Earlier  
Later
New Search
Posted Nick Remark
#openstack-cyborg - 2026-06-30
14:54:16 jgilaber that is all for today, thanks everyone!
14:54:19 opendevmeet Log: https://meetings.opendev.org/meetings/cyborg/2026/cyborg.2026-06-30-14.00.log.html
14:54:19 opendevmeet Minutes (text): https://meetings.opendev.org/meetings/cyborg/2026/cyborg.2026-06-30-14.00.txt
14:54:19 opendevmeet Minutes: https://meetings.opendev.org/meetings/cyborg/2026/cyborg.2026-06-30-14.00.html
14:54:19 opendevmeet Meeting ended Tue Jun 30 14:54:19 2026 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)
14:54:19 jgilaber #endmeeting
14:54:23 morenod thanks!!
14:54:28 rlandy jgilaber++ thank you
15:11:25 melwitt sean-k-mooney: a primer for the API samples testing patch ... I used claude and gave a goal for something much lighter weight than the usual .tpl model. if you hate it let me know and I can try something else
15:12:54 sean-k-mooney i hate our normal model so im sure ill like it :)
15:13:17 melwitt haha well, I have left the hate door open if you need it :)
15:13:55 sean-k-mooney so long term (in like 6-12 months time) i woudl love to not need this because we fully impleitne the OPENAPI schema goal for cybrog
15:14:15 melwitt next in the "plan" is gabbi API tests
15:14:29 melwitt yeah, good point, that takes some of the pressure off
15:14:41 sean-k-mooney i.e. it woudl be nice if the schmea was validate speratly form the sampels which jsut woudl help with docuemtion
15:15:16 sean-k-mooney i need ot start a wider converation on evovling the api
15:15:22 melwitt so openapi does both request and response schemas right?
15:15:28 sean-k-mooney yes
15:15:33 sean-k-mooney if you implemnt both of course
15:15:44 melwitt I mean, should I instead be starting with that?
15:16:00 sean-k-mooney no i think this is low haning fruit
15:16:08 sean-k-mooney and its good to get this in place to have any coverage
15:16:10 melwitt ok cool
15:16:21 melwitt yeah. it found some minor errors
15:16:41 sean-k-mooney but if you wanted to start on that after sure. we should ahve that converstaion about the other api issues that i was hopign to look at next cycle
15:16:43 melwitt well, I guess define "minor" but some parameters that should not have been in the samples and some missing
15:17:05 melwitt nah just wanted to make sure I stay aligned with the wanted plans
15:17:26 sean-k-mooney nope your totally aligned now worries on that front
15:17:40 sean-k-mooney so tempst ahas only ever had scemeas for nova and cindier i belive
15:17:56 sean-k-mooney i woudl perfer to avodi addig them for cybrog and jsut do the openapi feature isntead
15:18:00 melwitt the tempest schemas are so painful in my experience
15:18:09 sean-k-mooney espcially since cybnorg api is small
15:18:42 sean-k-mooney by the way i think one fo the api routes is missing form the api ref
15:19:19 melwitt oh hm
15:19:25 sean-k-mooney https://docs.openstack.org/api-ref/accelerator/#deployables or https://docs.openstack.org/api-ref/accelerator/#devices
15:19:33 sean-k-mooney has a update/program endpoint
15:19:40 sean-k-mooney for programign fpgas
15:19:50 sean-k-mooney i belive it sdeployables via a patch or put
15:20:21 melwitt ok I'll check on it
15:21:23 sean-k-mooney looking at my srbac serise its deployables https://review.opendev.org/c/openstack/cyborg/+/992295/5/cyborg/policies/deployables.py#41
15:21:46 sean-k-mooney ah and yes PATCH
15:22:32 sean-k-mooney chandan has a chagne up to test that with the fake driver via tempest
15:23:17 sean-k-mooney long term im not conviced that api shoudl exsit but that a sperate matter
15:23:18 melwitt so speaking of the tempest test stuff ... is your pci kernel module thing what we can use to do proper lifecycle coverage?
15:23:28 sean-k-mooney yep
15:23:36 melwitt ok great
15:23:45 sean-k-mooney the emulated device can be passed to the qemu isntance and actully work
15:23:52 melwitt super
15:23:56 sean-k-mooney so the uart is a simple loop back
15:24:07 sean-k-mooney chandan has extened it to emulate nvme ssds
15:24:20 sean-k-mooney i plan ot add nic emualtion evnutlly adn supprot for mdevs
15:25:03 sean-k-mooney so we will be able to emulate all of the devices we manage via the 3 generici driver and in the future htat will allow us to do sriov testing in nova as well if/whewn we want too
15:25:22 melwitt awesome
15:25:38 sean-k-mooney melwitt: i have been holding of adding that to the nova jobs until we are happy with the current feature set and the stableity
15:25:52 melwitt ++ I think that is wise
15:25:53 sean-k-mooney but im thinking of adding it to the new nova-whitebox job
15:27:01 melwitt speaking of that, we need your re-review on https://review.opendev.org/c/openstack/nova/+/898100 to land it
15:27:02 sean-k-mooney its pretty stable form what i can tell and compiling it is fast but i was goign to give it a month or two before goign town that route
15:27:21 sean-k-mooney oh yes cool ill look at that shortly
15:27:30 melwitt yeah ++ for adequate bake time
15:28:46 sean-k-mooney it also currently need debian 13 or ubuntu 26.04 which i dont think that whiteboxc job uses
15:29:00 melwitt oh right
15:29:07 sean-k-mooney so i want to get the whitebox tests running on the cybrog job first
15:29:21 sean-k-mooney and only look at updatign the nova oen after we have all the kinks worked out
15:29:36 melwitt ++
15:29:40 sean-k-mooney one of the optimisation we need to do for whitebox is use the serial decorator
15:29:47 sean-k-mooney instead of --concurancy=1
15:30:23 sean-k-mooney i think gmann was consideritng doign that refactor but if not i woudl like to do that before integrating pci-sim into it
15:30:54 melwitt ok so about that
15:31:30 melwitt there is one test in whitebox, a vtpm test, that restarts libvirtd. this is what messes up all ability to use parallel. we need to separate out that test for it to run on its own
15:32:13 melwitt I could work on that actually, I just didn't think about it at the time
15:32:56 melwitt I guess there might be other similar tests in whitebox too. if so, just all of those need to go in their own serial/ directory and run after everything else like normal tempest does
15:32:57 gmaan i have that in my list but I think I will not be able to start that for next 2 weeks, feel free to do it if you can before that
15:33:18 melwitt ah cool :)
15:35:48 gmaan sean-k-mooney: melwitt I was thinking to add it in meeting but missed. as you both are here, please check/test the enforce_scope removal for cyborg, I have proposed the example changes for other projects, melwitt has reviewed the nova/placement one already. https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/message/436HHMRVNUY3OSXE4ULPQ4P4DSHQEMES/
15:36:36 melwitt ok cool yes I will look
15:36:41 gmaan thanks
15:38:03 gmaan i can see only a few reference of that but you can add testing change with oslo.policy change and test if anything else other than those few test changes is needed https://github.com/search?q=repo%3Aopenstack%2Fcyborg%20enforce_scope&type=code
15:38:16 melwitt oh yikes the deadline is july 3
15:44:51 opendevreview Takashi Kajinami proposed openstack/cyborg master: Remove unused context properties https://review.opendev.org/c/openstack/cyborg/+/995205
15:46:38 opendevreview Takashi Kajinami proposed openstack/cyborg master: Remove unused context properties https://review.opendev.org/c/openstack/cyborg/+/995205
16:02:17 opendevreview melanie witt proposed openstack/cyborg master: Fix api-ref inaccuracies and add missing program endpoint https://review.opendev.org/c/openstack/cyborg/+/995533
16:02:46 opendevreview Takashi Kajinami proposed openstack/cyborg master: Remove unused _extract_attributes_v2 https://review.opendev.org/c/openstack/cyborg/+/995534
16:02:55 sean-k-mooney gmaan: yes i have been folowing that form afar
16:03:13 sean-k-mooney gmaan: basiclly waitign for you and stephen to figure out the way forward
16:03:37 gmaan sean-k-mooney: ack, thanks
16:03:42 opendevreview Takashi Kajinami proposed openstack/cyborg master: Remove unused method for image v1 API https://review.opendev.org/c/openstack/cyborg/+/995534
16:03:53 sean-k-mooney gmaan: so i need to remove my overried where i explicly default that to true
16:04:15 sean-k-mooney but i wrote teh securigy bug fixes assuming the enfoce scope
16:04:18 gmaan yes, true or false, basically all usage of enforce_scope
16:04:34 sean-k-mooney and that whaat my srbac serise is built around as well
16:04:50 gmaan enforce_new_defaults stay same to toggle the old/new defaults but not the scope
16:04:58 sean-k-mooney yep
16:05:31 sean-k-mooney ill tack a patch on the end of my srbac serise to drop our usage
16:05:54 sean-k-mooney gmaan: im hopign to merge that this week as well
16:06:09 gmaan ++
16:06:35 sean-k-mooney melwitt: sorry was on a call
16:06:49 sean-k-mooney melwitt: so there are a couple of whitebox tests that restart the agent
16:07:15 sean-k-mooney melwitt: we used ot have test that reconfigre the nova.conf setting to test spefic feature that were mutally exclivie
16:07:46 sean-k-mooney so for vtpm ya there is one that need special handleing but the test for offliniing cpus and some fo the pining test used to restart agents too

Earlier   Later