Good afternoon,
I have been toying with two different designs that will hit production shortly.
This is a single cloud DMVPN, that will be used to backup a number of point-to-point links. (private leased lines, a few different types, but irrelevant in this discussion)
I have simplified the diagram for the purposes of the discussion.
The core layer3 switches terminate the P2P links, and one EIGRP process is used for the routing. summary addresses (/16's) are advertised out these links. In the event of a link failure, the DMVPN is to be used.
For the DMVPN, There are two HUBS, one in the USA, and one in the UK. The USA has two spokes in different cities, and the UK has two spokes in different cities. Each spoke connects to BOTH hubs.
On the DMVPN network, when using the SAME EIGRP process as the P2P links, everything seems to work as expected, in phase 1 and phase 2 and Phase 3. I have tested all the configurations. Some of the delays had to be tweaked, and the bandwidth was set to 10Mbit on the tunnels.
When using a SEPERATE EIGRP process, and two way redistribution everything is also fine, in all the phases. The DMVPN is never used unless its the only link, less tweaking required as Admin distanced handles that.
My question that I would like design input on, is which method is better? When rolling out DMVPNs as backups - who is using a seperate process.
I'm not a huge fan of redistribution, when It can be avoided.
Thoughts?
