Open Stack

On 13/06/17 13:00, Ben Nemec wrote:
> 
> 
> On 06/12/2017 03:53 PM, Zane Bitter wrote:
>> On 12/06/17 16:21, Steven Hardy wrote:
>>>> I think we wanted to move to convergence anyway so I don't see a 
>>>> problem
>>>> with this.  I know there was some discussion about starting to test 
>>>> with
>>>> convergence in tripleo-ci, does anyone know what, if anything,
>>>> happened with
>>>> that?
>>> There's an experimental job that runs only on the heat repo
>>> (gate-tripleo-ci-centos-7-ovb-nonha-convergence)
>>>
>>> But yeah now seems like a good time to get something running more
>>> regularly in tripleo-ci.
>>
>> +1, there's no reason not to run a non-voting job against tripleo itself
>> at this point IMHO. That would allow me to start tracking the memory use
>> over time.
> 
> Do you have a strong preference multinode vs. ovb?  I would tend to 
> think we want this to be ovb since multinode stubs out a bunch of stuff, 
> but at the same time ovb capacity is limited.  It's better* now that 
> we've basically halved our ovb ci coverage so we could probably get away 
> with adding a job to a specific repo though (t-h-t seems logical for 
> this purpose).

The job I've been tracking memory usage on against t-h-t was 
tripleo-ci-centos-7-ovb-nonha... which I see no longer exists (doh!). So 
I guess we can pick whatever it makes sense to track over the long term.

We want something fairly representative, so that we can be confident to 
flip the switch early in Queens development. I don't think it especially 
matters whether we're using baremetal servers or VMs though... that 
should be fairly transparent to Heat.

cheers,
Zane.