[openstack-dev] [tripleo] scenario006 conflict

Emilien Macchi emilien at redhat.com
Mon Jul 17 16:10:25 UTC 2017

On Mon, Jul 17, 2017 at 9:04 AM, Michele Baldessari <michele at redhat.com> wrote:
> On Wed, Jul 12, 2017 at 02:23:25PM -0700, Emilien Macchi wrote:
>> Hey folks,
>> Derek, it seems like you want to deploy Ironic on scenario006
>> (https://review.openstack.org/#/c/474802). I was wondering how it
>> would work with multinode jobs.
>> Also, Flavio would like to test k8s on scenario006:
>> https://review.openstack.org/#/c/471759/ . To avoid having too much
>> scenarios and complexity, I think if ironic tests can be done on a
>> 2nodes job, then we can deploy ironic on scenario004 maybe. If not,
>> then please give the requirements so we can see how to structure it.
>> For Flavio's need, I think we need a dedicated scenario for now, since
>> he's not going to deploy any OpenStack service on the overcloud for
>> now, just k8s.
>> Thanks for letting us know the plans, so we can keep the scenarios in
>> good shape.
>> Note: Numans also wants to test OVN and I suggested to create
>> scenario007 (since we can't deploy OVN before Pike, so upgrades
>> wouldn't work).
>> Note2: it seems like efforts done to test complex HA architectures
>> weren't finished in scenario005 - Michele: any thoughts on this one?
>> should we remove it now or do we expect it working one day?
> I'm a bit on the fence on this. On one side I'd love to resurrect it
> and try and complete it (previous attempts were failing due to unrelated
> issues and we never circled back to it). What we could do is catch two
> birds with one stone: namely use scenario005 for both OVN and more
> complex HA. As long as we deploy a node which hosts OVN in a dedicate
> role (be it via full pacemaker or pacemaker remote) we are exercising
> both OVN *and* the composable HA work.
> What do you think? Does that sound feasible?

Yes but warnings:

* scenario005 is a 4nodes deployments in OpenStack Infra, it takes
quite a lot of CI resources.
* we would like to run the scenario which has OVN into
openstack/networking-ovn gate.

So I'm not sure this is really efficient. Scenario005 was created to
be run when touching pacemaker files in THT / puppet-tripleo and test
composable HA. Adding OVN to the stack is fine, but it also means the
job would run much more often (every time we patch OVN files in
TripleO and also for every patch in openstack/networking-ovn) - do we
really want that?

> cheers,
> Michele
> --
> Michele Baldessari
> Email:                        <michele at redhat.com>
> C2A5 9DA3 9961 4FFB E01B  D0BC DDD4 DCCB 7515 5C6D

Emilien Macchi

More information about the OpenStack-dev mailing list