On 2012-12-12, Loïc Blot <loic.b...@unix-experience.fr> wrote: > Hello to OpenBSD Community. > I am testing OSPF + BGP dynamic routing. > I'm happy to see OSPF learn BGP learnt routes natively. I have a problem > with my default route. > > > > As you see Pala1 (one of the two main router) learn default route from > 14.14.14.1 (which is my simulated backbone router). OSPF must > redistribute also default route to GWPala1 (and GWPala2) which are the > client gateways. > Accordingly, OSPF also redistribute the default route between Pala1 and > Pala2 (the second main router), and there is a routing loop. > In fact OSPF is prior than BGP and i want default BGP route be prior > than OSPF. > How can i proceed to break this priority for the default route only ? > At this moment i have thought: > - Block OSPF packets from the other router
Then OSPF won't work... > - Recompile the kernel to make BGP prior than OSPF This wouldn't be the kernel - daemons add routes with priority RTP_BGP / RTP_OSPF - this would be a wrong approach too. Say you have an IXP-facing /24 which is advertised from external sources over BGP and internally over OSPF, you wouldn't want to prefer the external route.. > Is there any option to block route from some peers on OSPF ? Have you > got another idea ? ospfd doesn't allow you to filter the routes imported to the kernel, In a similar scenario I ran bgpd on the internal gateways and fed a default route to them that way instead, then use OSPF for loopbacks, peer links (/30's / IXP subnets) and internal subnets (i.e. all external "internet" routes are carried in BGP). Another way to handle this would be to have a carp address on the inside of the main routers and on the client gateways, point a static default route to that carp address, you can again still use ospf for announcing networks behind the gateway.
