Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-07-19
16:03:29 mriedem cdent: not sure if you saw my comment on https://review.openstack.org/#/c/484162/
16:03:37 mriedem i tried backporting to ocata and the test needs some changes there
16:03:40 mriedem since we don't have 1.7
16:04:05 mriedem i got past that but there was one more failure in the allocation request which i couldn't dig into
16:04:24 cdent mriedem: i did yeah, but haven’t had a chance to react. if you have some wip code and want me to pick it up I can do that
16:04:41 mriedem i might, se
16:05:06 cdent I also so your comments on the dansmith’s functional test now being a real failure. I’m not sure if I own that by default now or if dan wishes to return to that. either is fine with me.
16:05:12 cdent s/so/saw/
16:06:09 mriedem let's just ignore it for now
16:06:21 mriedem cdent: https://review.openstack.org/485263
16:06:55 cdent cool, will inspect, thanks
16:08:55 mriedem bauzas: it would be good if you could start going through this series https://review.openstack.org/#/c/408955/ - takashin has been really patient with rebases and that's been around a few releases now, but hasn't gotten review
16:10:08 bauzas mriedem: I did it a couple of times but I can do other cycles, for sure
16:10:20 openstackgerrit Merged openstack/nova master: Updated from global requirements https://review.openstack.org/484218
16:11:14 openstackgerrit Sylvain Bauza proposed openstack/nova master: Accept any scheduler driver entrypoint https://review.openstack.org/484828
16:11:25 bauzas mriedem: sean-k-mooney: ^
16:12:06 bauzas mriedem: sean-k-mooney english grammar nits welcome
16:13:50 bauzas jianghuaw: thanks for updating https://review.openstack.org/#/c/450122/20/specs/queens/approved/virt-add-support-for-vgpu.rst
16:14:14 bauzas jianghuaw: jaypipes: I had thoughts on that whether it was requiring nested RPs
16:14:26 jaypipes bauzas: yes, it does.
16:14:37 bauzas jianghuaw: jaypipes: tbh, I don't think we really need to wait for nested RPs if we can use custom RCs
16:14:48 openstackgerrit Ed Leafe proposed openstack/nova master: WIP - Migrate Ironic Flavors https://review.openstack.org/484949
16:14:49 bauzas lemme explain
16:15:00 edleafe cdent: Fixed the warning ^^
16:15:29 bauzas jaypipes: if we say that drivers can provide their own resource classes
16:15:53 bauzas jaypipes: and if operators provide custom RCs in flavors for defining how many "foos" we need to consume
16:16:07 bauzas jaypipes: why should we then requiring a parent/child relationship ?
16:16:22 bauzas for example, libvirt is not having a tight parent/child relationship
16:16:39 bauzas it's just providing a list of mediated devices having types
16:17:06 jaypipes bauzas: that's fine for a non-interoperable solution of course.
16:17:20 jaypipes bauzas: as soon as you have custom anything, you throw interop out the window
16:17:28 bauzas jaypipes: I don't disagree with your statement
16:17:51 bauzas jaypipes: I'm just saying that we have all the tooling in place for the feature
16:18:04 bauzas nested resource providers will give us more
16:18:08 sean-k-mooney bauzas: haha the only grammer/spelling corrections that i usally give are wrong. but ill take a look
16:18:28 mriedem bauzas: rather than rely on extra specs,
16:18:42 mriedem we will need to focus priority on actually getting nested RPs done in quens
16:18:46 mriedem for vgpus support
16:18:48 jaypipes bauzas: jianghuaw and I were trying to get GPU traits standardized in os-traits. once we have standard traits, you need resource providers to associate those traits to. and since different vGPU providers (pGPU groups) can have different traits, you need a way of saying "this pGPU group has these traits and this pGPU group has those traits". without nested providers, you can't do that.
16:18:51 mriedem and fpga and everything else
16:19:03 bauzas jaypipes: pGPU group is only a xen thing
16:19:19 bauzas mriedem: again, I don't disagree
16:19:25 mriedem bauzas: then why even bring it up?
16:19:30 jaypipes bauzas: or physical GPU... if you have >1 on a compute node.
16:19:37 openstackgerrit Julia Kreger proposed openstack/nova master: Ironic: Support boot from Cinder volume https://review.openstack.org/215385
16:19:41 bauzas mriedem: I just feel that the vGPU spec doesn't require it
16:19:49 bauzas jaypipes: no
16:19:55 openstackgerrit John Haan proposed openstack/nova-specs master: Support volume_type with BDM paramter https://review.openstack.org/466595
16:20:04 mriedem bauzas: there are lots of things that can be implemented the wrong way with extra specs
16:20:09 mriedem that doesn't mean we should do them
16:20:41 sean-k-mooney bauzas: so that version is effectivly a straight revert of just the incidental change so that should work well for the backport.
