16:02:37 #startmeeting kolla 16:02:38 Meeting started Wed Feb 14 16:02:37 2018 UTC and is due to finish in 60 minutes. The chair is inc0. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:02:39 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:02:41 now it did:) 16:02:42 The meeting name has been set to 'kolla' 16:02:49 \o 16:02:50 #topic w00t! 16:03:01 O/ 16:03:03 good morning everyone 16:03:09 hi! 16:03:14 o/ 16:03:16 hi there! 16:03:28 hello 16:03:43 * gkadam_ 's first kolla meeting. Hello Everyone! :) 16:03:46 hi 16:03:54 welcome gkadam_ 16:04:14 w00t w00t 16:04:17 inc0, thanks! :) 16:04:36 welcome gkadam_ 16:04:56 happy Valentines day;) 16:05:14 ok let's move to actual meeting 16:05:18 #topic announcements 16:05:43 well, this is probably last meeting I'm leading. Everyone voted for next PTL? 16:05:53 elections ends in few hours 16:06:43 any community announcements? 16:06:56 thanks again for your service inc0 16:07:07 :) 16:07:07 inc0++ 16:07:11 inc0: thanks a lot for your leadership indeed! 16:07:33 just be as good to next ptl as you were to me 16:07:49 that's my request 16:07:56 allright, let's move to agenda 16:08:05 #topic Deploying a cloud with Kolla for the RDO test days - status? 16:08:16 dmsimard: ^ 16:08:29 I actually added that line, unsure if dmsimard is around 16:08:40 Hi, we actually just discussed it in our RDO meeting 16:08:44 ah there you are :) 16:09:22 so what are we doing and how do we do it?:) 16:09:27 Long story short we're going to push it until after the PTG. The timing is not really good right now with everyone scrambling to get Queens out the door 16:09:44 We're still very much interested in doing it (and I hope you are too) 16:10:16 that probably makes sense too in that we can test using the released queens rather than an rc 16:10:39 dmsimard: just please don't be shy about sending us reminders to the mailing list so everyone can be prepared 16:10:41 Something like that, the stable repos will be out by then as well 16:10:50 dmsimard: will you be at the PTG? 16:11:03 Unfortunately not this time around but some of my colleagues will be there 16:11:16 I'll hopefully be in Vancouver though 16:11:42 kk 16:12:18 Don't be shy to poke me for updates either, I'm on a lot of threads so sometimes it's hard to keep up :) 16:12:27 I hear you 16:12:31 (just like you pinged me right now!) 16:13:49 ok so we'll pick it up again after the PTG 16:14:15 ok, so let's move on then 16:14:25 thanks dmsimard for update 16:14:35 #topic keystone uuid removal in rocky (pbourke) 16:14:38 Anytime 16:14:50 pbourke: you're on 16:15:13 ok so as people may have noticed keystone is broken in kolla master 16:15:34 they moved onto rocky and removed uuid tokens 16:16:00 luckily for us, fernet works fine in kolla so it will be a small patch to update our defaults 16:16:06 that said, we have two tasks as I see it 16:16:26 1) reconfigure/upgrade for people coming from previous releases 16:16:42 2) what to do with our queens master as is before we branch 16:16:59 2) may not be an issue as the gates still seem to be working though im not sure how... 16:17:28 does this make sense to people? 16:18:17 well...how it's possible that gates work? oO 16:18:36 also, does that mean we can get rid of memcached? 16:18:36 yeah I think I need to figure that out :/ 16:18:51 I think horizon may still need it? 16:19:17 tbd, but memcache with uuid was huge security risk 16:19:23 anyway, just wanted to make people aware. Will try sync up with Jeffrey4l wrt branching 16:20:28 thanks for heads up 16:21:05 pbourke: anythin else on that topic? 