Problem with leaking connections

Hi,

we have an webapplication running on tomcat6 with hibernate.

We are using the apache DBCP connection pool (tomcat built in) and an
postgres 8.19 database server on an different host.

Every few days the pool is running out of connections.

The pool is configured to aggressively close connections when it is
exhausted, but the application gets very slow when this happens, mostly we
have to restart

the tomcat server.

#ps -aux | grep postgres on the database host shows me many open connections
in the select or in transaction state (that never get closed until tomcat or
postgres restart).

Is there a way to figure out which statement didn't closed the connections?

In the application we found a few places where the connections not closed
and fixed them.

But they are hard to find.

Is it possible to configure postgres to close connections after a timeout or
something like this (maybe statement_timeout)?

Are there any other possibilities?

Thanks

Mark