I have the following scenario going on for an MPLS Lab. The diagram below shows the topology:
R1 through R6 are an MPLS core. R1 and R6 are PE routers, R2, R3, R4 and R5 are P routers. Each Frame relay network is a separate subnet.
The core is running OSPF and MPLS is enabled on all interfaces.
My problem has to do with the FR between R1, R2 and R3. In this Configuration, R3 is a spoke, and R1 and R2 are hubs that do not have a direct circuit between them. There is an FR map pointing to R3 for each router to reach the other.
With the OSPF core converged, the I've noticed that the R1 decides to reach R6 through R2 to R5, despite that they are not OSPF neighbors as they do not have a direct FR circuit. When this happens, the CE devices cannot ping each other across the MPLS VPN, and even R1 cannot ping the vrf routes received from R6 and vice versa.
When I take that R2 interface so that it does not neighbor with R3, R1 will then decide to reach R6 thorugh R3 to R4. In this case CE devices can ping each other and R1 can also ping vrf routes received from R6 and vice versa.
I'm not sure what the issue is here. Is the problem related to MPLS switching, with Frame Relay routing, or is this not even a valid scenario to begin with and the FR between R1, R2 and R3 have to be fully meshed?
Also good to note that I've seen it happen that when R3 and R2 have already neighbored, and then I get R1 to neighbor with R3, R1 does not receive any OSPF routes from R3 at all. I had to take down R2 in order for R1 to receive routes.
Any help is appreciated. Thanks
