On Thu, Jan 15, 2015 at 10:38:50PM +0000, Pedro Caetano wrote: > Hi I have setup openbsd routers running dual homed with another pair of > upstream routers announcing a default route. > Each router has two interfaces, egress and ingress. > r1 - openbsd1 > r2 - openbsd2 > r3 - upstream1 > r4 - upstream2 > > vio0 is the external interface uses a /28 network to talk ibgp with two > upstream routers. > vio1 is the internal interface that also uses a /28 network to interconnect > both openbsd routers. > Each router has its own private AS, talking with both isp peers via the > vio0 (there is only one remote AS) to get a default-route. > The two openbsd routers are also connected via ibgp via the vio1 interface.
If both routers have own private AS numbers they will not have an ibgp session but a ebgp one. Only if they share a common AS number then it is an ibgp session. > BGP works as expected, yet there is a behaviour i find strange. > > By setting vio0 down on r1, shouldn't the local route be removed? No, this is (currently) not happening. The local route (which is staticly declared on the interface) is always valid no matter what. There is currently a lot of work going on in the routing code and maybe we get feature but no promises. > Although r2 announces a valid default-route, the local route in fib is > preferred rendering that network unreachable. Are r1 - r4 sharing on common network? A few things you can do to make the situation better: - set nexthop self (on the session between r1 and r2 over vio1) - use carp on vio1 and "demote carp" in the neighbor section for r3 & r4 -- :wq Claudio
