| *** belmoreira has joined #openstack-placement | 07:05 | |
| *** diga has joined #openstack-placement | 07:25 | |
| *** bhagyashris_ has joined #openstack-placement | 07:50 | |
| *** tetsuro has joined #openstack-placement | 08:18 | |
| *** rubasov has quit IRC | 08:28 | |
| *** helenafm has joined #openstack-placement | 08:29 | |
| *** giblet is now known as gibi | 08:33 | |
| *** e0ne has joined #openstack-placement | 08:34 | |
| *** tssurya has joined #openstack-placement | 08:47 | |
| *** ttsiouts has joined #openstack-placement | 09:11 | |
| *** bhagyashris_ has quit IRC | 09:34 | |
| *** e0ne has quit IRC | 09:40 | |
| *** rubasov has joined #openstack-placement | 09:56 | |
| *** cdent has joined #openstack-placement | 09:57 | |
| *** rubasov has quit IRC | 10:06 | |
| *** rubasov has joined #openstack-placement | 10:12 | |
| *** tetsuro has quit IRC | 10:15 | |
| *** e0ne has joined #openstack-placement | 10:16 | |
| *** takashin has quit IRC | 10:19 | |
| *** takashin has joined #openstack-placement | 10:45 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Factor listiness into an ObjectList base class https://review.openstack.org/637325 | 10:51 |
|---|---|---|
| openstackgerrit | Chris Dent proposed openstack/placement master: Move _set_objects into ObjectList https://review.openstack.org/637328 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move *List.__repr__ into ObjectList https://review.openstack.org/637332 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Clean up ObjectList._set_objects signature https://review.openstack.org/637335 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move RC_CACHE in resource_class_cache https://review.openstack.org/640114 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Use native list for lists of Usage https://review.openstack.org/639391 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make base test case file for object unit tests https://review.openstack.org/640406 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move reshape() into placement.objects.reshaper https://review.openstack.org/640540 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: ResourceProvider.increment_generation() https://review.openstack.org/640433 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move Allocation and AllocationList to own module https://review.openstack.org/640184 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Use native list for lists of Allocation https://review.openstack.org/640450 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Inline Consumer.increment_generation() https://review.openstack.org/640437 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Remove pep8 whitespace ignores https://review.openstack.org/640554 | 10:51 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Nit collector https://review.openstack.org/640555 | 10:51 |
| * cdent sighs | 10:52 | |
| cdent | rebase messiness | 10:52 |
| cdent | gibi: if you have any cycles that ^ stuff may be of interest. The real value is towards the end of the stack, but we're being incremental. | 10:54 |
| cdent | bbl | 10:54 |
| *** cdent has quit IRC | 10:54 | |
| *** ttsiouts has quit IRC | 11:04 | |
| *** ttsiouts has joined #openstack-placement | 11:05 | |
| *** ttsiouts has quit IRC | 11:09 | |
| *** e0ne has quit IRC | 11:10 | |
| *** e0ne has joined #openstack-placement | 11:11 | |
| *** diga has quit IRC | 11:14 | |
| *** cdent has joined #openstack-placement | 11:34 | |
| *** ttsiouts has joined #openstack-placement | 11:59 | |
| *** e0ne has quit IRC | 12:00 | |
| openstackgerrit | Merged openstack/osc-placement master: add python 3.7 unit test job https://review.openstack.org/610074 | 12:06 |
| *** e0ne has joined #openstack-placement | 12:08 | |
| *** ttsiouts_ has joined #openstack-placement | 12:36 | |
| *** ttsiouts has quit IRC | 12:40 | |
| *** edmondsw has joined #openstack-placement | 12:49 | |
| *** cdent has quit IRC | 13:03 | |
| *** e0ne has quit IRC | 13:04 | |
| openstackgerrit | Merged openstack/placement master: Refactor _get_trees_matching_all() https://review.openstack.org/639888 | 13:20 |
| openstackgerrit | Merged openstack/placement master: Adds debug log in allocation candidates https://review.openstack.org/639889 | 13:20 |
| *** cdent has joined #openstack-placement | 13:21 | |
| cdent | jaypipes or gibi: this is entirely optional, so if you like, it's not dangerous: https://review.openstack.org/#/c/619050/ | 13:35 |
| cdent | needs one more +2 | 13:35 |
| gibi | cdent: ack, I'm a bit swamped with FF | 13:35 |
| cdent | gibi: noted, there's no real rush on that stuff. I'm simply cruising outstanding stuff | 13:36 |
| cdent | efried: is your +1 on https://review.openstack.org/#/c/601614/ waiting on something | 13:36 |
| cdent | gibi: if there's something I can help with, let me know | 13:37 |
