15:02:32 <anteaya> #startmeeting third-party
15:02:32 <openstack> Meeting started Mon Aug 17 15:02:32 2015 UTC and is due to finish in 60 minutes.  The chair is anteaya. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:02:33 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:02:35 <openstack> The meeting name has been set to 'third_party'
15:03:29 <anteaya> so hello
15:03:39 <anteaya> and I hope everyone is well today
15:03:48 <anteaya> as I said I am about to board a plane
15:04:02 <anteaya> so if anyone has anything to share do speak up
15:05:20 <wznoinsk> hi anteaya, nothing hear, only tunning in
15:05:26 <wznoinsk> here* even
15:05:32 <anteaya> wznoinsk: hi
15:05:37 <anteaya> thanks for tuning in
15:05:48 <anteaya> I have no announcements
15:05:57 <tim_o> same here, just tuning in
15:06:09 <anteaya> wznoinsk: willing to allow me to make you chair in case I lose service?
15:06:23 <wznoinsk> yes, no probs
15:06:37 <asselin> o/
15:06:43 <wznoinsk> or asselin better
15:06:52 <anteaya> #chair wznoinsk asselin
15:06:53 <openstack> Current chairs: anteaya asselin wznoinsk
15:06:55 <anteaya> you both are
15:06:57 <anteaya> thank you
15:07:11 <anteaya> I lost service when I tried to make you chair?
15:07:20 <anteaya> s/?/:)
15:07:28 <anteaya> hi tim_o thanks for tuning in
15:07:32 <anteaya> asselin: welcome back
15:07:50 <asselin> good morning
15:07:53 <anteaya> asselin: anything you want to share today?
15:08:01 <anteaya> it is a good morning :)
15:08:24 <wznoinsk> a long shot but did anyone hear of/used himself,herself/saw someone using containers for a CI? asking because we did it and somehow works ok for use, I was wondering if there are others interested
15:08:35 <anteaya> wznoinsk: nice
15:08:42 <anteaya> wznoinsk: I think others would be
15:08:57 <anteaya> wznoinsk: would you like to share some of the information about your setup?
15:09:28 <anteaya> what would be the best? a mailing list post to -infra? an etherpad and a link to a mailing list post?
15:09:38 <anteaya> wznoinsk: what might make sense to you?
15:09:47 <wznoinsk> it's a simple one because we don't use any provisioning tool (like nodepool for VMs), we spawn containers on the jenkins slaves themselves
15:09:57 <anteaya> interesting
15:10:09 <anteaya> perhaps a post to the infra mailing list sharing the setup
15:10:10 * wznoinsk digs out an old irc meeting log
15:10:30 <wznoinsk> http://eavesdrop.openstack.org/meetings/third_party/2015/third_party.2015-03-18-15.00.log.html
15:10:40 <anteaya> and then if more info is required folks reading the mailing list post can share their thoughts about that
15:11:01 <wznoinsk> yes, it sounds like a good idea, I can send something to ML and see the response
15:11:07 <anteaya> is there a time stamp that is useful for that log?
15:11:15 <anteaya> wznoinsk: I support that direction
15:11:20 <anteaya> that would be great, thank you
15:11:25 <asselin> wznoinsk, I remember this presentation
15:12:25 <anteaya> asselin: what do you think, would others be interested?
15:12:29 <wznoinsk> asselin: good to hear ;-)
15:13:01 <rhe00> I would be interested
15:13:07 <asselin> anteaya, yes I believe so
15:13:08 <anteaya> rhe00: thank you
15:13:12 <anteaya> asselin: awesome
15:13:20 <anteaya> I look forward to your post wznoinsk
15:14:13 <anteaya> anything more on this topic?
15:14:18 <wznoinsk> ...
15:14:37 <anteaya> let's move on
15:14:49 <anteaya> does anyone have anything else they would like to discuss today?
15:14:50 <wznoinsk> one case CI may be facing is access to the physical hardware on the host to test specific functionality, would you guys know is it pci passthrough used or rather baremetal stacking?
