| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2025-04-03 | |||
| 13:50:02 | sean-k-mooney | that casuse the neutron ml2 driver to update the port | |
| 13:50:08 | sean-k-mooney | with extra data like the ip | |
| 13:50:53 | sean-k-mooney | however we dont know the mac adress of interface in nvoa unless we are looking that up form ironci or anytihng about the port groups again unless we are calling out to ironic | |
| 13:51:32 | sean-k-mooney | the ironic driver moved all fo that to ironic so they could have a singel code path for that in ironic for standaloen mode | |
| 13:52:07 | sean-k-mooney | mnaser: a simpler way to do this might be to modify ironic to just update the content of the config drive wehn it does all the networking setup | |
| 13:52:46 | sean-k-mooney | that doubling down on the wrong behavior but if they are modifying networkhing out of band then updating the metadta to reflect that makes sense | |
| 13:54:17 | dansmith | sean-k-mooney: https://meetings.opendev.org/irclogs/%23openstack-nova/%23openstack-nova.2025-04-01.log.html#t2025-04-01T14:09:06 | |
| 13:54:50 | dansmith | sean-k-mooney: to be fair, you were talking to gibi not me.. I did show up a few minutes later (first thing in the morning) asking gibi a question about a gerrit review, but I obviously didn't read back the scroll there | |
| 13:55:16 | sean-k-mooney | ya thats fair. | |
| 13:55:50 | sean-k-mooney | i expected Uggla to rely that into the release patch | |
| 13:56:11 | dansmith | also, what's not fair is how hard it is to grep logs for something sean-k-mooney said.. "grep sean-k-mooney.*release 2025*.log" .. no results :D | |
| 13:56:41 | sean-k-mooney | :) there use to be a hound instnace that someone ran that had all the irc logs | |
| 13:56:48 | sean-k-mooney | liek 8-10 year ago | |
| 13:56:54 | sean-k-mooney | it was kind of nice to be able to find thing | |
| 13:56:56 | dansmith | sean-k-mooney: honestly, every time I add something to os-traits I get really upset about how incredibly silly the process is that we've designed for getting string constants into nova | |
| 13:57:09 | sean-k-mooney | ya | |
| 13:57:27 | sean-k-mooney | so that was added by mikal as a ptg topic | |
| 13:57:33 | sean-k-mooney | askign if we can avoid the dance somehow | |
| 13:57:34 | dansmith | write nova patch, write os-traits patch, merge os-traits, release os-traits, merge upper-constraints, bump requirements.. only then can you go back to the first step and run tests on the nova patch :( | |
| 13:57:46 | sean-k-mooney | we coudl have nova just create the standard traits on start up if not present | |
| 13:58:02 | dansmith | sean-k-mooney: I proposed this per gmann to cut out one step at least https://review.opendev.org/c/openstack/requirements/+/946183 | |
| 13:58:31 | sean-k-mooney | hum that could help | |
| 13:58:45 | dansmith | feel free to +1 to show your support :D | |
| 13:58:46 | sean-k-mooney | we stil have an annoying coupleing dance with placeemtn too | |
| 13:59:06 | sean-k-mooney | this applies to os-ressouce classes too. | |
| 13:59:07 | dansmith | yes, honestly we should just squash placement and os-traits back into nova to solve it for good :D | |
| 13:59:40 | sean-k-mooney | well we solve it for us. but then we force neutorn et all to land a patch to nova to add a triat | |
| 14:00:07 | sean-k-mooney | what i was consdiring is allowing anytihng with a service role to add standard traits via the api | |
| 14:00:11 | dansmith | ...which is still less work than they have to do currently with os-traits, releases, requirements, etc :) | |
| 14:00:29 | dansmith | idk, that's a much larger step | |
| 14:00:55 | sean-k-mooney | it is | |
| 14:01:09 | sean-k-mooney | the other thing i codnered was sevador plugin | |
| 14:01:24 | sean-k-mooney | i.e. have an entry poirnt for addign traits | |
| 14:01:29 | sean-k-mooney | btu that feel more complex | |
| 14:02:47 | sean-k-mooney | dansmith: https://etherpad.opendev.org/p/nova-2025.2-ptg#L235 that where mikal still expressed there furstration with the current workflow | |
| 14:03:46 | dansmith | that's sort of a different complaint, but yeahj | |
| 14:09:27 | opendevreview | Masahito Muroi proposed openstack/nova master: Use dict object for request_specs_dict in the _list_view https://review.opendev.org/c/openstack/nova/+/939658 | |
| 14:11:29 | mnaser | sean-k-mooney: this all seems to be an ironic issue for now at the moment so maybe i'll take that there but thank you | |
| 14:12:39 | sean-k-mooney | mnaser: ack, we do have a related issue with the generic code path for port that never have ip adress | |
| 14:12:48 | sean-k-mooney | but that technially a differnt bug | |
| 14:13:28 | sean-k-mooney | we shoudl be providing the mac info from neutron if the ip_allcoation policy is none but that is nto what happens today | |
| 15:08:17 | dansmith | Uggla: by the way, on the otu session you scheduled.. even if we don't need to discuss otu specifically, we can discuss the placement bug fix that is related to it (and migration while over capacity) | |
| 15:08:30 | dansmith | that may be contentious (not sure) but might benefit from discussion at least | |
| 15:10:01 | Uggla | dansmith, sure. I will add a note about it. | |
| 15:10:03 | gibi | +1 to mention that | |
| 15:11:34 | sean-k-mooney | i think tehre are enough bugs related to it that it would be contorvisal | |
| 15:11:44 | sean-k-mooney | but sure good to highlight it in general | |