| gibi | cdent: there are 3 patches left for the Stein part of the bandwidth the current concern is with pci passthrough and pci claim :) | 13:38 |
| cdent | sadly that stuff is way outside my wheelhouse, but getting it down to 3 is a huge accomplishment | 13:39 |
| gibi | sean-k-mooney helped to go to the right direction so I'm fixing tests now | 13:39 |
| gibi | cdent: understood. There is not many devs who are happy to touch that part of nova | 13:39 |
| sean-k-mooney | gibi: happy may be a strech but im content or jaded enough to deal with the numa/pci/nfv code | 13:42 |
| gibi | sean-k-mooney: you have a very valuable knowledge | 13:43 |
| sean-k-mooney | gibi: did you push the auto name lookup patch by the way. feel free to add me to the review when you do | 13:45 |
| gibi | sean-k-mooney: still workig on it I will ping you when ready | 13:46 |
| jaypipes | cdent: will try my best. raging headache this morning... | 13:47 |
| cdent | jaypipes: don't strain yourself. make sure you're hydrated. etc etc. | 13:48 |
| *** tetsuro has joined #openstack-placement | 13:55 | |
| efried | cdent: just not feeling qualified to +2 it. | 13:57 |
| cdent | efried: roger that | 13:58 |
| efried | n-sch meeting now in #openstack-meeting-alt | 14:00 |
| *** diga has joined #openstack-placement | 14:01 | |
| openstackgerrit | Merged openstack/placement master: Add a vision-reflection https://review.openstack.org/630216 | 14:04 |
| *** diga has quit IRC | 14:12 | |
| *** ttsiouts_ has quit IRC | 14:18 | |
| *** ttsiouts has joined #openstack-placement | 14:19 | |
| *** ttsiouts has quit IRC | 14:21 | |
| *** ttsiouts has joined #openstack-placement | 14:21 | |
| cdent | efried: I'll resolve the merge conflicts on scrub-Lists after the meeting | 14:28 |
| efried | okay. | 14:29 |
| *** mriedem has joined #openstack-placement | 14:36 | |
| *** tetsuro has quit IRC | 14:53 | |
| efried | cdent: is https://storyboard.openstack.org/#!/project_group/placement going to be the best "top link" to see placement bugs in sb? | 14:56 |
| cdent | efried: I don't know yet. And I guess it depends on what you mean by "placement". That's the "group" which includes all 4 projecdts | 14:57 |
| efried | right, I'm just looking for a link to put next to the lp one in the meeting agenda. I'll use that for now; lmk if there's something better | 14:57 |
| cdent | All these "I don't know" is why I'm not rushing to say "we're soley using storyboard now" | 14:57 |
| cdent | seems fine | 14:57 |
| efried | yeah, but imo now that we've got the sb available, it's not too early to announce we're transitioning. | 14:58 |
| efried | I don't think anyone's suggesting we cut over exclusively and completely immediately. | 14:58 |
| efried | or really ever. I think a transition period is healthy. | 14:58 |
| *** e0ne has joined #openstack-placement | 14:59 | |
| efried | especially considering how loose the existing mechanism is, piggybacked on nova and broken tool-wise. | 14:59 |
| efried | so it's not like we'd really lose anything if somebody put something in sb and it wasn't perfect. | 14:59 |
| cdent | yeah, I don't disagree with any of that. I'm just not feeling any immediacy | 14:59 |
| cdent | it's in the queue | 14:59 |
| efried | ight | 14:59 |
| cdent | right, I agree with that too | 14:59 |
| *** 18WAAFJGJ has quit IRC | 15:00 | |
| *** guilhermesp has joined #openstack-placement | 15:01 | |
| *** takashin has left #openstack-placement | 15:01 | |
| *** e0ne has quit IRC | 15:02 | |
| *** e0ne has joined #openstack-placement | 15:04 | |
| *** ttsiouts has quit IRC | 15:06 | |
| *** ttsiouts has joined #openstack-placement | 15:07 | |
| *** ttsiouts has quit IRC | 15:11 | |
| alex_xu | efried: what can you do when we have NUMA, you already separate request group for different request group, and group_policy is isolate, how do you separate the VGPU request | 15:15 |
| alex_xu | efried: just like the case at https://review.openstack.org/#/c/552924/14/specs/stein/approved/numa-topology-with-rps.rst@282 | 15:18 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Factor listiness into an ObjectList base class https://review.openstack.org/637325 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move _set_objects into ObjectList https://review.openstack.org/637328 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move *List.