Am 23.02.2010 16:47, Oliver Boehmer (oboehmer) schrieb: > >> Hello Oli, thx for your support again. I have configured the HUB/PE as >> suggested: >> [..] >> I see that a traceroute from CPE1 to CPE2 now take the path over the > HUB >> and then back to the LNS as expected: >> [...] >> When I remove the def-route on the HUB, I'am still able to reach CPE2 >> from CPE1 directly over the LNS: >> >> cpe1-vrftest#traceroute >> Target IP address: 10.98.2.1 >> Source address: 10.98.1.1 >> Tracing the route to 10.98.2.1 >> 1 10.99.17.254 68 msec 60 msec 64 msec (Loopback102 LNS) >> 2 10.99.17.2 152 msec * 148 msec (CPE2) >> >> So I *can* re-direct the traffic from CPE to CPE through the HUB but > in >> the case the HUB fails, the CPEs are directly connected again through >> the LNS/SPOKE PE. Is that the expected behaviour? Or is there still > some >> thing I'am missing (RPF is enabled on the Vi's)? > > That's strange.. Can you open a TAC case to get this looked at?
Ok, I will do so if I can't get ahead soon. > I just > tried this with "regular" serial interfaces, and I don't see the issue, > i.e. without a default route, the CEs don't see each other. I assume even without any MP-BGP between the SPOKE and HUB PEs, it should be possible to isolate two interfaces on the SPOKE/PE with the Half Duplex VRF feature enabled. I'am right here? So how looks your SPOKE/PE test setup regarding the VRF configuration (VRF definition, interfaces and static routes for that VRF)? That would be interesting for me. Maybe I can build a similar setup with some unused FastEth's in my LNS/SPOKE/PE. > Can you remove urpf and try again? I've tried that, looks like uRPF has no influence. I get the same resluts with and without. Thx a lot so far! -- Gerald _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
