15:03:37 <scottda> #startmeeting cinder_testing 15:03:38 <openstack> Meeting started Wed Sep 28 15:03:37 2016 UTC and is due to finish in 60 minutes. The chair is scottda. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:03:39 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:03:39 <ntpttr> o/ 15:03:41 <openstack> The meeting name has been set to 'cinder_testing' 15:03:43 <_alastor_> \o 15:03:44 <xyang1> hi 15:03:47 <geguileo> dulek: Yeah, I had to make some changes to A/A for that as well :-( 15:03:54 <dulek> hi! 15:03:58 <geguileo> Hi! 15:04:01 <eharney> hi 15:04:04 <eharney> ! 15:04:34 <scottda> As usual, we've no agenda. 15:04:50 <scottda> So, feel free to speak up about any test efforts/issues/info 15:05:32 <scottda> It's be nice to figure out what we'd like to talk about at the Summit... 15:06:07 <scottda> #link https://etherpad.openstack.org/p/ocata-cinder-designsummit-planning 15:06:27 <scottda> I put a Subject for Test on the agenda. Feel free to put stuff there. 15:07:14 <dulek> If we're going to make Ocata a stabilization release (BTW - when we'll decide that? Was it already decided?), then this will be important working session. 15:07:57 <scottda> dulek: Agreed. That's why It'd be nice to start capturing ideas and details, prior to the scheduling of the sessions. 15:08:30 <scottda> dulek: I think it was decided when we re-elected smcginnis , who campaigned saying that's what he envisioned for O? 15:08:31 <smcginnis> dulek: Decided. ;) 15:08:46 <scottda> See? What Great Leader said^^ 15:08:58 <smcginnis> scottda: And yes, that will be important. I'm hoping to add some plans there. 15:09:09 * smcginnis is in another meeting as usual right now... 15:09:24 <dulek> Oh, is voting for smcginnis in TC elections a vote to make O a stabilization release for whole OpenStack? I'm convinced then. :D 15:09:33 <smcginnis> LOl 15:09:37 <scottda> geguileo: I think test efforts for AA/HA are important for getting it all merged into O. 15:09:45 <geguileo> scottda: I agree 15:10:02 <geguileo> scottda: I'm now banging my head against my desk to make replication work in A/A environments 15:10:17 <scottda> ouch 15:10:29 <geguileo> scottda: Because I somehow forgot about it :'-( 15:11:15 <geguileo> scottda: I'll try to get it fixed this week and update the post 15:11:23 <geguileo> scottda: But we have to decide how we want to test this 15:11:31 <geguileo> On a conceptual level I mean 15:11:35 <scottda> geguileo: We had great success merging a bunch of a/a patches at the midcycle. I think we should be prepared for another round of that at the summit. 15:11:59 <geguileo> scottda: That would definitely make me happy :-) 15:12:00 <dulek> scottda: +1! 15:12:20 <scottda> Yeah, testing A/A has proven to be tricky. Your manual steps are great, but we need efforts to do more of the automation, etc. 15:12:36 * scottda is guilty of not spending time on A/A tests. 15:12:45 <geguileo> scottda: Yeah, automation is key 15:13:03 <geguileo> scottda: But we have to decide to what level we want to test it 15:13:33 <scottda> geguileo: Well some things are probably out of scope, like the idea of error-injection. 15:13:59 <geguileo> scottda: Out of the scope of the initial tests or in general? 15:14:07 <scottda> That will probably have to come later, and start in someone's lab. 15:14:14 <scottda> Just outside initial tests. 15:14:36 <scottda> I think if we gate the merge of AA patches on those tests, we won't get it in O 15:14:51 <scottda> just my opinion. 15:15:45 <scottda> It seems within reach to create basic functional tests, running in a multi-node cinder devstack. 15:16:00 <scottda> I would guess start with the experimental jobs. 15:16:52 <geguileo> scottda: Sounds reasonable to me 15:17:20 <geguileo> scottda: So we should create test cases we'll cover initially 15:17:46 <geguileo> scottda: But I'm not so sure how we'll test that the A/A part is actually working 15:18:05 <scottda> yeah, that is tricky... 15:20:31 <scottda> Of course some of the testing is that normal operations don't get broken. We get that for free just having the A/A code running. 15:21:21 <dulek> Why don't we get same tactic like with rolling upgrades? It's experimental until we've set up CI testing it? 15:21:24 <scottda> I think the worrisome parts are around failure cases and cleanup. i.e. could we accidentally have a volume attached to 2 instances. 