22:04:14 <jeblair> #startmeeting zuul
22:04:15 <clarkb> hello
22:04:16 <openstack> Meeting started Mon Jun 12 22:04:14 2017 UTC and is due to finish in 60 minutes.  The chair is jeblair. Information about MeetBot at http://wiki.debian.org/MeetBot.
22:04:17 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
22:04:19 <openstack> The meeting name has been set to 'zuul'
22:04:20 <jeblair> #link agenda https://wiki.openstack.org/wiki/Meetings/Zuul
22:04:28 <clarkb> I have to pop out in a bit to run errands
22:04:43 <jeblair> #link previous meeting http://eavesdrop.openstack.org/meetings/zuul/2017/zuul.2017-06-05-22.03.html
22:05:01 <jeblair> first thing -- i just revised the status updates section of the agenda
22:05:48 <jeblair> these are the things that we want to take special care not to block, as they are the areas we're most focusing on in order to get into production
22:05:56 <jeblair> i listed:
22:06:00 <jeblair> Standard job library
22:06:00 <jeblair> Documentation
22:06:00 <jeblair> Github parity
22:06:01 <jeblair> Zuul test enablement
22:06:43 <jeblair> (i've dropped bubblewrap, added documentation, and morphed the 'jobs' related topic as well as the 'github' topic)
22:07:07 <jeblair> how does that list look to folks?
22:07:41 <fungi> what's the context around dropping bubblewrap? is that done, or deferred for now, or what?
22:07:48 <jeblair> fungi: oh, 'cause it's done
22:08:04 * fungi cheers quietly into his drink
22:08:06 <fungi> wfm!
22:08:34 <jeblair> and the main push on 'github' is done, but i morphed it into 'parity' since there's still some important things missing (but basic functionality is there)
22:09:05 <pabelanger> from ze01.o.o, I think we are just missing credentials today and a project to run against
22:09:09 <pabelanger> for github
22:09:27 <fungi> we have gtest-org
22:09:32 <jeblair> i'm excited since i think the first two -- standard jobs and documentation, are the only things that should be holding us back from beginning our scale-out/load testing on openstack.  :)
22:09:53 <clarkb> ya gtest-org or even just have zuul test itself against github too?
22:10:03 <clarkb> since we mirror there we should get the push events I think
22:10:04 <fungi> jeblair: seems very close now!
22:10:15 <pabelanger> I think we still have a little infrastructure to discuss too, but we likely can move that to tomorrows meeting?  eg: zookeeper scale out
22:10:19 <jeblair> okay, no objections, so let's go with that list for now.  we can update it pretty easily.  it's a wiki.  :)
22:10:30 <clarkb> pabelanger: do we need to scale out zk?
22:10:53 <clarkb> a single node zk is no worse than the single node geard we have been reliant on for years. I don't know that its necessary to do that first
22:10:58 <jeblair> do we just want to open the meeting agenda?
22:11:55 <clarkb> I was reading it as there may be a few infrastrucutre pieces we want to capture on the special care not to block but maybe I am mistaken
22:12:17 <fungi> i have no preference on whether the meeting agenda is open or structured
22:12:21 <pabelanger> Ya, I don't want to block the meeting now too, I can loop back to it for open discussion
22:12:51 <jeblair> #topic status updates:  Standard job library
22:13:19 <jeblair> okay, so pabelanger and i are going to be working on the zuul-jobs repo this week
22:13:38 <jeblair> that's where we're going to put jobs which may be of interest to any zuul installation
22:14:04 <jeblair> i expect openstack-infra to either run jobs from it or run jobs inherited from jobs in it
22:14:18 <jeblair> same for bonnyci, third party cis, and any other zuul
22:14:23 <fungi> have a short-list of example jobs?
22:14:49 <fungi> er, examples of what jobs would be of interest to any zuul installation i mean
22:14:50 <jeblair> fungi: the common python related jobs, for starters.  like a 'python27' job, a 'python35' job, etc.
22:15:09 <jeblair> so that anyone running a zuul can say "i want to run python 2.7 based unit tests against this repo"
22:15:22 <fungi> okay, thanks. i wasn't sure if that was too openstack/python-specific
22:15:33 <jeblair> fungi: yeah, i think that's the really interesting part of this
22:15:58 <fungi> but since zuul itself is written in python, i guess it makes sense to include some basic python-oriented jobs
22:16:06 <jeblair> fungi: i would have thought a while ago that writing such a job would be very hard.  but mordred has argued that it can be safely generalized.
