14:00:29 <efried> #startmeeting nova-scheduler
14:00:30 <openstack> Meeting started Mon Feb 11 14:00:29 2019 UTC and is due to finish in 60 minutes.  The chair is efried. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:31 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:33 <openstack> The meeting name has been set to 'nova_scheduler'
14:00:37 <gibi> o/
14:00:38 <takashin> o/
14:00:42 <mriedem> o/
14:00:42 <edleafe> \o
14:00:44 <alex_xu> \o
14:00:47 <bauzas> \o
14:01:41 <cdent> oh hai
14:02:30 <efried> Welcome one and all to the last n-sch meeting before the second full moon of 2019!
14:02:44 <efried> #link agenda https://wiki.openstack.org/wiki/Meetings/NovaScheduler#Agenda_for_next_meeting
14:02:57 <efried> #topic specs and review
14:03:02 <cdent> oooo special
14:03:12 <efried> ikr
14:03:15 <efried> #link latest pupdate http://lists.openstack.org/pipermail/openstack-discuss/2019-January/002055.html
14:03:15 <efried> #link blueprint tracking https://etherpad.openstack.org/p/nova-stein-blueprint-status
14:03:45 <efried> #link libvirt reshaper https://review.openstack.org/#/c/599208/
14:03:45 <efried> I saw some excitement in -nova just now about this. bauzas gibi anything worth sharing?
14:04:08 <gibi> I think
14:04:08 <bauzas> efried: well, just discussing about reshapes
14:04:09 <bauzas> https://review.openstack.org/#/c/599208/
14:04:13 <cdent> i reckon a key thing here is that we're open for business on the placement releated specs and tetsuro has started on at least one of them
14:04:35 <efried> (cdent getting there)
14:04:41 <bauzas> gibi helped me identifying a possible issue with VGPU allocations unrelated to real mdevs
14:05:06 <efried> gibi to the rescue again.
14:05:07 <cdent> (yea, I was typing that while you seemed to be waiting for something to happen, and then I realized where you were going, but decided I didn't want to store my buffer so flushed it)
14:05:23 <cdent> gibi keeps doing that
14:05:36 <gibi> :)
14:05:55 <efried> So bauzas we're expecting another respin soon?
14:06:03 <efried> tldr ^ ?
14:06:07 <bauzas> efried: yup, working on it
14:06:18 <efried> cool
14:06:19 <efried> #link xen reshaper (middle of series) https://review.openstack.org/#/c/521041
14:06:26 <efried> Assume still stalled, unless someone has an update
14:06:33 <bauzas> and when I say "working", I really mean 'I stopped to be paid by our customers'
14:07:04 <efried> customers schmustomers, *I* know what's best for Nova!
14:07:28 <efried> #link in_tree allocation candidates series starting at https://review.openstack.org/#/c/635722/
14:07:30 <mriedem> tip: ibm will pay you either way
14:07:37 <efried> tetsuro has a good start on this --^
14:08:04 <efried> tetsuro: anything we need to know about this?
14:08:19 <efried> (that isn't apparent from the spec/review?)
14:08:48 <efried> tetsuro might not actually be here. Moving on.
14:08:55 <efried> #topic Extraction
14:08:55 <efried> #link Extraction https://etherpad.openstack.org/p/placement-extract-stein-5
14:09:33 <efried> As cdent mentioned, key thing here is that we decided to
14:09:33 <efried> a) open extracted placement for business, viz aforementioned in_tree alloc cands patch from tetsuro
14:09:33 <efried> b) freeze in-tree placement, but keep it around for Stein
14:10:22 <efried> c) not write any nova code to use features developed in a), version discovery capabilities notwithstanding.
14:11:09 <efried> wrt b), we realized we had something of a testing hole (we weren't testing in-tree placement anymore) so mnaser is putting together a job to do that again.
14:11:10 * bauzas nods
14:11:44 <efried> #link CI job for in-tree placement with ansible https://review.openstack.org/#/c/635852/
14:11:44 <efried> (I think)
14:12:03 <cdent> yes that's it
14:12:18 <bauzas> with the NUMA/HPC/NFV firehose for Placement queries coming by at the PTG, I like the idea to start thinking about both testing *and* new features
14:12:32 <efried> ++
14:12:45 <efried> cdent, edleafe: anything else about extraction to highlight?
14:12:59 <edleafe> No, that about covers it
14:13:03 <cdent> aye
14:13:12 <efried> aye that covers it, or aye you have something else?
14:13:19 <cdent> covers it
14:13:23 <efried> cool, moving on.
14:13:35 <efried> #topic bugs
14:13:35 <efried> #link Placement bugs https://bugs.launchpad.net/nova/+bugs?field.tag=placement
14:13:35 <efried> Any bugs to highlight?
14:14:03 <efried> ight
14:14:05 <efried> #topic opens
14:14:30 <efried> Anyone?
14:14:40 <edleafe> PTL election season is just around the corner
14:14:52 <efried> mm
14:14:55 <edleafe> It might be time to revisit the governance change
14:15:07 <edleafe> So that we don't need a special election
14:15:21 <cdent> #link election details: https://review.openstack.org/#/c/629749/
14:15:50 <efried> where are we on the checklist? A matter of merging the VGPU reshaper patch and proving upgrades with tripleo and/or osa?
