Route Redistribution

When a router compares two identical prefixes, first it looks at the administrative distance. If it is the same (the same source protocol), then it compares the metric (distance/cost/hop-count/etc) to determine which of these two routes should be installed in the routing table. If the administrative distances are different, it chooses the route with smaller administrative distance. Here is the default administrative distances for various routing protocols:

Administrative distance (AD) has local significance meaning that it is only known on that particular router and is not propagated to other routers. Its value can be changed using the “distance” command:

• RIP distance <value>
• EIGRP distance eigrp <internal> <external> distance eigrp <internal> <src ip address> <wildcard> [access-list-number]
• OSPF distance ospf <external> <intra> <inter> distance ospf <value> <src router-id> <wildcard> [access-list-number]

OSPF does not allow to overwrite its rule for intra-inter-external routes: O is preferred over O IA which is preferred over O E1/E2. So the second OSPF distance command without the access-list-number only works for host routes.

AD can also be specified per route rather than per class of routes, except for EIGRP external routes. OSPF or EIGRP distance command accepts the access-list-number to match a particular route.

Preventing Routing Loops

The routing loops can be prevented by:

• using a metric in a “redistribute” command set to higher value than any metric in the routing domain (works well for RIP, for example) or
• by changing the administrative distance (OSPF is a good example since it has the same default AD for its routes) or
• by tagging routes and later filtering on that tag number if for some reason those routes get back to the routing domain via another ASBR, for example.

Tagging and Filtering

A tag can be assigned via route-map “set tag” command by distribute-list or redistribute commands. The tag value should survive further redistribution but not all routing protocols do it by default; for example, OSFP has to be forced to re-apply the tag. Router’s internal routes (advertised via a “network” command) do not have a field for a tag though, only redistributed routes do.

Path Manipulation

Path manipulation is useful when the network topology has two or more ASBRs for symmetric routing, load balancing, preferring one path over the other, etc. It can be done by one of the following methods:

• changing metric in redistribution via route-map set metric, redistribute or default-metric for some (IGPs)
• changing metric type if supported by a routing protocol; for example, OSPF prefers O E1 over O E2.
• route filtering, which lacks the redundancy
• route summarization, where the most specific route will match for packet forwarding; it preserves the redundancy