14:01:51 <jaypipes> #startmeeting nova-scheduler
14:01:52 <openstack> Meeting started Mon Jan  2 14:01:51 2017 UTC and is due to finish in 60 minutes.  The chair is jaypipes. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:01:53 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:01:55 <openstack> The meeting name has been set to 'nova_scheduler'
14:02:10 <jaypipes> cdent: ok, go for it. :)
14:02:29 <jaypipes> #topic review of December merges
14:02:52 * cdent defers to jaypipes who is at speed
14:03:01 * jaypipes just showin' cdent how change topic :)
14:03:27 <cdent> bah, I know that stuff, I'm just slow today,
14:03:35 <jaypipes> ok, well then, I'll do it then :)
14:03:48 <jaypipes> alright, so real topic is...
14:04:07 <jaypipes> #topic open reviews
14:04:26 <jaypipes> there are a number of series of patches still going through review
14:04:46 <jaypipes> the highest priority one of those is the "integrating placement API calls into the scheduler"
14:05:17 <jaypipes> #link https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/resource-providers-get-by-request
14:05:40 <jaypipes> if we could get quality reviews done on the above, that would be appreciated.
14:06:12 <jaypipes> I don't want to +W that one because I basically wrote it.
14:06:17 * cdent nods
14:07:09 <jaypipes> next patch series is the work on aggregate linkage.
14:07:53 <jaypipes> this patch is the lynchpin:
14:07:57 <jaypipes> #link https://review.openstack.org/#/c/415031/
14:08:20 <jaypipes> unfortunately, mriedem has a -2 on that for procedural reasons while we get a spec exception due to the API change needed.
14:08:41 <jaypipes> hopefully we will get to hear from bauzas, johnthetubaguy or dansmith today on that one.
14:08:56 <cdent> ah, interesting
14:09:25 <jaypipes> cdent: yeah, mlavalle noticed that slight "issue" when trying to work with the placement API :)
14:09:41 <jaypipes> OK, next series up that needs reviews is the custom resource classes + Ironic integration work:
14:09:44 <jaypipes> #link https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/custom-resource-classes
14:10:53 <jaypipes> another patch of mine that needs reviews from cores is the patch which tracks aggregate mapping in the scheduler report client:
14:10:56 <jaypipes> #link https://review.openstack.org/#/c/407309/
14:11:29 <jaypipes> followed by cdent's REST API filtering by aggregate(s):
14:11:33 <jaypipes> #link https://review.openstack.org/#/c/407629/
14:12:34 <jaypipes> #topic interesting patch series, not priority...
14:13:07 <jaypipes> OK, so outside of the priority items discussed above, we have a few patch series that should be looked at but represent lesser priority non-Ocata items.
14:13:25 <jaypipes> The first is the PoC code for traits that alex_xu and Yingxin have put together:
14:13:46 <jaypipes> #link https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/resource-provider-tags
14:14:08 <jaypipes> I'd love to get folks' feedback on the above work. I'm going to be reviewing that today.
14:14:23 <jaypipes> The next patch series of interest is around nested resource providers:
14:14:33 <jaypipes> #link https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/nested-resource-providers
14:14:57 <jaypipes> unfortunately, due to the design of the NUMA objects in Nova, I've had to take a bit of a tack in direction on this patch series.
14:15:13 <jaypipes> this change in direction requires some surgery to the virt driver API...
14:15:25 <jaypipes> will write a ML post about this later today.
14:15:29 <cdent> is this new strategy since xmas?
14:15:33 <jaypipes> yes.
14:15:53 <cdent> bumer
14:15:59 <cdent> or even bummer
14:17:22 <jaypipes> cdent: I was able to get the SRIOV physical function resource tracking done, but then when trying to integrate those SRIOV PFs with their affined NUMA cell, ran into issues with the fact that the libvirt virt driver code re-creates the NUMATopology object that it JSON-ifies into the compute_nodes.numa_topology field on each and every call to get_available_resource(). This meant there was no way of uniquely identifying a NUMA cell with a
14:17:22 <jaypipes> UUID, because that UUID kept being regenerated on every periodic call to update_available_resource() in the resource tracker.
14:17:50 <cdent> eep
14:18:03 <jaypipes> cdent: all of which is due to the fact that all the NUMA/hardware.py stuff originally didn't use Nova objects and instead created its own non-versioned, non-DB-stored data over and over again.
14:18:15 <jaypipes> cdent: because, you know, functional programming.
14:18:20 <cdent> ossum
14:18:20 <jaypipes> anyways..
14:18:35 <jaypipes> so that is going to need some craftiness to unroll.
14:18:41 <jaypipes> ok, next topic...
