16:01:02 #startmeeting Kolla 16:01:08 Meeting started Wed Nov 16 16:01:02 2016 UTC and is due to finish in 60 minutes. The chair is inc0. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:09 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:01:11 The meeting name has been set to 'kolla' 16:01:13 woot o/ 16:01:23 hello everyone:) 16:01:28 woot/ 16:01:30 helo 16:01:30 hi all 16:01:31 hello 16:01:33 #topic w00t for Kolla rollcall 16:01:33 o/ 16:01:37 0/ 16:01:39 o/ 16:01:40 woot o/ 16:01:40 woot 16:01:43 w00t 16:01:47 woot:) 16:01:49 wo0t/// 16:02:23 ok, let's move on, I bet this will be full time meeting 16:02:30 #topic Announcements 16:02:41 o/ 16:02:46 o/ 16:02:46 1. Kolla ansible is up! Our repo is split at last 16:02:47 o/ 16:02:58 we will cover logistics around that later in the meeting 16:03:14 2. 4.0.0b1 aka ocata-1 releases this week 16:03:37 any other announcements? 16:03:56 ok, let's move on to hard stuff 16:04:02 #topic repo split 16:04:09 \o 16:04:20 #link https://github.com/openstack/kolla-ansible 16:04:30 we have it:) 16:04:43 o/ 16:04:55 as you can see this is copy of kolla itself from weekend 16:05:22 today I'll propose patch removing ansible dir from kolla repo 16:05:47 inc0, can we do this until we fixed the gate? 16:06:04 Jeffrey4l, can we fix gate in kolla-ansible? 16:06:25 in ml, i propose leave the deploy gate in kolla. 16:06:39 inc0 i already did the removal of ansible from kolla repo 16:06:47 the patch is there https://review.openstack.org/#/c/398025/ 16:06:56 sorry, here now 16:07:05 ahh missed that, sorry sdake 16:07:13 Jeffrey4l, no, we need to remove ansible asap 16:07:19 ok. 16:07:21 i propose remove docker from kolla-ansible 16:07:24 because people will keep pushing code there 16:07:28 https://review.openstack.org/#/c/398320/ 16:07:34 mliima, that on the other hand will have to wair 16:07:44 until we prepare zuul-cloner 16:07:51 or docker registry in infra 16:08:00 ok inc0 16:08:10 as we can build without ansible, we can't deploy without build 16:08:38 ok 16:08:59 on that note, we'll create zuul-clone of kolla in kolla-ansible 16:09:13 we can also do git submodule of kolla in kolla ansible 16:09:19 to keep installation easy 16:09:28 which means we also need to change our docs 16:09:45 as deployment procedure will be slightly different 16:10:26 #link https://etherpad.openstack.org/p/kolla-repo-split 16:10:45 let's use this etherpad to list out all the logistics and cleanups we need to make 16:10:46 inc0 our docs don't recommend deploying gfrom git 16:10:57 pip will be diffetent too 16:11:03 inc0 they recommend deploying from pip becasuse of the pbr version problem 16:11:18 git submodules only help developers 16:11:24 and really, I personally find this an abomination that it's not equivalent 16:11:29 pretty sure everyone thats a dev can check out two repos :) 16:11:36 inc0 feel free to fix it 16:11:40 but that's discussion for another day 16:11:54 inc0 did best I could with the expert resources (dhellman in particular) available to me 16:13:03 Jeffrey4l i thought i saw christmas there for a moment ;) 16:13:30 let's focus on todo guys for now 16:13:40 we need this list to be comprehensive to distribute work effectively 16:13:48 hmm. sorry, do not get the point :/ 16:14:18 Jeffrey4l nm :) 16:16:29 sdake: would kolla-k8s arch spec is still open for input ? 16:19:00 did I lost connection or meeting is actually stalled? 16:19:08 im still here 16:19:11 no, we are in etherpad 16:19:20 doing todo list for cleanups 16:19:27 10 more minutes 16:19:30 coolsvap we are brainstorming - not sure what the window is 16:19:35 sorry missed the etherpad 16:19:45 sp__ yes its open until merged 16:20:05 coolsvap: https://etherpad.openstack.org/p/kolla-repo-split 16:20:18 thanks egonzalez90 got there 16:20:20 sp_ reference https://github.com/openstack/kolla-kubernetes/blob/master/specs/README.rst 16:20:47 sp_, yes it's still open 16:21:14 thanks sdake and rhallisey for info and link 16:22:11 hey could you give me a ping when your back 16:22:21 (from etherpad) 16:26:42 inc0: sup? 16:27:07 2 more minutes in our brainstorm topic (working on ehterpad now) and we'll move on to how gate kolla-ansible 16:27:19 but I appreciate having you around:) thanks 