22:16:16 <jeblair> i'm convinced it's worth a shot
22:16:49 <jeblair> so i think we can go as far as putting "our" python27 job into that repo (with all the "weird" things it does like sudo grep, subunity processing, etc)
22:17:03 <fungi> sure
22:17:14 <fungi> no objection here
22:17:16 <jeblair> and with some carefully placed conditionals, it should actually be pretty widely applicable
22:17:24 <fungi> gotta draw that line in the sand somewhere
22:17:46 <jeblair> (someday, it may grow the ability to run nose tests too, even if we don't use it (much))
22:18:01 <jeblair> anyway, that's my thinking on that at the moment.
22:18:22 <jeblair> anything that doesn't work out there, we'll still have the openstack-zuul-jobs repo to stick our own versions of things in.
22:18:54 <jeblair> there's a crossover topic here -- documentation for the standard library
22:19:04 <clarkb> fwiw our current job does have the ability to run nose tests (eg horizon and swift)
22:19:21 <jeblair> how close we are already :)
22:19:50 <jeblair> i have a change to create 'zuul-sphinx': https://review.openstack.org/472823
22:20:05 <jeblair> this will be a repo which holds a sphinx plugin that we can add to all of our "jobs" repos
22:20:26 <fungi> self-documenting jobs!
22:20:29 <jeblair> i've added a 'description' field to jobs in zuul.yaml
22:20:44 <jeblair> so that, yeah, we can put documentation about jobs right in the yaml
22:20:52 <jeblair> and we can generate nice sphinx docs for the whole repo
22:20:56 <jeblair> same thing applies to roles
22:20:59 <fungi> best place for it
22:21:08 <pabelanger> that would be great
22:21:33 <jeblair> i'd like to start having that documentation right from the start so we have good habits and encourage good copy pasta :)
22:22:34 <jeblair> the code for that is mostly written here: https://review.openstack.org/473544
22:22:40 <jeblair> i'll move it over there as soon as the repo exists
22:23:53 <jeblair> anything else stdlib related?
22:24:14 <jeblair> #topic status updates: Documentation
22:25:03 <jeblair> i have this draft from a couple weeks back: https://review.openstack.org/463328
22:25:22 <jeblair> i plan on continuing work on that soon
22:25:47 <jeblair> i definitely think it's time to get zuul's documentation in order, before we start exposing more folks in the community to it
22:25:59 <jeblair> that *mostly* addresses reference documentation
22:26:24 <jeblair> i think we may want some more narrative documentation as well
22:26:54 <jeblair> there's some narrative in there, but interspersed like that may not be the best way for someone to get a holistic view of how things work
22:28:15 <jeblair> if anyone else wants to pitch in on revising docs, that'd be great.
22:28:43 <jeblair> we need to establish a new baseline there, and once we do, we'll start being very strict about requiring docs as part of changes again
22:29:08 <jeblair> (the fact that we haven't recently is a temporary aberration, not our normal process)
22:29:13 <jeblair> anyone else have anything related to docs?
22:29:41 <mordred> I've got a todo list task personally to work on narrative docs this week
22:30:13 <fungi> nope, other than pitching in on docs might be slightly lower-hanging fruit task (probably needs you to at least be able to stand up a working zuulv3 though i guess)
22:30:44 <jeblair> mordred: neat! :)
22:30:49 <fungi> but maybe a good place for people who are less certain of their python skills to pitch in
22:31:19 <jeblair> fungi: it probably doesn't require that; but it may require reading the code or paying lots of attention.  :)
22:31:30 <fungi> fair
22:32:17 <fungi> still, convenient to be able to try out the things you're documenting to make sure they're correct
22:32:28 <jeblair> ya
22:32:38 <jeblair> happily, zuul is good at telling you if things are correct
22:33:08 <jeblair> #topic status updates: Github parity
22:33:24 <jeblair> i know jlk is working on this, but doesn't seem to be around for today's meeting
22:33:49 <jeblair> https://review.openstack.org/472468 is the current focus
22:34:03 <jeblair> the next big thing after that i think might be cross-repo dependencies
22:34:32 <jeblair> anything else about either this, or the next topic, zuul test enablement?
22:35:10 <pabelanger> is there any value in setting up github pipeline today?
22:36:03 <jamielennox> pabelanger: what value are you looking for?
22:36:08 <jamielennox> we are certainly testing it
22:36:20 <jamielennox> but i'm not sure if there's anything that -infra wants from github
22:36:34 <jeblair> pabelanger: yes, i think we should start using it asap.  probably we should start by pointing it at a gtest-org repo
22:36:50 <jeblair> pabelanger, jamielennox: we have our eyes on the ansible repo :)
22:36:52 <jamielennox> and yes, afaik cross-repo dependencies are the next big thing
22:37:09 <jeblair> shade has an immediate use for reporting on ansible prs
22:37:13 <mordred> yup
22:37:27 <fungi> we've talked about things like being able to test with and comment on pull requests to ansible, pip, et cetera in the past
22:37:30 <jeblair> so as soon as we think we can do something useful there without spamming errors, we should start working on hooking that up.
22:37:32 <pabelanger> okay, I assume mordred you have credentials for github intergration
22:37:45 <mordred> yes. I have created an OpenStack Zuul acount
22:37:55 <jeblair> s/acount/app/
22:38:03 <jamielennox> ok, ping with any questions on creds, because there's a few ways you can do that
22:38:06 <mordred> I also think we should make a job at some point which tests ansible prs that would affect zuul against zuul
22:38:17 <pabelanger> Cool, I'll follow up after meeting
22:38:22 <mordred> \o/
22:39:04 <jeblair> Shrews, mordred: should we add '(web) console streaming' to the agenda status updates section?  that seems like it might be ready for this?
22:39:07 <mordred> pabelanger, jamielennox, jeblair: I'd LOVE to have it working-ish enough to show people next week at ansiblefest london, fwiw
22:39:34 <mordred> jeblair: probably?
22:39:35 <pabelanger> mordred: Same, it would be great to show off what we have for sure
22:39:37 <jamielennox> mordred: you should definitely coordinate that with jlk as he's in london for that
22:39:37 <Shrews> maybe?
22:39:50 <mordred> jamielennox: yup. totes agree
22:40:00 <jeblair> mordred: i think that's going to be a pretty tight deadline.  i was hoping for a halfway working python job by the end of the week.
22:40:18 <Shrews> the whole web discussion on the ML left me wondering if we should reconsider current stuff
22:40:56 <jeblair> before we move on to web stuff and streaming --
22:41:21 <mordred> jeblair: agree. if it's not there, I won't be sad. but if it's magically actually ready, then awesome
22:41:26 <jeblair> mordred: how important do you think that is?  because i think we'll need to reprioritize our work this week if we want that to happen.
22:42:16 <jamielennox> mordred: bonny is running if you want it, it just obviously doesn't have tie in to -infra
22:42:18 <jeblair> okay.  let's focus on the standard library stuff as we were planning, but maybe try to keep github operational stuff moving and maybe we'll end up being able to 'hello world' or something.
22:42:30 <mordred> that's a hard question and I'm not sure I've got a solid answer. the desire is mostly driven by the fact that we'll be in the room with a set of folks
22:43:01 <mordred> jamielennox: cool, thanks. depending on where we're at that might be a good option too
22:43:15 <mordred> jeblair: so yah - I agree with what you said
22:43:33 <jeblair> cool.  i like the 'show off bonnyci' as an option too.  :)
22:43:51 <mordred> jeblair: stdlib is the important goal, but maybe everything will move magically quickly and we'll have a hello-world that'll work
22:44:19 <jeblair> #status updates: (web) console streaming
22:44:26 <jeblair> #topic status updates: (web) console streaming
22:44:41 <jeblair> Shrews: i sent a reply on this to the ML recently
22:45:00 <jeblair> Shrews: in short, i don't think we should delay any of your work with the web framework discussion
22:45:38 <Shrews> yep, read that, and i'm fine with that direction
22:45:57 <Shrews> b/c it does seem like a much larger topic for >3.0
22:46:10 <jeblair> Shrews: to me, it's more that i want to make sure tristanC is able to work on his thing starting with the thing we'd like to use in the end, and we can move the other stuff to it when we have time
22:46:28 <mordred> ++
22:46:46 <jeblair> we should, obviously, make sure whatever we select can handle websockets so that we *can* port the current work over
22:46:53 <jeblair> but i think we're doing that
22:47:55 <jeblair> mordred: you have a whole streaming related stack starting at https://review.openstack.org/472850 that's ready to go in i think
22:48:13 <mordred> the first 6 are at least
22:48:21 <mordred> patch 7 breaks spectacularly
22:49:03 <mordred> jeblair: although, before I debug patch 7 - I wouldn't mind feedback from you on whether or not you agree that it's a good idea in the first place
22:49:07 <jeblair> Shrews: and you said your test change - https://review.openstack.org/471079 is working now right?
