Punching a hole in policy routing

From: Earl Smith (ewsmith@parkcity.net)
Date: Thu Jan 06 2000 - 21:02:44 EST


I have a 2620 that has two T1s connected to two different
providers in preparation for multihoming (no, I don't intend to
take full views). Provider A routes me a /23. Provider B routes
me a /24. The 2620 has 2 WICs and 1 Ethernet.

I have added a secondary IP to the e0/0 so there are addresses
from both blocks on the e0. I have set up policy routing on the
e0 so that packets sourced from each provider's IP block
are routed to the appropriate interface using a simple set of
standard access list statements and a route-map. This all works
fine. Traffic is being routed properly from hosts on each
network to its proper T1. The route map is on e0.

My problem is that the source based policy routing is sending
traffic from each of the two networks (on the same segment) bound
for the other network out the appropriate T1 through the internet
and back through the other T1 back to the host. This is expected
actually.

But this is kinda dumb for accessing a mail server that is
connected to the same piece of wire. I've tried proxy-arp; s
static routes don't work. I know the way I have it set up is
dumb. I just need to make it work temporarily so I have
something to fall back on while I get BGP, etc. set up

What's the secret?

-- 
Regards,

Earl



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:08 EDT