This problem is common when running OSPF between a switch (i.e 3550 or 3560) and a router. The error message that is generated when this problem occurs is:
%OSPF-5-ADJCHG: Process 1, Nbr 150.8.5.5 on Vlan258 from DOWN to DOWN, Neighbor Down: Dead timer expired
%OSPF-5-ADJCHG: Process 1, Nbr 150.8.2.2 on Vlan258 from EXSTART to DOWN, Neighbor Down: Too many DBD retransmitions
The resolution is to either adjust the system MTU on the switch or have OSPF ignore the MTU.
A) Change the system MTU on the switch (system mtu 1500 or system mtu routing 1500)
B) Have the router and/or switch ignore the MTU (ip ospf mtu-ignore)
C) Change the interface MTU on the router (GigE only)
Note that using the system mtu 1500 option requires a reboot of the switch but the system mtu routing 1500 does not require a reboot.

Which method is preferred for production environment?
Changing switch system mtu (any side effects?) or “ip ospf mtu-ignore”?
(not taking into account that switch requires reboot to change system mtu, run into it once when IRB bridging on a router to a 3750 switch)
I usually use solution B, but in some cases, it was enough to configure “ip ospf mtu-ignore” on the switch, and in some other cases, I had to configure “ip ospf mtu-ignore” on both the switch AND the router.
I had not been able to identify why. Any hint ?
Thanks,
Navid
In answer to the question about which device to enter the command ‘ip ospf mtu-ignore’ on, you need it on the one with the smaller MTU.
Basically, the one with the smaller MTU is getting DBD (DataBase Descriptor) messages from the other device that indicate a bigger MTU than the local device is capable of handling. So the local device ignores the packets.
HTH
Andrew