[c-nsp] Weird BGP Routing Problem

[Paul Stewart]

Hi there...

Sometime this morning, we noted a sudden increase in outbound traffic to one
of our transit providers.  Have now realized that some routes that should
prefer peering are now going via transit.

What makes this very strange is that in our routing table, the best route
chosen is not being honoured - very confused about this...

Below is an example:

core1-rtr-to#sh ip bgp xxx.xxx.xxx.105
BGP routing table entry for xxx.xxx.xxx.0/18, version 36369947
Paths: (5 available, best #1, table default)
  Not advertised to any peer
  xxxx
    xxx.32.245.67 from xx.75.100.39 (xx.75.100.39)
      Origin IGP, metric 0, localpref 200, valid, internal, best
      Community: 5645:5000 11666:2000 11666:2001

Highest localpref, low metric and all kinds of other good reasons state this
is the best route.  But this isn't the route being chosen and I don't know
why...??

The route being chosen is one AS hop longer, local-pref of 100 (instead of
200 above) and for what it's worth a metric of 50 (not that it matters at
this point)

Feeling kinda silly here - but why is the route not the best route? ;)

Cisco IOS Software, c7600s72033_rp Software
(c7600s72033_rp-ADVENTERPRISEK9-M), Version 12.2(33)SRC, RELEASE SOFTWARE
(fc3)

Thanks for any input...

Paul