| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-02-08 | |||
| 18:33:09 | ganso | sean-k-mooney: hmmm my search didn't come up with anything. Do you remember why it was rejected? | |
| 18:33:11 | sean-k-mooney | ganso: we decieded not to change the existing behavior | |
| 18:34:15 | sean-k-mooney | bascilly becasue we dont really feel this type of api should be provided by nova and analtics of this type would be better done externally | |
| 18:34:26 | sean-k-mooney | by consuming the instance life cycle events | |
| 18:34:39 | sean-k-mooney | it came up in the nova meeting a few weeks ago | |
| 18:37:09 | sean-k-mooney | http://eavesdrop.openstack.org/meetings/nova/2020/nova.2020-12-03-16.00.log.txt | |
| 18:37:30 | sean-k-mooney | https://review.opendev.org/c/openstack/nova/+/711113 | |
| 18:37:34 | ganso | sean-k-mooney: shouldn't this API be deprecated? From what you said it sounds like you'd like to disencourage its usage | |
| 18:38:00 | sean-k-mooney | ganso: well we already do discuage usage :) but its not deprecated | |
| 18:38:39 | sean-k-mooney | we resovled to instead update the docs to call out its expected usage i think | |
| 18:40:08 | sean-k-mooney | ganso: this api has a pretty heavy perfomance impact too in comparision to outer apis so it does not scale well | |
| 18:41:39 | ganso | sean-k-mooney: thanks for the info, I will close the bug as invalid! | |
| 18:42:05 | sean-k-mooney | ganso: can you like to the previous meeting and that patch that was abandoned for context | |
| 18:42:15 | ganso | sean-k-mooney: yes, I will include those links | |
| 18:42:17 | sean-k-mooney | ganso: out of interest what was your usecase for this | |
| 18:42:32 | ganso | sean-k-mooney: it was fixing this bug: https://bugs.launchpad.net/nova/+bug/1913641 | |
| 18:42:32 | openstack | Launchpad bug 1913641 in OpenStack Compute (nova) "Incorrect Shelved_offloaded instance metrics on openstack usage show output" [Undecided,In progress] - Assigned to Rodrigo Barbieri (rodrigo-barbieri2010) | |
| 18:42:35 | ganso | see the last comment ^ | |
| 18:42:50 | ganso | I implemented the fix, but then I found that other issue | |
| 18:43:19 | sean-k-mooney | ah ok well that is likely invalid too | |
| 18:43:43 | ganso | yes. I will close that one too, because it is pointless fixing that without fixing the launched_at being reset | |
| 18:45:11 | sean-k-mooney | ya so i would clouse one of those two bugs as a duplicate of the other then close the scond with refrences to the meeting as wontfix or invalid | |
| 18:45:43 | sean-k-mooney | ganso: melwitt also commened on comment one that it was intended behavior too. | |
| 18:46:13 | sean-k-mooney | although that was more on the difference between counting quota with placement or not | |
| 18:47:25 | ganso | sean-k-mooney: my understading from her comment was that it was intentional in the past, but now with that new count_quota_from_placement that it could change, so I moved forward with a fix | |
| 18:48:03 | sean-k-mooney | not the intent is eventually to remove the current api and only count quota and usage via placment | |
| 18:48:14 | sean-k-mooney | that is part of teh unified limits work | |
| 18:48:50 | sean-k-mooney | ganso: placment has its own usage api https://docs.openstack.org/api-ref/placement/?expanded=list-usages-detail#usages | |
| 18:49:14 | sean-k-mooney | that wont give you historic usage over a time interval but it will tell you the current usage of a project | |
| 18:51:08 | ganso | sean-k-mooney: thanks! | |
| 19:07:29 | ganso | sean-k-mooney: I added the comment, but I lack the powers to mark both as "won't fix". I don't see them as duplicates, neither invalid ones. I do think they are legit, but considering the usage API is discouraged, they will not be fixed. Could you please mark both as "won't fix" ? | |
