[Starlingx-discuss] Attempt to PCI Passthrough in Stx ended up with "no valid host available" error

open infra openinfradn at gmail.com
Mon Sep 5 14:33:01 UTC 2022


Hi,

I recently noticed that newly added workers not able to host any openstack
instance but only few workers (which created very early in our setup) able
to host instanes created in openstack side.

system application-apply stx-openstack is failed at downloading docker
images though controller can ping outside. No specific information found on
system inventory log and no armada log at all for stx-openstack-apply.

When I looking at stx active alarms, there was a failure of stx-openstack
apply, which was on the day apply changes to enable PCI Pasthrough. [1].
I can see all the workers when I check nodes from kubernetes side.

01. How to get the stx-openstack application apply without fail at docker
image download?
I tried to remove the line added into helm override yaml to add pci
passthrough device and apply the application but it failed.
02. I need to enable Nvidia T10 cards in selected nodes as PCI Passtrough.
Should we add all selected gpu pci addresses to controller?  Is there a
complete guide for enabling PCI Passthorugh with GPU cards in starlingx
side as well as openstack setup as well?

[1]
https://docs.starlingx.io/node_management/openstack/configure-pci-passthrough-interface-to-nvidia-gpu.html

Regards,
Danishka
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.starlingx.io/pipermail/starlingx-discuss/attachments/20220905/dba6761a/attachment.htm>


More information about the Starlingx-discuss mailing list