BGP Issue

xzevallos · ‎11-02-2007

I have an MPLS VPN based Campus LAN in my lab. It consists of two P core routers connected to two switch blocks A and B, each switch block containing two PE routers. The P routers are acting as iBGP route reflectors for the PE routers. The primary PE router in each switch block has a higher local preference value.

In the PE routers of switch block A, I see the VRF routes of switch block B learned by iBGP, all pointing to the loopback address of the primary PE router in switch block B, as expected. When I turn off the primary PE router in switch block B, it takes a long time for the routes to point to the loopback address of the secondary PE router of switch block B. When EIGRP is used as the global routing protocol, it takes 45 seconds. When OSPF is used, it takes 90 seconds.

Why does it take 45 seconds for BGP to converge with EIGRP and 90 seconds with OSPF? Is there any way to reduce the BGP convergence time in this scenario?

royalblues · ‎11-04-2007

Will it be possible for you to post a brief diagram of your lab setup

Narayan

cisco_lad2004 · ‎11-04-2007

EIGRP hold timer are 15s where as OSPF dead interval to 40s.

The convergence time is dictated by this sequence of events:

1-Primary PE fails.

2-IGP to declare LoopBack dead at same time BGP timer holdtimer is also ticking default 180s.

However, "BGP next-hop address tracking" is default and tracks changes in RIB, which means we are back to IGP convergence.

So to answer ur 1st Qs, please confirm ur IGP timers.

As far as increasing BGP convergence,test ur config by adding "Fast Peering Session Deactivation" or fine tune ur IGP timers.

HTH

Sam