We appear to having some issues with the hold down routes on our c6509e border routers. We use a collapsed core model, so the border routers also act as IBGP route reflectors to the core routers.
The core routers advertise (IBGP) routes to the border routers, which then advertise (EBGP) upstream to the internets.
If the core stops advertising internal routes to the borders, the border router hold down routes continue to advertise our networks upstream to the internets, without the risk of being dampened(if there is any flapping).
When this happens the IBGP routes learned from the core are removed from the border router FIB's, and replaced with static-hold down routes. The problem we are seeing is that when the BGP routes are re-learned from the core, they are placed into the RIB, but do not replace the static-hold down routes in the FIB table. The main issue is that any external ingress traffic destined for hosts in the core hit the border routers, but obviously goes no further.
As soon as I remove the static hold-down routes from the borders, the IBGP routes get added back into to the FIB, and external ingress traffic can reach hosts in the core. After I re-add the hold-down routes, the IBGP routes continue to remain in the FIB.
Has anyone else struck this problem before? Sounds almost like an IOS bug.
 e.g. âip route 192.168.0.0 255.255.252.0 loopback0 250â admin distance = 250