Options

Confusion regarding iBGP advertisements...

CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
in CCNP
I was working though a setup in GNS3 that consisted of three routers participating in iBGP together. Routers A, B, and C exist in the topology and both routers B and C only have direct connections to A (Like a hub spoke sorta setup). If I advertise a route from router B, router A learns it but router C does not. Whats more is when I do a "show ip bgp nei [addr] advertised-routes" I get 0 prefixes being advertised to router C. I read that iBGP routers will not take routes they learned from an iBGP peer and then advertise them to another iBGP peer which would clear up some of the confusion. If this is the case though (which I think it is), how is it done on internet routers? Are you telling me that every internet router in the world has a different BGP AS configured on it?
Currently reading: Network Warrior, Unix Network Programming by Richard Stevens
· Share on FacebookShare on Twitter

Comments

  • Options
    CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
    I suppose this can be deleted... I got my answer from a really bright colleague (CCIE material kinda guy) who has done work for us.
    Currently reading: Network Warrior, Unix Network Programming by Richard Stevens
    · Share on FacebookShare on Twitter
  • Options
    NetworkVeteranNetworkVeteran Member Posts: 2,338 ■■■■■■■■□□
    If this is the case though (which I think it is), how is it done on internet routers?

    It's a common question. You'd make Routers B and C neighbors, or configure Router A as a route reflector.
    Are you telling me that every internet router in the world has a different BGP AS configured on it?
    Hehe! Nope. :)
    · Share on FacebookShare on Twitter
  • Options
    CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
    Thanks for the response, that's what a buddy of mine mentioned to me as well (not long after I created this thread). He mentioned that my options would be a route reflector or confederation. So it's safe to assume that iBGP is gonna be full mesh if there is no "route reflector" setup huh?
    Currently reading: Network Warrior, Unix Network Programming by Richard Stevens
    · Share on FacebookShare on Twitter
  • Options
    NetworkVeteranNetworkVeteran Member Posts: 2,338 ■■■■■■■■□□
    CodeBlox wrote: »
    So it's safe to assume that iBGP is gonna be full mesh if there is no "route reflector" setup huh?

    Yes. Also note that, unlike IGPs, you don't always need to run BGP on every device in your domain. I.e., in many cases you can get away with turning BGP off on RouterA, so only RouterB and RouterC need to be neighbors.
    · Share on FacebookShare on Twitter
  • Options
    powmiapowmia Users Awaiting Email Confirmation Posts: 322
    NetworkVeteran wrote: »
    Yes. Also note that, unlike IGPs, you don't always need to run BGP on every device in your domain. I.e., in many cases you can get away with turning BGP off on RouterA, so only RouterB and RouterC need to be neighbors.

    Unless your physical topology traverses RouterA
    · Share on FacebookShare on Twitter
  • Options
    NetworkVeteranNetworkVeteran Member Posts: 2,338 ■■■■■■■■□□
    powmia wrote: »
    Unless your physical topology traverses RouterA
    I was thinking along the lines of running MPLS in the core. In which case, BGP is not needed on RouterA, despite traffic transiting that node. Of course, assuming RouterA doesn't need to peer externally. :)
    · Share on FacebookShare on Twitter
  • Options
    powmiapowmia Users Awaiting Email Confirmation Posts: 322
    A whole nother topic :) I do the same thing. "Why the heck would you be running BGP on that P router?!?!!.... oh, you're not using MPLS."
    · Share on FacebookShare on Twitter
Sign In or Register to comment.