IP Networking

I have a Red Hat linux server X on a x.x.0.0 network. This machine also has to communicate with another server Y on a network called y.y.0.0

Server X has two network interfaces.

eth0 is configured on the x.x.0.0 network and has a default gateway on the x.x.0.0 network.

In order to communicate server Y on the y.y.0.0 network I'm planning on configuring eth1 with a y.y.0.0 IP address and gateway. I then want to add a static route on eth0 that uses eth1's IP address as a gateway.

Example (using fictional IP addresses):

SERVER X -> eth0 192.168.0.1 (add static route 20.10.0.0 gw 20.10.0.1)
-> eth1 20.10.0.1

SERVER Y -> 20.10.0.10

My logic tells me that if I then request to talk to server Y the request will go to eth1 via the static routed on eth0. eth1 is on the same network as server Y, so the request should then be able to reach server Y

Could someone please confirm that his is correct (or incorrect for that matter).

Cheers

If server X has an interface on the 192.168.0.0 network then it knows how to easily reach 192.168.0.0 addresses. A second inferface on another network won't change that. It should be able to reach either network as needed with out any routes.

You need an "interface route" for every interface, that is: a route with the network your interface is in as destination and your interfaces adress as gateway.

Example: you have an interface with the adress 192.168.1.100 which is part of the network 192.168.1/24. Add a route to destination network 192.168.1.0 with gateway 192.168.1.100.

As you have two interfaces you should have two of these interface routes, one for each interface. You then should define a "default route", which is a catch-all for every packet not already dealt with by the interface routes. This route should point to a gateway router on one of the networks you already participate in with your interfaces.

I can't understand what a static route between your interfaces (from eth0 to eth1 or vice versa) is supposed to achieve. If you want to send packets coming in on one interface over to the other network: this is the functionality of a router/gateway. Switch on IP-forwarding to accomplish this but be careful to understand what you are doing when you do this. You might easily compromise any network security by doing so.

bakunin

Hi all

I think the bit where I'm getting confused is the default gateway. I understand it's a catch-all for requests that can't be resolved on the two network interfaces. However, the default gateway is connected to eth0 as specified in /etc/sysconfig/network-scripts/ifcfg-eth0. Shows you how little I understand of networking on a linux system.

Thanks all for answering my question. I shall steer clear of static routes :-)

The "default route" may be shown as an attribute for eth0 but *every* IP host has one and exactly one default route.

A route is telling the IP stack (and, regardless of the number of interfaces, there is only one IP stack per host) that "everything directed to network <destination> is to be sent to <gateway>, <gateway> will know how to handle it". "<network>" in this regard is a bit of a stretch, it can be a "network" consisting of only one host, if the subnet mask is all bits set to 1.

A "default" route is nothing different from that, just that "destination" is not a network or a host, but "everything not covered by other rules".

For instance:

your IP networks: 192.168.1/24, 192.168.2/24 in each of them your own IP adress will be 100. There is a router on network 192.168.1, which connects the network(s) to the world and has the adress 1 in the network.

Your IP address layout:

eth0: 192.168.1.100 / 255.255.255.0
eth1: 192.168.2.100 / 255.255.255.0

Your routing table (i do that on an AIX box, so it looks a bit different from Linux):

Code:

# netstat -rn
Routing tables
Destination        Gateway           Flags   Refs     Use  If   Exp  Groups

Route Tree for Protocol Family 2 (Internet):
default            192.168.1.1       UG        9    229328 en0      -      -   =>
192.168.1/24       192.168.1.100     U         0         0 en0      -      -   =>
192.168.2/24       192.168.2.100     U         0         0 en1      -      -   =>

The IP stack will now work like that: if it has to send a packet to a destination on one of "its own" networks (192.168.1/2.x) it will do so using eth0 or eth1, depending on rule 2 or 3.

If it has to send a package to 192.168.3.100 (or any other network), it will first consult all the rules it knows (2&3), finding out that no one fits. Now it falls back on its default route, which tells it to send it faithfully to 192.168.1.1, a host who (hopefully) will know what to do with it. Now it again looks up how to find 192.168.1.1 and - bingo! - it knows from rule 2 how to reach that host.

If the router would be moved from 192.168.1.1 to 192.168.2.1 you would have to change the default route and in the same way the packet would leave the host via eth1 instead of eth0.

I hope this clears things up.

bakunin