Tuesday, 2026-06-23

« Monday, 2026-06-22 Index Wednesday, 2026-06-24 »
opendevreviewsean mooney proposed openstack/cyborg-specs master: Add consistent and secure RBAC spec for Cyborg  https://review.opendev.org/c/openstack/cyborg-specs/+/99193213:17
opendevreviewMerged openstack/cyborg-specs master: Add consistent and secure RBAC spec for Cyborg  https://review.opendev.org/c/openstack/cyborg-specs/+/99193213:37
sean-k-mooneyfyi we will start the meeting at the top of the hour13:53
sean-k-mooney#Start-Meeting14:01
sean-k-mooney#start-meeting14:01
sean-k-mooneyah no -14:01
raukadahit should startmeeting cyborg with #14:01
sean-k-mooney#startmeeting14:01
opendevmeetsean-k-mooney: Error: A meeting name is required, e.g., '#startmeeting Marketing Committee'14:01
sean-k-mooneyping list sean-k-mooney amoralej bogdando jgilaber rlandy chandankumar14:02
*** raukadah is now known as chandankumar14:02
sean-k-mooneywe will give everyone a minitue to join14:02
rlandyo/14:02
jgilabero/14:02
chandankumaro/14:02
skovilio/14:02
sean-k-mooneyas alwasys if you have topic feel free to add them to the adgenda14:03
sean-k-mooney#link https://etherpad.opendev.org/p/openstack-cyborg-irc-meeting14:03
melwittI don't think the meetbot started it (have to give meeting name when starting)14:03
chandankumarsean-k-mooney: we need to re-run the command for start the meeting `#startmeeting cyborg`14:03
sean-k-mooneyoh right14:03
sean-k-mooney#startmeeting cyborg14:03
opendevmeetMeeting started Tue Jun 23 14:03:35 2026 UTC and is due to finish in 60 minutes.  The chair is sean-k-mooney. Information about MeetBot at http://wiki.debian.org/MeetBot.14:03
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.14:03
opendevmeetThe meeting name has been set to 'cyborg'14:03
sean-k-mooney#link https://etherpad.opendev.org/p/openstack-cyborg-irc-meeting14:03
sean-k-mooneyi think we are all here now14:04
sean-k-mooneyso first topic14:04
sean-k-mooney#topic abandoned specs14:04
amoralejo/14:04
sean-k-mooneyso i did a clean up of olad specs a while ago and i decied to continue that today14:04
sean-k-mooney3 specs of note 14:05
sean-k-mooneyRefactor API Structure Proposal14:05
sean-k-mooney#link https://review.opendev.org/c/openstack/cyborg-specs/+/841077/3/specs/zed/approved/refactor-api-structure.rst14:05
sean-k-mooneyAdd NVIDIA MIG management spec14:05
sean-k-mooney#link https://review.opendev.org/c/openstack/cyborg-specs/+/88577814:05
sean-k-mooneyAdd api to support change device vgpu_type14:05
sean-k-mooney#link https://review.opendev.org/c/openstack/cyborg-specs/+/88543914:05
sean-k-mooneythese are all 2-3 years old and while some of the fucntionaltiy or refactor may be desireable14:06
sean-k-mooneyi closed thse as i think we will need to aproch them form first principals 14:06
sean-k-mooneythe nvidia mig management is actully kind fo superceaded14:07
sean-k-mooney#link https://github.com/openstack/cyborg-specs/blob/master/specs/2026.2/approved/support-vfio-variant-driver-managed-mode.rst14:07
sean-k-mooneythe basic supprot for mig mode will be enabled by genericly supprotign vfio-variant drivers14:08
sean-k-mooneyi do think we should spend some time before the next ptg considering larger api refactoring but again i think that need some careful tought so i have abandoned that spec for now14:09
sean-k-mooneyany concerns with that?14:09
jgilabernot really, all those specs were quite short and looked like wip anyway14:09
jgilaberif we need them at any point we could recover them14:09
sean-k-mooneywe have them in gerrit anyway so we can alwasy find them again if needed14:09
sean-k-mooneyexactly14:10
sean-k-mooneyok if there is noitng else lets move on to the next topic14:10
sean-k-mooney#topic spec freeze14:10
sean-k-mooneytradtionally for projects that use spec the default deadlien is milestone 214:11
sean-k-mooney#link https://releases.openstack.org/hibiscus/schedule.html#hibiscus-2-milestone14:11
