22:02:35 <jeblair> #startmeeting zuul
22:02:36 <openstack> Meeting started Mon Dec 19 22:02:35 2016 UTC and is due to finish in 60 minutes.  The chair is jeblair. Information about MeetBot at http://wiki.debian.org/MeetBot.
22:02:37 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
22:02:39 <openstack> The meeting name has been set to 'zuul'
22:02:49 <jeblair> #link agenda https://wiki.openstack.org/wiki/Meetings/Zuul#Agenda_for_next_meeting
22:03:16 <jeblair> #link previous meeting http://eavesdrop.openstack.org/meetings/zuul/2016/zuul.2016-12-12-22.03.html
22:03:20 <auggy> o/
22:03:35 <jeblair> #topic Actions from last meeting
22:03:47 <jeblair> jeblair send email announcing impending merge and release jeblair switch production to run from master after merge
22:03:52 <jeblair> er, that's two things
22:03:55 <jeblair> both of them done!
22:03:59 <pabelanger> o/
22:04:14 <jeblair> #topic Status updates: Nodepool Zookeeper work
22:04:29 <phschwartz> afternoon
22:04:43 <jeblair> we switched production to running the master branch
22:04:48 <adam_g> o/
22:05:05 <jeblair> so changes related to maintaining the running zookeeper-based nodepool-builder should go there
22:05:05 <fungi> and it's working quite well
22:05:16 <pabelanger> ++
22:05:19 <jeblair> as well as changes preparing to make a release of it
22:05:23 <Shrews> fungi: surprising!
22:05:32 <jeblair> i wrote some doc changes last week to that end
22:05:34 * fungi is always surprised when computers work
22:06:19 <mordred> computers work?
22:06:31 <Shrews> speaking of that
22:06:39 <Shrews> doc fix: https://review.openstack.org/411844
22:06:54 <Shrews> for builder docs
22:07:12 <morgan> mordred: lies
22:07:27 <jeblair> Shrews: thanks!
22:07:59 <jeblair> i'm considering a revert of this change: https://review.openstack.org/396749  i'm working on that now
22:08:22 <jeblair> i bring that up because it affects a config-file format change
22:08:31 <jeblair> so i'd like to wait on that before we actually release
22:08:50 <jeblair> we also may want to land some race-condition mitigating changes that we were just talking about in #zuul
22:09:17 <jeblair> does anyone else have suggestions for what we might want to do before a release?
22:09:43 <Shrews> test the single zk change?
22:10:00 <jeblair> oh, i think that's in prod since friday?
22:10:02 <fungi> what's the argument behind the revert?
22:10:03 <Shrews> single zk connection, that is
22:10:32 <Shrews> jeblair: oh? are the builders automatically updated?
22:11:56 <jeblair> fungi: i'll write extensively in the commit message, but short version: that change cost us the ability to have the same image with different flavors.  i still want to rework that mapping (because it's *very* confusing), but i think 396749 may have been a step backwards -- *assuming* i can rework it the way i think i can.
22:12:09 <jeblair> fungi: if i fail at the revert, then nevermind anything i'm saying.  :)
22:12:46 <jeblair> fungi: (i hope that at least justifies why i'm looking into it, even if, at this point, i haven't fully gotten it in hand)
22:13:29 <fungi> okay, makes sense. thanks!
22:13:51 <jeblair> Shrews: they have to be manually restarted, which was done on friday
22:14:07 <jeblair> i *think* with that change
22:14:25 <jeblair> Shrews: when did it land?
22:14:58 <Shrews> https://review.openstack.org/411360
22:14:59 <nibalizer> o/
22:15:12 <jeblair> 16 dec 20:27 is the restart time
22:15:26 <jeblair> so about 3 hours after it landed
22:15:37 <Shrews> jeblair: cool. glad it works  :)
22:15:46 <jeblair> yay!
22:15:51 <Shrews> i wonder if the cpu load dropped
22:16:02 <Shrews> but, that's for #zuul
22:16:18 <jeblair> Shrews: i couldn't check right after i restarted because it was busy building/uploading.  we should look.  :)
22:16:45 <jeblair> any other pre-release items?
22:17:11 <jeblair> and perhaps we should not actually release until january
22:17:27 <jeblair> as a holiday gift to all our users :)
22:17:52 <clarkb> ++
22:18:03 <clarkb> setuptools released again the other day and we are lucky the world didn't burn down around us
22:18:05 <jeblair> (especially the ones who are running unpinned CD)
22:18:10 <clarkb> (it only slightly burned down)
22:18:34 <nibalizer> jeblair: as of last night, we are on master
22:18:35 <jhesketh> Not releasing is the gift? :-)
22:19:02 <jeblair> #agreed release nodepool master in early january
22:19:15 <jeblair> nibalizer: yay!  how's it workin out for you?