16:27:46 I'll use this time to give you full context mordred 16:28:01 so as you probably know, we split kolla into kolla and kolla-ansible 16:28:06 we also have kolla-kubernetes 16:28:13 kolla now holds all the images 16:28:38 inc0 you mean gate kolla :) 16:28:45 sdake, hold on 16:28:55 that'll be topic part;) 16:29:03 oh right 16:29:05 just giving mordred state we are now 16:29:33 inc0: nod. kolla has the images, kolla-ansible and kolla-k8s want to consume them, yeah? 16:29:39 yeah 16:29:41 but... 16:29:45 portdirect, ping 16:29:52 #topic gates in brave new world 16:29:55 cheers 16:30:08 so mordred fungi and everyone, we want 2 things really 16:30:23 kolla-ansible and k8s needs to build images today 16:30:30 as we don't have docker registry for them 16:30:52 so I expect to use zuul cloner there to pull kolla and build images locally 16:31:15 but at the end of the day, we would like to have some smoketest gates in kolla to see if our deployments hasn't been broken after a change 16:31:45 so correct me if I'm wrong, but zuul cloner will allow cross repo gates as in adding gates from kolla to kolla-ansible, but not other way around right? 16:32:00 they can totally be bi-directional 16:32:22 so zuul clones kolla to kolla-ansible and kolla-ansible to kolla? 16:32:28 so you can say "this job needs the kolla and kolla-ansible repos" then the job will clone both repos in the appropriate state and will run the job content 16:32:49 yeah, it's just a convenience utility to clone arbitrary refs of an arbitrary number of repos, taking advantage of local repo caches and refnames provided by zuul in our ci context 16:33:01 let's say there is a job called "gate-kolla-images" 16:33:11 that job will be configured to clone kolla, kolla-ansible and kolla-kubernetes 16:33:15 Hi 16:33:33 and then in the content of the job, the job will expect those three repos to exist and will do $stuff 16:33:52 then, you can configure the kolla repo to run that job in its gate pipeline, and you can also configure kolla-ansible to do the same 16:34:15 (and kolla-kubernetes too if you want) 16:34:17 yah 16:34:27 yeah we will want to have smoketest for both 16:34:37 by smoketest I assume gates we have today;) 16:34:59 ok, so short term, we need to configure zuul-cloner on both ends 16:35:07 and keep gates working as they work today 16:35:26 that will also allow us to remove docker directory from kolla-ansible immediatly 16:35:51 #link http://docs.openstack.org/infra/zuul/cloner.html 16:36:18 Jeffrey4l, sdake, others - is this solution acceptable? 16:36:28 yep. i like that. 16:36:34 +1 16:36:43 sorry two meetings at once - leemme catch up 16:36:47 DST FTL 16:36:48 looks fine 16:36:50 I would refrain of having full fledged deploy gates in kolla as gates will take forever and still be not-voting 16:37:14 but our current gates are ok for that purpose imho 16:37:15 looks good 16:37:17 yup lgtm 16:37:26 infra cats are experts, follow their lead imo :) 16:37:38 so we just don't change anything aside of making build gates voting (yes, yay, I think we can do that( 16:38:07 i think mordred suggested having them right? 16:38:47 the reason for corss repo gating is to make sure the kolla imgaes are GTG when committed 16:39:04 the only plalce I know to do that is in the kolla repo 16:39:08 yah - the main thing is just to make sure that the build job does the right thing with the multiple repos 16:39:10 but yes 16:39:16 so another question since you're here:) 16:39:27 * mordred hides 16:39:28 (can we wrap up this? I think we know what to do) 16:39:33 yeah, i mostly didn't want to see you roll out something confusing involving hacky git submodules that subsequently locks you out of a lot of convenient options from better testing across your repos 16:39:36 #topic git submodule 16:39:45 run away screaming 16:39:45 so, we'll use zuul cloner 16:39:50 fungi ;) 16:39:57 \o/ 16:40:24 but there are people asking for submodule in kolla-ansible for kolla just for purpose of easier installation for users 16:40:37 git clone kolla-ansible --recursive and you have all you need locally 16:40:42 git submodules are not a substitute for an installer 16:40:51 +1 16:41:04 heh 16:41:17 I'm asking because other teams well, uses it like that it seems;) 16:42:15 I guess putting kolla into requirements.txt of kolla-ansible should do the trick? 