IBGP issue

Unanswered Question
Apr 13th, 2010

Hi Expert,

I want to ask as I know, ibgp need to have fully mesh concept.
But according to my attached diagram, my ibgp has no fully mesh (R27 no dedicate link to R5, just R27 can reach R5 loopback address by OSPF) and (R28 no dedicate link to R6, just R28 can reach R6 loopback address by OSPF). but it also work after during UAT (disconnect / power down ) one by one


Do you have any comment ? Do I actually need to add physical red dot link in order to form the ibgp (R27 to R5, R28 to R6)

diagram information
1) All router use ospf routing, use route port interface to form neighbor
2) R27 already form ibgp neighbor with R28, R6, R5 (R27 no dedicate link to R5, just R27 can reach R5 loopback address by OSPF)
3) R28 already form ibgp neighbor with R27, R5, R6 (R28 no dedicate link to R6, just R28 can reach R6 loopback address by OSPF)
4) R5 already form ibgp neighbor with R6, R27, R28
5) R6 already form ibgp neighbor with R5, R27, R28

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
Giuseppe Larosa Tue, 04/13/2010 - 23:15

Hello AcBennyma,


>> my ibgp has no fully mesh (R27 no dedicate link to R5, just R27...


in iBGP context unless using BGP route reflector servers and/or BGP confederations you need to implement a full mesh of iBGP sessions not a full mesh of links.

in iBGP TTL of sessions is 255 and so as you have seen there is no need for a full mesh link topology.


What you see is normal. However, I would deploy two RRS one per site


configuration of RRS is quite easy:

on RRS only you use

neigh x.x.x.x route-reflector-client


you need also to peer the two RRS (normal BGP session without the above command)


of  course you don't need to add the dotted red link in the picture unless you want to add redundancy between the two sites, but it is not a requirement of iBGP.


Hope to help

Giuseppe

acbennyma Thu, 04/15/2010 - 09:13

hi Gluseppe,


Thanks for your explaination, but I still not understand. Is it means that.


1) In ibgp environment, ibgp neighbor need to connect via fully mesh session, not physically fully mesh link. So according to my diagram,
   because ospf routing is used so that all loopback address for all router are reachable. As a result, ibgp fully mesh session can be build up.
  
2) Otherwise, if OSPF routing is not exist, Base on physical link and only use bgp, then red dot link need to add to achieve fully mesh session. Right?


Thanks!

Jon Marshall Thu, 04/15/2010 - 09:20

In ibgp environment, ibgp neighbor need to connect via fully mesh session, not physically fully mesh link. So according to my diagram,
because ospf routing is used so that all loopback address for all router are reachable. As a result, ibgp fully mesh session can be build up.


Correct ie. you don't need to physically link all routers but you must have an IBGP session between all routers if you are not running router-reflectors. So the loopbacks must be reachable to all routers and that is what OSPF is doing in yout setup.


Jon

Giuseppe Larosa Thu, 04/15/2010 - 09:22

Hello AcBennyma,


a) yes


B) if no OSPF or other IGP is running you will need  a lot of static routes for other nodes loopbacks  not that red dotted link


B) is not a real world example and as I wrote you should consider the use of BGP route reflector server


Hope to help

Giuseppe

Actions

This Discussion