22:00:17 #startmeeting qa 22:00:18 Meeting started Thu Oct 16 22:00:17 2014 UTC and is due to finish in 60 minutes. The chair is mtreinish. Information about MeetBot at http://wiki.debian.org/MeetBot. 22:00:19 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 22:00:22 The meeting name has been set to 'qa' 22:00:36 hi, who's here today? 22:00:42 o/ 22:00:46 o/ 22:00:59 o/ 22:01:12 today's meeting is dedicate to discussing proposed session topics 22:01:27 #topic Summit Topic Discussion 22:01:29 the brainstorming etherpad is located here: 22:01:32 #link https://etherpad.openstack.org/p/kilo-qa-summit-topics 22:02:25 hi 22:02:45 so just a summary, we have 7 40min session slots 22:03:09 and then a 1 day shared room with infra and rel-mgt for a contributor meetup 22:03:47 I have an etherpad link for topics in that room 22:03:52 one sec let me pull it up 22:04:23 #link https://etherpad.openstack.org/p/kilo-infrastructure-summit-topics 22:05:06 the rule of thumb is that we can use that meetup day for any topics which we'd like to discuss or collaborate on in person, but we don't think warrants a dedicated 40min block 22:05:39 also we should really be using the session slots for topics where we think it'll be good to get people from outside to participate in 22:05:47 o/ 22:05:49 because that probably won't be the case with the meetup day 22:06:31 so before we dive into the topics on the etherpad are there any questions or comments about the format for summit? 22:06:54 mtreinish: for the day-long, are we sharing topics with infra or just the room? 22:07:36 dkranz: well we're sharing the room with infra, we can also share topics but we don't have to 22:07:43 k 22:08:16 I imagine if there are any topics with overlap and it doesn't really require outside discussion using the meetup would be better 22:08:56 for everyone who was at the midcycle in Darmstadt the meetup day will probably be similar to that 22:08:57 I agree. Some topics have overlapping part. 22:10:21 ok, if there aren't any other questions about the summit format then let's start going through the etherpad 22:10:37 how about we start from the bottom 22:10:47 so, Gap analysis for using Tempest in production (not DevStack) environments 22:11:25 masayukig: so I think this is a great topic, it's similar to what I was trying to address with the ux one 22:11:39 but I feel it's a distinct issue which we should dedicate time to 22:11:48 +1 22:11:59 did you have a feeling for the format you wanted? 22:12:28 for example did you think it would be better as a dedicated session or a more informal thing as part of the meetup? 22:12:47 perhaps a session would be good here 22:12:52 because I'd be fine with either, although I think the need for external comments probably means this makes more sense as a session 22:12:55 andreaf: +1 22:13:04 because it's really a topic where we could get input from people 22:13:12 mtreinish: Unless we go into the room with some concrete ideas to discuss, a session may be too short 22:13:26 I have one question. What is the difference from above topics like Web-GUI or Tempest UX? 22:13:31 mtreinish, andreaf: Yeah, we should discuss that in ML too. 22:13:37 andreaf: but yes, we could gather data, and then discuss on Friday 22:13:44 dkranz: yeah that is a concern 22:14:19 dkranz: sounds like a plan :) 22:14:47 andreaf: I think there is potentially a lot of overlap with the other make-tempest-friendly-for-users items in the pad 22:15:30 dkranz: agree. 22:15:56 dmorita: I think Web-GUI and Tempest UX are one of this topic. 22:16:08 masayukig: ok 22:16:27 masayukig: so you think we should combine those 3 topics and make that a single session? 