16:21:13 dont think so :) 16:21:17 kk 16:21:22 #topic yankcrime - haproxy config 16:21:30 yankcrime: you have the floor 16:21:34 wow 16:21:35 hi! 16:21:42 hello:) 16:21:57 hello yankcrime :) 16:21:59 ok so here's the change i proposed: https://review.openstack.org/#/c/534834/ 16:22:29 it's been languishing for a while, mostly because of a subject which pbourke brought up on openstack-dev 16:22:50 yeah im sort of to blame here... 16:23:18 essentially the patch looks fine, but it sort of re-raised the whole 'kolla only provides essential config' philosophy we have going 16:23:36 so in my case, the suggestion isn't arbitrary - it's driven by an actual use case, and one which i think is fairly common 16:23:42 hence the associated blueprint: https://blueprints.launchpad.net/kolla-ansible/+spec/haproxy-listener-extra-options 16:23:54 pbourke: well we don't have config overrides for haproxy 16:24:00 yeah I get that 16:24:20 though in the past we've gone with the 'with_first_found' or whatever the ansible directive is for these kind of files 16:24:31 overall I dont mind hugely I just want to be consistent 16:24:32 is there any sane way that config overrides could work for haproxy? 16:24:37 as others have had similar patches rejected in the past 16:24:42 right, and i'm guessing non-ini merge_config is a long ways off being viable for stuff like haproxy 16:24:42 which doesn't seen fair 16:24:48 *seem 16:24:55 mgoddard_: you can use your own config completely 16:25:10 yeah, but I don't want to have to maintain a clone of the template 16:25:14 so worst case you can do kolla-ansible genconfig, get default, edit and make ansible use it 16:25:21 haproxy's configuration is probably worst-case for having to completely manage your own configuration outside of kolla-ansible though, no? 16:25:36 yeah it's not ideal 16:26:23 that being said, as pbourke mentioned, having completelty extensible config is also hard/impossible 16:26:50 it is, but in cases where config can't be overridden, it's more justified right? 16:27:10 well in .ini confs we wouldn't accept that 16:27:14 here, we debate:) 16:27:31 I'm actually for accepting this one 16:27:50 what if the haproxy config was more modular, and autogenerated for each service from snippets 16:28:00 well the conversation is more about whether or not you could see this feature, which is justified with a use-case, is worth merging 16:28:15 mgoddard_: if you could implement sth like that, it sounds great 16:28:53 my current thinking on it is there's no one right answer 16:29:09 the real solution is probably somewhere in between the two extremes 16:29:33 yeah I agree pbourke that's why I would accept this one and debate proper full solution 16:29:36 PTG?:) 16:29:44 that sounds like a good compromise 16:29:45 yeah I think that's a good course of action 16:30:05 +2 so for your change yankcrime, sorry again for the procrastination 16:30:19 thanks for sticking with it 16:30:25 no worries pbourke, thanks for helping it move along 16:31:01 ok 16:31:05 #topic open discussion 16:31:10 anyone? 16:31:12 Hi there 16:31:21 hey:) go ahead paken 16:31:24 since this is my first meeting, I would like to introduce myself 16:31:33 hey paken , same here 16:31:51 My actual name is Paco Bernabé (spaniard living/working in The Netherlands) 16:32:03 Working at SURFsara, where we have several clusters 16:32:20 We have the plan of using Openstack to virtualize all these clusters 16:32:35 and we are looking at Kolla-Ansible as an option for deployment 16:32:51 I have submitted last week an small change request 16:33:05 but I have some other related the Ansible playbooks 16:33:18 since this is where I have started looking at 16:33:41 paken: so right now we are in feature freeze - close to release and only bufixes are accepted 16:33:46 for next week or so 16:33:46 so, I hope to keep on contributing on this. 16:33:55 ok, I understand 16:34:14 but awesome, thank you 16:34:29 when we branch out queens, we're open for business again 16:34:39 paken: feel free to post a link to your change(s) if you'd like to get more progress with it 16:35:04 let me introduce my self as well. I am currently working in to openstack tech support at Red Hat 16:35:05 ok, great. I’ll keep on making the list and submitting the