19:01:05 <jeblair> #startmeeting infra 19:01:05 <openstack> Meeting started Tue Sep 10 19:01:05 2013 UTC and is due to finish in 60 minutes. The chair is jeblair. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:01:06 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:01:08 <openstack> The meeting name has been set to 'infra' 19:01:29 <jeblair> #link https://wiki.openstack.org/wiki/Meetings/InfraTeamMeeting 19:01:55 <jeblair> #link http://eavesdrop.openstack.org/meetings/infra/2013/infra.2013-09-03-19.03.html 19:02:18 <anteaya> o/ 19:02:25 <jeblair> #topic Salt (UtahDave) 19:02:51 <fungi> #link https://review.openstack.org/#/c/45898/1 19:03:04 <fungi> that's up for review 19:03:19 <jeblair> neat. i wasn't around much yesterday; what do we need to discuss here? 19:03:26 <UtahDave> jeblair: Should I describe the purpose, or is the commit clear enough? 19:03:32 <fungi> we discussed the runaway processes/memory leak and there was apparently an issue with the message queue and unaccepted minion certs 19:04:14 <clarkb> one big thing that came up was if and how we should determine which nodes need puppet kicked when a config change merges 19:04:29 <clarkb> I think we settled on not worrying about it initially and just kicking everything 19:04:35 <jeblair> clarkb: sounds good to me 19:04:45 <fungi> UtahDave: do you happen to know which salt release fixed the unaccepted minion certs leak? (or did you already tell me and i missed it?) just want to make sure that won't be an issue for us long-term 19:04:45 <jeblair> fungi: are those things resolved or do we need to make some decisions? 19:05:23 <fungi> i think it's mostly just getting confirmation on that situation before we turn it back on everywhere 19:05:28 <UtahDave> fungi: that issue should be fixed in the 0.16 branch. If I recall correctly, your server was running 0.15.3 19:05:36 <mordred> o/ 19:05:37 <fungi> UtahDave: thanks! 19:05:58 <fungi> so i'll check to make sure we don't end up with older salt anywhere 19:06:12 <fungi> and if the issue we were seeing crops back up, then it's something else 19:06:35 <UtahDave> fungi: sure. If it crops back up we can dedicate some engineering resources to help track down the problem 19:06:51 <mordred> so, if I understand the patch correctly, it means we would run a job on the jenkins master which would execute that salt command, yeah? 19:07:04 <fungi> UtahDave: keen. i'll need to look back through how we were installing it to make sure we have channels to new enough versions on our systems 19:07:31 <fungi> mordred: it wouldn't have to be the on the jenkins master--that was just a first stab i think 19:07:42 <UtahDave> mordred: correct. The jenkins user should run sudo salt-call event.fire_master 'some data' 'jenkins' 19:07:50 <mordred> fungi: ok 19:07:52 <fungi> mordred: it could be a specific node-label or even a dedicated slave if we really wanted 19:07:57 <mordred> kk 19:08:15 <mordred> UtahDave: what are "some data" and "jenkins" 19:08:35 <UtahDave> 'jenkins' is the tag that the Reactor is scanning for. 19:08:49 <mordred> UtahDave: like, if the thing I want to acheive on each node is "puppet agent --test" ... would I do sudo salt-call event.fire_master 'agent --test' 'jenkins' ? 19:09:19 <UtahDave> The first item is the "data" field in which you can put any arbitrary pertinent data. 19:09:35 <zaro> ~.~. 