Route redistribution: seed metrics, route tagging and avoiding loops between protocols

When and why you redistribute

Ideally a network runs one interior routing protocol. In practice you will meet several: after a company merger, during a move from EIGRP to OSPF, where a supplier uses a different protocol, or where static and connected routes must be advertised. Redistribution takes routes from one source (another protocol, static or connected routes) and injects them into a routing protocol. It only redistributes routes that are actually in the routing table from that source.

Seed metrics and configuration

Each protocol measures paths differently: RIP counts hops, OSPF uses cost and EIGRP uses bandwidth and delay. There is no way to convert one into another, so the receiving protocol needs a starting value, the seed metric.

router eigrp 100
 redistribute ospf 1 metric 100000 100 255 1 1500 route-map OSPF-TO-EIGRP
router ospf 1
 redistribute eigrp 100 subnets route-map EIGRP-TO-OSPF
router rip
 redistribute ospf 1 metric 3

On older IOS releases OSPF only redistributes classful networks unless you add subnets; newer releases include subnets automatically, but adding the keyword does no harm. You can also set a default for all redistributed routes with default-metric under the process. Check results with show ip route (EIGRP external routes appear as D EX, OSPF ones as O E2) and show ip ospf database external.

The danger of two-way redistribution at several points

With one redistribution router, one-way redistribution is simple. Real networks often redistribute in both directions on two routers for redundancy, which opens the door to routing feedback. A RIP route is redistributed into OSPF by router A, travels through OSPF to router B, and B redistributes it back into RIP with a good metric. Now RIP routers may prefer the path that loops through OSPF, or two protocols keep re-advertising a route after it has failed.

Administrative distance makes this worse. If a router learns the same network from RIP (AD 120) and OSPF (AD 110), it installs the OSPF route even when the RIP path is the real, shorter one. EIGRP gives external routes an AD of 170, which protects it a little, but OSPF uses 110 for both internal and external routes by default.

Route tagging and other fixes

The standard fix is to tag routes as they are redistributed and refuse to redistribute tagged routes back again:

route-map RIP-TO-OSPF deny 10
 match tag 110
route-map RIP-TO-OSPF permit 20
 set tag 120
route-map OSPF-TO-RIP deny 10
 match tag 120
route-map OSPF-TO-RIP permit 20
 set tag 110
router ospf 1
 redistribute rip subnets route-map RIP-TO-OSPF
router rip
 redistribute ospf 1 metric 3 route-map OSPF-TO-RIP

Apply the same route-maps on every redistribution router. Other tools are prefix lists that only allow the intended networks across, adjusting administrative distance for external routes (distance ospf external 170) so internal routes are preferred, and setting higher seed metrics so redistributed routes look worse than native ones. Plan redistribution on paper first, and use it as a temporary bridge where you can. See routing policy tools for route-map syntax.

Real-world example: two charities merged. One ran EIGRP, the other OSPF, and both head office routers were set to redistribute in both directions. Within a day, staff in one regional office reported that their file server became unreachable for a minute every few hours. Traceroutes showed traffic going in circles between the two head office routers whenever a WAN link flapped. The consultant added route tags with deny statements on both routers, raised the administrative distance of OSPF external routes, and drew up a six-month plan to move everything to OSPF. The loops stopped immediately.