A little background on the application that I am gonna talk about in the next few lines:

XYZ is a data masking workbench eclipse RCP application: You give it a source table column, and a target table column, it would apply a trasformation (encryption/shuffling/etc) and copy the row data from source table to target table. Now, when I mask n tables at a time, n threads are launched by this app.

Here is the issue:

I have run into a production issue on first roll out of the above said app. Unfortunately, I don't have any logs to get to the root. However, I tried to run this app in test region and do a stress test.

When I collected .hprof files and ran 'em through an analyzer (yourKit), I noticed that objects of oracle.jdbc.driver.T4CPreparedStatement was retaining heap. The analysis also tells me that one of my classes is holding a reference to this preparedstatement object and thereby, n threads have n such objects. T4CPreparedStatement seemed to have character arrays: lastBoundChars and bindChars each of size char[300000].

So, I researched a bit (google!), obtained ojdbc6.jar and tried decompiling T4CPreparedStatement. I see that T4CPreparedStatement extends OraclePreparedStatement, which dynamically manages array size of lastBoundChars and bindChars.

So, my questions here are:

1. Have you ever run into an issue like this?
2. Do you know the significance of lastBoundChars / bindChars?
3. I am new to profiling, so do you think I am not doing it correct? (I also ran the hprofs through MAT - and this was the main identified issue - so, I don't really think I could be wrong?)

I have found something similar on the web here: http://forums.oracle.com/forums/thread.jspa?messageID=2860681

Appreciate your suggestions / advice.