BGP Question

Unanswered Question
Nov 29th, 2007

Using Route Reflectors allows an administrator to send iBGP updates without the need of a fully meshed topology. When the reflector router sends iBGP updates to reflector clients are individual TCP/IP session opened on 179 then tore down after the update, or how does that work?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 3 (1 ratings)
jcarrabine Wed, 11/21/2007 - 22:08

OK. I guess since I was seeing one RR there could only be one port 179 opened, but did not think of each individual interface connection being capable of having a TCP/179 connection. I guess I was a bit quick to hop on the boards. Probably the best way to see this on a large scale to to see RR and clustering in action...right?

Thanks for the help!!

Kevin Dorrell Thu, 11/29/2007 - 07:38

Best way is to lab it. Even with three routers it suddenly becomes clear when it is in front of you.

Don't worry about being a bit quick to hop on the boards. I do it all the time! (My last posting started with "Ooops!" for precisely that reason.) It keeps us on our toes.

A useful command here is show tcp, which should allow you to see all those sessions.

Good luck!

Kevin Dorrell


Kevin Dorrell Thu, 11/29/2007 - 06:30

No, the RR keeps a TCP/179 session open all the time to each of its clients. It's just that each time he receives an route update from any client, he propagates it to all the other clients.

Its a bit like disabling split-horizon in RIP.

Kevin Dorrell


jcarrabine Thu, 11/29/2007 - 06:40

So if a RR was acting as a hub connection to many spokes (reflector clients)does that mean there can be as many TCP/179 sessions as there are interfaces between RR and client?

Kevin Dorrell Thu, 11/29/2007 - 06:57

That is correct. From the hub's p.o.v. there are many sessions, each with my port TCP/179, his port TCP/179, but a different remote IP address. In fact, if you watch the console log, you can see when the neighbor relationships are established or taken down. Ideally, they come up once and then stay up for ever.

Kevin Dorrell


paul.matthews Fri, 11/30/2007 - 02:44

Kevin is bang on - I will just add one little bit of info.

Different RPs have different methods t make sure the routing info is up to date, for example protcols like RIP repeatedly send all the info out. OSPF ages the LSAs out, so individual LSAs need to be refreshed. To make sure the neighbour is still there OSPF hellos are used. Neither of these would be great if you had many thousands of routes, as per the internet and BGP, so BGP uses TCP at the transport mechanism for routing updates, and uses the TCP connection as the keepalive to make sure a neighbour is still there.



This Discussion