22:49:13 <jeblair> mordred: ack, will do
22:49:20 <jeblair> Shrews: the websocket stuff proper is still WIP?
22:49:23 <mordred> https://review.openstack.org/#/c/473090 is the one that can use a directional nod
22:49:56 <Shrews> jeblair: yes, and ran 100's of times to make sure there were no sneaky failures in there. It is needed before I can start working on tests for 473090
22:50:20 <Shrews> (though I'm not guaranteeing NO sneaky failures)
22:50:51 <Shrews> there were several races around directories and files and data structures
22:50:51 <jeblair> Shrews: you meant s/473090/456353/ ?
22:50:56 <jeblair> (090 is mordred's change)
22:51:12 <Shrews> right, 353
22:51:41 <jeblair> okay, i've starred things
22:52:17 <jeblair> #topic open discussion
22:52:33 <jeblair> several of us are in london for ansiblefest next week, myself included.
22:52:42 <jeblair> i'm thinking maybe we should skip the meeting
22:52:49 <mordred> ++
22:52:58 <mordred> I believe I will be in an airplane during the meeting time
22:53:30 <pabelanger> I think we need to focus on log publishing for zuulv3.o.o. It is becoming difficult to debug ansible playbooks :)  Have we figured out our log URL issues now?
22:53:45 <pabelanger> ++ I am also on PTO 2 weeks after ansiblefest
22:54:07 <fungi> i am curious if/when SpamapS's zuul v3 security spec modification to talk about ssh is ready for council vote, but that doesn't necessarily need a zuul meeting to decide i guess
22:54:10 <mordred> jeblair: on an unrelated note - is it time to discuss removing our py27 testing? we're deploying on py3 and have landed v3-only code now
22:54:45 <jeblair> fungi: this spec has escaped my attention?
22:54:46 <mordred> jeblair: I ask mostly because I have hit at least one issue that was caused by the habit of running local unittests under py2
22:54:52 <jamielennox> landed v3 only code outside the websocket stuff?
22:55:16 <pabelanger> mordred: that might leave centos-7 installs in a bad place
22:55:24 <fungi> #link https://review.openstack.org/462207 Revise security spec to discuss SSH keys
22:55:31 <jamielennox> I haven't transitioned our stuff to v3 because i haven't deployed the log streaming yet
22:55:47 <jeblair> pabelanger: is there a centos-7 install of zuulv3?
22:56:16 <pabelanger> software factory
22:56:34 <pabelanger> the main blocker right now is zookeeper too
22:56:34 <jeblair> pabelanger: i really hope no one is running v3 in production
22:56:49 <pabelanger> jeblair: they've started work towards it
22:57:06 <jeblair> pabelanger: that's great!  we should make sure they know it's v3 only.
22:57:26 <pabelanger> okay, I don't know if that is known
22:57:31 <jeblair> pabelanger: from the way you said it earlier, it sounded like you were worried about us leaving existing v3 users in the lurch.
22:57:40 <jeblair> pabelanger: probably not, we don't have release notes yet.  :)
22:58:09 <pabelanger> Right, or leave out some centos-7 users from adopting zuulv3 :)
22:58:55 <jeblair> i hope that python3 is not an insurmountable obstacle.
22:59:07 <jeblair> implementing zuulv3 in python2 certainly is one
22:59:10 <pabelanger> FWIW: this issue also applies to openstack in general
22:59:15 <pabelanger> so they need to solve it there first
22:59:46 <jeblair> anyway, for better or worse, zuulv3 does require python3
22:59:53 <pabelanger> okay, good to know
22:59:57 <pabelanger> will pass that along
23:00:03 <jeblair> and mordred's question is worth considering -- should we drop the py27 tests
23:00:12 <jeblair> i'd say let's do it soon, but not right away
23:00:35 <jeblair> i like having multiple python jobs running while we're using zuul itself as a guinea pig
23:00:47 <jeblair> and we're out of time
23:00:49 <jeblair> thanks everyone!
23:00:51 <jeblair> #endmeeting