We are running an old version of Discoverer (10.1.2.3) w/ the users on Discoverer Plus (10.1.2.55.26). When originally installed, the server was running RHEL 4 and the client PCs were running 32-bit XP, IE5.5, and a very old Java plugin (1.4.2 or older). The database that they were reporting against in those days was 10gR2. Over the years we have upgraded just about everything in our infrastructure except Discoverer. We are now on 11gR2, RHEL5, IE9, Java 1.6.0_35, and Windows 7 64-bit.
Since making all these changes in our environment we have started seeing an odd issue where users start a report, and it never finishes (at least from the client perspective). If I monitor their DB session in Grid Control, I see their queries run to completion, but the results never make it back to the client. If I examine the user session in GC, I see that the session is in a wait state (SQL*Net message from client) and the wait duration just keeps increasing no matter what we do. The only way to get such sessions to end is to restart the Discoverer processes on the server (i.e. opmnctl stopproc ias-component=Discoverer;opmnctl startproc ias-component=Discoverer).
Has anyone else seen this? What can we do to diagnose the cause of this issue?
I found document ID 414786.1 that describes issues resulting from the JVM heap being too small. We tried changing that with mixed results - sometimes it helps, and sometimes not.
Have you enabled the java console on a client's side to see what errors are being thrown? This enabled me to work through some errors I had with CP 10 of 10.1.2.3 (Oracle provided a bug fix patch to be applied after CP10 was applied).
Also, I would examine a log file on the Application Server side. Next, I would do a search in My Oracle Support in the Knowledge Base.