sean-k-mooneythat would be july 2nd this year14:11
sean-k-mooneysince we have not elected a difent date im prposing we work towards that goal adn we can review the status of the spec/blueprints in the irc meeting 14:12
chandankumaroh that is one week from now14:12
sean-k-mooneyon july 7th14:12
sean-k-mooneyyes thrusday week14:12
sean-k-mooneyok if there is no objection lets proceed with that plan14:14
sean-k-mooneyand we will check in 2 weeks form now in the irc meeting14:14
chandankumarsounds good to me!14:14
jgilaber+114:15
sean-k-mooney#agreed spec and blueprint freeze will be july 2nd (milestone 2) and we will check in in the july 7th irc meeting14:15
sean-k-mooney#topic completed spec14:15
sean-k-mooneyatirbutes api14:15
sean-k-mooney#link https://review.opendev.org/c/openstack/cyborg-specs/+/896076 14:16
sean-k-mooneyso this is more or less just paperwork and markign it implemented14:16
sean-k-mooneyit was added in 2023.214:16
sean-k-mooneythe reason i call thi sout is i think we still could do with more tempest testing of this14:16
sean-k-mooneyso we may decied to file a tracker bug for the tempst plugin14:17
sean-k-mooneybut for all practical cases this is done so i appoved https://review.opendev.org/c/openstack/cyborg-specs/+/896076 to reflect that14:17
sean-k-mooneyi think we can move on to reviews14:18
sean-k-mooney#topic reviews14:18
sean-k-mooneyok first one consistent and secure RBAC spec for Cyborg14:19
sean-k-mooney#link https://blueprints.launchpad.net/openstack-cyborg/+spec/consistent-and-secure-rbac14:19
sean-k-mooney#link https://review.opendev.org/q/topic:%22bp/consistent-and-secure-rbac%22 14:19
sean-k-mooneychandankumar: and jgilaber: have reviewd the spec so i approved it to merge this morning14:19
sean-k-mooneyi have a complete impleaiton aviable and im hoping to merge that in the next 1-2 weeks.14:20
sean-k-mooneyobviously spec freeze has precedence so we should focus on those14:20
sean-k-mooneybut if you have extra review bandwith feedback is welcome14:20
sean-k-mooneyany questions on the topic of SRBAC that anyone want to ask now?14:21
jgilaberI hope to start reviewing this series this week14:21
sean-k-mooneyi kicked of my review ci on it overnight14:22
sean-k-mooneyim gogn to go though and triage its fineidng later in the week14:22
sean-k-mooneyproably firday14:22
opendevreviewMerged openstack/cyborg-specs master: move attribute spec to impletment  https://review.opendev.org/c/openstack/cyborg-specs/+/89607614:22
sean-k-mooneyso ill also look at any other feedback14:22
sean-k-mooneyi expect at leat one revison in the next week14:22
sean-k-mooneywhile i belive its mergable as is im sure there are things that can be impvoed14:22
sean-k-mooneyone thing ill call out14:23
sean-k-mooneyhttps://review.opendev.org/c/openstack/cyborg/+/99272214:23
sean-k-mooneythe first patch is baisclly dead code removal and a minor bug fix14:23
sean-k-mooneyso i pulled that out to the front14:23
sean-k-mooneythe rest is the actual srbac implementation14:24
sean-k-mooneyi think we can move on ?14:25
sean-k-mooney#topic Add generic NVMe driver spec with secure cleanup14:25
sean-k-mooney#link https://review.opendev.org/c/openstack/cyborg-specs/+/985349/14/specs/2026.2/approved/generic-nvme-driver-with-secure-cleanup.rst14:25
sean-k-mooneythis is chandans feature. i hope to do another pass on that on thrusday14:25
sean-k-mooneychandankumar: anything you want to raise on this topic?14:26
sean-k-mooneychandankumar: i dont think you have updated this with the dicussion we had yestready correct?14:26
chandankumarnope, I hope to finish the cleanup rewrite with clear_mode and clear_method by today14:26
sean-k-mooneyack14:27
sean-k-mooneyi guess if there are no questions we can move on14:27
sean-k-mooney#topic generic mdev driver14:27
sean-k-mooney#link https://review.opendev.org/c/openstack/cyborg-specs/+/98227614:27
sean-k-mooneyjgilaber: anythign you want to rasie on your spec?14:28
