multiple NIC / same subnet

Hi
Kernel : 2.4.19

I have 2 NICs on same subnet with IP 192.168.0.76(eth0) and
192.168.0.72(eth1).

I initiated a data transfer from another machine to a server running
on the above machine. I gave the server IP as 192.168.0.72 on the
client side. During the transfer if i unplug eth0,
- the data transfer is hanging
- iam unable to reach the server machine even thru eth1 which is
connected
- iam not able to reach any machine on 192.168.0.0 network from the
server machine

I understand this is bcos the route to 192.168.0.0 is using the device
eth0 which is the first entry in the routing table. is there anyway to
avoid this , so that linux will automatically failover to the second
route?

thanks
ssk

ssk wrote:
> I have 2 NICs on same subnet with IP 192.168.0.76(eth0) and
> 192.168.0.72(eth1).

Having two NICs on the same subnet is never a good idea; this has
been discussed here multiple times over...


> I initiated a data transfer from another machine to a server running
> on the above machine. I gave the server IP as 192.168.0.72 on the
> client side. During the transfer if i unplug eth0,
> - the data transfer is hanging
> - iam unable to reach the server machine even thru eth1 which is
> connected
> - iam not able to reach any machine on 192.168.0.0 network from the
> server machine

There are two reasons that You cannot get to Your server/service on
that box: Firstly, depending on the service, if it is configured to
only listen on the eth0 IP, it will not listen on eth1. Secondly,
even if it serves the other IP, it will try to send the reply via
eth0, which will fail when eth0 is no longer connected. So on the
remote end, it looks like "unreachable", although it is the other
way around: You can reach the server very well, but the server can-
not reach You any more.


> I understand this is bcos the route to 192.168.0.0 is using the device
> eth0 which is the first entry in the routing table. is there anyway to
> avoid this , so that linux will automatically failover to the second
> route?

Correct. - If You remove eth0 from Your routing table, things will work
again.
As for automatic failover and things, see the Linux Advanced Routing
HowTos which You will find spread everywhere over the net.


Cheers, Jack.

--
----------------------------------------------------------------------
My personal reading of the string "MicroSoft" expands to "NanoWeak"...