16:01:23 <rhallisey> #startmeeting kolla 16:01:24 <openstack> Meeting started Wed Jul 6 16:01:23 2016 UTC and is due to finish in 60 minutes. The chair is rhallisey. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:26 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:28 <openstack> The meeting name has been set to 'kolla' 16:01:31 <rhallisey> #topic rollcall 16:01:33 <rhallisey> hello 16:01:34 <inc0> o/ 16:01:37 <wirehead_> o/ 16:01:38 <mliima_> hi :) 16:01:39 <Jeffrey4l_> o/ 16:01:39 <coolsvap> o/ 16:02:13 <rhallisey> alrighty then 16:02:17 <rhallisey> #topic Announcements 16:02:39 <rhallisey> Summit talks are due on the 12th 16:02:56 <rhallisey> if anyone is planning a talk, there are 6 days left to submit 16:03:22 <inc0> Midcycle is next week 16:03:34 <vhosakot> o/ 16:03:41 <rhallisey> if you have anything you are thinking about, mention it with some people and maybe we can help the idea along 16:03:46 <rhallisey> inc0, yes thanks 16:03:52 <rhallisey> midcycle is also next week 16:03:55 <pbourke> o/ 16:04:05 <rhallisey> in North Carolina 16:04:07 <wirehead_> If you haven’t gotten tickets for midcycle and get plane tickets now, you will probably be in trouble with your boss. 16:04:20 <rhallisey> :) 16:04:24 <rhallisey> #topic Customization of dockerfiles 16:04:35 <rhallisey> inc0, anything to add for this? 16:04:47 <inc0> yeah 16:05:08 <inc0> so we need to start doing work of adding placeholders for customizations 16:05:14 <inc0> let me show you an example 16:05:36 <inc0> https://review.openstack.org/#/c/329651/ 16:06:14 <coolsvap> #link https://review.openstack.org/#/c/329651/ 16:06:29 <akwasnie> Hi, sorry for being late! 16:06:31 <rhallisey> this work can be divided among the community 16:06:41 <rhallisey> inc0, is thet BP setup for that? 16:06:42 <Jeffrey4l_> inc0, how to add a new package? 16:07:12 <inc0> rhallisey, we don't have bp with work items 16:07:16 <inc0> I'll do it after meeting 16:07:16 <Jeffrey4l_> for example, add `tmux` to the heat container. ( even though it is meaningless) 16:07:34 <rhallisey> inc0, ok let's get that out there. Fix up that patch and merge it. 16:07:36 <inc0> Jeffrey4l_, hold on 16:07:51 <rhallisey> once we have that one done the work can be spread out 16:08:09 <Jeffrey4l_> ok 16:08:18 <inc0> you'll need to add {{ heat_api_packages_append = ["tmux"] }} in customization file 16:09:12 <inc0> ok, so I'll start with writing docs:) 16:09:13 <Jeffrey4l_> inc0, where is the customization file locate? 16:09:32 <inc0> build.py --template-override path-to-customization-file 16:09:47 <rhallisey> inc0, yes please :) 16:10:08 <inc0> #action inc0 to provide docs for customization 16:10:28 <rhallisey> ok cool 16:10:29 <inc0> #action inc0 to provide blueprint with work items 16:10:45 <inc0> ok, that's on me, after that take on items guys 16:11:04 <rhallisey> inc0, ya just make a writeup of the work items and we can spread them out 16:11:10 <rhallisey> #topic Gating 16:11:30 <rhallisey> dmsimard write another email about gating that I think brought up some good points 16:11:35 <rhallisey> s/write/wrote 16:12:14 <rhallisey> as for moving towards voting gates, I've seen a lot more green lately 16:12:33 <rhallisey> Jeffrey4l_, I know you've been working hard on that :) 16:12:35 <rhallisey> nice job 16:12:51 <coolsvap> Jeffrey4l_ +1 nice 16:12:53 <rhallisey> there is a new issue that I've seen: http://logs.openstack.org/51/298451/21/check/gate-kolla-dsvm-deploy-centos-binary/f212710/console.html#_2016-07-06_01_49_11_762175 16:12:54 <Jeffrey4l_> yes. the gate is more stable now. 16:13:08 <rhallisey> issue with neutron-ovs-agent 16:13:12 <wirehead_> I’m assuming we can’t dip a toe in by creating an artificial gate that votes no if all of the ‘deploy’ or ‘build’ gates are red. 16:13:39 <Jeffrey4l_> rhallisey, this fix is under review https://review.openstack.org/336352 16:13:46 <rhallisey> awesome 16:14:06 <Jeffrey4l_> one question: should we move the xenial in the CI for ubuntu? 