22:16:41 Good! and, you can be the session leader 22:16:58 you -> masayukig 22:17:06 dkranz, masayukig: yes there's a lot of overal, but I think the last topic on the pad is really specific issues like devstack specific defaults, cannot run without admin credentials, rate limiting and so forth 22:17:06 i think this can be a good session which can provide much input for UX, GUI topic 22:17:22 s/overal/overlap 22:17:50 mtreinish: My concern is the time. Is 40 minutes enough for them..? 22:17:53 I think that would be good but we should set expectations that unlike many sessions, this one is not about coming out with a decision 22:18:03 And then continue on Friday 22:18:12 With the intent of coming out of that with a plan 22:18:45 So the session would be more about gathering input 22:19:01 masayukig: hmm, dunno. I think dkranz is right we can always spill into friday if we need extra time 22:19:13 dkranz: I like that idea, we just need to make sure we advertise the session well 22:19:23 and probably make it earlier in the schedule 22:19:29 ++ 22:19:38 dkranz: +1. Thats what i was thinking, to have this session for more on gathering input 22:20:05 mtreinish: It would actually be good if all projects advertised particular sessions where they were looking for outside input 22:20:15 maybe send to ml 22:20:26 mtreinish: ok. 22:21:40 ok, well that sounds like a plan then 22:21:51 does anyone have any other comments on this topic? 22:23:00 ok then next up is: 22:23:02 Localization test: how to ensure qualities of localized OpenStack with translations 22:23:20 I was talking to Daisy about this earlier today 22:24:00 the basic issue is that occasionally (very infrequently) the tests (both tempest and unit tests) break with a specific locale enabled 22:24:15 it exposes real bugs in openstack, but we don't have a mechanism to test it 22:24:32 all the tests have been run manually 22:24:43 I think this is a good candidate for a periodic job 22:24:57 but I'm not sure we need a full session for discussing it 22:25:25 mtreinish: well we run 26 (?) checks on each tempest change, we could inject a random localization in each of them 22:25:50 mtreinish: "random" meaning the same for each job so that it's still ok to troubleshoot 22:25:52 on one hand it would be a good time for interaction between qa, infra, and i8ln 22:26:43 andreaf: well aren't the log messages translated too? I'm not comfortable debugging logs from a random failure in a language I don't know :) 22:26:43 mtreinish: it doesn't feel like something worth a full session though 22:27:20 mtreinish: ah right, not tempest logs but the rest is 22:27:42 We could spend a little time on Friday 22:28:20 yeah, I'm thinking this would be a good meetup day topic. Or maybe a split session (20min on this 20min on something else) 22:29:34 well let's say it's a meetup topic for now, and depending on how the schedule looks at the end we maybe allocate a half slot for it 22:29:37 any objections? 22:29:58 it sounds good 22:30:11 nope :) 22:30:27 ok cool, let's move onto the next one then 22:30:36 Auth interface design for testing various clusters (plus tempest-lib migration) 22:30:43 It is my suggestion :) 22:30:59 dmorita: yep, I'm guess it's related to your spec I -2'd :) 22:31:26 I -2'd ? 22:31:45 dmorita: the spec to add non-keystone config to tempest that you wrote 22:31:52 I have a -2 sitting on it IIRC 22:32:11 I see. Partially related. 22:32:45 But main motivation is to enable Tempest "really" run tests to any OpenStack cluster. 22:33:08 dmorita, mtreinish: the work I've done with the multi-auth bp should help switching auth provider 22:33:16 Sub-motivation is migrating auth part to tempest-lib . 22:33:23 andreaf: yeah that's what I was thinking 22:33:46 dmorita: sure I think if we make the auth portion of tempest easy to consume and replace as part of tempest-lib that would make this easy 22:33:47 dmorita: so do you mean clusters not running keystone for identity? 22:34:09 I still don't think we should support the swift auth mechanism in tempest, because it's swift specific 22:34:24 but if we have the lib interface stable you should be able to make one easily 22:34:45 and just plug it in for your use case 22:34:57 mtreinish: yes. But, now it is difficult to write a specific driver for Tempest's auth by their own. 22:35:10 ++ I think until OpenStack should decide to de-integrate keystone this is outside of our core mission 22:35:31 dmorita: anyway this feels like a subtopic on tempest-lib, do you agree? 