I have a fresh install of EOS-Plasma on an Intel based 11th gen NUC. I create a network bridge using nmcli and that way I can create my VMs all on my home network subnet.
I created a VM of Ubuntu 20.04 server as a test and once it was working where I could ssh into it I checked the Virt-manager box under boot options for “Start virtual machine on host boot up”. However, it does not work. When EOS boots up Virt manager show it shutdown. Not sure how to fix it.
To collect more data I tested this on a Linux Mint 21.1 host and it works there. Also I tried it on a Debian 12 RC2 host and it behaves just like on EOS.
Okay, making some progress. I found that using the command virsh autostart "vmName" has the same negative effect.
I check the logs and found this related to libvirtd.service
den-pc libvirtd[580]: Cannot get interface MTU on 'br0': No such device
den-pc libvirtd[580]: ethtool ioctl error on vnet0: No such device
den-pc libvirtd[580]: internal error: Failed to autostart VM 'ubuntu20.04': Cannot get interface MTU on 'br0': No such device
den-pc libvirtd[580]: ethtool ioctl error on vnet0: No such device
So it doesn’t connect with my br0 bridge that is setup. If I start the VM with virt-manager it works fine and using br0.
I was thinking that it could be a race condition. libvirtd.service is trying to start up the VM’s marked for autostart before NetworkManager has gotten the br0 setup?
Maybe need some systemd “wait for” statement in libvirtd to delay it until the network is up
It has a section that talked about setting the default network which I’ll not be using. I didn’t do this originally, but when I did the autostart VMs came up at host boot.
sudo virsh net-autostart default
Once I ran this command, the autostart VMs booted at host boot even though none of them use the default NAT NIC. All mine use br0