16:59:37 #startmeeting ironic 16:59:38 Meeting started Mon Sep 26 16:59:37 2016 UTC and is due to finish in 60 minutes. The chair is jroll. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:59:40 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 16:59:42 The meeting name has been set to 'ironic' 16:59:51 * jroll waits until it's actual time 16:59:58 o/ 16:59:58 3 seconds 17:00:01 o/ 17:00:02 o/ 17:00:03 o/ 17:00:05 o/ 17:00:10 o/ 17:00:14 o/ 17:00:17 o/ 17:00:17 agenda is here, as always 17:00:18 o/ 17:00:19 #link https://wiki.openstack.org/wiki/Meetings/Ironic#Agenda_for_next_meeting 17:00:27 o/ 17:00:28 * jroll jumps right in 17:00:30 o/ 17:00:34 o/ 17:00:34 #topic announcements 17:00:55 o/ 17:00:58 so, we released newton last week 17:01:03 \o/ 17:01:05 o/ 17:01:05 it's PACKED full of features 17:01:11 I think we did very well last cycle 17:01:12 Congrats to returning PLT jroll 17:01:17 so thanks to all of you who contributed :) 17:01:28 o/ 17:01:30 o/ 17:01:32 ++ good work everyone in this release 17:01:35 \o/ 17:01:38 yes, big pat on the back to all of us :D 17:01:42 ++ 17:01:43 o/ 17:01:56 yeah, you've successfully dodged the risk of having me as a ptl :D 17:01:57 we are in ocata cycle o/ 17:01:58 I'm glad we had two highly qualified candidates running for PTL :) 17:02:16 thanks for running dtantsur, and keeping jroll honest ;) 17:02:16 +1, competition is good 17:02:22 :) 17:02:24 dtantsur: thanks for running ;) 17:02:27 :) rather 17:02:36 ha ha 17:02:46 shift finger slipped :P 17:02:48 yeah, your ;) looked suspicious 17:02:51 heh 17:03:12 so, moving onto ocata, a couple of things 17:03:28 1) the grenade job is broken, that should be our top priority right now so we can do work 17:03:53 2) we need to be talking about summit sessions, the etherpad for suggestions is here 17:03:55 #link https://etherpad.openstack.org/p/ironic-ocata-summit 17:04:01 jroll: is 1) broken cuz of ironic? 17:04:08 o/ 17:04:19 rloo: that's unlikely 17:04:24 3) we should start looking at specs for this cycle 17:04:30 Any ideas what caused breakage? 17:04:52 rloo: unlikely, the instances created on newton code are no longer pinging after upgrading control plane to master 17:04:52 vdrok: ok thx. i was wondering how/what we could do about that, given that jroll said it should be our top priority 17:04:58 * jlvillal wonders if that is what he saw on Friday evening. He thought it was a transient issue. 17:05:00 help debug :) 17:05:06 I've looked at recent devstack and grenade commits and nothing caught my eye 17:05:09 jroll: gotcha 17:05:14 I've just skimmed trhough the logs (see whiteboard) we need more debugs, but tl;dr we can't ping the instance 17:05:15 obviously there's diminishing returns on how many people jump on that 17:05:29 causes could be many 17:05:37 yeah 17:05:49 nova grenade had a similar issue, which they fixed in grenade code 17:06:50 any other announcements or reminders? 17:07:11 o/ 17:07:11 jroll: as far as summit, when do we need to decide on the sessions? 17:07:24 small announcement - i'm working on ironic full time starting now-ish, in the OSIC group :) 17:07:35 yay mariojv! 17:07:36 so looking forward to working with everyone more upstream 17:07:59 rloo: I haven't seen a deadline yet, but I'm out the week leading up to it, so let's say by october 14 17:08:04 mariojv: \o/ 17:08:36 jroll: so you want to spend the Monday meeting Oct 10, discussing/deciding? 17:08:39 mariojv: Woot woot! :) 17:08:46 rloo: how about finalizing? :) 17:08:54 maybe start deciding in the oct 3 meeting 17:09:04 mariojv, ++ welcome :-) (tho you've been around already) 17:09:09 which is... next one. wow. 17:09:10 jroll: ok, that is fine with me. i'm going to put that down in the etherpad. 