| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-04-16 | |||
| 19:49:13 | mriedem | optional in newton, required starting in ocata | |
| 19:49:16 | cfriesen | mriedem: apparently "nova show" on the instance will result in adding the uuid onto the service, but "nova list" hits the ServiceNotFound error | |
| 19:49:20 | artom | mriedem, thank you :) | |
| 19:49:33 | melwitt | you'd have to upgrade your compute nodes to at least newton to make them report in to placement. might even be ocata, not sure | |
| 19:49:44 | mriedem | cfriesen: those should both go down the same code path as far as i know, but maybe not | |
| 19:50:48 | melwitt | I think it would be random depending on the ordering of the returned dupe services | |
| 19:54:33 | melwitt | nevermind, it looks like it should try to generate a UUID for both services that match the instance via host, so it should fail reliably | |
| 19:54:35 | sean-k-mooney | dansmith: just finished looking at the patch. the broad strokes look fine. i left some nits inline. main concern is the name attribute of the driver field. i think we should just not set it for safty across livemirate in the future. anyway i left comments in the patch but over all it looks ok | |
| 19:56:10 | sean-k-mooney | dansmith: i might try an apply it locally tomorow and see if i have any issues. anyway im going to head for the evening. | |
| 19:57:12 | dansmith | sean-k-mooney: okay just to be sure, you know we're already setting it to vhost in almost all cases now right? | |
| 19:57:24 | dansmith | sean-k-mooney: it's done in _get_virtio_mq_settings() | |
| 19:57:43 | sean-k-mooney | dansmith: am i dont think we do that for vhost-user ports however | |
| 19:57:51 | dansmith | sean-k-mooney: we do | |
| 19:57:56 | dansmith | sean-k-mooney: even though it is wrong | |
| 19:57:59 | dansmith | that's the point :) | |
| 19:58:09 | sean-k-mooney | when we request vhost multiqueue only right | |
| 19:58:16 | dansmith | no | |
| 19:58:33 | dansmith | if we're in kvm, and MODEL_VIRTIO, we call that method and it sets it to vhost always I think | |
| 19:58:56 | TheJulia | efried: your python-ironicclient patch, Would you be okay if I updated the commit message? | |
| 19:59:08 | dansmith | sean-k-mooney: oh, actually, I guess you're right | |
| 19:59:16 | dansmith | sean-k-mooney: sahid said the opposite and I looked but didn't see | |
| 19:59:19 | sean-k-mooney | dansmith: oh we did not used too | |
| 19:59:32 | dansmith | sean-k-mooney: so yeah, this does change it to be more often, which is what I was trying to avoid when we talked last week | |
| 19:59:36 | sean-k-mooney | dansmith: ya this is a normal vhost-user xml http://paste.openstack.org/show/719324/ | |
| 19:59:39 | dansmith | but I missed that we only do it for multiqueue | |
| 19:59:43 | efried | TheJulia: Sure, you can do whatever you like with it. I would be happy to keep working on it too if you prefer. (I saw you made a comment, but I haven't gotten to reading it yet.) | |
| 19:59:45 | dansmith | sean-k-mooney: so good call on that, thanks | |
| 20:00:18 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add wrinkle to FailedEvacuateStateTests https://review.openstack.org/561707 | |
| 20:00:31 | sean-k-mooney | dansmith: well i dont think it would break anything today since its ignored but if they ever stop ignoring it i dont know if that would break us | |
| 20:00:47 | dansmith | sean-k-mooney: yeah | |
| 20:01:03 | dansmith | sean-k-mooney: I figured since I don't know the details, it'd be better to just avoid changing it | |
| 20:01:16 | efried | TheJulia: Oh, it was rloo's note. Anyway, just let me know whatever you want me to do there. | |
| 20:01:20 | dansmith | sean-k-mooney: I was asking him to just set it explicitly where he knows it should be vhost or vhostuser | |
| 20:01:33 | dansmith | sean-k-mooney: did you see the cleanup patch below? | |
