Due to default Idle time out and port collision facing slow speed

I explain:

 

RMI idle time-outs: RMI is a java protocol for RPC. Most time-outs in the MDM configuration have 600sec and some 300s, where the F5 has 300 I understand. For complex actions users get an "in process" window for more than 5 minutes and then when the action completed the TCP socket is already closed and the client is not updated and stays "in process" for extended periods of time. When the user checks for the action it may have completed or may failed, but they are not updated as expected in the client side.

 

Port Collision: Normally from client side a dynamic port is assigned for tcp communications in the range from 32768 and 65536. As the F5 is doing sNAT the cluster always see the same IP from the client, and the port may be duplicated from different clients in the other side effectively using the same port and one of them getting a TCP reset. The probability is low but given the amount of activity in production is does happens at peak hours.

 

Can we update the default idle timeouts in the F5 to 610 seconds (or at least to 310 for some of the cluster time-outs hit first)? Is there another way to prevent the closing idle tcp sockets?

 

Can we resolve the port collision in any way? can we use more IPs maybe to avoid port collisions?

 

I look forward to your feedback

 

Thank you Kamlesh Yadav