changes 16:35:06 also if you run into issues during your deployments, make sure to look for help on our channel 16:35:26 sure, thanks 16:35:52 gkadam_: we have quite a few hatters in our community: 16:35:57 Introductions are a nice idea actually, nice to get to know new members 16:36:10 just came accross the dployment tool called infrared which i tried this week , and learnt that it somehow is using Kolla , tyring to learn more in to container world 16:36:37 also, i am looking for some easy fixes to start with 16:36:39 agree pbourke we should make it standard part of meeting 16:36:42 gkadam_: by 'openstack tech support' you mean rhosp? 16:36:43 as m not a seasoned programmer 16:36:48 hrw, yep 16:37:10 gkadam_: one day I have to check how rhosp looks 16:37:15 gkadam_: I'm not familiar with infrared, but tripleo uses kolla images 16:37:18 hrw, :D 16:37:29 gkadam_ search for wish list bugs in launchpad 16:37:38 egonzalez, ack 16:37:58 inc0, http://infrared.readthedocs.io/en/infrared_v1/introduction.html 16:38:09 But now we need for help in testing upgrades to queens 16:41:14 egonzalez: I'm trying that out a bit actually at the moment, maybe we can chat some more afters 16:42:50 jmccarthy sure 16:43:00 ok, anyone else wants to talk about anything?:) 16:43:07 one question only 16:43:27 although you are feature freeze, changes can still be submitted? 16:43:33 paken: yes 16:43:38 Yes but not merged 16:43:42 they'll just be held / -1'd until branching 16:43:49 cool, got it, thanks 16:43:51 Only bug fixes 16:44:08 i have a small announcement also, I finally got ceph-ansible working with kolla :) 16:44:19 oh, cool! 16:44:20 Yay! 16:44:23 was it hard? 16:44:23 nice 16:44:24 \o/ 16:44:28 very minimal changes actually 16:44:34 which Im very pleased about 16:44:49 so maybe I'll prepare a demo for the PTG and we can go from there 16:44:55 cool, then, do we want to deprecate ceph in Rocky? 16:45:03 Hi, I'm also new here. I work at Ubisoft in Montreal. I have a question regarding what pbourke mentioned regarding 'with_first_found' for custom configs. What was wrong with this approach? 16:45:29 inc0: unless objections I think it would be a good objective 16:45:37 mchlumsky: well you'd need to have your own copy of whole config file and keep it updated as defaults change 16:45:38 inc0: does that mean that that the tasks in external_ceph.yml won’t be used anymore? 16:45:40 pbourke: let's discuss at PRG 16:45:41 PTG 16:45:49 paken: we could keep external ceph stuff 16:45:52 what the changes are and how realistic it is for production 16:45:56 pbourke: well I'd go and mark it as deprecate now 16:46:02 now begin before Q is released 16:46:08 good point 16:46:08 pbourke: I thought we had until the end of luminous to move :) 16:46:09 paken ceph will likely be external ceph 16:46:20 ok 16:46:30 hrw: we'll need to look into adding ARM64 images to ceph-ansible sources asap 16:46:35 while it can take more than one release, it will make it possible to remove it during rocky 16:46:37 gema: make a card 16:46:43 hrw: you got it 16:46:46 pbourke, the final goal is to use ceph-ansible started by kolla ? 16:46:48 gema: hrw: ceph-ansible can use any images afaik 16:46:49 gema: ok 16:47:01 pbourke: but will kolla keep producing them? 16:47:08 We can deprecate and remove when ready, not need to be next release 16:47:17 well we can keep producing them tbh 16:47:20 ktibi: the goal is to stop having to maintain ceph orchestration in kolla. ceph-ansible may be a means to this end 16:47:33 right egonzalez 16:47:54 pbourke, inc0: ok, understood, let's talk about it at PTG and maybe do it in Rocky 16:47:59 that's why i'd be ok with marking it deprecated now and saying that "near future" 16:48:08 inc0: yep, sounds good 16:48:11 +1 16:48:11 near future could be Rocky 16:48:32 pbourke, right but is it kolla which use ceph-ansible ? or need to deploy a ceph cluster before kolla. 