sean-k-mooneyim hoping to review that tomorow if its ready for review?14:28
opendevreviewchandan kumar proposed openstack/cyborg-specs master: Add generic NVMe driver spec with secure cleanup  https://review.opendev.org/c/openstack/cyborg-specs/+/98534914:28
jgilabernothing to raise, I think I addressed all comments 14:28
chandankumarOk I have updated it here, But a one more pass on my side before asking for review14:29
jgilaberthe only point left I think is how to handle the multiple mdev types with placement14:29
sean-k-mooneyack anythin you want to call out on the nova side?14:29
jgilaberon the nova side the mtty series is close to merge I think14:30
jgilaberit's working well in CI14:30
jgilaberand I pushed a patch for the OWNER_NOVA trait14:30
jgilaber#link https://review.opendev.org/c/openstack/nova/+/99429914:30
jgilaberon the last one I'm still working on the tests, but hopefully will be ready for review by the end of the week14:30
sean-k-mooneyour initall plan was to reused that for the cyborg ci. we may want to revaluate if we will instead port mdev supprot into the pci sim14:31
sean-k-mooneybut i think we choudl continue with the inial plan for now14:31
sean-k-mooneyack ill add the owner trait patch to my review list14:31
jgilaberyes, we can evaluate once I'm done with the nova part if we want to use pci sim, but for now using the nova devstack plugin should work and be easy to use14:32
sean-k-mooneywe may need to think about the upgrade impact of https://review.opendev.org/c/openstack/nova/+/994299/2/nova/compute/pci_placement_translator.py14:32
sean-k-mooneythat likely need to be gated behind a min compute service version check14:33
sean-k-mooneywell no14:33
sean-k-mooneythat where your addign the trait14:33
sean-k-mooneythat  proably fine14:33
jgilaberthe pre filter is behind a service check14:33
jgilaberadding the trait should be fine I think since it would not prevent any allocation14:33
sean-k-mooneyah yes https://review.opendev.org/c/openstack/nova/+/994299/2/nova/scheduler/request_filter.py14:33
sean-k-mooneyok ill do a proper review later14:34
jgilaberack, thanks14:34
sean-k-mooneyfor the prefilter i think we want to add @functools.cache to the min version check14:34
sean-k-mooneyan all cell scatter gather on ever schdule is expensive14:35
sean-k-mooneybut we can do it once per scdhuler restart14:35
sean-k-mooneywe woudl just need to document that in the upgrade release note14:36
sean-k-mooneyi.e. that a schduler restart is needed after all compute are upgraded14:36
sean-k-mooneybut that kind of needed anyway for other easons14:36
sean-k-mooneyim not conviced the prefilter is correct by the way14:37
jgilaberack, I think I'll need to update the release note for that14:37
sean-k-mooneywe can go into this in more detail in teh review14:37
jgilabersure14:37
sean-k-mooneyeffectivly you will ned to include the triat request in the request group14:37
sean-k-mooneynot just root reqiures 14:37
sean-k-mooneybut ya lets move to bugs unless there are other questions?14:37
sean-k-mooneymelwitt: ^ might be good to get your input on too14:38
sean-k-mooney#topic bugs14:38
sean-k-mooneyfirst one14:39
sean-k-mooney#link https://bugs.launchpad.net/openstack-cyborg/+bug/215758614:39
sean-k-mooneyCyborg Device Profile API rejects OWNER_CYBORG trait14:39
sean-k-mooneythis came up while reviewing chandands spec i belvie14:39
sean-k-mooneychandankumar: do you want to expand on it14:39
chandankumarit came from here https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/992210/4//COMMIT_MSG14:40
sean-k-mooneyoh yes your right14:40
chandankumarI was adding scenario test for pci driver to create device profile there I found that14:40
sean-k-mooneyso the tl;dr is14:40
sean-k-mooneycybrog device profiles14:41
sean-k-mooneycannot currently request any standard traits14:41
sean-k-mooneysuch as OWNER_CYBORG14:41
sean-k-mooneythis is a limitation that we should adress14:41
