19:02:51 #startmeeting 19:02:52 Meeting started Tue Jul 19 19:02:51 2011 UTC. The chair is mtaylor. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:02:53 Useful Commands: #action #agreed #help #info #idea #link #topic. 19:03:04 dprince: you are quite welcome! 19:03:32 * jaypipes thinks mtaylor and jeblair doing a good job of creating blueprints in openstack-ci now. nice work. 19:03:42 yay! 19:03:55 thanks! i updated some this morning 19:03:55 jaypipes: but I'll have to point out that we are not to that topic yet. :) 19:03:58 very, very useful for folks to see what's up. 19:04:06 mtaylor: ah, sorry mate. pls continue! 19:04:11 #topic Actions from last meeting 19:04:30 alright -the first two are about machines for bare metal testing. 19:04:38 we don't have them yet, so that hasn't been accomplished 19:04:56 but - I did talk to folks about getting them, and we should have them in a couple of weeks. yay 19:05:39 the next ones are all really items that should be blueprint workitems (and there is no new update on them) 19:05:57 in a win for progress - we did get dprince set up with a dns entry! yay! 19:06:23 and nati doesn't seem to be here, so we can't ask him about list of existing tests 19:06:33 Yay! 19:06:46 any questions on those before we move on? 19:07:31 On the GitHub blueprint. 19:07:40 #topic Blueprints are up 19:07:56 What is the expected date for Nova/Glance conversion to GitHub? 19:08:05 #link https://blueprints.launchpad.net/openstack-ci 19:08:31 dprince: I'm guessing we're still probably a couple of weeks out 19:08:50 jaypipes: how solid do you want keystone testing to be before we work on migrating glance? 19:09:23 dprince: we're wanting to shake out process/code/implementation issues with smaller projects before throwing the mass of nova devs at the thing 19:09:36 mtaylor: heh, not sure. tbh, there isn't a huge chorus of glance contribs itching to move to GH ;) 19:09:45 hehe 19:09:55 mtaylor: other than dprince, that is ;) 19:09:58 jaypipes: are you planning on moving then? or staying? 19:10:26 mtaylor: we will use whatever system/platform is recommended by your team. 19:10:26 as I do believe the current policy is "project can decide to use approved environments", yeah? 19:10:32 jaypipes: hehe. ok 19:10:45 mtaylor: we're not sticklers about tooling. 19:10:54 jaypipes: To be clear I'm not pushing it. Just want to be ahead of the game. Thats all. 19:10:56 jaypipes: honestly, at the moment, the git/gerrit setup is going to be more robust and better integrated with jenkins 19:11:07 dprince: I'm just pulling your chain, mate. no worries :) 19:11:27 mtaylor: yes, that's what I've seen, and I've been impressed with the gerrit setup so far. 19:11:28 the gerrit plugin does essentially everything I wanted the tarmac replacement plugin for launchpad to do - and I didn't have to write it. :) 19:11:34 right. 19:11:53 jaypipes: I think my favorite thing is that it listens actively to a stream of events - so the INSTANT something happens in gerrit jenkins is testing stuff 19:12:06 no waiting for poll cycles 19:12:13 mtaylor: so... on a practical side of things, Glance could probably move to GH soon. Though it would be good to get through the existing code reviews. About 5 of them outstanding. 19:12:14 mtaylor: One question about your Git pulls. Do you plan on cloning or using the GitHub tarball URLs. 19:12:42 jaypipes: I plan on taking a look at the caching stuff today/tomorrow. 19:12:52 jaypipes: Rick's branch. 19:12:59 dprince: I'm not sure what you mean by "your git pulls" ... but I do not believe that github tarball urls provide any value 19:13:36 mtaylor: I just meant each time Gerrit runs does it use a local repo (clone) or a tarball. 19:13:46 mtaylor: I think you answered my question. 19:13:50 dprince: :) 19:13:59 dprince: so, they way it works is actually this: 19:14:10 dprince: yeah, I've already read through it. Looks pretty good, though I wish folks would stick to doing just the feature in a single merge proposal, and not random refactoring and style cleanups, too ;) 19:14:42 gerrit holds a repo internally for each project 19:14:52 code reviews are triggered by pushing to gerrit 19:15:01 once it's reviewed and merged, gerrit replicates to github 19:15:09 so it's all VERY git-y 19:15:45 great 19:16:25 jeblair: where's that wiki page? 19:16:44 #link http://wiki.openstack.org/GerritWorkflow 19:17:03 yay. thanks 19:17:11 docs ++ 19:17:41 we're working on more of them - I should have a http://ci.openstack.org up and going soon for us with docs on the whole CI infrastructure 19:17:51 well - I mean, that's what WILL go there 19:18:03 the site will be up soon, then we can ADD docs on the whole infrastructure to it :) 19:18:27 mtaylor: excellent. 