14:00:11 <edleafe> #startmeeting nova_scheduler
14:00:12 <openstack> Meeting started Mon Apr 16 14:00:11 2018 UTC and is due to finish in 60 minutes.  The chair is edleafe. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:16 <openstack> The meeting name has been set to 'nova_scheduler'
14:00:19 <jaypipes> o/
14:00:20 <efried> ō/
14:00:20 <mriedem> o/
14:00:20 <tssurya> o/
14:00:20 <takashin> o/
14:00:29 <edleafe> #link Agenda: https://wiki.openstack.org/wiki/Meetings/NovaScheduler#Agenda_for_next_meeting
14:00:33 <cdent> hello
14:01:19 <edleafe> Hope that everyone had an enjoyable weekend
14:01:50 <jaypipes> icing my elbow was just ... awesome.
14:02:06 <edleafe> oooh, sounds like fun!
14:02:43 <edleafe> #topic Specs
14:03:03 <edleafe> Once again I'll bulk list them here (courtesy of cdent)
14:03:05 <edleafe> #link VMware: place instances on resource pool https://review.openstack.org/#/c/549067/
14:03:08 <edleafe> #link Proposes NUMA topology with RPs https://review.openstack.org/#/c/552924/
14:03:11 <edleafe> #link Account for host agg allocation ratio in placement https://review.openstack.org/#/c/544683/
14:03:14 <edleafe> #link Spec for isolating configuration of placement database https://review.openstack.org/#/c/552927/
14:03:18 <edleafe> #link Support default allocation ratios https://review.openstack.org/#/c/552105/
14:03:20 <edleafe> #link Spec on preemptible servers https://review.openstack.org/#/c/438640/
14:03:23 <edleafe> #link Handle nested providers for allocation candidates https://review.openstack.org/#/c/556873/
14:03:27 <edleafe> #link Proposes Multiple GPU types https://review.openstack.org/#/c/557065/
14:03:29 <edleafe> #link Standardize CPU resource tracking https://review.openstack.org/#/c/555081/
14:03:32 <edleafe> #link Network bandwidth resource provider https://review.openstack.org/#/c/502306/
14:03:36 <edleafe> #link Propose counting quota usage from placement https://review.openstack.org/#/c/509042/
14:03:38 <edleafe> #link Add history behind nullable project_id and user_id https://review.openstack.org/#/c/560174/
14:03:42 <edleafe> #link Return resources of entire trees in Placement https://review.openstack.org/#/c/559466/
14:03:45 <edleafe> #link Numbered request groups use different providers https://review.openstack.org/#/c/560974/
14:03:48 <edleafe> Are there any specs anyone would like to discuss?
14:04:33 <cdent> only to reiterate: I'd really like to see " Spec for isolating configuration of placement database" merge.
14:04:50 <cdent> it doesn't have to mean anything changes, it just means things _can_ change
14:05:02 <bauzas> \o (even if in another private corp discussion too)
14:05:25 <edleafe> Anything else on specs?
14:05:34 <efried> jaypipes seems like a good second +2 on that spec
14:05:55 <efried> watch out for that bus overhead, jaypipes
14:05:56 <jaypipes> efried: I will review it.
14:06:22 <jaypipes> efried: you are referring to cdent's placement db spec, yes?
