| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-05-30 | |||
| 13:59:20 | jichen | dansmith: and could you please check this -2 on https://review.openstack.org/#/c/523387/ ? it's been reviewed by some core and non-cores, will there be some criteria to remove the -2 so I can get more core to help (efried has reviewed the patch and it's dependency patches previously thoroughly) | |
| 14:00:09 | dansmith | jichen: that -2 is just to hold the base patch until that patch and the rest have +2+W on them so we don't land a driver with no function | |
| 14:00:25 | dansmith | jichen: so yes, I'll drop that procedural -2 when the approvals are stacked up behind it | |
| 14:00:48 | dansmith | I said that on the patch, so it shouldn't be preventing anyone from reviewing it | |
| 14:00:53 | jichen | dansmith: ok, I will get more people to review and have +2 +w on them then ping you later, thanks a lot | |
| 14:00:57 | dansmith | yep | |
| 14:05:12 | mriedem | sahid: some comments in your neutron docs patch for trusted vf https://review.openstack.org/#/c/571167/ | |
| 14:06:38 | mriedem | dansmith: just noticed this https://review.openstack.org/#/c/570685/ above, but wasn't there something similar that bnemec did for tripleo? | |
| 14:07:11 | dansmith | I think he proposed something | |
| 14:08:34 | bhagyashris | kashyap: any idea about ^ | |
| 14:09:52 | kashyap | bhagyashris: Sorry, which one? | |
| 14:11:15 | openstackgerrit | Moshe Levi proposed openstack/nova stable/pike: Don't overwrite binding-profile https://review.openstack.org/563501 | |
| 14:11:55 | mriedem | so something that's been bothering me about the FFU sessions which i don't think anyone has brought up, | |
| 14:12:07 | bhagyashris | kashyap: actually i have written functional test here: https://review.openstack.org/#/c/560459/11/nova/tests/functional/test_servers.py the patch is having libvirt related changes but in the functional test case i have not initialized the actual libvirt driver | |
| 14:12:07 | mriedem | besides just saying, the config options churn a lot, | |
| 14:12:34 | mriedem | is that for the upgrade support tag, we can't drop config options within a release, but we can across boundaries, | |
| 14:12:45 | mriedem | which includes dropping support for deprecated config option names, | |
| 14:13:09 | mriedem | sometimes we leave those deprecated config option aliases around for much more than 1 release, like pci_passthrough_whitelist and the scheduler_default_filters | |
| 14:13:17 | bhagyashris | kashyap: so as per the review comments I am trying to initialized the actual libvirt driver so that the change in code will be tested but i am facing some issues | |
| 14:13:24 | mriedem | i'm kind of surprised the FFU folk haven't pushed for config option drop/renames to last more than 1 release | |
| 14:13:58 | mriedem | lyarwood: have you heard anything like ^ in your FFU travels? | |
| 14:14:09 | kashyap | bhagyashris: If you don't mind, I'm in the middle of something. This requires me to see all the context, will get to it tonight. | |
| 14:15:11 | bhagyashris | kashyap: ok np but i am not sure will be present because i am working in IST time zone so here it's almost EOD | |
| 14:15:27 | mriedem | e.g. pci_passthrough_whitelist was deprecated in ocata https://review.openstack.org/#/c/356604/ and renamed to [pci]/passthrough_whitelist, certainly that's a fair enough amount of time to drop the old deprecated name | |
| 14:15:43 | kashyap | bhagyashris: I'll comment on the review tomm. I need to be AFK soon | |
| 14:16:16 | bhagyashris | kashyap: ok thank you :) | |
| 14:18:40 | lyarwood | mriedem: yeah some ops/diy folks have asked for that but we (TripleO) don't need it, heat/puppet takes care of everything for us under the covers. | |
