Multiple Autonomous Systems using EIGRP Named Mode

Question

I have an EIGRP network that has multiple AS numbers. Lets says they are AS 200 and 201. AS 200 exists only in the default VRF and AS 201 exists in both the default VRF and well as a dedicated server VRF. I need to configure both AS numbers to run concurrently on our core 6500 switch.
With the older way to configure EIGRP I would just create 2 EIGRP intstances as follows:
router eigrp 200
router eigrp 201
With named mode, would I create 1 or 2 named instances for EIGRP? And would every address family need a separate router-id or can both AS nubmers in the default VRF share a router ID?
example: 1 name mode instance
router eigrp named1
 address-family ipv4 unicast autonomous-system 200
  af-interface default
   passive-interface
  exit-af-interface
  network 10.10.0.0 0.0.255.255
  eigrp router-id 10.1.1.1
 address-family ipv4 unicast autonomous-system 201
  af-interface default
   passive-interface
  exit-af-interface
  network 10.20.0.0 0.0.255.255
  eigrp router-id 10.1.1.1
address-family ipv4 unicast vrf server autonomous-system 201
  af-interface default
   passive-interface
  exit-af-interface
  network 10.30.0.0 0.0.255.255
  eigrp router-id 10.3.1.1
Example 2: 2 named mode instances
router eigrp named1
 address-family ipv4 unicast autonomous-system 200
  af-interface default
   passive-interface
  exit-af-interface
  network 10.10.0.0 0.0.255.255
  eigrp router-id 10.1.1.1
router eigrp named2
 address-family ipv4 unicast autonomous-system 201
  af-interface default
   passive-interface
  exit-af-interface
  network 10.20.0.0 0.0.255.255
  eigrp router-id 10.1.1.1
address-family ipv4 unicast vrf server autonomous-system 201
  af-interface default
   passive-interface
  exit-af-interface
  network 10.30.0.0 0.0.255.255
  eigrp router-id 10.3.1.1
Any Thoughts? Any help would be appreciated.
Ben

Answer 1

Hi Ben,
Personally, I do not see a significant difference between the two options you have, i.e. having multiple per-VRF EIGRP processes under a single router eigrp instance-name, as opposed to having a separate instance for each VRF. Recall that even in classic numbered configuration mode, you can have multiple per-VRF processes configured under a single numbered EIGRP instance so there is really no specific difference here.
That being said, I think that you would like the second option better, that is, having a separate EIGRP name mode section for each VRF. In fact, I have found it confusing in the numbered mode to have several VRFs grouped under a single numbered instance whose autonomous system number did not even relate in any way to the autonomous system number in the per-VRF processes.
Regarding the uniqueness of Router IDs - that's a good question. In EIGRP, the Router ID is used to prevent a router from processing information originated by itself, possibly causing a routing loop. I have always found this explanation somewhat strange, as EIGRP has different mechanisms for loop prevention, and I could never come up with an example where this mechanism would actually be useful. In any case, in newer EIGRP implementations, the RID is attached to each internal and external route as it is advertised throughout the EIGRP domain. This has two important consequences:
By looking at the show ip eigrp topology X.X.X.X/M.M.M.M output, you can always find out which router originates that particular network
If a router receives an update about a network marked with its own RID, it will ignore the update.
If your network and your VRFs are intended to remain perfectly isolated at all times, i.e. no routes from a VRF will ever be advertised across the network so that they leak from one VRF to another, or between a global routing table and a VRF, then a single router can use the same RID in all its EIGRP processes, both in global table and in a VRF. However, if there is an intended possibility of a route existing in one VRF to be advertised over a series of routers and being intentionally received by the same router in a different VRF then using the same RID in multiple EIGRP processes on that router would prevent it from accepting the update. In such case, you would need to use unique RID per each VRF process.
I hope this helps - please feel welcome to ask further!
Best regards,
Peter