14:06:31 <efried> jaypipes: https://review.openstack.org/#/c/552927/
14:06:44 <jaypipes> ya
14:06:46 <jaypipes> on it
14:06:50 <efried> thanks
14:06:52 <edleafe> cool
14:07:08 <edleafe> Next up:
14:07:11 <edleafe> #topic Reviews
14:07:39 <edleafe> Keeping with our decision last week to use the placement update as our priority tracker, here are the main themes:
14:07:52 <edleafe> #link Update Provider Tree https://review.openstack.org/#/q/topic:bp/update-provider-tree
14:07:55 <edleafe> #link Nested providers https://review.openstack.org/#/q/topic:bp/nested-resource-providers
14:07:58 <edleafe> #link Nested providers in allocation candidates https://review.openstack.org/#/q/topic:bp/nested-resource-providers-allocation-candidates
14:08:01 <edleafe> #link Mirror nova host aggregates to placement https://review.openstack.org/#/q/topic:bp/placement-mirror-host-aggregates
14:08:04 <edleafe> #link Forbidden Traits https://review.openstack.org/#/q/topic:bp/placement-forbidden-traits
14:08:07 <edleafe> #link Consumer Generations https://review.openstack.org/#/q/topic:bp/add-consumer-generation
14:08:10 <edleafe> # Extraction https://review.openstack.org/#/q/topic:bp/placement-extract
14:08:41 <edleafe> I am currently working on the consumer_generation stuff. I have some POC for using the Consumer object posted, but still need to add tests
14:09:10 <efried> Granular should be on that list.
14:09:26 <efried> I'm working on that - very sparsely, desperately trying to make time for it.
14:09:47 <edleafe> efried: want to add a #link?
14:09:55 <efried> https://review.openstack.org/#/q/topic:bp/granular-resource-requests+status:open
14:10:02 <efried> Or if you prefer
14:10:08 <efried> #link Granular Resource Requests https://review.openstack.org/#/q/topic:bp/granular-resource-requests+status:open
14:10:14 <edleafe> much better
14:10:49 <cdent> I'm happy to add Granular as a theme on the weekly report, but that list of main themes was a result of some post-ptg summaries. Have we declared granular a proper (whatever that means) theme? (we probably should, but just checking to be sure)
14:10:50 <edleafe> cdent: please add that on the next update. I'll update the meeting page
14:11:15 <edleafe> It seems pretty critical for nrp to work
14:11:44 <efried> It's been a central assumption of pretty much every placement-using spec we've got approved or proposed for Rocky and beyond.
14:12:09 * cdent nods
14:12:54 <cdent> I think mentally I just had it lumped under nested, but can make it standalone for clarity
14:13:25 <cdent> there's room with forbidden already merged
14:13:30 <efried> :)
14:13:31 <cdent> (thanks for that, btw)
14:14:04 <efried> has anyone considered how we're going to consume that from the nova side?
14:14:18 <bauzas> consuming what?
14:14:22 <bauzas> request groups ?
14:14:39 <bauzas> or forbidden traits ?
14:14:58 <bauzas> if the former, my own spec tries to use it
14:15:32 <efried> I saw a spec somewhere that mentioned using forbidden traits too, but I think it was tangential.  Do we have a real use case for it yet?
14:16:08 <cdent> the ptg sure made it seem so
14:16:31 <cdent> but don't know how it's flowed since
14:16:36 <efried> It was an ironic thing, maybe?
14:16:37 <edleafe> efried: one of the suggested use cases was a flavor where the instance should not land on an SSD host
14:16:55 <efried> okay, so if any of our use cases involve "flavor" then we need a new spec.
14:17:05 <edleafe> why?
14:17:22 <efried> because right now we don't have anything specified that translates flavor syntax to placement syntax.
14:17:32 <efried> The work cdent did was just on the placement side.
14:17:42 <efried> We never decided how the flavor would look
14:17:50 <efried> Whether it was going to be with ! or =forbidden or what.
14:18:05 <edleafe> I thought we had settled on !
14:18:09 <efried> and have nothing lined up to decide that, or implement it once decided.
14:18:14 <efried> We settled on ! for the placement side, yes.
14:18:19 <cdent> I thought we decided =forbidden in flavors and ! on the placement side
14:18:23 <efried> But we never even discussed the flavor side, to my recollection.
14:18:25 <efried> See?  :P
14:18:26 <cdent> different because the language was already different
14:18:49 <cdent> mostly because extra specs are already kinda weird
14:18:51 <efried> But even if we did "already decide", there's nothing lined up to actually implement it.
