BGP Route Selection - oldest

[CHARLES PARISH]

Is there any way to determine the age of a route/prefix in a bgp table?

Once a path has been selected as "best", is there a direct way of determining the criteria that actually drove the selection?

Most of the criteria can be seen in a "sh ip bgp prefix". However with all higher precidence criteria being the same, or ignored (in the case of MED if you haven't coded "always compare med"), how can you analyze the route selection? If I just go on age of peer, I have routes that appear to have the "wrong" one selected. But it would seem that peer age would not necessarily be accurate representation of the age of the route.

We have large descripancy in the amount of outbound traffic to 2 major ISP's, and want to better understand route selection and possible do some tweaking to better distribute the traffic.

[Giuseppe Larosa]

Hello Charles,

this is a good question.

Age of currently installed BGP route can be  seen in sh ip route and this could give some insight on a possible route flapping on oldest eBGP peer that has caused the route of another eBGP to be installed.

Age of non best path routes can be only guessed as minor equal to eBGP session established state uptime.

What we do for upstream providers is to graph with MRTG the number of IP prefixes received from each, this allows to detect temporary variations on this number over time for each eBGP peer.

It is likely that if something happens on parts of internet more then one prefix is involved, also if the session fails the received prefixes falls down to 0 giving visual evidence of this fact.

Excluding any form of debug  that cannot be kept on all time and could tear down the router in case of full BGP tables you could use debug ip routing with a specific access-list to monitor some specific IP prefix this should give some messages when best path for the monitored IP prefixes change.

there is also a feature called ip route profile that allows to track changes to IP routing table over time. Again it should be tested if this is compatible with multiple BGP full tables on the router.

Possible corrective actions include the use of route-maps to selectively increase local preference based on AS path access-lists to distribute load in the outbound direction.

This requires monitoring and corrections/tuning over time.

Hope to help

Giuseppe