19:09:40 <UtahDave> The current setup is not using the data field 19:09:57 <jeblair> i don't think we want to pass parameters 19:10:06 <UtahDave> mordred: I would avoid allowing the jenkins server to pass in commands to be run 19:10:08 <jeblair> i think we want jenkins to say "run puppet" and have salt know how to do that 19:10:31 <mordred> UtahDave: great. and I agree 19:10:38 <UtahDave> So right now, when the reactor sees the 'jenkins' tag it just executes the /srv/reactor/tests.sls 19:10:38 <fungi> yeah, from a security perspective we just want to make sure that the slave where this job runs can tell the salt master to do one thing and one thing only (for now, and expand to a vetted list later if desired) 19:10:44 <mordred> so we'd want to do salt-call event.fire_master '' 'jenkins' 19:10:58 <UtahDave> yes, exactly. 19:11:01 <mordred> coool 19:11:17 <jeblair> this seems like a safe thing to do on the jenkins master. is that where we should run it? 19:11:28 <mordred> jeblair: seems like a safe place to me 19:11:37 <mordred> and also less work than other things 19:11:57 <UtahDave> I think it's pretty safe on the jenkins server based on the sudo privileges the jenkins server has 19:11:57 <fungi> agreed. it's flexible enough we could put it wherever we want slave-wise, but should be fine on a jenkins server as well 19:12:03 <anteaya> will it run on all jenkins masters, or just one? 19:12:26 <fungi> the way it's written now, all i think 19:12:32 <anteaya> k 19:12:45 <fungi> so whichever one zuul picks at random 19:13:12 <fungi> though the job itself is not written yet 19:13:41 <UtahDave> fungi: correct. There would need to be a jenkins job written that executed the above mentioned salt-call command when appropriate 19:14:00 <jeblair> this is a review comment, but i'd imagine we don't want that sudo command defined everywhere, so we'll probably want to put a sudoers.d fragment just on whatever jenkins master/slave will run this 19:14:13 <fungi> jeblair: yeah, i was thinking the same 19:14:24 <fungi> right now this sets it on every server where we install sudo 19:14:35 <fungi> but easily addressed 19:14:52 <jeblair> (which is actually making me lean slightly toward having a slave for this; i'd like to trust the \d\d masters less in the future) 19:15:26 <fungi> anyway, mostly just wanted to sync up on comfort level for turning salt back on and making sure it's the right version to theoretically avoid the previous issue we were seeing 19:15:38 <fungi> sounds like we're cool with that? 19:15:51 <jeblair> fungi: sounds like it; and we can go over the finer points in reviews 19:15:58 <fungi> perfect 19:16:12 <clarkb> ++ 19:16:14 <mordred> ++ 19:16:21 <jeblair> #topic Marconi migration from stackforge -> openstack (flaper87) 19:16:27 <jeblair> flaper87: hi there! 19:17:00 <jeblair> marconi was accepted for incubation 19:17:11 <jeblair> and i think they would like an org move 19:17:26 <clarkb> yes, that is my understanding 19:17:52 <clarkb> #link https://review.openstack.org/#/c/44963/ 19:18:20 <clarkb> that is a WIP change that can be merged after the manual steps of moving a project are completed 19:18:24 <jeblair> also, it would be cool to know what kind of testing they're planning on 19:18:57 <jeblair> will they be doing devstack-gate tests, etc... 19:19:14 * mordred would love to know that 19:19:19 <jeblair> but since flaper87 doesn't seem to be around anymore (though he was here at the beginning of the meeting).... 19:19:30 <jeblair> i guess we'll shelve this for now 19:19:32 <clarkb> jeblair: maybe come back to this when we are done with the other agenda items? 19:19:46 <jeblair> #topic Trove testing (jeblair) 19:19:53 <jeblair> also, mordred, hub_cap ^ 19:20:04 <jeblair> real quick: 19:20:05 <hub_cap> heloo helooo 19:20:15 <jeblair> i've put a couple of project testing related topics on the agenda 19:20:28 <jeblair> trove, tripleo, and xen.... 19:20:37 * mordred supports this 19:20:37 <jeblair> because there are efforts to get upstream ci testing going for all of those 19:20:49 <mordred> so - hub_cap - how's upstream ci testing going for trove? 