Thursday, 2026-03-05

manideep#openstack-neutron03:41
opendevreviewNope proposed openstack/nova master: scheduler: optional re-query on claim failure  https://review.opendev.org/c/openstack/nova/+/97790803:58
opendevreviewDominik proposed openstack/nova master: Regression test for Placement allocations remaining during failed schedule  https://review.opendev.org/c/openstack/nova/+/96925108:41
opendevreviewDominik proposed openstack/nova master: Remove Placement allocations in the broken build cleanup  https://review.opendev.org/c/openstack/nova/+/96844608:41
opendevreviewDominik proposed openstack/nova master: Regression test for Placement allocations remaining during failed schedule  https://review.opendev.org/c/openstack/nova/+/96925110:36
opendevreviewDominik proposed openstack/nova master: Remove Placement allocations in the broken build cleanup  https://review.opendev.org/c/openstack/nova/+/96844610:37
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Preserve vTPM state between power off and power on  https://review.opendev.org/c/openstack/nova/+/97890810:38
opendevreviewDominik proposed openstack/nova master: NUMA Topology with Resource Providers: Scheduler changes  https://review.opendev.org/c/openstack/nova/+/97117613:04
opendevreviewribaudr proposed openstack/nova master: Update compute rpc alias for Gazpacho  https://review.opendev.org/c/openstack/nova/+/97900814:31
opendevreviewribaudr proposed openstack/nova master: Add service version for Gazpacho  https://review.opendev.org/c/openstack/nova/+/97901614:57
opendevreviewribaudr proposed openstack/nova master: Add Gazpacho prelude section  https://review.opendev.org/c/openstack/nova/+/97902315:19
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Preserve vTPM state between power off and power on  https://review.opendev.org/c/openstack/nova/+/97890815:35
nicolairuckelHow much am I allowed to deviate from an original patch for a backport? I needed to make two small adjustments that were not part of the original patch but came from some earlier patches.15:54
*** sambork_ is now known as sambork15:55
dansmithnicolairuckel: as little as possible16:05
dansmithcall out the fixes in the commit message16:06
elodillesnicolairuckel: as little as possible o:) if the conflict is due to another bug is not backported, then the easiest is to backport that as well16:06
dansmithif it's too invasive, that's grounds for saying "maybe this isn't suitable for backporting"16:06
elodillesjinx16:06
opendevreviewStephen Finucane proposed openstack/nova master: objects: Remove unnecessary type aliases, exceptions  https://review.opendev.org/c/openstack/nova/+/73824016:47
opendevreviewStephen Finucane proposed openstack/nova master: objects: Use imports instead of type aliases  https://review.opendev.org/c/openstack/nova/+/73801816:47
opendevreviewStephen Finucane proposed openstack/nova master: objects: Remove wrappers around ovo mixins  https://review.opendev.org/c/openstack/nova/+/73801916:47
opendevreviewStephen Finucane proposed openstack/nova master: objects: Remove 'NovaObjectDictCompat' from 'Service'  https://review.opendev.org/c/openstack/nova/+/83559516:47
stephenfingmaan: sean-k-mooney: What do we want to do about the failures on https://review.opendev.org/c/openstack/python-novaclient/+/978254?16:54
stephenfinnova docs state that nova only supports the 'default' quota class set and the JSON schemas now enforce this. However, based on that now-failing test, it looks like this wasn't enforced previously and non-'default' quota class sets were allowed even if they did nothing16:56
stephenfinMy guess is that we need to back out the schema restriction that prevents us using non-default values but I just wanted to check you were happy with me doing so16:56
opendevreviewStephen Finucane proposed openstack/nova master: api: Don't restrict return value of quota class set name  https://review.opendev.org/c/openstack/nova/+/97903117:00
stephenfingmaan: sean-k-mooney: ^17:00
stephenfinah, tkajinam already beat me to it https://review.opendev.org/c/openstack/nova/+/97849417:02
sean-k-mooneyim aware and i dont think that is the correct fix17:10
