Different prefixes in BGP on Spines switches

[kwojtyra]

Hi,

I would like to start by describing our topology: we have a leaf-spine architecture with BGP configured between the two spine switches and the leaf switches. The leaf switches are in VLT, while the spine switches are not.

My question is: if we have the same BGP configuration on both spine switches and have followed the guide precisely, should the prefixes on the spines be different?

When we run the command show ip bgp summary, we can see that on the first spine switch, we have the same state/prefix for the neighbors, whereas, on the second spine switch, we observe a different state/prefix for each leaf and significantly fewer prefixes than on the first.

Is this the expected behavior?

[MHM Cisco World]

your run ibgp between spine-leaf?

it usually that the traffic from all leafs and one spine point to other spine, 

in spine check the prefix is it point to other spine or not ?

MHM

[kwojtyra]

Hi,

we use eBGP.

On the spine, we only point to each leaf.

[MHM Cisco World]

show ip bgp summary <<- is not so accurate 

use 

sh ip bgp neighbor x.x.x.x received-routes <<- check the prefix 

also share the config of BGP

MHM