The dangers of route summarization
By stretch | Wednesday, April 9, 2008 at 5:27 p.m. UTC
I recently encountered a problem where misconfigured route summarization resulted in suboptimal routing. Consider the following scenario.
A small LAN of two subnets has two equal-cost paths to its parent WAN through routers A and B, both of which run EIGRP in AS 123. By default, EIGRP will load balance these paths and all will be good. However, each subnet will be advertised separately. This isn't a big deal for just the two we have, but what about ten? Or a hundred?
Enter route summarization. A proactive network engineer may attempt to manually summarize contiguous subnets for maximum efficiency. This limits the overall number of routes in the EIGRP (or other applicable IGP) topology, and contributes to the stability of the network. The two LAN subnets in our scenario can be easily summarized on the WAN-facing interfaces:
Router(config-if)# ip summary-address eigrp 123 192.168.0.0 255.255.254.0
The two /24 routes are now advertised to the WAN as a single /23 route. Here's how the WAN saw our LAN subnets before the summarization:
D 192.168.0.0/24 [90/30720] via 172.16.0.6, 00:01:08, FastEthernet1/1 [90/30720] via 172.16.0.2, 00:01:08, FastEthernet1/0 D 192.168.1.0/24 [90/30720] via 172.16.0.6, 00:00:13, FastEthernet1/1 [90/30720] via 172.16.0.2, 00:00:13, FastEthernet1/0
D 192.168.0.0/23 [90/30720] via 172.16.0.6, 00:00:01, FastEthernet1/1 [90/30720] via 172.16.0.2, 00:00:01, FastEthernet1/0
Nice and efficient. But now assume that router A fails for some reason and its replacement isn't configured with the summary statement. Here are the routes the WAN sees with only router B summarizing:
D 192.168.0.0/24 [90/30720] via 172.16.0.2, 00:00:04, FastEthernet1/0 D 192.168.1.0/24 [90/30720] via 172.16.0.2, 00:00:04, FastEthernet1/0 D 192.168.0.0/23 [90/30720] via 172.16.0.6, 00:00:04, FastEthernet1/1
As you can see, the new router A is advertising two more-specific routes than router B. As a result, the WAN will always prefer the path through router A over B, and load balancing no longer takes place. When implementing route summarization, always make sure to implement it equally across border routers.
About the Author
Jeremy Stretch is a network engineer living in the Raleigh-Durham, North Carolina area. He is known for his blog and cheat sheets here at Packet Life. You can reach him by email or follow him on Twitter.
Posted in Routing
April 12, 2011 at 9:13 a.m. UTC
Thanks for the post
September 2, 2011 at 7:30 a.m. UTC
Thanks for the post.. I'm able to understand the concepts of sub optimal routing better now...
I tried simulating this in my dynamips but could not get see the results.I did not understand when you mean " But now assume that router A fails for some reason and its replacement isn't configured with the summary statement. Here are the routes the WAN sees with only router B summarizing:
D 192.168.0.0/24 [90/30720] via 172.16.0.2, 00:00:04, FastEthernet1/0 D 192.168.1.0/24 [90/30720] via 172.16.0.2, 00:00:04, FastEthernet1/0 D 192.168.0.0/23 [90/30720] via 172.16.0.6, 00:00:04, FastEthernet1/1 "
When R1 fails, then its neighborship to Router B and Wan Router fails. So RouterA will never be able to advertise any routes. how come your example has routes to 192.168.0.0/24 and 192.168.1.0/24 via 18.104.22.168 if router A fails?
Correct me if i misunderstood..
Thanks and Eagerly awaiting for reply for better clarity..
I'm sorry if the question sound silly.
September 3, 2011 at 10:13 a.m. UTC
Here router A has been replaced with a new router and in the new router, administrator has forgotton to summarize the network.
September 3, 2011 at 8:37 p.m. UTC
Thanks Anshu for your reply.
When router A is replaced with a new router, then the source of summary route for the WAN router is only from Router B. so If I'm not wrong, the summary address will have the next hop as Router B.
Correct me if I misunderstood..
June 9, 2013 at 2:15 p.m. UTC
Yes,source will be router B.172.16.0.6
June 25, 2014 at 11:52 a.m. UTC
Actually A is going to be preferred over B because it has the most specific prefixes