Because all the NAS servers permit only the default route via Enhanced IGRP, the only router that would query for 10.1.13.0/24 would be R1 when the primary connection goes down, and R2, and NAS1 in this case, when the PVC is restored between R1 and the remote router. During the initial failure, R1 loses the specific route to 10.1.13.0. It will query all its neighbors—in this case, NAS1 through NAS7, R2, R3, and R4 because none of them had the specific route to 10.1.13.0/24.
In such a case, R2, R3, and R4 are receiving summaries, and NAS servers are not installing another route other than the default via Enhanced IGRP. They will all send "immediate unreachable" back to R1 without propagating the queries.
After the remote router has dialed into NAS1, NAS1 will leak the specific route to 10.1.13.0/24 to all its neighbors—NAS2 through NAS7, and R1 and R2. Although all the NAS servers did receive the specific route, because distribute list permits only the default route, the NAS servers will not install any specific route. R2 has also received the specific route. After the original PVC between R1 and the remote site is restored, the connection between NAS1 and the remote site is brought down. Now R2 has lost the specific route to 10.1.13.0/24. R2 will query all its neighbors again. None of R2's neighbors, other than R1, had the specific route and R1 is sending the summary to R2. All of R2's neighbors will send unreachable back to R2 and not propagate the query further.
The entire network is redundant, except for the Ethernet that connects the hub routers to the regional core routers. An extra layer of protection could be added by connecting a hub router with a core router in a second region.
At this point, you may wonder: What if I simply connect all four of the routers on another LAN interface? Although this possibility may seem logical, it would simply be too expensive. This would be a good alternative if your service provider is willing to connect the two regional core connections to different ATM switches. This merely eliminates the single point of failure.
The final design with Enhanced IGRP acting as the IGP is shown in Figure 16-44.
Was this article helpful?