15:58:32 <inc0> #startmeeting kolla
15:58:33 <openstack> Meeting started Wed Oct 18 15:58:32 2017 UTC and is due to finish in 60 minutes.  The chair is inc0. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:58:34 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:58:36 <openstack> The meeting name has been set to 'kolla'
15:58:52 <inc0> #topic w00t for kolla!
15:58:55 <duonghq> wOOt \o -o- \o/
15:59:03 <inc0> o/
16:00:04 <Jeffrey4l> o/
16:00:21 <gema> o/
16:00:48 <chason> o/
16:00:51 <duonghq> hi gema
16:01:00 <gema> duonghq: :)
16:01:33 <vhosakot> o/
16:01:42 <duonghq> hi Vikram
16:01:46 <hrw> o/
16:03:07 <britthouser> 0/
16:03:44 <inc0> #topic announcements
16:03:55 <inc0> 1. zuulv3 is here
16:04:10 <inc0> which means we don't have gates until we finish working on them
16:04:25 <inc0> so please don't approve any changes until that happens
16:04:54 <Jeffrey4l> #link https://review.openstack.org/508661
16:05:03 <Jeffrey4l> #link https://review.openstack.org/508759
16:05:09 <Jeffrey4l> any review is welcome
16:05:11 <inc0> any community announcements? (we'll discuss gates later)
16:05:37 <inc0> guess not:)
16:05:41 <inc0> #topic gates
16:05:46 <inc0> ^ links
16:05:50 <inc0> so current status
16:06:18 <inc0> https://review.openstack.org/#/c/508661/ <- build gates are green, I guess we could merge it
16:06:23 <inc0> just to have build gates
16:06:41 <inc0> for deploy gates we need to first merge kolla-ansible gates
16:07:40 <inc0> so please cores review and merge
16:08:13 <inc0> https://review.openstack.org/#/c/508759/ <- this is almost finished, should be done today, also please merge when gates gets green
16:08:25 <inc0> so when these two gets merged
16:08:42 <inc0> https://review.openstack.org/#/c/512779/ <- this is wip, multinode gates
16:09:20 <inc0> and we'll need one more to do deploy gates in kolla
16:09:32 <inc0> after that plan go as follows:
16:09:48 <inc0> 1. Increase scenario amount in kolla-ansible
16:10:05 <rwellum> o/
16:10:10 <inc0> and per discussion in ptg, we won't run binary images in ansible regular gates
16:10:20 <inc0> (I'll get to that in a second)
16:10:35 <inc0> so 3 options, all source, per scenario
16:11:47 <inc0> 2. make periodical gates for kolla
16:11:52 <inc0> now that's going to be interesting
16:12:05 <gema> how periodical? daily? weekly?
16:12:20 <inc0> dunno yet, doesn't really matter, daily ideally
16:12:29 <gema> ok, so all scenario tests should run daily
16:12:30 <britthouser> yeah daily sounds good
16:12:44 <inc0> yeah, that will build images, run all scenarios available
16:12:51 <gema> k
16:12:54 <inc0> and it will be connected to dockerhub publisher
16:13:11 <inc0> we'll also need to figure out way to easily observe results
16:13:24 <inc0> maybe irc bot?
16:13:27 <inc0> dunno yet
16:13:40 <Jeffrey4l> irc bot is a good idea.
16:14:54 <inc0> we will want to publish images afterwards
16:15:07 <Jeffrey4l> and failed periodic jobs will trigger a special email in ML.
16:15:24 <inc0> dunno yet, we don't want to spam
16:15:30 <inc0> but we'll figure it out later
16:15:42 <Jeffrey4l> this is already exist ;)
16:16:09 <inc0> ok, well, anyway, discuss when we're there;)
16:16:18 <inc0> so comments?
16:16:52 <Jeffrey4l> about push images. will we push branch images?
16:17:02 <inc0> master and all stable, yes
16:17:05 <inc0> no tags yet
16:17:27 <inc0> top of branch is important because priority usage is gates
16:17:35 <inc0> but it will be very useful for everyone
16:17:40 <Jeffrey4l> to docker hub or openstack infra registry?
16:17:45 <inc0> both
16:17:51 <inc0> latter doesn't exist yet
16:18:11 <gema> inc0: do you want a spec for all this work? or do I keep it on the blueprint?
16:18:13 <Jeffrey4l> tags is more stable to use especially in docker hub
16:18:19 <inc0> I'm thinking about pushing straight to dockerhub now, for gates
16:18:22 <gema> inc0: haven't had time to start digging but I am going to start now
16:18:40 <sdake> o/
16:18:44 <inc0> yeah, but one thing at the time, any images will be useful and tarballs might stop working
16:18:50 <sdake> late start today - apologies
16:19:19 <inc0> gema: if you want to help, start by reviewing these 2 changes we linked
16:19:23 <Jeffrey4l> i am thinking push tag images to docker hub and branch images to openstack infra registry.
