| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-cyborg - 2026-03-10 | |||
| 13:33:44 | chandankumar | Cyborg meeting starts in 27 mins, Please add your agenda here: https://etherpad.opendev.org/p/openstack-cyborg-irc-meeting#L43 | |
| 13:59:22 | opendevreview | chandan kumar proposed openstack/cyborg-tempest-plugin master: [WIP] Add cyborg pci nvme test https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/979847 | |
| 14:01:55 | chandankumar | Hello, who all are around for cyborg meeting? | |
| 14:02:02 | rlandy | o/ | |
| 14:02:19 | jgilaber | o/ | |
| 14:02:23 | opendevmeet | The meeting name has been set to 'cyborg' | |
| 14:02:23 | opendevmeet | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | |
| 14:02:23 | opendevmeet | Meeting started Tue Mar 10 14:02:23 2026 UTC and is due to finish in 60 minutes. The chair is chandankumar. Information about MeetBot at http://wiki.debian.org/MeetBot. | |
| 14:02:23 | chandankumar | #startmeeting cyborg | |
| 14:02:29 | bogdando[m] | o/ | |
| 14:03:09 | chandankumar | #courtesy ping sean-k-mooney amoralej jgilaber | |
| 14:03:31 | chandankumar | feel free to add topics to the agenda | |
| 14:03:57 | amoralej | o/ | |
| 14:04:20 | chandankumar | #link https://etherpad.opendev.org/p/openstack-cyborg-irc-meeting#L43 | |
| 14:05:10 | sean-k-mooney | hi o/ | |
| 14:05:27 | chandankumar | starting with the first topic | |
| 14:05:50 | chandankumar | #topic testing nvme pci device tempest tests in Upstream CI | |
| 14:06:37 | chandankumar | I was working on adding nvmevirt devstack plugin for cyborg to test nvme pci device in CI https://review.opendev.org/c/openstack/cyborg/+/977083 | |
| 14:07:03 | chandankumar | I was working on adding the tempest scenario test https://review.opendev.org/c/openstack/cyborg-tempest-plugin/+/979847 where we can expose the nvme pci device to the vm | |
| 14:07:28 | chandankumar | But while creating the vm, On nove compute side, I am hitting unsupported configuration: host doesn't support passthrough of host PCI devices | |
| 14:08:02 | sean-k-mooney | ya so nova can virutalize a iommu but you need to enable it in the image uploaded to the provider | |
| 14:08:03 | chandankumar | Later on I realized that I need to enable IOMMU on the host to make sure pci passthroug works | |
| 14:08:13 | sean-k-mooney | althernitivly i think there are 2 other options | |
| 14:08:26 | sean-k-mooney | 1 enable the softiommu on the kernle command line | |
| 14:08:41 | sean-k-mooney | 2 set teh mdoule option for vfio-pci to allow "unsafemode" | |
| 14:09:19 | sean-k-mooney | modprobe vfio enable_unsafe_noiommu_mode=1. | |
| 14:09:24 | chandankumar | thanks sean-k-mooney Do we have any job upstream using any of these options? | |
| 14:09:32 | sean-k-mooney | no | |
| 14:09:40 | chandankumar | Do we need to change the nodeset label also? or current one will work fine | |
| 14:09:46 | sean-k-mooney | without those we could still do some limited testing | |
| 14:09:50 | sean-k-mooney | just not passhtough to a vm | |
| 14:10:05 | sean-k-mooney | i..e we could bind an arq and unbindi it formn tempet | |
| 14:10:37 | sean-k-mooney | in a whitebox style plugin we coudl aslo ssh in write some data to the device and hten asert that its cleaned when unbound | |
| 14:10:45 | sean-k-mooney | it would be preferable to test it via nova howver | |
| 14:10:55 | sean-k-mooney | with standard tempest | |
| 14:12:29 | chandankumar | bind and unbind arg sounds good | |
| 14:13:28 | chandankumar | I will break the existing test into multiple tests to have bind/ubind as well as passthrough to vim via standard tempest tests | |
| 14:13:53 | chandankumar | Thank you! I will update it in next meeting how it goes. | |
| 14:14:10 | bogdando[m] | a naive question, can't we reboot zuul node in CI? that breaks devstack, right? | |
| 14:14:26 | sean-k-mooney | yes and no | |
| 14:14:31 | sean-k-mooney | reboots dont break devstack | |
| 14:14:36 | sean-k-mooney | and we can remboot the vms | |
| 14:14:43 | sean-k-mooney | we woudl do that before installing devstack however | |
| 14:16:07 | sean-k-mooney | if the kernel is compiled with supprot for it | |
| 14:16:09 | bogdando[m] | so worth exploring? | |
| 14:16:09 | sean-k-mooney | echo 1 | sudo tee /sys/module/vfio/parameters/enable_unsafe_noiommu_mode | |
| 14:16:21 | sean-k-mooney | is the simples option in the exsitng pre playbook | |
| 14:17:43 | sean-k-mooney | without that we can still do some testign via whtiebox we just can boot a vm. lets see if ubuntu 24.04 or debian 13 ship with that kernel module partmer | |
| 14:17:52 | sean-k-mooney | if not we can revisit how we approch this | |
