14:00:22 <edleafe> #startmeeting nova_scheduler 14:00:23 <openstack> Meeting started Mon Jan 22 14:00:22 2018 UTC and is due to finish in 60 minutes. The chair is edleafe. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:24 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:26 <openstack> The meeting name has been set to 'nova_scheduler' 14:00:31 <edleafe> I am lacking caffeine! 14:00:33 <efried> Dang dude, do an espresso shot or something. 14:00:35 * bauzas yawns 14:00:36 <efried> \o 14:00:38 <takashin_> o/ 14:00:41 <ttsiouts> o/ 14:00:45 <edleafe> I'm back from my hellish commutes to Austin. Thanks to efried for running things in my absence. 14:01:00 <efried> edleafe I was about to say - you would drive in this condition?? 14:01:35 <edleafe> Sure - doesn't take much brainpower to stare at red tail lights 14:01:44 <alex_xu> o/ 14:02:24 <jroll> \o 14:02:35 <bauzas> edleafe: Austin commutes ? hah 14:02:53 <edleafe> bauzas: yeah - IBM brilliance 14:03:06 <bauzas> that's what I understood :-) 14:03:47 <edleafe> Well, let's get started 14:03:47 <edleafe> #topic Feature Freeze 14:03:50 <edleafe> Note that Feature Freeze for Queens is this coming Thursday, January 25 14:04:11 <edleafe> Please spend any and all available cycles reviewing the patches that are still in progress so that we can get them into Queens. 14:04:23 <efried> Now what does this mean, exactly? 14:04:38 <efried> That anything blueprint-related that's not already merged won't be merged? 14:04:39 <edleafe> Anything not merged will require an exception 14:04:52 <efried> Even if already proposed and mostly up to snuff 14:04:56 <edleafe> and if it isn't super-close, it will be put off until Rocky 14:05:49 <edleafe> It's gotten much stricter in recent cycles, so don't count on getting an exception - get it merged! 14:06:02 <bauzas> yup, in general, we only accept exceptions for things missing because of gate issues 14:06:11 <bauzas> like a +W waiting to be merged 14:06:12 <efried> Well, there's been plenty of that lately. 14:06:19 <efried> But okay, got it. 14:06:32 <edleafe> Any other questions about Feature Freeze? 14:06:33 <bauzas> except super-mega-urgent concerns that really require a merge 14:06:49 <edleafe> And, as always, bug fixes are always accepted 14:06:59 <bauzas> edleafe: tbc, I'll spend tomorrow reviewing scheduler changes 14:07:08 <edleafe> awesome 14:07:08 <bauzas> so, hold your breath 14:07:13 <efried> Bug: my feature doesn't work because it needs more patches. 14:07:27 <bauzas> I'm myself stopped by libvirt 14:07:32 <edleafe> efried: Matt isn't that gullible :) 14:07:35 <bauzas> for the vGPU changes 14:07:43 <bauzas> so, let's see what others do 14:07:58 <edleafe> Moving on... 14:07:59 <edleafe> #topic Specs & Reviews 14:08:09 <edleafe> #link update_provider_tree series - starting with https://review.openstack.org/#/c/533244/ 14:08:14 <edleafe> The bottom patches close to ready. 14:08:23 <edleafe> efried: comments? 14:08:53 <efried> Bottom three have gibi +2, expect jay to +A pretty much right away since they've had his comments addressed at this point. 14:09:05 <efried> Sorry, bottom two. 14:09:15 <efried> But the rest of the series up to the WIPs is also ready 14:09:26 <efried> and similarly has had Jay's concerns addressed. 14:09:50 <efried> I guess I will spend the next day or so getting the WIPs un-WIP'd. 14:10:01 <edleafe> ok, we'll be sure to keep those tabs open for review 14:10:13 <edleafe> and let us know if there is anything we can help with 14:10:19 <efried> Not sure where we would stand wrt the xen and vmware impls on top of those. Do those count as "features" for the sake of feature freeze? 