[c-nsp] MTU on XR

[Nick Hilliard]

On 25/08/2014 12:36, Mikael Abrahamsson wrote:
> So which one is correct? I have talked to numerous people and they have
> different experience and results are contradicting.

looks like you're correct.  Given the following test configuration:

> interface GigabitEthernet0/0/0/10
>  mtu 9200
>  ipv4 address 10.232.1.1/24
> interface GigabitEthernet0/0/0/10.100
>  ipv4 address 10.232.100.1/24
>  encapsulation dot1q 100
> interface GigabitEthernet0/0/0/10.200
>  ipv4 address 10.232.200.1/24
>  encapsulation dot1q 200 second-dot1q 201

The mtu config for each of these is (in order):

gi0/0/0/10:
>   Protocol        Caps (state, mtu)
>   --------        -----------------
>   None            ether (down, 9200)
>   arp             arp (down, 9186)
>   ipv4            ipv4 (down, 9186)
>   ether_sock      ether_sock (down, 9186)
>   vlan            vlan_target (down, 9186)

gi0/0/0/10.100:
>   None            vlan_jump (down, 9204)
>   None            dot1q (down, 9204)
>   arp             arp (down, 9186)
>   ipv4            ipv4 (down, 9186)

gi0/0/0/10.200:
>   None            vlan_jump (down, 9208)
>   None            dot1q (down, 9208)
>   arp             arp (down, 9186)
>   ipv4            ipv4 (down, 9186)

i.e. the ipv4 / ipv6 / arp MTU in each case is 9200 - 14 = 9186.  In order 
to compensate for the single vlan tagging on the .100 interface and the 
double tagging on the .200 interface, the vlan MTU is increased to 9204 and 
then to 9208 bytes respectively.  There is no need to adjust for the vlan tags.

Nick