20:00:22 <asalkeld> #startmeeting heat 20:00:26 <openstack> Meeting started Wed Dec 3 20:00:22 2014 UTC and is due to finish in 60 minutes. The chair is asalkeld. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:00:27 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:00:30 <openstack> The meeting name has been set to 'heat' 20:00:47 <asalkeld> #topic rollcall 20:00:48 <zaneb> o/ 20:00:51 <skraynev_> o/ 20:00:52 <shardy> o/ 20:00:52 <stevebaker> \o 20:00:53 <pas-ha> o/ 20:00:56 <tspatzier> hi 20:00:58 <ryansb> hi folks 20:01:02 <inc0> o/ 20:01:06 <jpeeler> hey 20:02:00 <asalkeld> whilst we are waiting for others, lets think of some topics 20:02:25 <inc0> I have one:) Status of zero downtime upgrades 20:02:29 <asalkeld> #topic Adding items to the agenda 20:02:37 <asalkeld> ok 20:02:37 <shardy> asalkeld: I have one, Grenade testing 20:02:40 <stevebaker> kilo-1 status 20:02:48 <shardy> looking for a volunteer to push that forward 20:03:37 <ryansb> also kilo-1 exceptions 20:04:00 <shardy> ryansb: exceptions? 20:04:16 <asalkeld> i thought that was for ff 20:04:25 <skraynev_> shardy: you mean grenade job or what? 20:04:26 <stevebaker> ryansb: I don't think we do that for milestones, its just a snapshot 20:04:55 <asalkeld> what about a quick summary of convergence 20:04:58 <shardy> skraynev_: Yeah, I started a patch adding Heat to grenade testing, EmilienM then did some work on it, but it's still not ready/merged 20:05:21 <ryansb> err, I got wires crossed. Was thinking of 2014.2.1 20:05:22 <shardy> need someone to take ownership of it, I never have time to look at it lately 20:05:23 <pas-ha> https://review.openstack.org/#/c/86978/ 20:05:26 <ryansb> scratch it 20:05:30 <asalkeld> urm.... 20:05:31 <skraynev_> shardy: I may take a look ;) 20:05:36 <asalkeld> #topic Review action items from last meeting 20:05:39 <pas-ha> skraynev, ^^^^^ 20:05:45 <shardy> ryansb: Ah, yeah I think we landed all we needed for the stable freeze 20:05:53 <asalkeld> #link http://eavesdrop.openstack.org/meetings/heat/2014/heat.2014-11-26-12.00.html 20:06:20 <inc0> I think last meeting ended early due to thanksgiving...is there anything to review? 20:06:22 <asalkeld> did this happen "Convergence folks to organise call/hangout to define direction re various PoC efforts" 20:06:32 <shardy> skraynev_: thanks, lets action you when we get to that item :) 20:06:48 <zaneb> asalkeld: it's organised for tomorrow 20:06:53 <inc0> it did -> tomorrow at 9am east coast time 20:06:54 <skraynev_> shardy: ok 20:06:55 <asalkeld> cool 20:06:58 <skraynev_> pas-ha: thx 20:07:10 <asalkeld> ok, moving on 20:07:14 <asalkeld> #topic Status of zero downtime upgrades 20:07:34 <asalkeld> so this is waiting on the oslo versioned objects 20:07:36 <inc0> so, do we have issues listed out? blueprints to implement/draft? bugs to fix? 20:07:49 <asalkeld> inc0, we have a bp 20:08:15 <asalkeld> #link https://review.openstack.org/#/c/132157/ 20:08:26 <pas-ha> and seems like nova versioned objects falls into that bucket too 20:08:47 <inc0> is that only issue we have? if that lands we'll be able to do upgrades? 20:08:51 <stevebaker> asalkeld: can't we have zero downtime upgrades now as long as db schema changes are backwards compatible? 20:09:08 <asalkeld> stevebaker, yeah 20:09:16 <inc0> is it always? 