We currently have two dedicated servers connected to each other with a crossover cable, and each one also has its own external connection. One is running Server 2k3, one is running CentOS with cPanel. What we want to do is configure the so they can run backups to each other through the crossover cable, saving our bandwidth
On the windows machine we have configured it with a static IP of 192.168.0.1 and subnet of 255.255.255.0. This works fine and the machine is running well, on the cPanel machine, when configuring eth1, like so:
the windows and cpanel machines can talk to each other fine through the crossover cable (pinging 192.168.0.1 + 2) but then the cPanel machine becomes unaccessable externally, i.e. ping requests time out and no websites on it can be hosted untill eth1 is disabled and the server rebooted.
Any ideas on how to get this to work?
You might want to go over your config files again and ensure that your machine isn't trying to route out of the secondary interface. Since you can hit your CentOS machine from the backup network, I'd suggest bringing up the second nic, (which should drop your primary connection if I read right), then SSHing into your machine and checking your route table.
'route -n' will do this for you. See what's in the "Gateway" column and go from there. Also if you could paste the output of 'ifconfig -a && route -n' then we can further diagnose this via the forums.
I wouldn't add eth1 as a persistent route simply because you're not doing any routing with that IP at all.
Out of curiosity, what's in /etc/sysconfig/network (the file, that is).
I think there's a rogue gateway statement somewhere that's overriding your main default gw. Also, just to be 100% sure, you don't have *any* references to 'eth0' in /etc/sysconfig/network-scripts/ifcfg-eth1, do you? Sometimes people copy ifcfg-eth0 and modify the IPs, but leave the device name as eth0 which causes the same problem you're having now.
Also if you could SSH into the machine while eth0 is down and run the above 'ifconfig -a; route -n' commands that would give us a 'before and after' dataset which would probably resolve this pretty quickly. Although I understand if you don't want to drop eth0 during the day.