19:00:47 <jeblair> #startmeeting ci
19:00:48 <openstack> Meeting started Tue Aug 28 19:00:47 2012 UTC.  The chair is jeblair. Information about MeetBot at http://wiki.debian.org/MeetBot.
19:00:49 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
19:00:50 <openstack> The meeting name has been set to 'ci'
19:01:31 <jeblair> so, i think we've talked about most of the current projects on the ML...
19:01:54 <jeblair> clarkb is working on some issues with the SCP plugin that copies logs to the log server
19:01:56 * LinuxJedi should really read that this week
19:02:29 <jeblair> when we're satisfied with that, we'll cut back on the build history that jenkins keeps
19:02:34 <jeblair> which should hopefully speed things up.
19:02:52 <LinuxJedi> ++
19:03:15 <clarkb> as part of that work I have also gotten the nosetests html output to include skipped tests
19:03:43 <jeblair> i'm working on getting the jobs to use the refs that zuul is composing, which should make gerrit-git-prep simpler, and actually allow us to switch to cherry-pick mode if needed.
19:04:17 <jeblair> though it looks like there's a bug in zuul that i need to look at first
19:04:36 <jeblair> (related to dequeuing failed changes)
19:05:15 <jeblair> after that, i'd like to do some work to enable us to run multiple jenkins masters...
19:05:29 <jeblair> i think dividing the load will help some....
19:05:51 <clarkb> jeblair: is that something zuul would need to be aware of?
19:05:53 <jeblair> part of our problems, i think, stem from having so many slaves; this should cut down on the number of threads in use.
19:05:57 <jeblair> clarkb: yep
19:06:10 <jeblair> clarkb: and maybe the jenkins job builder too.
19:06:31 * LinuxJedi is looking forward to the multiple masters
19:06:41 <LinuxJedi> should make things easier to find :)
19:06:50 <jeblair> i'm thinking that zuul should use gearman to dispatch jobs, and then there should be a little zuul agent on each master that handles interacting with jenkins.
19:07:41 <LinuxJedi> so a project goes to a random jenkins rather than a jenkins per-project?
19:08:02 <LinuxJedi> run goes to a random jenkins even
19:08:32 <jeblair> maybe, or, i've been thinking we could try having an oneiric, precise, and devstack jenkins... so split them up by node type
19:09:33 <jeblair> but i think the general approach should be flexible, so we can split up the jenkinses however we need
19:10:17 <jeblair> (side effect, this will tie zuul to jenkins slightly less -- things other than jenkins could join the gearman system)
19:11:14 <jeblair> we also have some new zuul features -- the non-voting tests we talked about last week are live, and running for nova pylint
19:11:27 <jeblair> and i'm testing out running coverage that way too
19:12:07 <jeblair> fungi has joined us
19:12:12 <jeblair> welcome fungi!
19:12:17 <fungi> thanks!
19:12:48 <jeblair> fungi is looking into the trivial-rebase hook
19:12:54 <jeblair> a long requested feature.
19:13:05 <fungi> yeah, claimed on lp and mostly working on my dev vm
19:13:21 <jeblair> awesome!
19:13:29 <LinuxJedi> also mawagon1 and dkehn have joined us
19:13:29 <fungi> my lack of prior exposure is biting me there, so i'm brushing up on gerrit hooks at the moment
19:13:57 <jeblair> fungi: i think you'll become a gerrit expert fast with this one.  :)
19:14:06 <fungi> looks trivial enough
19:14:12 <jeblair> heh
19:14:16 <fungi> ;)
19:15:56 <jeblair> anyone else have things they're working on they'd like to talk about?
19:16:26 <fungi> i ran into a handful of things so far
19:16:31 <fungi> http://wiki.openstack.org/Meetings/CITeamMeeting
19:16:41 <fungi> i assume that's not updated regularly any longer?
19:16:51 <fungi> or is there a more proper home for it?
19:17:39 <jeblair> nope, our normal chair does not regularly update it.
