Hi folks,
I am stuck with a weird phenomenon.
I have set up two servers as xen servers with bonded interfaces (mode 1). The bond interface works fine on both servers as long as xen is not used.
When using xen's network-bridge script (with netdev=bond0), one server works fine without problems, and the other starts upping and downing the slave interfaces of the bond in a very fast manner (several runs per second). Both configurations are identical, the only difference is: - the working server has tg3 nic drivers - the dysfunctional server has e100 nic drivers.
Does it really depend on the nic driver if bonding works? And why does it not work in xen, but work fine without xen? What can I do to analyse the problem?
I know I could ask this on the xen list, but I fear that this is a redhat/centos specific phenomenon.
Any hint or help is appreciated.
Dirk