Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-01-26
16:52:51 edleafe jaypipes_: any idea about mriedem's comment here: https://review.openstack.org/#/c/404472/20/nova/scheduler/client/report.py@174
16:53:01 edleafe That got changed in PS 8
16:55:58 openstackgerrit Dan Smith proposed openstack/nova: WIP: Allow graceful service exits from compute https://review.openstack.org/425751
16:57:12 jaypipes_ edleafe: I don't see that changed in PS8. but anyway, it's likely just an oversight. There were so many patches that changed that line that I've lost track at this point which introduced latest and which introduced 1.X specified explicitly.
16:57:32 jaypipes_ edleafe: in this particular case, it should be the microversion that added resource classes, so 1.2, yes?
16:58:02 edleafe jaypipes_: not sure, but that sounds about right
16:58:54 bauzas johnthetubaguy: dansmith: so I made the spit
16:58:57 bauzas split even
16:59:13 bauzas johnthetubaguy: dansmith: are you suggesting me other way to verify if the placement section is correct?
16:59:16 bauzas I'm all ears
16:59:20 edleafe jaypipes_: https://review.openstack.org/#/c/404472/8..20/nova/scheduler/client/report.py@174
16:59:30 edleafe jaypipes_: that's where it got added
16:59:31 dansmith bauzas: I said check os_region right?
16:59:38 johnthetubaguy bauzas: I was thinking of using auth_url CONF instead
16:59:43 dansmith bauzas: that's our config value, should be there by the time you need it
17:00:07 dansmith johnthetubaguy: is that going to be different than password from an order of loading perspective? because his patch fails on password not even being a thing by the time we're loading
17:00:12 dansmith s/loading/running/
17:00:40 bauzas dansmith: you mean os_region_name?
17:00:48 bauzas that's something coming from us ?
17:01:05 johnthetubaguy dansmith: unsure, but I like your os_region suggestion better
17:01:12 johnthetubaguy https://github.com/openstack/nova/blob/master/nova/conf/placement.py#L22
17:01:44 johnthetubaguy dansmith: bauzas: password is part of the pluggable auth and not not actually required, so it shouldn't be that one
17:02:08 bauzas okay, let's try os_region_name then
17:02:19 bauzas dansmith is right, given it's the only one we have in tree
17:02:30 johnthetubaguy (assuming we don't later decide os_region_name should be defaulted to RegionOne, but lets just be sure not to do that this cycle!)
17:02:32 bauzas and password means a password auth, which could be changing in the future
17:03:00 bauzas johnthetubaguy: I could write an unittest for preventing that
17:03:11 johnthetubaguy bauzas: cool
17:03:16 bauzas johnthetubaguy: meaning that I'm expecting a None value
17:03:20 johnthetubaguy ack
17:03:27 openstackgerrit Claudiu Belu proposed openstack/nova: hyper-v: Adds support for VMs with nested virtualization https://review.openstack.org/398509
17:03:28 bauzas so someone defaulting the opt would have to cut that
17:03:50 bauzas okay, let's try again
17:07:22 bauzas shit, I need a fixturre
17:07:25 jaypipes_ edleafe: I must be seeing something different than you... I don't see placement latest being added in that patch?
17:07:31 bauzas dansmith: johnthetubaguy ^
17:07:54 jaypipes_ edleafe: i.e. that line isn't green in gerrit.
17:08:28 jaypipes_ edleafe: in any case, not sure it matter, right? solution is straight-forward. just s/latest/1.2/, yesh?
17:09:07 sfinucan erm, vladikr - you're looking at this, right? https://review.openstack.org/#/c/424786/ :)
17:09:24 mriedem jaypipes: i'd prefer the calling code is explicit about the microversion it needs, so GET and POST for resource classes,
17:09:28 mriedem not globally in the init
17:09:34 mriedem edleafe: ^
17:09:56 vladikr sfinucan, yeah, but it's not really possible to test this with tempest
17:10:19 sfinucan vladikr: Oh, wait - that was the thing that required hardware, right?