| 15:11:59 | dansmith | sean-k-mooney: ? | |
| 15:12:06 | sean-k-mooney | the placement bug fix | |
| 15:12:18 | sean-k-mooney | like we have at least 3 bugs form diffent people related to it | |
| 15:12:22 | dansmith | "enough bugs" makes it controversial? | |
| 15:12:40 | sean-k-mooney | sorry not controversial | |
| 15:13:00 | dansmith | ah | |
| 15:13:49 | sean-k-mooney | the approch to fixign it might be, i dont know but the fact the current bavhior is surpising to poeple is demonstarted by the exising bug reports | |
| 15:14:23 | dansmith | right, I meant the actual approach might be controversial | |
| 15:14:35 | dansmith | but I'd definitely like the opportunity to make my case for why I think this approach is the right one | |
| 15:14:56 | sean-k-mooney | i dont really see many other options to fixing it that would be backportable so that a vote in favor of the appoch | |
| 15:15:07 | sean-k-mooney | but ya lets see what others think next week | |
| 15:15:38 | dansmith | yeah, I agree of course.. but people may have religious/idealistic objections, idk | |
| 15:15:52 | sean-k-mooney | as an aside i might call it soon. im feellign progressively worse. soim going to take some panadol and get a hot drink if that does nto help ill call it a day | |
| 15:16:28 | dansmith | ack, melwitt just approved the traits patch | |
| 15:18:00 | opendevreview | Merged openstack/os-traits master: Add traits for sound and USB devices on instances. https://review.opendev.org/c/openstack/os-traits/+/940418 | |
| 15:24:51 | sean-k-mooney | cool melwitt ++ | |
| 15:31:31 | melwitt | sean-k-mooney: just in case you missed it, from your comment I added a repro func test under this and also a reno https://review.opendev.org/c/openstack/nova/+/880399 | |
| 15:32:36 | melwitt | note that the func test doesn't cover the "old" cinder API path, I wasn't sure if it needs to i.e. I wasn't sure if the old Cinder path is something we expect to be possible at this point/release | |
| 15:34:59 | sean-k-mooney | i had not seen that | |
| 15:35:03 | sean-k-mooney | but cool | |
| 15:35:45 | sean-k-mooney | on a related not i have a cinder volume multi attch pathc i need to clean up and push again.. | |
| 15:37:12 | sean-k-mooney | https://review.opendev.org/q/topic:%22bug/2048837%22 but its in merge conflict so no need ot look now. | |
| 15:37:38 | sean-k-mooney | there are a couple of volume related pathces in diffetn state aroudn at the moment | |
| 15:39:39 | melwitt | ack | |
| 15:40:26 | sean-k-mooney | weried os before we alwasy faked the volujme was attached even fi it was not assocated ot teh vm | |
| 15:41:51 | melwitt | what? :) | |
| 16:14:33 | sean-k-mooney | melwitt: https://review.opendev.org/c/openstack/nova/+/946222/1/nova/tests/fixtures/cinder.py#271 | |
| 16:14:42 | sean-k-mooney | i was commenting on the addtion of the if | |
| 16:16:17 | sean-k-mooney | not that what i orginally type convayed that particallary well | |
| 16:26:44 | melwitt | sean-k-mooney: oh haha | |
| 16:27:37 | melwitt | yeah the cinder fixture is a little wonky | |
| 16:47:34 | sean-k-mooney | +1 but im not entirlly sure how stable the looping is | |
| 16:47:41 | sean-k-mooney | comments inlien | |
| 16:48:05 | sean-k-mooney | we often use threading.evetns of locs when we want to encfoce a specific ordering | |
| 16:49:18 | sean-k-mooney | melwitt: im going to run your repoducer in a loop for a bit and see if its sable | |
| 16:49:24 | sean-k-mooney | melwitt: have you done that? | |
| 16:49:37 | melwitt | sean-k-mooney: no I didn't | |
| 16:50:50 | melwitt | I'll adjust things based on your comments. I haven't really done a func test like this before where it is more about checking the state of a resource in a different component, so I may have done it not the best way | |
| 16:51:36 | sean-k-mooney | well i think yoru test is fine. i just know that we have had issues with retry loops in fucntional tests before | |
| 16:51:49 | sean-k-mooney | so if we can force the ordering more explcitly that can help | |
| 16:52:10 | sean-k-mooney | ill let this run for a while and if it passes 50-100 times its proably ok | |
| 16:52:49 | sean-k-mooney | i think we can impove things if we reuse some of the helpers but im no tsure we strictly need too for this to prove what it needs too | |
| 16:53:36 | sean-k-mooney | ya the cidner fixutre behvaior is a litttel odd in places | |
| 16:53:50 | sean-k-mooney | but its also non trivial to emulate os im not really surpised | |
| 16:54:24 | sean-k-mooney | melwitt: ah https://review.opendev.org/c/openstack/nova/+/919979/4/nova/tests/functional/integrated_helpers.py | |
| 16:54:56 | sean-k-mooney | i _attch_volume for my multi attach repoducer | |
| 16:55:01 | melwitt | yeah I think my concern with it is if it could possibly "skip" the status "detaching" due to timing differences. so I won't be surprised if need to do the wait a different way | |
| 16:55:12 | sean-k-mooney | i was very confused why i did nto find it when i went looking | |
| 16:56:29 | sean-k-mooney | melwitt: so there are some exmples form gibi useing treading.event | |
| 16:56:36 | sean-k-mooney | to get thisn to pause in a specific state | |
| 16:56:41 | melwitt | oh my bad, I either didn't see that or there was some reason it didn't work for me and I forgot. but more likely the former | |
| 16:56:42 | sean-k-mooney | mainly for networking i think | |
| 16:57:01 | sean-k-mooney | no _attach_volume is not merged | |