> Yes, I understand this. But if I read your previous post correctly (note
> the "if"), then I assume you'll be IP forwarding across a couple of ether
> cards which serve as their respective network gateways. What happens if
> you have a kernel panic due to some other application and your kernel
> freaks? No traffic...
> If this doesn't apply to your situation, please elaborate on what you
> mean to do.
Just one ethernet card. There is no second wire. Just aliased
interfaces. Ie. eth0 is on one Class C, eth0:0, eth0:1, eth0:2 etc.
are all on another class C but they are still on the same machine
and root device. If the kernel panics on this machine all web sites
hosted by that server go down. But since the mail for that customer
is MX'd to another server on our internal Class C, that customer still
gets mail, all terminal servers still work, core router works, news
box works etc. In short, I don't know how on a virtual web server, you
can prevent all its hosts from going down if it crashes. IP forwarding
is turned on. The server pretends to be a router from its eth0
interface to all the aliased interfaces eth0:0, eth0:1 etc. You could
even subnet lets say 62 hosts on one web server, 62 on another, etc.
The second Class C on each machine is set up, one IP for each web site.