16:21:46 bauzas mriedem: the spec itself uses extra specs for traits, y'know :)
16:21:56 cdent presumably if we do something the quick and dirty way now, we have to allow that way to be present until the end times (or thereabouts) and we’d rather just do it right
16:22:21 bauzas mriedem: AFAIK, the only possibility we'll get for asking qualititative resource types will be by using extra specs, correct?
16:22:42 bauzas mriedem: and the same goes for some quantitative resources, if we speak of custom resource classes, right?
16:23:28 bauzas after all, we implemented the custom resource classes series because we knew that we needed some flexible model
16:25:12 mriedem we talked about putting something into the api for qualitative things at the newton midcycle
16:25:18 mriedem that didn't involve extra specs
16:25:25 mriedem but that would all have to be flushed out
16:26:07 mriedem unrelated, -1 on takashi's bottom change https://review.openstack.org/#/c/408955/ since it doesn't account for not allowing forced cold migrations between cells
16:26:20 mriedem but there is all sorts of request spec and filter properties goodies in there for you to look at
16:27:15 bauzas mriedem: I remember about that API
16:27:58 bauzas mriedem: after all, the spec from jianghuaw is just an other usecase of how to use placement, right?
16:28:01 openstackgerrit John Haan proposed openstack/nova-specs master: Support volume_type with BDM paramter https://review.openstack.org/466595
16:28:24 sean-k-mooney mriedem: you can migrate between cells? i taught you were normally limited to mightions within the same availablity zone. can you disable that check with a force flage?
16:28:25 bauzas mriedem: most of the spec is describing not some implementation proposal, just the workflow how it should work
16:29:01 bauzas mriedem: jaypipes: the only bits requiring implementation is internal to the drivers, right?
16:29:27 bauzas given we already have the virt driver interface for all of that
16:29:40 openstackgerrit Merged openstack/nova master: api-ref: Verify parameters in os-migrations.inc https://review.openstack.org/474879
16:30:29 jaypipes bauzas: sorry, reading back, was on a call
16:30:38 openstackgerrit Merged openstack/nova master: Stop using mox stubs in test_remote_consoles.py https://review.openstack.org/478423
16:30:39 mriedem sean-k-mooney: you can't migrate between cells
16:32:03 sean-k-mooney mriedem: oh sorry missread your comment. https://review.openstack.org/#/c/408955 currently allows it when it should not
16:33:17 jaypipes bauzas: so, the vgpu spec indeed mentions traits in extra specs, but it doesn't mention *custom* traits or *custom* resource classes :) jianghuaw has been good about pushing for standardized resource classes and traits.
16:34:23 mriedem sean-k-mooney: correct
16:36:20 mriedem jaypipes: bauzas: let's leave the vgpus stuff for after FF please
16:36:22 mriedem ffs
16:36:56 bauzas mriedem: jaypipes: sure didn't really wanted to make it clearly for now
16:37:03 bauzas just sharing thoughts
16:38:49 cdent mriedem: I think I fixed https://review.openstack.org/#/c/485263/
16:39:16 sean-k-mooney jangutter: can you click the rebase button on https://review.openstack.org/#/c/485125 so that the commit id wont change when it merges
16:40:07 openstackgerrit Jan Gutter proposed openstack/os-vif master: Improve OVS Representor VF Lookup https://review.openstack.org/485125
16:40:19 sean-k-mooney jangutter: im going to create a patch to the release repo to tag that commit for the 1.6.1 release of os-vif
16:40:31 jangutter sean-k-mooney: done
16:41:10 sean-k-mooney jangutter: cool ill just check locally but now it should be able to merge without a merge commit being created
16:44:15 openstackgerrit John Haan proposed openstack/nova-specs master: Support volume_type with BDM paramter https://review.openstack.org/466595
16:47:04 ralonsoh cdent, edleafe: thanks for your reviews in https://review.openstack.org/#/c/483451/. I posted my reply
16:48:27 melwitt counting instances is in the gate \o/ \o/ \o/
16:49:24 mriedem f yeah
16:49:36 mriedem melwitt: are you going to cleanup the 'remove refresh-quotas cli' change todya?
16:49:43 mriedem if not, i can whack at that
16:50:17 melwitt mriedem: yes, gonna hit that next
16:50:29 mriedem awesome
16:51:55 mriedem sheet, cellsv2 meeting in 9 minutes
16:51:58 mriedem dansmith: you want me to run it?
16:52:35 dansmith ugh
16:52:36 dansmith mriedem: melwitt do we have things to talk about other than the obvious/
16:52:55 dansmith obvious being "get the quotas shit merged"
16:52:59 mriedem i don't have anythign that's not in the etherpad
16:53:14 mriedem https://etherpad.openstack.org/p/nova-pike-cells-v2-todos

Earlier   Later