RIP and directly connected routes
I'm studying RIP and have a really hard time understanding the rationals behind to automatically pulling directly connected routes. If I connect 4 RIP routers in a circle and neither is advertising it's connected routes, it's going to be very quiet ... Please help me understand :-) Secondly, I use 'protocol direct' to have these connected routes pulled from kernel tables to RIP and advertised. Is that the right way to do it ? Lastly, these 'protocol direct' routes override the split horizon rule. Is there a way to prevent an interface to multicast its own subnet ? Thanks, I 'm just discovering BIRD, I may be asking clumsy questions ;-)
Hi Olivier! On 2015-03-10 10:52, olivier a wrote:
I'm studying RIP and have a really hard time understanding the rationals behind to automatically pulling directly connected routes. If I connect 4 RIP routers in a circle and neither is advertising it's connected routes, it's going to be very quiet ... Please help me understand :-)
Yes, you have to configure directly connected routers using 'protocol direct' or 'protocol static' for each router otherwise it will be quiet.
Secondly, I use 'protocol direct' to have these connected routes pulled from kernel tables to RIP and advertised. Is that the right way to do it ?
Yes, this is a fine solution.
Lastly, these 'protocol direct' routes override the split horizon rule. Is there a way to prevent an interface to multicast its own subnet ?
The RIP uses a split-horizon routing with poison reverse, so the routes that are not in own subnet will send with metric infinity (in default with metric 16). This works well even with 'protocol direct' for me now. Are you sure the split horizon rule is overriden in your case? Maybe try connect routers in a simple line not in a circle. Best Regards, Pavel
[ due to my misunderstanding of mailing lists, the end of the discussion went offline. I copy here for other people / traceback ] Hi, Olivier! On 2015-03-10 14:46, olivier a wrote: I retried with 2 routers inline : 192.168.4.0/24---[R1]---192.168.1.0/24-[BIRD]----192.168.5.0/24 Shouldn't 192.168.1.0/24 be not announced or announced with metric infinity on the 192.168.1.0/24 network ? No :-) The network 192.168.1.0/24 is directly connected network to the [BIRD] router and it knows this route from the bird.conf configuration ('proto static' or 'proto direct'). Split-horizon rule is applied in update message only for routes that are sending to the router from which the [BIRD] router was this routes learned. BIRD routing table : ------------------------------------------------------------------- $ ip -4 route show default via 192.168.1.1 dev eth0 proto static 192.168.1.0/24 [2] dev eth0 proto kernel scope link src 192.168.1.240 metric 1 192.168.4.0/24 [3] via 192.168.1.230 dev eth0 proto bird 192.168.5.0/24 [4] dev eth1 proto kernel scope link src 192.168.5.1 metric 1 ------------------------------------------------------------------- BIRD announces on the 192.168.5.0/24 network : 0.0.0.0 Metric 1 192.168.1.0 Metric 1 192.168.4.0 Metric 2 192.168.5.0 Metric 1 BIRD announces on the 192.168.1.0/24 network : 0.0.0.0 Metric 1 192.168.1.0 Metric 1 192.168.4.0 Metric 16 Here (192.168.4.0 Metric 16) we see a split-horizon usage. Becouse we are sending a update to 192.168.1.0/24 network from which we learned about the network 192.168.4.0. 192.168.5.0 Metric 1 Okay? :-) Yours Truly, Pavel
Thank you for the explanations pavel !! On Mar 10, 2015 10:52 AM, "olivier a" <oatech7402@gmail.com> wrote:
I'm studying RIP and have a really hard time understanding the rationals behind to automatically pulling directly connected routes. If I connect 4 RIP routers in a circle and neither is advertising it's connected routes, it's going to be very quiet ... Please help me understand :-)
Secondly, I use 'protocol direct' to have these connected routes pulled from kernel tables to RIP and advertised. Is that the right way to do it ?
Lastly, these 'protocol direct' routes override the split horizon rule. Is there a way to prevent an interface to multicast its own subnet ?
Thanks, I 'm just discovering BIRD, I may be asking clumsy questions ;-)
participants (2)
-
olivier a -
pavel.tvrdik＠nic.cz