See https://bugzilla.redhat.com/show_bug.cgi?id=1652845 for discussion with the virt team. The requirements are kernel 4.20 and qemu 3.1 on the host.
When do you need this? (YYYY/MM/DD) ASAP
When is this no longer needed or useful? (YYYY/MM/DD) N/A
If we cannot complete your request, what is the impact? containers and images for ppc64le can't be delivered
@kevin @mohanboddu @puiterwijk
+1
@sharkcz Have we tested it first to make sure that KVM HV mode works as expected with latest mentioned kernel and qemu version on power9 nested virt?
Could we apply similar bypass as for cloud build as per issue 7463 ? https://pagure.io/fedora-infrastructure/issue/7463#comment-547863
This request should fix workaround done in #7463 . Builders in the image channel are the one used to build Fedora cloud and container images.
We have to get our power9 machines setup first, but sure.
Metadata Update from @kevin: - Issue priority set to: Waiting on External (was: Needs Review)
Metadata Update from @mizdebsk: - Issue tagged with: koji
Hi, can you give me an update about the status? The recent updates to the kernel made the existing workaround ineffective.
So, we have the hardware in place and all ready to go.
However, do we want to send a freeze break on this and do it during freeze? It seems risky to me since it will affect composes if something is wrong.
Thoughts?
I think deploying the builders for the "image" channel should be doable even during the freeze, it won't be worse IMHO. But have no problem waiting either. For adding regular builders I would wait after the freeze is lifted off.
ok. I have moved the image builders over to a power9 box. I had to reinstall it as fedora 30 and then add:
<features> <nested-hv state='on'/> </features>
to each buildvm and 'options kvm-hv nested=1' in /etc/modprobe.conf.d/kvm_hv.conf
Metadata Update from @kevin: - Issue close_status updated to: Fixed - Issue status updated to: Closed (was: Open)
Login to comment on this ticket.