14:18:54 <jaypipes> #topic diga and rfolco efforts
14:19:09 <rfolco> o/
14:19:18 <jaypipes> So, both diga and rfolco reached out to me about areas they could contribute to with the placement service.
14:19:38 <diga_> o/
14:19:59 <jaypipes> rfolco is going to be working on the item I posted to the mailing list about the lack of a DELETE /r-p/{uuid}/inventories call. diga is going to be working on having the placement service support event notifications.
14:19:59 <cdent> excellent
14:20:21 <diga_> +1
14:20:22 <jaypipes> rfolco: you want to say a few things about your effort, including #link'ing your blueprint/spec?
14:20:34 <jaypipes> followed by some words from diga :)
14:20:49 <rfolco> I am just struggling with gabbits, but this is minor :)
14:21:04 <jaypipes> rfolco: well cdent invented gabbi, so you have a resource here :)
14:21:20 <cdent> yeah, rfolco, feel free to ping me about it or anything else
14:21:37 <rfolco> #link https://review.openstack.org/415885
14:21:57 <cdent> that's in my review-asap queue
14:22:49 <diga_> jaypipes: yes, I am going through placement API code now
14:23:11 <diga_> jaypipes: setup is ready with devstack
14:23:24 <rfolco> I am just unsure about creating a new method for delete call or re-using set_inventories(), which does the same with non-empty resource list....
14:24:57 <cdent> rfolco: I'd suggest a new handler at the http api layer that calls existing objects-level code?
14:25:19 <cdent> btw, this stack of pending docs may be of use: https://review.openstack.org/#/c/408313/
14:25:29 <jaypipes> cdent: or calls set_inventories() with an empty InventoryList.
14:25:44 <jaypipes> cdent: that way it behaves the same as the PUT /inventories call
14:25:47 <cdent> jaypipes: isn't that what I said in my second point?
14:26:06 <jaypipes> cdent: doh, yes.
14:26:08 <jaypipes> sorry mate
14:26:13 <rfolco> ok we are on the same page then, I hope :)
14:26:19 * jaypipes steps back, bitten by cdent.
14:26:38 <cdent> no way man it's not a bite if I put a ? on the end
14:26:42 <jaypipes> diga_: you want to tell folks what you're working on?
14:26:47 <jaypipes> cdent: :)
14:27:43 <diga_> guys, I am working nova placement event notification
14:28:17 <cdent> diga_: that's great to hear, I started something on that a long time ago and then stalled out because it was too early in the process:
14:28:29 <cdent> #link cdent's early notification stuff: https://review.openstack.org/#/c/308503/
14:28:30 <diga_> cdent: :)
14:28:41 <diga_> cdent: will go through it :)
14:28:47 <jaypipes> diga_: #link to your blueprint?
14:28:48 <cdent> not much there
14:29:20 <diga_> cdent: jaypipes https://blueprints.launchpad.net/nova/+spec/placement-notifications
14:29:27 <diga_> cdent: okay
14:29:35 <jaypipes> #link https://blueprints.launchpad.net/nova/+spec/placement-notifications
14:29:45 <jaypipes> diga_: rock on.
14:29:59 <jaypipes> alright, well I don't have a whole lot else on the agenda.
14:30:06 <jaypipes> #topic opens
14:30:16 <jaypipes> anybody have anything to bring up or highlight?
14:30:22 <cdent> I wanted to highlight those docs I pointed a moment ago:
14:30:25 <diga_> jaypipes: will do my best :) finally got interested work in openstack
14:30:28 <cdent> #link pending dev docs: https://review.openstack.org/#/c/408313/
14:30:33 <jaypipes> diga_: ++!
14:30:50 <jaypipes> cdent: awesome. will review shortly.
14:31:04 <jaypipes> cdent: right after another review I need to do... ;)
14:31:19 <cdent> there are still some to dos, reviewers should feel free to suggest additional sections that may not yet be marked as to do
14:31:27 <jaypipes> sounds good.
14:31:36 <cdent> there's also the start at some api-ref docs, but that's much less formed, so far
14:31:44 <jaypipes> cdent: have we thought about an api-ref for placement?
14:31:48 <jaypipes> doh.
14:31:50 <jaypipes> jinx.
14:32:07 <cdent> #link api-ref start: https://review.openstack.org/#/c/409340/
14:32:12 <jaypipes> nice.
14:32:38 <jaypipes> awesomesauce. anyone have anything else?
14:32:43 <jaypipes> 5
14:32:45 <jaypipes> 4
14:32:47 <jaypipes> 3
14:32:49 <jaypipes> 2
14:32:51 <jaypipes> 1
14:32:56 <jaypipes> #endmeeting