22:19:17 <jeblair> jhesketh: exactly
22:19:34 <nibalizer> jeblair: we think it working!
22:19:36 <jhesketh> We are generous
22:19:57 <jeblair> cool!
22:20:02 <Shrews> nibalizer: how adventurous of you
22:20:09 <fungi> no challenge getting zk set up then?
22:20:12 <nibalizer> it was all jamielennox
22:20:33 <nibalizer> im mostly reporting this to say we won't be affected by a relaese
22:20:39 <nibalizer> but I support the delay until january
22:20:47 <Shrews> nibalizer: that sounds like a challenge to us
22:20:48 <jamielennox> slightly burned on the removal or --no-images but an easy fix and it seems to be fine otherwise
22:21:25 * Shrews codes something to affect nibz
22:21:57 <jeblair> oh yeah, --no-images was sort of a desperate temporary move on our part; --no-deletes --no-launches and --no-webapp are all in the same boat and not long for this world.
22:21:58 * nibalizer codes something to effect Shrews
22:22:39 <fungi> --no-lunches
22:23:03 <pabelanger> om
22:23:14 <jeblair> and i guess one final thing -- the nodepool feature/zuulv3 branch is open for destructive work on the zuulv3 spec.  we'll periodically merge master into it to keep it up to date with operational improvements there.
22:23:42 <jamielennox> jeblair: i was thinking i'd have a crack at --no-deletes -> nodepool-deleter if no one else is?
22:24:51 <jeblair> jamielennox: well, in reality i don't think those should be separate.  rather the idea is to have multiple nodepool launchers (which also delete), possibly divided up by provider.
22:26:07 <jamielennox> ok, i'll leave it
22:26:08 <fungi> yeah, the delete worker split was done more because the old codebase made it a quick (if temporary) scaling solution
22:26:10 <jeblair> splitting deletes like that is really just because our install has grown so large we couldn't handle it in one process, and it was a fairly easy thing to do
22:26:25 <jeblair> fungi: right -- a quick solution, not a good one.  ;)
22:26:45 <fungi> saved more significant refactors under pressure
22:26:56 <jamielennox> it was mostly because our unit files had gotten weird, so i can wait for grander plans
22:27:32 <jeblair> jamielennox: if you have less than 1k instances available to you, i would recommend ignoring the "--no-" options and run them all in one daemon
22:29:19 <jeblair> let's move on to devstack-gate roles
22:29:27 <jeblair> #topic #status updates: Devstack-gate roles refactoring
22:29:40 <jeblair> that was weird but you get the idea
22:30:12 <jeblair> rcarrillocruz: how's it going?
22:30:19 <rcarrillocruz> so
22:30:22 <rcarrillocruz> good progress
22:30:28 <rcarrillocruz> we landed a few merges last week
22:30:35 <rcarrillocruz> https://review.openstack.org/#/c/401975/
22:30:41 <rcarrillocruz> https://review.openstack.org/#/c/402107/
22:30:45 <rcarrillocruz> https://review.openstack.org/#/c/402208/
22:30:48 <rcarrillocruz> ...
22:30:56 <rcarrillocruz> and nothing is burning
22:31:09 <mordred> yay for lack of fire!
22:31:18 <rcarrillocruz> special thanks to clarkb, for being extra vigilant on his reviews
22:31:44 <rcarrillocruz> i've closed the sb tasks to those commits where appropiate
22:32:10 <clarkb> rcarrillocruz: the next round of these will be pushed up soon ya? I think you have one up now?
22:32:13 <rcarrillocruz> and assigned to me the write localrc and some other ansible d-g refactor tasks that were unasssigned
22:32:34 <rcarrillocruz> clarkb: yeah, just waiting for the remove openvz thing to go thru, then i'll rebase the chain on it
22:32:38 <clarkb> kk
22:32:43 <rcarrillocruz> since that removal is in the middle of the refactor logic
22:33:16 <jeblair> rcarrillocruz: btw, i think if you add Story: and Task: commit message headers, gerrit will update the tasks automatically
22:33:28 <fungi> yep, it should do
22:33:33 <rcarrillocruz> ah nice, haven't tried, thanks
22:34:42 <jeblair> are third-party folks relying on the openvz support?
22:35:15 <fungi> i don't think the openvz implementation for nova ever went anywhere, did it?
22:35:25 <rcarrillocruz> i asked mriedem, and he agreed to the 3rd party guy that 'presumably' would be affected to just put that via a devstack plugin
22:35:37 <fungi> wasn't devananda working on that in the long-long ago, in the beforetime?
22:35:39 <clarkb> we notified them (they are the openvz devs) and they were fine with it
22:35:41 <clarkb> fungi: ya
22:35:50 <jeblair> cool
22:35:52 <Shrews> fungi: that may have been me?
