15:00:36 #startmeeting gantt 15:00:36 Meeting started Tue Apr 14 15:00:36 2015 UTC and is due to finish in 60 minutes. The chair is n0ano. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:37 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:39 The meeting name has been set to 'gantt' 15:00:44 anyone here to talk about the scheduler? 15:00:53 \o 15:01:18 o/ 15:02:34 not much to talk about today, mostly reviews for the release, but... 15:02:44 #topic Vancouver sessions 15:03:01 o/ 15:03:07 I know we're on the list for a scheduler session, do we have any more details we want to put up? 15:03:57 n0ano: I updated the etherpad 15:04:04 https://etherpad.openstack.org/p/liberty-nova-summit-ideas 15:04:08 L50 15:04:17 I saw that, it looks good for getting a discussion going 15:04:18 oops L60 15:04:30 bauzas, yeah, others updated the pad 15:05:10 n0ano: yeah johnthetubaguy added a very good question about the opportunity to split out the sched 15:05:40 the one about external API? 15:05:51 n0ano: technically, everything can be done within Nova without splitting the sched 15:06:01 n0ano: even the cross-project stats 15:06:18 n0ano: but we should raise the point about it not being only a tech thing 15:06:28 bauzas, +1 15:06:36 n0ano: but also a scaling out development team 15:06:44 bauzas, all good ideas that are appropriate for the summit session 15:07:29 n0ano: that said, that's still an excellent question - ie. I don't care if we don't split the scheduler, I'm just concerned about making sure we can fix our problems quickly 15:07:36 from a session perspective, I think what's on the pad is good enough so I think we're set 15:07:50 so we can focus on a cross-project thing soon 15:08:41 n0ano: I'm also in the cells V2 effort and those guys desesperatly need a scalable scheduler 15:08:58 bauzas: +1 15:09:09 cells, nova, cinder, ironic - there are lots of people that need a scheduler 15:09:31 n0ano: so I'm beginning to draft a few things in my mind to see how we can achieve a shared-state scheduler 15:09:50 n0ano: cells primarily needs the current scheduler to scale, not for it to split 15:09:56 n0ano: technically cells is just a nova thing 15:09:59 bauzas, beyond what we were supposed to get with the resource tracker? 15:10:03 n0ano: it doesn't it to be split 15:10:08 ergh alaski burned me 15:10:52 n0ano: the resource tracker is just an ugly thing for providing food for thoughts to the scheduler 15:11:19 n0ano: I raised an old BP led by jay about allocation ratios which need to move on to the RT 15:11:29 I 15:11:35 n0ano: we also need to consider how we can put claims to the scheduler 15:11:49 hi guys. 15:11:57 but I don't think those two tasks are necessary for scaling out the sched 15:11:59 jaypipes: \o/ 15:12:10 sorry, been on vacation until about 2 hours ago. 15:12:12 jaypipes: you were under radar :) 15:12:18 jaypipes: np 15:12:20 yes, deliberately :) 15:12:24 I'm a little worried that the scheduler was ugly so we create the RT and now the RT is ugly so we create something else, a lot of spinning going on 15:12:56 n0ano: we reduced the tech debt by a good level in Kilo 15:13:47 jaypipes: so, we could discuss further on but I was basically saying that I provided a list of things to do for Liberrty 15:13:54 bauzas, for the priority tasks, there's still a lot of other tech debt to deal with 15:14:09 jaypipes: do you think you will have time to look at bp/resource-usage ? 15:14:14 ergh 15:14:18 bp/resource-objects 15:14:49 jaypipes: I also hijacked bp/allocation-ratios-to-RT because we need it soon :) 15:15:01 n0ano: which are ? 