17:09:14 * jroll will send an email early this week 17:10:06 #topic subteam status reports 17:10:18 as always, these are on the whiteboard 17:10:20 #link https://etherpad.openstack.org/p/IronicWhiteBoard 17:10:23 probably not much to review here? 17:11:12 yeah, looks about right, I'll give folks a couple minutes to ask questions though 17:11:27 jroll: can we take keystone policy support out? 17:11:32 reminder, we'll shuffle these around after priorities are decided at the summit 17:11:34 rloo: yes! 17:11:42 rloo: we just merged that :P 17:11:50 I think active node can go too, as long as we remember to get the tempest test done 17:11:53 lucasagomes, you don't have to wait to remove -2 from https://review.openstack.org/#/c/366742/ 17:11:55 hi folks! sorry for being late - had trouble getting on the wifi 17:12:04 jroll, jlvillal: sweet. jlvillal -- do you know the patch that got merged? maybe add to the subteam report. 17:12:07 lucasagomes, it won't pass anyway without ironic-lib, assuming you have tests 17:12:23 rloo: I was joking about you taking it out. 17:12:29 jlvillal: whoa. someone already deleted it :-( 17:12:33 rloo: I just removed it from ... yeah 17:12:34 dtantsur, right, well that's what I mean by "blocked" 17:12:39 * jlvillal was not involved 17:12:43 rloo: not okay? that's been done for a couple weeks 17:12:51 dtantsur, the -2 is just a flag to people not waste time reviewing it for now 17:12:56 jroll: s'okay, not worth adding back in again :) 17:12:58 lucasagomes, why not review it? 17:13:05 rloo: heh, okay :) 17:13:10 lucasagomes, also W-1 is for letting people know it's not worth reviewing 17:13:31 dtantsur, right yeah I can lift the -2 and leave WIP only 17:13:33 * lucasagomes do that 17:13:36 thanks! 17:13:49 yeah, I thought -2 meant "this entire approach is wrong, please abandon this patch" 17:14:00 * lucasagomes done 17:14:13 mgould: well, we also use them for process things like "don't merge until ocata" 17:14:14 mgould, we also have technical -2s, but this is not one any more 17:14:21 or spec approval or whatever 17:14:21 ah, OK 17:14:33 jroll: it isn't here, and maybe it is addressed in qa meeting? do we have to do anything about 3rd party CI? 17:14:54 krtaylor: ^^^ 17:14:55 mgould: also it's kind of hard to put w-1 on every patch set 17:15:06 rloo: as a subteam report thing? 17:15:21 rloo: there's some documentation we need to write up for driver devs / reviewers 17:15:32 write up / figure out 17:15:32 jroll: oh, i dunno. i was just wondering now that newton is 'out', was there something we needed to do. kick stuff out or what? 17:16:23 rloo: 3 months from deprecation we can drop things, we also need to talk about process for e.g. requiring driver CI to pass when changing that driver, etc 17:16:50 jroll: it does make me wonder if we should have a status thing or pointer to the other ironic-related meetings. 17:17:12 rloo: which sort of status thing? 17:17:32 jroll: dunno, anything of importance from those meetings. eg, ironic-api one, deva's specs. 17:17:59 jroll: just a thought that isn't fully formed yet. 17:18:02 ah, right, maybe 17:18:07 I'd like to rethink subteams as a whole 17:18:25 jroll: oh, that sounds interesting! 17:18:49 :) 17:18:58 * jroll still needs to give it some thought 17:19:01 anything else on this topic? 17:19:26 * krtaylor reads scrollback 17:19:37 not sure, was there a question for me? 17:20:18 krtaylor: Was wondering if there was any thing to put in about 3rd Party CI in the subteam section 17:20:35 krtaylor: we were talking about "what needs to happen for 3rd party CI in ocata" 17:21:18 I think we've got it for now, but feel free to chime in 17:21:23 #topic open discussion 17:21:25 REMOVE ALL THE DRIVERS \o/ 17:21:27 ok, we did bring up rethinking deprecation policy, but thats all I can think of, other than policing the current policy 17:21:57 dtantsur wants to remove the drivers, so he won't need to do the driver composition reform stuff :) 17:22:01 dtantsur: not _all_ the drivers... i hope ;) 17:22:10 rloo, you knew my plan, how?? 17:22:19 hehheh 17:22:21 dtantsur: lucky guess 17:22:21 TheJulia, let's leave "fake", agreed 17:22:29 * dtantsur shuts up 17:22:33 dtantsur: and noop ;) 17:23:12 should we start looking at s/_ssh/_ipmitool/ for the tests in gate ? 