equal load balancing in multi tier senario :

Can you post your sanitized virtual server and pool configurations from your internal LTM? Is this just HTTP traffic or something else?

Its HTTPs traffic both (ssl offload , and ssl server side)

Vs is https , with default tcp prof and cookie persistence

Pool : LB mechanism is least connection

Do you have an HTTP profile applied to the virtual server?

How much client traffic is there (how many users)?

Do you have OneConnect enabled on the first tier LBs?

OK, very odd behaviour. A few more questions;

1. Do the clients use a proxy?
2. Any NTLM involved?
3. Does the Persistence profile have any 'Match Across' options configured?
4. Are the real servers in any other Pools?
5. Are the Pool Members actually different servers, or the same but listening on different ports?
6. I assume health monitors show all the members as up?

Thanks. So, this leaves perhaps two avenues of investigation;

1. A proxy is in use and is doing some sort of request multiplexing. A tcpdump on the external interface of the internal F5 might help you prove this. You should see initial requests from 'new' clients without the cookie being present, as it won't have been set yet. If you don't then there's a proxy doing something. Hope that makes sense.

    

2. Perhaps switch to the Round Robin load balancing method, this may at least prove something odd is going on with the connection counting when using Least Connections.

    

Also, I was wondering if you are using a session cookie or a particular time period before expiry?

The problem that all the traffic is going to one or two servers from the ten servers !!!!!!!!!

have you tried another load balancing algorithm such as round robin?

The problem that all the traffic is going to one or two servers from the ten servers !!!!!!!!!

have you tried another load balancing algorithm such as round robin?

Well, I'm pretty stuck but I still think doing a tcpdump would be very useful at this stage.