sean-k-mooneyits a workaorund but you were correct in restriting the resopcne schema, the supprot for classes other then defualt was an api exstion and it was removed in ocata17:11
sean-k-mooneythe respocne schdmea validation expsoed and input validate bug17:12
stephenfinright, but restricting that would be an API change17:13
stephenfinand therefore needs a microversion17:13
sean-k-mooneythe ablity to use this was remvovd with the ablity to confirue out of tree quota drivers17:13
stephenfinright, but the API still works17:14
sean-k-mooneyit does not need a microverion if we retrun an existing error code that we allow in our api17:14
stephenfineven if it does nothing17:14
sean-k-mooneyso no this does not need a microverison17:14
stephenfinAssuming that's true, can we merge that patch then work on the request schema change? If we *do* allow that, I suspect we also want to go and delete all the non-default entries from the database too17:15
sean-k-mooneyyes17:16
sean-k-mooneyim not sure we want to clean up the db17:16
stephenfinLooks like the quota class show API doesn't have any expected error codes :/ https://github.com/openstack/nova/blob/968965cc77c6932eb67cc36a539f108c73d50ea0/nova/api/openstack/compute/quota_classes.py#L8917:17
sean-k-mooneyhttps://docs.openstack.org/nova/latest/contributor/microversions.html#id317:18
sean-k-mooney"""17:18
sean-k-mooneyThe exception to not needing a microversion when returning a previously unspecified error code is the 400, 403, 404 and 415 cases. This is considered OK to return even if previously unspecified in the code since it’s implied given keystone authentication can fail with a 403 and API validation can fail with a 400 for invalid json request body. Request to url/resource that17:18
sean-k-mooneydoes not exist always fails with 404. Invalid content types are handled before API methods are called which results in a 415."""17:18
sean-k-mooneythis should be a 40017:18
stephenfinthere wasn't a previously unspecified error code though17:18
stephenfinas in, we don't see HTTP 50017:19
stephenfinwhich I thought was the only case where a microversion wasn't needed17:19
sean-k-mooneythat isnt requried17:19
stephenfin(that and returning HTTP 410 (Gone) where we could no longer support a service)17:19
sean-k-mooneywe have estiblishted several time sthat you shoudln not need ot opt into a bug fix17:19
sean-k-mooneystephenfin: by the way if you read the comment thread https://review.opendev.org/c/openstack/nova/+/978494/comments/6ba93b6c_a0707c41 i said i was ok if we add a fixeme comment to adress this proplry later17:22
sean-k-mooneyi would also prefer if it was not closes-bug but related-bug17:22
sean-k-mooneybuecause it does not close it in my view17:22
sean-k-mooneybut i care about that less17:22
opendevreviewStephen Finucane proposed openstack/nova master: WIP: api: Block non-default quota class sets  https://review.opendev.org/c/openstack/nova/+/97903317:27
stephenfinsean-k-mooney: mock up of what I think you're after ☝️17:27
sean-k-mooneyya basiclly i was also debating if we woudl encoded it in the open api schema as well17:28
nicolairuckeldansmith, elodilles: I'm not sure where exactly the changes came from though. One is this (https://opendev.org/openstack/nova/commit/650772d97efe98e578c18b8268b3bbed63ac4d18) and the other one removed a line so I can't just use git blame (or I just don't know how)17:28
sean-k-mooneystephenfin: but something like that 17:29
stephenfinwe don't currently have schemas for request parameters (only ironic does for now), but if we did then that should be added, yes17:29
stephenfin*request path parameters17:30
sean-k-mooneyah ok17:30
sean-k-mooneythen ya ignore that for now17:30
nicolairuckelah, this is the other one: https://opendev.org/openstack/nova/commit/880019baafe23f46e59dee7debadf49b9177dc7017:34
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Preserve vTPM state between power off and power on  https://review.opendev.org/c/openstack/nova/+/97890817:40
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Preserve vTPM state between power off and power on  https://review.opendev.org/c/openstack/nova/+/97890817:40