19:20:53 <jeblair> and i want to make sure that we're being supportive of those, and they don't slip through the cracks 19:21:01 <mordred> ++ 19:21:10 <hub_cap> going as in, how is it going w/ us running it? 19:21:35 <hub_cap> the only problems we have is the plugin hp uses to spin up builds... which wiould be much nicer if done by yall 19:22:00 <hub_cap> or do you mean, hows the integration w/ teh gate going, mordred? (cuz thats not happened yet) 19:22:18 <mordred> how's the integration with the gate going? 19:22:25 <hub_cap> for me, i see our devstack integration as gating for me to get the integration w teh gate 19:22:39 <hub_cap> https://review.openstack.org/#/c/38169/ 19:22:51 <hub_cap> its been going back and forth and SlickNik is doing it on free time 19:22:52 <mordred> awesome. I'll go star that review 19:22:54 <hub_cap> so i might take it over 19:22:59 <hub_cap> and push it forward 19:23:18 <mordred> I think that getting trove into devstack is a valid first step for sure 19:23:23 <hub_cap> yes yes 19:23:34 * flaper87 is here 19:23:35 <hub_cap> then i can focus on the special steps for our tests in teh gate 19:23:46 <mordred> anything you need from us this week (I'm guessing no, since you're waiting on devstack itself right now) 19:24:06 <hub_cap> mordred: correct. ill come to you when i need to start integrating. 19:24:14 <hub_cap> lets say late this wk, early next 19:24:21 <hub_cap> depoending on the reviews for devstack 19:24:43 <hub_cap> <feel free to pull me in next wk to check my status jeblair 19:25:08 <mordred> hub_cap: I believe we're goign to pull you in weekly until such as time as you're integrated 19:25:35 <jeblair> hub_cap: cool, thanks 19:25:48 <hub_cap> mordred: jeblair good by me 19:25:52 <hub_cap> <3 19:25:58 <hub_cap> itll keep me workin on it ;) 19:26:25 <jeblair> yay! ok, back to marconi 19:26:31 <jeblair> #topic Marconi migration from stackforge -> openstack (flaper87) 19:26:58 <jeblair> flaper87: hi, so one of the things we want to discuss (in addition to the org move) is testing for marconi 19:27:06 <flaper87> sorry, got disconnected 19:27:06 <flaper87> did I miss my chance ? 19:27:06 <flaper87> :( 19:27:39 <jeblair> flaper87: will you be doing integration tests with devstack, or something similar? 19:27:48 <jeblair> flaper87: (what are marconi's integration points with the rest of openstack?) 19:29:03 <jeblair> this isn't going very well, is it? 19:29:36 <clarkb> :/ 19:29:44 * fungi mails flaper87 more internets 19:29:54 <jeblair> #topic Tripleo testing (jeblair) 19:30:06 <jeblair> ok, so tripleo is a program now 19:30:30 <mordred> yup. so probably stuff should get tested and stuff 19:30:35 <clarkb> ++ 19:30:43 <jeblair> and while it isn't part of the integrated release, it would still be great if whatever testing that is done could be done with this neato infra we have 19:30:56 <pleia2> so with the baremetal stuff I have a sketched out plan to use portions of tripleo with lxc 19:31:09 <pleia2> but still slogging through some issues running openstack in lxc 19:31:20 <jeblair> pleia2: how does this relate to "toci"? 