20:09:23 <asalkeld> tho' i doubt that is normally the case 20:09:47 <asalkeld> so can a heat-engine talk to an old db 20:10:33 <asalkeld> inc0, i could start with just coping the versioned objects code in tree 20:10:42 <asalkeld> there in not much code 20:10:55 <asalkeld> and then swith over when we have it in oslo 20:11:05 <asalkeld> so it's not holding us back 20:11:19 <stevebaker> thats a bit risky 20:11:27 <asalkeld> stevebaker, why? 20:11:43 <asalkeld> i don't believe it's going to change 20:11:46 <stevebaker> the api might change as it moves to oslo 20:11:50 <skraynev_> asalkeld: can we wjust wait this functional from oslo? 20:12:10 <skraynev_> s/wjust/just 20:12:19 <asalkeld> we can wait, but i am not sure if it would make it in kilo 20:12:21 <shardy> stevebaker: Sometimes the API changes when it moves out of incubator anyway, so I don't see that as a huge barrier 20:12:32 <asalkeld> which would be disapointing to me 20:13:00 <skraynev_> asalkeld: got it. Agree, will be nice to have it in kilo 20:13:22 <asalkeld> ok, moving on 20:13:27 <shardy> asalkeld: what's holding it up landing in oslo? 20:13:27 <asalkeld> #topic Grenade testing 20:13:35 <stevebaker> ok, I wouldn't block it in the short term 20:13:46 <asalkeld> shardy, spec approval + code landing 20:13:49 <shardy> Ok, so here's the patch I started a while back: 20:13:58 <shardy> https://review.openstack.org/#/c/86978/ 20:14:03 <shardy> #link https://review.openstack.org/#/c/86978/ 20:14:19 <stevebaker> shardy: vaguely related, do you have a working local instack? 20:14:42 <shardy> EmilienM: helped get it working, now it's broken again, and I never seem to have time to look at it 20:15:01 <shardy> skraynev_: if you have time to help that would be great 20:15:04 <EmilienM> shardy: I should have a look if you want 20:15:14 <asalkeld> shardy, is the setup tricky? 20:15:14 <EmilienM> s/should/could/ 20:15:30 <shardy> we need to land initial grenade support, then define javelin stuff so we can actually test persisting stacks over upgrade 20:15:51 <shardy> Oh hey EmilienM, yeah if you have time to revisit it that would be great too 20:16:01 <asalkeld> shardy, does that test a rolling upgrade? 20:16:09 <skraynev_> shardy: sure, I'd interesting in it. 20:16:17 <EmilienM> shardy: I can manage to have time. I don't want to abandon something I helped a bit 20:16:27 <zaneb> asalkeld: aiui grenade tests a stop-start upgrade 20:16:33 <shardy> EmilienM: could you work with skraynev_ and somehow divide the work of defining what's tested and implementing it? 20:16:35 <stevebaker> asalkeld: looks like it kills heat, db upgrades, then starts heat again 20:16:45 <EmilienM> shardy: sure. 20:16:48 <shardy> EmilienM: Ok, great, thanks 20:16:50 <asalkeld> #action skraynev take over https://review.openstack.org/#/c/86978/ 20:16:58 <asalkeld> ok 20:17:02 <stevebaker> thanks skraynev_ 20:17:07 <shardy> stevebaker: I don't have instack locally atm 20:17:30 <stevebaker> k 20:17:41 <asalkeld> ok, anything more on this? 20:17:51 <shardy> asalkeld: Last time I tried setup was very tricky, but that may have improved 20:18:03 <shardy> asalkeld: not from me, thanks 20:18:07 <asalkeld> ok 20:18:09 <asalkeld> #topic kilo-1 status 20:18:16 <skraynev_> EmilienM: I suggest connect via email ;) because near of midnight for me. So I start looking tomorrow. 20:18:27 <asalkeld> #link https://launchpad.net/heat/+milestone/kilo-1 20:18:42 <EmilienM> skraynev_: emilien@redhat.com 20:18:59 <asalkeld> there are a couple of bp i am not sure of the status there 20:19:28 <skraynev_> EmilienM: k 20:19:36 <asalkeld> any one know multi region status? 20:19:54 <stevebaker> shouldn't that be assigned to Qiming? 20:19:55 <asalkeld> and " Templates outputs need to be lazy loaded" 20:20:07 <jpeeler> working on that ^ 20:20:08 <tspatzier> asalkeld: I think Qiming is working in the multi region one 20:20:21 <asalkeld> ok, i'll update the bp 20:20:39 <tspatzier> asalkeld: IIRC, stevebaker had some good comments so Qiming is working on an update 20:21:13 <asalkeld> ok as long as it is moving forward 20:21:35 <asalkeld> i have taken shardy decouple stack bp 20:21:56 <asalkeld> we need to focus on reviewing bugs/bps on that list 20:22:01 <asalkeld> to make sure they land 20:22:22 <shardy> asalkeld: I rebased decouple-nested, and fixed the first test failure today, help with the remaining tests would be great, thanks! 