| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-07-12 | |||
| 22:17:41 | mriedem | *re-learned | |
| 22:28:47 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add note about reschedules and num_attempts in filter_properties https://review.openstack.org/582412 | |
| 22:49:50 | openstackgerrit | melanie witt proposed openstack/nova stable/ocata: [stable only] Always commit quota decrement if build request found https://review.openstack.org/582413 | |
| 22:53:42 | melwitt | mriedem: I'd appreciate a sanity check from you at your convenience ^ before I open a bug. this came from a customer seeing behavior where they delete an instance and don't see quota usage decrease (in the logs I saw racing creates and deletes) | |
| 22:54:43 | melwitt | I think I've convinced myself that it makes sense to commit the quotas whether we succeed in looking up the instance or not but it would be good to know if you think that's legit too | |
| 22:55:57 | mriedem | seems not crazy, but i'd have to load up all of that terrible context from conductor and such into my brain | |
| 22:56:18 | mriedem | i wish we could remove a lot of that "in the future" and "soon this will happen" stuff in api and conductor | |
| 22:56:23 | melwitt | yeah :( it was painful for me to re-learn that stuff too | |
| 22:57:14 | melwitt | yeah, same. for all I know, we can. honestly don't know where we're at on those | |
| 23:02:17 | cfriesen | re: the issue with service UUIDs and deleted services...you might want to take a look at https://bugs.launchpad.net/nova/+bug/1764556 | |
| 23:02:18 | openstack | Launchpad bug 1764556 in OpenStack Compute (nova) ""nova list" fails with exception.ServiceNotFound if service is deleted and has no UUID" [Medium,In progress] | |
| 23:04:18 | cfriesen | not sure if it'll cause problems or not, but it's mucking around in the same area | |
| 23:04:27 | melwitt | yeah, thanks for the heads up. reading through it | |
| 23:05:12 | openstackgerrit | Julia Kreger proposed openstack/nova master: virt/ironic: Implement rescue and unrescue https://review.openstack.org/416487 | |
| 23:05:13 | openstackgerrit | Julia Kreger proposed openstack/nova master: ironic: provide facilies to gracefully navigate versions https://review.openstack.org/582416 | |
| 23:05:18 | mriedem | melwitt: i've had these forever https://review.openstack.org/#/q/topic:cleanup-update-instance+(status:open+OR+status:merged) | |
| 23:05:29 | mriedem | not in the same stuff you're removing, bu related idea on cleaning up transition code | |
| 23:05:31 | TheJulia | mriedem: ^^^ | |
| 23:05:42 | mriedem | TheJulia: ack | |
| 23:05:59 | TheJulia | mriedem: I'm going to go kick the two jobs to recheck | |
| 23:06:14 | mriedem | good idea | |
| 23:06:25 | melwitt | mriedem: gah, I didn't know about those. I've starred them to remind myself to review | |
| 23:08:13 | TheJulia | heh, need to rebase one anyway | |
| 23:09:16 | melwitt | cfriesen: unless I'm all mixed up, it sounds like the patch I've proposed would fix that bug too | |
| 23:14:36 | cfriesen | melwitt: because it will no longer try to generate the new UUID? | |
| 23:14:44 | melwitt | yes | |
| 23:15:05 | melwitt | I'm assuming there's no reason to generate a new UUID for a deleted service, but maybe I'm wrong | |
| 23:16:21 | melwitt | if we want to do that, we'd have to add some conditionals in the Service.save() method to skip things like the version check if self.deleted and do something to handle ServiceNotFound | |
| 23:16:36 | melwitt | *if we want to generate UUID for deleted service | |
| 23:16:39 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Pass request_spec from compute to cell conductor on reschedule https://review.openstack.org/582417 | |
| 23:16:49 | cfriesen | melwitt: seems plausible that it'd fix the problem. | |
| 23:17:44 | melwitt | I'm gonna add another functional regression test to the stack for the bug you linked to verify it | |
| 23:18:26 | melwitt | maybe we'll have to do a different approach to fix things but the functional tests should be useful either way | |
| 23:27:17 | mriedem | TheJulia: couple of small things in https://review.openstack.org/#/c/582416/ | |
| 23:29:01 | TheJulia | I can revise once the ironic jobs finish, otherwise zuul will flag everything as mergefail and then even more fun :( | |
| 23:29:22 | mriedem | yeah | |
| 23:31:15 | mriedem | +2 on the rescue driver change itself now at least | |
| 23:32:33 | TheJulia | mriedem: awesome! Thanks! | |
| #openstack-nova - 2018-07-13 | |||
| 00:30:19 | openstackgerrit | Merged openstack/nova master: Update process doc to be more generic about point of contact https://review.openstack.org/582312 | |
| 00:43:13 | openstackgerrit | Merged openstack/nova master: Handle rebuild of instances with image traits https://review.openstack.org/569498 | |
| 01:04:13 | openstackgerrit | Vu Cong Tuan proposed openstack/nova-specs master: Switch to stestr https://review.openstack.org/581284 | |
| 01:17:52 | mriedem | https://review.openstack.org/569498 merged so we can mark that bp complete and move it out of the runway slot | |
| 01:18:59 | melwitt | mriedem: ack, thanks | |
| 01:19:32 | openstackgerrit | Vu Cong Tuan proposed openstack/python-novaclient master: Switch to stestr https://review.openstack.org/582090 | |
