Re: Increasing CPU on every execution of an anonymous block sent from WebLogic

  • From: Wolfgang Breitling <breitliw@xxxxxxxxxxxxx>
  • To: paul.baumgartel@xxxxxxxxxxxxxxxxx
  • Date: Tue, 20 Mar 2007 15:40:45 -0600

Just a shot in the dark. Are you using JDBC with thin client or OCI. The JDBC thin client has a problem in that it does not send the bind variables. I don't see how that would cause your cpu increase but I thought I mention it. It also may be fixed in 9.2.0.8. It is fixed in 10g but I don't know in which release exactly. Christian?


At 10:04 AM 3/20/2007, Baumgartel, Paul wrote:
Oracle 9.2.0.8 on Solaris 8.

I've asked development to rewrite Java code that issues SQL with literals so that it uses binds. The new code dynamically constructs an anonymous PL/SQL block with binds, binds values to it, and executes it. The block consists solely of insert statements. I am aware that this would be better to implement this as a packaged procedure, but because this is a heavily-used function of a critical system, development is not willing to move the code from one tier to another.

Each business transaction executes this block a few hundred times.

In testing we found that transactions using this code run for over 100 seconds, at which point WebLogic times them out. (In the test, the dynamically-constructed PL/SQL block is identical every time.) The same code implemented with literals in its dynamic PL/SQL anonymous blocks performs just fine.

Analysis of the trace files for the bind-variable version reveals that the CPU time to EXEC the anonymous block (and, to a lesser extent, its constituent INSERT statements) steadily increases. The first time through the CPU time is about 0.02 second. By the last execution before the timeout, the CPU time is just about 1.0 second. Interestingly, the logical I/O stats for each execution are quite low (<100 current reads and consistent reads) and quite consistent.

We've made a rough copy of the anonymous block and run it several hundred times from SQL*Plus; this setup doesn't show the increasing CPU time. It's not an exact copy of the Java code, so not conclusive, but I wonder if it's possible that interaction between Oracle and WebLogic, maybe in getting the bind variable values, is somehow involved here.

Has anyone seen such behavior?  Any suggestions?

Thanks in advance.

Regards

Wolfgang Breitling
Centrex Consulting Corporation
www.centrexcc.com
______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email ______________________________________________________________________

Other related posts: