14:00:12 <edleafe> #startmeeting nova_scheduler
14:00:12 <openstack> Meeting started Mon Jul 31 14:00:12 2017 UTC and is due to finish in 60 minutes.  The chair is edleafe. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:16 <openstack> The meeting name has been set to 'nova_scheduler'
14:00:19 <mriedem> o/
14:00:19 <edleafe> Who's here today?
14:00:25 <dansmith> kinda
14:00:47 <gibi_> o/
14:00:48 * edleafe hands dansmith some coffee
14:01:17 <bauzas> \o
14:01:17 <alex_xu> o/
14:01:24 <cdent> o/
14:02:37 <edleafe> Jay Pipes doesn't seem to be around
14:02:56 <edleafe> Oh well, let's get started
14:02:57 <edleafe> #topic Specs & Reviews
14:03:05 <edleafe> #link Correct allocations https://review.openstack.org/#/c/488510/
14:03:40 <edleafe> This series is correcting (by removing) allocations from computes
14:04:09 <edleafe> Since Jay isn't around, anyone else have comments on this?
14:04:19 <dansmith> just to be clear,
14:04:20 <jaypipes> hey I'm here now, sorry for being late
14:04:43 <dansmith> I think the plan is to land that resize_confirm one, but not the full allocate-ectomy patch for pike
14:04:47 * edleafe smacks jaypipes on the wrist for tardiness
14:05:02 <jaypipes> dansmith is correct. the last patch was an exploratory one.
14:05:07 <edleafe> ah
14:05:41 <edleafe> I had read them as a one-two punch
14:05:52 <gibi_> based on the functional test in https://review.openstack.org/#/c/487958/ the confirming the resize does not remove the allocation from the source host
14:06:14 <edleafe> gibi_: that was next up:
14:06:15 <edleafe> #link Test resize with placement API https://review.openstack.org/#/c/487958
14:06:34 <gibi_> I moved that patch to of https://review.openstack.org/#/c/488510/
14:06:47 <gibi_> and made the periodic tasks run in the test
14:06:50 <edleafe> #undo
14:06:50 <openstack> Removing item from minutes: #link https://review.openstack.org/#/c/487958
14:06:59 <edleafe> #link Test resize with placement API https://review.openstack.org/#/c/488510
14:07:05 <edleafe> sorry about that
14:07:17 <gibi_> and it is now fails at a stable point
14:07:39 <gibi_> it seems that resource clean up is missing both after confirm and revert
14:08:02 <edleafe> gibi_: any idea where this is going wrong?
14:08:38 <gibi_> what I see is that the source host still has allocations after the resize has been confirmed
14:08:47 <gibi_> even if I let the periodic tasks run
14:08:59 <gibi_> I mean the update_available_resources task
14:09:19 <gibi_> similarly in case of revert the target host keeps the allocation
14:09:24 <dansmith> gibi_: both hosts have allocations or just the source?
14:09:46 <dansmith> both keeping the allocations means the periodic is doing the right thing (not interfering with a  migration)
14:10:01 <gibi_> dansmith: I have to check but I think both has the allocation
14:10:05 <dansmith> so I would suspect the confirm/revert logic that is supposed to collapse the double allocation
14:11:06 <dansmith> gibi_: your test is on top of jay's first patch above yes?
14:11:19 <dansmith> yeah
14:11:26 <dansmith> I'll look at those when we're done here
14:11:47 <cdent> does that code also address resize on same host?
14:11:51 <cdent> or did we punt on that?
14:11:59 <dansmith> we can't punt on that,
14:12:07 <dansmith> but that's a problem with the scheduler I think,
14:12:16 <dansmith> which I think happens before this
14:12:18 <jaypipes> correct
14:12:18 <gibi_> dansmith: yes
14:13:16 <edleafe> geez, I need more coffee. The original link was correct
14:13:18 <edleafe> #undo
14:13:19 <openstack> Removing item from minutes: #link https://review.openstack.org/#/c/488510
14:13:24 <edleafe> #link Test resize with placement API https://review.openstack.org/#/c/487958
14:14:17 <edleafe> So what's the plan to investigate this?
