| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-11-15 | |||
| 10:22:10 | mriedem | ok i thought we had a functional test to go from flat to nested and then schedule another workload on that provider | |
| 10:22:49 | openstackgerrit | Chris Dent proposed openstack/nova master: WIP: Use external placement in functional tests https://review.openstack.org/617941 | |
| 10:23:55 | cdent | gibi: that ^ is the latest. I've needed to do quite a bit to make unit tests not think about placement | |
| 10:25:01 | melwitt | gibi: oh, so that same test will exercise the virt integration code? that's where I'm not clear on | |
| 10:26:14 | gibi | cdent: ack | |
| 10:27:50 | gibi | melwitt: to have rehape in the test we need a virt layer returning a reshaped tree. I don't think the actal vgpu virt integration needs to be covered in the same test | |
| 10:28:56 | melwitt | what I mean is, does test_reshape verify that https://review.openstack.org/599208 works correctly I wonder | |
| 10:32:19 | gibi | melwitt: it does not cover the VGPU resource specifically therefore it does not cover the change in https://review.openstack.org/#/c/599208/8/nova/virt/libvirt/driver.py | |
| 10:33:49 | melwitt | thanks | |
| 10:49:46 | openstackgerrit | Chris Dent proposed openstack/nova master: WIP: Use external placement in functional tests https://review.openstack.org/617941 | |
| 11:22:08 | openstackgerrit | Chris Dent proposed openstack/nova master: WIP: Use external placement in functional tests https://review.openstack.org/617941 | |
| 11:23:00 | sean-k-mooney | stephenfin: https://review.openstack.org/#/c/568999 was rebased care to take a look | |
| 12:38:34 | frickler | coreycb: latest nova stable/pike released switched from ostestr to stestr, which draws a bunch of dependencies in, did someone start working on getting these into pike uca yet? I'd like to avoid duplicating the work | |
| 12:53:33 | coreycb | frickler: i haven't gotten to the latest stable release. we are due for an update for pike. | |
| 12:53:59 | coreycb | frickler: so haven't considered it. but if we need to backport packages that are only used for testing that shouldn't be a problem. | |
| 12:55:27 | sean-k-mooney | coreycb: frickler is this not somthing that should have been done prior to the pike release on ubuntu | |
| 12:55:53 | coreycb | sean-k-mooney: if other projects were using stestr then yes | |
| 12:56:59 | sean-k-mooney | well assuming ubuntu shipped nova it should still have used it but yes i belive several project made the swtich and i think all use it at this point | |
| 12:57:10 | sean-k-mooney | at least the python ones | |
| 12:57:12 | coreycb | sean-k-mooney: it's a little odd to switch dependencies for a stable release though | |
| 12:57:45 | sean-k-mooney | coreycb: yes which is why im questioning backportingt stestr if it was not already done | |
| 12:58:30 | coreycb | sean-k-mooney: just checked and we don't have stestr backported for pike | |
| 12:58:35 | sean-k-mooney | anyway i gues that up to ye to decided. its a test package so it should be a much lower risk | |
| 12:59:27 | sean-k-mooney | the change to stestr was made in pike right so its not that its not backprot it was never packaged when the release was made | |
| 13:00:14 | sean-k-mooney | or are you saying pike used ostester and frickler was wondering could you backport stestr to use instead on pike | |
| 13:01:20 | coreycb | sean-k-mooney: it appears that no projects in the cloud archive for pike currently use stestr so we may not have packaged it for pike | |
| 13:03:19 | sean-k-mooney | coreycb: so in that case the unit and functional tests would not have been runable for nova on pike without installing stestr from pip | |
| 13:04:09 | coreycb | sean-k-mooney: right, well s/wouldn't/won't/ in our case because we're not at 16.1.6 yet | |