| 14:20:02 | mriedem | i imagine the pain is really just needing to read through the release notes to see which config options are deprecated/dropped each release | |
| 14:20:10 | mriedem | and digesting that across all projects per release you're FFing through | |
| 14:23:18 | lyarwood | mriedem: yeah, something this group is trying to address for ops/diy folks AFAIK - https://www.openstack.org/videos/sydney-2017/making-fast-forward-upgrade-openstack-possible-with-configuration-mapping-tool | |
| 14:25:18 | lyarwood | https://review.openstack.org/#/c/526314/ maybe? | |
| 14:31:38 | mriedem | yup i heard about that for the first time at the summit | |
| 14:38:45 | efried | bhagyashris: Looks like the libvirt library isn't a [test-]requirement. If you install it into your functional venv, you should be able to get around that import error. | |
| 14:39:08 | openstackgerrit | Jay Pipes proposed openstack/nova master: mirror nova host aggregate members to placement https://review.openstack.org/553597 | |
| 14:39:25 | mriedem | jaypipes: figured out massimo's cold migrate issue in the ML | |
| 14:39:34 | efried | bhagyashris: As for the second thing, not sure I can help without seeing the code. Can you upload your patch set? | |
| 14:41:50 | bhagyashris | efried: here is my second approach test case http://paste.openstack.org/show/722352/ | |
| 14:42:47 | bhagyashris | efried: under /nova/tests/functional/libvirt/test_rt_servers.py | |
| 14:42:47 | jaypipes | mriedem: I saw that. :) | |
| 14:43:01 | openstackgerrit | Kashyap Chamarthy proposed openstack/nova master: libvirt: Remove usage of migrateToURI{2} APIs https://review.openstack.org/567258 | |
| 14:43:32 | bhagyashris | efried: can you help me how i should install libvirt library in my functional venv | |
| 14:43:45 | efried | bhagyashris: Are you still getting that import error? | |
| 14:44:58 | stephenfin | bhagyashris: IIRC, you need to install the libvirt and python-libvirt packages using your package manager (apt, dnf, yum...) | |
| 14:45:06 | efried | bhagyashris: You would do it like this: | |
| 14:45:06 | efried | bash # enter a new subshell | |
| 14:45:06 | efried | source .tox/functional/bin/activate # enter the `functional` venv | |
| 14:45:06 | efried | pip install libvirt # install the lib into the venv | |
| 14:45:06 | efried | exit # leave the venv by leaving the subshell | |
| 14:45:18 | efried | stephenfin: Oh, they're apt, not pip? | |
| 14:45:20 | openstackgerrit | Stephen Finucane proposed openstack/nova-specs master: Add 'numa-aware-vswitches' spec https://review.openstack.org/541290 | |
| 14:45:32 | efried | bhagyashris: Listen to stephenfin, not me. He would know better. | |
| 14:46:20 | stephenfin | efried: I think so, yeah. It's installed by default on Fedora, anyway | |
| 14:46:29 | dansmith | bhagyashris: what stephenfin said | |
| 14:47:54 | efried | stephenfin: If you're in a helpful mood, would you take a look at bhagyashris's paste above? What we're going for here is a functional test setup that uses the actual libvirt driver as much as possible, so that the compute node RP is created in placement, and update_provider_tree gets called by the RT on compute startup and periodics. | |
| 14:48:29 | mriedem | all of this request spec hydration code just needs to be burned out with acid/fire; i've had a todo on my list for awhile to actually do that, but... | |
| 14:48:41 | efried | and I don't know from libvirt, so don't really have a concept of how/where stuff needs to be mocked or whatever. | |
| 14:48:58 | stephenfin | efried: I thought functional tests were supposed to use fakelibvirt everywhere? | |
| 14:49:35 | stephenfin | i.e. you're testing everything in nova but nothing outside it, if possible. Integration tests (Tempest) are for testing the whole system | |
