Wednesday 12 June 2019

OSPF Series 11: Everything on Redistribution: Part 2- Redistribution into OSPF, Pitfall of Redistribution (Administrative Distance, Route Feedback)

Everything on Redistribution: Part 2- Redistribution into OSPF, Pitfall of Redistribution (Administrative Distance, Route Feedback)

 

Redistributing into OSPF

OSPF is a standardized Link-State routing protocol that uses cost (based on bandwidth) as its link-state metric. An OSPF router performing redistribution automatically becomes an ASBR.

To redistribute all EIGRP routes into OSPF:

RouterB(config)# router ospf 20

RouterB(config-router)# network 172.16.0.0 0.0.255.255 area 0

 

RouterB(config-router)# redistribute eigrp 15

RouterB(config-router)# default-metric 30

First, the router ospf process was enabled with a process-ID of 20. Next, OSPF was configured to place any interfaces in the network of 172.16.0.0/16 into area 0. Then, OSPF will redistribute all eigrp routes from AS 15. Finally, a default-metric of 30 was applied to all redistributed routes.

If the default-metric or a manual metric is not specified for the redistributed routes, a default metric of 20 will be applied to routes of all routing protocols except for BGP. Redistributed BGP routes will have a default metric of 1 applied by OSPF.

By default, OSPF will only redistribute classful routes into the OSPF domain. To configure OSPF to accept subnetted networks during redistribution, the subnets parameter must be used:

RouterB(config)# router ospf 20

RouterB(config-router)# redistribute eigrp 15 subnets

Routes redistributed into OSPF are marked external. OSPF identifies two types of external routes, Type-1 (which is preferred) and Type-2 (which is default). To change the type of redistributed routes:

RouterB(config)# router ospf 20

RouterB(config-router)# redistribute eigrp 15 subnets metric-type 1

Pitfalls of Route Redistribution – Administrative Distance

Route redistribution introduces unique problems when there are multiple points of redistribution. Consider the following diagram:

The first issue is caused by Administrative Distance (AD), which determines which routing protocol is “trusted” the most. By default, OSPF routes have an AD of 110, whereas RIP routes have an AD of 120. Lowest AD is preferred, thus making the OSPF routes the most trusted.

Assume mutual redistribution has been performed on RouterC and RouterD.

The following networks will be injected from RIP into OSPF: 10.1.1.0/24,

10.1.2.0/24, 10.1.3.0/24, 10.1.4.0/24, and 10.1.5.0/24.

RouterC will eventually receive OSPF routes to the above networks from RouterD, in addition to the RIP routes already in its table. Likewise, RouterD will receive OSPF routes to these networks from RouterC.

Because OSPF’s AD is lower than RIP’s, both RouterC and RouterD will prefer the sub-optimal path through OSPF to reach the non-connected networks. Thus, RouterC will choose the OSPF route for all the 10.x.x.x/24 networks except for 10.1.1.0/24, as it is already directly connected.

This actually creates a routing loop. RouterC will prefer the OSPF path through RouterA to reach the 10.x.x.x networks (except for 10.1.1.0/24), and RouterA will likely consider RouterC its shortest path to reach those same networks. Traffic will be continuously looped between these two routers.

Even if RouterC managed to send the traffic through RouterA and RouterB to RouterD, the preferred path to the 10.x.x.x networks for RouterD is still through OSPF. Thus, the routing loop is inevitable.

 

 

There are two methods to correct this particular routing loop. The first method involves filtering incoming routes using a distribution-list, preventing RouterC and RouterD from accepting any routes that originated in RIP from their OSPF neighbors.

RouterC’s configuration would be as follows:

RouterC(config)# access-list 10 deny 10.1.2.0 0.0.0.255

RouterC(config)# access-list 10 deny 10.1.3.0 0.0.0.255

RouterC(config)# access-list 10 deny 10.1.4.0 0.0.0.255

RouterC(config)# access-list 10 deny 10.1.5.0 0.0.0.255

RouterC(config)# access-list 10 permit any

RouterC(config)# router ospf 20

RouterC(config-router)# distribute-list 10 in fastethernet0/0

An access-list was created that is denying the RIP networks in question, and permitting all other networks. Under the OSPF process, a distribute-list is created for routes coming inbound off of the fastethernet0/0 interface. The access-list and distribute-list numbers must match. RouterD’s configuration would be similar.

