Open Stack

On 8/20/2015 10:42 AM, Matt Riedemann wrote:
>
>
> On 8/20/2015 5:33 AM, John Garbutt wrote:
>> On 20 August 2015 at 03:08, Matt Riedemann
>> <mriedem at linux.vnet.ibm.com> wrote:
>>> After spending a few hours on
>>> https://bugs.launchpad.net/nova/+bug/1370590
>>> I'm annoyed by the fact we don't yet have a CI system for testing
>>> libvirt +
>>> LXC.
>>
>> Bit thank you for raising this one.
>>
>>> At the Juno midcycle in Portland I thought I remember some guy(s) from
>>> Rackspace talking about getting a CI job running, whatever happened with
>>> that?
>>
>> Now you mention it, I remember that.
>> I haven't heard any news about that, let me poke some people.
>>
>>> It seems like we should be able to get this going using community infra,
>>> right?  Just need some warm bodies to get the parts together and
>>> figure out
>>> which Tempest tests can't be run with that setup - but we have the
>>> hypervisor support matrix to help us out as a starter.
>>
>> +1
>>
>>> It also seems unfair to require third party CI for libvirt + parallels
>>> (virtuozzo) but we don't have the same requirement for LXC.
>>
>> The original excuse was that it didn't bring much value, as most of
>> the LXC differences were in libvirt.
>> But given the recent bugs that have cropped up, that is totally the
>> wrong call.
>>
>> I think we need to add a log message saying:
>> "LXC support is untested, and will be removed during Mitka if we do
>> not get a CI in place".
>>
>> Following the rules here:
>> https://wiki.openstack.org/wiki/HypervisorSupportMatrix/DeprecationPlan#Specific_Requirements
>>
>>
>> Does that make sense?
>>
>> John
>>
>> PS
>> I must to kick off the "feature classification" push, so we can get
>> discuss that for real at the summit.
>>
>> Really I am looking for folks to help with that, help monitor what
>> bits of the support matrix are actually tested.
>>
>> __________________________________________________________________________
>>
>> OpenStack Development Mailing List (not for usage questions)
>> Unsubscribe:
>> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>
> To recap from today's nova meeting, apmelton provided the simple localrc
> [1] for getting devstack setup with lxc.
>
> He noted that there are some known issues, however:
>
> 1. nbd isn't installed by default in devstack - I think this goes back
> to the nbd + neutron ubuntu kernel panic on 12.04 back in the
> havana/icehouse timeframe.  He also said that nbd appears to leak
> resources.  I'm not entirely sure at this point if we can use something
> other than nbd like guestfs or the loop mount stuff if the image format
> is 'raw'.  We'll have to tinker.  It also sounds like Rackspace has some
> patches to workaround the nbd issues and apmelton was going to look at
> upstreaming those.
>
> 2. There is some weird intermittent issue where the network on the
> public interface just drops.
>
> --
>
> The rough plan is for me to try and get a project-config change started
> for an lxc job that we can run in nova's experimental queue.  We'll keep
> the blacklisted tempest tests in the nova tree like we do for cells.
>
> I'll probably need help with any devstack changes required.
>
> [1] https://gist.github.com/ramielrowe/081deaf0c6b79aec6890
>

I've started an etherpad to track the work needed to get an lxc job 
going in nova's experimental queue:

https://etherpad.openstack.org/p/nova-lxc-ci

-- 

Thanks,

Matt Riedemann