| 20:01:38 | efried | TheJulia: Caveat: Never used storyboard before... | |
| 20:01:43 | dansmith | that was also based on ignoring, but it's vhostuser and you said vhost-user, so is it wrong? | |
| 20:01:58 | TheJulia | efried: I can create it, its not a big deal | |
| 20:02:07 | TheJulia | efried: but your welcome to if you want | |
| 20:02:31 | TheJulia | Story clean-up is taking... many braincells. | |
| 20:02:39 | efried | TheJulia: Go for it, co-author away. | |
| 20:02:56 | sean-k-mooney | dansmith: well neighter vhostuser or vhost-user is documented explictly as an allowed value but they used vhost-user when refing to it. | |
| 20:03:13 | TheJulia | efried: don't need that :) But thanks. | |
| 20:03:16 | dansmith | sean-k-mooney: okay you wanna comment on the patch below with that info then? | |
| 20:03:18 | sean-k-mooney | dansmith: ill try and take a look at the libvirt source code tomorrow and see whent there parse has defiend | |
| 20:03:56 | dansmith | sean-k-mooney: okay | |
| 20:11:17 | openstackgerrit | Artom Lifshitz proposed openstack/nova master: Add (Same|Different)HostFilter to defaults https://review.openstack.org/561651 | |
| 20:13:32 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add trusted_image_certificates to REST API https://review.openstack.org/486204 | |
| 20:13:33 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add certificate validation docs https://review.openstack.org/560158 | |
| 20:18:03 | openstackgerrit | Lance Bragstad proposed openstack/nova master: Set scope_types for admin password policies https://review.openstack.org/553613 | |
| 20:18:05 | openstackgerrit | Lance Bragstad proposed openstack/nova master: Add scope_types to server policies https://review.openstack.org/525772 | |
| 20:19:54 | lbragstad | johnthetubaguy_: jaypipes ^ look another shot at working a rough test structure into those | |
| 20:20:36 | lbragstad | they will fail, but i'm curious about the approach you'd like me to take with the testing | |
| 20:20:52 | lbragstad | or if what's proposed even makes sense | |
| 20:22:04 | jaypipes | lbragstad: cool, thanks | |
| 20:22:12 | jaypipes | lbragstad: will re-review shortly. | |
| 20:22:30 | lbragstad | awesome - ty | |
| 20:54:35 | openstackgerrit | Matt Riedemann proposed openstack/nova stable/queens: List instances performace optimization https://review.openstack.org/561715 | |
| 20:54:49 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add trusted_image_certificates to REST API https://review.openstack.org/486204 | |
| 20:54:50 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add certificate validation docs https://review.openstack.org/560158 | |
| 20:56:23 | openstackgerrit | Eric Fried proposed openstack/nova-specs master: Support forced separation of providers in granular https://review.openstack.org/561717 | |
| 21:03:19 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Remove vestigial system_metadata param from info_from_instance() https://review.openstack.org/561724 | |
| 21:27:00 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add trusted_image_certificates to REST API https://review.openstack.org/486204 | |
| 21:27:01 | openstackgerrit | Jackie Truong proposed openstack/nova master: Add certificate validation docs https://review.openstack.org/560158 | |
| 21:32:42 | jackie-truong | mriedem dansmith jaypipes melwitt johnthetubaguy_: Thanks for reviewing the certificate validation patches. 479949 and 561262 should just need one last look. The API patch is going through Zuul now. | |
| 21:33:00 | jackie-truong | https://review.openstack.org/#/c/479949/ | |
| 21:33:05 | jackie-truong | https://review.openstack.org/#/c/561262/ | |
| 21:39:39 | melwitt | thanks jackie-truong. I've rechecked the tempest tests so we can look at that too for the latest revision of the series https://review.openstack.org/#/c/515210 | |