14:14:43 <dansmith> finish here so we can go look?
14:14:44 <cdent> I've got a failing tempest+gabbi test that demonstrates the double allocations not happening on same host resize. probably overkill, but I was wanting to flesh out that side of things as a way to start doing multi-node stuff, but seems gibi has figured out a way to do that functionally (yay!)
14:14:54 <jaypipes> yeah, I'm currently looking at gibi's work
14:15:11 <edleafe> OK, as long as this is being investigated.
14:15:15 <edleafe> Let's move on
14:15:24 <edleafe> #link Amend spec for Custom Resource Classes in Flavors https://review.openstack.org/#/c/481748/
14:15:27 <edleafe> This is still not approved, although it seems fine as is. Related to this spec:
14:15:38 <edleafe> #link Ironic Flavor Migration https://review.openstack.org/#/c/487954/
14:15:41 <edleafe> #link Improve handling of ironicclient failures https://review.openstack.org/#/c/487925/
14:16:05 <dansmith> edleafe: you mean the spec amendment isn't approved?
14:16:23 <edleafe> dansmith: Nope
14:16:43 <dansmith> nope it's not approved or nope that's not what you mean?
14:16:52 <edleafe> Nope, it's not approved
14:17:15 <edleafe> it just needs some love from nova-specs cores
14:17:36 <dansmith> aye
14:18:29 <edleafe> Just for completeness, these two patch series are on hold until Queens
14:18:32 <edleafe> #link Devstack to use resource classes by default https://review.openstack.org/#/c/476968/
14:18:35 <edleafe> #link Nested Resource Providers series starting with https://review.openstack.org/#/c/470575/
14:18:48 <dansmith> why is the devstack one on hold?
14:18:59 <dansmith> I feel like we're not testing what we need to test if we don't do that now
14:20:01 <edleafe> dansmith: dtantsur put a -2 on it, so I'm assuming that it is on hold
14:20:08 <edleafe> it would be good to get that in
14:20:25 <dansmith> he said to just avoid breaking anything
14:20:31 <dansmith> presumably so it didn't merge before it was ready
14:21:19 <edleafe> Well, no activity in over a week, so I assumed it was not going to be in Pike
14:22:16 <edleafe> Anything else for specs/reviews?
14:22:47 <alex_xu> #link https://review.openstack.org/#/c/489206/
14:23:23 <alex_xu> ^ I'm working on the common method for allocation candidates and 'GET /resource_providers' about traits
14:23:52 <alex_xu> but not finish yet
14:24:11 <edleafe> alex_xu: oh, thanks - I skipped a whole section!
14:24:19 <alex_xu> edleafe: np
14:24:22 <edleafe> The following are all related to filtering of RPs by Traits:
14:24:22 <edleafe> #link Traits support in the Allocation Candidates https://review.openstack.org/478464/
14:24:25 <edleafe> #link Add traits to the ResourceProviders filters https://review.openstack.org/#/c/474602/
14:24:28 <edleafe> #link Refactor of trait filtering SQLA https://review.openstack.org/#/c/489206/
14:24:31 <edleafe> #link Correct resource provider set_traits() https://review.openstack.org/#/c/489205/
14:24:44 <edleafe> These are all for Queens
14:25:00 <alex_xu> cool
14:25:59 <edleafe> #topic Bugs
14:25:59 <edleafe> #link Placement bugs https://bugs.launchpad.net/nova/+bugs?field.tag=placement
14:26:25 <edleafe> A few new ones this week
14:27:18 <edleafe> All related to the ongoing allocation work
14:27:35 <edleafe> Anything else to discuss about bugs?
14:28:12 <edleafe> #topic Open discussion
14:28:47 <edleafe> What's on you mind? Anything?
14:29:02 <edleafe> Or are we all anxious to get back to what we were doing?
14:29:39 <cdent> edleafe: seems the latter, conversations rage over in #openstack-nova
14:30:00 <edleafe> yeah, what I was thinking too
14:30:01 <edleafe> OK then. Thanks everyone!
14:30:01 <edleafe> #endmeeting