14:10:25 <efried> I.e. do they stand any chance of getting into Q? 14:10:29 <edleafe> Since the work on my series is done 14:10:38 <efried> braggart 14:11:02 <edleafe> efried: generally yes, they would count as features 14:11:27 <edleafe> it all depends on how close they are, and especially how comprehensive the testing around them are 14:11:47 <edleafe> Confidence, and all 14:11:55 <efried> Okay, that'll be something to note for the authors of those (rgerganov naichuans_) 14:12:03 <bauzas> efried: fasten your seatbelt, I'm planning to use update_p_t by Rocky 14:12:30 <efried> bauzas Good deal. Hopefully me too (PowerVM). 14:12:30 <bauzas> the fact is, I'm a poor lone cow-boy for implementing it w/ libvirt :p 14:13:05 <efried> bauzas et al, as a first step, it should be easy as pie to implement exact parity with get_inventory 14:13:12 <edleafe> bauzas: https://goo.gl/images/Wbn9Df 14:13:25 <efried> Then we can do the more intricate/powerful stuff from there. 14:13:49 <efried> edleafe Guy on the left looks a little bit like me 20 years ago. 14:15:04 <edleafe> Next up: 14:15:05 <edleafe> #link Nested RP traits selection https://review.openstack.org/#/c/531899/ 14:15:08 <edleafe> Starts with a test to assert correct behavior, though the implementation for it isn't there yet 14:15:32 <edleafe> Guess Jay isn't around to comment 14:16:12 <efried> Looks like the top of the series has a WIP for that, but looks pretty raw. 14:17:22 <edleafe> Do you know if the plan is to squash that into the implementation, or reverse the order of the patches? 14:17:34 <edleafe> Because that patch won't merge before the implementation 14:17:46 <efried> Sorry, I don't follow. 14:18:17 <efried> The patches with test cases have the actual expected results commented out. Then the real impl patch will uncomment that bit so the assertions hold. 14:18:17 <edleafe> If the implementation is later in the series, the functional test patch won't ever pass 14:18:42 <edleafe> efried: ok, I haven't looked into it that closely 14:18:42 <bauzas> edleafe: my reference was more https://en.wikipedia.org/wiki/Lucky_Luke 14:18:43 <efried> I believe we're following what shall henceforth and forevermore be known as The Gibi Process. 14:19:04 <edmondsw> test driven development? 14:19:09 <efried> Shhhhh! 14:19:27 <efried> gibi totally invented it. 14:20:00 <efried> edleafe Since this wasn't added to the agenda explicitly, it should be noted that... 14:20:19 <edleafe> Favoring functional tests over unit tests?? What a concept!! 14:21:00 <efried> Traits affordance *was* going to be folded into the granular work, BUT since the last stage of the granular work prereqs the above (NRP in alloc cands), it's not likely to make Q at this point. So alex_xu has a series up to expose (single-request-group) trait consumption. 14:21:21 <alex_xu> #link https://review.openstack.org/#/q/topic:bp/request-traits-in-nova+(status:open+OR+status:merged) 14:21:27 <efried> Series starts: https://review.openstack.org/#/c/535642/ 14:21:31 <alex_xu> yea, here is link 14:21:44 <edleafe> ok, thanks efried and alex_xu 14:21:54 * efried adds to agenda... 14:21:54 <alex_xu> thanks to efried, I just need to a very thin code to expose the traits out 14:22:27 <edleafe> #link Single-request-group trait consumption: https://review.openstack.org/#/c/535642/ 14:23:09 <edleafe> And the link to the granular series efried mentioned: 14:23:12 <edleafe> #link Granular resource requests review://openstack.https.