14:00:36 <edleafe> #startmeeting nova_scheduler 14:00:36 <openstack> Meeting started Mon Aug 1 14:00:36 2016 UTC and is due to finish in 60 minutes. The chair is edleafe. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:37 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:39 <openstack> The meeting name has been set to 'nova_scheduler' 14:00:45 <edleafe> Who's here today? 14:00:46 <_gryf> o/ 14:00:55 <alex_xu> \o 14:00:57 <Yingxin> o/ 14:01:05 <macsz> O/ 14:01:24 <cdent> I'm sort of nearby 14:01:37 <edleafe> cdent: I know that feeling :) 14:02:32 <edleafe> Let's give everyone another minute... 14:02:49 <jaypipes> o/ 14:02:52 <edleafe> I know that bauzas_off is on PTO this week and next 14:02:55 <jaypipes> morning everyone. 14:03:04 <edleafe> Hey Hey Jay 14:03:29 <edleafe> OK, let's get started 14:03:34 <edleafe> #topic Mid-cycle recap 14:03:46 <edleafe> I didn't know how to summarize this 14:03:56 <edleafe> so I figured that questions might be best 14:04:29 <edleafe> We did agree on the resource provider tags for handling qualitative aspects of a request 14:04:52 <alex_xu> \o/ 14:05:08 <edleafe> And that we would continue to abuse^H^H^H use flavor extra-specs to present them to the user 14:05:21 <jaypipes> yay netsplits. 14:05:42 <alaski> o/ 14:05:43 <_gryf> edleafe, so the extra specs will not go away? 14:05:55 <edleafe> _gryf: probably not in our lifetime 14:05:57 <edleafe> :) 14:05:57 <jaypipes> edleafe: can you repeat the last few things. I think a number of netsplits just occurred. 14:06:02 <_gryf> oh my… 14:06:38 <edleafe> jaypipes: ok 14:06:49 <edleafe> 09:04 < edleafe> We did agree on the resource provider tags for handling qualitative aspects of a request 14:06:52 <edleafe> 09:04 < alex_xu>| \o/ 14:06:54 <edleafe> 09:05 < edleafe> And that we would continue to abuse^H^H^H use flavor extra-specs to present them to the user 14:06:57 <edleafe> 09:05 < jaypipes>| yay netsplits. 14:07:00 <edleafe> 09:05 < alaski>| o/ 14:07:02 <edleafe> 09:05 < _gryf> edleafe, so the extra specs will not go away? 14:07:05 <edleafe> 09:05 < edleafe> _gryf: probably not in our lifetime 14:07:07 <edleafe> 09:05 < edleafe> :) 14:07:10 <edleafe> jaypipes: did you get that ok? 14:07:45 <jaypipes> edleafe: yeah. didn't agree on the specifics yet (doing a review of the spec now actually) but the overall proposal was ++ 14:08:23 <alex_xu> edleafe: I thought the require prefer tags can be added as atrribute of flavor, then leave the extra_spec alone, maybe keep existed extra_spec and freeze to add more? 14:08:27 <edleafe> jaypipes: exactly. We agreed that tags were the way to go, and that we should proceed along those lines 14:09:14 <edleafe> alex_xu: not sure that I follow. How would you add a flavor attribute? 14:09:15 <alex_xu> jaypipes: thanks :) 14:09:53 <alex_xu> edleafe: i mean flavor.required = [....a set of tags ...] 14:10:26 <Yingxin> edleafe: alex_xu: we may need to draft another spec for it 14:11:24 <edleafe> alex_xu: so would it be a json blob in the db? 14:11:41 <edleafe> alex_xu: *another* json blob? 14:12:08 <_gryf> edleafe, it might be one-to-many relation… 14:12:32 <alex_xu> edleafe: i think it isn't json blob, I remember jaypipes write a db scheme in the maillist 14:13:33 <edleafe> alex_xu: well, I generally don't like stuffing multiple values into a string and then relying on parsing to extract them 14:14:09 <edleafe> hence my preference for tag attributes over big_long_honking_names_with_many_meanings 14:14:35 <edleafe> so I guess I'd have to see a spec on how that would work. 14:14:53 <Yingxin> edleafe: http://lists.openstack.org/pipermail/openstack-dev/2016-July/099675.html fyi 14:15:11 <Yingxin> the table flavor_tags 14:15:36 <edleafe> But in any case, the notion that we could separate the qualtative aspects of the request from the flavor was not possible 14:16:20 <edleafe> Yingxin: thanks 14:16:28 <alex_xu> Yingxin: thanks for the link 14:16:32 <alex_xu> edleafe: yea 14:16:36 <edleafe> Yingxin: I see, it's sort of the way extra_specs *should* have been done :) 14:17:13 <edleafe> FYI, I abandoned my spec for separating out qualitative request items from flavor 14:17:32 <edleafe> since it will probably be at least two more cycles before we can consider it 14:18:25 <edleafe> Anything else from the midcycle that we should cover? 14:18:51 <alex_xu> edleafe: library os-capabilities? 14:19:05 <edleafe> alex_xu: ah yes! thanks 14:19:18 <edleafe> alex_xu: do you want to summarize? 14:20:14 <alex_xu> edleafe: I'm not sure I clear about that, I only know it is libraty to define and catalog the capabilities, probably a set definitions for tags. 14:21:04 <jaypipes> sorry, netshits... 14:21:05 <edleafe> alex_xu: it was also to provide a standard interface for reporting capabilities to the placement api, right? 14:21:09 <alex_xu> and there should be some convenient function like mapping the existed extra_spec(which is capabilities) to capabilities tags 14:21:56 <alex_xu> edleafe: emm...i'm not sure, maybe jaypipes have better explain 14:22:32 * edleafe send jaypipes some network connectivity karma 14:23:16 <edleafe> Well, we can discuss the particulars when it comes time to start designing it. 14:23:21 <edleafe> Let's move on for now 14:23:25 <Yingxin> is it going to cover Standardize capabilities using Enums? 14:23:49 <Yingxin> the standardize part? 14:24:14 <edleafe> Yingxin: probably not through Enums, but standardize on the format/naming for things 14:24:38 <Yingxin> edleafe: thanks 14:24:52 <edleafe> #topic Placement API 14:25:04 <edleafe> #link https://review.openstack.org/#/c/329149 14:25:13 <edleafe> ^ cdent's spec 14:25:40 <jaypipes> edleafe: merged. 14:25:55 <edleafe> jaypipes: yup 14:26:22 <edleafe> but it's been a few weeks since our last meeting, and wanted to make sure everyone was familiar with it 14:27:40 <edleafe> There are also some follow-up patches dealing with the separate placement db issues 14:28:41 <edleafe> #topic Generic Resource Pools 14:28:50 <jaypipes> edleafe: yup, will be reviewing those today as soon as I finish review on alex_xu's tags spec. 14:28:56 <edleafe> Looks like the current stack of patches begins with: 14:28:59 <edleafe> #link https://review.openstack.org/#/c/334031 14:29:52 <edleafe> jaypipes: anything to add about that series? 14:31:13 <jaypipes> edleafe: no, just that reviews on it would be very useful. 14:31:48 <edleafe> got it 14:31:58 <edleafe> Next up 14:32:10 <edleafe> #topic Populate Allocation Fields 14:32:22 <edleafe> #link https://review.openstack.org/#/c/300177 14:32:30 <jaypipes> _gryf poo-poo'd my spec :( 14:32:46 <_gryf> jaypipes, sorry… 14:32:51 <jaypipes> _gryf: just kiddin' :) 14:32:57 <_gryf> jaypipes, i know ;) 14:33:04 <jaypipes> _gryf: I'll fix those typos and push another. 14:33:18 <jaypipes> _gryf: and answer your query about the pci devs 14:33:53 <_gryf> jaypipes, ok! 14:34:10 <edleafe> Yeah, that spec looks very close 14:35:26 <edleafe> Nobody posted any items to review, so... 14:35:30 <edleafe> #topic Opens 14:35:38 <edleafe> Anyone have anything to discuss? 14:36:44 <Yingxin> I submitted a spec related to scheduler performance improvement 14:36:52 <Yingxin> https://review.openstack.org/#/c/345874/ 14:37:25 <edleafe> #link https://review.openstack.org/#/c/345874/ 14:37:44 <Yingxin> if we continue need to use in-memory python filtering and weighing, it is worthwhile to look at 14:37:56 <edleafe> Yingxin: thanks, I haven't looked at that yet. Added to my queue 14:38:18 <Yingxin> thanks ed 14:38:18 <alex_xu> Yingxin: cool, added to my queue also 14:38:41 <Yingxin> alex_xu: \o/ 14:39:07 <edleafe> Anything else? Or should we get back to work? 14:39:09 <edleafe> :) 14:39:28 <alex_xu> or get back to sleep :) 14:39:43 <edleafe> alex_xu: Heh, forgot about timezones :) 14:40:05 <edleafe> OK, everyone, back to work/sleep!! 14:40:11 <edleafe> #endmeeting