Problem getting virt manager running

[manfredlotz]

When I switched from ghcr.io/ublue-os/bluefin-dx-nvidia-open:stable -> ghcr.io/projectbluefin/bluefin-nvidia:stable the virt manager/qemu related packages wered gone. This is simply intended.

I tried ujust devmode and installed Virtual Machines (virt-manager + QEMU)"

Now I can start virtmanager and get an error message

Could not detect a default hypervisor. Make sure the appropriate QEMU/KVM virtualization and libvirt packages are installed to manage virtualization on this host.

A virtualization connection can be manually added via File->Add Connection

Of course, there are no services active here.

I need some documentation how to get this working.

[castrojo]

It's supposed to work out of the box please file these as issues and I'll batch and ship today, thanks!

[manfredlotz]

Ok. projectbluefin/bluefin#1222

[sebjag]

Of course, there are no services active here.

Did you select QEMU/KVM user session in the add connection window?

[manfredlotz]

@sebjag In the past I could access my virtual machines from my host machine by using ssh. Do you know how to configure things that this is possible with QEMU/KVM user session. Did not find a documentation? Thank.s

[sebjag]

My isos are outside of my home directory

You are right, I think you need to copy or move the images and change ownership, then create a new VM with Import existing disk image.

access my virtual machines from my host machine by using ssh

Should be possible. With usermode networking, you need to port forward from vm to host (e.g. 22 -> localhost:2222). I haven't done this, but I suggest asking an llm for some xml config that sets that up.

[manfredlotz]

There are more issues. I have a situation where I previously ran /usr/bin/qemu-system-x86_64 in a script and I don't know how to translate this into flatpak.

As I have no time to deal with these issues I decided that for the time being it is best to add in my custom image

dnf5 install -y @virtualization
dnf5 install -y swtpm

Now all is fine here. The time will come when I will try again the flatpak way...

[sebjag]

how to translate this into flatpak

When the time comes, you can try something like this:

# qemu-system-x86_64
flatpak run --command=qemu-system-x86_64 org.virt_manager.virt-manager --version

# see also:
flatpak run --command=sh org.virt_manager.virt-manager
[📦 org.virt_manager.virt-manager ~]$ ls /app/bin/
[📦 org.virt_manager.virt-manager ~]$ ls /app/lib/extensions/bin/

[manfredlotz]

Thanks very much. I have added it to my documentation.

[znmeb]

I've been down this path - probably opened an issue. There are QEMU packages and groups that need to be installed on the image and the user needs to be part those groups. The Virtual Machine Manager RPM in Silverblue sets everything up except making sure the user's groups are set.

[manfredlotz]

The ublue-os based ghcr.io/ublue-os/bluefin-dx-nvidia-open had those packages installed as RPMs. All was working fine.

But the new way in projcetbluefin is to get away from this. I could install those RPMs easily as I have a custom image but prefer to follow the choosen path.