[c-nsp] asr 920 - lower mpls mtu?

[Lukas Tribus]

Hi Mike,


> Trying to set up an EoMPLS tunnel, the mtu allowed for
> 'l2 vfi somename manual' is a bit short.. only 9180 bytes as 
> opposed to 9216 for all the rest of my me3600's for example.
> 
> asr920(config-vfi)#mtu ?
>    <1500-9180>  MTU size in bytes
>   <cr>
>
>
> I am trying to figure out why this is... am I being silly and this is 
> too high and I should select a lower global default like 9100 or ?

I would consider setting this value to something lower, a value that is likely supported by future platforms as well, with a bit of margin. Otherwise you would have reconfigure the VPLS instance or the pseudowires each time you have a platform not supporting the MTU. The ASR920 has a lower MTU than the ME3600, the c6840x platform has an even lower MTU (9154) for example.

Setting the vfi mtu to the absolute maximum doesn't seem like a good idea (especially if we are talking about VPLS, where you don't want to take down everything to reduce the mtu so you can enable the instance on a new node).


Slightly related: the maximum MTU on an ASR920 interface seems to be 9216, however it appears that packets get drop if higher than 9214 (ip ping to the next-hop IP, non-labeled). Anyone seen this already? If its another limitation I didn't find any documentation about it and it certainly breaks Path MTU discovery/IP fragmentation because the box thinks it can handle 9216 while it can only handle 9214.


Lukas