| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-11-28 | |||
| 17:32:06 | cdent | so yeah, this sounds kind of 400 ish | |
| 17:36:13 | dansmith | wait, what? | |
| 17:36:25 | openstack | Launchpad bug 1805164 in OpenStack Compute (nova) rocky "Confusing error message when trying to disable non-nova-compute service" [Low,Triaged] | |
| 17:36:25 | mriedem | https://bugs.launchpad.net/nova/+bug/1805164 | |
| 17:36:36 | dansmith | disabling a non-compute service is still a 404 right? since the url to a non-compute service wouldn't exist? | |
| 17:37:04 | mriedem | the service id isn't in the URL for older microversoins | |
| 17:37:18 | mriedem | e.g. PUT /os-services/disable | |
| 17:37:23 | mriedem | the host/binary is in the body | |
| 17:37:41 | dansmith | ah yeah, was just pulling up the ref | |
| 17:37:45 | dansmith | yeah, okay | |
| 17:37:48 | mriedem | and anyway, we return 400 if you try doing PUT /os-services/{service_id} with a non-compute service with microversion >= 2.53 | |
| 17:37:54 | mriedem | so this would be consistent with that | |
| 17:38:40 | dansmith | yn | |
| 17:38:44 | dansmith | um | |
| 17:38:50 | dansmith | can you GET the service by id? | |
| 17:38:55 | dansmith | for non-computes? | |
| 17:39:21 | dansmith | api-ref doesn't show that you can, but I wonder if it's a missing doc | |
| 17:39:22 | mriedem | you can GET /os-services | |
| 17:39:28 | mriedem | not a specific service | |
| 17:39:35 | dansmith | weird | |
| 17:39:35 | dansmith | but you can PUT it? | |
| 17:40:06 | mriedem | there was lots of weird to be dealt with in https://specs.openstack.org/openstack/nova-specs/specs/pike/implemented/service-hyper-uuid-in-api.html | |
| 17:40:07 | dansmith | point being PUT /os-services/id where id is not a compute is kinda 404ish right? if you can't update non-compute services? | |
| 17:40:36 | dansmith | if you can't GET them it's hard to say either way | |
| 17:40:42 | mriedem | shrug | |
| 17:40:53 | mriedem | you can definitely GET them via list, but not directly via show right | |
| 17:41:00 | dansmith | yeah I mean get by id | |
| 17:54:30 | mriedem | here it comes | |
| 17:54:34 | mriedem | for all your ridicule | |
| 17:54:34 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Provide a useful error message when trying to update non-compute services https://review.openstack.org/620667 | |
| 17:57:08 | dansmith | mriedem: I commented on the rebuild spec.. I can fix my own nits but had a legit question | |
| 17:57:18 | dansmith | if you can answer I'll fix my nits and approve | |
| 17:59:52 | openstack | Launchpad bug 1793255 in OpenStack Compute (nova) rocky "nova-consoleauth missing from Rocky install guide; unable to use VNC" [High,In progress] - Assigned to melanie witt (melwitt) | |
| 17:59:52 | mriedem | melwitt: i've triaged at least 2 more bugs about the rocky install guide missing the mention of installing nova-consoleauth b/c it's still needed, i've duplicated them against https://bugs.launchpad.net/nova/+bug/1793255 but not sure what your plans were for putting some kind of note in the install guides about, "hey, you might still need to install nova-consoleauth, see the workaround option for details" | |
| 18:00:07 | mriedem | dansmith: before i look, let me guess: what do we do about running it through the scheduler? | |
| 18:00:15 | dansmith | no | |
| 18:05:36 | mriedem | dansmith: replies inline | |
| 18:10:15 | dansmith | mriedem: thanks, replied | |
| 18:10:54 | dansmith | mriedem: so I can fix my mechanical nits, but it seems like both our meatier comments deserve some actual new words. agree? | |
| 18:11:59 | mriedem | i replied again to make sure i'm following you | |
| 18:12:04 | mriedem | if we're on the same page, i'll update it | |
| 18:13:03 | dansmith | oh, so, re-reading I may have missed something | |
| 18:13:12 | dansmith | the procedure says "delete the existing volume attachment", | |
| 18:13:27 | dansmith | which I thought meant "the BDM" but that's not the case -- that means the attachment on the cinder side right? | |
| 18:13:52 | mriedem | correct | |
| 18:13:55 | dansmith | I was worried we might lost the linkage to the volume if we failed to create a temporary BDM at the right spot, and/or fail to allow re-replacing that | |
| 18:13:57 | dansmith | but this makes more sense | |
| 18:14:00 | mriedem | and it's what we already do today for rebuild | |
| 18:14:01 | dansmith | okay fair enough | |
| 18:14:09 | dansmith | for non-root you mean | |
| 18:14:13 | mriedem | the rollback procedure wasn't called out | |
| 18:14:19 | mriedem | yes | |
