Dynamic inter-VRF routing between the global VRF domain and a VRF instance
For both these examples all BGP neighbor relationships involve peering between IP local addresses, not to VLAN interface IP addresses within the same subnet.
BGP configuration tips
The following BGP configuration tips are included to explain the use of some BGP specific commands used in the i-BGP and e-BGP example configuration files below.
neighbor x.x.x.x update-source lo
The command neighbor x.x.x.x update-source lo is used to ensure the lo is used as the update source when establishing the BGP neighbor relationship instead of the egress VLAN interface IP.
neighbor x.x.x.x ebgp-multihop 2
The command neighbor x.x.x.x ebgp-multihop 2 is not applicable for an i-BGP connection, but is required for e-BGP when peering to an IP address in a remote network. For example, when forming an e-BGP neighbor relationship to the IP local address configured in a remote peer, the command is required.
This command above, is automatically generated when using e-BGP peering in conjunction with the neighbor x.x.x.x update-source command. The command defaults to hop count of 2 when automatically generated, but it can be explicitly configured to allow e-BGP peering to devices up to 255 hops away.
I-BGP doesn't default to peers being in the same subnet, as it supports multi-hop automatically. This is because the default configuration of i-BGP is a full mesh of all the routers in the AS and there’s no expectation that all i-BGP peers within the mesh will be in the same subnet. So, unlike e-BGP it can be quite common for an i-BGP TCP connection to be formed to IP address in a remote network, instead of peering to an IP address in same subnet.
In the case of e-BGP, it is uncommon to peer to a local loopback address, and similarly, the connection is not typically via a multi-hop L3 routed path - and the concept of a full mesh between all peers doesn’t apply. Hence e-BGP defaults to not allowing peering beyond a single hop.
neighbor x.x.x.x next-hop-self
I-BGP does not change the next hop address contained in BGP routes. To get i-BGP to change the nexthop IP, you need to use the neighbor x.x.x.x next-hop-self command.
neighbor x.x.x.x route-map <xx> out
The command neighbor x.x.x.x route-map <xx> out is used to reference and apply a route map. The route map in turn references an access-list. The out parameter in the command neighbor x.x.x.x route-map <xx> out specifies that the access list (used to filter routes), applies to outgoing advertisements.