| 13:05:40 | sean-k-mooney | coreycb: well its listed in the test requirements.txt https://github.com/openstack/nova/blob/stable/pike/test-requirements.txt#L20 so i you have the deb installed and you go to the site pacages directory you should just be able to run tox to run the tests and it will install it in the venv | |
| 13:06:13 | sean-k-mooney | that said i guess that would not be testing with the distro packages | |
| 13:06:35 | sean-k-mooney | i belive you can still run the test with ostester so its not a hard requirement | |
| 13:06:46 | coreycb | sean-k-mooney: yeah but that's not how debian packages work | |
| 13:07:35 | coreycb | sean-k-mooney: i'll raise an issue with you all if we have any problems. | |
| 13:07:50 | coreycb | sean-k-mooney: but i think we can probably figure it out | |
| 13:08:22 | coreycb | sean-k-mooney: still it's probably a change that shouldn't have landed in a stable branch | |
| 13:09:09 | openstackgerrit | Chris Dent proposed openstack/nova master: WIP: Use external placement in functional tests https://review.openstack.org/617941 | |
| 13:09:20 | cdent | gibi: let's see how that one goes (the depends-on has been updated too) | |
| 13:10:35 | frickler | sean-k-mooney: fyi this was introduced in https://review.openstack.org/580492 . result is I cannot build latest release within pike UCA currently | |
| 13:11:06 | sean-k-mooney | coreycb: oh it didnt ship this way https://github.com/openstack/nova/commit/5939ae995fdeb2746346ebd81ce223e4fe891c85 you are right that should not have landed in a stable branch | |
| 13:11:07 | gibi | cdent: testing it locally... | |
| 13:12:04 | cdent | i've been doing spot checks locally, haven't got reasonable cpu for the whole set | |
| 13:12:54 | sean-k-mooney | lyarwood: https://review.openstack.org/#/c/580492/ added a new dependcy to stable pike why was this allowed to be backported? | |
| 13:16:07 | lyarwood | sean-k-mooney: test dependcy that was already widely used by other projects | |
| 13:16:41 | gibi | cdent: I have access to a pretty big machine. `tox -e functional` now all green | |
| 13:16:58 | cdent | woot | |
| 13:17:21 | cdent | let's see what the gate wants to break | |
| 13:17:32 | lyarwood | dependency* | |
| 13:18:02 | cdent | gibi: do you agree with my choice to remove tests but not code (yet)? That made things easier, but I'm not sure it is proper. | |
| 13:20:57 | sean-k-mooney | lyarwood: still we normally dont even allow version bumps and in pike im not sure it was widely used yet. | |
| 13:21:21 | sean-k-mooney | lyarwood: infact on stable/pike keystone,glance,neutron and cinder still use ostestr | |
| 13:21:24 | lyarwood | sean-k-mooney: did this break anything/anyone? | |
| 13:21:55 | sean-k-mooney | yes it broke debian distros as its not packaged in uca in ubuntu or in based debian im guessing | |
| 13:21:59 | sean-k-mooney | stestr that is | |
| 13:22:07 | gibi | cdent: I agree with that compromise. I think the patch is alreasy complex. I think there will be a code removal patch later, after nova switched to use split placement repo | |
| 13:23:27 | gibi | cdent: I think the only problematic situation when we fix bugs in the placement code that is in the nova repo while nova still depends on that copy of the placement code | |
| 13:24:20 | sean-k-mooney | lyarwood: so far of all the project i have looked at nova is now the only project using stestr on pike | |
| 13:25:24 | lyarwood | sean-k-mooney: I was sure mtreinish was doing this across stable/pike tbh | |
| 13:27:06 | sean-k-mooney | nope https://review.openstack.org/#/q/owner:%22Matthew+Treinish%22++stestr+branch:stable/pike | |
| 13:27:21 | sean-k-mooney | look like it was nova only | |
| 13:27:35 | lyarwood | urgh | |
| 13:33:04 | sean-k-mooney | coreycb: frickler are ye going to adress this by packaging stester for pike or do we need a revert of jsut the stestr change? | |
| 13:34:28 | coreycb | sean-k-mooney: i wouldn't be opposed to a revert of stestr usage. if you didn't we could probably figure it out on our end it's just a bit of a pita. :) | |