15:15:13 bauzas: I will tackle resource-usage this week. 15:15:19 jaypipes: \o/ 15:15:21 bauzas, just in general, a lot of overload on the core team 15:15:38 bauzas: cool on allocation-ratio one. 15:15:42 n0ano: so we need to get a priority for Liberty 15:15:58 n0ano: because it damned worked good 15:16:10 worked *well 15:16:22 fat fingers and a f* English 15:16:22 bauzas, well, it'll be interesting to see what other items want to be a priority 15:16:29 n0ano: sure 15:16:53 n0ano: but wrt cells, scheduling is just a necessary for them too 15:16:54 the process works great if you're a priority, not so much if not 15:17:05 n0ano: so let's us become a priority thing 15:17:10 * PaulMurray o/ - sorry I'm late 15:17:27 PaulMurray, NP 15:17:47 n0ano: honestly, we're one of the priorities which kinda succeeded on the last cycle 15:18:20 n0ano: but sure, we can't bet on that - at least I wouldn't 15:18:32 bauzas, +1 15:19:03 well, I think we're actually ready for Vancouver, what a surprise 15:19:16 soooo, I would just say that we need to continue focusing on improving the scheduler while considering why a split would be important at the same time 15:19:39 n0ano: well, we need some housekeeping stuff 15:20:00 jaypipes: would you mind resubmit bp/resource-objects for Liberty as a Previously-approved spec ? 15:20:05 what housekeeping are you thinking of, other than re-submitting specs for liberty 15:20:16 n0ano: exactly that one :D 15:20:27 good seque... 15:20:35 #topic Specs for Liberty 15:20:47 so I made a few uploads 15:20:54 https://review.openstack.org/173252 15:21:01 now that L is open we shoul re-submit all our open specs again, this is mainly to bauzas and PaulMurray 15:21:09 https://review.openstack.org/173316 15:21:12 bauzas: yes, I will do that. 15:21:24 n0ano: and jaypipes's bp/resource-objects 15:21:32 jaypipes: excellent, ty 15:21:49 and PaulMurray's make-rt-use-objects 15:22:10 then, I'll sort it out with johnthetubaguy for the ones who are consensual 15:22:12 n0ano, johnthetubaguy approved the rt-objects as a trivial-not-needing spec 15:22:17 do ping me if you need help getting stuff approved 15:22:33 PaulMurray, reall? cool, that makes your life simpler 15:22:33 johnthetubaguy, will I need to bother with the spec ^^^ 15:22:37 PaulMurray: cool, then let's resume your work on that bp 15:22:57 johnthetubaguy: I made a few updates for the L specs 15:23:04 PaulMurray: treating it as an objects tidy up / bug fix, given we discussed that spec enough before now 15:23:21 PaulMurray: lets see if everyone agrees, but I would ignore the spec thing for now 15:23:24 johnthetubaguy: I wouldn't bet it could be fast-approved as I needed to make some adjustements 15:23:45 bauzas, adjustments to the spec or the patches? 15:23:51 n0ano: on the specs 15:23:59 bauzas: good heads up 15:24:27 n0ano: some content was outdated because we merged a few things and some content was wrong because the implementation draft showed it was an error 15:24:40 unless the bauzas adjustments are major we should still be able to fast track the spec 15:24:58 n0ano: I leave that to the nova-drivers 15:25:16 n0ano: that doesn't block me to work on the implementation 15:26:09 so, bottom line, PaulMurray doesn't need to re-submit, hopefully we can fast track bauzas & jaypipes specs 15:26:32 that leaves others think about other things they want to shape for L 15:26:41 bauzas, that 15:26:48 * n0ano fat fingers also 15:26:49 n0ano: on my side, I'm pretty staffed for Liberty 15:27:01 bauzas, that's always the case of other developtment 15:27:30 agreed 15:27:43 OK, moving on... 15:27:47 #topic opens 15:27:51 anything new for today? 15:28:33 hearing crickets 15:28:48 * n0ano is afraid he will have to go back to re-wiring his house (long story) 15:29:09 re-wiring ? 15:29:26 is it what I understand ? 15:29:44 100Mbit needs 4 wires from a Cat-5 cable, 1G needs all 8 leaving nothing for the phone line 15:29:47 n0ano: invest on IoT 15:30:23 getting 1G ethernet & phone is `challenging` 15:30:42 ? 15:31:16 anyway, that's off-topic but I can help you, I have a personal home wired network that can do both 15:31:17 bauzas, my house only sent 1 single Cat5 cable to each room, I want 1G 15:31:42 n0ano: got it, move to Cat6 but that shouldn't impact your phone line 15:31:52 I have the solution, just requires re-wiring 15:32:01 we can compare notes at Vancouver 15:32:04 or buy a SIP hone 15:32:08 phone 15:32:19 OK, I think we're done on the scheduler 15:32:21 lolo 15:32:22 tnx everyone 15:32:26 \o 15:32:28 #endmeeting