15:21:43 <openstackgerrit> David Sariel proposed openstack/cinder: Hacking checks for H105 https://review.openstack.org/377714 15:21:45 <scottda> dulek: Yes, that seems like the right path. 15:21:55 <openstackgerrit> David Sariel proposed openstack/cinder: Hacking checks for H105 https://review.openstack.org/377714 15:22:04 <dulek> The only problem would be that it can break something that's working already, but well… This means our Tempest tests set is too narrow. 15:22:05 <openstackgerrit> Bartek Żurawski proposed openstack/cinder: Add volume type name into volume notification https://review.openstack.org/378700 15:22:44 <scottda> dulek: Right. IF our tempest coverage were complete, we'd at least know that we hadn't broken anything. 15:24:02 <scottda> But what about the idea of what worries people have? Is there anything besides the data corruption case where a mistake in cleanup of 'attaching' or 'detaching' volumes results in an unwanted multi-attach? 15:24:12 <dulek> Yeah, but what I mean is that we're breaking stuff that's not tempest-tested all the time anyway. Getting A/A early in Ocata is the best course of action for us. 15:24:17 <scottda> I've had a hard time thinking of other bad scenarios. 15:24:30 <scottda> dulek: +1 to getting it in early 15:24:31 <openstackgerrit> David Sariel proposed openstack/cinder: Hacking checks for H105 https://review.openstack.org/377714 15:24:35 <smcginnis> dulek: +1 15:24:54 <scottda> I'm for planning a big Merge-Fest for this Friday :) 15:25:01 <erlon> dulek: +1 15:26:03 <scottda> Maybe we should bring this subject up at the main Cinder meeting? Seems like there's only a few of us here ATM. 15:26:19 <DuncanT> scottda: Definitely 15:26:46 <scottda> I'll just put it on the agenda now. You OK with that geguileo ? 15:27:08 <geguileo> scottda: Having a merge-fest this Friday? 15:27:17 <openstackgerrit> David Sariel proposed openstack/cinder: Updated from global requirements https://review.openstack.org/378705 15:27:55 <scottda> haha...no, I meant OK with discussing the state of A/A and getting it in soon and what's left and maybe the Merge Fest ; all at today's Cinder meeting. 15:28:07 <scottda> Basically this discussion we're having here and now. 15:28:38 <scottda> I don't want to wait for Barcelona to start this discussion with the wider community. 15:28:59 <geguileo> scottda: Yes, I think it's a great idea 15:29:21 <scottda> And, yes, I think if, as a community, we were committed to getting this in the O release, we should get it in soon. Like, before the Summit. 15:29:48 <geguileo> scottda: I agree, we want it asap so we can detect bugs, corner cases, etc. 15:29:56 <geguileo> Otherwise it's only going to get half in 15:29:58 <DuncanT> If we merge very early in the cycle, we've plenty of time to fix things up. Hopefully we'll have some initial testiing results for the summit.... those are far more interesting than reitterating plans 15:30:35 <scottda> OK, I'm putting it on the agenda now. 15:32:01 <dulek> Next topic? 15:32:07 <scottda> dulek: Sure 15:32:23 <dulek> https://goo.gl/Vmwifv - multinode grenade job is voting quite consistently for ~a month now. 15:33:02 <dulek> Big red bar is the moment when scheduling of typeless volumes was broken - this job creates typeless volumes actually. 15:33:28 <scottda> dulek: Looks good. Are you proposing making it voting? 15:34:10 <dulek> scottda: Yup, I'll send an email to openstack-dev soon unless someone has an objection. 15:34:42 <erlon> dulek: is it possible to add other jenkins jobs in the graph to compare? 15:34:43 <dulek> So - is anything wrong that I don't know about? :) 15:35:12 <dulek> erlon: Yeah, I need to sit down and get friendly with Graphite before posting on ML. 15:35:31 <erlon> dulek: nice 15:36:50 <scottda> dulek: There don't seem to be any objections. 15:36:54 <openstackgerrit> Eric Harney proposed openstack/cinder: extract_host: Handle empty host string input https://review.openstack.org/378719 15:37:05 <dulek> I'll just write the email tomorrow and get more broad feedback. Thanks! 15:37:19 <scottda> Next subject? or is there another subject? 15:39:30 <scottda> OK, talk again in 20 minutes 15:39:33 <scottda> #endmeeting