I've got two routers, A and B. A gets some routes over BGP over a VPN connection, and has a leg on It also can access the whole via its default gateway.

B is on It can access via its default gateway.

There is a router between A and B that I don't control (ie I can't peer with it). I can't use multicast traffic between A and B. I could potentially use a GRE tunnel between A and B to get multicast across, though.

I would like:

  • A to have a route to
  • B to have a route to + the routes A got off the VPN connection

  • How could I set this up?
  • Could I use eBGP, iBGP or OSPF?
  • Can I use OSPF at all when there's that router in the middle and I can't do multicast?
  • If using OSPF was possible, should I use multiple areas? Is it necessary (as I need to do route summarization)?

For the curious, the routers are on Amazon on a VPC each, and the VPCs are joined with a peering connection.

  • 1
    Sorry, don't have time for a long answer at the moment... personally I would create a GRE tunnel and route through it (create OSPF ajacency between A and B over the tunnel). Another possibility would be multihop BGP (that would assume the Router in the middle knows the routes though, which at least for the VPN routes sounds unlikely).
    – Daniel
    Commented Oct 14, 2016 at 22:25
  • 1
    GRE tunnel. That would work. Commented Nov 14, 2016 at 2:39
  • VPC Peering Overview Commented Dec 14, 2016 at 3:47
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can post and accept your own answer.
    – Ron Maupin
    Commented Nov 20, 2022 at 0:10

3 Answers 3


You need a tunnel of some sort.

Even if you could get the routing protocol to run via the uncooperative router, it wouldn't really help you. While your routers would know about the routes they wouldn't be able to actually use them.

IP routing maps a destination address to an interface and a next-hop IP address. If the packets will be sent over a multipoint link layer like Ethernet then the next hop IP address will be mapped to a link layer address and used to send the frame.

But the next hop IP address itself does not become part of the packet. Each router along the path must normally make it's own routing descisions. The uncooperative router would have no idea what to do with your packets.

Hence you neeed a tunnel, you routers can peer with each other over the tunnel and send data down the tunnel to each other. The uncooperative router only sees the outer headers of the encapsulated packets.


First, you don't normally configure a default gateway for a router; the router could be a default gateway for end-hosts. You could configure a default route for a router, but that route is, not something like You could configure a static route for

As waza-ari points out, you could use a GRE tunnel to peer the routers with OSPF if it is a broadcast medium, or you could configure the interface for OSPF as a non-broadcast medium, but you would need a static route to reach the interface of the corresponding router. Also, eBGP could be used with multihop, or iBGP could be used, but you need a static route to reach the corresponding interface of the other router.

  • Comments are not for extended discussion; this conversation has been moved to chat.
    – Ron Maupin
    Commented Jan 7, 2019 at 17:04

If I understand the question correctly.

Try implementing a Point-to-point protocol and bridge the VPN using a secure socket tunneling protocol.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.