Just solved a Xen problem for a new client : his Xen Dom0 host had two interfaces. Both were configured with an IP address on the same subnet. While this worked on the regular Linux kernel, as soon as he booted into the Xen kernel, the machine became unreachable on the network.

Together, we figured out that when Xen boots, it does the “interface shuffle”, switching eth0 to peth0, and creating a bridge. It seems that Linux didn’t like having two interfaces on the same subnet, one of whom was a member of a bridge. It became confused about which NIC to use to respond to network traffic.

The resolution was to disable eth1, and reboot. The Xen host is now happily talking to the network.

Related posts:

1. Serial Port (ttyS0) pass-thru to Xen domU on CentOS 5
2. How to convert a (CentOS 5) VMWare image to Xen