Rate and give feedback:  Feedback Received. Thank You!

Rate and give feedback: 

X

This document helped resolve my issue

Yes No

Additional Comments

800 characters remaining

May we contact you if necessary?

Name:  

E-mail: 

Enter a valid Email ID

Need product assistance? Contact Juniper Support

Submit

This site is protected by hCaptcha and its Privacy Policy and Terms of Service apply.

English  

English

Understanding Internal BGP Peering Sessions

When two BGP-enabled devices are in the same autonomous system (AS), the BGP session is called an internal BGP session, or IBGP session. BGP uses the same message types on IBGP and external BGP (EBGP) sessions, but the rules for when to send each message and how to interpret each message differ slightly. For this reason, some people refer to IBGP and EBGP as two separate protocols.

Figure 1: Internal and External BGP

In Figure 1, Device Jackson, Device Memphis, and Device Biloxi have IBGP peer sessions with each other. Likewise, Device Miami and Device Atlanta have IBGP peer sessions between each other.

The purpose of IBGP is to provide a means by which EBGP route advertisements can be forwarded throughout the network. In theory, to accomplish this task you could redistribute all of your EBGP routes into an interior gateway protocol (IGP), such as OSPF or IS-IS. This, however, is not recommended in a production environment because of the large number of EBGP routes in the Internet and because of the way that IGPs operate. In short, with that many routes the IGP churns or crashes.

Generally, the loopback interface (lo0) is used to establish connections between IBGP peers. The loopback interface is always up as long as the device is operating. If there is a route to the loopback address, the IBGP peering session stays up. If a physical interface address is used instead and that interface goes up and down, the IBGP peering session also goes up and down. Thus the loopback interface provides fault tolerance in case the physical interface or the link goes down, if the device has link redundancy.

While IBGP neighbors do not need to be directly connected, they do need to be fully meshed. In this case, fully meshed means that each device is logically connected to every other device through neighbor peer relationships. The neighbor statement creates the mesh. Because of the full mesh requirement of IBGP, you must configure individual peering sessions between all IBGP devices in the AS. The full mesh need not be physical links. Rather, the configuration on each routing device must create a full mesh of peer sessions (using multiple neighbor statements).

Note: The requirement for a full mesh is waived if you configure a confederation or route reflection.

To understand the full-mesh requirement, consider that an IBGP-learned route cannot be readvertised to another IBGP peer. The reason for preventing the readvertisement of IBGP routes and requiring the full mesh is to avoid routing loops within an AS. The AS path attribute is the means by which BGP routing devices avoid loops. The path information is examined for the local AS number only when the route is received from an EBGP peer. Because the attribute is only modified across AS boundaries, this system works well. However, the fact that the attribute is only modified across AS boundaries presents an issue inside the AS. For example, suppose that routing devices A, B, and C are all in the same AS. Device A receives a route from an EBGP peer and sends the route to Device B, which installs it as the active route. The route is then sent to Device C, which installs it locally and sends it back to Device A. If Device A installs the route, a loop is formed within the AS. The routing devices are not able to detect the loop because the AS path attribute is not modified during these advertisements. Therefore, the BGP protocol designers decided that the only assurance of never forming a routing loop was to prevent an IBGP peer from advertising an IBGP-learned route within the AS. For route reachability, the IBGP peers are fully meshed.

IBGP supports multihop connections, so IBGP neighbors can be located anywhere within the AS and often do not share a link. A recursive route lookup resolves the loopback peering address to an IP forwarding next hop. The lookup service is provided by static routes or an IGP, such as OSPF.