[c-nsp] Synchronize / mutal redistribute global and vrf BGP routing table

[Dennis Breithaupt]

Hello folks,

I'm a bit confused with a bgp-problem and request some hints from you :)

We're using multiple 38xx and 28xx routers as headends for an
large-scale DMVPN design using Server Load Balancing to balance the
VPN-sessions to mGRE-tunnels on the headendrouters.

Between outposts (1200+) and central office we run EIGRP and
redistribute it to iBGP connecting to the internal core routers.

Now we need to run either full-mesh iBGP or route reflection, so that
every headend router knows every route from another headend, so that
traffic could be routed directly between headends, without going over
the central router. (I need to adjust next-hops a bit using route-maps,
but I'm aware of this...)

Both variants do work well, except we've a problem with vrf-processes on
our headends. To segregate the outposts and keep EIGRP-stub-updates
lower, we're using two mGRE-interfaces on the headends. One in global
routing and one in an vrf.

As the physical router has only one BGP router id of course BGP from vrf
ABC would not peer with the "global BGP" on the same router. Also route
reflection from the global router over a RR to the "vrf-BGP" does not
work, because it says that itself was the originator and denies to
install the route.

So every physical BGP table misses the routes from the vrf-table on the
same system and every vrf-table misses the BGP-routes from the main table.

I also thought, that I could import/export prefixes between global and
vrf-routing table, but that seems not to work, because I can only import
from global to vrf but not the other way round.


-> How can I synchronize BGP-routing tables between all physical and all
vrf-routing tables in my setup? - probably using RR or import/export on
the router?


Thank you very much in advance,

Dennis