| 19:08:08 | sean-k-mooney | am sure. you need to be in the nova bug group. its an open team in launchpad so anyone can join but ill update them now | |
| 19:08:27 | ganso | oh cool, anyone can join. Will join them now | |
| 19:10:10 | sean-k-mooney | ganso: done if you want to join just join here https://launchpad.net/~nova-bugs | |
| 19:10:19 | ganso | sean-k-mooney: just did =) | |
| 19:10:40 | sean-k-mooney | cool i think you need to renew it every 1-2 years but its simple to do | |
| 19:36:43 | melwitt | ganso, sean-k-mooney: fwiw ganso's interpretation of my comment is correct, I think it could be debated whether usage should count shelved offloaded instances (same way it could be debated whether quota should count it). that said, I could see it making sense to keep both behaviors consistent (as long as default quota behavior counts shelved offloaded, so does usage). if/when default quota behavior turns to placement counting, usage | |
| 19:36:43 | melwitt | removes counting of shelved offloaded | |
| 19:38:23 | sean-k-mooney | melwitt: where you in the team meeting when https://review.opendev.org/c/openstack/nova/+/711113 was discuessed | |
| 19:39:02 | sean-k-mooney | that attempted to cahnge the meaning of the usage api to only cound it when the vm was active and running | |
| 19:39:05 | melwitt | no, I missed that | |
| 19:39:12 | melwitt | ah ok | |
| 19:39:48 | sean-k-mooney | we decieded not to basically because we did not want to change its current meaning | |
| 19:40:12 | sean-k-mooney | it could be done but we did not really want too do it as a bug fix | |
| 19:40:46 | sean-k-mooney | its really an api change and im not sure how open we are to adding more uses of or enhanceing the simple tenant usage api | |
| 19:40:49 | melwitt | thanks for that link, that helps | |
| 19:40:59 | sean-k-mooney | personally i would be in favor of eventually removing it | |
| 19:41:17 | melwitt | I see, was just reading through the irc discussion linked on the abandoned patch | |
| 19:41:39 | sean-k-mooney | we could revisit this dicussion | |
| 19:41:50 | sean-k-mooney | if you think there is merit in it | |
| 19:42:13 | sean-k-mooney | but it feels to me that this should be a spec if we were to change the api behavior | |
| 19:42:30 | sean-k-mooney | with a new microverion to request the new behaivor | |
| 19:43:42 | melwitt | eh.. I don't feel that strongly about it. I hadn't considered the broader view that the simple tenant usage api is obsolete | |
| 19:44:25 | sean-k-mooney | melwitt: do you know what the status of unified limits is by the way? | |
| 19:44:32 | melwitt | I'd be ok with tweaking it to make more sense but I'm also ok if most people think it should be left alone as obsolete | |
| 19:45:11 | melwitt | sean-k-mooney: no, I have seen johnthetubaguy around occasionally but haven't seen any activity on unified limits | |
| 19:46:00 | ganso | melwitt, sean-k-mooney: AFAIK it is widely used, despite its falws. If there is no intention in enhancing it further or addressing it flaws, IMO the best way is to deprecate it is usage. Surely other sources of info are better for billing, but a lot of cloud admins use it anyways for billing because it is there and it is convenient | |
| 19:51:26 | melwitt | yeah, that's been my understanding as well. another layer about it is that horizon uses it | |
| 19:59:46 | sean-k-mooney | ganso: i belive cloudkitty was ment for billing | |
| 20:00:05 | sean-k-mooney | melwitt: it does but they are the ones that told us the performacne sucked | |
| 20:00:12 | sean-k-mooney | they use it for the tenant overview page | |
| 20:01:15 | sean-k-mooney | ganso: https://github.com/openstack/cloudkitty | |