16:42:15 fungi right - we use pip to install 16:42:35 sdake, everyone uses something different tbh 16:42:47 well devs use git obviously 16:42:54 that's my point 16:42:56 operators use pip 16:43:02 other project like puppet or openstack ansible are different from kolla. they depend on the code itself. But kolla-ansible depend on the docker images which kolla produced. 16:43:31 you're right Jeffrey4l, so we don't make any dependency at all? 16:43:48 and just ask people to install both projects separately? 16:43:49 i prefer to no. 16:43:52 sorry - laptop crashed 16:44:06 on the deployment node, why i must install kolla and its dependency? 16:44:07 that's another approach 16:44:15 ansibe is enough. 16:44:16 i think the seperation between containers, and deployment makes sense i think from an ops perspective (ie no submodule) - but thats just me. 16:44:17 which I actually like 16:44:44 anyone is against having completely unrelated projects as far as installation goes? 16:45:03 inc0 could you restate the question, it didn't pass my parser 16:45:25 if we add kolla-ansible as kolla' submodule, how about the kolla-kubernetes projects? 16:45:25 sdake, when kolla-ansible is installed in any way, kolla is not 16:45:30 wfm 16:45:41 ok, for now we keep things totally separate then 16:45:47 kolla-ansible depends on an image built by kolla, not depends of kolla at all 16:45:47 maybe we need separate "build-node" from "deployment-node" 16:46:06 we can 16:46:12 and it's quite elegant tbh 16:46:24 duonghq, +1 16:46:28 if you do wind up wanting to tie things together, I would strongly suggest that submodules will cause you more pain than they will solve, so trying other approaches first, I would suggest, will make you all much happier 16:46:30 +1 16:46:43 duonghq, +1 16:46:45 mordred, got it, thanks for warning 16:46:48 mordred: +1 to that :) 16:46:51 cool. 16:46:52 but it seems for now we keep things totally separate 16:47:15 if you do end up getting to the place where it becomes essential - ping us, there are additional features/gotchas in gerrit-land :) 16:47:23 but I like where you're going :) 16:48:01 ok 16:48:08 anything else on that note? 16:48:20 #topic open discussion 16:48:28 still have few minutes 16:49:21 http://lists.openstack.org/pipermail/openstack-dev/2016-November/107429.html 16:49:26 the spec 16:49:35 it's almost done 16:49:41 will have another review up soon 16:49:42 we need push some goals for ocata. 16:50:23 it breaks our release. 16:50:25 https://review.openstack.org/398433 16:50:27 block 16:51:13 rhallisey: thanks for keeping on top of it! 16:51:19 no problem 16:51:21 it doesn't seem to be the case for us tho 16:51:31 just be sure everyone gets their comments in 16:51:42 :) 16:52:15 also the impromptu hangouts have been really usefull - I really think we should keep them up when we hit issues 16:53:10 gotta meeting, so need to head - catch you guys later 16:53:34 ok, I guess we're done for today?:) 16:53:39 woot 16:53:39 for meeting* 16:53:46 today is still ongoing, don't stop working 16:53:50 re impromptu meetings 16:53:53 can we please schedule them 16:53:56 if we are going to have them 16:54:02 they are not inclusive 16:54:03 TIA :) 16:54:20 schedule impromptu meetings? that's not impromptu 16:54:27 right oxymoron 16:54:41 the problem with improtu meetings is not everyone is aorund 16:54:47 we jump on hangouts when we can't communicate over irc 16:54:54 4 guys in a meeting do not make up the communities best judgement 16:55:03 but they can get into agreement 16:55:24 right - then gotta convince everyone else 16:55:30 not everyone 16:55:56 not inclusive... 16:56:10 i constantly have conflicts with the spur of the moment hangouts 16:56:22 we had something called "liberum veto" in Polish history, where single person could throw out voting...spoiler alert - didn't end well 16:57:00 anyway, will be reflected in spec and totally optional 16:57:03 i'm not asking for much 16:57:09 so feel free to drop it there 16:57:45 anyway, I think we exhausted this meetings time 16:57:54 thanks everyone for comming! 16:58:03 #endmeeting kolla