20:22:36 <asalkeld> i might kick this back to k2:https://bugs.launchpad.net/bugs/1319813 20:22:39 <uvirtbot> Launchpad bug 1319813 in heat "no event recorded for INIT_* resources" [Medium,Triaged] 20:22:43 <asalkeld> thx, shardy 20:23:01 <asalkeld> zaneb, ^ 20:23:07 <asalkeld> you set it to k1 20:23:11 <zaneb> that isn't even a bug 20:23:33 <zaneb> but it will probably happen as part of convergence anyway 20:23:36 <asalkeld> to 'fix' it looks like a non-trivial amount of work to me 20:23:42 <zaneb> not that it will change anything 20:24:06 <asalkeld> moved it to k2 20:24:34 <asalkeld> anyone know if jpeeler is working on the lazy loading code? 20:24:50 <jpeeler> asalkeld: yes i am 20:25:03 <asalkeld> ok, you think it can land in k1? 20:25:08 <jpeeler> i do 20:25:17 <asalkeld> cool 20:25:19 * zaneb should check the review again 20:25:31 <jpeeler> zaneb: nothing yet, will let you know 20:25:38 <zaneb> oh good :) 20:25:46 <asalkeld> well i think we are in good shape then 20:25:52 <zaneb> I hate when I'm holding everything up 20:26:11 <asalkeld> are you? 20:26:24 <zaneb> sometimes 20:26:28 <asalkeld> well half time bell 20:26:34 <asalkeld> and we are out of topics 20:26:58 <shardy> https://review.openstack.org/#/c/138800/ 20:27:14 <shardy> heads-up we need to land that (or a sample conf sync) to unblock our gate 20:27:27 <shardy> oslo.messaging released this afternoon 20:27:54 <stevebaker> that needs devstack changes first, devstack starts with heat.conf.sample 20:28:19 <asalkeld> o - i quite like having a sample in tree to point people to 20:28:20 <shardy> stevebaker: Hmm, OK, I'll propose a sync and mark that WIP while we work that out 20:28:33 <shardy> asalkeld: It's constantly breaking us 20:28:39 <zaneb> other projects are ditching it though I think 20:28:40 <stevebaker> I thought I made that comment in one of the other delete heat.conf.sample patches 20:28:46 <asalkeld> not *that* often 20:29:10 * shardy shrugs 20:29:17 <asalkeld> shrug, seem we dont' have much of a choice 20:29:20 <pas-ha> shardy - looks like some change must be made to devstack first 20:29:23 <shardy> well I find it annoying anyway 20:29:30 <shardy> pas-ha: Yeah 20:29:37 <pas-ha> it tries to stat the heat.conf.sample 20:29:40 <shardy> stevebaker: evidently I missed those, apologies 20:30:00 <stevebaker> shardy: there is a tox env to generate a sample, devstack can call that instead 20:30:01 <asalkeld> zaneb, you and inc0 want to fill us in on the convergence status 20:30:31 <inc0> well, we've spoke with HP guys today, and agreed to meet up tomorrow 20:30:32 <shardy> stevebaker: Yeah, I'm just thinking of the immediate step to un-break things 20:30:52 * shardy waited 6 weeks to land his last devstack patch 20:30:54 <asalkeld> #topic open discussion 20:31:03 <stevebaker> shardy: maybe just this would be fine for now https://review.openstack.org/#/c/138800/2/tox.ini 20:31:23 <inc0> from my point of view: I hope I've answered questions from mailing list...now I think we just need to decide on approach 20:31:32 <zaneb> I think we are close to settling on an approach 20:31:36 <shardy> stevebaker: Ok, I'll break it into two patches, thanks 20:31:56 <zaneb> just figuring out the last differences in how to store the graph efficiently 20:32:07 <asalkeld> ok 20:32:16 <inc0> zaneb, so you've already dropped out my approach? 