| 14:18:06 | sean-k-mooney | liek addign a new image to the ci and nodepool lable with the viommu enabled | |
| 14:18:27 | sean-k-mooney | we dont wnat to do that by defualt as there is a non trivial performace hti to that | |
| 14:18:31 | chandankumar | I will check with my existing nvmevirt preplaybook | |
| 14:19:52 | chandankumar | yes, we can revisit this topic in next meeting based on testing how it goes. | |
| 14:20:06 | chandankumar | Any more questions on this topic? | |
| 14:20:23 | sean-k-mooney | ubuntu@openclaw:~$ grep CONFIG_VFIO_NOIOMMU /boot/config-$(uname -r) | |
| 14:20:25 | sean-k-mooney | CONFIG_VFIO_NOIOMMU=y | |
| 14:20:42 | sean-k-mooney | so ubuntu 24.04 appears to have it enabeld at least if we install the mainline kernel | |
| 14:20:50 | bogdando[m] | \o/ | |
| 14:20:51 | sean-k-mooney | but now lets see what your ci run shows | |
| 14:21:20 | sean-k-mooney | if we need to install the mainline (upstream kernel) we can do that but we need to verify it in a vanilla 24.04 | |
| 14:22:13 | chandankumar | yup, I will verify it locally with mainline kernel | |
| 14:22:22 | chandankumar | good point! | |
| 14:23:00 | chandankumar | moving to next topic | |
| 14:23:09 | chandankumar | #topic Reviews | |
| 14:23:25 | chandankumar | there is one review | |
| 14:23:29 | chandankumar | #link 979805: Rename devstack CYBORG_CONF_FILE to CYBORG_CONF for consistency | https://review.opendev.org/c/openstack/cyborg/+/979805 | |
| 14:24:18 | chandankumar | While I was working on nvmevirt devstack patch, I found that In cyborg we use CYBORG_CONF_FILE var for cyborg conf in devstack | |
| 14:24:41 | sean-k-mooney | hum i think its currently CYBORG_CONF_FILE to disambiguate with CYBORG_CONF_DIR | |
| 14:24:53 | chandankumar | but in Nova or other services we use <SERVICE>_CONF, so I thought about changing it to nova format | |
| 14:25:37 | sean-k-mooney | well nova has more then one conf it uses with mutlipe variables | |
| 14:25:48 | sean-k-mooney | i.e. NOVA_CPU_CONF | |
| 14:25:50 | sean-k-mooney | i think | |
| 14:26:14 | sean-k-mooney | we could make this change im just not sure it by us much | |
| 14:27:10 | sean-k-mooney | if we do this i woudl make it fallback to the old name | |
| 14:28:08 | sean-k-mooney | so CYBORG_CONF=${CYBORG_CONF:-${CYBORG_CONF_FILE}} | |
| 14:28:10 | chandankumar | ok make sense, It will not break existing user | |
| 14:28:25 | chandankumar | I will update the review after this call. | |
| 14:28:51 | sean-k-mooney | ack | |
| 14:29:05 | chandankumar | since there is no more reviews moving to next topic | |
| 14:29:09 | chandankumar | #topic Bugs | |
| 14:29:30 | chandankumar | #link https://bugs.launchpad.net/openstack-cyborg/+bug/1954888 Docs: state how to use bound accelerators | |
| 14:29:39 | chandankumar | we discussed this bug in the last meeting | |
| 14:30:44 | sean-k-mooney | ya so im incliend to ether use this to track descirbing how to create the nova flavor and device profile | |
| 14:30:51 | sean-k-mooney | and that workflow | |
| 14:30:51 | bogdando[m] | looks like it should stay deferred unless standalone cyborg gets on the radar | |
| 14:31:08 | sean-k-mooney | or using it to track a contibutor doc or closing as wont fix | |
| 14:31:30 | sean-k-mooney | right standalon cyborg is not really a thing today so we shoudl nto have end user docs for that | |
| 14:31:46 | sean-k-mooney | so either we have a conceptal do for devleoper/contibutors | |
| 14:32:02 | sean-k-mooney | or we docuemnt the nova flow for end admins ro end users | |
| 14:32:13 | sean-k-mooney | but i dont think we shoudl spend time on stand alone | |
| 14:32:35 | sean-k-mooney | we could close this and create 2 seperate bugs for those other two docs gaps | |
| 14:32:41 | jgilaber | +1 I think there is value in documenting the standard workflow using nova, but I would not go further than that for now | |
| 14:33:41 | sean-k-mooney | i.e. 1 for admin doc for configuing and testign the standard wrokflow and 2 a contibutor doc to serve as a technial deep dive/refence of howt he services interact? | |
| 14:34:44 | chandankumar | Do we have a reference for nova flow doc link which we can add as a reference in the bug? | |
| 14:35:32 | sean-k-mooney | i dont think this really exist out side fo the spec today | |
| 14:35:50 | sean-k-mooney | i dont recall htere being a nova doc for ti but there may be | |
| 14:36:56 | chandankumar | ok | |
| 14:36:58 | sean-k-mooney | i will file 2 seperate bugs and close this when done | |
| 14:37:15 | chandankumar | thanks sean-k-mooney! | |
| 14:37:20 | sean-k-mooney | i can likely create a draft for the docs in the next few weeks if no one else does | |