07-16-2013 10:48 PM - edited 03-04-2019 08:28 PM
I was reading up on BGP route reflection and I read that a Route Reflector (RR) will advertise all received routes to all clients and non-clients. A RR will also advertise (or reflect) routes received from clients as well as routes received from non-clients. So I'm wondering what's the difference between a client and non-client? If a RR receives and reflects routes to/from everyone, what's the point of defining some of the peers as clients and leaving others as regular iBGP peers.
The example toplogy in the book I was reading had several iBGP peers configured in a RR cluster and one peer configured as a regular iBGP peer. The regular iBGP peer was peered with the RR. As far as I could tell, the regular iBGP peer was receiving all the same routes as the RR clients and advertising its routes to the RR, which the RR reflected to its clients. So why even bother with the client/non-client distinction?
Thanks.
Solved! Go to Solution.
07-16-2013 11:45 PM
Hi Venison,
If a RR receives and reflects routes to/from everyone, what's the point of defining some of the peers as clients and leaving others as regular iBGP peers.
Imagine that an RR has two clients, C1, C2, and two non-clients, N1 and N2. If an update comes from C1, RR will reflect this to C2, N1 and N2. The same would happen if C2 sent an update - it would be reflected to C1, N1 and N2.
However, if an update comes from N1, this will be reflected to C1 and C2 but not to N2. Similarly, if an update comes from N2, it will be reflected only to C1 and C2 but not to N1. This requires N1 and N2 to have their own peerings apart from the peering with the RR.
Would this help?
Best regards,
Peter
07-16-2013 11:45 PM
Hi Venison,
If a RR receives and reflects routes to/from everyone, what's the point of defining some of the peers as clients and leaving others as regular iBGP peers.
Imagine that an RR has two clients, C1, C2, and two non-clients, N1 and N2. If an update comes from C1, RR will reflect this to C2, N1 and N2. The same would happen if C2 sent an update - it would be reflected to C1, N1 and N2.
However, if an update comes from N1, this will be reflected to C1 and C2 but not to N2. Similarly, if an update comes from N2, it will be reflected only to C1 and C2 but not to N1. This requires N1 and N2 to have their own peerings apart from the peering with the RR.
Would this help?
Best regards,
Peter
07-29-2013 09:51 PM
Yeah, that makes sense, good explanation. Thanks for the reply.
07-29-2013 10:34 PM
So one more thought on that - I guess the RR client relationship effectively allows you to override the BGP split-horizon rule?
07-30-2013 12:50 AM
Hello Venison,
Can you explain more precisely what do you mean by "the BGP split-horizon rule"? Although this term occurs every now and then, the BGP-4 RFC 4271 never uses that term, and so its meaning is ambiguous to me.
Best regards,
Peter
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide