dominik Holler <dholler@redhat.com> wrote:

On Tue, Apr 14, 2020 at 10:25 AM Dominik Holler <dholler@redhat.com> wrote:

On Fri, Apr 10, 2020 at 8:13 AM Michal Skrivanek <michal.skrivanek@redhat.com> wrote:

On 9 Apr 2020, at 10:35, Dominik Holler <dholler@redhat.com> wrote:

On Wed, Apr 8, 2020 at 5:34 PM Michal Skrivanek <michal.skrivanek@redhat.com> wrote:

On 8 Apr 2020, at 14:55, Dominik Holler <dholler@redhat.com> wrote:

On Wed, Apr 8, 2020 at 2:48 PM Michal Skrivanek <michal.skrivanek@redhat.com> wrote:

On 8 Apr 2020, at 13:52, Dominik Holler <dholler@redhat.com> wrote:

On Wed, Apr 8, 2020 at 1:35 PM Michal Skrivanek <michal.skrivanek@redhat.com> wrote:

> On 8 Apr 2020, at 11:32, Michal Skrivanek <michal.skrivanek@redhat.com> wrote:
>
> Eh, so no, still not correct. Steven/Shmuel, I guess it could be your [1] or maybe other related recent patches from you, but OST is now broken (again).
> With [2] finally fixing the cluster creation OST now runs with a Q35 seabios (as it was supposed to, but wasn’t until now), and the vm2 which has a custom emulated machine of i440fx-rhel-7.4.0 doesn’t run anymore, because apparently it is launched as a q35 VM for the first time, and then failing restart ever since.

Sorry, my bad, it’s really getting confusing with the amount of breakages:)
It should be fixed just in OST because using a custom i440fx type in a Cluster with q35/seabios is invalid.
Well, that’s easy...

If I run networking-suite-master with additional repo

https://jenkins.ovirt.org/job/ovirt-engine_standard-on-merge/865/

the VMs will use

pc-i440fx-rhel8.1.0 type [1].

Is this the same problem, or is this another one?

I don’t know, you didn’t say what problem you've seen

pc-i440fx-rhel8.1.0 type seems to be not valid.

it’s not. I don’t know how/where you create the VMs or the Cluster in network suite. You need to check it’s using the default and not something hardcoded…

It is using defaults.

It is also easy to reproduce, just import the cirros image as a template and create a VM with "other OS" from this template.

ah, glance import, yeah, that seems to be creating wrong templates. you could tell easily in UI it’s asking you that there’s a disrepancy between template’s and cluster’s chipset.

- glance import should use q35 because that’s the new default

- vm from template should drop the devices when there’s a difference in cluster

Needs to be fixed.

Is there already a bug reported, or should I report one?

I just reported https://bugzilla.redhat.com/1823674, to have a justification for the required code change in OST network suite to introduce the workaround.

Dominik,
we(me manually, basic suite, QE manually and using REST API) were not able reproduce your issue. Can you confirm it’s happening on a recent build. Not anything in CQ, I mean a recent nightly or recently merge artifacts of ovirt-engine. I looked at your logs and it looked like all the patches should have been there...but I don’t have any other explanation.

Today I built engine e946364377a49b1ff019dc6b2f77c61fb27e48c5 as developer installation, did a clean build, cleaned the db, run engine-setup, imported the CentOS 8 image as template.

I also tried CentOS 7 image on the latest rpm version 4.4.0-0.5.beta3.20200408120550.gitf94f968ca81.el8 , which results in the same behavior.

The VM created from the template did not boot because of the issue.

On creating the VM from the template on UI, I get the suspicious warning:

"""

Devices Will Be Changed

In order to create a VM from a template with a different chipset, device configuration will be changed. This may affect functionality of the guest software. Are you sure you want to proceed?

Even if I neither modify the template and set only a name for the new VM.

"""

I attached some db tables which might be relevant, please let me know if I can provide further information.

Thanks.