| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-cyborg - 2026-08-18 | |||
| 14:08:38 | rlandy | #action: prepare etherpad for PTG topics | |
| 14:09:03 | sean-k-mooney | #link https://etherpad.opendev.org/p/cyborg-2027.1-ptg | |
| 14:09:12 | sean-k-mooney | that is empty but that follow our normal naming scheme | |
| 14:09:23 | rlandy | thank you sean-k-mooney for standing as PTL this cycle | |
| 14:10:08 | rlandy | anything more on PTG? | |
| 14:11:03 | rlandy | ok ... any other topics for today? | |
| 14:11:24 | sean-k-mooney | am the only thing is once we are pass FF we shoucl assess how much progress we made on the last ptg's items | |
| 14:11:56 | sean-k-mooney | on other topics i guess we could quickly update on the progress of open reviews/work | |
| 14:12:18 | sean-k-mooney | we are 1 weeks form FF thursday the 27th | |
| 14:12:30 | rlandy | original work items listed here for reference: | |
| 14:12:36 | rlandy | #link: https://etherpad.opendev.org/p/cyborg-2026.2-status | |
| 14:13:07 | sean-k-mooney | i need to add my vfio managed mode serse to that as well | |
| 14:13:23 | sean-k-mooney | i have started doing a detailed review of the nvme series today | |
| 14:13:34 | sean-k-mooney | i hope to complete my first pass on it end ot end today | |
| 14:13:41 | rlandy | if FF is the 27th - that mean the IRC meeting in two weeks should be the review? | |
| 14:13:46 | sean-k-mooney | ill try and deploy it later in the wek | |
| 14:14:02 | sean-k-mooney | rlandy: yes | |
| 14:14:30 | sean-k-mooney | so we will decied in 2 weeks if any feature is close enough to have a FF excption | |
| 14:14:39 | sean-k-mooney | until the 4th | |
| 14:14:46 | sean-k-mooney | but ideally we do not need that | |
| 14:15:03 | sean-k-mooney | so in 2 weeks time we should really know what in and what moves to next cycle | |
| 14:15:17 | rlandy | #agreed FF item review - two weeks | |
| 14:15:37 | sean-k-mooney | we will likely try and do the RC1 tagging early if we can | |
| 14:16:07 | chandankumar | sean-k-mooney: I have updated 994299: Report OWNER_NOVA trait on resource providers | https://review.opendev.org/c/openstack/nova/+/994299 for cyborg vgpu support. I will try to update rest of the mdev patches soon. | |
| 14:16:34 | sean-k-mooney | RC1 is due septmber 10th | |
| 14:17:05 | sean-k-mooney | so in 3 weeks time we shoudl review that we have everything ready for that | |
| 14:17:28 | rlandy | #action - three weeks add RC1 review | |
| 14:17:38 | sean-k-mooney | so 2 weeks time review Feature compeltion, 3 weeks time review bug completion for rc1 | |
| 14:18:09 | sean-k-mooney | chandankumar: thanks ill see if i can review it again in the next day or two | |
| 14:18:51 | chandankumar | thank you! | |
| 14:18:58 | sean-k-mooney | does anyone want to highlight progress on teh series they are working on ? | |
| 14:19:15 | gamio | 1000240 (Intel NIC device_addresses allow-list, Closes-Bug 2161354) has melwitt's CR+1. I'm pushing PS2 today with her feedback addressed. It's a small feature so I'd like to land it before FF on the 27th - could a core take a look this week? | |
| 14:19:45 | sean-k-mooney | gamio: sure can you add it to https://etherpad.opendev.org/p/cyborg-2026.2-status | |
| 14:20:07 | sean-k-mooney | i agree it would be nice to include that in this release | |
| 14:20:16 | gamio | sure i will | |
| 14:21:06 | rlandy | I am making slow progress with the documentation redesign - next set of moves/rework should be available for review next week | |
| 14:21:32 | chandankumar | I was working on nvme series. It works with pci-sim nvme implementation (which test sanitize cleanup) as well as libvirt nvme qemu setup (testing write zero setup). | |
| 14:21:41 | sean-k-mooney | ack the docs changes can continue to be reviewed/merge up until RC1 | |
| 14:22:22 | chandankumar | The nvme docs needs to be some modification, I have -w not ready for review, except everything is in good shape | |
| 14:22:38 | sean-k-mooney | chandankumar: ack i have given some feedback on that this morning there are some unsafe operations that we will need to remove | |
| 14:23:13 | chandankumar | sure, I will take a look and update it. | |
| 14:23:13 | sean-k-mooney | specificly detecting if a device supprot namespace managmenet by trying to delete the namespace on the device is not really something we can do | |
| 14:23:44 | sean-k-mooney | instead if the qemu device supprot it and it does not work later we need to trust what was reported and handel the failure later gracefully | |
| 14:24:19 | chandankumar | ah ok, I pushed this part today morning. | |
| 14:24:34 | sean-k-mooney | i just pushed my comments | |
| 14:24:40 | sean-k-mooney | i realised i had them still pending | |
| 14:25:02 | chandankumar | ok, make sense! | |
| 14:25:27 | rlandy | It looks like we are already in the Reviews section - so please raise any other reviews that require attention | |
