Question

A service provider running IOS XR has a large IBGP domain with over 200 routers. To reduce the full-mesh requirement without deploying route reflectors, the design team implements BGP confederations, dividing AS 65000 into sub-ASes 65001, 65002, and 65003. After configuration, an operator notices that when routes are advertised between the sub-ASes, the AS_PATH shows sub-AS numbers in parentheses but external neighbors in AS 100 never see these sub-AS numbers. When these routes reach the external EBGP peer, how does the confederation affect BGP path selection and loop prevention?

Accepted Answer

The confederation sub-AS numbers are stripped when advertised to the true EBGP peer, so the external AS only sees AS 65000; internally the AS_CONFED_SEQUENCE segment is used for loop prevention and does not count toward AS_PATH length in best-path comparison.

Answer

The sub-AS numbers are prepended to the AS_PATH and advertised to the external peer, increasing the AS_PATH length seen by AS 100 and making the route less preferred.

Answer

Confederation peers behave exactly like IBGP peers, so the local preference must be manually set on every inter-sub-AS boundary to prevent routing loops between sub-ASes.

Answer

The AS_CONFED_SEQUENCE is treated identically to a normal AS_SEQUENCE and is counted in AS_PATH length comparisons, so intra-confederation paths appear longer than they physically are.