21:00:00 <dansmith> #startmeeting nova_cells 21:00:01 <openstack> Meeting started Wed Apr 5 21:00:00 2017 UTC and is due to finish in 60 minutes. The chair is dansmith. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:02 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:05 <openstack> The meeting name has been set to 'nova_cells' 21:00:23 <dansmith> hoodat 21:00:25 <melwitt> ahoy 21:00:30 <dtp> hoo dee hoo 21:00:43 <mriedem> o/ 21:01:03 <dansmith> #topic testing and boogs 21:01:06 <dansmith> mriedem: I'm sure you want to say something here 21:01:20 <mriedem> nope 21:01:21 <mriedem> well, 21:01:22 <mriedem> bugs? 21:01:25 <dansmith> bugs yeah 21:01:25 <mriedem> yeah, we've got'em! 21:01:39 <dansmith> um 21:01:39 <melwitt> we have a treasure trove of bugs 21:01:57 <mriedem> i'm back 21:02:00 <dansmith> heh 21:02:01 <mriedem> https://review.openstack.org/#/c/453859/ and down 21:02:14 <mriedem> fixes the quota double decrement and shelved offloaded local delete thing 21:02:20 <mriedem> once those are in i'll backport to ocata 21:02:26 <dansmith> woot, thanks for your work on those 21:02:33 <mriedem> and then i'm going to start poking on cleaning up that turd 21:02:36 <dansmith> once those are merged, I'll make some more bugs 21:02:43 <mriedem> peeling back the 'temporary' transitional stuff 21:03:03 <melwitt> me too, I'll make more bugs 21:03:14 <mriedem> you bastards 21:03:18 <dansmith> heh 21:03:25 <melwitt> ;D 21:03:37 <dansmith> so, I'll put this in testing since it kinda is: 21:03:43 <dansmith> https://review.openstack.org/#/c/436094 21:03:53 <dansmith> that's a successful run with multiple cells in devstack 21:04:17 <dansmith> I asked sdague this morning to start reviewing the carnage in there so I can get the devstack part closer to acceptable 21:04:33 <dtp> nice 21:04:39 <dansmith> and I haven't done the actual fleet part of moving the conductors to the second node 21:05:03 <dansmith> not sure that would pass nova's full complement of tests, but it's something 21:05:25 <mriedem> well, 21:05:32 <mriedem> didn't we talk about some specific test scenarios there? 21:05:41 <mriedem> we have some notes from the ptg, i just don't have it in my head right now 21:05:51 <mriedem> like, 21:05:55 <mriedem> live migration with multiple cells 21:06:00 <mriedem> the scheduler restricts that now though 21:06:03 <dansmith> mriedem: well, we're going to need to make the live migration job run the cold migration tests 21:06:11 <dansmith> to make sure we keep our coverage 21:06:26 <dansmith> I guess I should dig into that this week 21:06:29 <mriedem> the normal dsvm multinode job runs cold migration 21:06:34 <mriedem> via resize i think 21:06:45 <dansmith> right but once it is multicell it won't test actual moves 21:07:11 <mriedem> hmm, 21:07:11 <dansmith> the thinking i had at ptg was to make the live migration job just one cell, and have it run the cold migration jobs too, 21:07:17 <dansmith> which will be our coverage of actually moving things 21:07:27 <mriedem> so we always just have 2 computes right? 21:07:37 <dansmith> and then once we are multicell on the main job, things won't actually move because the two computes will be in separate cells 21:07:40 <mriedem> maybe i was thinking with the fleet we grew a 3rd compute 21:07:45 <mriedem> but that's a 3-node job 21:07:48 <mriedem> which we're not doing 21:07:52 <mriedem> b/c $$$ 21:07:56 <dansmith> right 21:08:03 <mriedem> ok i'm following now 21:08:26 <mriedem> so, 21:08:32 <mriedem> multicell on the main job == single-node dsvm? 21:08:39 <mriedem> or dsvm multinode? 21:08:45 <mriedem> b/c it has to be the latter right? 21:08:51 <dansmith> just the latter yeah 21:08:52 <mriedem> single-node dsvm is just single cell 21:09:03 <dansmith> we'll never create another cell on the single job obviously 21:09:07 <mriedem> yeah 21:09:10 <dansmith> then on the live migration job we'll force it to be single-cell even though it's two computes 21:09:15 <dansmith> and then we'll keep our coverage 21:09:20 <mriedem> this just gets weird, because the multinode dsvm job you'd think you could run migrations between hosts 21:09:21 <mriedem> but nope 21:09:37 <mriedem> we'll probably need a treasure map at some point 21:09:39 <mriedem> for QA and infra 21:09:40 <dansmith> yeah, well, 21:09:50 <dansmith> we could grow another multinode job just for this, but...