IP load balancing sticky connections

I am using ipvsadm and arptables_jf service for configuring a 4 node cluster with load balancing capability. Nodes are running on RHEL 6.7. I am using RR scheduling algorithm.
When I try to mount multiple NFS shares with different client nodes, all the requests are served by different cluster node depending on the scheduling algorithm.
However, when I try to mount multiple NFS shares using a single client, all the requests are served by same cluster node, for particular time (~300 seconds) frame from first connection.
I suspect this is due to persistence time and/or sticky connection.
How can I reduce this persistence time? and sticky connection?

I have done "sbin/ipvsadm --set 10 10 10" which is supposed to set FTP connections time limit to 10 seconds, but this is not working.

Please educate me if I am doing something wrong, and how I can achieve zero/minimal persistence/sticky time.

Thank you.