L2TPv3 Redundancy

[jfraasch]

I have a rather complicated design I am using that involves L2TPv3.

I have about 200 Cisco 1921 routers that will originate an L2TPv3 tunnel.  These tunnels will terminate on a pair of Cisco ASR1001s.

First, I don't believe there is a way to do this in high redundancy mode, ie, I have to create a single tunnel from one 1921 to one of the ASRs and I will try to use routing to use the second tunnel to the second ASR in case of failure of the first router.

My question is this though.  All of my 1921s are connected to a cloud (managed by another company).  The 1921's and the devices hanging off the other side of the 1921, all need to be on the same subnet.  I need to make sure that ARPs are passed through the cloud from one 1921 to another through the L2TPv3 tunnel.  It seems to me that this means the tunnles would need to terminate on the same interface on the ASR1001s but I can't figure out if I can do that or not.

Has anyone run into this issue before and if so, how was it solved?

James Fraasch

[Giuseppe Larosa]

Hello James,

sorry for late answer your questions are interesting.

L2TPv3 provides a point - to -point L2 transport service.

the question of redundancy on the central site IP endpoint could be fixed with the use of loopbacks one with /32 mask and one with a less specific mask with traffic reaching the primary device when the host route /32 is alive.

How to make it possible to have an ARP request to start from a remote site, to reach the central site and to be re-circulated on all the other p2p L2TPv3 tunnels is really difficult.

You should use a different vlan-id for each p2p, then you would need a way to remove the vlan tag in order to create a single broadcast domain on the central site.

Then from that merged broadcast domain you should take the broadcast frame in order send it in all other locations.

shortly you would need a VPLS service that cannot run on the C1921, so you should look for a VPLS service provided by a provider.

in your current network scenario this controlled multicast replication cannot be achieved unless you had an external switch with advanced capabilities in SP switching.

What you can give is a VRF lite separation with remote sites in their own IP subnets carried over the L2TPv3 tunnels.

This is feasible.

By the way the requirement to have a single IP subnet spanning over 200 remote sites is simply wrong for whatever reasons.

Hope to help

Giuseppe