16:14:08 <rhallisey> so tieing into what I first mentioned about that email, getting out gates to stable is one step in the right direction 16:14:18 <rhallisey> beyond that we need to extend out gate coverage 16:14:40 <rhallisey> it's something that's going to propel the general stability of the project 16:14:49 <Jeffrey4l_> move the/move to 16:15:17 <inc0> Jeffrey4l_, we should be good 16:15:27 <inc0> I didnt experience any issues with xenial yet 16:16:00 <Jeffrey4l_> inc0, if we move to xenial, all the branch will use it. 16:16:09 <Jeffrey4l_> what do you guys think this? 16:16:11 <inc0> yeah, but as base system 16:16:25 <Jeffrey4l_> include liberty, mitaka and newtwon 16:16:47 <rhallisey> I think that sounds reasonable 16:16:54 <Jeffrey4l_> now, one issue from ubuntu kernel block the ubuntu 16.04 as the base image. 16:17:03 <Jeffrey4l_> https://review.openstack.org/329454 16:18:27 <rhallisey> ok good to know 16:18:39 <rhallisey> one more think per the email 16:18:42 <rhallisey> #link https://github.com/openstack/packstack#packstack-integration-tests 16:18:54 <rhallisey> those are the tests linked that packstack does 16:19:11 <rhallisey> it worth to think about in terms of ideas for gate coverage 16:19:16 <Jeffrey4l_> yes. 16:19:52 <rhallisey> 1) stability 2) getting minimal coverage 16:19:57 <Jeffrey4l_> but additional, kolla has more matrix beyond that. like ubunt+source cent0S+source centos+binary etc 16:20:04 <rhallisey> yes it does 16:20:07 <rhallisey> which is awesome 16:20:45 <rhallisey> in total we have a lot of gate, but we have some hold when it comes to basic coverage 16:20:58 <rhallisey> well talk a bunch more about this at the midcycle 16:21:04 <rhallisey> #topic Kolla-kubernetes 16:21:20 <rhallisey> #link https://review.openstack.org/#/c/335279/ 16:21:26 <rhallisey> #link https://review.openstack.org/#/c/334255/ 16:21:47 <rhallisey> the first link is a spec to ansible as a deployment tool that hands off to kube 16:22:10 <rhallisey> there is some good discussion ongoing there that could be worth jumping into if you haven't already 16:22:26 <rhallisey> wirehead_, did you want to mention anything 16:22:32 <rhallisey> or anyone else for that matter? 16:22:54 <wirehead_> Well, Kube 1.3 has init containers that might solve some of our workflow issues in a better fashion. 16:23:07 <wirehead_> I think I wrote enough paragraphs for the next day or two on the spec. :D 16:23:11 <rhallisey> hehe 16:23:32 <rhallisey> wirehead_, think we can get a poc of that? To compare? 16:23:36 <wirehead_> Yeah. 16:24:00 <inc0> yeah, init containers might actually be good for you guys 16:24:08 <rhallisey> wirehead_, something to keep in mind too is that the init containers would need to work for both bootstrap and upgrades 16:24:10 <wirehead_> gherlien submitted the doc patch for kube 1.3 instructions yesterday and I’m running my tests. 16:24:32 <inc0> rhallisey, I think that's still doable 16:25:21 <rhallisey> I mean it's technically do able with init containers, but we want to be sure it's the best possible method 16:25:33 <wirehead_> We might end up taking both options. 16:25:41 <rhallisey> avoid auto magic and make sure were flexibile 16:25:43 <inc0> I'd be opposed to do both 16:25:53 <inc0> as we want to make one but good 16:25:56 <rhallisey> inc0, so init + ansible? 16:26:10 <wirehead_> c.f. MariaDB using init containers to get started, because it’s a snowflake, but database migrations using ansible. 