Local Route Leaking Agent

Route Leaking Requires An Interface IP:

It is possible to route leak from VRF-1, which has no Route Targets and does not send or receive routes via BGP, into VRF-2 which does have RTs and sends/receives routers via BGP. An example is route leaking from a local VRF into a VRF that is global, to make those local routes globally reachable, without having to deploy the local VRF everywhere, or to filter which routes are leaked back into the local VRF-1 from the global VRF-2.

With routes from VRF-1 present in VRF-2, the PE allocates a label and advertises the routes over BGP to other routers as is they originated in VRF-2. And example of this is shown below:

router general
   vrf 1
      leak routes source-vrf 2 subscribe-policy LEAK_1_TO_2
      exit
   !
   vrf 2
      leak routes source-vrf 1 subscribe-policy LEAK_2_TO_1
      exit
   !
   exit

When a packet is received on the PE which is destinated for route in VRF-2, the PE pop's the incoming label and needs to perform an IP lookup in order resolve the next-hop of the incomming packet to the leaked route.

EOS is not able to perform an IP lookup on a received packet inside VRF-2 and resolve it to a route leaked from VRF-1, if the device does not have an interface with an IP address in the receiving VRF (VRF-2 in this case).

An example might be that all interfaces in VRF-2 are down, and a packet comes in to the PE destined for a route leaked from VRF-1 into VRF-2. VRF-2 has no active interface with an IP in VRF-2 to perform the recursive IP look-up for the leaked route. In this case a Loopback interface is needed inside VRF-2 so that the PE can always perform an IP look-up for the leaked routes.

---