14:01:28 <bauzas> #startmeeting nova_scheduler
14:01:28 <openstack> Meeting started Mon Nov 16 14:01:28 2015 UTC and is due to finish in 60 minutes.  The chair is bauzas. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:01:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:01:31 <openstack> The meeting name has been set to 'nova_scheduler'
14:02:10 <edleafe> o/
14:02:16 <sudipto> o/
14:02:22 <PaulMurray> o/
14:02:37 * PaulMurray synchronized swimming ^^
14:02:56 <bauzas> nice
14:03:01 <markus_z> o/
14:03:03 <bauzas> lxsli: around ?
14:03:28 <PaulMurray> hold on
14:03:36 <bauzas> jaypipes: hola ?
14:03:36 <PaulMurray> I'll nudge him
14:03:56 <lxsli> o/
14:03:59 <bauzas> well, we can start, just trying to see who's around
14:04:03 <bauzas> okay, we can start
14:04:07 <lxsli> congratulations on core bauzas!
14:04:13 <bauzas> shhhhhtttttttttttttttt
14:04:31 * johnthetubaguy is lurking with intent
14:04:40 <edleafe> \o/
14:04:41 <bauzas> okay, moving on
14:04:48 <bauzas> #topic last actions
14:05:05 <bauzas> so, that shouldn't take too much time but...
14:05:08 <bauzas> http://eavesdrop.openstack.org/meetings/nova_scheduler/2015/nova_scheduler.2015-11-09-14.00.html
14:05:39 <bauzas> - bauzas to update https://etherpad.openstack.org/p/mitaka-nova-spec-review-tracking
14:05:50 <bauzas> well, that has been done by someone else, so thanks to him :)
14:06:23 <bauzas> for the moment, we are pretty gathering all stuff, so in case you want to drop something for reviewing, you know where to go
14:07:03 <bauzas> but we can talk on that right after
14:07:06 <bauzas> - bauzas to open a low-hanging-fruit for using stevedore with scheduler driver
14:07:21 <bauzas> that's done, see https://etherpad.openstack.org/p/nova-low-hanging-fruit
14:08:01 <bauzas> in case anyone wants to step up, he can sneak peek on https://blueprints.launchpad.net/nova/+spec/scheduler-driver-use-stevedore
14:08:20 <bauzas> johnthetubaguy: I guess I should mention it for a specless BP approval ?
14:08:40 <lxsli> A Chinese person offered to work on that I think? On the ML
14:08:47 <Yingxin> bauzas: I'm trying to implement low-hanging-fruit?
14:08:53 <Yingxin> It's me
14:08:58 <lxsli> Hi :D
14:09:03 <Yingxin> o/
14:09:17 <johnthetubaguy> bauzas: yeah, something for that spec etherpad, I can get that approved
14:09:36 <bauzas> johnthetubaguy: I updated https://etherpad.openstack.org/p/mitaka-nova-spec-review-tracking to include it in the sched section
14:10:07 <bauzas> Yingxin: okay, you could ping me offline your launchpad ID, I can put your name on it
14:10:20 <bauzas> Yingxin: thanks for helping
14:10:28 <Yingxin> ok np
14:11:00 <bauzas> ok, moving on
14:11:11 <bauzas> - bauzas to write an ML thread about how to step up for the scheduler bits
14:11:17 <bauzas> that's done
14:11:39 <bauzas> #link http://lists.openstack.org/pipermail/openstack-dev/2015-November/078888.html
14:11:43 <bauzas> voila
14:11:49 <bauzas> that's it for the last actions
14:11:55 <bauzas> any questions so far ?
14:12:15 <bauzas> guess no, moving on
14:12:23 <bauzas> #topic bugs
14:12:32 <bauzas> markus_z: your call
14:12:46 <markus_z> right, thanks
14:12:55 <jaypipes> sorry guys, still on calls... and it's 6am here.. need coffee :(
14:13:11 <markus_z> so, it would be great if you could have a look at these new/incomplete bugs: https://bugs.launchpad.net/nova/+bugs?field.status%3Alist=NEW&field.status%3Alist=INCOMPLETE_WITH_RESPONSE&field.status%3Alist=INCOMPLETE_WITHOUT_RESPONSE&field.tag=scheduler
14:13:12 <jaypipes> bauzas: I will get updated resource-providers and pci-generate-stats specs up today.
