On Wed, May 01, 2013 at 11:45:04PM +0200, Loïc BLOT wrote: > My border routers obtain a default route in fact, and OSPF must > redistribute this route to LAN Routers. Here is a scheme > > > |-------------- R1 site 1------------ R3 Site 1 > | BGP AS 650XX | OSPF a3 | > |-------------- R2 site 1------------ R4 Site 1 > | | > WAN | GRE (OSPF a3) > | | > |-------------- R1 site 2 ------------ R3 Site 2 > | BGP AS 650YY | OSPF a3 | > |-------------- R2 site 2------------- R4 Site 2 > > Each BGP AS redistribute a default route. > you are right, OSPF should redistribute default route (it's the case) > for R3/R4 routers on each site. The problem is between between the two > border routers and on GRE. > Please note R1 and R2 are full mesh GRE (R1S1 -> R1S2 / R1S1 -> R2S2 / > R2S1 -> R2S1 -> R2S1 / R2S1 -> R2S2). > When i said priority it's not route priority but protocol priority (BGP: > 48/OSPF: 40) > > Any idea ? I think the only and the best solution is to filter installed > routes >
OK I see your problem now. The BGP feeds are not sending you a full view but just a default route and so you clash on the default routes. For now the only hack that I know which would work is to have BGP distribute to /1 (0/1 and 128/1) networks instead of the default route because those will be more specific and win over the OSPF default route. I know this is a ugly hack. My plan is two make the routing priority configurable per daemon and I will also look into a Ext-LSA filter option for ospfd. Ext-LSA are strict leave nodes in the LS graph and can be pruned without risk. This will take some time so don't hold your breath. -- :wq Claudio