[c-nsp] xconnect on ASR9k - Type 5 PW, what happens with VLAN tag?

[Adam Vitkovsky]

Hello Marcin,
> Marcin Kurek
> Sent: 04 June 2015 15:37
> 
> I didn't change the MTU of TenGig interfaces of ASRs towards 4948, so
> it's default 1514.
Right that's because XR doesn't include L2 overhead in the MTU so the 14B there are to account for the L2 header.

> 
> RP/0/RSP1/CPU0:ASR9006#sh l2vpn xconnect detail
> Group TEST, XC A, state is up; Interworking none
>    AC: TenGigE0/0/1/5.2001, state is up
>      Type VLAN; Num Ranges: 1
>      VLAN ranges: [2001, 2001]
>      MTU 1504; XC ID 0x1080007; interworking none
>      Statistics:
>        packets: received 402, sent 23797
>        bytes: received 361925, sent 1909567
>        drops: illegal VLAN 0, illegal length 0
>    PW: neighbor 2.2.2.2, PW ID 2001, state is up ( established )
>      PW class PW_CLASS, XC ID 0xc0000003
>      Encapsulation MPLS, protocol LDP
>      Source address 1.1.1.1
>      PW type Ethernet, control word disabled, interworking none
>      PW backup disable delay 0 sec
>      Sequencing not set
> 
> I have two questions.
> 1. PW type Ethernet means Type 5 PW (port-based PW). Cisco docs says:
> "The ingress PE transports frames received on a main interface or after
> the subinterface tags have been removed when the packet is received on a
> subinterface. There is no requirement to send a tagged frame over a type
> 5 PW, and no dummy tag is added by the EVC-based platforms".
And the last sentence of the paragraph says: The results of that VLAN manipulation are transported over the type 5 PW, whether tagged or untagged.

> So in this scenario, is the dot1q tag popped by ingress PE even though I
> didn't use "rewrite ingress tag pop 1"?
-no it's not 
-since you mention just the top tag in your example -the top tag might be of a significance at the remote end so the ingress node can't make an assumption that the tag can be POPed safely.


> What is the difference between these two configs assuming Type 5 PW?
> 
> interface TenGigE0/0/1/5.2001 l2transport
>   encapsulation dot1q 2001
> 
The frame with the top most VLAN tag of 2001 is accepted by this sub-interface for further processing 
And since the sub-interface is mapped to a PW the frame is transported as is (with the top tag 2001 and any underlying VLAN tags) over the PW to the remote end.

> interface TenGigE0/0/1/5.2001 l2transport
>   encapsulation dot1q 2001
>   rewrite ingress tag pop 1
The frame with the top most VLAN tag of 2001 is accepted by this sub-interface for further processing
The 1st topmost VLAN tag is removed from the frame 
And since the sub-interface is mapped to a PW the resulting frame is transported as is (with any remaining VLAN tags) over the PW to the remote end.

> 
> What factors determine PW type?
I thing by default the PWs are of type 5
And you change them to type 4 with "transport-mode vlan" 

> 
> 2. How to interpret Access Circuit MTU in the above output? I noticed
> that it changes depending on subinterface configuration (whether I use
> rewrite command or not) and MTU.
Yes unfortunately the PW MTU on XR changes automatically depending on the rewrite operation.
But obviously the system can account for only what it sees so if I know there are double tagged frames coming through e.g. an Ethernet NNI and I only pop the service tag on ingress I have to account for the additional c-tag when configuring the interface MTU.
  
Basically always configure the MTU manually yourself on both ends and account for the rewrite operations performed at each end so that the resulting PW MTU is the same on both ends.

adam