We have recently signed up for Metro Ethernet for our main campus and 8 branches. It's a "mesh" with the carrier providing a virtual layer-2 switch in the cloud. I have a Cisco router at each site running EIGRP and they all have neighbor relationships with one another on their "WAN-facing" interface.
The carrier has provided a second circuit into the main campus, which comes into a different building via a completely diverse path. That's good because there's a lot of construction going on at the main campus. I'm trying to think of the best way to provide failover. Load balancing isn't necessary since each branch has 100 Mbps and the main campus has 1 Gbps, so even without load balancing, there's no way that the 1 Gbps could be saturated. That connection is currently terminating in an ASR1001-X at the main campus.
Each campus also has a backup Internet connection running between 7 and 60 Mbps terminating via DMVPN into an ISR-4431. I mention that because I'm fine if terminating the second Ethernet WAN connection is done in the same ASR1001-X as the “primary” line if that makes it easier.
Here are some possibilities...
- Both connections into the ASR
- Have a normal interface and a backup interface. However, since the WAN connections terminate into a CPE, the link light probably wouldn't go off even if the back-end connection is experiencing trouble. I saw an article on Ethernet CFM but haven't really looked into it yet.
- See if the carrier supports VLAN-trunk interfaces. That way I could have two virtual connections at each branch and let EIGRP manage what's up and down based on keepalives/neighbor relationships. This would also provide load balancing (which I don't really need). I'd rather not screw around with dot1q sub-interfaces or worry about tagging to the carrier, but if need be...
- Terminate the second connection into a different router (e.g., the aforementioned 4431) and adjust stuff like bandwidth and delay so EIGRP doesn't mark it as equal-cost.
Anybody had this scenario and found an ideal solution?