| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2020-05-14 | |||
| 14:09:42 | openstackgerrit | Jiri Suchomel proposed openstack/nova master: Bring back allowed_direct_url_schemes in support of RBD image download https://review.opendev.org/728095 | |
| 14:12:12 | stephenfin | artom: replied | |
| 14:12:44 | stephenfin | artom: tl;dr: I'd suggest blindly approving that and focusing your efforts on https://review.opendev.org/#/c/714703/ instead :P | |
| 14:15:35 | artom | stephenfin, hehe - yeah, I'm slowly making my way up the stack. | |
| 14:16:12 | stephenfin | gmann: Before you do any work on it, I already have a hacking 3.1.0 proposed https://review.opendev.org/#/c/728016/ | |
| 14:17:47 | stephenfin | Using semver for hacking (or flake8 and its ilk) is weird. On one hand, you normally think of minor versions as "not breaking anything" and that entails ensuring changes are purely additive, but for hacking a change can be purely additive and still break stuff | |
| 14:21:13 | stephenfin | gmann: in any case, I've opted for 3.1.0 because the release doesn't remove anything (so it's correct in a literal semver sense) and bumping the major version every time we add something new to hacking (or get it for free via flake8) seems insane | |
| 14:32:47 | stephenfin | lyarwood: Do you think you'd be able to take a look at the bottom of this series at some point? https://review.opendev.org/#/q/topic:bp/use-pcpu-and-vcpu-in-one-instance+status:open | |
| 14:34:10 | stephenfin | The bottom patches are almost all cleanup and shuffling of code. Things only get interesting higher up (currently in merge conflict) and I'm happy to review those ones | |
| 14:34:28 | lyarwood | stephenfin: yup but likely not today | |
| 14:34:40 | stephenfin | oh yeah, just at some point | |
| 14:34:58 | lyarwood | ack added | |
| 14:35:32 | lyarwood | gerrit really doesn't like long series like that | |
| 14:35:42 | lyarwood | it took a few clicks to work out where the bottom of the series actually was | |
| 14:38:14 | gmann | stephenfin: i see, thanks. I was waiting for hacking patch to merge before release. i agree on versioning things, it is difficult to make it consistent with other as per its nature of breaking things. 3.1.0 is ok for me. | |
| 14:41:46 | gmann | stephenfin: thanks you pinged me, i was about to push my local patch to release it :) | |
| 14:42:00 | stephenfin | heh, no problem :) | |
| 14:48:16 | lyarwood | weirdness, I'm getting the following in a functional test despite using microversion = 'latest' | |
| 14:48:18 | lyarwood | nova.tests.functional.api.client.OpenStackApiException: Unexpected status code: {"badRequest": {"code": 400, "message": "'host' is a required property"}} | |
| 14:49:16 | lyarwood | self.admin_api.put_service(test_compute0_service_id, {"forced_down": True}) <- from that, looks like it's using an earlier microversion but I can't see any obvious reason why, has anyone seen this before? | |
| 14:51:01 | gmann | lyarwood: which test. 'despite using microversion = 'latest'' you mean particular test right not overall for all request | |
| 14:51:48 | lyarwood | gmann: all requests, self.admin_api.microversion == 'latest' | |
| 14:53:55 | gmann | lyarwood: that might not work as test are written per microversion cap they would(might) not run for 'latest'. | |
| 14:55:17 | lyarwood | gmann: you're talking about https://github.com/openstack/nova/blob/ef3b5707328f460d0d03759061c01b1a20f28743/nova/api/openstack/compute/services.py#L380-L383 right? | |
| 14:55:26 | lyarwood | gmann: so you have to use 2.53 to use this? | |
| 14:55:34 | gibi | dansmith: hi! one complication of the cache bug is that the cache management is virt driver specific but the placement resource management is up in the compute manager. So for the fix I need to pass new information through the virt driver interface, either by returning disc occupancy from manage_image_cache() or by adding some new get method to the virt interface | |
| 14:57:23 | gmann | lyarwood: yes, but i am saying if you run all the func test with 'latest' then it would not work. or you are running only update service tests ? | |
| 14:57:58 | dansmith | gibi: yeah, I was saying to sean-k-mooney yesterday how it's kinda split between them in an unhelpful way | |
| 14:58:02 | lyarwood | gmann: ah sorry no this is a specific test class that's set to 'latest' | |
| 14:58:29 | dansmith | gibi: doing it during manage_image_cache() means the allocations will lag reality, which isn't going to work well I think | |
| 14:58:43 | dansmith | gibi: IIRC, that runs very slowly, like every 24h or something | |
| 14:58:48 | dansmith | (by default) | |
| 14:59:06 | dansmith | gibi: unfortunately, I think it kinda has to happen synchronously during boot to really be effective | |
| 14:59:07 | openstackgerrit | Takashi Natsume proposed openstack/nova master: Remove six.text_type (1/2) https://review.opendev.org/728109 | |
| 14:59:53 | gibi | dansmith: yeah for the boot I need a sync approach. for the periodicy resynchronization I'm thinking of reusing manage_image_cache() | |
| 15:00:08 | dansmith | gibi: sure | |
| 15:00:43 | gibi | dansmith: OK so we agree that a virt interface impact is unavoidable | |
| 15:00:45 | gmann | lyarwood: ohk, then it is taking old 2.1 schema which need host. i think we read 'latest' correctly in code- https://github.com/openstack/nova/blob/ef3b5707328f460d0d03759061c01b1a20f28743/nova/api/openstack/compute/schemas/services.py#L46 | |
| 15:01:05 | dansmith | gibi: yeah I dunno how else you would do it.. like I had to add driver.cache_image() for my thing | |
| 15:01:39 | dansmith | gibi: and probably some signaling about the filesystem same-ness | |
| 15:03:03 | gibi | dansmith: OK, I will continue working through this path then | |
| 15:04:36 | lyarwood | gmann: ah got it thanks | |
| 15:04:45 | lyarwood | gmann: I was missing api_major_version = 'v2.1' | |
| 15:06:08 | gmann | lyarwood: ohk, and it was microversion request with 2.0 so ignored | |
| 15:08:18 | lyarwood | gmann: yeah that wasn't at all obvious, thanks again :) | |
| 15:26:56 | openstackgerrit | Takashi Natsume proposed openstack/nova master: Remove six.text_type (2/2) https://review.opendev.org/728117 | |
| 15:28:24 | lyarwood | stephenfin / artom: random question, do either of you recall how nova.tests.functional.libvirt.base.ServersTestBase stops the instance directories from being created? | |
| 15:29:00 | artom | lyarwood, I'd have to go digging - there's a presumably a mock/monkeypatch somewhere | |
| 15:29:04 | stephenfin | lyarwood: Not off the top of my head, unfortunately | |
| 15:29:10 | lyarwood | I've stopped applying ImageBackendFixture and the other obvious file_open fixtures etc but it's still not happy | |
| 15:29:17 | lyarwood | ack np | |
| 15:30:39 | openstackgerrit | Stephen Finucane proposed openstack/nova master: WIP: Add emulated TPM support to Nova https://review.opendev.org/631363 | |
| 15:30:40 | openstackgerrit | Stephen Finucane proposed openstack/nova master: Add support for resize and cold migration of emulated TPM files https://review.opendev.org/639934 | |
| 15:37:54 | lyarwood | doh, it was nova.virt.libvirt.LibvirtDriver._create_image | |
| 15:39:33 | lucadelmonte90 | hello, | |
| 15:40:40 | lucadelmonte90 | i couldn't find anything online, is there a way to disable instance resize from flavor a to flavor b, but keep it enabled for flavor a to flavor c? | |
| 15:42:01 | stephenfin | efried: Can you drop your -2 on https://review.opendev.org/#/c/631363/ ? | |
| 15:42:56 | efried | stephenfin: dangit, I thought I had removed all of my -2s. Done. | |
| 15:43:15 | stephenfin | Thanks | |
| 15:45:39 | artom | lucadelmonte90, no | |
| 15:46:22 | lucadelmonte90 | artom thanks :D | |
| 15:47:16 | artom | lucadelmonte90, if only all questions were that easy :) | |