19:18:30 i'm working on adding docs on the installation / integration process for gerrit, github, and jenkins 19:18:58 also - in case anyone's been following along at home, I think we've got a solution for the problem of needing keys on the jenkins slaves. once that's sorted, we can move the rest of the jobs to all build on slaves 19:19:05 so how we set up this process will be fully documented and repeatable, and the recipe is out there if others want to do the same thing 19:19:44 then we can move jenkins to a new server ... and then I can stop dorking around with that and help jeblair with getting the +2 events and or the openid issue sorted (which I believe are our two main pain points with gerrit right now, yeah?) 19:19:56 jeblair: woot. 19:20:05 that and the launchpadlib sync script 19:20:12 I _definitely_ want to figure out the tests-run-after-+2 thing before nova gets brought over. 19:20:14 oh yes. 19:20:24 so, in a perfect world, we will solve those three issues before nova migrates 19:20:51 everything past that is bunnies, rainbows and gravy 19:20:54 i'm wondering whether all the keystone devs will register with gerrit before i get around to writing the script that would obviate their need to register ;) 19:21:04 mtaylor: for us morons in the audience, the run-after-+2 thing is because Gerrit runs tests when a pull request is received, not after approvals, right? 19:21:05 * mtaylor is guessing no 19:21:15 jaypipes: that is correct 19:21:18 k 19:21:44 mtaylor, jeblair: rough timeframe on solving those three issues? 19:21:44 and, although that is nice for parallelism ... it's problematic from a "any damn person can now submit code and have that run on our farm of servers" perspective 19:21:58 mtaylor, jeblair: is there anything I can do to assist you? 19:22:41 well - two of them involve java. the launchpadlib sync script is python and shouldn't be too hard to do - but I'm guessing it would take jim more time to knowledge dump what's needed there 19:23:11 mtaylor: a week of work? two days? what's your feeling about time needed for that? 19:23:26 mtaylor: and is there anything we can do to help? 19:24:06 I'm going to defer to jeblair there ... what do you think jeblair? 19:24:25 launchpadlib is probably just a couple of days. i may get to start on it this afternoon 19:25:13 +2 is really waiting on some changes to the gerrit jenkins plugin -- we may start on it before then, but we shouldn't plan on having something final with that for, what did they say, a few weeks? 19:25:43 yeah - although if it's not hard to do, it might be worth having a divergent version of the plugin for a few weeks 19:25:58 jeblair: can we deploy gerrit before the +2 thing is fixed? or is that a full blocker? 19:26:09 I think if we put the nova guys off for a _few_ weeks (i.e. closer to a month) that they'll get antsy 19:26:21 mtaylor: right. agreed. 19:26:45 the openid change is a relatively simple change to gerrit; probably a few days. 19:27:11 then yeah, we should hack something together. i got the impression that the event refactoring they were doing wasn't the highest priority, so it could slip on their side. 19:27:15 the +2 thing will mean a workflow change, and I'd prefer to not introduce the confusion of "here's the new workflow" and then a week later "oh! we changed it" 19:27:25 jeblair: agree 19:27:42 mtaylor: agree 19:27:43 just to be clear, we can continue with keystone in our current state, as soon as they finish signing up 19:27:53 yes. and glance, for that matter 19:28:14 it's easier to use a temporary workflow with more caveats with a couple of small teams 19:28:14 I just don't want to do workflow changes like that for 100 devs 19:28:20 ++ 19:28:37 (even if the 2nd workflow is easier) 19:29:43 so, I'd say actually though - we should do the other two first - openid and launchpad sync script - because those affect the signup/initial process - and that way we give ericsson a brief chance to make their changes before we briefly fork 19:30:12 unless folks think I'm nuts 19:31:03 that was my thinking. 19:31:09 fine by me. 19:32:01 sweet. should we target launchpadlib script and openid as actions for next week? (jeblair I could probably get the launchpadlib sync script going since I wrote the ssh user one - unless you really want to do it) 19:33:11 that works for me 19:33:15 cool 19:33:39 #action jeblair Fix OpenID support in Gerrit 19:34:01 #action mtaylor Write script to sync launchpad users/teams to gerrit users/groups 19:34:19 and just for completeness 19:34:33 #action mtaylor finish migrating jenkins jobs from running on master to running on slaves 19:36:18 that's about all I've got... any other discussion on the openstack-ci blueprints? 19:37:13 #topic Open Discussion 19:37:23 how about anything that isn't about openstack-ci blueprints? 19:37:58 nothing here 19:38:20 sweet. 19:38:46 than I'ma gwanna call this bad boy done. early even! 19:38:48 #endmeeting