I guess i know the reason as seen from the forum (it was suggested that the no.of db connections should be more or equal to the no.of threads).
The job hung thrice in the 6 months but every time it failed in the parallel step (split/flow). A restart worked.

My question is: I want to turn on the log.For what packages should I be enabling the log so as to prove that it is because of exhausted db connections?

Note: I turned log for DBCP but it never reached the MAX Connections.

Thanks