19:31:34 <jeblair> (i don't really know what any of these things are as i've never seen them) 19:31:38 <pleia2> jeblair: I'll be using portions of toci 19:32:04 <mordred> toci is basically a scripted version of https://git.openstack.org/cgit/openstack/tripleo-incubator/tree/devtest.md 19:32:10 <pleia2> but toci is designed to run on actual bare metal, whereas we're all virtual (so lxc and qemu) 19:32:12 <mordred> which is the walkthrough on what it takes to install tripleo 19:32:20 <jeblair> so a thing that got me thinking about this is this bug: https://bugs.launchpad.net/openstack-ci/+bug/1217815 19:32:21 <uvirtbot> Launchpad bug 1217815 in openstack-ci "Tripleo ci service account in gerrit" [Undecided,New] 19:32:32 <pleia2> so I'm writing patches for tripleo scripts to support lxc, and eventually will have to patch toci to do the same 19:32:51 <mordred> lifeless, SpamapS: around? we're talking about you in here 19:33:16 <jeblair> which got my attention because most openstack programs don't have their primary testing infrastructure hosted outside of openstack 19:33:24 <pleia2> but full tripleo is more complicate than what I'd doing (since my goal is testing baremetal nova driver, not tripleo) 19:33:29 <pleia2> complicated 19:33:39 <pleia2> I just happen to be using tripleo to do it 19:33:50 <ttx> o/ 19:34:03 * flaper87 is here 19:34:05 <lifeless> mordred: hi, yes in tuskar meeting just now 19:34:07 <flaper87> sprry, I got disconnected 19:34:08 <lifeless> mordred: then OSRB 19:34:10 <lifeless> mordred: then physio 19:34:11 <flaper87> did I miss my chance? 19:34:16 <lifeless> mordred: then maybe work ::P 19:34:19 <flaper87> :) 19:34:26 <mordred> lifeless: well, we're talking about infra testing of tripleo 19:34:30 <lifeless> cool 19:34:38 <anteaya> flaper87: I think jeblair will try to give you another shot 19:34:40 <lifeless> it needs to be openstack-infra'd as soon as possible 19:34:48 <lifeless> was talking with derekh about it last night 19:34:52 <mordred> lifeless: we'd like that - but we don't really know what that means 19:35:03 <jeblair> so we want to find out who to talk to about that 19:35:32 <pleia2> I will be at the tripleo sprint next week, so I can have some chats then 19:35:37 <lifeless> ok, so me 19:35:41 <clarkb> I will be there as well 19:35:46 <clarkb> (ish) 19:35:53 <lifeless> derekh is more familiar with the toci plumbing, but he's on leave for 2 weeks. 19:36:00 <pleia2> clarkb: cool, maybe we schedule some time to talk specifically about testing with them? 19:36:08 <clarkb> pleia2: that sounds like a good idea 19:36:14 <mordred> lifeless: aiui, that runs on some metal that is laying around somewhere, right? 19:36:15 <pleia2> lifeless: can we add this to sprint schedule? 19:36:22 <pleia2> mordred: yeah, I think it's at redhat 19:36:24 <lifeless> pleia2: it's an etherpad... :P 19:36:31 <lifeless> mordred: yes, which is a big scaling problem. 19:36:31 <pleia2> lifeless: oh right :) 19:36:37 * pleia2 digs up the etherpad 19:36:49 <lifeless> mordred: I want to remove all the redundancy between it and the gerrit /zuul/jenkins infra 19:37:05 <lifeless> mordred: turn it into a focused test runner script 19:37:57 <jeblair> i think engineering this is far too large of a topic for this meeting 19:38:01 <mordred> totally 19:38:08 <pleia2> clarkb: penciled in for thursday https://etherpad.openstack.org/tripleo-havana-sprint 19:38:11 <jeblair> so the useful things to know are who's leading the effort 19:38:12 <mordred> I think the outstanding question is qhat to do about the toci service account request 19:38:20 <clarkb> jeblair: agreed. I think if pleia2, mordred and I sit in a session at their sprint we should be able to get somewhere next week 19:38:21 <mordred> and who to talk to in general 19:38:27 <jeblair> and where/how should we track the design? 19:38:48 <jeblair> clarkb, pleia2: thank you 19:39:08 <mordred> I'd say the goal for next week shoudl be an etherpad or somethign with a design on it 19:39:15 <pleia2> ++ 19:39:17 <mordred> that we all fel comfortable we can communicate to jeblair 19:39:29 <mordred> without saying "oh, I guess you needed to have been there" 19:39:33 <clarkb> ++ 19:39:40 <lifeless> waiting for the sprint would be a mistake :). derekh's not there, lets get rolling on discussions. 