| 14:49:37 | efried | stephenfin: You mean nova.tests.unit.virt.libvirt.fakelibvirt.FakeLibvirtFixture ? | |
| 14:49:40 | stephenfin | Yup | |
| 14:49:47 | sahid | dansmith: what was the agreement regarding incrementing objects version for os-vif? | |
| 14:49:49 | stephenfin | Wait, lemme double check | |
| 14:49:50 | sahid | https://review.openstack.org/#/c/570959/1/os_vif/objects/vif.py | |
| 14:50:40 | stephenfin | efried: See nova/tests/functional/libvirt/test_numa_servers.py | |
| 14:50:58 | dansmith | sahid: what do you mean? that patch looks correct to me (wrt the version) | |
| 14:51:08 | dansmith | sahid: if it's an os-vif-specific thing, then I guess ask jaypipes | |
| 14:51:46 | sahid | dansmith: no it's not it's really about objects verion. I think you did not want that we incrmeent the version because that is not going to be shared by rpc | |
| 14:51:50 | sahid | or something like that, no? | |
| 14:52:06 | sahid | at least it's not right now | |
| 14:52:16 | openstackgerrit | Elod Illes proposed openstack/nova stable/pike: placement: Fix HTTP error generation https://review.openstack.org/571218 | |
| 14:52:26 | mriedem | i thought we tried to follow the normal object versioning in os-vif even if it's not going over the wire yet | |
| 14:52:33 | dansmith | sahid: but it could be serialized somewhere else at some point, like between neutron and nova | |
| 14:52:44 | dansmith | mriedem: right, I'm not sure why not, but maybe I'm missing something | |
| 14:53:22 | sahid | mriedem, dansmith ok so let's do that, seems reasonable yes | |
| 14:53:32 | stephenfin | mriedem, dansmith: Reworked the numa-aware-vswitch spec per our discussion in Vancouver. Lemme know if I missed anything | |
| 14:53:41 | stephenfin | Will start reworking the patches themselves shortly | |
| 14:53:48 | bhagyashris | efried, stephenfin: thanks checking for the same. | |
| 14:54:02 | mriedem | stephenfin: ok thanks, | |
| 14:54:08 | mriedem | got a couple of patches ahead of me yet | |
| 14:54:18 | stephenfin | ack | |
| 14:54:49 | dansmith | stephenfin: cool. given the complexity here, I think it probably makes sense to evaluate the spec and the code somewhat in parallel | |
| 14:55:21 | stephenfin | dansmith: Just to make sure, there isn't a deadline I need to wary of, in that case? | |
| 14:55:36 | dansmith | oh, maybe | |
| 14:55:58 | stephenfin | I'm also nervous of forgetting our learnings about not solving everything in the spec | |
| 14:56:17 | stephenfin | I think the code that I already have proves out the idea. Now I just need to shuffle about where I'm storing stuff | |
| 14:56:40 | stephenfin | (also, no spec = no runway slot) | |
| 15:01:00 | bauzas | gentle reminder for folks that the spec about multiple vGPU types has been updated https://review.openstack.org/#/c/557065/ | |
| 15:01:06 | bauzas | jaypipes: dansmith: ^ | |
| 15:02:21 | mriedem | stephenfin: spec freeze is june 7 | |
| 15:02:25 | mriedem | so you've got about a week | |
| 15:02:33 | stephenfin | to...? | |
| 15:02:38 | mriedem | get the spec approved | |
| 15:03:13 | stephenfin | That's in your hands then, I guess? I'll work on updating the code in the interim | |
| 15:03:17 | mriedem | probably not a bad idea to do another spec review sprint/flush in the coming week to get anything we want into rocky through before spec freeze | |
| 15:03:34 | mriedem | my volume multiattach enhancement spec has been waiting too | |
| 15:04:48 | bauzas | mriedem: you can review this easy spec for just a conf opt https://review.openstack.org/#/c/557065/ | |
| 15:05:11 | mriedem | i can add it to the queue | |
| 15:05:15 | cfriesen | mriedem: what are your thoughts on getting https://review.openstack.org/#/c/465787/ merged? | |
| 15:05:29 | bauzas | mriedem: fair enough | |
| 15:05:41 | cfriesen | mriedem: I guess you've got a todo in there still | |