| 18:14:20 | mriedem | well, | |
| 18:14:25 | mriedem | you can rebuild a volume-backed server today too, | |
| 18:14:28 | mriedem | as long as the image doesn't change | |
| 18:14:36 | dansmith | yeah | |
| 18:14:45 | mriedem | we detach all the bdms, destroy the guest, spawn the guest and attach the volumes again | |
| 18:14:46 | dansmith | okay I'm with you, I was just projecting too much bdm in there | |
| 18:14:55 | mriedem | the last part 'completes' the attachment by giving cinder the host connector | |
| 18:15:01 | dansmith | detach or delete/ | |
| 18:15:08 | mriedem | detach | |
| 18:15:10 | mriedem | the bdms are fixed | |
| 18:15:18 | mriedem | the volume attachment record is transitory | |
| 18:15:29 | dansmith | right, so we still have the linkage between the instance and the volume it should be attached to yes? | |
| 18:15:33 | mriedem | so we'll update the bdm.attachment_id during the rebuild | |
| 18:15:36 | mriedem | yes | |
| 18:15:41 | dansmith | right, okay, gotcha | |
| 18:15:44 | mriedem | we do that dance to keep the volume 'ours' | |
| 18:15:49 | dansmith | yeah | |
| 18:21:41 | mriedem | dansmith: do you agree that if the image changes, we should run it through the scheduler as we do for image-backed servers? | |
| 18:21:51 | dansmith | mriedem: yeah I said that didn't I?\ | |
| 18:21:58 | mriedem | you said...something, | |
| 18:22:00 | mriedem | i was trying to confirm | |
| 18:22:33 | dansmith | I said "same policy-enforcment as the image-backed ones" | |
| 18:22:34 | mriedem | but ok i'll update the spec with more wordz | |
| 18:22:39 | dansmith | I ain't speakin' no jive | |
| 18:22:45 | mriedem | you said conflictory | |
| 18:22:52 | mriedem | my nose started bleeding | |
| 18:22:55 | dansmith | haha, okay I was speakin' jive | |
| 18:22:57 | dansmith | haha | |
| 18:50:51 | openstackgerrit | Merged openstack/nova master: Give drop_move_claim() correct docstring https://review.openstack.org/620170 | |
| 18:58:45 | mriedem | dansmith: having thought about this over lunch, i'm not really sure if we can/should try to update the host connector back into the attachments if rebuild fails - the volume would be in 'error' status, so i'm not sure if we should mess with it. rebuilding the server again later would just delete the empty volume attachment and start over with a new one | |
| 18:59:40 | dansmith | mriedem: okay I didn't think we would, because we'd be putting attachment info back in that is no longer valid (i.e. couldn't be deleted again when you retry) right? | |
| 19:00:20 | mriedem | we can delete it again when we retry | |
| 19:00:29 | mriedem | it's just a CRUD operation on the volume attachment record | |
| 19:40:41 | openstackgerrit | Matt Riedemann proposed openstack/nova-specs master: Support volume-backed server rebuild https://review.openstack.org/532407 | |
| 19:40:44 | mriedem | dansmith: updated ^ i'll wait to +2 | |
| 20:06:23 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Restore nova-consoleauth to install docs https://review.openstack.org/605154 | |
| 20:14:07 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Restore nova-consoleauth to install docs https://review.openstack.org/605154 | |
| 20:22:13 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Restore nova-consoleauth to install docs https://review.openstack.org/605154 | |
| 20:30:55 | mnaser | simple backport needing some votes - https://review.openstack.org/#/c/619351/ (stable/rocky already merged) | |
| 20:48:44 | melwitt | mriedem: by default, people shouldn't have to install nova-consoleauth on a fresh install _unless_ they've enabled the [workarounds]enable_consoleauth option. I added a comment to the review | |
| 20:49:35 | melwitt | I'm going to compare when those bugs were opened vs when the change that made nova-consoleauth optional merged to stable/rocky | |
| 20:50:44 | mriedem | both were opened this month | |
| 20:50:47 | mriedem | the duplicates | |
| 20:52:47 | melwitt | ok. I had tested that nova-consoleauth is no longer needed via this devstack change https://review.openstack.org/607070 back when I worked on the patch that made it optional | |
| 20:52:58 | melwitt | going to look at the duplicates now | |
| 21:14:50 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Mention size limit on user data in docs https://review.openstack.org/620700 | |
| 21:15:19 | flwang | mriedem: does nova support configuring the volume type when booting? | |
| 21:15:49 | mriedem | flwang: with microversion 2.67 in stein yes https://docs.openstack.org/nova/latest/reference/api-microversion-history.html#id60 | |