14:00:08 <edleafe> #startmeeting nova_scheduler
14:00:09 <openstack> Meeting started Mon Jan 23 14:00:08 2017 UTC and is due to finish in 60 minutes.  The chair is edleafe. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:10 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:12 <openstack> The meeting name has been set to 'nova_scheduler'
14:00:15 <cdent> o/
14:00:19 <_gryf> o/
14:00:20 <edleafe> Good UGT morning everyone!
14:00:20 <macsz> o/
14:00:45 <jroll> \o
14:00:49 <diga> o/
14:01:21 <edleafe> I saw cdent mention that jaypipes is on PTO today
14:01:31 <cdent> all week
14:01:35 <edleafe> ah
14:01:38 <alex_xu> o/
14:02:32 <edleafe> Let's get started
14:02:34 <edleafe> #topic Specs & Reviews
14:02:42 <bauzas> hola
14:02:45 <edleafe> Sorry I didn't update the agenda last week
14:02:48 <bauzas> just going quick here
14:03:01 <bauzas> given I'm mostly rushing for implementing the scheduler upcall
14:03:08 <edleafe> Nested Resource Providers: series beginning with:
14:03:09 <edleafe> #link https://review.openstack.org/#/c/415920/
14:03:33 <edleafe> Jay isn't here, so unless anyone objects I'll take care of rebasing those
14:03:47 <cdent> +1
14:03:50 <edleafe> Next up is bauzas's series
14:03:50 <edleafe> Resource Providers Scheduler DB Filters: series beginning with:
14:03:51 <edleafe> #link https://review.openstack.org/#/c/424110/
14:04:06 <bauzas> we merged a few patches last week
14:04:08 <edleafe> bauzas: anything you'd like to point out?
14:04:20 <bauzas> mostly about the functional tests modifications needed
14:04:31 <bauzas> so we created a new PlacementFixture
14:04:44 <bauzas> we also merged an object method for getting the list of Compute nodes
14:04:54 <bauzas> now, I have a last prereq for fixing the fake driver
14:05:08 <bauzas> and once all of that is merged, the func tests in the main change should be green
14:05:22 <bauzas> I'm now investigating why our tempest jobs are telling that the disk usage is 0
14:05:29 <bauzas> that should be the last problem
14:05:55 <bauzas> so, hopefully all of that should be sorted tomorrow so we could rush out for merging that very important change
14:06:04 <bauzas> that's it for me
14:06:39 <edleafe> OK, let's all put some eyes on that series
14:07:18 <_gryf> +1
14:07:21 <edleafe> Next up: CORS support:
14:07:24 <edleafe> #link https://review.openstack.org/#/c/392891/
14:07:33 <edleafe> that one's been sitting a while
14:07:41 <edleafe> cdent: anything needed to push it through?
14:07:56 <cdent> just review
14:08:26 <cdent> it hasn't changed in any significant way since jay's +2
14:08:28 <edleafe> OK, you heard the man, people
14:08:37 <edleafe> Review that thing!! :)
14:08:51 <bauzas> I forgot something important
14:08:55 <edleafe> go ahead
14:09:05 <bauzas> I need eyes on a ML thread I wrote this week-end
14:09:13 <bauzas> I'll provide a follow-up on it
14:09:22 <edleafe> link?
14:09:31 <bauzas> but the fact is that we need to optionnally call resources given the existing filters
14:09:48 <bauzas> #link http://lists.openstack.org/pipermail/openstack-dev/2017-January/110661.html
14:10:00 <bauzas> the WIP change is updated consequently
14:10:16 <edleafe> ok, I was offline most of the weeked, so I'll take a look at that this morning
14:11:04 <cdent> After talking about it with bauzas in IRC, I ended up thinking the idea was sane
14:11:23 <cdent> (even if the behavior by the deployers was weird to me)
14:12:10 <edleafe> Finally, my series removing Allocation.create() and validating allocation amounts:
14:12:13 <edleafe> #link https://review.openstack.org/#/c/416751/
14:12:16 <edleafe> There is a comment from johnthetubaguy about object versioning, but I agree that a major bump is more or less irrelevant for non-remotable objects.
14:12:19 <edleafe> Comments? Opinions?
14:12:21 <edleafe> Please add them to the review
14:12:55 <edleafe> Just need some +2s on the removal of create() - the second patch is already +W'd
14:13:39 <edleafe> Any other specs or reviews anyone wants to discuss?
14:13:48 <_gryf> yup
14:13:56 <_gryf> #link https://review.openstack.org/#/c/422780
14:13:58 <diga> edleafe: https://review.openstack.org/#/c/423872/
14:14:05 <_gryf> it's a bugifx
14:14:49 <diga> #link https://review.openstack.org/#/c/423872/ - spec on placement notification
14:14:51 * alex_xu updates the poc for traits, both for server and client side https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/resource-provider-tags but it isn't the focus for now
14:14:52 <edleafe> _gryf: ok - anything controversial, or is it pretty clear?
14:15:10 <_gryf> edleafe, it's straightforward
14:15:18 <edleafe> _gryf: cool
14:15:39 <_gryf> only eyes and +2s needed (jay already gave hes +2)
14:15:54 <edleafe> diga: specs for pike will probably get ignored for a while due to upcoming Feature Freeze
14:16:08 <diga> edleafe: Okay
14:16:19 <edleafe> alex_xu: same with traits
14:16:37 <edleafe> Any other reviews to mention?
