Martin Michlmayr wrote:
First I can ping all ip addresses from a remote pc. eth0 is the built-in nic on the Qube. If I unplug eth1 I can still ping the ip 10.32.8.112. If I unplug eth0 all ip addresses loose connection. I have never seen this behavior before.
Where are the pings originating? I presume that you are logged on to the qube via eth0, and pinging the eth1 address from there?
If you tried to ping the eth1 address from an external machine, I don't see any way that this could happen :-)
The exact same thing happens on my Ubuntu desktop machine - the "cable fault" isn't detected, the interface is not stoped, and local pings to the interface address continue (because after all, pings on the internal machine do not travel out onto the cable)
It's possible that unplugging the cable *should* have caused the interface to stop, but it doesn't appear to be anything specific to the qube/debian.
No, it's a configuration error. I'm not a network expert, but I did the same once and I was told you cannot have two devices on a machine which are in the same sub-net (or have the same broadcast address or whatever).
I hope you've confused this with something else - having multiple NICs with same-subnet (e.g. adjacent) IP addresses is normal practice in many environments. In fact, having multiple aliased interfaces works too (i.e. eth0, eth0:0, eth0:1 and so on)
The situation could get more confused in environments where it's easy to rewrite MAC addresses, like with Sun equipment. But otherwise, ARP will resolve the situation happily.
-jim