chandankumarhttps://github.com/openstack/cyborg/blob/master/cyborg/api/controllers/v2/device_profiles.py#L21114:41
sean-k-mooneyhttps://docs.openstack.org/api-ref/accelerator/#create-device-profile14:42
sean-k-mooneyif we look at the api ref14:42
sean-k-mooneythis is the descripton of the group field14:43
sean-k-mooney"""This is a list of dictionaries to describe returned accelerator resource by users, where users use keys to describe the resource_classes or traits and values to indicate its quantity or property. This is intentionally similar to extra_specs in nova flavor, and uses the same keywords for resources and traits. The key-value pair can either be a resource/trait or a Cyborg14:43
sean-k-mooneyproperty. Cyborg property is of the form “accel:<key>”: “<value>”. The valid key-value pairs can be found below."""14:43
sean-k-mooneythe grousp is a dict of string key to string value14:43
sean-k-mooneyso from an api poirnt of view this is unversioned14:43
sean-k-mooneyand the limitation to CUSTOM_ triats is undocuemnted14:44
sean-k-mooneythere is some docuemition fo the valid accel: keys https://opendev.org/openstack/cyborg-specs/src/branch/master/specs/train/implemented/device-profiles.rst#valid-accel-keys14:44
sean-k-mooneyin the orginal spec14:44
sean-k-mooneywhic we really shoudl move into the api ref14:45
sean-k-mooneybut i think we can treat this as a valid bug14:45
sean-k-mooneydo others agree?14:45
sean-k-mooneynote requesting a standard trait via the device profile i think is ok. allowing you to add standard trait via the atirbute api is not14:46
sean-k-mooneyor at least not genericlly ok so this si really about the request14:46
sean-k-mooneyauto requestign OWNER_CYBORG while releated is also sperate 14:46
chandankumaryes, you mean to say device profile should accept any triat14:46
sean-k-mooneyas that has an upgrade impact14:46
sean-k-mooneyyes14:46
sean-k-mooneyand if the triat does not eixsti in the delooymen then you will just get a schduler fialure14:47
sean-k-mooneybut we shoudl jsut check it conforms to placment trait nameign shape14:47
chandankumargood idea, first check about trait existance and then accept14:48
sean-k-mooneyi.e [A-Z_]* 14:48
sean-k-mooneywell i dont think we need to check with placmeent if the trait exists although we could14:48
sean-k-mooneybut we should validate that it conforms to placment requirement for the name of a valid triat14:49
chandankumarok14:49
chandankumarI will propose a patch14:49
sean-k-mooneywe coudl decied if we ant to check with the placment api in teh implemation review14:49
sean-k-mooneywoudl folks agree this shoudl be medium?14:49
chandankumaryes14:50
jgilaberyes, since this can be worked around14:50
sean-k-mooneyok done feel free to assign this to yourself if you start working on it14:50
sean-k-mooneynext bug14:51
sean-k-mooney#topic Set literals used instead of dicts in placement_client.py LOG calls14:51
sean-k-mooney#link https://bugs.launchpad.net/openstack-cyborg/+bug/215789214:51
sean-k-mooneymelwitt: you found this one14:51
melwittoh yes that was opened by me yesterday. just something I noticed while going over code in general14:52
sean-k-mooneyhttps://github.com/openstack/cyborg/blob/master/cyborg/common/placement_client.py#L31314:53
sean-k-mooneythat looks pretty valid to me14:53
jgilaberyes, this looks valid and a good find14:53
melwittyeah that should also be "%(rc_name)s" there too14:53
sean-k-mooneyits even correct like 4 lines up14:54
sean-k-mooneyhttps://github.com/openstack/cyborg/blob/master/cyborg/common/placement_client.py#L30814:54
melwittheh yeah. not sure what happened there14:54
sean-k-mooneyim feeling like traged, low14:54
sean-k-mooneygiven its only a log issue14:54
melwitt++14:54
jgilaber+114:55
sean-k-mooneycool i tagged it low-haning-fruit as well14:55
sean-k-mooneymelwitt: so ya if you paln to work on it feel free to assign it to your self as normal but anything else on this?14:55