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/granular-resource-requests 14:23:30 <alex_xu> except I was tried by version 1.12 14:23:36 <cdent> I think it's probably useful to point out that the "single-request-group" trait stuff is still quite useful without the numbered groups 14:23:51 <edleafe> So is the plan to get alex_xu's patch in Queens, and then continue efried's series in Rocky? 14:24:27 <efried> edleafe I think so. 14:24:51 <efried> I don't see a way we're going to get Jay's series finished and merged in time to finish up granular and get THAT merged. 14:25:02 <efried> I'm frankly not convinced even the former is going to happen. 14:25:09 <edleafe> understood 14:25:46 <edleafe> Last one I have is the alternate hosts series: 14:25:47 <edleafe> #link Use alternate hosts for resize https://review.openstack.org/#/c/526436/ 14:26:09 <edleafe> One patch left, just waiting for a +W 14:26:18 <efried> edleafe Looks like needs a rebase 14:26:28 <edleafe> Heh, wouldn't doubt it 14:26:58 <edleafe> Anything other reviews to discuss? 14:27:02 <edleafe> geez 14:27:14 <edleafe> s/Anything/Any 14:28:26 <edleafe> OK, moving on then 14:28:38 <edleafe> #topic bugs 14:28:44 <efried> edleafe FYI, that rebase is going to be manual 14:28:59 <edleafe> #link Placement bugs: https://bugs.launchpad.net/nova/+bugs?field.tag=placement 14:29:24 <edleafe> efried: I do 'em all manually - lots of practice this cycle! 14:29:38 <cdent> i had added something to the agenda for another pending review 14:29:40 <edleafe> A few new ones, but nothing critical 14:30:01 <edleafe> #undo 14:30:01 <openstack> Removing item from minutes: #link https://bugs.launchpad.net/nova/+bugs?field.tag=placement 14:30:04 <edleafe> #undo 14:30:05 <openstack> Removing item from minutes: #topic bugs 14:30:11 <cdent> #link removing microversion fallback: https://review.openstack.org/#/c/528794/ 14:30:16 <edleafe> #topic Reviews 14:30:23 <edleafe> go ahead, cdent 14:30:27 <cdent> basically we should do that sooner than later if we can 14:31:16 <edleafe> cdent: would that be considered a bug? Or potential bug? 14:31:34 <cdent> it's a thing that matt asked me to do 14:31:42 <cdent> so that's kind of like a bug 14:32:09 <edleafe> more like a cleanup of dead code 14:32:56 <edleafe> ok, thanks 14:33:02 <cdent> it's also a month old 14:33:19 <edleafe> Sorry I missed that - I prepared my notes ahead of time and didn't check for agenda updates 14:33:51 <edleafe> #topic bugs 14:33:53 <edleafe> #link Placement bugs: https://bugs.launchpad.net/nova/+bugs?field.tag=placement 14:34:09 <edleafe> As noted earlier, a few new ones, but nothing critical 14:34:26 <edleafe> Anyone have any bug-related comments? 14:35:08 <edleafe> #topic Open Discussion 14:35:21 <edleafe> The floor is open... 14:35:35 <gibi> one thig. I did not invented TDD :) 14:35:42 <efried> Shhhh! 14:35:52 <cdent> gibi: claim it. that way we can use it and people won't resist 14:35:54 <efried> All hail gibi, inventor of The Gibi Process 14:35:57 <cdent> and the world will be so much better 14:36:13 <bauzas> just in case people wanna use virtual gpus for mining, there are some changes still up :p 14:36:26 <edleafe> TDD is too unit-testy compared to the Gibi Process 14:36:32 <bauzas> if you want money, think of my patches 14:36:34 <bauzas> (c) me 14:36:36 <cdent> edleafe: bollocks 14:36:51 <cdent> unit tests are too unit-testy 14:37:10 <efried> edleafe (Translation, in San Antonio terms: No way, Jose) 14:38:37 <edleafe> I think we're done here. 14:38:49 <edleafe> Everyone start reviewing patches! 14:38:56 <edleafe> Let's get all of this in Queens! 14:39:07 * edleafe waves his pom-poms 14:39:20 * cdent blinks 14:39:22 <edleafe> And with that... 14:39:25 <edleafe> #endmeeting