Hello Tarun,
the link provided is a good reference.
From a networking point of view iBGP sessions are TCP flows that are originated and terminated at loopback-M's ip addresses.
The requirement is that the IGP protocol OSPF is publishing the routing information for all these loopbacks' addresses in the OSPF domain.
The combined use of OSPF and BGP allows for a divide and conquer approach:
- the infrastructure ip subnets including physical links and loopbacks used as BGP and OSPF router-ids are routed by the IGP OSPF
- customer routes can be learned by eBGP and depending if using MPLS or not can be left in the BGP table or they should be redistributed into OSPF (if not using MPLS)
In order to get a successful redistribution is good practice to have OSPF router-id = BGP router-id.
iBGP requires some specific commands for scalability: the usage of BGP Route Reflector Servers and/or BGP confederations.(less used).
A simple example
Let's suppose that 10.76.x.y is the loopback address range you are using for loopbacks.
Let's suppose that 10.128.0.0/16 is used for infrastructure physical links
int loop 5
desc OSPF and BGP router-id
ip address 10.76.1.12 255.255.255.255
!
router ospf 5
router-id 10.76.1.12
network 10.76.1.12 0.0.0.0 area 0
network 10.128.0.0 0.0.255.255 area 0
!
router bgp 65012
router-id 10.76.1.12
neighbor 10.76.2.20 remote-as 65012
neighbor 10.76.2.20 update-source loop5
neighbor 10.76.2.20 send-community
neighbor 10.76.2.21 remote-as 65012
neighbor 10.76.2.21 update-source loop5
neighbor 10.76.2.21 send-community
no auto-summary
!
! where 10.76.2.20 and 172.76.2.21 are two RR servers
! what makes them RRS are the following lines:
router bgp 65012
router-id 10.76.2.20
neighbor 10.76.1.12 remote-as 65012
neighbor 10.76.1.12 update-source loop5
>>neighbor 10.76.1.12 route-reflector-client
This just for start, if not using MPLS you need to redistribute customer eBGP routes into OSPF to be able to route to them.
If you use MPLS and you don't redistribute routes into OSPF, or all of your internal routers run both OSPF and BGP, you need to disable the syncronization in the BGP process with
router bgp 65012
no sync
For edge routers that have eBGP sessions with customers you may need to overwrite the BGP next-hop towards the core to be sure that the customer prefixes will pass the BGP next-hop check.
router bgp 65012
neighbor 172.16.2.20 next-hop-self
neighbor 172.16.2.21 next-hop-self
hope to help
Giuseppe