nicolairuckelI added my additional changes to the commit message. I'm not sure if it would make sense to backport those two patches as well.17:47
nicolairuckelBut I guess that can be discussed in the patch. :)17:47
sean-k-mooneyi commeted in the patch17:49
sean-k-mooneyyou shoudl backport https://review.opendev.org/c/openstack/nova/+/961734/4 and the fix https://review.opendev.org/c/openstack/nova/+/96200717:49
sean-k-mooneyat least that woudl be my inclination as i think we will want to backprot that to fix https://bugs.launchpad.net/nova/+bug/2125030 ingeneral17:51
sean-k-mooneymelwitt: is there a reason ^ has not been backported beyound time?17:51
melwittsean-k-mooney: no there is no reason and in fact I think it should be backported. there is also a func test patch that goes with it17:53
sean-k-mooneyyep nicolairuckel was pullingin part of the funtional test patch17:53
melwittI just didn't think of it at the time17:53
sean-k-mooneyto resolve a conflict in backing the fix for vtpm rebot17:53
yuval_hey I checked the iothread thing - machine show no difference in iops. is there some flag I need to enalble? I see the iothread in the xmldump17:54
melwittyeah that bug fix should be backported bc that bug can permanently damage a vtpm instance if a resize revert is done. it will destroy the barbican secret and then it's gone forever17:54
yuval_talking with claud - we are using not in the right way we need to assign the iothread to a block device17:54
sean-k-mooneyyuval_: the cange that was done this cycle we expect to not significaly increate th iops but rather impove the managablity of the vm (attach/detach infaces, start stop ectra) whne the qemu process is under load from high io17:56
sean-k-mooneythe disk io is still limited to the perfromace of 1 thread 17:56
sean-k-mooneybut now that thread is not also runign the qemu monitor and other background tasks in addtion to io17:56
sean-k-mooneyso it shoudl impove a litell but not massively17:57
sean-k-mooneyyuval_: the block deivces will be assigne dby qemu automaticly17:57
sean-k-mooneyyuval_: its possibel to futher increase perfomce if we do allcoate more io thread and provide a dedicated one or multiple to block devices17:58
sean-k-mooneythat left for future work in the upcoming cycles17:58
yuval_I have tested iothreads in the past - it can improve performance alot17:58
yuval_thats why I am testing it - but the xml configuration was difference17:59
yuval_I need to assign iothread to the block device17:59
yuval_its a question how we assign those threads when block device is attach/detach18:00
sean-k-mooneyyep, and as i said we agred not to do that for this release18:00
yuval_got it thanks18:01
yuval_I will try to dig in see how that might be possible18:02
sean-k-mooneythe issue is there are several ways to do it with diffeern lvel of complexity18:02
sean-k-mooneymy current inlcidation is to keep it simple and use the defautliothread element https://libvirt.org/formatdomain.html#iothreads-allocation18:03
sean-k-mooneyalhtough you can also map disks indivutally18:03
sean-k-mooneyor at least storage contolers in the case of virtio scsi18:04
sean-k-mooneywe could for now explcitly add the iothread affinity to iothrad 1 to all disk18:05
sean-k-mooneybut i belive that shoudl be the effecite behvior we alerady get when we say nothign in this case18:05
sean-k-mooneyyuval_: if you fidn that is not the case file a bug and we can look to expclity do that mappign18:07
yuval_I will do some extra tests18:09
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Add functional reproducer for bug 2125030  https://review.opendev.org/c/openstack/nova/+/97904118:19
nicolairuckelsean-k-mooney: First backport is ready18:42
nicolairuckelSince they depend on each other I have to wait for the next one.18:42
dansmithnicolairuckel: no, you should stack them all together18:44
dansmithnot squash, but stack18:44
dansmithmeaning, create your topic branch, cherry-pick the first, then the second, then the third.. then git-review and it will say "Submitting 3 commits, right?"18:45
