EBGP Multihop Misconception

MVP Expert
MVP Expert
Q:

 

The neighbor relationship for the BGP routing  protocol is such that,

1) For the iBGP neighborship, the peer could be upto 255 hops away (considering we have routing to reach that peer)

2) For the eBGP neighborship, the peer has to be directly connected by default. If the neighbor is more than 1 hop away, then the TTL value has to be changed from the default.

 

As a practice, the neighborship is generally formed with the loopback interface for obvious reasons.

 

For the following network diagram, if the neighborship is built between the 2, 8400 series switches using the loopback interface, the BGP neighborship will not be established as we require to modify the TTL value.

 

IS THE LOOPBACK INTERFACE ON THE NEIGHBORING DEVICE ACTUALLY MORE THAN ONE HOP AWAY ?

 



A:

 

The reason we need to change the TTL value from the default (TTL = 1) is because there is some check which BGP protocol performs.

The check is basically a requirement to see if the peer IP with which it tries to form the relationship with is part of its own network, basically a "Connected Check" behavior. The loopback is still one hop away, but since the local BGP process check that the loopback IP is not part of its own subnet, it doesn't even initiate the TCP 3-way handshake on that link.

 

In some other vendors ( not in HPE ), there is a command to disable the connected check feature and make the neighborship without changing the default.

As a reference, in Cisco routers, this can be disabled with the following command :

 

Version history
Revision #:
1 of 1
Last update:
‎02-26-2019 09:32 AM
Updated by: