| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-cyborg - 2026-07-14 | |||
| 14:02:16 | jgilaber | courtesy ping: sean-k-mooney amoralej bogdando chandankumar skovili | |
| 14:02:40 | jgilaber | while we wait, feel free to add topics to the agenda https://etherpad.opendev.org/p/openstack-cyborg-irc-meeting#L49 | |
| 14:03:22 | sean-k-mooney | o/ | |
| 14:04:25 | jgilaber | let's get started with the first topic | |
| 14:04:34 | jgilaber | #topic Tempest testing of multiple drivers | |
| 14:05:03 | jgilaber | I wanted to raise this to have a quick discussion on how we envision we should test the different drivers in tempest | |
| 14:05:36 | jgilaber | we're working on supporting mdevs and nvme emulation of pci-sim so we could test the pci, and future nmve and mdevs drivers in first party ci | |
| 14:05:55 | jgilaber | should we test them in the same job? or have a separate ci job for each? | |
| 14:05:57 | sean-k-mooney | yep | |
| 14:06:03 | sean-k-mooney | yes ideally | |
| 14:06:31 | sean-k-mooney | the pci-sim is desitned so that in the future each PF can present VF with diffent personas | |
| 14:06:45 | sean-k-mooney | i.e. you can create up to 16 PFs | |
| 14:07:02 | sean-k-mooney | and the intent was that some cold be UARTs some NVME devices adn other MDEVs | |
| 14:07:15 | sean-k-mooney | then we can assign those as we see fit to cybrog nova ectra | |
| 14:07:30 | sean-k-mooney | so i dont think we need multiple jobs in this case | |
| 14:07:34 | jgilaber | I'm skimming the document added to the agenda | |
| 14:07:35 | jgilaber | #link https://paste.openstack.org/raw/bQVCnAE5DfkR8Tky99g7/ | |
| 14:07:49 | jgilaber | I agree with that direction, I was thinking something like that | |
| 14:08:22 | jgilaber | I left a comment to discuss that in the pci lifecycle tests patch | |
| 14:08:22 | melwitt | yeah I would think same job if it is technically possible. only separate the jobs if there is some technical limitation | |
| 14:08:24 | jgilaber | #link https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/994580/comment/361c56ef_7480909b/ | |
| 14:09:59 | jgilaber | ok, that looks clear to me, any other comment question on this topic¿ | |
| 14:10:08 | sean-k-mooney | jgilaber: that does not mean we wont have dedicated test for each | |
| 14:10:37 | sean-k-mooney | the genicp pci lifecslcle test are intnede to work with real pci devices as well not just the sim | |
| 14:10:54 | sean-k-mooney | my expecation is each set fo driver will have there won set of lifecycle tests | |
| 14:11:10 | sean-k-mooney | and we will run the subset that can be supproted in a given job | |
| 14:11:33 | sean-k-mooney | but ideally the same test coudl be run with real hardware aswell only by chanign the config options for the tempst plugin | |
| 14:11:36 | jgilaber | sure, I'm planning on working on a similar patch for mdev soon, hence my original question | |
| 14:12:09 | sean-k-mooney | i shared https://chatgpt.com/share/6a3e9241-53a8-83ed-b58f-ab1c933b05af | |
| 14:12:35 | sean-k-mooney | with chandan ealriesr it was my one of my intial planning sessiosn with creating the pci sim | |
| 14:13:01 | sean-k-mooney | i had started on teh design of the persona/personbatly work but put it on hold while i refeind the core module | |
| 14:13:53 | sean-k-mooney | if we need sperate jobs for a speicifc reason we can certely do that when required | |
| 14:14:09 | sean-k-mooney | but for now i think we will be ok | |
| 14:14:32 | jgilaber | to be clear, I think we can test the new drivers along with the pci in the same job, just wanted to have the discussion here | |
| 14:14:54 | sean-k-mooney | yep | |
| 14:15:01 | sean-k-mooney | if we get to the point where the job is takign too long | |
| 14:15:09 | sean-k-mooney | or there are other reason to split im happy to revisit this | |
| 14:15:49 | sean-k-mooney | until we actully ahve a job with multipel diffent device active at the same teim this is still just teroetical | |
| 14:15:53 | sean-k-mooney | but i think we are getting close to that | |
| 14:16:05 | jgilaber | sounds good, moving to the next topic unless there are other questions | |
| 14:16:41 | jgilaber | #topic Enable PCI lifecycle test in cyborg | |
| 14:16:53 | jgilaber | this one is from Chandan, who could not join today | |
| 14:17:10 | jgilaber | it highlights three patches for review | |
| 14:17:19 | jgilaber | #link https://review.opendev.org/c/openstack/cyborg/+/996595 | |
| 14:17:30 | jgilaber | this one enables the tests to run in CI | |
| 14:17:40 | jgilaber | #link https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/994580 | |
