[nsp] eBGP routes not balancing

[Jeff Chan]

Hi Dmitri,
Sounds like it could work, and I appreciate your documenting it
for us, but I agree it would not be the prettiest way to accomplish
this.  :)

Cheers,

Jeff C.
__

On Monday, November 17, 2003, 1:52:13 PM, Dmitri Kalintsev wrote:
> Hi Jeff,

> You may want to try to overwrite the next hop for both of your upstreams to
> something obscure (a couple of rfc1918 address for instance) and have this
> obscure thing statically pointing out of your uplink interfaces.

> i.e.:

> 1) create a route-map for AT&T, that would change the next-hop received to
> say 10.0.0.1 for the destination prefixes you want to load-balance to;

> 2) create similar map for Sprint, that would change next-hop to say
> 10.0.1.1;

> 3) create two static routes FOR EACH obscure next-hop:

> ip route 10.0.0.1 255.255.255.255 hssi4/0
> ip route 10.0.0.1 255.255.255.255 hssi11/0

> AND

> ip route 10.0.1.1 255.255.255.255 hssi4/0
> ip route 10.0.1.1 255.255.255.255 hssi11/0

> That should take care of it. Of course you will have to advertise the routes
> to 10.0.0.1 and 10.0.1.1 to the parts of your network where there is no
> default.

> This setup (or should I say "dirty hack"?) should perform a cef load
> balancing for you.

> Hope this helps.

> On Sun, Nov 16, 2003 at 11:50:31PM -0800, Jeff Chan wrote:
>> Hi all,
>> We're running multihomed with two DS3s and have been for a while.
>> After we reset our peering session with one upstream our outbound
>> traffic no longer balances correctly.  Both are large providers
>> and we're getting similarly full routes from both.  Our route maps
>> are stamping identical MEDs and localprefs on all the incoming
>> routes but the bestpath selection is clearly skewed towards one
>> (AT&T which was reset least recently).  I'm wondering if anyone
>> has any ideas how to correct this. 

> [skip to the end]

> ---end quoted text---

> SY,

-- 
Jeff Chan
mailto:cisco-nsp at jeffchan.com
http://www.supranet.net/