17:10:23 edleafe jaypipes_: this is what I see: http://imgur.com/a/ocO8X
17:10:33 vladikr sfinucan, yeah, I should abandon that
17:10:36 edleafe jaypipes_: just so you don't think I crazy (well, crazier)
17:10:41 sfinucan vladikr: Do please
17:10:48 sfinucan though I'd also suggest submitting a patch for https://github.com/wznoinsk/hw-tests
17:11:02 dansmith bauzas: I'm confused -- is os_region_name what you're going to use instead of password?
17:11:08 vladikr yeah
17:11:09 bauzas dansmith: yup
17:11:11 sfinucan vladikr: wznoinsk should be able to provide more information on that project, if you ask nicely
17:11:15 dansmith okay I thought you were saying you weren't
17:11:19 dansmith so.. good
17:11:21 bauzas dansmith: sorry then
17:11:22 dansmith :)
17:11:27 edleafe mriedem: ok, I'll look for where those are needed.
17:11:31 bauzas dansmith: so, it works with devstack (like the t-shirt)
17:11:51 bauzas dansmith: but for UTs and functional tests, I need to make a fixture and add it to nova.test
17:12:06 bauzas dansmith: in order to set the config opt
17:12:19 mriedem bauzas: just throw it in the global conf patcher
17:12:22 bauzas unless I'm just setting the opt straight
17:12:23 mriedem that is run on all tests by default
17:12:25 mriedem it's a 1 line change
17:12:32 bauzas mriedem: oh right
17:12:38 bauzas I forgot it, but yeah, much better
17:13:14 jaypipes_ edleafe: yes, that's what I see too. Notice that the line he is referring to is white, meaning the patch has not changed that line.
17:14:26 jaypipes_ mriedem: you mean passing in a specific microversion header for each GET/POST call?
17:14:55 mriedem jaypipes_: yes, like we're already doing for 1.1 with aggregates
17:15:12 jaypipes_ mriedem: got it. edleafe, got it? :)
17:15:47 edleafe jaypipes_: yeah, but if you compare to PS7, it's green
17:15:58 openstackgerrit Eric Fried proposed openstack/nova: Introduce base PowerVM driver (cfg drv) https://review.openstack.org/409404
17:16:00 edleafe meaning it's been in the code since PS8
17:16:09 edleafe IAC, I'm working on the changes
17:17:23 mriedem bauzas: yeah in here somewhere https://github.com/openstack/nova/blob/master/nova/tests/unit/conf_fixture.py#L30
17:17:34 bauzas mriedem: already did
17:17:40 mriedem dansmith: oh nice https://review.openstack.org/319379 is approved
17:17:43 bauzas running tests and uploading in a sec
17:17:44 mriedem was just going to check on that
17:17:53 dansmith mriedem: #10 in the gate MOFO
17:18:01 dansmith mriedem: I'm watching it like my life depends on t
17:18:29 johnthetubaguy melwitt: I sent a note on that quota patch, its not a blocker I am just curious what you think though.
17:18:44 johnthetubaguy melwitt: looking at the latest one now.
17:19:07 mriedem dansmith: was melwitt good with it?
17:19:14 dansmith mriedem: she pushed the last change
17:19:28 dansmith so I'm guessing
17:19:34 mriedem oh excellent
17:19:39 melwitt mriedem, dansmith: yes, all good
17:20:39 melwitt johnthetubaguy: I just saw, thanks. I do like the idea of making the limit check atomic but I'm not familiar with the generation id and how passing a deltas dict helps?
17:21:12 johnthetubaguy melwitt: I was really thinking, we do the generation thing much later, but the new interface would allow us to do that without changing all the callers
17:21:27 melwitt johnthetubaguy: ah, gotcha
17:21:51 jaypipes_ edleafe: danke
17:24:53 melwitt johnthetubaguy: okay, so you mean moving the counting call inside of Quotas.check_deltas for now. that makes sense
17:25:12 johnthetubaguy melwitt: yeah, squash the two calls into a single call
17:25:48 mriedem vladikr: question in https://review.openstack.org/#/c/405954/
17:26:01 vladikr mriedem, looking
17:26:26 melwitt johnthetubaguy: it would clean things up a lot too
17:26:34 johnthetubaguy melwitt: its a nit, I was wondering if get counts should be on the list object?
17:28:01 melwitt johnthetubaguy: it could. I wasn't opinionated about it. I feel like there might be an example of a count on an existing list object already

Earlier   Later