martian source and multiple interface

[kireito]

Hello!

I've got a DNS server witch has 2 NICs connected on 2 differents networks.
The first interface, eth0 is used for administration usage (ssh, snmp, backup...)
The second interface eth1 is used for user service (DNS request).

The problem is that when I want to reach eth1 from a device on the same LAN as eth0, it doesn't work and I get a "martian source" log in /var/log/messages.

In fact the server detects a source adress on eth1 which should come from eth0 so the packet is drop by the kernel.

Does anyone as a solution to permit such trafic?

I've tried to play with ip routes, but that didn't give a solution...

[RDU]

This is quite unusual. It's a best practice to have a management network but your management station should be in it.

In this case, it's normal that your kernel detect and block what it's called martian source.

This need to do advanced routing (giving different route depending of the protocol used).
I do that long time ago and don't remember exactly.
Take a look at http : // lartc . org / howto

[kireito]

Hi!

I've worked on this subject. The link you gave is quite a good solution and permit me to resolve some problems with other stations. But in the case of my DNS server, advanced IP routing didn't get a solution.

In fact, I think table "local" is prefered to any other table...