16:19:27 <gema> inc0: ack
16:19:28 <inc0> everything else depends on them
16:19:54 <inc0> Jeffrey4l: reason I want to have branch images is that we want images asap, and we need branch images anyway
16:20:16 <inc0> also, I'm reluctant to say that tag images are more stable than stable branch ones
16:20:27 <inc0> yes, some more testing there, but less bugs fixed
16:21:17 <inc0> and when we tag our code and fix bug, fix goes to stable branch
16:21:26 <inc0> so it'll always be more stable than tagged code
16:24:21 <Jeffrey4l> aha.
16:24:24 <Jeffrey4l> what's the status of infra registry? any process ?
16:24:25 <inc0> only scenario when thats not true is when our bugfix destabilizes code
16:24:33 <inc0> that's why we'll have a lot of gates there
16:24:49 <inc0> nothing, they were busy with zuulv3
16:24:59 <inc0> I keep asking and I think we'll get to it soon
16:25:31 <inc0> but until we do it, I was thinking of pushing straight to dockerhub/quay
16:25:56 <inc0> and use this + cache proxy in nodepools
16:26:08 <inc0> we'll see, I'll need to discuss this with infra
16:26:23 <Jeffrey4l> roger, thx
16:27:17 <inc0> any questions guys?
16:28:00 <rwellum> I have an unrelated question for the open forum.
16:28:16 <inc0> #topic open discussion
16:28:20 <inc0> shoot:)
16:29:29 <inc0> rwellum:
16:29:34 <rwellum> Well my question is about kolla-k8s - I've spoken to a few people privately. We have a long roadmap to get to 1.0, formalized at the PTG - but very few people working on it. As regulars have backed away. Sydney summit coming up - where do we stand?
16:29:45 <rwellum> (I'm a slow typer inc0 )
16:30:42 <inc0> well, nothing much changed from PTG - we need to finish up orchestration
16:31:02 <rwellum> We have the one playbook in review - yes.
16:31:31 <rwellum> I have my tool in review too - but it's unstable because I can't seem to get a working pike+ OS.
16:31:59 <inc0> we need ansible, yes
16:32:09 <pbourke> folks seem to be moving ahead with openstack-helm
16:32:17 <pbourke> whats the main difference between it and kolla-k8s as of today
16:32:28 <inc0> there is some difference
16:32:32 <sdake> rwellum back away is the wrong terminooogy - have other things to do is more accurate
16:32:38 <inc0> main one is control over what's happening to your deployment
16:32:55 <inc0> as in, with kolla-k8s you have much more detailed control over it
16:33:06 <inc0> super useful for brownfield deployments
16:33:36 <gema> rwellum: how many people are there in kolla-k8s?
16:33:56 <rwellum> gema: that's more a question to inc0 really
16:34:09 <rwellum> I know that the three people I worked with are all very hard to get hold of now.
16:34:10 <inc0> that was main reason openstack helm wanted their own project - disagreement regarding granularity of charts
16:34:25 <rwellum> Each of them has told me they are working on other things
16:34:26 <inc0> if you ask me it's not too big difference, but we're in mess we are
16:34:55 <inc0> yeah, series of unfortunate events in corporate worlds too
16:35:44 <gema> well, I have a hire in flight, once we get it we'll be able to pitch in as well
16:35:56 <inc0> that'd be awesome
16:36:21 <rwellum> inc0: I think the granularity of charts has coalesced though. Both projects can do micro and macro services.
16:36:30 <inc0> thing about kolla kubernetes vs os-helm is that I, for one, believe we will have better architecture at the end
16:36:59 <inc0> took me a while to understand where it'll shine but that's what I believe now
16:37:10 <rwellum> The difference is in the orchestration - os-helm does it via helm charts - we generate the helm charts using the config method that kolla-ansible uses.
16:37:20 <rwellum> So they claim readibility
16:37:32 <inc0> you mean config files?
16:37:40 <rwellum> Yes.
16:37:43 <sdake> the other main difference is that kolla-k8s has a very complete implmeentation of ironic
16:38:03 <sdake> it is possible, for example, to spawn new nodes via ironic - and have that dynamically expand the control and compute planes
16:38:04 <inc0> no, biggest difference is microcharts
16:38:18 <inc0> in osh you have chart neutron
16:38:26 <inc0> we have chart neutron-database-init
16:38:38 <inc0> which means you can *just* initialize database
16:38:50 <inc0> something extreamally useful for brownfield-like scenarios
16:39:08 <inc0> this granularity gives us control over flow of deployment
16:39:19 <inc0> problem is it hurts readibility
16:39:31 <inc0> (golang template language doesn't help too)
16:39:59 <rwellum> +1 inc0 for that explanation.
16:40:34 <pbourke> ++
16:40:52 <pbourke> appreciate it
16:41:36 <inc0> I'd really love to merge projects at some point, but reason we don't is that our condition is this brownfield compatibility
16:41:36 <rwellum> Big difference though is the community. I have an issue with kolla-k8s and post it on os-kolla - usually crickets when people realize I'm talking k8s. On os-helm you can see they are really reactive. Because everyone there is working on the same project - have the same goals.