14:13:30 <bauzas> jaypipes: no worries, life is short and you can read logs :)
14:14:09 <markus_z> These 3 bugs need more eyes if they are valid or not.
14:14:16 <bauzas> markus_z: ouch, pretty old bugs for at least https://bugs.launchpad.net/nova/+bug/1477098
14:14:16 <openstack> Launchpad bug 1477098 in OpenStack Compute (nova) "When you unshelve an instance, AggregateImagePropertiesIsolation filter doesn't work." [Undecided,New]
14:14:37 <markus_z> yepp
14:14:44 <bauzas> I can see one of them being incomplete
14:14:55 <bauzas> https://bugs.launchpad.net/nova/+bug/1498367
14:14:56 <openstack> Launchpad bug 1498367 in OpenStack Compute (nova) "Invalid host found when nova boot" [Undecided,Incomplete]
14:15:19 <markus_z> yes, I left a comment when I did a triage back then, but there was no response.
14:15:41 <markus_z> IMO it's fair to let this one expire.
14:15:46 <bauzas> markus_z: agreed
14:16:01 <edleafe> probably fixed already in liberty, no?
14:16:09 <edleafe> at least the logging info part
14:16:16 <bauzas> markus_z: okay, I'll dig into https://bugs.launchpad.net/nova/+bug/1477098 and https://bugs.launchpad.net/nova/+bug/1510395
14:16:18 <openstack> Launchpad bug 1477098 in OpenStack Compute (nova) "When you unshelve an instance, AggregateImagePropertiesIsolation filter doesn't work." [Undecided,New]
14:16:19 <openstack> Launchpad bug 1510395 in OpenStack Compute (nova) "resize vm across azs" [Undecided,New]
14:16:28 <markus_z> bauzas: thanks!
14:16:31 <bauzas> #action all to bug scrub https://bugs.launchpad.net/nova/+bug/1477098 and https://bugs.launchpad.net/nova/+bug/1510395
14:16:41 <markus_z> That's all from my side
14:16:54 <bauzas> edleafe: probably, but we can leave the expiry do its work :)
14:17:19 <bauzas> markus_z: cool, thanks for helping us with the bug triage, really awesome
14:17:30 <bauzas> #topic blueprints and specs
14:18:15 <bauzas> even if we have been identified as a priority, my biggest concern is still to try to get all the main specs approved by M-1
14:18:41 <bauzas> not really because of the deadline, but rather because the Code Freeze will be like 2 months from that time
14:18:59 <bauzas> so, my question is
14:19:26 <bauzas> except the resource-providers BP that I'll mention later, do we have any other BP that requires review ?
14:19:55 <edleafe> Can we talk about https://blueprints.launchpad.net/nova/+spec/configurable-resource-tracker
14:19:55 <edleafe> johnthetubaguy asked for a spec, but I want to make sure that everyone's on board with this]
14:20:01 <edleafe> Code-wise it will be a fairly trivial change
14:20:30 <bauzas> edleafe: there is a big operator impact, so I'm +1 with mriedem's -2
14:20:44 <edleafe> the -2 was simply procedural
14:20:54 <edleafe> because the bp isn't approved
14:20:57 <bauzas> edleafe: yup
14:21:01 <bauzas> edleafe: that's my point
14:21:02 <edleafe> which is why I want to discuss the bp
14:21:10 <bauzas> edleafe: I'm +1 with asking for a spec
14:21:26 <edleafe> sure, I can write a spec for it
14:22:00 <edleafe> but imo, you can't have a pluggable scheduler without it
14:22:10 <bauzas> edleafe: even if that sounds a trivial change, it means that we'd be enabling a new entrypoint, and I'd love to see some operator's feedback
14:22:19 <edleafe> sure
14:22:29 <edleafe> of course, defaults would remain unchanged
14:22:48 <bauzas> edleafe: that I understood
14:24:10 <bauzas> edleafe: I'd pretty be concerned by what could be the pluggable mechanism, given that there is a 2-side mechanism with the Scheduler's HostManager
14:24:11 <edleafe> #action edleafe to write spec for https://blueprints.launchpad.net/nova/+spec/configurable-resource-tracker
14:24:26 <bauzas> not sure you can put an action on your name
14:24:28 <edleafe> bauzas: the host manager is pluggable
14:24:40 <edleafe> the only part of the process that isn't is the RT
14:24:48 <edleafe> that's the part that writes the data
14:25:20 <edleafe> without being able to vary that, you can't use any alternate way of passing resource info to the scheduler
14:25:20 <bauzas> edleafe: you know that we're working towards removing the Ironic HostManager, right?
