Hi,
I did some labs to understand what should we use if we have two WAN and our goal is to have one primary and another secondary in case if primary Service Provider failed.
Lab scenario:
1. Campus site with Collapsed Core design.
a. Core Switch connected to remote sites via MPLS and has eBGP peering PE.
b. Core Switch also runs EIGRP for its SVIs. Design is collapsed core, so there is no any L3 network device (like Distribution Switch) downstream of the Core Switch, so Core Switch doesn't need to send any EIGRP routes downstream to Distribution Switch and etc.
c. Router connected to remote sites via DMVPN and has eBGP peering with Spoke Routers.
2. The same routes are being received from both MPLS and DMVPN and your goal is to make MPLS primary egress point and DMVPN secondary - this if for egress traffic. The same for ingress traffic - remote sites should prefer to use MPLS over DMVPN when they want to reach this site.
3. I'll write here my several solutions, how I understand it:
For egress traffic:
a. Configure only EIGRP between Core Switch and Router. Do mutual redistribution (BGP -> EIGRP and EIGRP -> BGP) on both Core Switch and Router. Manipulate with EIGRP metric in that way, so Core Switch (MPLS) is preferred over Router (MPLS). In that solution there is no iBGP between Core Switch and Router. On both Core Switch and Router's routing table there will be routes from remote sites as eBGP routes.
b. Configure EIGRP (to advertise loopbacks) and iBGP between Core Switch and Router. Manipulate with local preference on Core Switch, make it higher to make the routes from MPLS more preferred. Redistribute EIGRP to BGP on Core Switch BGP to EIGRP on Router. In that case, on Core Switch's routing table there will be eBGP routes from remote site and on Router's routing table there will be external EIGRP routes from remote sites with next hop pointing to Core Switch.
c. Configure only iBGP between Core Switch and Router. Manipulate with local preference on Core Switch, make it higher to make the routes from MPLS more preferred. Redistribute EIGRP to BGP on Core Switch. Configure static routes on both Core Switch and Router towards each others Loopbacks to establish iBGP. In this case, on Core Switch's routing table there will eBGP routes from remote site and on Router's routing table there will be internal BGP routes from remote sites with next hop pointing to Core Switch. Probably, this is less preferable solution due to static routing needed for loopbacks.
For all three scenarios, to manipulate ingress traffic as-prepend will be used. So, ingress traffic manipulation is pretty easy.
So, what do you think? Are those solutions I provided correct? If they're all correct, which one from your opinion is better? Maybe you have better solution, could you please describe it?