__repr__ into ObjectList https://review.openstack.org/637332 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Clean up ObjectList._set_objects signature https://review.openstack.org/637335 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move RC_CACHE in resource_class_cache https://review.openstack.org/640114 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Use native list for lists of Usage https://review.openstack.org/639391 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make base test case file for object unit tests https://review.openstack.org/640406 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move reshape() into placement.objects.reshaper https://review.openstack.org/640540 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: ResourceProvider.increment_generation() https://review.openstack.org/640433 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move Allocation and AllocationList to own module https://review.openstack.org/640184 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Use native list for lists of Allocation https://review.openstack.org/640450 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Inline Consumer.increment_generation() https://review.openstack.org/640437 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Remove pep8 whitespace ignores https://review.openstack.org/640554 | 15:38 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Nit collector https://review.openstack.org/640555 | 15:38 |
| cdent | efried: that was more complicated that I expected | 15:38 |
| cdent | biab | 15:48 |
| *** cdent has quit IRC | 15:48 | |
| *** e0ne has quit IRC | 16:06 | |
| *** e0ne has joined #openstack-placement | 16:12 | |
| *** e0ne has quit IRC | 16:20 | |
| *** rubasov has quit IRC | 16:21 | |
| *** e0ne has joined #openstack-placement | 16:22 | |
| *** e0ne has quit IRC | 16:23 | |
| efried | alex_xu: If you care enough to isolate your CPUs, presumably you care enough to isolate your VGPUs as well. | 16:24 |
| efried | alex_xu: I assume you would want to affine (some of?) your VGPUs with (some of?) your CPUs, in which case you would put them together in the same request group | 16:25 |
| efried | alex_xu: So like, if you wanted one VGPU in the "smaller" group and two in the "larger": | 16:26 |
| efried | group_policy=isolate | 16:26 |
| efried | &resources1=VCPU:2,MEMORY_MB:1024,VGPU:1 | 16:26 |
| efried | &resources2=VCPU:6,MEMORY_MB:3072,VGPU:2 | 16:26 |
| *** cdent has joined #openstack-placement | 16:34 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Docs: extract testing info to own sub-page https://review.openstack.org/639628 | 16:41 |
| *** helenafm has quit IRC | 16:43 | |
| *** rubasov has joined #openstack-placement | 17:03 | |
| efried | cdent: I'm back | 17:05 |
| efried | I'm confused as to why you needed to rebase anything. | 17:05 |
| efried | I rebased everything on Friday. | 17:06 |
| efried | I guess tetsuro's refactors merged this morning, that would have needed some conflict resolution. | 17:06 |
| efried | But before that, was just an accident? | 17:06 |
| cdent | merge conflicts with code that | 17:06 |
| cdent | yes that | 17:07 |
| cdent | Earlier today I got a reject when sending code, because the code was based on stuff that was already "closed" and in the process of resolving that I over-rebased | 17:08 |
| efried | cdent: I've finished re-reviewing those and (re-)approved the ones I didn't author. | 17:18 |
| cdent | great, thanks | 17:19 |
| openstackgerrit | Merged openstack/placement master: Optionally migrate database at service startup https://review.openstack.org/619050 | 17:25 |
| cdent | efried: how do we feel about me voting on your changes at the start of that stack? seems legit, yeah? | 17:36 |
| efried | cdent: Yeah, your changes to them were negligible, right? | 17:36 |
| * cdent nods | 17:37 | |
| efried | Consider me +2 on your changes to my changes, then, which adds up to your +2. | 17:37 |
| cdent | gack, sometimes I really really hate the opt in nature of microversions | 17:44 |
| *** tssurya has quit IRC | 17:58 | |
| *** e0ne has joined #openstack-placement | 17:59 | |
| *** mriedem has quit IRC | 18:05 | |
| *** e0ne has quit IRC | 18:17 | |
| *** mriedem has joined #openstack-placement | 19:11 | |
| *** e0ne has joined #openstack-placement | 19:13 | |
| *** openstackgerrit has quit IRC | 19:23 | |
| *** openstackgerrit has joined #openstack-placement | 20:07 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make the PlacementFixture usable without intercept https://review.openstack.org/640852 | 20:07 |
| *** e0ne has quit IRC | 20:22 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make the PlacementFixture usable without intercept https://review.openstack.org/640852 | 20:36 |
| cdent | efried: I was thinking of intercept as a verb, but I'll adapt to be more clear | 20:39 |
| cdent | also that typo is just hilarious | 20:40 |
| efried | I thought so | 20:40 |
| efried | cdent: you're using it as a noun in the code later, which if consumers are just looking at the interface is maybe okay, but that happens never. | 20:41 |
| efried | cdent: Is this reproducible locally? | 20:46 |
| cdent | yes, pull down aspiers patch, (or likely just nova master), run `tox -r -efunctional-py37 test_report_client`, should get some failures | 20:47 |
| efried | cdent: and enginefacade blows up if there are conf options it doesn't recognize? That's... weird. | 20:49 |
| cdent | that's a slightly different thing | 20:50 |
