i have a process model as attached , my aim is to show the user input task (UI)

AS per my understanding below are the points that might be breaking chaining in your process model
1. You have 'Execute process report' node in process model before user input task, which might be taking long time to get execute and breaking the chaining.
2. All nodes in 'Get Repayment scanned Document' sub process should be chained.
3. Sub process should not have any node which required to be wait (e.g. times event, user input task)
And one more thing that I have observed in your process model is, you can probably remove parallel flows and 'And' gateway as it seems that both flows are assigned to system. (Assuming that Get repayment Document does not have user input task)
PFA screen shot which depicts the PM after removing parallel flows and ‘AND’ gateway

Thanks for your kindly help.
Regarding to point 2 & 3, i've checked that all nodes are chained and no times events/user input task in sub process.
For the point 1, the 'Execute process report' executed around few seconds only, how can i check if it breaks the chaining ?

Thanks a lot!

It shouldn't be issue if it is taking seconds,Can you give a try by removing parallel flows?

Check the db_PX*.log, broken chains will show up there. Just make sure you analyze on what engine the process that showed the behavior was running on so you open the right db_PX*.log

@ajinkab277 : Removed the parallel flows but it doesn't help.
@Eduardo: Thanks for your advice. I will taks a look.

how about i set the value of "server.conf.exec.CHAINED_EXECUTION_NODE_LIMIT"?
Does it help ?

@Eduardo: Here's the log in db_PX*.log:
"Could not chain to attended activity. Limit of 50 unattended nodes executed was reached"

Seems it is due to conf problem, isn't it?

Default value of CHAINED_EXECUTION_NODE_LIMIT is 50, so that's not the case with your process model.

Might be your sub process has more than 50 nodes.

If the log is saying that then it means you're indeed exceeding the limit which makes sense if the flow is going through the sub-process, we can't count the nodes because we don't see the whole model .

Changing this value is not a best practice, redesigning your application is the best approach. Either you put an intermediate task that shows the message "Processing your request, submit to continue" or move away from the wizard like approach for this particular scenario.

When you have more than 50 nodes a poor user experience is created, the user might think the environment is slow.