16:48:34 pbourke: will you make a change?:) release note with deprecation? and maybe deprecation warning 16:48:40 external ceph may one day mean end of ceph packaging for me ;D 16:48:41 inc0, that's true but I found it very helpful when I needed to add custom keystone configuration to support openid connect in keystone 16:48:54 ktibi: kolla itself can install ceph 16:48:56 inc0: sure lets do it 16:49:12 we're talking about removing this ability because we have hard time supporting it 16:49:12 inc0, ok great :D 16:49:13 inc0, by custom keystone configuration, I meant the wsgi config 16:49:18 mchlumsky: I would totalyl be ok with allowing with_first_found for every config 16:49:32 mchlumsky: the issue is how we handle the ones we do provide 16:50:04 mchlumsky: right, you can do it, issue was with haproxy which itself is very long file 16:50:12 hard to keep track of changes 16:50:44 Re haproxy i was thinking on doing a macro to generate automatically the file, allowing some extra options 16:50:51 ktibi: I think that it can be "deploy ceph using ceph-ansible, deploy openstack using kolla-ansible" or "deploy ceph+openstack using kolla-ansible which calls ceph-ansible to do ceph part" 16:51:01 also, don't tell me Anthem will run on Kolla mchlumsky ;) 16:51:06 Something like kolla image install macro 16:51:31 hrw: yeah, I haven't thought hugely about how we'd integrate the two projects so thoughts here are appreciated 16:51:40 inc0: lol 16:52:02 pbourke: I think that first step would be first method and then doing second way 16:52:27 correction, anthem is EA 16:52:36 soo...assasins creed?;) 16:52:44 i'll be ok with a free copy of origins 16:52:45 hrw yes I think all options are good. But the configuration of pools or anythink about openstack need to be in kolla no ? 16:53:00 anyway, I think config issue should be discussed on PTG 16:53:09 ktibi: no idea how ceph is deployed. my biggest setup is all-in-one+compute 16:53:14 pbourke, the best part about overriding wsgi-keystone.conf (for me) was that I could override it with a slightly modified jina2 template and it would still do the rendering. That was awesome! :) 16:53:18 pbourke: would you mind running this session? 16:53:28 inc0: sure thing 16:53:34 cool 16:53:44 hrw I use kolla + ceph in production so it's very important to me ^^ 16:53:56 ktibi: 99% of deployments do probably 16:53:58 ktibi: we'll be sure to have a migration path 16:54:03 that is essential 16:54:22 ktibi: look how external ceph is handled today 16:54:26 when do we expect ceph to be removed from kolla? 16:54:30 that's going to be the thing 16:54:44 mgoddard_: we're discussing rocky 16:54:56 is it deprecated? 16:54:57 but that's only if we manage to implement robush upgrade path 16:55:00 inc0, I can't discuss such things... but I'm here so.... 16:55:23 haha, no worries, I'll keep that in mind when playing next game;) 16:55:47 (number of cool use cases for kolla grows!) 16:55:48 :) 16:55:55 inc0, yep I waited production ready for ceph-ansible for use it. Now I think I'll maybe use a external ceph (external but on compute :p) 16:56:22 ktibi: external only means "not deployed by us" 16:56:33 so we should be clear about this deprecation 16:56:42 yes, release note and warning 16:56:44 there will be a full cycle of deprecation 16:56:59 at least one full cycle 16:57:01 and it prob should not be done if a lot of community members have reservations 16:57:19 inc0, yes it's why I say on compute ^^ 16:57:20 so I'll make sure to circulate this review widely for feedback 16:57:36 maybe should have a spec, dunno 16:57:39 well after Boston PTG I think everyone was kinda on board that we need to be careful but should do it 16:57:51 pbourke: review it on PTG 16:58:08 pbourke: spec sounds good 16:58:41 allright, we're at the end of meeting 16:58:59 have a great day/evening everyone and remember to vote for PTL if you haven't done so 16:59:06 thanks all! o/ 16:59:06 #endmeeting kolla