| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-08-09 | |||
| 20:08:12 | mriedem | "lemme put the bong down and i'll be right over" | |
| 20:08:37 | mriedem | hmm, setting cpu_allocation_ratio=1.0 in gibi's test doesn't make it fail | |
| 20:08:55 | dims | mriedem : ouch! | |
| 20:09:04 | dansmith | mriedem: nice | |
| 20:09:16 | dansmith | mriedem: set it before anything gets started? | |
| 20:09:24 | mriedem | yeah, in setUp | |
| 20:09:27 | mriedem | before setting the virt driver | |
| 20:09:51 | dansmith | mriedem: before compute or scheduler starts? | |
| 20:10:07 | mriedem | yes before everything starts | |
| 20:10:19 | dansmith | but wait, | |
| 20:10:23 | dansmith | before _everything_ starts? like all thethings? | |
| 20:10:26 | mriedem | http://paste.openstack.org/show/617983/ | |
| 20:11:03 | mriedem | so you're just <1 from 4 day weekend and no longer caring, i see | |
| 20:11:04 | dansmith | no mriedem I mean BEFORE EVERYTHING | |
| 20:11:07 | dansmith | haha | |
| 20:11:13 | mriedem | *1 hour | |
| 20:11:42 | dansmith | don't worry, this four day weekend will be a special kind of hell where I'll be begging to work on unit tests by mid-day tomorrow | |
| 20:15:13 | openstackgerrit | Eric Fried proposed openstack/nova master: WIP: Use ksa adapter for placement conf & requests https://review.openstack.org/492247 | |
| 20:16:27 | mriedem | dansmith: i'm +2 on this https://review.openstack.org/#/c/492242/ | |
| 20:19:59 | mriedem | ok bottom 2 are approved | |
| 20:26:32 | mriedem | jaypipes: easy one https://review.openstack.org/#/c/488510/ | |
| 20:28:32 | dansmith | mriedem: I bet it's because of the fake virt driver | |
| 20:28:42 | dansmith | 2017-08-09 13:27:58,682 INFO [nova.compute.claims] Total vcpu: 10 VCPU, used: 0.00 VCPU | |
| 20:28:55 | dansmith | 2017-08-09 13:27:58,682 INFO [nova.compute.claims] vcpu limit not specified, defaulting to unlimited | |
| 20:31:33 | edmondsw | efried see the comment I just added to https://review.openstack.org/#/c/485121/7 and let me know if you have any ideas there | |
| 20:31:34 | dansmith | actually we're getting an empty dict of limits in filter_properties | |
| 20:32:33 | efried | edmondsw Ack. | |
| 20:32:45 | openstackgerrit | Jay Pipes proposed openstack/nova master: Remove provider allocs in confirm/revert resize https://review.openstack.org/488510 | |
| 20:32:46 | openstackgerrit | Jay Pipes proposed openstack/nova master: Resource tracker compatibility with Ocata and Pike https://review.openstack.org/491012 | |
| 20:32:49 | jaypipes | mriedem: fixe.d | |
| 20:32:53 | mriedem | dansmith: yeah, because the only limits we'd get are from NUMATopologyFilter | |
| 20:33:02 | dansmith | but our limit for vcpus is in there | |
| 20:33:07 | dansmith | otherwise we won't check anything, AFAICT | |
| 20:34:48 | mriedem | oh i see, | |
| 20:34:50 | dansmith | mriedem: we're hitting this code: https://github.com/openstack/nova/blob/master/nova/compute/claims.py#L240-L244 | |
| 20:34:52 | mriedem | CoreFilter puts the vcpu limit in there | |
| 20:34:58 | mriedem | right | |
| 20:34:58 | dansmith | hah | |
| 20:35:15 | dansmith | lemme try with ram or something else | |
| 20:35:17 | mriedem | since the CoreFilter isn't enabled, as noted in the commit message, we don't care about filtering on cores... | |
| 20:36:02 | dansmith | must be same with ramfilter? | |
| 20:36:09 | dansmith | because obviously I don't hit that either since no limit | |
| 20:36:36 | mriedem | the flavor he's using fits the ram on the fake driver | |
| 20:36:46 | dansmith | I mean making it not | |
| 20:36:59 | dansmith | yeah novalidhost when I do that and enable ramfilter | |
| 20:38:28 | mriedem | ok +2 on his test then https://review.openstack.org/#/c/490814/ | |
| 20:38:39 | mriedem | i didn't realize the enabled filters impacted the claim code in the compute | |