14:17:16 <diga> edleafe: will continue to work on it so that it will be ready for pike cycle :)
14:17:45 <edleafe> diga: that's fine
14:17:53 <edleafe> Let's move on
14:17:56 <edleafe> #topic Bugs
14:17:57 <diga> edleafe: ok
14:18:13 <edleafe> Placement Bugs:
14:18:13 <edleafe> #link https://bugs.launchpad.net/nova/+bugs?field.tag=placement
14:18:13 <edleafe> Scheduler Bugs:
14:18:13 <edleafe> #link https://bugs.launchpad.net/nova/+bugs?field.tag=scheduler
14:18:49 <edleafe> Nothing new jumps out at me from there
14:18:57 <edleafe> But bug fixes are always welcome!
14:19:30 <edleafe> Any more on bugs?
14:20:10 <edleafe> moving on...
14:20:12 <edleafe> #topic Open discussion
14:20:29 <edleafe> Speak now, or forever hold...
14:20:30 <jroll> ohai
14:20:38 <edleafe> ohai yourself
14:20:48 <jroll> just a quick thing, I've been distracted and haven't been following this stuff super close
14:20:58 <jroll> I'm working on CI to run ironic with the custom resource class stuff
14:20:58 * cdent gives cheeseburger to jroll
14:21:12 <jroll> (and docs etc etc)
14:21:29 <jroll> a question came up, will it be required for folks to be using custom resource classes things in ocata for a smooth upgrade to pike?
14:21:41 <jroll> (e.g. are we going to drop old-style scheduling for ironic in pike)
14:21:48 <jroll> s/e.g./i.e./
14:22:17 <edleafe> jroll: as far as I know, yes, it will not be an optional thing
14:22:30 <cdent> jroll: that's a good question. The current code allows for both but expresses a hope for pike for it not be both
14:22:45 <edleafe> we're trying to remove the never-ending optional stuff, because otherwise we never move forward
14:22:49 <jroll> so... "probably"
14:22:51 <jroll> :)
14:22:52 <cdent> If I remember right there's a few todos that say something like "we can stop doing this in pike"
14:23:04 <jroll> I believe that answers my question
14:23:12 <edleafe> jroll: if we had our way it would not be optional
14:23:31 <edleafe> jroll: but there are always others who bring up "oh, it will be too hard"
14:23:50 <jroll> edleafe: I won't be. just wondering if our docs say "you should do this" or "you must do this"
14:24:00 <jroll> and tripleo folks were asking if they should do the thing
14:24:02 <jroll> etc
14:24:08 <edleafe> +1 on "must"
14:24:36 <cdent> I'm +1 on must too, but that's probably not universal
14:24:38 <edleafe> jroll: but unfortunately I'm not the release god
14:24:46 <jroll> yar
14:24:48 * cdent adjusts his perceptions
14:24:49 <jroll> thanks cdent and edleafe
14:25:56 <edleafe> So... anything else to discuss, or shoudl we all get back to our day?
14:26:04 <cdent> this is the most important part of my day!
14:26:04 <alex_xu> edleafe: cdent may I ask what is purpose of generation?
14:26:20 <edleafe> confllict checking
14:26:20 <cdent> alex_xu: it is a compare and swap marker
14:26:32 * edleafe can't type all of a sudden
14:26:42 * cdent has given edleafe his disease
14:27:00 <alex_xu> cdent: for the client side? when I write the traits client side, I found it just try to keep the generation update-to-date
14:27:16 <cdent> the client side should not be changing the generation
14:27:19 <edleafe> alex_xu: So if you have a view of resources, and someone changes it before you make your change, yours will fail
14:27:28 <cdent> it should use the generation the server tells it
14:27:32 <jroll> generation ~= etag?
14:27:38 <cdent> jroll: effectively yes
14:27:57 <cdent> but etags were considered sort of at the wrong layer
14:27:59 <alex_xu> do we have case for update inventory or traits from two client at sametime?
14:28:00 <jroll> nod
14:28:29 <edleafe> alex_xu: first update wins; second update fails
14:28:44 <cdent> alex_xu: a test case? not real ones, but there are some faked examples of providing bad generations, or changing generations in the backend
14:29:25 <alex_xu> cdent: the real one, for rp which for compute node, there is only one client to report that
14:30:26 <cdent> ?
14:30:49 <edleafe> alex_xu: it's part of the RP object (and database record)
14:31:10 <diga> need to go home, leaving
14:31:25 * edleafe waves to diga
14:31:36 <diga> :)
14:31:57 <alex_xu> edleafe: cdent emm...let me catch you guys later. I need to construct my word first. avoid to waste everybody's time
14:32:27 <edleafe> alex_xu: Sure. If you can give an example of your concern, that would help too
14:32:44 <alex_xu> edleafe: got it, thanks
14:32:59 * edleafe admires anyone who can speak more than one language
14:33:10 <alex_xu> :)
14:33:45 <edleafe> OK, anything else for Opens?
14:34:35 * _gryf have no other things to discuss
14:35:14 <edleafe> Thanks everyone! Now go back to work/play/sleep!
14:35:14 <edleafe> #endmeeting