Regarding EBGP multihop

[ncnaveen_arasu]

Hi,

In BGP multihop concept how do we define the exact multihop count in the sample scenario.

R_A--------R_B---------R_C------------R_D

Here i want to establish a neighbporship between R_A loopback to R_D loopback.

So what should be the exact multihop count.

And please let me know wether EBGP and IBGP neighbor should be directly connected ot not.

Thanks in advance

Naveen

[Harold Ritter]

Hi Naveen,

You would need an "ebgp-multihop" value of 3 to get from A to D.

And to answer you second question, both iBGP and eBGP can be directly or non directly connected but generally, iBGP sessions are to non directly connected peers, using an IGP to get from one to the other, and eBGP session are directly connected.

Regards

[ncnaveen_arasu]

Hi Ritter,

Thanks for the reply.

Still my doubt is not clear. when R_A want to reach R_D loopback, it has to reach R_B, R_C, R_D and Loopback of R_D right. then it will be 4 if iam not wrong, how it's 3 please clarify how to calculate this multihop.

for 2nd question if we run IBGP in all our routers and if they are directly  still IGP is required if yes why.

Please clarify my doubt.

[Harold Ritter]

Hi Naveen,

It is actually 3 hops. 

B receives the packets with TTL of 3, decrements it and forwards to C.

C receives the packets with TTL of 2, decrements it and forwards to D.

D receives the packet with TTL of 1 and does not decrement it since it is destined to itself.

TTL is only decremented before the packet is forwarded to the next hop.

iBGP is not meant to be run as an IGP. It generally requires a full mesh between all routers in your network. You could make it work the way you are describing but it would be complicated and certainly not following best practices.

Regards

[ncnaveen_arasu]

Hi Herold

Thanks for providing useful information.

When the packet sent to D with ttl of 1 it will expire before it sends to loopback of router D right?

A sample of ebgp multihop is given in Wendel Odom Routing book as follows.

With two router A and B, When EBGP neighborship forms between loopback of A & B, First it sends a packet to D interface with TTL-2 and as the destination will be loopback it will decrement the TTL by 1 and sends to loopback of B. If the TTL was 1 when it was sending by A, packet would have not been delivered to Loopback of B as the default TTL1 will expire to 0 before reaching loopback.

If that is corrrect in my 1st question with 4 routers it has to send with a TTL value of 4.

Please correct me if iam wrong.

Regards,

Naveen

[Harold Ritter]

Hi Naveen,

> When the packet sent to D with ttl of 1 it will expire before it sends to loopback of router D right?

> A sample of ebgp multihop is given in Wendel Odom Routing book as follows.

TTL is not decrement if the packet is destined to any of the locally assigned IP addresses. It is only decremented if the packet needs to be forwarded to another device.

> With two router A and B, When EBGP neighborship forms between  loopback of A & B, First it sends a packet to D 

> interface with TTL-2  and as the destination will be loopback it will decrement the TTL by 1  and sends to loopback of > B. If the TTL was 1 when it was sending by A,  packet would have not been delivered to Loopback of B as the default  > TTL1 will expire to 0 before reaching loopback.

This is incorrect. This misunderstanding might come from the fact that, in cisco IOS, ebgp-multihop 1 is the default and assumes a session to a directly connected peer IP address. So if you want to establish a BGP session to a directly connected neighbor loopback address, you need to use "ebgp-multihop 2". There is a command that will allow you to establish that session without using the default TTL of 1. This command is "neighbor disable-connected-check".

> If that is corrrect in my 1st question with 4 routers it has to send with a TTL value of 4.

ebgp-multihop 3 is sufficient for A to reach D in your scenario.

Regards