14:15:50 <mriedem> so there is about 3 weeks to determine the governance change
14:16:17 <efried> seems like from what I heard from lyarwood, the upgrades deal was going to happen too late for ^
14:16:21 <mriedem> yes
14:16:22 <mriedem> it is
14:16:28 <mriedem> which is why placement in-tree is staying for stein,
14:16:39 <mriedem> and the compromise was lifting the extracted placement api freeze in stein
14:16:41 <edleafe> that isn't the only reason
14:17:14 <mriedem> i think the governance checklist probably needs amending due to the change in plans for stein given the deployment tooling stuff
14:17:39 * cdent agrees
14:18:09 <mriedem> at this point i think we're just waiting on the reshaper thing
14:18:25 <efried> one way to look at it would be that, as long as we have an in-nova placement, placement should be governed under nova
14:18:25 <efried> but
14:18:25 <efried> the other way to look at it is that governance is about shaping the direction of placement, and since in-nova placement is frozen, that's n/a
14:18:25 <efried> so
14:18:25 <efried> I for one would be in favor of amending as stated.
14:18:30 <mriedem> deployment tools are targeting base install for stein with tripleo being "there" and OSA working on it
14:18:40 <bauzas> which is one of the reasons why I shifted gears recently to make sure I'm not blocking you
14:19:15 <bauzas> I just feel a bit uncomfortable to discuss governance change here and now
14:19:30 <bauzas> this timeslot isn't convenient for all parties
14:19:31 <cdent> changing the governance helps to signal that the extracted things is _the_ thing, hopefully encouraging related projects to move along as needed
14:19:46 <mriedem> for any deployment tools that aren't working on supporting deploying extracted placement for stein, they should probably be kicked in the butt because they are going to be surprised in train when the in-tree nova code is dropped
14:20:09 <edleafe> The signaling to others is important, IMO
14:20:13 <efried> So we need to repoen this discussion with melwitt present, proposal being to have a placement PTL election with the rest, contingent upon VGPU reshaper being merged by the time campaigning starts?
14:20:14 <mriedem> bauzas: cdent did ask melwitt about this last week after the call
14:20:18 <mriedem> she said she'd think abou tit
14:20:19 <mriedem> *about it
14:20:23 <bauzas> ok
14:20:28 <efried> okay
14:20:44 <cdent> yeah we want neither nova nor placement to have their progress gated forever on deployment tools
14:21:28 <efried> so is there an #action from here?
14:21:49 <cdent> somebody following up with melwitt ?
14:21:50 <mriedem> looks like train PTL nominations are right after stein feature freeze
14:22:06 <mriedem> well,
14:22:22 <mriedem> could just post something to the ML for follow up discussion in the nova meeting this week
14:23:00 <bauzas> that said, I'm not sure about the fact that a new governance change will make deployment tools more committed
14:23:01 <edleafe> Let's check in with melwitt first
14:23:16 <edleafe> If she's a no, then it's kind of moot
14:23:19 <mriedem> edleafe: by post to the ML i meant post "this is an idea"
14:23:23 <bauzas> just to be clear, it's - I think - not a signaling problem, more a resources problem :)
14:23:38 <efried> mriedem: ++, as long as that won't be the first time she hears of it, which it sounds like it won't.
14:23:44 <mriedem> it's not
14:23:52 <mriedem> briefly discussed in -placement last week
14:23:58 <efried> right
14:24:05 <cdent> bauzas: right, the goal isn't to signal encouragement, it is to avoid overly tight coupling, in the face of limited resources
14:24:18 <mriedem> bauzas: governance change won't speed up deployment tooling adoption of extracted placement,
14:24:23 <mriedem> but nova dropping the in-tree code in train will
14:24:29 <bauzas> oh yeah
14:24:43 <mriedem> or should
14:24:52 <cdent> dangerous word "should"...
14:24:53 <bauzas> well, if you break, you loose
14:25:04 <cdent> who has the ml action? edleafe or efried ?
14:25:16 <edleafe> I can do it
14:25:22 <cdent> cool
14:25:26 <bauzas> I just feel most of the deployment team know about the story
14:25:39 <mriedem> i wouldn't assume that
14:25:50 <mriedem> e.g. chef and juju
14:25:54 <mriedem> don't know how active they are
14:25:56 <cdent> "assume" also dangerous word
14:26:02 <bauzas> okay okay
14:26:04 <mriedem> but juju is financed so i'm sure they'll put resources on it
14:26:06 <efried> Thanks edleafe
14:26:06 <efried> #action edleafe to post to ML suggesting placement PTL election with the rest in the next cycle
14:26:47 <efried> Anything else before we close?
14:26:59 <bauzas> I just feel that if chef or juju isn't that active for knowing placement being splitted, then I guess they won't probably bother about a new PTL or even a code break until they run it :)
14:27:07 <bauzas> but meh
14:27:22 <cdent> there's a reason there is a cycle-trailing release mode
14:27:35 <edleafe> It's a signal. It can be ignored. :)
14:28:34 <efried> As mriedem says, given the current strategy (in-nova freeze in stein, remove in train), governance changes should be pretty irrelevant for deployment
14:29:27 <cdent> it's gonna be nova-net all over again!
14:29:37 <mriedem> i was already thinking that :)
14:29:40 <mriedem> and god i hope not
14:30:08 <mriedem> hell we haven't even dumped cells v1 yet
14:31:16 <efried> Ready to wrap up?
14:31:33 <efried> Thanks all
14:31:33 <efried> o/
14:31:33 <efried> #endmeeting