| 14:25:27 | sean-k-mooney | melwitt: i belive i approved your docs changes for the intel nic driver this morning | |
| 14:25:48 | sean-k-mooney | rlandy: sorry yes i jumped ahead | |
| 14:26:15 | melwitt | yes thank you sean-k-mooney | |
| 14:26:31 | sean-k-mooney | melwitt: are there any other open reviews you want to highlight? | |
| 14:27:04 | melwitt | no, not at the moment | |
| 14:27:10 | gamio | quick one on 1000240 - i checked the gpu driver too while looking at melanie's comment. | |
| 14:27:41 | gamio | both share the same root cause: device_addresses is compared exact-match against lowercase sysfs/lspci values, so a wrong-case or absent address never matches. | |
| 14:27:59 | sean-k-mooney | right | |
| 14:28:10 | sean-k-mooney | we need to normalsie the adresses when comparing | |
| 14:28:25 | chandankumar | one more thing https://review.opendev.org/c/openstack/cyborg/+/999939 when I added nvme device state , I think I have added a regression in fake driver, jsut wanted to highlight. | |
| 14:28:37 | gamio | the nic side just doesn't discover the device on the gpu side the lookup returns None and the card is registered as a pgpu instead of a vgpu _get_vgpu_type_per_pgpu's docstring is off too | |
| 14:29:00 | gamio | it calls the param a config address but the caller passes the discovered one. | |
| 14:29:17 | gamio | so rather than fixing only the nic half here, i'd keep 1000240 as is and open a bug to fix both in a follow-up. does that sound ok? | |
| 14:29:35 | sean-k-mooney | sure | |
| 14:29:54 | sean-k-mooney | im not fully folloing the code you are refint too without links | |
| 14:29:59 | sean-k-mooney | but if you write it up as a bug | |
| 14:30:08 | sean-k-mooney | we can proceed with fixing it in both places | |
| 14:31:00 | sean-k-mooney | also ideally were approcate we can consdier using your recenly intoduced fucntion for listing devices ectra if that makes sense | |
| 14:31:12 | gamio | sean-k-mooney: thanks, will do. i'll write it up as a bug with the code links and drop the number on the review. | |
| 14:31:25 | sean-k-mooney | i know the intel driver currenly does not use lspci and use direct sysfs access to enermerate teh devices | |
| 14:31:44 | sean-k-mooney | chandankumar: that one of hte comment i left on the nvme serise | |
| 14:32:02 | sean-k-mooney | chandankumar: i dont think you shoudl write yoru own pci device enumeration logic | |
| 14:32:15 | sean-k-mooney | you shoudl reuse teh get_devices funciton for the acclerator utils | |
| 14:33:05 | sean-k-mooney | gamio: for these bug fixes im not sure we need to rewite them to use your new util fuctions if that is invaise | |
| 14:33:11 | chandankumar | ok, let me go through comments | |
| 14:33:31 | sean-k-mooney | gamio: but next cycle im hopeing to add a PCIBusManager | |
| 14:33:46 | sean-k-mooney | to handel all of this for all driver so we can defer some of that to that work | |
| 14:34:15 | sean-k-mooney | gamio: so use your best judgement on what to keep local to the drvier for now in an easy backportabvly patch vs what tech debt to adress | |
| 14:34:16 | gamio | nice, that'll be a good home for all this. happy to help on the PCIBusManager next cycle, it's right in my area. | |
| 14:35:04 | rlandy | since we're in Bugs now ... any additional bugs to raise? | |
| 14:35:22 | rlandy | reference ... | |
| 14:35:25 | rlandy | https://bugs.launchpad.net/python-cyborgclient | |
| 14:35:25 | rlandy | https://bugs.launchpad.net/openstack-cyborg | |
| 14:36:45 | rlandy | There are two new bugs | |
| 14:37:05 | rlandy | #link: https://bugs.launchpad.net/openstack-cyborg/+bug/2017513 | |
| 14:37:17 | rlandy | sync data to placement raise resource in use exception | |
| 14:37:20 | rlandy | from 2023 | |
| 14:38:04 | rlandy | undecided | |
| 14:38:45 | rlandy | can we mark that triaged? | |
| 14:39:19 | sean-k-mooney | yes we can mark it as triaged | |
| 14:39:20 | rlandy | importance? | |
| 14:39:36 | sean-k-mooney | its medium to high | |
| 14:39:43 | sean-k-mooney | because this is a data integrety bug | |
| 14:39:57 | rlandy | medium to start | |
| 14:40:06 | sean-k-mooney | as i noted in the comments | |
| 14:40:14 | sean-k-mooney | nova had this exact same bug years ago | |
| 14:40:25 | sean-k-mooney | the fix jsut never got ported to cyborg | |
| 14:41:09 | sean-k-mooney | the tl;dr is even if you remove a device form the pci whitelist | |
| 14:41:25 | sean-k-mooney | we shoudl not remove it form placment or the devices api if it is currently allcoated | |
| 14:41:51 | sean-k-mooney | so we will need to modify the condocutor/cybrog-agent to do this detection on startup | |
| 14:42:11 | sean-k-mooney | and prevent this form happening but we can leave it there for now | |
| 14:42:51 | rlandy | ack - thank you ... the second untriaged bug: #link https://bugs.launchpad.net/openstack-cyborg/+bug/2163240 | |