ospf mtu-ignore + multi mtu + routers&switches = RUN!

I'm pretty sure most people on this list remember the big debate on whether
or not opsf mtu-ignore is a bad thing. I can tell that it is a very, very
bad thing.

I had a very simple ospf network with two backbone/abr sites and three edge
routers. All of this is connected on a common vlan.

One backbone (a router) is set for mtu 1500 and no ignore.
One backbone (a 3560 switch) is set for mtu 1546 and no ignore.
Three edge sites running 2801's with ethernet switch modules (damn those
things!). Mtu-ignore and mtu 1500 set there.

Basically the sites were in a rolling DR/BDR election. Nobody agreed on who
was DR/BDR. When the backbone site with mtu 1546 would start sending DBDs
there would be a one-way situation and timeout. Then another election and
round and round we go.

An easy fix right... Wrong!

You can't change the int vlan mtu individually on those switches or ESMs.
The bubble gum patch for now is an ip ospf priority 0 on the 3560. It at
least prevents the flooding of DBDs with a high mtu that no one can use. It
appears to have stopped the rolling elections.

This would have been nearly impossible to troubleshoot without debug ip
ospf adjacency.

Blogs and organic groups at http://www.ccie.net
Received on Fri Dec 16 2011 - 10:45:29 ART