15:15:37 <wznoinsk> rhe00: check the above log and http://pastebin.com/972cE2mc and shout if questions
15:15:38 <asselin> wznoinsk, not sure I understand your question, but we use pci passthrough for cinder fibre channel driver tests
15:15:55 <anteaya> sorry I should link that url
15:15:57 <wznoinsk> ok, is it anyhow problematic/needs workaround or not?
15:15:58 <anteaya> #link http://eavesdrop.openstack.org/meetings/third_party/2015/third_party.2015-03-18-15.00.log.html
15:16:16 <wznoinsk> <- asselin
15:16:20 <anteaya> #link http://pastebin.com/972cE2mc
15:17:05 <asselin> wznoinsk, it's been working fine for us and patrickeast. not sure who else.
15:18:07 <wznoinsk> asselin: it sounds easy then, we on the other hand, can't use VMs as they have only one numa node and we run test using at least two
15:18:13 <anteaya> time for me to board
15:18:15 <anteaya> thank you
15:18:19 <anteaya> please continue
15:18:23 <anteaya> I will read the log
15:18:27 <wznoinsk> anteaya: have a good flight
15:18:29 <anteaya> see you next week
15:18:31 <anteaya> thank you
15:18:52 <asselin> anteaya, bon voyage
15:19:25 <lennyb> Hi
15:19:27 <wznoinsk> I'm then wondering what other cases can be 'not possible' with VMs that could potentially be workable when using containers for the builds
15:20:20 <wznoinsk> hi lennyb
15:20:34 <asselin> wznoinsk, I see containers as a way to possible improve ci performance
15:21:31 <wznoinsk> asselin: that's for sure, you have to be carefull with some 'shared' resources when you use containers tho, kernel/modules, hugepages etc.
15:21:53 <wznoinsk> all depends whether you run some exotic tests obviously, but that's probably not your case
15:23:23 <wznoinsk> if anyone has any questions about ci in containers (docker) I'm happy to sort you out or die trying
15:23:33 <asselin> it would be nice if nodepool could also provision containers
15:24:16 <wznoinsk> asselin: do you know whether there's any movement for it there?
15:24:38 <asselin> wznoinsk, I don't think there is.
15:24:56 <asselin> I'm not sure upstream can use containers with their public clouds
15:26:12 <mmedvede> In theory, if openstack supports docker containers, we should be able to reuse nodepool for third-party maintained cloud?
15:28:34 <wznoinsk> mmedvede: you mean provision vms in 3rd party cloud using nodepool and openstack to start containers inside?
15:29:21 <mmedvede> wznoinsk: yes. In theory, it should look the same to nodepool, so no reason it should not be possible
15:29:52 <mmedvede> I did not try docker OpenStack hypervisor. There might be quirks
15:30:54 <wznoinsk> I think the ideal solution is if nodepool would support containers itself, then less failure points and overhead
15:32:12 <asselin> wznoinsk, I believe the nodepool design can support containers. It already supports vms & baremetal provisioning.
15:33:48 <mmedvede> asselin: was not aware of baremetal provisioning - so it does not use OpenStack for that?
15:34:06 <wznoinsk> it looks like it could be a quick win here
15:34:09 <asselin> mmedvede, it should use ironic
15:35:10 <mmedvede> afaik, nodpool only can do what OpenStack can
15:35:12 <wznoinsk> would you know how much/less difficult it would be to have openstack provisioning + cloudinit script instead of nodepool?
15:36:49 <asselin> wznoinsk, I think they're the same
15:37:19 <wznoinsk> I realized it just now, thanks
15:38:22 <asselin> nodepool tries to keep a supply of nodes available. It uses openstack apis to manage node lifecycle
15:42:06 <asselin> any other topic?
15:46:24 <asselin> wznoinsk, just saw this
15:46:29 <asselin> #link https://www.linkedin.com/pulse/dynamically-add-jenkins-slaves-using-kubernetes-yolanda-robla
15:47:20 <wznoinsk> looks interesting
15:47:34 <wznoinsk> btw. is nodepool in (or planned) in downstream ci?
15:48:02 <asselin> nodepool is planned to downstream-puppet
15:49:53 <wznoinsk> it looks like I have enough to read about now, that's all from me for this meeting
15:50:49 <asselin> anyone else have topics/questions to bring up?
15:54:29 <asselin> #end-meeting
15:54:47 <asselin> #endmeeting