I've been doing some labs that involve the use of MPLS TE primary and backup auto-tunnels. This is a somewhat new concept for me, so I have not had much exposure to this technology.
So far, from what I've learned, this is my concept:
Enableing primary/backup auto-tunnels (next-hop only), creates a "mesh" of TE tunnels and gives the given topology FRR on each link. Then, once this "mesh" accross all links is enabled, we are able to run other tunneling methods ontop of the auto-tunnels. For example, we can run LDP on all links and create LSPs between all PEs...we could also create "manual" TE tunnels between PEs, etc. LSPs built ontop of the auto-tunneled network are all protected by FRR. The auto-tunnels themselves do not actually create end-to-end LSPs between the loopbacks of 2 PEs though. The auto-tunnels by default will enable "autoroute announce" on each tunnel interface, so each node will use each it's primary tunnel to reach far end LSPs. This will by default (with no other label allocation methods riding ontop of the tunnels) not create end to end LSPs. An optional command will enable "mpls ip" on each auto-tunnel interface and LDP adj will be built between all of the links . Enableing "mpls ip" will allow for end to end LSPs without using other label allocation methods ontop of the auto-tunnels.
Are any of you using this feature in production?
Thanks