struggling with xen on an updated fedora
I have a pc which I use to run various test of software. A few months ago, I did install a Fedora Core 6 (or 5?) system to test Xen (since I already had a good feedback about this configuration) with 2-3 domU also running Fedora. But for some time now, I didn’t gave it the love it deserved, and now that the Sun is coming back, I’m trying to update it so that can I spend more time behind my screen instead of going in the Big Blue room.
So, I did a local mirror of the fedora repo and here I go. The update (first to fedora 7 and then 8) itself went almost ok. But the first already installed domU would boot ok but not have any network. Not good.
So, there may be things up with the libvirt thingie that (I guess, or just bad-mouthing again) redhat people try to force everybody else to use. The system fails to correctly configure the bridge and leaves a tmpbridge around (which sounds bad enough). Being a bit dumb those days, I try to force the bridge name or dev inside the xend-config.sxp but no joy.
After spending a few hours relying on The Internet, I had to admit Google failed to help me. So I disabled libvirt on boot and I did what another guy did and just added traces to the network-bridge xen script. let’s reboot… look… reboot… look again.
Ah! The dumb script tries to detect the correct interface by finding the one with the default gateway, but my system, being a test system and having all it needs to have on a local mirror, does not have such a gateway. So, ok, let’s try to add such a gateway if it can please it. Hum, better but still a virbr0 laying around. Where does that one come from? Apparently, for some reason, Xen tries to be smart and did remember somewhere that this was a name it needed to use.
Uninstall everything, purge cache/config files and reinstall everything, reboot. And, Victory! At last. And creating the dom0 works with the network being ok.