| 01:21:19 | melwitt | mriedem: I put next in line into a runway and made the end date July 26 FF (instead of July 27) | |
| 01:21:36 | mriedem | melwitt: comment on https://review.openstack.org/#/c/582312/ if you wanted to make that more actionable | |
| 01:22:05 | melwitt | thanks | |
| 01:34:42 | openstackgerrit | melanie witt proposed openstack/nova master: Add functional regression test for bug 1778305 https://review.openstack.org/582407 | |
| 01:34:43 | openstack | bug 1778305 in OpenStack Compute (nova) "Nova may erronously look up service version of a deleted service, when hostname have been reused" [Undecided,In progress] https://launchpad.net/bugs/1778305 - Assigned to melanie witt (melwitt) | |
| 01:34:44 | openstackgerrit | melanie witt proposed openstack/nova master: Don't generate service UUID for deleted services https://review.openstack.org/582408 | |
| 01:34:44 | openstackgerrit | melanie witt proposed openstack/nova master: Add functional regression test for bug 1764556 https://review.openstack.org/582435 | |
| 01:34:46 | openstack | bug 1764556 in OpenStack Compute (nova) ""nova list" fails with exception.ServiceNotFound if service is deleted and has no UUID" [Medium,In progress] https://launchpad.net/bugs/1764556 | |
| 01:36:38 | mriedem | Kevin_Zheng: i see you didn't change anything about the rpc api version handling in https://review.openstack.org/#/c/568542/ - i'll see if i can talk with dansmith about that tomorrow morning so we can sort that out (he's on vacation next week) | |
| 01:37:13 | mriedem | Kevin_Zheng: are you going to rebase https://review.openstack.org/#/c/573136/ back onto the top of that series? | |
| 01:38:02 | Kevin_Zheng | mriedem, yeah, I was planning to collect more ideas on version checks | |
| 01:39:29 | Kevin_Zheng | mriedem, I did all the updates to the API patch, but somehow the branch got messed up when I upload it, so I'm going to do it again today | |
| 01:39:30 | mriedem | i guess dansmith is saying the version checks would be redundant in the compute api and rpc api, | |
| 01:39:38 | mriedem | because if you can't send 5.1, that's all the check you need | |
| 01:39:50 | Kevin_Zheng | yeah, they are suggesting do it in API only | |
| 01:39:59 | Kevin_Zheng | I mean Alex and Gmann | |
| 01:40:03 | mriedem | dan is saying doing it all in rpcapi | |
| 01:40:10 | mriedem | b/c that's what has the can_send_version check | |
| 01:40:13 | mriedem | and knows the upgrade_levels pin | |
| 01:40:26 | mriedem | https://docs.openstack.org/nova/latest/configuration/config.html#upgrade-levels | |
| 01:40:41 | mriedem | the API could tell us that the compute is running version 33 but the rpc api might be pinned to rocky, | |
| 01:40:49 | mriedem | at which point can_send_version would still fail the 5.1 check | |
| 01:41:08 | mriedem | also good lord that's a wall of text ^ | |
| 01:41:08 | Kevin_Zheng | hmmm this doc helps | |
| 01:41:40 | gmann | mriedem: what is case where node has compute 33 and RPC API not pinned ? | |
| 01:42:37 | gmann | mriedem: Kevin_Zheng the only reason of checking in api we thought is early revert back the request. check in rpc is needed as you and dansmith mentioned | |
| 01:43:11 | mriedem | you can fail fast and return an error to the user in either place | |
| 01:43:17 | mriedem | since the check happens before the cast to compute | |
| 01:43:41 | mriedem | i'm trying to sort out if we can *move* the !running check from api to rpcapi | |
| 01:43:45 | mriedem | otherwise it's redundant | |
| 01:44:00 | mriedem | this https://review.openstack.org/#/c/568542/17/nova/compute/api.py@4427 | |
| 01:44:11 | gmann | mriedem: you mean from rpc to api ? | |
| 01:44:38 | mriedem | api to rpc | |
| 01:44:38 | gmann | agree to have check of !running in single place, either one | |
| 01:44:50 | gmann | humm | |
| 01:44:51 | Kevin_Zheng | Hmm, that is for microversion | |
| 01:44:52 | mriedem | nova.compute.api.API can't do the rpc version check | |
| 01:44:58 | mriedem | it doesn't have the can_send_version logic | |
| 01:45:05 | mriedem | that's in nova.compute.rpcapi | |
| 01:45:11 | Kevin_Zheng | I mean, we should check for microversion first | |
| 01:45:22 | mriedem | right that's what is confusing me | |
| 01:45:29 | mriedem | because we need to build in the microversion logic as well | |
| 01:45:35 | Kevin_Zheng | yeah | |
| 01:46:12 | mriedem | but couldn't we just pass support_abort_in_queue to rpcapi? | |
| 01:46:14 | mriedem | and do the logic there? | |
| 01:46:35 | mriedem | then your acceptable status values are based on (1) the microversion and (2) the rpcapi version | |
| 01:46:57 | mriedem | allowed_statuses = ['running'] | |
| 01:46:59 | mriedem | if support_abort_in_queue: | |
| 01:47:16 | Kevin_Zheng | and can_send_version? | |
| 01:47:20 | mriedem | yeah, | |
| 01:47:29 | mriedem | if support_abort_in_queue and can_send_version(5.1): | |
| 01:47:41 | mriedem | allowed_statuses.extend(['queued', 'preparing']) | |
| 01:47:51 | gmann | microversion check is on APOI controller side not enough ? | |
| 01:47:52 | Kevin_Zheng | Then should I merge patch 2 and 4? | |
| 01:47:53 | mriedem | if migration.status not in allowed_statuses: | |
| 01:47:54 | gmann | API | |