21:02:35 <strigazi> #startmeeting containers 21:02:35 <openstack> Meeting started Tue Oct 30 21:02:35 2018 UTC and is due to finish in 60 minutes. The chair is strigazi. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:02:36 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:02:38 <openstack> The meeting name has been set to 'containers' 21:02:44 <strigazi> #topic Roll Call 21:03:01 <strigazi> who is here? 21:03:16 <eandersson> o/ 21:03:39 <imdigitaljim> o/ 21:04:38 <strigazi> hello guys 21:05:13 <strigazi> flwang seems offline 21:05:26 <strigazi> #topic Announcements 21:05:47 <strigazi> magnum 7.0.2 is ready to released, thanks tobias-urdin for pushing 21:06:00 <strigazi> https://review.openstack.org/#/c/614214/ 21:06:33 <strigazi> #topic Stories/Tasks 21:06:57 <imdigitaljim> im looking into possible fixes for what flwang was looking at last week 21:07:14 <imdigitaljim> our week at blizzard will be slow we have major event this weekend (blizzcon) 21:07:16 <strigazi> for disk storage? 21:07:19 <imdigitaljim> yeah 21:07:26 <imdigitaljim> the sandbox issue he was having 21:07:30 <imdigitaljim> when using docker volume storage 21:07:45 <strigazi> Could be mount propagation? 21:08:04 <strigazi> since the kubelet is in a chroot 21:08:18 <imdigitaljim> interesting 21:08:28 <imdigitaljim> ill look at that as well 21:08:38 <imdigitaljim> good callout 21:08:54 <strigazi> I've been burnt 21:08:58 <strigazi> multiple times 21:08:58 <colin-> o/ 21:09:00 <tobias-urdin> o/ 21:09:17 <strigazi> hello colin- tobias-urdin 21:09:54 <strigazi> I created a placeholder for the config file option for cluster creation here: 21:10:06 <strigazi> https://storyboard.openstack.org/#!/story/2004215 21:10:14 <strigazi> And I was wondering 21:10:43 <strigazi> This options will be a cli convenience or the config file will reach the API? 21:11:27 <strigazi> imdigitaljim: thoughts? 21:11:51 <imdigitaljim> yes 21:11:56 <imdigitaljim> that looks good 21:11:59 <imdigitaljim> thats a good start 21:12:04 <imdigitaljim> add a flag --config 21:12:11 <imdigitaljim> and begin deprecation soon after they all work 21:12:49 <imdigitaljim> i think as far as cli vs api 21:12:50 <strigazi> yes but, as an API payload or just cli util? 21:12:53 <imdigitaljim> thats up to us 21:13:01 <imdigitaljim> we could have the cli map the config to flags 21:13:27 <imdigitaljim> however 21:13:37 <imdigitaljim> you couldnt send a config file in an api call 21:13:42 <imdigitaljim> so we may want it in the api 21:14:04 <strigazi> I think we need more written input I guess 21:14:09 <imdigitaljim> yeah 21:14:13 <imdigitaljim> maybe POC a little of it too 21:14:22 <strigazi> Others? what is your feeling about this? 21:14:24 <imdigitaljim> experiment and see what feels more appropriate 21:15:04 <strigazi> kubectl however doesn't send the raw file 21:16:14 <cbrumm__> What sort of options would go in the config file? 21:16:17 <strigazi> kubectl does validation and sends objects to the API server 21:16:29 <strigazi> for sure all labels 21:16:47 <strigazi> and I think the cluster options too. 21:17:09 <strigazi> we just need to figure out the API microversion. 21:17:38 <cbrumm__> ok cool 21:18:09 <strigazi> If we follow the k8s example, the client must some work. 21:18:15 <strigazi> If we follow the k8s example, the client must do some work. 21:18:53 <strigazi> It is also a better option for performance. 21:19:07 <imdigitaljim> and maintainability 21:19:51 <strigazi> I'll write this down. 21:21:22 <strigazi> I brought the following up, last week as well. Can you give some input on: 21:21:57 <strigazi> Introduce magnum nodegroups https://review.openstack.org/#/c/607363/ 21:22:35 <strigazi> Also, k8s_fedora: Deploy tiller https://review.openstack.org/#/c/612336/ 21:23:25 <eandersson> We really need to figure out a good way of handling versioning in those containers. 21:24:01 <strigazi> eandersson: which one? agent or tiller? 21:24:21 <strigazi> tiller's tag is each version. 