19:17:59 <jeblair> at least partially in response to that (there ore some others with similar habits), i updated meetbot to require a meeting nam
19:18:11 <fungi> that works
19:18:15 <LinuxJedi> and a name
19:18:19 <jeblair> name, so future meetings will be findable in the directory structure
19:18:29 <clarkb> on eavesdrop.openstack.org
19:18:32 <jeblair> #link http://eavesdrop.openstack.org/meetings/
19:19:02 <jeblair> this one is already in progress.  :)  http://eavesdrop.openstack.org/meetings/ci/2012/ci.2012-08-28-19.00.log.txt
19:19:15 <fungi> another trivial item... the gerrit initscript. should we add lsb headers, do you think?
19:20:17 <jeblair> fungi: that might be a good patch for upstream gerrit
19:20:25 <fungi> i would assume so
19:20:32 <fungi> good point
19:20:44 <fungi> i'll try to get a patch to them when i get a moment
19:21:09 <clarkb> I think mtaylor may need to do something with their CLA
19:21:36 <clarkb> seems like he had to add me after I got an account on their gerrit server
19:21:55 <jeblair> yep, but that's an hp-specific thing we don't need to talk about in the openstack meeting.  :)
19:22:09 <clarkb> nevermind then
19:22:37 <jeblair> fungi: what else ya got?
19:22:50 <fungi> on a similar gerrit note, the instructions on the ci site assume debian (ubuntu) package management, but then use useradd instead of adduser
19:23:05 <fungi> maybe just a personal nitpick
19:23:09 <jeblair> fungi: we'll probably just drop that since it's all managed by puppet now
19:23:35 <fungi> point taken
19:23:38 <LinuxJedi> fungi: one is an alias of the other
19:23:57 <jeblair> fungi: in the long run, we don't need installation instructions we don't use.  that was how it was initially set up before we puppetized everything.
19:24:01 <fungi> really? thought adduser was a wrapper with different semantics
19:24:21 <fungi> jeblair: makes sense
19:25:17 <fungi> anyway, that's all i have so far
19:25:22 <jeblair> #topic open discussion
19:25:29 <jeblair> (as if it hasn't been)
19:25:39 <jeblair> anyone else?
19:26:27 <LinuxJedi> MRaaS has been re-added to Stackforge
19:26:34 <LinuxJedi> that is about all I remember
19:27:06 <jeblair> LinuxJedi: cool; you brought the jobs up to current standards, right?
19:27:22 <jeblair> LinuxJedi: and are they committing yet?
19:27:43 <LinuxJedi> jeblair: well 'job' only the merge one for now.  I haven't seen them committing yet, I have been teaching them how git/gerrit/everything works...
19:28:04 <LinuxJedi> jeblair: it is a Java project so we have to figure out how to test that at some point
19:28:59 <jeblair> lovely
19:29:35 <jeblair> oh, keystone has a feature branch
19:29:46 <jeblair> so there is now a feature/keystone-v3 branch in gerrit
19:30:12 <jeblair> we're experimenting with letting the keystone-drivers group land merge commits there, to keep it going as a long running feature branch
19:30:28 <jeblair> that will eventually merge into master, some time during grizzly i expect
19:30:47 <clarkb> they are pushing directly to the branch?
19:31:01 <jeblair> no, it's still code reviewed in gerrit
19:31:19 <jeblair> the difference is that group has an acl that lets them upload a merge commit
19:31:38 <jeblair> #link http://wiki.openstack.org/GerritJenkinsGithub#Merge_Commits
19:31:46 <jeblair> there's the preliminary documentation on that.
19:32:09 <jeblair> this is something we discussed a bit at the last summit, but this is the first time it's been requested
19:32:13 <jeblair> so it's still pretty experimental.
19:32:47 <jeblair> but because of the way we write all of the jenkins jobs, etc, it shouldn't require any extra support.
19:34:51 <jeblair> anything else?
19:38:12 <jeblair> i reckon that's it then, thanks!
19:38:14 <jeblair> #endmeeting