17:23:17 lucasagomes+++ 17:23:19 lucasagomes: ++ 17:23:20 +1 17:23:36 we could then drop at least 2 jobs on check which is being used for ipmitool now 17:23:48 sounds good 17:23:49 cool, I will rpopose a patch 17:23:51 propose* 17:23:56 lucasagomes: yes please :) 17:24:03 also about gate: wdyt about merging some jobs by launching >1 instances per job? e.g. testing both pxe_<> and agent_<> things in one run? 17:24:14 dtantsur: I would like that 17:24:18 dtantsur, ++ 17:24:36 okie, I'll look into what's required for that (I guess some devstack hackery) 17:24:37 do we have any measurements of which code is covered by integration tests? 17:24:38 either both drivers or things like pxe_ partition image and whole disk image in the same run 17:24:46 with 2 VMs each one deploying one image type 17:25:05 dtantsur: probably 17:25:11 mgould: don't believe so 17:25:25 how many "bare metals" can we run in CI with tinyipa now? 17:25:31 we already have jobs that create / launch >1 instance, right? and we default to 3 Nodes in each job anyway, so I think we could just enable >1 driver and assign driver A to node 1, driver B to node 2, ... and then consolidate several jobs that way 17:25:44 devananda, this is exactly what I want to do 17:25:49 +1 17:25:51 I like :) 17:25:53 +1 17:25:57 Mirants QA folks once wrote a wrapper under which to start services and measure test coverege 17:26:10 dtantsur: I did some math recently, I think we can do 6-7 nodes with tinyipa 17:26:10 pas-ha, cool! is it public? 17:26:11 I can ask around if for iy 17:26:20 jroll, wow, that's a lot! 17:26:23 dtantsur: remember multinode is also coming, so we can scale out too :) 17:26:27 pas-ha: awesome 17:26:36 most probably yes, but dug in some obscure repo 17:26:43 * jroll would like to try 10x devstack nodes running 5 vms each, just for fun 17:26:46 that would really ehlp to decide where to focus test effort, IMHO 17:26:48 jroll, yeah 17:26:55 *help 17:27:02 pas-ha, that's amazing. milan may be interested on it 17:27:19 * milan reads scroll back 17:27:30 jroll, I'd only leave different jobs for things that don't easily co-exist like PXE vs iPXE, BIOS vs UEFI (maybe) 17:27:40 but this is a goal for the future :) 17:27:44 BIOS and UEFI *should* coexist 17:27:46 dtantsur: indeed, one thing at a time :) 17:27:58 yep 17:28:08 mgould, yeah, my bad. but PXE and iPXE don't 17:28:16 * mgould nods 17:28:41 mgould, however, BIOS vs UEFI requires different configuration of VMs... chances to test some capabilities discovery? ;) 17:28:50 heh, why not :-) 17:29:03 mgould, ++ 17:29:08 dtantsur: we could make pxe and ipxe co-exist 17:29:17 yeah it's hard to setup as dtantsur said 17:29:29 make an iPXE boot a separate interface, and use dynamic iPXE 17:29:46 and configure iPXE-ish driver per node 17:29:47 pas-ha, even without dynamic iPXE, but I don't imagine a use case 17:30:04 pas-ha, cool, we used to have a tool for remote/multi-node coverage stats: https://github.com/RedHatQE/python-moncov (though it collects to redis not mongo ;) 17:30:20 dtantsur: a zoo of hardware 17:30:36 pas-ha, probably.. worth an RFE? 17:30:49 * milan didn't run it in a while, most likely the entropy ate it already 17:31:08 milan, same for our scripts 17:31:13 dtantsur: will do 17:31:24 thanks! 17:31:27 fyi a couple of related patches - https://review.openstack.org/369021 and a job already there - https://github.com/openstack-infra/project-config/blob/master/jenkins/jobs/ironic.yaml#L1068 17:31:46 oh cool! 17:31:54 ^ that's multinode 17:32:19 or rather multitest 17:33:14 yeah, we can do the tests with pxe, then reconfigure and do with ipxe, kind of similar thing 17:34:51 anything else we should chat about here, or shall we head back to channel? 17:35:38 alright, thanks everyone 17:35:41 thanks 17:35:44 #endmeeting