$$$ 21:09:55 <mriedem> maybe a devref doc on cells v2 testing permutations 21:10:00 <dansmith> yeah 21:10:02 <mriedem> i'm fine with the plan, 21:10:09 <mriedem> we just need to write it down publicly 21:10:14 <dansmith> well, once we have things up to make those changes and people are happy, writing it down would be fine 21:10:15 <dansmith> yeah 21:10:27 <dansmith> we just need to make sure our devref doesn't get out of sync with infra 21:10:30 <dansmith> which could easily happy 21:10:32 <dansmith> happen 21:10:42 <dansmith> and older releases will codify incorrect details about infra 21:10:45 <dansmith> anyway 21:11:02 <dansmith> anything else for testing/bugs? 21:11:07 <mriedem> yeah branches.. 21:11:11 <mriedem> that's going to get extra nasty 21:11:24 <mriedem> but we'll see 21:11:48 <dansmith> #topic open reviews 21:11:52 <dansmith> got plenty of these too 21:12:02 <dansmith> progress on my set has kinda stalled out, 21:12:11 <dansmith> melwitt has a thousand quotas ones open (thanks to me) 21:12:24 <dansmith> mriedem has can't-stop-writing-specs-itus 21:12:24 <mriedem> the quota ones need work don't they? 21:12:30 <dansmith> they do 21:12:32 <mriedem> i'm burned out on the specs bullshit 21:12:38 <dansmith> oh good 21:12:40 <melwitt> I'm close to the end, I'm going through each seam and making sure unit tests pass, that sort of thing 21:12:49 <mriedem> dtp has started https://review.openstack.org/#/c/453025/ 21:12:49 <dansmith> melwitt: excellent 21:12:58 <mriedem> i -1ed immediately 21:13:05 <mriedem> just to be friendly 21:13:05 <dtp> heh 21:13:10 <dansmith> good, keep that self esteem from getting out of control 21:13:32 <mriedem> dtp and i will have to pow-wow on splitting up the api changes probably 21:13:40 <mriedem> b/c it's 2 apis, a bunch of changes each, but a single microversion 21:13:50 <dansmith> ack 21:13:51 <melwitt> what could go wrong 21:13:58 <mriedem> i'm thinking we'll work them as separate changes stacked, then collapse once they are ready 21:14:02 <dansmith> yeah 21:14:31 <dansmith> could do the divorced parents thing and only have the kids every other day 21:14:42 <dansmith> and fight over the holidays 21:14:51 <mriedem> maybe 21:15:12 <mriedem> you also have to cleanup your faster host mapping discovery series 21:15:19 <dansmith> oh 21:15:21 <dansmith> yeah I bet I do 21:15:29 <dansmith> that thing gets lost behind the cells one 21:15:31 <mriedem> and the nova-status check for min compute 21:15:51 <dansmith> I think I'm annoyed with (you on) that one and have been ignoring it 21:15:54 <mriedem> i know 21:15:58 <dansmith> hah 21:15:59 <mriedem> i could feel it 21:16:02 <dansmith> that obvious? 21:16:06 <mriedem> yes 21:16:07 <dansmith> lol 21:16:08 <dansmith> damn 21:16:13 <dansmith> thought I was sneaky-er than that 21:16:21 <melwitt> hah 21:16:50 <dansmith> okay anything else here? 21:16:54 <mriedem> nope 21:17:03 <dansmith> I have to go get my mid-nova-gathering haircut when we're done here 21:17:17 <mriedem> it's a bit early... 21:17:19 <mriedem> you're a month away 21:17:21 <dansmith> #topic open discussion 21:17:25 <dansmith> yeah, but I'm hairy 21:17:37 <dansmith> I mean I have to get the mid one so I can get another right before boston 21:17:38 <mriedem> i think poofy is the word 21:17:58 <dtp> dansmith thanks for writing this > http://superuser.openstack.org/articles/upgrades-in-nova-database-migrations/ 21:17:59 <dtp> it helped 21:18:10 <dansmith> dtp: heh cool 21:18:38 <dansmith> okay anything else? 21:18:39 <mriedem> i'll poke on https://review.openstack.org/#/c/439891/ before the kid gets home 21:19:03 <dansmith> thanks 21:19:15 <dansmith> going twice... 21:19:36 <dansmith> #endmeeting