[c-nsp] Pseudowire VC Types on Cisco

[James Bensley]

Hi All,

Can someone please clarify for me why the following pseudowires ARE working?

This is a lab scenario, here is a link to the topology, I want to test
a pseudowire that will carry multiple tagged C-VLANS:
http://null.53bits.co.uk/uploads/networking/cisco/mpls/pwe3-testing/EVC-Mac-Issue.png

The 1941 is acting as two CPEs, it has gi0/0.11 in vrf1 and gi0/1.11
in vrf2 configured as 192.168.0.1 and 192.168.0.2 respectively. To
ping from gi0/0.11 (192.168.0.1) to gi0/1.11 (192.168.0.2) the traffic
needs to pass over a pseudowire between two PEs, the ME3800 and
ASR1001.

This is the configuration and results (same config and "show" output
on both PEs) when configuring the EoMPLS pseudowire on the PEs using
an xconnect on an EVC with dot1q encapsulation set:
http://null.53bits.co.uk/uploads/networking/cisco/mpls/pwe3-testing/IOS%20to%20IOS-XE%20EVC%20tagged.txt


interface GigabitEthernet0/1
 description ME3600-Gi0/3
 switchport trunk allowed vlan none
 switchport mode trunk
 no cdp enable
 service instance 11 ethernet
  encapsulation dot1q 11
  xconnect 10.0.0.2 123 encapsulation mpls

ME3800#show mpls l2transport binding 123
  Destination Address: 10.0.0.2,VC ID: 123
    Local Label:  19
        Cbit: 1,    VC Type: Ethernet,    GroupID: 0
        MTU: 1500,   Interface Desc: ME3600-Gi0/3
        VCCV: CC Type: CW [1], RA [2]
              CV Type: LSPV [2], BFD/Raw [5]
    Remote Label: 20
        Cbit: 1,    VC Type: Ethernet,    GroupID: 0
        MTU: 1500,   Interface Desc: ME3600-Gi0/4
        VCCV: CC Type: CW [1], RA [2], TTL [3]
              CV Type: LSPV [2]


Ping works fine. My question is why is the VC type “Ethernet” (which I
assume is VC type 5?) when I haven’t configured a “rewrite” operation
to PoP the VLAN tag before the frame enters the xconnect? Since I want
to carry the dot1q tag I want this to be VC type 4 surely? Is it
simply that IOS negotiates to VC type 5 by default?

If I reconfigure the xconnect’s as follows the VC type becomes “Eth
VLAN” (which I assume is VC Type 4?) so now I assume the C-VLAN is
being carried across the pseudowire:
http://null.53bits.co.uk/uploads/networking/cisco/mpls/pwe3-testing/IOS%20to%20IOS-XE%20EVC%20tagged%20VC-4.txt


pseudowire-class EoMPLS-dot1q
 encapsulation mpls
 interworking vlan

interface GigabitEthernet0/1
 description ME3600-Gi0/3
 switchport trunk allowed vlan none
 switchport mode trunk
 no cdp enable
 service instance 11 ethernet
  encapsulation dot1q 11
  xconnect 10.0.0.2 123 pw-class EoMPLS-dot1q


In both cases ping works. However as you can see in the configs, the
ME3600 which sits in the middle of the lab tying everything together
at layer 2, this is configured with EVCs that expect tagged frames.

In the first scenario with VC type "Ethernet", is the dot1q tag being
stripped by the PE despite the lack of “rewrite” command because it is
VC type 5? And when the frame lands at the other PE is is added back
on again because the EVC is configured with “encapsulation dot1q 11”
again, even though there is no “rewrite” command here either? Or
because it is VC type 5 and the VLAN ID is signalled during pseudowire
setup so the remote PE and push it onto the frame before transmission
onto the remote access circuit?

Sadly this is a remote lab so I can't easily go there and plug my
laptop in and run wireshark

In both scenarios the frames are coming out of the PE towards the
ME3600 layer 2 devices as tagged, I can see the dot1q frame count
increasing with "show int gi0/4 controller | i Dot"

Any help is appreciated.

Cheers,
James.