14:00:27 <edleafe> #startmeeting nova_scheduler
14:00:27 <openstack> Meeting started Mon Sep 18 14:00:27 2017 UTC and is due to finish in 60 minutes.  The chair is edleafe. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:29 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:31 <openstack> The meeting name has been set to 'nova_scheduler'
14:00:37 <mriedem> o/ kind of
14:00:41 <edleafe> Good UGT morning!
14:00:51 <edleafe> mriedem: I know the feeling :)
14:00:56 <edleafe> Who else is around?
14:01:26 <ralonsoh> hi
14:01:32 * johnthetubaguy lurks without much intent
14:02:10 * edleafe suspects PTG burnout on a Monday
14:03:24 <gibi> o/
14:03:27 <edleafe> I guess that there isn't much to discuss that wasn't discussed last week.
14:04:34 <ralonsoh> can I ask some questions? if there is no agenda today
14:05:08 <edleafe> ralonsoh: sure
14:05:25 <ralonsoh> about scheduler and placement: is there any plan to move PCI to placement?
14:05:50 <ralonsoh> to remove PCI filter and move all this logic to placement
14:06:22 <edleafe> ralonsoh: yes, and that was the subject of a lot of the discussion at PTG
14:06:27 <johnthetubaguy> who said they were writing up the generic PCI device hander stuff? efried?
14:06:40 <ralonsoh> sorry, I couldn't attend
14:06:48 <ralonsoh> do you have any spec or bp?
14:06:49 <edleafe> It will rely on the nested resource providers stuff being completed
14:06:59 <ralonsoh> edleafe: thanks!
14:07:14 <edleafe> ralonsoh: as johnthetubaguy said, the spec will probably be drawn up in the next few days
14:07:33 <ralonsoh> edleafe: I'll be waiting for it
14:07:41 <mriedem> the generic pci device handler stuff is all post queens at this point
14:07:52 <edleafe> ralonsoh: it will be led by either be efried or jaypipes
14:08:29 <ralonsoh> I'll ping to help them
14:08:35 <edleafe> mriedem: yeah, good point. We need to complete a lot of the structures in Queens in order to make PCI and other complex RPs work in Rocky
14:09:00 <edleafe> Any other questions? Or should we get back to whatever it was we were doing?
14:09:22 <johnthetubaguy> maybe have an ironic question
14:09:33 <edleafe> johnthetubaguy: go for it
14:09:36 <johnthetubaguy> so traits get set on ironic nodes, etc
14:09:46 <johnthetubaguy> the custom ones need to be registered by someone in placement
14:10:00 <johnthetubaguy> I am tempted to say the operator should register those, or is that bad?
14:10:15 <johnthetubaguy> do we want the driver to register any it "sees" instead
14:10:27 <edleafe> no, that's the general approach
14:10:35 <edleafe> so that they are consistently named
14:10:47 <edleafe> since the name has to match the extra_specs key
14:10:53 <johnthetubaguy> yeah, makes sense, just wanted to check I wasn't going off on one
14:10:58 <edleafe> s/key/value
14:11:13 <edleafe> it may be possible for an agent to do that
14:11:21 <johnthetubaguy> I guess the driver can log a warning on not register any unrecognised traits
14:11:35 <edleafe> but it would have to know the trait naming conventions
14:11:57 <edleafe> i.e., what are standard and what are custom
14:12:00 <johnthetubaguy> I think the flow of set it in placement first, and offer validation seems fine
14:12:10 <edleafe> yup
14:12:15 <johnthetubaguy> yeah, worried more about CUSTOM_XXX ones I guess
14:12:32 <johnthetubaguy> anyways, happy to keep it simple for the v1 of the integration
14:12:36 <edleafe> the custom stuff is 100% on the operators
14:13:30 <johnthetubaguy> yeah, totally
14:13:36 <johnthetubaguy> its just in ironic when you set the trait on a node
14:13:36 <edleafe> So... unless there is anything else, we can make this a quick meeting
14:13:48 <johnthetubaguy> I was figuring just ensure its CUSTOM_xxx or a known one
14:14:02 <johnthetubaguy> the virt driver can check with placement for validity, and skip, etc
14:14:15 <johnthetubaguy> (i.e. don't require Ironic to talk to placement to do validation)
14:14:16 <edleafe> exactly
14:14:19 <johnthetubaguy> cools
14:14:24 <johnthetubaguy> I am all done then.
14:14:50 <edleafe> standard traits are also highly cacheable
14:15:04 * johnthetubaguy shrugs and nods
14:15:20 <edleafe> OK, everyone - thanks for playing!
14:15:22 <edleafe> #endmeeting