14:18:54 <efried> so
14:18:59 <edleafe> You may be right, but that wasn't my recollection
14:19:02 <efried> do we need to add THAT for rocky?
14:19:03 <cdent> trait:FOO={required,forbidden}
14:19:21 <jaypipes> efried: what cdent says ^^
14:19:25 <cdent> it's small thing
14:19:34 <cdent> adding great power
14:19:45 <cdent> if nobody else has the time/cycles I can do it
14:19:54 <jaypipes> I thought we'd settled on required:FOO=forbidden?
14:20:12 <edleafe> cdent: so something like: required=foo&forbidden=bar ?
14:20:13 <efried> flavor trait:FOO=forbidden translates to placement qstring ?required=!FOO
14:20:18 <jaypipes> w.r.t to the flavor extra spec format.
14:20:32 <efried> edleafe: No, the placement syntax is definitely already settled.
14:20:33 <jaypipes> right, sorry trait:FOO=forbidden
14:20:45 <efried> So let's not have this design discussion here.
14:20:46 <cdent> efried: yes, that translation
14:20:52 <efried> Cause we need to write a spec for it anyway.
14:21:05 <jaypipes> efried: haven't we already had the design discussion on this?
14:21:09 <cdent> we have
14:21:18 <cdent> and I'm not sure we need a spec, maybe just a blueprint
14:21:19 <jaypipes> efried: it should be trait:FOO=forbidden.
14:21:20 <efried> jaypipes: Informally, verbally, but to my recollection never completely landed.
14:21:34 <cdent> becquse all we're doing is implementation a translation of extra spec to query string
14:21:35 <efried> I'm fine with that.  Needs a spec regardless, so let's do that there.
14:21:43 <efried> but
14:21:51 <efried> do we have room for it in rocky?
14:22:15 <efried> and/or is it important enough to prioritize over other things we would rather enslave^W task cdent with?
14:22:39 <jaypipes> I don't see a) why this needs another spec (it's an implementation detail) and b) why we can't do this in Rocky
14:22:50 <edleafe> It would seem to be a low priority, though
14:23:16 <efried> needs spec because has all the impacts that the spec template was written to make us address.
14:23:27 <efried> and because we need it written down how it's going to work.
14:23:44 <efried> that's my take, anyway.
14:24:21 * jaypipes burned out on writing specs... someone else can throw one up.
14:24:50 <cdent> like I said I'll take this action if we say it should be done. I don't agree that it needs a spec, but if people think it does, I'm happy to do that part too.
14:24:59 <edleafe> When the spec creation/discussion takes longer than the coding would...
14:25:19 <bauzas> btw. I have a spec that I'll probably abandon and just write code
14:25:20 <jaypipes> edleafe: right....
14:26:43 <cdent> i'll take this ball, in whatever form, let's move on
14:26:53 <edleafe> thanks cdent
14:27:12 * alex_xu waves late
14:27:23 <edleafe> Anyone else want to discuss one of these Main Theme reviews?
