[c-nsp] BFD and EoMPLS

[Michael Robson]

I have a test network of three Cisco 6500(Sup720s) running SXI5 connected together in a triangle. Each is iBGPed to the others using loopbacks advertised around this testbed using OSPF and Label switching is also turned on. I then have two clients connected one to router A and the second to router B with an EoMPLS pseudowire connecting them (the pseudowire is terminated against the same loopback of the 6500s used for the BGP peerings). If I then get the clients to continuously ping each other and break the link directly connecting routers A and B, there is loss of connectivity for about 6 or 7 seconds before the pseudowire re-routes via router C. If I configure BFD for each link and also attach it to OSPF and repeat the experiment, the time to re-route is still 6-7 seconds.

I can reason that this might work or that it shouldn't: does this not work because there is no direct hook between BFD and LDP or should it work because the pseudowire terminating loopbacks are advertised in OSPF, OSPF has been hooked into BFD and ultimately the LFIB is built via the FIB and LIB (or perhaps with this testbed because the size of the network and Penultimate Hop Popping it means that the FIB is used anyway)?

Ta,

Michael.
--