| 21:42:50 | jackie-truong | melwitt: Awesome, thanks | |
| 22:15:02 | mriedem | artom: gotta say, i'm not a fan of this https://review.openstack.org/#/c/561651/ | |
| 22:15:31 | mriedem | if there is a problem with tempest, let's fix tempest | |
| 22:15:31 | cfriesen | mriedem: I reproduced the Service UUID issue with vanilla stable/pike devstack. Writing up a bug now. | |
| 22:15:50 | mriedem | artom: as noted inline, those tempest tests for these filters probably shouldn't even live in tempest | |
| 22:18:57 | mriedem | jackie-truong: besides dumping on artom's patch i'm done for the day, sorry | |
| 22:19:04 | mriedem | cfriesen: ok | |
| 22:19:34 | jackie-truong | mriedem: np, thanks for all of the comments! | |
| 22:21:16 | melwitt | jackie-truong: looks like there's a problem :( http://logs.openstack.org/10/515210/31/check/barbican-simple-crypto-devstack-tempest/0ff6bcc/logs/testr_results.html.gz | |
| 22:21:23 | melwitt | Additional properties are not allowed (u'trusted_image_certificates' was unexpected)", u'code': 400 | |
| 22:21:31 | melwitt | looks like a schema validation fail? | |
| 22:24:08 | jackie-truong | melwitt: Saw that. Looking into it | |
| 22:47:56 | cfriesen | melwitt: mriedem: here's the Service-without-UUID exception.ServiceNotFound issue: https://bugs.launchpad.net/nova/+bug/1764556 | |
| 22:47:56 | openstack | Launchpad bug 1764556 in OpenStack Compute (nova) ""nova list" fails with exception.ServiceNotFound if service is deleted and has no UUID" [Undecided,New] | |
| 22:48:13 | melwitt | thanks | |
| 22:48:30 | cfriesen | reproducing is straightforward, just needs a fake DB entry in the services table | |
| 23:01:44 | cfriesen | why does "openstack server --live" require a hostname? | |
| 23:02:00 | cfriesen | make that "openstack server migrate --live" | |
| 23:09:50 | artom | mriedem, yeah, I get it | |
| 23:10:32 | artom | We were having downstream discussions about it, and the argument that sort of got me was "if every CI has to change it, shouldn't it be in defaults?" | |
| 23:11:10 | artom | Devstack enables it, downstream OSP CI needs to enable it, so... | |
| 23:11:28 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in test_serversV21.py (1) https://review.openstack.org/560821 | |
| 23:11:33 | artom | But yeah, "CI needs it" isn't a massively compelling argument | |
| 23:12:15 | artom | So I threw it up, just to see what discussion would come out of it, and to be able to say "upstream said" as opposed to "I said no" :) | |
| 23:12:34 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in test_serversV21.py (2) https://review.openstack.org/560829 | |
| 23:15:53 | melwitt | artom: I think it makes sense that CIs would enable most, if not all of the filters that exist, to cover the features and make sure they work with their products | |
| 23:16:56 | melwitt | for deciding whether to include certain filters in default_filters, I think that would be mostly about whether the filters are used by "most users" or are useful in a typical deployment | |
| 23:17:20 | artom | melwitt, yeah, apparently it's really complicated for them to enabled it in their thousand+ jobs? | |
| 23:17:33 | artom | melwitt, but I agree - CI and users have different needs | |
| 23:19:26 | melwitt | I didn't think it's complicated in that I thought you just set enabled_filters to the filters you want enabled in nova.conf | |
| 23:19:57 | artom | melwitt, yeah, but apparently they need to do that for *every* job? As in, there's no common deployment that covers all jobs? | |
| 23:20:31 | melwitt | that said, I don't really have an opinion about whether SameHost and DifferentHost belong in default filters because TBH I don't know if those are really commonly used by end users or not | |
| 23:20:34 | artom | https://review.gerrithub.io/#/c/407794/ is the infrared change | |