BIGIP LTM Persistence

I have 1 VServer referencing a pool with 2 nodes in it. The Nodes are monitored for availability via a specific port 1533. As long as the port 1533 service is active on the node, the node is flagged as up and data is pushed to this node accordingly.

FTP data is sent to this VServer.

In my scenario, I rely on the app people to enable the 1533 service or disable it depending on which node is currently acting as the primary, so essentially I only have 1 node active at any one time in the pool.

Recently, i had the non-active node crash and auto reboot and upon the restart, the port 1533 service started automatically, hence bringing up the node in the F5 pool and data started to traverse to this node. I have round robin enabled by default and the data was flowing to both pool members as they were both flagged as active.

I need to determine a way to ensure that all data will be persistent to the 1 real active node if this issue happens again.

But when we manually stop the port 1533 service on the current primary node and then start the service on the alternate node, I need all traffic traverse to the newly activated node and not stay persistent on the old node that just had its port 1533 service disabled.

I have tried destination persistence (and multiple other persistence profiles) but this did not meet my need as the data would always be sent to the most recent used dest address regardless of it state, so long as the current session was within the relevant timeout period, hence some of the session remain on the non-active node indefinitely.

Any suggestions on how to resolve this issue would be greatly appreciated?