Indirect Next-hop Address Space and Route Reflectors

If you attempt to allocate indirect next-hop indexes for all the active routes in the routing table, the indirect next-hop address space can be exhausted. This occurs most frequently in topologies in which a Juniper Networks router acts as a VPN route reflector and a BGP peer with other vendors' routers. The other routers advertise unique label stacks per prefix, instead of sharing common label stacks across many prefixes.

You can configure a forwarding table export policy to prevent routes from being installed in the forwarding table, even when those routes are active in the routing table. With this change, the routes that are omitted from the forwarding table do not have additional indirect next-hop indexes allocated to them.

The following configuration avoids indirect next-hop address space exhaustion, but does not allow the router to forward traffic for the BGP learned prefixes:

routing-options {

    forwarding-table {

        export kern;

    }

}

policy-options {

    policy-statement kern {

        from protocol bgp;

        then reject;

    }

}