>On Wed, 2005-05-25 at 19:02 +0100, T. Horsnell wrote: >> Hi all, >> I have a system with dual gigabit ethernet ports. >> I've assigned them each an ip address, and am using >> tcpdump and ping to check that they work as expected. >> They dont. >> >> If I ping, from another box, the address associated with eth1, >> the traffic appears on eth0. This only happens when I ping from >> another Linux box - if I ping from one of my Alpha boxes, everything >> works as expected, i.e. pings addressed to the ipaddress >> allocated to eth0 appear on eth0, and pings to the ipaddress >> allocated to eth1 appear on eth1. >> >> I'm running FC3 with kernel 2.6.10 on the dual-eth box, and I've >> tried pinging from RH7.3, RH9 and FC3 boxes, all with the same >> result. >> >> It appears from the tcpdump output below, that ls1 is returing the >> mac address of the eth1 interface *and* the address of the eth0 >> interface. Is this proper? What am I doing wrong? >> >> Cheers, >> Terry. >> >You are using 2 interfaces on the same subnet on this machine. This has >historically never worked reliably. > >If you put one of the interfaces on one subnet and the other on a >different subnet I expect it will work better (and as expected). I see. I think that will require a lot of re-jigging but I'll keep it in mind. For the moment, I've added a pile of static routes to the ls1 host to force the use of eth1 when talking to the subset of machines for which I want to use eth1. Thanks for your help, Terry.