| 13:36:06 | sean-k-mooney | well thats something that i guess the nova stable team should discuss. proably sooner rather then later | |
| 13:36:15 | sean-k-mooney | am im just going to grab lunch so brb | |
| 13:36:30 | coreycb | sean-k-mooney: thanks for pushing on this | |
| 13:41:26 | mriedem | coreycb: hey can we get some extracted placement debs? https://etherpad.openstack.org/p/BER-placement-extract | |
| 13:45:55 | coreycb | mriedem: placement's been split from nova in stein, is that right? | |
| 13:46:19 | coreycb | mriedem: if so, yes we can do that in the stein cycle | |
| 13:47:14 | cdent | coreycb: yeah, that's right | |
| 13:47:45 | cdent | coreycb: people are hoping to have packages sooner, if possible, to make testing with ansible+packages more possible | |
| 13:47:50 | cdent | not a blocker though | |
| 13:48:22 | coreycb | cdent: mriedem: ok thanks for the nudge. we'll try to get it done soon. feel free to keep nudging if we don't get to it soon. | |
| 13:50:01 | cdent | awesome, thanks | |
| 14:00:47 | openstackgerrit | Jack Ding proposed openstack/nova master: Improve libvirt image and snapshot handling https://review.openstack.org/616692 | |
| 14:01:02 | aspiers[m] | mriedem: got any time for a 5 min sync up on SEV in the next few hours? | |
| 14:02:42 | mriedem | aspiers[m]: i've got about 10 minutes right now outside of M-4 on level 3 | |
| 14:02:48 | mriedem | or inside | |
| 14:03:13 | openstackgerrit | Ghanshyam Mann proposed openstack/nova stable/rocky: DNM: Testing tempest-full keep testing on xenial on stable brnach https://review.openstack.org/618176 | |
| 14:03:20 | aspiers[m] | I think I'm next door | |
| 14:03:58 | aspiers[m] | I'll nip out of this Vitrage session | |
| 14:18:45 | openstackgerrit | Ghanshyam Mann proposed openstack/nova stable/pike: DNM: Testing tempest-full keep testing on xenial on stable brnach https://review.openstack.org/618181 | |
| 15:09:18 | openstackgerrit | Emmet Hikory proposed openstack/nova master: Execute Nova functional jobs on arm64 https://review.openstack.org/618191 | |
| 15:25:44 | openstackgerrit | Chris Dent proposed openstack/nova master: Use external placement in functional tests https://review.openstack.org/617941 | |
| 15:26:08 | cdent | gibi, mriedem, dansmith ^ that's fairly real now. | |
| 15:31:48 | sean-k-mooney | cdent: so this remvoes all the placement api unit and fucntional tests | |
| 15:32:07 | sean-k-mooney | cdent: but i dont se the removal of the placement api code | |
| 15:32:24 | sean-k-mooney | was that intentional | |
| 15:33:23 | cdent | sean-k-mooney: yes, I left the placement code in place so that devstack/tempest will still work without needing to depends on yet more stuff, but leaving the tests in would have meant many changes in those files to keep them working in the face of the "use the external placement fixtures" changes | |
| 15:33:35 | cdent | I explain that in the commit message, but probably not well enough | |
| 15:33:58 | sean-k-mooney | i have not read it yet :) | |
| 15:34:05 | sean-k-mooney | i just was looking at the file list | |
| 15:36:07 | sean-k-mooney | i dont know why i would feel more comfortable with haveing a second patch that jsut deleted the nova/api/openstack/placement folder after this but i would | |
| 15:36:33 | cdent | sean-k-mooney: happy to do that | |
| 15:43:37 | gibi | cdent: great work! I left one comment inline | |
| 15:43:53 | cdent | thanks gibi | |
| 15:47:47 | cdent | gibi: I left a response, summary is: not quite sure what to do | |
| 15:49:47 | gibi | cdent: ohh so tox-siblings is the problem there. I did not get that earlier when you mentioned a similar thing | |
| 15:51:09 | cdent | well, I tried it using git and one of the test runs in the gate did not work, and when I switched it back to package, it did | |
| 15:51:28 | cdent | but I _may_ have misinterpreted the problem, lemme look at the logs | |
| 15:51:56 | gibi | cdent: I think it worth to ask the zuul devs | |