16:26:11 <inc0> wirehead_, are you going to midcycle? 16:26:13 <wirehead_> Yes. 16:26:19 <inc0> let's talk there shall we? 16:26:31 <rhallisey> I'm not sure if I'm opposed.. it could move us closer to a state were kube could one day handle the whole thing 16:26:40 <rhallisey> idk 16:26:45 <inc0> rhallisey, opposed for both at same time 16:26:56 <inc0> I think we need to make a call 16:27:08 <rhallisey> ok let's have a poc of it to see how it works 16:27:09 <wirehead_> Opposed to both at the same time for no reason, but OK with using a different option to handle the snowflakes. 16:27:13 <inc0> if its hybrid of both - cool, but single arch 16:27:17 <rhallisey> then we can discuss what we like 16:27:36 <wirehead_> Yeah, I’m fully acknowledging that my opinions have been flipping as the code develops. 16:27:52 <rhallisey> wirehead_, no worries, I've been all over the place 16:28:04 <rhallisey> the tech is changing rapidly and we have to adjust on the fly 16:28:12 <inc0> wirehead_, we've rewritten kolla couple of times ourselves;) 16:28:22 <rhallisey> ya not to mention that ^ 16:28:25 <wirehead_> This is also probably a good time to consider if there’s something we need in upstream kube 1.4 16:29:04 <rhallisey> wirehead_, the only thing I can think of is additional HA. But since we haven't gotten there yet I can't quantify it 16:29:22 <wirehead_> PetSets. :) 16:30:05 <rhallisey> I briefly read about them 16:30:09 * rhallisey will read further 16:30:23 <rhallisey> #topic Open Discussion 16:31:06 <wirehead_> So, we’re still blocked on merging https://review.openstack.org/#/c/327925/ 16:31:40 <wirehead_> To clarify, to inc0 and maybe kfox1111… we want a new variable that dictates that we’re using kolla-kubernetes 16:32:30 <rhallisey> wirehead_, interesting 16:32:33 <rhallisey> k will comment on later 16:32:52 <inc0> wirehead_, why not set 0.0.0.0 if this variable is not set? 16:33:05 <Jeffrey4l_> need guy to reply my concern https://review.openstack.org/#/c/327925/8/ansible/roles/common/tasks/config.yml 16:34:07 <wirehead_> It’s getting really bike-shed-y. 16:34:07 <Jeffrey4l_> is anyone interested in ansiblize dockerfile? spec is here https://review.openstack.org/336757 , a poc is here https://review.openstack.org/334208 16:34:46 <inc0> Jeffrey4l_, I'm really opposed to this 16:34:47 <pbourke> Jeffrey4l_: i feel its too big of a change for the project 16:34:54 <inc0> if you do everything in single ansible run 16:34:55 <wirehead_> inc0: So, you mean that we set one variable that’s kolla_kubernetes=yes, and then from that set 0.0.0.0 as the interface address? 16:35:02 <inc0> you effectively lose docker layers 16:35:16 <pbourke> most arguments are a repeat of the dsl 16:35:34 <Jeffrey4l_> pbourke, is is ansible script. not dsl. 16:35:55 <pbourke> i know, but many of the arguments that were made against the dsl also apply here 16:35:56 <inc0> in reality it's worse than dsl, as again, you lose bunch of docker mechanisms 16:36:02 <Jeffrey4l_> and i will make some change to reduce the repeat. 16:36:31 <inc0> wirehead_, I'd love to not have kolla_kubernetes=yes in ansible... 16:36:34 <Jeffrey4l_> inc0, now. docker image cache is useless actually in kolla. we use the parent image mechanisms more. 16:36:41 <Jeffrey4l_> s/now/no. 16:37:08 <inc0> not true 16:37:37 <inc0> also I'm ok with refactoring of dockerfiles 16:37:40 <Jeffrey4l_> btw, I will change the current implemention to use only one ansible playbooks to generate all the images. 16:37:50 <inc0> but not doing total repave of everything 16:39:46 <rhallisey> shall we carry over to #kolla? 16:39:52 <inc0> yeah, I guess 16:39:55 <Jeffrey4l_> ok 16:40:01 <rhallisey> #end meeting 16:40:09 <rhallisey> #endmeeting