14:27:32 * edleafe waves back at alex_xu
14:28:40 <edleafe> Coming up: a dump of all the other reviews listed in the placement update:
14:28:43 <edleafe> #link Purge comp_node and res_prvdr records during deletion of cells/hosts https://review.openstack.org/#/c/546660/
14:28:46 <edleafe> #link A huge pile of improvements to osc-placement https://review.openstack.org/#/q/topic:bp/placement-osc-plugin-rocky
14:28:49 <edleafe> #link Add compute capabilities traits (to os-traits) https://review.openstack.org/#/c/546713/
14:28:52 <edleafe> #link General policy sample file for placement https://review.openstack.org/#/c/524425/
14:28:55 <edleafe> #link Provide framework for setting placement error codes https://review.openstack.org/#/c/546177/
14:28:58 <edleafe> #link Get resource provider by uuid or name (osc-placement) https://review.openstack.org/#/c/527791/
14:29:01 <edleafe> #link placement: Make API history doc more consistent https://review.openstack.org/#/c/477478/
14:29:04 <edleafe> #link Handle agg generation conflict in report client https://review.openstack.org/#/c/556669/
14:29:07 <edleafe> #link Remove usage of [placement]os_region_name https://review.openstack.org/#/c/557086/
14:29:10 <edleafe> #link Add unit test for non-placement resize https://review.openstack.org/#/c/537614/
14:29:13 <edleafe> #link Address issues raised in adding member_of to GET /a-c https://review.openstack.org/#/c/554357/
14:29:16 <edleafe> #link cover migration cases with functional tests https://review.openstack.org/#/c/493865/
14:29:19 <edleafe> #link Update check to ensure compute is using placement https://review.openstack.org/#/c/558089/
14:29:22 <edleafe> #link Bug fixes for sharing resource providers https://review.openstack.org/#/q/topic:bug/1732731
14:29:25 <edleafe> #link normalize_name helper (in os-traits) https://review.openstack.org/#/c/560107/
14:29:28 <edleafe> #link Fix issues with unicode uppercasing in normalizing resource classes https://review.openstack.org/#/q/topic:bug/1762789
14:29:31 <edleafe> #link WIP at granular in allocation candidates https://review.openstack.org/#/c/517757/
14:29:34 <edleafe> #link Fix a bug with syncing traits. It can fail, ruining the whole service.  https://review.openstack.org/#/q/topic:bug/1760322
14:30:02 <cdent> as a point of order, since these are "other" can we just leave them to email. repeating them here seems
14:30:25 <cdent> seems an ack that people don't read email or look at their gerrit dashboards, which we should encourage or accept
14:31:05 <edleafe> I add them for the sake of completeness to anyone reading the meeting minutes.
14:31:06 <cdent> let's establish the precedent that people do read and are reviewing, and will raise issues here based on thinks they can link in if necessary
14:31:59 <edleafe> I think that that precedent was established when we switched from one-by-one listing of specs and reviews to the overall listing
14:32:00 <efried> That wfm.  The repetition of the whole list of links makes it more like noise than like actionable
14:32:46 <efried> The comprehensive list was good for a week or two of "whoah, we've got a lot of stuff on the table"
14:34:02 <edleafe> ok, if that's the consensus, next week I'll replace that with link to the most recent placement update email. How does that sound?
14:34:29 <cdent> +1
14:34:48 <efried> +1
14:34:53 <edleafe> kewl
14:35:07 <efried> and of course folks can still link in specific ones they want to discuss in the meeting.
14:35:39 <edleafe> #agreed Stop posting all the links to reviews in the meeting, and replace that with a link to the most recent placement update email
14:36:33 <edleafe> Let's move on
14:36:40 <edleafe> #topic Bugs
14:36:43 <edleafe> #link Placement bugs https://bugs.launchpad.net/nova/+bugs?field.tag=placement&orderby=-id
14:37:03 <edleafe> One new one this week, with backport potential
14:37:43 <edleafe> There's already a fix proposed
14:37:52 <edleafe> Anything else bug-related?
14:38:05 <cdent> no sir
14:38:37 <edleafe> #topic Open Discussion
14:38:47 <edleafe> What's on your respective minds?
14:39:09 <cdent> Spring is late this year.
14:39:41 <edleafe> Not here it ain't
14:40:09 <edleafe> Probably more like your summer - days going up to the 80-90F range
14:40:24 <efried> oh?  It was mid-40s up here in Central TX
14:40:25 <mriedem> we got 8 inches of snow this weekend
14:40:33 <efried> What is this snow of which you speak?
14:40:36 <mriedem> i played xmas music on saturday morning
14:40:38 <efried> Oh, that stuff we had in Dublin.
14:40:52 * cdent considers tracking snow in placement
14:40:58 <cdent> are we done?
14:41:09 <edleafe> with snow? I sure am!
14:41:22 <edleafe> OK, thanks everyone!
14:41:25 <edleafe> #endmeeting