20:32:19 <inc0> :/ 20:32:49 <asalkeld> inc0, have you worked on it more? 20:32:53 <zaneb> inc0: I... haven't changed my mind since my email last week 20:33:06 <asalkeld> does it fill the gaps that zaneb pointed out 20:33:08 <zaneb> I haven't seen any new patches as of yesterday 20:33:34 <inc0> I didn't really have time for coding unfortunately, but I hope I clarified that on email? 20:33:56 <zaneb> and I don't really want to wait another month to just come up with something equivalent to what we already have 20:34:15 <inc0> zaneb, it will never be equivalent because well..its different 20:34:21 <zaneb> what would change my mind at this point is a demonstration that phase 2 won't work with the other approaches 20:34:31 <asalkeld> inc0, i think the problem we have with that approach is that there are some nagging questions that remain unanswered 20:34:47 <inc0> asalkeld, ask then;) 20:35:04 <asalkeld> the rescheduling 20:35:07 <inc0> zaneb, have you guys solved things like concurrent update? 20:35:19 <inc0> asalkeld, what do you mean by rescheduling? 20:35:35 <zaneb> inc0: yeah. that is pretty much the entire point. 20:35:36 <asalkeld> when to re-call converge() 20:35:52 <asalkeld> and it's effect on the remaining actions 20:36:06 <inc0> asalkeld, it doesn't matter, at first we can do call every few seconds 20:36:08 <asalkeld> also the db load 20:36:40 <inc0> I'm pretty sure db load can be dropped to n*logn and it can be heavily optimized by queries 20:36:53 <inc0> its hard to do optimization on dummy db... 20:37:03 <asalkeld> inc0, agree 20:37:11 <shardy> stevebaker: https://review.openstack.org/138850 20:37:29 <inc0> for first version I was tinking of calling converge every few seconds 20:37:35 <zaneb> well, it's easy to see what does lots of db calls and what doesn't 20:37:38 <inc0> or when resource state change 20:38:05 <inc0> zaneb, when you have things like joins, index lookups and so on you can optimize that 20:38:24 <inc0> also my initial mistake of using name as unique key took its toll 20:38:43 <stevebaker> shardy: +2, waiting for gate pass 20:38:48 <asalkeld> i guess i am in favour of something at least somewhat logically proven - compared to an arguement 20:39:06 <zaneb> asalkeld: +1 20:39:11 <shardy> stevebaker: thanks, I'll look at the devstack fix for the next patch now 20:39:18 <inc0> well, it does solve test cases;) 20:39:25 <zaneb> we all agree that this stuff can _probably_ be solved given enough time 20:39:37 <asalkeld> inc0, dont' get me wrong: i like what you have done 20:39:43 <inc0> I mean I'm not mentally bound to this one, but it does seem to solve biggest space of problems 20:39:52 <zaneb> why would we take 90% chance of a solution in a month over 99% change of a solution right now? 20:39:59 <stevebaker> speaking of which, porting the test cases to functional tests now would benefit all approaches, and would be hugely useful 20:40:13 <asalkeld> stevebaker, +1 20:40:28 <inc0> zaneb, because it solves concurency problems 20:40:37 <inc0> in an elegant, non-hacky way 20:40:43 <asalkeld> stevebaker, i might have to do that iwth shardy's decouple patches 20:40:47 <inc0> concurrent updates I mean 20:41:21 <inc0> or this stuff https://bugs.launchpad.net/heat/+bug/1260167 20:41:23 <uvirtbot> Launchpad bug 1260167 in heat "heat autoscale scale up does not work if delete one instance manually " [Medium,Confirmed] 20:41:46 <asalkeld> so inc0 that is really phase 2 right? 