QinQ Tunneling - basic question

[Robert Saurer]

Imagine, customer XY wants to connect his two sites A and B over a SP's network with QinQ tunneling technology in the SP's network.

The customer has two vlans (10 and 20) spanning both sites. Each site is connected with a dot1q trunk to the SP.

There is an active HSRP router in vlan 10 at customer site A sending hello frames with src mac 00-00-0c-07-ac-00.

There is an active HSRP router in vlan 20 at customer site B sending hello frames with the same src mac 00-00-0c-07-ac-00.

The SP switches will receive the customer dot1q frames and add an additional dot1q label between the source mac and the existing dot1q label. Let's say the tunnel Vlan for customer AB is Vlan 30.

Now the SP switches will populate their forwarding tables with endcustomer MAC addresses. Normally MAC addresses are unique even across multiple VLANs.

There are two "hosts" with the same MAC in different vlans.

So the SP switches will update their forwarding table each time a frame is received with src 00-00-0c-07-ac-00.

The problem that I see is that it will flap between sites A and B, depending from where the frame was received last.

Can someone please tell me if I am misunderstanding the technology or if this is really the case.

Regards,

Robert

[ebreniz]

I think configuring multiple HSRP groups for different VLANs can solve your problem. check this document

http://www.cisco.com/univercd/cc/td/doc/cisintwk/ics/cs009.htm#xtocid122333

[rwcrowe]

the last 2 digits of the mac address (in your case 00) are taken from the HSRP group number. the hsrp they configured did not specify a group number, ie "standy ip x.x.x.x" instead of "standby 1 ip x.x.x", 1 being the hsrp group number. if you have then just reconfigure hsrp so they use a group number and not the same group number at both sites you will be fine. i know this is a work around, but like you I dont know QinQ very in-depth so I have the same question you do.

[blakelyholman]

HSRP hello messages come from the real mac on each router, not the virtual one, right? Otherwise, I concur with the previous poster's answer that if different HSRP groups are used with different IP subnets, there shouldn't be a MAC confict.