| 20:39:46 | dansmith | so I didn't realize the filters controlled the compute node claiming behavior | |
| 20:39:54 | dansmith | not sure that really makes sense | |
| 20:40:07 | mriedem | did you just say the same thing as me? | |
| 20:40:12 | dansmith | however, it's nice to know that removing those base filters will stop the compute node from defeating placement's decisions | |
| 20:40:29 | dansmith | two things to set limits with different logic will definitely lead to confusion-based bugs | |
| 20:41:13 | dansmith | now I just gotta find a typo nit in here to get my stat back and we're good to go | |
| 20:42:15 | dansmith | found one | |
| 20:42:31 | mriedem | which change are you talking about? | |
| 20:42:35 | mriedem | oh gibi's patch | |
| 20:42:38 | dansmith | the test | |
| 20:43:33 | mriedem | removed my +2 just in time | |
| 20:43:37 | openstackgerrit | Jay Pipes proposed openstack/nova master: replace chance with filter scheduler in func tests https://review.openstack.org/491529 | |
| 20:43:43 | dansmith | I was just bluffing | |
| 20:43:49 | mriedem | d'oh | |
| 20:43:49 | dansmith | I did find one but it's not worth delaying over | |
| 20:44:05 | mriedem | heh replaying my +2 | |
| 20:44:39 | mriedem | are you happy with cfriesen's wording here then? https://review.openstack.org/#/c/491854/ | |
| 20:45:50 | dansmith | mriedem: I wouldn't say I'm happy | |
| 20:46:05 | mriedem | jaypipes: i'm not sure why i'm the author on this https://review.openstack.org/#/c/488510/3 | |
| 20:46:07 | mriedem | https://review.openstack.org/#/c/488510/ | |
| 20:46:32 | dansmith | mriedem: I've noticed that on a couple things you've updated recently | |
| 20:46:38 | mriedem | i didn't update this | |
| 20:46:38 | dansmith | resetting the author for some reason I mean | |
| 20:46:45 | jaypipes | yeah, me too... | |
| 20:46:51 | jaypipes | haven't cared much, though. :) | |
| 20:47:04 | dansmith | mriedem: you did, PS15 | |
| 20:47:09 | dansmith | and 16 | |
| 20:47:19 | mriedem | yeah it was PS15 | |
| 20:47:31 | mriedem | jaypipes: you want to fix that quick? | |
| 20:47:34 | jaypipes | mriedem: I certainly didn't change the author. | |
| 20:47:42 | jaypipes | mriedem: how do I do that? | |
| 20:47:52 | mriedem | git commit --amend --author "Jay Pipes <jaypipes@gmail.com>" | |
| 20:47:57 | dansmith | are we re-pushing this anyway? | |
| 20:48:07 | dansmith | because ... six hour check run | |
| 20:48:11 | jaypipes | gimme a sec, will do. | |
| 20:48:31 | mriedem | dansmith: i brought it up in the ML http://lists.openstack.org/pipermail/openstack-dev/2017-July/119801.html | |
| 20:48:51 | mriedem | i don't know what causes it automatically, something to do with git review -d + git rebase -i + git commit | |
| 20:48:53 | openstackgerrit | Jay Pipes proposed openstack/nova master: Resource tracker compatibility with Ocata and Pike https://review.openstack.org/491012 | |
| 20:48:53 | openstackgerrit | Jay Pipes proposed openstack/nova master: Remove provider allocs in confirm/revert resize https://review.openstack.org/488510 | |
| 20:48:59 | jaypipes | ok, mriedem done | |
| 20:49:22 | dansmith | mriedem: I do that all the time and it doesn't reset for me | |
| 20:49:30 | dansmith | I've had my way with all of jay's patches that way | |
| 20:49:52 | mriedem | git version 2.7.4 ? | |
| 20:50:20 | dansmith | indeed actually | |
| 20:50:23 | mriedem | git-review Version: 1.25.0 | |
| 20:50:34 | dansmith | same | |
| 20:50:45 | dansmith | I'm not running Windows98 or whatever you run though | |
| 20:51:02 | mriedem | this is xenial 16.04 | |
| 20:52:46 | mriedem | stretch/sid | |
| 20:52:46 | mriedem | user@ubuntu:~/git/nova$ cat /etc/debian_version | |
| 20:52:46 | mriedem | huh | |
| 20:52:58 | jaypipes | mriedem: quick summary of the cause of the "claim mystery" from that resize too big patch please? | |
| 20:52:59 | dansmith | expected :) | |