22:35:48 I want to show the analysis of current implementation, and to-be interface design. 22:35:59 IMO, auth architecture is a little complicated so it seems necessary to describe a current implementation in detail. Therefore, a dedicated session is preferred. 22:36:24 dmorita: A written description to the ml would make that much easier 22:36:43 dmorita: I don't think we use these sessions to describe things 22:37:46 I think we need many eyes to discuss more other use cases and interface design. 22:37:46 yes I agree, if we prepare any explanatory material ahead of the sessions that will make the face-to-face time more productive 22:38:01 mtreinish: we are on topic 3 out of 11 with 20 min left 22:38:23 mtreinish: that sounds like something I should work on 22:38:38 dkranz: oh great, well lets cross of some easy ones then 22:38:57 tempest-lib is going to be a session I don't think anyone disagrees with that, right? 22:39:03 +1 22:39:05 +1 22:39:09 +1 22:39:11 ++ 22:39:11 I'll take the session leader on that, because I've been driving that effort so far 22:39:16 +1 22:39:23 then the tempest scope one also will be getting a session 22:39:39 I don't think it needs to be a double session necessarily 22:39:57 I plan to start a ML to prempt the session tomorrow or early next week 22:40:04 mtreinish: with Friday available I don't think we need double sessions for anything 22:40:37 dkranz: fair point 22:40:46 I was just echoing your earlier concern :) 22:40:59 mtreinish: right, but I did not know the format then 22:41:49 ok after that we have 4 more topics 22:41:56 QA and CI after merge 22:42:06 ++ 22:42:08 andreaf: what did you have in mind for talking about in this session? 22:42:32 what tests could be moved to post-merge, how we track and maintain them 22:42:58 andreaf: that is related to the in-project functional testing as it will have the same issues 22:43:07 andreaf: ok, so like a dashboard view on test results 22:43:13 and how do we let deployers know which versions to pick to have something passing certain post merge tests 22:43:51 andreaf: ok I'm fine with giving that a topic, although it does feel like a split topic with infra 22:44:02 because we don't have a good story around post tests at this point 22:44:36 mtreinish: sure there are two aspects, one is how to make people care about post-merge tests 22:44:44 does anyone have an objection to that as a topic? 22:44:56 mtreinish: the other one is really what can we move out of the main gate 22:45:37 andreaf: well I think that's more a question of configs not necessarily individual tests 22:45:47 mtreinish: which is like saying something is core openstack always working, other is "class B" most of the time working, perhaps working on releases but not all the time 22:46:23 mtreinish: yes indeed 22:46:24 andreaf: yeah that part feels more like a cross-project topic :) 22:46:42 but anyway I think that's a good topic to have 22:46:45 I agree > cross-project 22:47:22 mtreinish: agree 22:47:44 what I think will happen is we can work out the technical aspects there and avoid policy decisions about what is tested everywhere vs post merge 22:47:51 ok I added that to the list of sessions 22:47:53 let's move on 22:48:05 Response objects of Tempest clients 22:48:14 I agree that we can just handle that in the spec review 22:48:26 +1 22:48:29 any objections to that? 22:48:31 +1 22:48:34 +1 22:48:35 yeah, agree 22:49:09 ok then we have 22:49:11 Implications of moving functional tests to projects 22:49:24 dkranz: you proposed this as a cross-project session right? 