19:39:51 <lifeless> suggest, either a dedicated etherpad, or ml discussion, or both 19:40:00 <mordred> I'd say etherpad 19:40:05 <mordred> ml discussion wrong scope level 19:40:24 <mordred> and also some IRC outside of this meeting 19:41:01 <clarkb> jeblair: so I think lifeless is the person to talk to now, derekh becomes the person when back. And an etherpad will be the place to track the design 19:41:09 * fungi is bowing out to drive to red hat hq. if we discuss the marconi org move scheduling, i'm free to help basically any saturday/sunday for the forseeable future 19:41:21 <pleia2> ok, here we go: https://etherpad.openstack.org/tripleo-initial-testing 19:41:27 <jeblair> fungi: have fun, thanks 19:41:31 <clarkb> #link https://etherpad.openstack.org/tripleo-initial-testing 19:41:40 <pleia2> thanks clarkb 19:41:43 <clarkb> jeblair: does that cover what we need to do in this meeting? 19:41:56 <jeblair> clarkb: yep. thanks 19:42:10 <jeblair> flaper87: around? 19:42:13 <flaper87> yup 19:42:18 <jeblair> #topic Marconi migration from stackforge -> openstack (flaper87) 19:42:33 <jeblair> 19:28 < jeblair> flaper87: will you be doing integration tests with devstack, or something similar? 19:42:33 <jeblair> 19:29 < jeblair> flaper87: (what are marconi's integration points with the rest of openstack?) 19:42:52 <flaper87> jeblair: I already have a patch ready for devstack 19:43:03 <flaper87> so we'll be doing it w/ devstack 19:43:09 <jeblair> flaper87: awesome! do you have a link to that? 19:43:26 <flaper87> jeblair: yup, https://github.com/FlaPer87/devstack/tree/marconi 19:43:33 <flaper87> I haven't submited it for review 19:43:40 <flaper87> because I was waiting for marconi to be migrated 19:43:46 <flaper87> and for another patch in requirements to land 19:43:59 <flaper87> which alread landed 19:44:21 <flaper87> we're already integrated with the rest of the infrastructure 19:44:38 <jeblair> flaper87: ok. are there any unusual requirements for running it in devstack? 19:45:10 <flaper87> don't think so, the most unusual would be mongodb but ceilo already uses it 19:45:24 <clarkb> more mongo? :/ 19:45:36 <flaper87> plus, we can run tests on sqlite 19:45:37 <jeblair> ok. well, the're about to (they actually use mysql atm), but that should be in place by the time your stuff lands 19:46:06 <flaper87> so, mongodb is not a "requirement" for tests 19:46:24 <jeblair> flaper87: all right, that all sounds pretty easy then. 19:46:25 <clarkb> jeblair: that depends on zul getting newer mongodb into cloud archive right? 19:46:30 <flaper87> but it would be nice to be able to run tests against mongodb, anyway 19:47:10 <jeblair> clarkb: jd__ pushed up some changes that lead me to believe that's happened. 19:47:15 <clarkb> jeblair: cool 19:47:50 <jeblair> flaper87: so when would be a good time to perform the repo rename? we usually try to do it during a quiet period 19:48:06 <jeblair> at this point, we could probably do a friday afternoon US-time, or weekend 19:48:25 <flaper87> jeblair: either work for us 19:48:40 <jeblair> i don't think i could help this friday or this weekend, but am available next weekend (though fungi said this weekend was fine) 19:48:54 <flaper87> ok, this weekend it is 19:48:56 <flaper87> :D 19:48:58 <mordred> I cannot help this weekend either 19:49:34 <jeblair> clarkb: thoughts? 