This prevents each router from building OSPF routes for the networks that originated in RIP, and thus eliminates the possibility of a loop. However, redundancy is also destroyed – if RouterC’s fa0/1 interface were to fail, it could not choose the alternate path through OSPF.

 

 

The second method involves using the distance command to adjust the AD of specific routes. This can accomplished two ways:

• —> Lowering the AD of the local RIP-learned routes

• —> Raising the AD of the external OSPF-learned routes

To force the RIP routes to be preferred, RouterC’s configuration would be as follows:

RouterC(config)# access-list 10 permit 10.1.2.0 0.0.0.255

RouterC(config)# access-list 10 permit 10.1.3.0 0.0.0.255

RouterC(config)# access-list 10 permit 10.1.4.0 0.0.0.255

RouterC(config)# access-list 10 permit 10.1.5.0 0.0.0.255 RouterC(config)# access-list 10 deny any

RouterC(config)# router rip

RouterC(config-router)# distance 70 10.1.1.0 0.0.0.255 10

An access-list was created that is permitting the RIP networks in question, and denying all other networks. Under the RIP process, an administrative distance of 70 is applied to updates from routers on the 10.1.1.0 network, for the specific networks matching access-list 10. RouterD’s configuration would be similar.

Thus, the RIP-originated networks will now have a lower AD than the redistributed routes from OSPF. The loop has again been eliminated. Another solution would be to raise the AD of the external OSPF routes. OSPF provides a simple mechanism to accomplish this:

RouterC(config)# router ospf 20

RouterC(config-router)# distance ospf external 240

Pitfalls of Route Redistribution – Route Feedback

A routing loop is only one annoying issue resulting from the above design. Route feedback is another problem that must be addressed.

OSPF routes redistributed into RIP on RouterC will eventually reach RouterD, and then be redistributed again back into OSPF. This is a basic example of route feedback.

Depending on the metrics used, this could potentially cause RouterB to prefer the route through RouterD (and through the RIP domain), to reach the 192.168.2.0/24 network. This is an obvious example of suboptimal routing.

Thus, routes that originated in a routing domain should not to be re-injected into that domain. Distribution-lists and the distance command can be utilized to accomplish this, but route tags may provide a more robust solution.

Tagging routes provides a mechanism to both identify and filter those routes further along in the routing domain. A route retains its tag as it passes from router to router. Thus, if a route is tagged when redistributed into RIP on RouterC, that same route can be selectively filtered once it is advertised to RouterD.

Route tags are applied using route-maps. Route-maps provide a sequential list of commands, each having a permit or deny result:

RouterC(config)# route-map OSPF2RIP deny 5

RouterC(config-route-map)# match tag 33

RouterC(config-route-map)# route-map OSPF2RIP permit 15

RouterC(config-route-map)# set tag 44

Route-maps are covered in great detail in a separate blog post.

The full configuration necessary on RouterC would be as follows:

RouterC(config)# route-map OSPF2RIP deny 5

RouterC(config-route-map)# match tag 33

RouterC(config-route-map)# route-map OSPF2RIP permit 15

RouterC(config-route-map)# set tag 44

RouterC(config)# router rip

RouterC(config)# redistribute ospf 20 route-map OSPF2RIP

RouterC(config)# route-map RIP2OSPF deny 5

RouterC(config-route-map)# match tag 44

RouterC(config-route-map)# route-map RIP2OSPF permit 15

RouterC(config-route-map)# set tag 33

RouterC(config)# router ospf 20

RouterC(config)# redistribute rip route-map RIP2OSPF

Thus, OSPF routes being redistributed into RIP are set with a tag of 44. When RIP is redistributed back into OSPF, any route with a tag that matches 44 is denied.

Similarly, RIP routes being redistributed into OSPF are set with a tag of 33. When OSPF is redistributed back into RIP, any route with a tag that matches 33 is denied.

The net result: routes originating from a routing domain will not redistributed back into that domain.

No comments:

Post a Comment