Round-robin method not load balanced

This is often a common misunderstanding of how round robin works in F5. 

It's TMM handle the round robin independently 

So if for example you have 4 TMMs, first 4 requests might most probably forwarded to server1, one request from each TMM.

Try disabling server1 and check if traffic is forwarded to server2. Or simulate a lot of requests and check statistics.

Typically when this happens it is most likely because the volume of traffic hitting your virtual server is not sufficient to show an traffic being sent to all the pool members. I recommend increasing the volume of traffic that you use to test the virtual server as well as the duration of the test.

You can give this a shot. Think this might work:

1. Enable SNAT Automap on the virtual server and retest connectivity.
2. Validate real-server routing to force replies via the BIG-IP (use traceroute).
3. Confirm VLAN/subnet assignments and Self IP coverage for all pool member subnets.
4. Tighten health monitors to verify full application functionality, not just TCP.
5. Capture and review packets on both BIG-IP and real servers to trace asymmetric drops.

By systematically addressing SNAT and asymmetric routing first—and then verifying VLAN, health-monitor, and firewall settings—you can restore true round-robin distribution across both servers.

vserver fallback persistence is by default set to source address.

there are many probable causes of the imbalance.
vserver config must match app characteristics to get correct load balancing.

Hi,

After ensuring that the server is up and running, as mentioned above in the comments (using a curl command or by applying an advanced monitor), the next step is to check whether any persistence method is applied at the F5 virtual server level.

In some cases, a persistence method such as Source Address Affinity might be configured. If users are connecting through a proxy, the F5 device may only see a single source IP address accessing the service. As a result, due to the persistence configuration, all traffic could be directed to a single backend server instead of being distributed across the pool members.