22:49:38 mtreinish: yes, but I don't know how/when they are chosen 22:49:42 do you want to hold off on having this in qa until we see if that gets picked up? 22:49:51 dkranz: it should be within the next week 22:49:55 mtreinish: sure 22:50:16 so I'll put that as a maybe, I do think it's related to the tempest scope topic but that's fine 22:50:36 then next up is 22:50:38 Tempest/Rally Integration 22:51:07 which is something I'm not sure warrants a separate topic, because we did this in Atl, and nothing was produced from it 22:51:19 mtreinish: so we have three other items that are about user-experience,ui,database,etc. 22:51:20 despite having a concrete list of work items from the session 22:51:44 mtreinish: and we also did nothing with the other proposals either 22:51:54 mtreinish: perhaps it does not need a session 22:52:09 mtreinish: but it is really about the best way to move forward on the UX front 22:52:29 dkranz: well we had masayukig's session in atl about the tempest server plan which had good discussion and a spec proposal 22:52:35 which I created subunit2sql from 22:52:53 the rest of the spec was blocked because of overlap concerns with jenkins 22:52:57 mtreinish: my only "point" is that a lot of people have started using rally 22:53:17 mtreinish: and we are either going to compete with that, or join forces, because I believe the end goal is the same 22:53:50 mtreinish: just something to think about 22:53:54 dkranz: I think there is definite overlap which is something I've said many times, but the scope of projects is very different. 22:54:04 I think baking it into a ux discussion is a good idea 22:54:30 I really would like to see collab here, it's just something which hasn't happened so far 22:54:31 mtreinish: I think the current feature focus is different but I don't see any UX goals we have that rally does not also have 22:55:36 mtreinish: anyway, I am not asserting an answer to this, just something to think about as we move forward on UX 22:55:59 dkranz: I agree, but I just don't think it warrants another slot 22:56:14 mtreinish: ok 22:56:27 does anyone else have an opinion on this topic? 22:56:46 because my opinion isn't the only one here :) 22:57:15 because that also leaves us with 2 (maybe 3) unallocated slots 22:57:23 How is auth interface topic? It must be also helpful to just consider migration of auth module to tempest-lib. 22:57:28 Actually, I already started surveying the auth implementation and found some points to be improved. 22:57:33 Sorry, I am persistent. 22:58:23 dmorita: if you have a concrete proposal for a better auth interface that we can use in tempest-lib and think it'll fill up enough time for a session I guess that's fine 22:58:40 but we really don't want to spend session time going over what's there now 22:58:52 and reviewing the current interface 22:59:33 Of course, I can prepare material before the session to explain current architecture quickly. 22:59:38 dmorita: feel free to share any proposal for improvement - the multi-auth bp is not finished anyways 23:00:24 does anyone object to having a session on the future of the tempest auth interface? 23:00:34 also we're at time, but I don't think there is a meeting after us 23:00:38 so we can go over a bit :) 23:01:14 andreaf: your effort is good. 23:01:28 mtreinish: oh.. :) 23:01:34 is anyone in favor of a session like that? 23:01:38 But, i think it is a little different from migration to tempest-lib. 23:02:45 dmorita: well considering we have at least 1 free slot I guess we'll make this a session 23:03:13 ping me if you've got some concerns on how we should be framing the discussion 23:03:20 andreaf: is that ok? 23:03:44 sure, ok 23:03:52 ok 23:04:13 ok well I think that's it for the topics unless I missed one 23:04:23 Got to run now 23:04:29 dkranz: ok 23:04:49 I also should go, it's past midnight 23:05:04 we have 1 unused slot, if there are no other ideas I can fill that with an administrative topic, or talk to ttx and give that back to the pool 23:05:13 or we can use it as an extension of the meetup time 23:05:25 andreaf: ok yeah I was surprised you joined :) 23:05:26 Ok. Thanks! 23:06:10 may be we can extend Gap analysis in 2 sessions which may need more time 23:06:16 just a option :) 23:06:27 oh, I know dtroyer said he didn't want a devstack/grenade topic, but we could use that for the 1 slot :) 23:06:35 gmann: yeah that was the other thing I was conisdering too 23:06:54 anyway we're ~5 min over time so let's end the meeting and we can pick this up in -qa 23:06:55 gmann: that's sounds good. 23:07:07 thanks everyone 23:07:08 mtreinish: ok 23:07:09 #endmeeting