| 15:50:24 | gibi | nova meeting going to start in 10 minutes on #openstack-meeting-3 | |
| 15:55:24 | artom | stephenfin, question on https://review.opendev.org/#/c/468203/14 | |
| 15:55:29 | artom | Not confident enough to -1... | |
| 15:56:37 | bauzas | just in time for the meeting | |
| 15:56:49 | artom | For the meating | |
| 15:57:39 | bauzas | and yelp doesn't help for getting doctor's votes :( | |
| 15:58:01 | bauzas | artom: too early for me | |
| 15:58:11 | bauzas | I'm not hungry but we're close to beer time tho | |
| 15:58:26 | artom | Not that kind of meating, muahaha | |
| 16:00:04 | artom | stephenfin, and another question on https://review.opendev.org/#/c/461456/13 | |
| 16:02:47 | artom | bauzas, I beet myself up every day | |
| 16:02:57 | artom | (Let's not pollute the meating) | |
| 16:04:06 | bauzas | artom: my doctor told me I should heel | |
| 16:31:25 | bauzas | bear time ! | |
| 16:31:32 | gibi | :D | |
| 16:33:48 | gmann | gibi: https://blueprints.launchpad.net/nova/+spec/policy-defaults-refresh-deprecated-apis | |
| 17:07:58 | melwitt | sean-k-mooney: why do you think we need the bigger list here? https://review.opendev.org/#/c/727898/3/bindep.txt just wondering bc it works without them being explicitly named | |
| 17:30:09 | openstackgerrit | melissaml proposed openstack/os-traits master: Remove translation sections from setup.cfg https://review.opendev.org/728160 | |
| 17:56:51 | NobodyCam | Good morning Nova folk's could someone help point me to which database this would come from: `Placement API returning an error response: Conflicting resource provider name: <hostName> already exists. call_func /openstack/venvs/nova-18.1.5.dev5/lib/python2.7/site-packages/nova/api/openstack/placement/wsgi_wrapper.py:31` | |
| 18:01:28 | melwitt | NobodyCam: what version? (has to be older bc placement was removed from nova in train) database would be the placement database, and if the version is old enough, it would be the nova_api database | |
| 18:01:58 | NobodyCam | Morning melwitt :) I'm on rocky | |
| 18:02:59 | melwitt | ok, yeah so if you have a database called 'placement' it will be in there, if not it will be in the database called 'nova_api' | |
| 18:04:27 | NobodyCam | I have nova_api | |
| 18:04:27 | melwitt | there's a unique constraint on resource provider name https://github.com/openstack/nova/blob/stable/rocky/nova/db/sqlalchemy/api_models.py#L292-L306 | |
| 18:04:56 | melwitt | resource_providers table | |
| 18:07:22 | melwitt | and if you want to change the name the recommended way is to use the placement cli https://docs.openstack.org/osc-placement/rocky/cli/index.html#resource-provider-set | |
| 18:08:22 | melwitt | although I guess it depends on how you're getting the error. if it's nova trying to do something then you will need to change something in nova. sorry I spoke too soon | |
| 18:10:49 | NobodyCam | we had a hypervisor crash and was brought back with the same name | |
| 18:11:38 | NobodyCam | the error is from a openstack server create with --AZ nova:<Host_fqdn> blah | |
| 18:12:20 | NobodyCam | I only see one entry for the host in resource_providers | |
| 18:14:01 | melwitt | oh, this sounds familiar | |
| 18:14:23 | NobodyCam | hum seem I'm missing something: | |
| 18:14:28 | NobodyCam | https://www.irccloud.com/pastebin/GVwbM6LK/ | |
| 18:15:53 | melwitt | the placement cli is a osc plugin so you have to install the osc-placement package to get the commands | |
| 18:16:33 | melwitt | but it sounds like you have a situation where nova is trying to create the resource provider but it already exists | |
| 18:18:55 | NobodyCam | yea | |
| 18:19:40 | NobodyCam | I can delete the record from resource_providers | |
| 18:20:32 | melwitt | this is likely be related to a WIP patch https://review.opendev.org/678100 (just fyi) but to recover, you will need to delete the resource provider. but before you do it, please check if it has any allocations on it by using the placement cli | |