| 20:01:25 | sean-k-mooney | CloudKitty aims at filling the gap between metrics collection systems like ceilometer and a billing system. | |
| 20:01:26 | melwitt | yeah... the performance is tied to the number of deleted instances (it uses those to do the usage over time) | |
| 20:01:52 | sean-k-mooney | melwitt: yep have you seen the performace of that page on PSI downstream | |
| 20:01:58 | melwitt | yes | |
| 20:02:03 | sean-k-mooney | its not pretty | |
| 20:02:15 | melwitt | that's why I told them they really need to have their archive cron dialed in | |
| 20:02:35 | sean-k-mooney | well yes that is one reason for that | |
| 20:02:44 | melwitt | yes, one of many | |
| 20:02:52 | sean-k-mooney | did you see the os-archive topic on the mailing list | |
| 20:03:23 | sean-k-mooney | https://github.com/ovh/osarchiver | |
| 20:03:33 | melwitt | ah, yeah | |
| 20:04:03 | sean-k-mooney | assuimg that becomes an offila project we shoudl perhaps exploer if that would be a better approch | |
| 20:04:17 | sean-k-mooney | e.g. move all archiving external to the services | |
| 20:05:00 | sean-k-mooney | they apparently dont use shadow tabels at all and that operats on a retention policy archiving/deleteing deleted instance after a period of time | |
| 20:05:19 | melwitt | re: usage, I think that one is a challenge for operators because if you don't use the "easy" way of nova's tenant usage which just uses soft deleted instances, you need something else to store the info on instances from the past, which means another service as you said. so simple tenant usage appeals in that way | |
| 20:05:58 | sean-k-mooney | ya the placment verion is point in time not a historic view | |
| 20:06:59 | sean-k-mooney | that said im still not conviece this should be an api provide by nova or placment | |
| 20:07:06 | sean-k-mooney | its nice to have | |
| 20:07:09 | melwitt | I wonder if horizon would be ok to have the tenant overview be point in time only | |
| 20:07:30 | sean-k-mooney | if they are then plamcent solves that usecase | |
| 20:07:36 | melwitt | right | |
| 20:07:38 | sean-k-mooney | that is more or less the only think i used that for | |
| 20:08:07 | sean-k-mooney | melwitt: placement does not have soft delete or shaddow tabels for alocations right | |
| 20:08:13 | melwitt | right | |
| 20:08:23 | sean-k-mooney | that proably the right choice | |
| 20:08:33 | sean-k-mooney | but also means we cant look at deleted allcoations | |
| 20:09:04 | melwitt | yeah | |
| 20:09:52 | sean-k-mooney | at the end of the day i think a small service that subsribes to the instance events versioned notifications | |
| 20:10:18 | sean-k-mooney | and maintains the usage externally is proably the way to go | |
| 20:10:30 | sean-k-mooney | but ya extra service | |
| 20:10:36 | melwitt | yeah, I agree. I assume that's what cloudkitty did/does | |
| 20:11:29 | sean-k-mooney | melwitt: im not sure if it pulled form celomiter or not | |
| 20:11:45 | sean-k-mooney | it has a data colector | |
| 20:12:21 | sean-k-mooney | https://github.com/openstack/cloudkitty/tree/master/cloudkitty/collector | |
| 20:12:35 | sean-k-mooney | so it pulls form gnocchi monasca or prometheus | |
| 21:39:14 | openstackgerrit | Merged openstack/nova master: Add regression test for bug #1908075 https://review.opendev.org/c/openstack/nova/+/766976 | |
| 21:39:14 | openstack | bug 1908075 in OpenStack Compute (nova) "Nova allows a non-multiattach volume to be attached to multiple instances *if* its volume state is reset by an admin" [Low,In progress] https://launchpad.net/bugs/1908075 - Assigned to Lee Yarwood (lyarwood) | |
| 22:45:14 | openstackgerrit | Merged openstack/nova master: api: Reject volume attach requests when an active bdm exists https://review.opendev.org/c/openstack/nova/+/768472 | |