20:41:52 <inc0> I mean it all can be solved by graph approach...but these problems simply won't happen 20:41:55 <asalkeld> having a 'check' 20:41:56 <inc0> with my approach 20:42:11 <asalkeld> and that should be solved 20:42:17 <zaneb> inc0: it really doesn't. to the extent that the other approaches have a problem with concurrent updates (which is only when something is in mid-update and we don't know what to do), yours has exactly the same problem afaict 20:42:55 <inc0> well it doesn't have notion of mid-upgrade 20:43:32 <inc0> so I don't think that would be problem at all 20:43:37 <zaneb> and yet the real world does, with existing plugins 20:43:43 <asalkeld> yeah, there will be tasks working that you won't know what to do with 20:43:58 <stevebaker> shardy: we have a regression in the functional tests http://logs.openstack.org/19/138819/1/check/check-heat-dsvm-functional-mysql/e9d51bc/console.html#_2014-12-03_19_10_38_802 20:44:10 <stevebaker> shardy: I think Qiming raised a bug, let me find it 20:44:20 <inc0> zaneb, we can simulate in_progress, but not build on it 20:44:28 <inc0> I mean in progress == actions left to perform 20:44:33 <asalkeld> shardy, i can look at that 20:44:50 <asalkeld> stevebaker, i mean 20:45:15 <stevebaker> asalkeld: ok, lets make that job voting as soon as its fixed 20:45:24 <asalkeld> totally 20:45:32 <inc0> still, we can work on graph approach, no problem with me 20:45:38 <shardy> stevebaker: interesting, great to see real issues getting caught :) 20:45:45 <inc0> it may give us something faster 20:46:11 <inc0> and we may learn more in the process 20:46:13 <stevebaker> shardy: although I recall Qiming commenting that the test might need fixing, not the code 20:46:45 <asalkeld> stevebaker, i'll raise a bug for it 20:47:19 <stevebaker> https://bugs.launchpad.net/heat/+bug/1397945 20:47:22 <uvirtbot> Launchpad bug 1397945 in heat "resource group fails validate a template resource type" [Undecided,New] 20:47:23 <stevebaker> found it! 20:47:28 <inc0> I just hope we won't get ourselves into some hack-loop with concurrent stuff 20:47:44 <stevebaker> I'll mark it critical, since it breaks the gate job 20:48:05 <asalkeld> #link https://bugs.launchpad.net/heat/+bug/1398973 20:48:07 <uvirtbot> Launchpad bug 1398973 in heat "test_stack_update_provider_group integration test is failing" [Undecided,New] 20:48:32 <shardy> stevebaker: Possibly, I saw his bug but could not reproduce locally 20:49:43 <stevebaker> shardy: so running heat_integrationtests.functional.test_update didn't reproduce? maybe its a race 20:50:14 <inc0> anyway, whichever approach we take, I prefere to start implementing than arguing about PoC 20:50:14 <inc0> s 20:50:26 <asalkeld> inc0, agree 20:50:34 <asalkeld> we need to get cracking on this 20:51:20 <asalkeld> we really don't want to land this to late (like right before release) 20:51:57 <inc0> we can get back to this discussion at phase 2 20:52:05 <inc0> when we won't have fire in our house 20:52:47 <asalkeld> stevebaker, we have bugs for templest 20:53:02 <asalkeld> and we have a spec for functional tests 20:53:09 <asalkeld> do we need both? 20:53:13 <zaneb> inc0: what makes you think the pressure ever lets up? ;) 20:53:56 <inc0> zaneb, I know it never is, thats why I'll be secretly preparing one big stateless fire extinguisher;) 20:54:44 <stevebaker> asalkeld: we need to rename all those tempest tags, maybe to 'functional'. They could stay as wishlist bugs though. 20:54:53 <asalkeld> ok 20:55:01 <asalkeld> i'd like to target some to k2 20:55:16 <asalkeld> and get people commited to them 20:55:35 <asalkeld> 5mins 20:55:51 <stevebaker> asalkeld: things have likely moved on a bit since then. I'd much rather see functional tests for native resources than the AWS ones 20:55:58 <inc0> asalkeld, any chances we could talk tomorrow about getting few actions drafted to start working on upgrades? 20:56:14 <asalkeld> sure inc0 20:56:21 <asalkeld> i'd love to progress that 20:56:47 <inc0> I'll bug you about that tomorrow then (for you it will be later today;)) 20:56:53 <asalkeld> k 20:57:06 <inc0> o/ 20:57:42 <asalkeld> i'll end this 2 early then 20:57:47 <asalkeld> #endmeeting