21:00:15 #startmeeting nova_cells 21:00:16 Meeting started Wed May 3 21:00:15 2017 UTC and is due to finish in 60 minutes. The chair is dansmith. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:17 um 21:00:18 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:20 The meeting name has been set to 'nova_cells' 21:00:25 HI! 21:00:26 heh 21:00:39 o/ 21:00:47 it's the first amazing-weather day outside so I want to make this quick 21:00:50 alaski: ohai :) 21:01:19 thought I would drop in today 21:01:29 no dtp 21:01:31 melwitt: ? 21:01:40 oops 21:01:43 FYI, when I quit openstack, don't expect me to come back around.. ever :) 21:01:44 she's still unicoding like crazy 21:02:08 omg alaski is here, sweet 21:02:14 dansmith: :) it's like a bad habit, hard to drop completely 21:02:15 #topic testing/bugs 21:02:22 openstack: you're slow today 21:02:35 so this is a bug fix https://review.openstack.org/#/c/460616/ 21:02:36 so we had a cellsv1 thing, that's the unicode thing for some reason? 21:02:47 yep, that too 21:03:10 sean was thinking the use_journal=True stuff was causing the unicode problem somehow maybe 21:03:29 okay, he had a change to revert that.. is it queued or landed? 21:03:36 dansmith: yeah, I think it's not directly related. that the deadlock issue was making an already existing problem cause job failures by chance 21:03:39 https://review.openstack.org/#/c/462163/ isn't merged yet 21:03:43 use_journal seemed to be having interesting affects on my stuff too 21:03:54 so we won't know until that merges 21:03:59 okay 21:04:03 until then we're at 50% fail rate on the cells v1 job 21:04:13 which is fing awesome 21:04:22 yeah that sucks 21:04:32 we're always had the unicode problem when logging whole Instance objects, it just never caused job failures before I guess 21:04:38 right 21:04:41 we still need to fix that bug 21:04:46 and by we i mean mel 21:04:48 seems odd that that never popped up before 21:04:54 dansmith: it's been around for awhile 21:04:59 just hasn't caused failures for whatever reason 21:05:01 it's been tracing this whole time but not causing fails in jobs 21:05:08 ah, okay 21:05:13 what I meant was, 21:05:26 if it was fatal, not sure how it wasn't a red hot issue for anyone in, say, china 21:05:30 I have a patch up that fixes it. sdague wanted me to explore the possibility of a bug in ovo, but at least to me it doesn't seem like a bug in ovo 21:05:53 I put up a unit test that shows what ovo __repr__ does at: https://review.openstack.org/462288 21:06:01 we have to make sure we only change repr code in ovo if we do, because otherwise we'll affect on-the-wire formats 21:06:51 dansmith: yeah. what I did has nothing to do with changing any repr code. I just converted the instance object string repr to unicode before calling the logging function 21:07:12 melwitt: yeah I meant _if_ we make a change in ovo 21:07:16 that unit test seems fine to me 21:07:26 like, slight change between py2 and py3 for logging is meh 21:07:40 same thing we've been doing in these cases, just that for the Instance object I found I had to specify the encoding 'utf-8' explicitly to six.text_type() otherwise it wouldn't work (it blew up otherwise) 21:07:42 er, slight difference I mean 21:07:58 dansmith: oh, okay 21:08:24 so sounds like we should merge the nova fix? 21:09:06 it's a wip still 21:09:06 https://review.openstack.org/#/c/461912/ 21:09:16 seems like it. I'm not crazy about the change since it's like whack-a-mole trying to catch things that could be unicode, and writing unit tests is similarly meh 21:09:49 the whack a mole nature does suck, but meh 21:09:49 as in, if someone adds a logging of Instance object somewhere later nothing is going to catch that if I write unit tests for only the specific places that log it that way 21:09:52 i don't hate it 21:10:08 we whack a mole leaking bdm.connection_info too 21:10:08 but I don't see a better way around it 21:10:10 which has credentials 21:10:28 that's why i thought we could do something in the stringify method for the String field type in ovo 21:10:32 but i guess not? 21:10:37 well, merging it for now seems prudent to stop the current bleeding regardless 21:10:41 so, the todo on that change is: rewrite the commit message again bc it's wrong bc I got confused, and add unit tests for those Instance object logging areas 21:11:59 mriedem: the ovo unit test I wrote might make it clear, but the ovo repr is intentionally returning ascii encoding. so that things like str() will not blow up in a python2 env. otherwise people would have to have foresight that something could be unicode 21:12:53 oh 21:13:01 okay, well, let's not bog down the cells meeting with ovo details 21:13:26 merging the change in the cells code to stop the bleeding seems not terrible 21:13:40 and if we pursue something generalized then that's cool, but... 21:14:00 we could also make a simple utility method in nova for the workaround, 21:14:04 i just commented on that in the change 21:14:08 that's less bad then at least 21:14:41 okay, any other bugs or test-related things to cover besides these two? 21:14:46 thanks. that's a good idea. I'll do that and add unit tests and then it should be good to go 21:15:29 #topic open reviews 21:15:46 So probably the biggest thing is melwitt's quotas set 21:15:56 I've seen some reviews on that so far, but it needs more 21:16:15 is it awaiting any changes for latest reviews, or fresh round is ready? 21:16:15 I'll likely try to spend some time on that on the plane if I'm able 21:16:17 it is a beast. johnthetubaguy and jaypipes added some comments today, which is good. I have replied to them already 21:16:22 ok 21:16:27 #link https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/cells-count-resources-to-check-quota-in-api 21:16:36 i've gotten a decent start on the service/compute node uuid stuff in api here https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/service-hyper-uuid-in-api 21:16:47 ^ are very small and easy 21:16:48 * dansmith #link https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/service-hyper-uuid-in-api 21:16:50 cool 21:16:51 building up to the REST API changes 21:16:54 I like small and easy 21:16:56 which are going to be suck 21:17:29 wouldn't be so bad if i didn't have to rewriting both apis basically completely, but that's what the specs people wanted... 21:17:47 effing specs people 21:18:03 friday might be quotas review day for me 21:18:08 b/c i'm stressed about summit prep 21:18:32 where are we on the cells-aware-api changes? 21:18:40 https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/cells-aware-api 21:18:42 oh just that 21:18:46 and the devstack patch 21:18:47 if you want to jump on a hangout to ask questions or have me explain things about the quotas changes, I'm more than happy to do that 21:18:53 o 21:18:54 ok 21:19:22 like, whenever. that will make things a lot easier I think, because going back and forth on the review is pretty slow. I tried to comment the hell out of it but I think it's still hard to understand 21:19:30 yeah, the cells aware api patches are pretty good.. the ones I've got pending are cleanups and some fixes, but I kinda want to get the devstack stuff working so I can poke at some of them 21:19:57 ok so maybe a hangout on friday morning ish 21:20:15 so for melwitt that's like noon your time 21:20:18 which mel's morning is.. 21:20:18 yeah 21:20:19 ha 21:20:22 :D 21:20:28 friday she has to be up early for a team thing 21:20:31 so you might get her earlier 21:20:37 I think she does that call from bed though 21:20:45 yeah, just saying if you want clarification on anything just ask and I'm happy to get on hangouts whenever you have a question 21:21:04 lol, verrry funny 21:21:14 21:21:21 anything else on open reviews? 21:21:25 that's it from me 21:21:25 we're really sucking at making this quick 21:21:39 #topic open discussion 21:22:00 I was just going to ask for volunteers to look at my slides, but everyone here has already done that 21:22:15 well, aside from laski, but I shan't ask him :) 21:22:15 * alaski has not 21:22:25 I'm not sure I can be useful though :) 21:23:16 I'll be glad to float you a url if you care, but I'd not expect you to really look 21:23:22 I've had several people comment already 21:23:25 so I think I'm pretty good 21:23:54 I'd take a look, but may not be able to make useful comments 21:23:59 anything else for open discussion? 21:24:40 alaski: I think you could, fwiw. it goes over how it works, how cells v1 works, why cells v2, etc 21:24:57 ahh, cool. I'll check it out 21:25:29 last call... 21:25:33 end it 21:25:35 (...for alcohol) 21:25:43 it's a nice day, go outside 21:25:48 already am :) 21:25:53 nice 21:25:53 #endmeeting