14:25:32 <edleafe> bauzas: sure, but for different reasons
14:25:48 <edleafe> because the model wasn't right
14:26:24 <bauzas> edleafe: agreed, but you see that I could be concerned by what is the plug and how
14:26:42 <edleafe> bauzas: as you should be
14:26:55 <bauzas> #action edleafe to write spec for https://blueprints.launchpad.net/nova/+spec/configurable-resource-tracker
14:27:03 <edleafe> bauzas: but there really cannot be anything to evaluate until it's possible
14:27:27 <bauzas> edleafe: okay, let's defer that discussion up to the next-to-be spec :)
14:27:38 <edleafe> cool
14:27:56 <bauzas> so, back to the resource-providers BP
14:28:32 <bauzas> #link https://review.openstack.org/#/c/225546/
14:28:50 <bauzas> jaypipes: do you want to address some concerns now, or are you planning to update the spec directly?
14:29:29 <jaypipes> bauzas: update spec directly. all of your feedback has been very valuable. just need to get the updates done (especially on the REST API impact and solving the numa topology field thing.
14:29:43 <bauzas> jaypipes: ack, wfm
14:30:02 <bauzas> okay, we can then move on ?
14:30:39 <jaypipes> sure
14:31:17 <bauzas> #topic open discussion
14:31:32 <bauzas> nothing from my side
14:32:10 <bauzas> well, except that I'd love to see how we could pinpoint any trivial change we'd like to address
14:32:27 <edleafe> like what?
14:32:52 <bauzas> some people expressed interest in contributing to the scheduler codebase, and we know that there are some stuff that could be trivially fixed
14:33:21 <edleafe> sure, that would be great
14:33:26 <bauzas> so I'm thinking of some way to write down my thoughts so it could be a base document for working on
14:33:38 <edleafe> just wondering what fixes you had in mind
14:33:56 <bauzas> for example, how we load the filters
14:34:09 <bauzas> we're still using nova.loadables thru the nova.filters layer
14:34:40 <edleafe> yeah, that would be a good clean-up
14:34:41 <bauzas> we're also missing some functional test for all filtres
14:35:03 <bauzas> that, I'm thinking of writing the foundational bits for that, so anyone could contribute
14:35:48 <Yingxin> would like to see that :)
14:36:08 <bauzas> documentation about filters and weighters is also highly unsufficient
14:36:21 <bauzas> I like what the api subteam does by iterating on docs
14:36:54 <bauzas> for the moment, the only stuff we have is that one-pager http://docs.openstack.org/developer/nova/filter_scheduler.html
14:37:17 <bauzas> that's just a docstring for each filter, which doesn't really provide the bits
14:37:42 <bauzas> like, good luck with using the MetricsWeigher if you don't read code :)
14:38:43 <bauzas> #action bauzas to write his thoughts on low-hanging-fruits for Mitaka (filters docs, loadables, functional coverage)
14:40:12 <bauzas> okay, sounds a massive plebicite :)
14:40:33 <bauzas> any other things to share/complain/yell/rant about?
14:41:32 <bauzas> fair enough, either I faced a netsplit or I'm more powerful than melatonine
14:41:40 <edleafe> or both :)
14:42:15 <bauzas> in case anyone suffers from sleeping troubles, lemme know
14:42:26 <bauzas> calling it a wrap :)
14:42:31 <bauzas> #endmeeting