| 14:17:54 | jgilaber | this adds the tests for pci driver in the plugin | |
| 14:18:03 | jgilaber | #link https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/996821 | |
| 14:18:23 | jgilaber | and this one adds api microversion support for the tempest plugin (is the first in the chain) | |
| 14:18:40 | sean-k-mooney | sort of | |
| 14:18:47 | sean-k-mooney | it is | |
| 14:19:13 | sean-k-mooney | but there is also sort of this https://review.opendev.org/c/openstack/cyborg/+/994574 | |
| 14:19:51 | sean-k-mooney | cyborg microverion handling is incorrect | |
| 14:19:54 | sean-k-mooney | server side | |
| 14:20:32 | sean-k-mooney | it was correct but it was broken by https://opendev.org/openstack/cyborg/commit/279a260bb2d3b91136ce71c1feb29bc68a7aa962 | |
| 14:20:34 | jgilaber | is there a dependency between them? | |
| 14:20:38 | sean-k-mooney | for https://bugs.launchpad.net/openstack-cyborg/+bug/2049489 | |
| 14:21:15 | sean-k-mooney | yes and no | |
| 14:21:26 | sean-k-mooney | https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/996821 | |
| 14:21:38 | sean-k-mooney | is adding the infra to supprot micoversion but not actully using them | |
| 14:22:53 | sean-k-mooney | the pci life cycle test also do not use them | |
| 14:23:14 | sean-k-mooney | well not beyound declaring min_microversion = '2.0' | |
| 14:23:52 | jgilaber | but the plugin patch could merge before https://review.opendev.org/c/openstack/cyborg/+/994574 right? | |
| 14:24:30 | sean-k-mooney | it could yes | |
| 14:25:01 | sean-k-mooney | im just pointing out that there is no actual depenonc eon the microversion supprot for ti to work to my knowldage | |
| 14:25:14 | sean-k-mooney | if you dont pass any microverion header its ment to defautl to 2.0 anyway | |
| 14:25:55 | sean-k-mooney | anyway we can proceed with it in this order | |
| 14:26:13 | sean-k-mooney | but it shoudl work without the microverison patch | |
| 14:26:29 | jgilaber | I see what you mean | |
| 14:26:49 | jgilaber | I'll try to review both the pci tests and the cyborg bug fix patch later today | |
| 14:27:03 | jgilaber | does anyone have any question comment on this patches? | |
| 14:28:05 | jgilaber | ok, moving on then | |
| 14:28:14 | jgilaber | #topci Generic Mdev driver | |
| 14:28:20 | jgilaber | #undo | |
| 14:28:20 | opendevmeet | Removing item from minutes: #link https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/996821 | |
| 14:28:39 | jgilaber | ahh I did not want to do that sorry | |
| 14:28:46 | jgilaber | #topic Generic Mdev driver | |
| 14:29:12 | jgilaber | I want to mention as fyi that I have a series of patches to add the generic mdev driver | |
| 14:29:14 | jgilaber | #link https://review.opendev.org/c/openstack/cyborg/+/997047/1 | |
| 14:29:24 | jgilaber | it also adds mdevs to pci-sim | |
| 14:29:48 | jgilaber | I'm not asking for reviews yet, but I've tested it manually on a devstack env and it seems to work well | |
| 14:30:05 | jgilaber | I've been able to create an instance with Nova | |
| 14:30:21 | jgilaber | I plan to start with tempests test before asking for reviews | |
| 14:32:02 | sean-k-mooney | cool | |
| 14:32:25 | sean-k-mooney | did you hit any roadblocks or painpoints along the way | |
| 14:32:40 | sean-k-mooney | i.e. anything that could/should be impvoed or highlighted | |
| 14:33:29 | jgilaber | not really no, it's been quite quick | |
| 14:33:59 | jgilaber | I had claude write the pci-sim part since I did not know much about the kernel side and it did not get it right on the first try | |
| 14:34:12 | jgilaber | but it fixed it quickly | |
| 14:35:09 | sean-k-mooney | so until we add the ablity ot have diffent pf types | |
| 14:35:20 | sean-k-mooney | there is gong to be a conflcit between your work and chandans | |
| 14:35:32 | sean-k-mooney | that oen of the thing that need to be doen with care | |
| 14:35:50 | sean-k-mooney | as chandands inital verison was leaking device speific info into the generic stuctures | |
| 14:36:10 | sean-k-mooney | currently your adding enable_mdev=1 | |
| 14:36:14 | sean-k-mooney | as a module parmater | |
| 14:36:27 | sean-k-mooney | buyt that not compatible with teh per PF device personaltiy approch | |
| 14:36:43 | sean-k-mooney | still the poc is still useful | |
| 14:36:57 | sean-k-mooney | but we will have to rework the kernel part before it merges | |
| 14:36:57 | jgilaber | ack, that's not a problem I can rebase and improve the approach | |
| 14:37:16 | jgilaber | my goal was to have something working end to end | |