It's because intra-area LSAs are preferred compared to inter-area LSAs. The super-backbone(MP-BGP->OSPF) introduces LSAs from the remote site as Type-3 into the LSDB of a local PE and if the local PE happens to have a Type-1 LSA from a directly connected link than that one is preferred. So you can either make both LSAs to appear as Type-1 using sham-link or you can make both to be type-3 using different areas on each site.
Or you can use BGP as everyone else and can forget about all this mess above. adam > -----Original Message----- > From: cisco-nsp [mailto:[email protected]] On Behalf Of > CiscoNSP List > Sent: Wednesday, June 18, 2014 7:12 AM > To: [email protected] > Subject: [c-nsp] BGP vs OSPF (CE -> PE) > > Hi Everyone, > > We typically use OSPF (CE/PE) so customer can advertise routes into their > VRF - We have issues with failover (When customer site has 2 links) but the > links go to different PE"s of ours (We only have agg's from carriers on > certain > PE's).. > > eg. > > Customer(vrf) has a site(foo) connected to PE A + B (PE B is "failover" link) > > Same customer has another site(bar) connected to PE B. > > Traffic from site "bar" to "foo" will go via PE B, which is not what we > want...we have manipulated this to work via longer subnets (i.e. failover link > advertising a /23 instead of a /24), but this isnt always feasible. > > Would BGP(Instead of OSFP) help in this situation...i.e. Can we manipulate > how the routes are advertised(PE A/B) within the vrf more easily if the CE > advertises via BGP vs OSPF? Or any other suggestions? > > Cheers > > > _______________________________________________ > cisco-nsp mailing list [email protected] > https://puck.nether.net/mailman/listinfo/cisco-nsp > archive at http://puck.nether.net/pipermail/cisco-nsp/ _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
