2

I wonder if there are any conceptual disadvantages of redistributing routes learned through eBGP peerings into an IGP, like OSPF.

I have read of course about performance issues and the risk of creating routing loops in the case of two-way redistribution between EGP and IGP, which make perfect sense to me. However those don't really apply to the network I am running (very small number of prefixes received on the eBGP peerings, in the range of 3000, and no IGP -> EGP redistribution).

In a context where performance is not really a concern, are there any more fundamental reasons not to redistribute those routes into OSPF and go for iBGP or another strategy ?

Or put another way, was performance the only motivation that led to specify iBGP when BGP was specified ?

1
  • The primary disadvantages are router resources and stability/convergence time when you receive routes from outside your control. What happens when the network advertising to you makes a mistake and sends the full Internet table to you? The other network instability can cause network instability in your own network. You probably do not want to trust a different AS to the point you risk your own network.
    – Ron Maupin
    Sep 28 at 12:23

2 Answers 2

6

To answer your specific question: no, there's no "conceptual" or fundamental reason not to redistribute. There are a few practical considerations, but you seem to already be aware of most of them.

If the question behind your question is Should I redistribute?, the answer is highly specific to the topology and network requirements. IGPs and BGP operate differently because they were designed for different purposes. The decision to redistribute routes or not comes down to the requirements of your network. There is no one right answer.

0

As already @Ron implied, general rules are for general/common scenarios. Specific topology may even benefit for infringing a best practice. IMHO the main reason comes down to performance issues, especially if (some of) the internal routers are not able to effectively handle the load because were not designed to do so in the first place. The overall design aim is also to maintain effectiveness, order and cleanliness in a messy world; common and easy to understand topologies are also easier to debug for anyone.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.