Showing results for 
Search instead for 
Did you mean: 

L2protocol forwarding on 9k


today we were performing some tests in our lab to verify that l2protocols such as LACP,STP,CDP were being forwarded to the end of a circuit. For this purpose we set up two ASR 9906 equipped with a A9K-RSP5-SE RSP and a  A9K-24X10GE-1G-SE line card.
The configuration was just two service instance under a physical interface cross-connected with a pseudowire, I used IXIA to generate traffic to test that l2 control traffic (namely STP) was being correctly forwarded. In order to do so, I generated traffic that had as destination mac 01:80:C2:00:00:00.

The configuration was rather basic:

interface TenGigE0/1/0/1
description *** TO IXIA 5/11 ***
mtu 9212
load-interval 30

interface TenGigE0/1/0/1.86 l2transport description Test L2 Forwarding encapsulation dot1q 86

On the other end it was basically the same, the mpls pseudowire used, has a pw-class applied to it that enables the control word and nothing more

According to this document:

In order to ensure that l2protocol control frames are being forwarded, I need to configure the "l2protocol cpsv tunnel" command. However, I observed that even without the command, the traffic was getting to the other end of the circuit.

Does the 9k platform forward these by default (and if that's the case, why have the command?) or did I set up my test scenario wrong?

VIP Guru

Mark Pace Balzan

Hi Thomas,


my understanding of using "l2protocol cpsv tunnel" is that this command will change dst mac of cpsv frames (so CDP, PVST, STP, VTP). the reason for this is that some platforms like the ME3400E changed tthe dst mac when l2tunnel was used to encapsulate such traffic in QinQ, so if you have this situation you need the 9k to recognise it and work with it.


However when you have two 9ks doing the forwarding, then I dont believe it is a strict requirement.


For the PW on the 9k the default type (dont recall if it is typ4 or type5 by default) should work just fine, it does for me !


I see you have a subinterface with vlan86 connected to your IXIA, so presumably your IXIA is sending vlan tagged frames. Do you want to preserve that vlan across the PW or not ?


You may want to use 'rewrite ingress tag pop 1 symmetric'  or alternatively set the main interfaces as l2transport like so:


interface TenGigE0/1/0/1

Hope this helps





MHM Cisco World



I ended up asking the TAC about this, the forwarding of l2 protocols control frames, even when tagged, is implicit on 9k platforms. The CLI detailed in that document I linked in the first post, is for backward compatibility with legacy catalyst switches