22:36:03 <fungi> oh! indeed
22:36:13 <Shrews> but i'm old and forget things
22:37:49 <jeblair> anything else on this?
22:37:54 <rcarrillocruz> i'm good
22:38:01 <jeblair> cool, thanks!
22:38:10 <rcarrillocruz> ++
22:38:14 <jeblair> #topic Status updates: Zuul test enablement
22:38:57 <jeblair> we enabled more tests!  also deleted some that aren't necessary.
22:40:22 <jeblair> i believe adam_g is in progress on re-enabling the merger for all changes, before sending them to the launcher, mostly because we have a lot of functionality that depends on that.
22:40:58 <jeblair> it's come up in a few tests
22:41:02 <adam_g> ive been making some slow progress the last couple of days reenabling a handful of merger tests. should hopefully have a couple things ready to review in the next couple of days
22:42:26 <jeblair> adam_g: thanks
22:42:32 <jeblair> anyone else on this topic?
22:42:35 <adam_g> one thing ive noticed in working with v3 against v2.5, the testing upstream git repos no longer seem to get updated with merged changes in v3.
22:43:00 <adam_g> i assume this is not an issue since merging is a function of external things, but maybe i haven't seen the test yet that might interrogate those repos
22:43:15 <jeblair> adam_g: hrm, that sounds like a bug
22:43:15 <adam_g> or is it worth fixing that in the v3 test suite?
22:43:46 <jeblair> adam_g: i'm pretty sure there are some tests somewhere that rely on it
22:43:46 <adam_g> i haven't hit any tests yet that actually assert anything about the state of the repos, instead using the builds JobDirs
22:44:58 <jeblair> adam_g: it's possible it's zuul-cloner tests that do that, or something... if so, they may need significant alteration for v3.
22:45:14 <adam_g> jeblair: k, ill take a closer look when i get back to the merger stuff im hacking on
22:45:40 <jeblair> adam_g: thanks!
22:46:02 <jeblair> #topic Progress summary
22:46:43 <jeblair> SpamapS: around?
22:47:02 <jeblair> #link https://storyboard.openstack.org/#!/board/41
22:47:17 <adam_g> i think clint is offline today with a sick baby
22:47:21 <jeblair> :(
22:47:26 <jeblair> i updated the big pile of things in new last week
22:47:44 <jeblair> so things are moved to backlog/todo -- there's some more things in todo if peopel want to grab them
22:48:21 <jeblair> any other updates to the board?
22:50:05 <jeblair> #topic Open discussion
22:50:13 <jeblair> shall we cancel next week's meeting?
22:50:31 <Shrews> i do not plan to attend
22:50:39 <pabelanger> ya, will be traveling too
22:50:49 <jhesketh> I'm on leave from the 22nd-> 2nd of Jan (just as a general FYI), but if there's a meeting I can turn up :-)
22:50:52 <jeblair> i don't plan on attending or chairing :)
22:51:08 <fungi> i don't expect to be around. other stuff going on that evening
22:51:38 <jeblair> come to think of it, should we cancel jan 2 as well?
22:51:45 <pabelanger> Since nodepool-builder has been working so well, I've started using it as my front-end to diskimage-builder. Would love to get some eyes on https://review.openstack.org/#/c/412160/ . Biggest reason for this, fast way to get YAML support for diskimage-builder
22:52:00 <fungi> i am around then, but fine with not having a meeting if there's little interest
22:52:05 <clarkb> I will not be around
22:52:14 <clarkb> er for next week, not sure jan 2
22:52:31 <rcarrillocruz> i should be around the 2nd, but fine if we don't have meeting too
22:52:50 <jeblair> okay, let's cancel next week, but leave jan 2 on the books.  if it's short, it's short.  :)
22:52:54 <Shrews> pabelanger: couldn't you just set --upload-workers=0 and achieve the same thing?
22:53:06 <jhesketh> sounds good to me
22:53:11 <rcarrillocruz> wfm
22:53:24 <jeblair> #agreed there will be no dec 26, 2016 meeting
22:53:57 <pabelanger> Shrews: Hmm, not sure. But, the biggest reason for 412160 is to avoid also adding the providers section
22:54:06 <pabelanger> since I don't have any provider to upload too
22:54:18 <pabelanger> otherwise, I need to add a fake provider
22:54:23 <pabelanger> which is what I am doing today
22:54:25 <morgan> jeblair: haha, i would  not show up for a dec 26 meeting ;)
22:54:29 <morgan> jeblair: good call ^_^
22:54:41 <morgan> jeblair: also, i will be missing jan 2, fwiw
22:56:27 <mordred> I agree with morgan
22:57:33 <jeblair> let's wrap this up
22:57:35 <jeblair> thanks everyone!
22:57:37 <jeblair> #endmeeting