19:49:43 <clarkb> I think I can do this weekend. Why don't we plan for Saturday at like 1700UTC and check with fungi when he is back? 19:49:59 <flaper87> sounds good to me 19:50:37 <jeblair> #action fungi clarkb move marconi saturday sept 14 1700 utc 19:50:44 <flaper87> w000000000000000000000t 19:50:47 <flaper87> thanks guys 19:50:49 <jeblair> clarkb: thanks 19:50:56 <flaper87> clarkb: fungi thanks :) 19:51:04 <jeblair> flaper87: thanks for being on top of things! 19:51:13 <flaper87> my pleasure 19:51:24 <jeblair> #topic Xen testing (jeblair) 19:51:45 <jeblair> I also put this on the agenda, but not with enough notice to make sure that BobBall could be here 19:52:04 <jeblair> because i want to make sure we don't lose track of his amazing effort to test xen 19:52:10 * mordred is interested in his amazing efforts 19:52:21 <jeblair> so we'll try to catch up with him later 19:52:45 <jeblair> #topic puppet-dashboard (pleia2, anteaya) 19:52:54 <jeblair> what's the latest? 19:53:02 <anteaya> I have a sodabrew-dashboard up using ruby 1.9.3 19:53:11 <anteaya> though the package is called ruby1.9.1 19:53:13 <pleia2> on a test vm for now 19:53:23 <anteaya> and a puppet client server 19:53:32 <anteaya> I used these instructions: http://paste.openstack.org/show/46510/ 19:53:42 <anteaya> now I am trying to get them talking to each other 19:54:04 <pleia2> once we have everything running, we'll dive into what we need to change in the puppet-dashboard module to support sodabrew instead 19:54:49 <anteaya> separate servers since puppet client uses ruby1.8 19:55:05 <anteaya> I think 19:55:11 <pleia2> yeah 19:55:24 <jeblair> what do you mean 'separate servers'? 19:55:27 <anteaya> yup/l $ ruby -v 19:55:28 <anteaya> ruby 1.8.7 (2011-06-30 patchlevel 352) [x86_64-linux] 19:55:29 <pleia2> need to make sure puppet with ruby1.8 can talk to dashboard with 1.9 19:55:38 <pleia2> jeblair: testing infrastructure 19:55:51 <anteaya> they each have their own vm 19:55:54 <pleia2> have a puppet dashboard server and a client that looks like some of our regular clients 19:56:02 <jeblair> ah, gotcha 19:56:14 <clarkb> a little mini infra 19:56:19 <pleia2> very little :) 19:56:45 <jeblair> sounds promising 19:56:50 <anteaya> yay 19:56:55 <jeblair> #topic Open discussion 19:57:08 <jeblair> #action jeblair send email update about asterisk testing 19:57:25 <clarkb> jeblair: please revwiew https://review.openstack.org/#/c/45928/1 19:57:32 <pleia2> I'm completely unreachable on saturday (no marconi for me!) and as mentioned flying to seattle sunday for the tripleo sprint 19:58:08 <mordred> I'll be in New Orleans over the weekend and early next week. I will then be in Seattle late next week, I will then be back in NYC 19:58:29 <mordred> it's possible that next week's meeting might be difficult... 19:58:39 <clarkb> jeblair: are you in New Orleans as well? 19:58:39 <ttx> mordred: I was thinking the same 19:58:40 <jeblair> i'm flying to nola on friday, so won't be around then 19:58:47 <mordred> if pleia2 is going to be in Seattle and jeblair and I will both be in nola 19:58:55 <jeblair> let's cancel it? 19:59:02 <clarkb> I can run a short one to do testing updates 19:59:08 <clarkb> to keep hub_cap et al honest :) 19:59:12 <pleia2> and just so we don't get bored, the following weekend anteaya, RyanLane and I running this on Sunday the 22nd: http://codechix-openstack1-rss.eventbrite.com/ 19:59:37 <zaro> Not sure how i should proceed with gerrit WIP patch. tried RFC on the patch. been there for 2 weeks without any comments. 19:59:39 * pleia2 is going to need a nap after all this 19:59:40 <jeblair> clarkb: ok all yours if you want it. :) 19:59:49 <jeblair> time's up 19:59:55 <clarkb> jeblair: ok, I will try wrangling the three involved paries 19:59:57 <jeblair> thanks all! 20:00:00 <jeblair> #endmeeting