dansmiththat's what you want, so we can see them all together as they'll land18:45
nicolairuckeloh, I see18:46
gmaansean-k-mooney: stephenfin on non-default quota class, if it was allowed then schema change makes it 200->400 which is an API change. that is why i commented on tkajinam fix that let's fix the test + revert the schema change and later we will see how we can improve the API by seeing what exactly we are changing for users18:52
gmaansean-k-mooney: stephenfin: I am ok to land the nova fix but we need to fix the novaclient test also which is wrong https://review.opendev.org/c/openstack/python-novaclient/+/97825418:55
nicolairuckeldansmith: I'm running the tests now and if that's successful I'll push my branch.18:59
gmaantkajinam: stephenfin +w on nova fix but let's fix the novaclient test also to test the 'default' quota class https://review.opendev.org/c/openstack/nova/+/97849418:59
opendevreviewZhan Zhang proposed openstack/nova-specs master: Allow Operators to Launch VMs on Disabled Hosts  https://review.opendev.org/c/openstack/nova-specs/+/97904919:07
bryanfraschettiHi folks, I hope you're doing well! A few weeks ago I mentioned this bug: https://bugs.launchpad.net/nova/+bug/2141355, and was advised to try restarting the instances with the problematic volume attachments as well as try nova-manage volume_attachment refresh while the instances are turned off. I have found that this did not work as expected and my best guess is that the cleanup doesn't quite work because the 19:17
bryanfraschettiattachment record is not present in the cinder database at all.19:17
bryanfraschettiThe current way out of this situation is to modify the db and mark the volume attachment as deleted in the Nova block_device_mapping table. To avoid manual DB access, I'm working to add this functionality to nova-manage via a command such as nova-manage volume_attachment remove <server_uuid> <volume_uuid>. I'd like to upstream the functionality and am wondering if there's any glaring problems with this idea or if 19:17
bryanfraschettithere would otherwise be interest in accepting this functionality19:17
*** erlon3 is now known as erlon19:19
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Add functional reproducer for bug 2125030  https://review.opendev.org/c/openstack/nova/+/97904119:19
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Preserve vTPM state between power off and power on  https://review.opendev.org/c/openstack/nova/+/97890819:19
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Move cleanup of vTPM secret from driver to compute  https://review.opendev.org/c/openstack/nova/+/97905219:19
nicolairuckeldansmith: Like this?19:19
melwittnicolairuckel: the stack is good but on the last backport afaict there was a manual merge conflict resolution needed and those are to be referenced in the commit message of the backport i.e. https://docs.openstack.org/project-team-guide/stable-branches.html#conflicts I will mention this on the review19:27
nicolairuckelinteresting, I didn't merge anything manually there19:36
nicolairuckelI wonder if its because of some of my Git settings.19:37
melwittyeah I noticed after I replied to you I see that the merge conflict content is included in your backport which it shouldn't be. so I wonder what kind of thing could have caused that to happen automatically? I have not experienced something like that before19:39
nicolairuckelme neither19:42
nicolairuckelI need to experiment a bit with that tomorrow19:42
nicolairuckelbut for now I want to just fix the patch :D19:42
opendevreviewNicolai Ruckel proposed openstack/nova stable/2025.2: Preserve vTPM state between power off and power on  https://review.opendev.org/c/openstack/nova/+/97890819:42
melwittsure. please do look into it later :)19:43
nicolairuckelThank you for looking into the patch19:44
nicolairuckelI'll be busy with dinner now but I'll check later again.19:44
opendevreviewZhan Zhang proposed openstack/nova-specs master: Allow Operators to Launch VMs on Disabled Hosts  https://review.opendev.org/c/openstack/nova-specs/+/97904920:43

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!