| cdent | The bug that aspiers revealed is that the CONF in test_report_client has no sync_on_startup registered on it, because it's a nova conf, not a placement conf | 20:50 |
| efried | well, right, but it's the reason that del existed in the first place. | 20:51 |
| cdent | yes | 20:51 |
| cdent | and yes, eginefacade, when taking a dict, will blow up when it gets keys it doesn't know | 20:55 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make the PlacementFixture usable without intercept https://review.openstack.org/640852 | 20:56 |
| openstackgerrit | Kashyap Chamarthy proposed openstack/nova-specs master: cpu-model-selection: s/cpu_models/cpu_model_list/ https://review.openstack.org/640866 | 20:59 |
| efried | cdent: afaict, test_report_client is the only thing using PlacementDirect. And there's exactly one place in there where we *aren't* wrapping the entire test case in the interceptor. And that one place we could achieve similar functionality by temporarily re-mocking, say, SchedulerReportClient._client.request to blow up. So I think we could "solve" this by replacing use of PlacementDirect in test_report_client with | 21:16 |
| efried | PlacementFixture. | 21:16 |
| efried | which kind of begs the question as to what we should do with PlacementDirect - kill it? Does it have the utility we thought it should? | 21:16 |
| cdent | efried: as I recall the recent we used placement direct there was because we need to have a wsgi intercept under our direct control, because we have to shut down the application for a duration, right? | 21:17 |
| cdent | [t 6017] | 21:18 |
| purplerbot | <cdent> however I have seen zero evidence that PlacementDirect ever should have existed, so I'm not inclined to put a lot of energy into it [2019-03-04 20:36:22.844291] [n 6017] | 21:18 |
| efried | Yeah, "shut down" sort of. Like I said, we can simulate that by re-mocking out the request method in a context manager. | 21:18 |
| efried | I'll work on that fix while we wait for the gate here. | 21:18 |
| cdent | I don't know if we should kill PlacementDirect or just let it languish until someone comes along and wants it and finds it doesn't quite work | 21:18 |
| cdent | I think it _will_ work for the non-test case (where you make a command line tool that read /etc/placement/placement.conf and make a database conection, etc) | 21:19 |
| cdent | but does anyone want that? | 21:19 |
| efried | don't know. I was excited about being able to use it for test, but much less so when I found that you had to instantiate the interceptor in every test case. And you explained to me why that was desirable so I let it go. | 21:20 |
| efried | You agreed with me that Direct and Fixture were very similar and could potentially be consolidated. | 21:21 |
| efried | and I looked at doing that at some point, but it was going to be nontrivial. | 21:21 |
| efried | Mainly because what I wanted was to preserve Direct and get rid of Fixture, which would have entailed changing a lot more test cases. | 21:21 |
| efried | Now looking at doing it the other way, that would be a smaller surface area (just test_report_client, I believe), so... | 21:21 |
| * cdent nods | 21:23 | |
| cdent | If there's stuff that needs to be done in the morning, somebody leave me a note on the reviews or email or something. I need to be done for today. | 21:25 |
| cdent | good night all | 21:25 |
| *** cdent has quit IRC | 21:25 | |
| *** mriedem has quit IRC | 21:50 | |
| *** zzzeek has quit IRC | 21:54 | |
| *** zzzeek has joined #openstack-placement | 21:54 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Fix typo in db-auto-sync release note https://review.openstack.org/640886 | 22:19 |
| alex_xu | efried: but say, we have two pGPU on NUMA node2, and you request 2 vGPUs on node2, and each pGPU on node2 only left 1 vGPU :) | 23:28 |
| * alex_xu can't stop annoying | 23:28 | |
| *** openstackgerrit has quit IRC | 23:28 | |
| efried | alex_xu: You will always be able to come up with scenarios that we can't manage perfectly with existing syntax. | 23:30 |
| efried | alex_xu: But before you start coming up with those scenarios involving NUMA, we should have some idea of how we're going to solve the NUMA affinity problem in placement in the first place. | 23:30 |
| alex_xu | efried: you mean we still need more think on https://review.openstack.org/552924 | 23:31 |
| efried | alex_xu: yes, absolutely. | 23:32 |
| efried | though I do need to give that another read. | 23:32 |
| alex_xu | thanks, will keep thinking | 23:32 |
| efried | sorry, this sucks that I'm having to leave just as you're arriving, but I need to run :( | 23:32 |
| alex_xu | no worries, i already get answer | 23:33 |
| *** dklyle has quit IRC | 23:39 | |
| *** mriedem has joined #openstack-placement | 23:43 | |
| *** mriedem has quit IRC | 23:52 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!