21:24:35 <strigazi> eandersson: can you elaborate? 21:24:38 <eandersson> I was referring to the heat-container-agent helm in this case (v2.11.0), but also things like pip packages in the 21:24:47 <eandersson> the same Docker file. 21:25:04 <eandersson> pip should always use upper-constraints etc 21:25:37 <eandersson> https://review.openstack.org/#/c/612336/5/magnum/drivers/common/image/heat-container-agent/Dockerfile 21:25:41 <strigazi> the OS dependencies can be managed with g-r and u-c 21:27:23 <strigazi> tiller, as long as we use helm 2.x.y can be deployed via any version of helm so far. 21:28:39 <strigazi> eandersson: if start using a version of th heat-agent we can always much it with templates version. 21:28:39 <imdigitaljim> i think you can provide to a file running in the container 21:28:51 <imdigitaljim> a parameter to use to curl out and get the specified version 21:29:00 <imdigitaljim> part of the 'startup' 21:29:13 <imdigitaljim> so you dont tie in a helm version in the heat-container 21:29:40 <eandersson> This is probably a larger discussion we should have. 21:29:49 <imdigitaljim> ^ 21:30:17 <eandersson> But for tiller, it was the main thing that I didn't like (hard coded version and url). 21:30:34 <imdigitaljim> realistically the heat-agent should work on *any* version 21:30:38 <imdigitaljim> and just be 'latest' 21:30:41 <imdigitaljim> and never change 21:31:43 <imdigitaljim> its primary job should just be interact with heat for software deployment and not be so tied to kubernetes 21:31:50 <imdigitaljim> imho 21:33:14 <strigazi> ok, got it, I'll propose another option then. Pass the url as a parameter or a script. 21:34:20 <strigazi> the heat-agent should be python OS deps only. 21:35:41 <strigazi> For nodegroups, any input? 21:36:12 <eandersson> I'll try to take a look this week, but more realistically next week (due to Blizzcon) 21:36:33 <imdigitaljim> ^ 21:36:45 <strigazi> fair enough 21:36:49 <colin-> i didn't have immediate feeedback after the comments that have been left but will look over it again strigazi 21:37:21 <strigazi> thanks colin- imdigitaljim eandersson 21:37:34 <eandersson> I pushed a couple of minor patches for devstack, one of the changing default auth_uri to www_auth_uri. 21:37:47 <imdigitaljim> also strigazi: im gonna try kubelet as just a binary on the host and see if the problem persists 21:38:06 <imdigitaljim> (the sandbox issue) 21:38:23 <eandersson> Trying to modernize the devstack config. 21:38:37 <strigazi> eandersson: I saw them, thanks for pinging too. 21:38:46 <strigazi> imdigitaljim: ack 21:39:03 <eandersson> It's possible kolla and magnum is currently not fully working, but haven't tested it myself yet, so it's speculation. 21:39:53 <strigazi> eandersson: rocky? 21:39:56 <eandersson> Maybe someone else has used kolla with magnum recently. 21:40:32 <strigazi> eandersson: do they have a devstak setup? any pointers? 21:40:54 <eandersson> https://github.com/openstack/kolla-ansible/commit/ae246945a6f9a803bdd6ccfe3d3b758deef813c5#diff-786f05c66ee3d151fb1e18e6be9ce71c 21:41:26 <strigazi> hmm, ok. that is not look good. 21:41:37 <eandersson> Should be pretty easy to get started with kolla. I can hit you up with some guides later. 21:41:41 <eandersson> Just never tried it with Magnum. 21:41:58 <eandersson> At least our devstack failed with www_authenticate_uri, but I might just have missed something. 21:44:30 <strigazi> eandersson: I'll review it and try it. 21:45:43 <strigazi> tobias-urdin: is there anything you want to discuss? 21:46:02 <eandersson> Thanks strigazi 21:48:06 <strigazi> If anyone doesn't have something else, I propose to end the meeting early. At least for me, tmr is a big day, Nova queens to rocky at 0700 UTC \o/ 21:49:16 <colin-> sounds good for me 21:50:22 <strigazi> thanks colin- 21:50:47 <strigazi> others must be drifting away :) 21:51:15 <strigazi> see you next week guys :) 21:51:31 <colin-> ttyl 21:51:38 <strigazi> #endmeeting