16:41:55 <inc0> well, that's true
16:42:01 <inc0> ATT funds project
16:42:05 <inc0> nobody funds kolla-k8s
16:42:24 <rwellum> But inc0 kolla-k8s is not even at 1.0. Why do we care about browfield at this point? Get greenfield working.
16:42:51 <gema> rwellum: do you have time to put into it?
16:42:56 <inc0> reason I believe brownfield is even more important than greenfield
16:43:03 <gema> cos I am sure inc0 would love that :D
16:43:18 <inc0> is that most of people who wants openstack, already does have openstack
16:43:24 <inc0> gema: rwellum puts a lot of time in it
16:43:28 <rwellum> gema: my management is expecting me to deliver. If no kolla-k8s then os-helm or pivot as needed.
16:43:29 <gema> inc0: I know
16:43:39 <gema> rwellum: understood
16:44:18 <inc0> community is this one company puts money in one project
16:44:33 <inc0> that's main reason why kolla-k8s have issues
16:44:55 <inc0> that's kinda the case for most of kolla forever, kolla-ansible had to go against fuel with no corporate support really
16:45:26 <sdake> inc0 i htink what rwellum is say8iing is kolla-ansible is healthy - kolla-k8s less so
16:45:41 <inc0> yup, I know and agree
16:45:53 <inc0> I'm iterating reasons why we're in space we are
16:46:09 <gema> inc0: back of the envelope calculation, how far is kolla-k8s from being ready , in terms of effort?
16:46:21 <inc0> and why I'm not super eager to kill project (it might actually be better at the end of the day)
16:46:40 <inc0> https://review.openstack.org/#/c/473588/ this review
16:46:47 <inc0> and few more, but when this one works
16:46:55 <inc0> we'll be in much better place
16:47:10 <gema> ok, will have a look tomorrow
16:47:23 <inc0> thanks
16:47:44 <sdake> ya - if people can taek that review on - kolla-k8s is pretty close technically to go
16:47:53 <rwellum> Me too - I have been following but haven't reviewed for a while
16:47:57 <gema> rwellum, inc0: let's keep this conversation alive in the channel and keep pushing for that review
16:48:03 <inc0> right, it might not seem so, but really it's not that far
16:48:16 <gema> will have kevin look into it too
16:48:20 <gema> (linaro's kevin)
16:48:28 <rwellum> Ok - that'd be appreciated. And if kolla-k8s questions come in, kolla-ansible folk should also answer if possible IMO.
16:48:43 <rwellum> gema: ^^
16:48:45 <inc0> rwellum: problem is, not many ansible folk can understand
16:48:53 <inc0> kolla-k8s
16:48:59 <rwellum> My argument is they need too :)
16:49:01 <inc0> that's one more thing I'm struggling with
16:49:16 <inc0> you can't force understanding
16:49:17 <gema> maybe we should organise a training session
16:49:28 <inc0> as much as you don't fully understand ansible piece
16:49:47 <rwellum> You can't but also a lot of the problems are OS or container based - k8s usually stays out of the way.
16:49:59 <rwellum> Those anyone of the cores on here could answer.
16:50:44 <rwellum> For example, cinder is crashing in Pike for me. I am sure that's no a k8s problem. Now getting to the logs is different and has to be learned.
16:51:12 * rwellum not sure why I keep typing 'no' instead of 'not'...
16:51:42 <rwellum> Anyway I am glad I raised my concerns here. Be happy if we can keep this alive.
16:52:17 <inc0> yeah, and getting logs out is actually k8s specific
16:52:31 <inc0> anyway, your concerns are correct
16:52:48 <rwellum> Thanks inc0, gema and sdake
16:52:56 <inc0> what we'd really need is one or two people full time in kolla-k8s
16:53:21 <rwellum> My hand is raised, but not at the level of sbezverk kfox1111 sdake yet
16:53:29 <inc0> that's main difference between us and osh in terms of community
16:53:54 <inc0> let's get you to that level then
16:53:57 <inc0> and make this a priority
16:54:18 <inc0> gema: if Kevin would want to join in, let's do this too
16:54:28 <rwellum> Sounds good to me
16:54:28 <inc0> I mean your idea about training session is great imho
16:54:58 <inc0> I could use it as well tbh;)
16:55:06 <inc0> maybe series of hangouts with kfox1111?
16:55:10 <rwellum> gema: my toolin review has a training mode - where it stops and attempts to explain what is going on.
16:55:34 <rwellum> inc0: +1
16:56:10 <inc0> ok, I'll talk with kfox1111 when he's around and we'll organize something
16:56:13 <rwellum> gema: are you going to Sydney?
16:56:17 <rwellum> thanks inc0
16:56:21 <gema> rwellum: nope
16:56:26 <rwellum> Gah.
16:56:39 <gema> rwellum: too long trip for three days, had to pass :)
16:56:52 <rwellum> Understood
16:56:54 <gema> gotta run guys, we'll keep talking
16:57:12 <inc0> ok, we're at the time
16:57:15 <inc0> thank you all!
16:57:25 <inc0> and remember them gate reviews
16:57:35 <inc0> #endmeeting kolla