sean-k-mooneyif not we will move on14:56
melwittsure, no that was all on this14:56
sean-k-mooneyi need to file a bug for this but ill just note it as an fyi14:56
sean-k-mooneyi found out that our api microverison parsing is incorrect14:57
sean-k-mooneyi have the fix locally14:57
sean-k-mooneybut tl;dr cybrog was incorrectly updated to expec Openstack-Api-Version: 2.x14:57
sean-k-mooneyinstead of Openstack-Api-Version: accelerators 2.x14:57
sean-k-mooneyand we later removed microvsrion pasrse as a dep14:58
sean-k-mooneythat was change 2 years ago14:58
sean-k-mooneythat was wrong and it breaks the sdk ectra14:58
opendevreviewchandan kumar proposed openstack/cyborg-specs master: Add generic NVMe driver spec with secure cleanup  https://review.opendev.org/c/openstack/cyborg-specs/+/98534914:58
sean-k-mooneyso ill file a bug to track that14:58
sean-k-mooneylast itme for today14:58
sean-k-mooneywe have notign in open dicuss14:59
melwittouch. so I guess you mentioned this before but that means we lack openstacksdk job running on cyborg repo right14:59
sean-k-mooneyare there any volunteers for the next meeting14:59
sean-k-mooneyso the sdk case might have a workaround14:59
sean-k-mooneyi noticed that nova is not using the sdk its using the wraw keystone auth client15:00
sean-k-mooneyand when i tried to use a new microveriosn it broek15:00
sean-k-mooneyso i need to validate if the sdk has a workaroudn or if its only broken for keystonatuth client15:00
sean-k-mooneybut we ill need to support both shapes for the forseable future15:00
jgilaberI can chair next week's meeting15:01
sean-k-mooney#link https://specs.openstack.org/openstack/api-wg/guidelines/microversion_specification.html#client-interaction 15:01
sean-k-mooney#agreed jgilaber will chair the next meeting15:01
sean-k-mooneyjgilaber: thanks15:01
sean-k-mooneyok we are over time so lets wrap there15:02
sean-k-mooney#endmeeting15:02
opendevmeetMeeting ended Tue Jun 23 15:02:09 2026 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)15:02
opendevmeetMinutes:        https://meetings.opendev.org/meetings/cyborg/2026/cyborg.2026-06-23-14.03.html15:02
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/cyborg/2026/cyborg.2026-06-23-14.03.txt15:02
opendevmeetLog:            https://meetings.opendev.org/meetings/cyborg/2026/cyborg.2026-06-23-14.03.log.html15:02
sean-k-mooneymelwitt: ya we lack the job validitaing it but also we have no microverison test in tempest15:02
melwittah gotcha15:02
sean-k-mooneymelwitt: and well the unit test were changed to match the new code when it was broken ...15:02
sean-k-mooneyhttps://review.opendev.org/c/openstack/cyborg/+/90564615:03
sean-k-mooneyi dont know really what the orgianl error was https://launchpad.net/bugs/204948915:04
sean-k-mooneysicne that is all that is in the bug15:04
melwittah yeah ... that is a fairly common thing I have seen, when the test just mirrors the code so much15:04
sean-k-mooneyso we have the downstream call now. but we dont have any api sample tests or juctiona or tempst tests for microverions15:04
sean-k-mooneyso we will need to build out those going forward15:05
sean-k-mooneywe have api sampels for the api-ref15:05
melwittI think i saw "some" api samples. oh ok15:05
sean-k-mooneyya some exist but they are not hooked into tests15:05
sean-k-mooneyso they wont catch if the shape changes15:05
melwittah ok15:06
sean-k-mooneyso we also dont have any of the openapi stuff that stpen did that would also adress this gap 15:07
sean-k-mooneybut i think for next cycle i want to prospe a genral api refacotrign spec to core some ot this15:08
sean-k-mooneywith that said we can also jsut impvoe this when we have time15:08
sean-k-mooneywe dont need a spec for adding more tests15:08
melwittcool15:08
sean-k-mooneyas long as we update the contibutor docs to document how the tests shoudl work15:08
« Monday, 2026-06-22 Index Wednesday, 2026-06-24 »

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