Weblogic 10, WLST and Admin Console Classpath Exceptions

Question

I've been deploying an application to Weblogic 10 (not 10.3 from Oracle, the last version under the BEA logo) for a few weeks now without a problem. We're going to upgrade to 10.3 from Oracle in our development environment and play around with that hopefully today but I can't imagine a dot-version being much different so I figured I'd post the issue anyway.
Recently, we've been having some issues with the WLST script we use to deploy/redeploy the application. It seems to have just started happening randomly which is odd because the script has not been modified since the get-go. What's happening is the application is compiled, staged and then the WLST script kicks off, log into the application server, looks to see if the app is already deployed, if it is, it redploys it with the latest code and then starts it up and then exits. Every now and again, the next time we go to deploy the application, the WLST fails saying that someone currently has a lock on the admin console. When we log in to the admin console, we see that the user the WLST script uses to log into the admin console currently has a lock on the console and we have the option to take the lock from this user.
When we click the button to take the lock, we get a HUGE exception in the lefthand panel where the Lock & Edit button typically is that says something along the lines of:
Error opening ...
The source of this error is javax.servlet.ServletException: javax.xml.transform.TransformerFactoryConfigurationError: Provider net.sf.saxon.TransformerFactoryImpl not found at weblogic.servlet.internal.ServletStubImpl.execute(ServletStubImpl.java:324) at weblogic.servlet.internal.ServletStubImpl.execute(ServletStubImpl.java:175) at weblogic.servlet.internal.RequestDispatcherImpl.invokeServlet(RequestDispatcherImpl.java:528) at weblogic.servlet.internal.RequestDispatcherImpl.include(RequestDispatcherImpl.java:454) at org.apache.beehive.netui.pageflow.scoping.internal.ScopedRequestDispatcher.include(ScopedRequestDispatcher.java:119) at com.bea.netuix.servlets.controls.content.JspContent.beginRender(JspContent.java:551) at ...
If I click on the 'Home' button along the upper righthand corner to go back to the home page of the admin console, the error goes away, the 'Activate Changes' button reappears, and then I click that and the Lock & Edit button reappears, then if I go back to my WLST script and run it, it runs successfully. If I log back into the admin console, everything appears normally. This happens intermittently and I'm not sure why it's happening some times and not others. If my WLST script is incorrect I would assume it would behave this way EVERY time.
I also don't get why the xml transformer factory SOMETIMES throws the exception and sometimes it doesn't. It also happens when I try to click on a JDBC data source. Same exact exception is thrown on the default configuration tab, if I click on any of the other tabs available after clicking on my JDBC data source, those all appear fine, just the initial default Configuration > General tab shows this huge exception. If I stop and restart the application server, these exceptions magically go away...the odd thing is this exception being thrown seems to be referencing libraries made available by my deployed application (the whole net.sf.saxon.TransformerFactoryImpl class is in a library we deploy with our application). These jar files containing these libraries should ONLY be available to the deployed application, I'm not putting these files on the classpath of the application server, so why the application server's admin console have any knowledge of these libraries? Could these be mucking with the overall functionality of the admin server's behavior?
Here is my WLST script I use the deploy the application. If you see something wrong with it, any feedback would be greatly appreciated. Any thoughts on the whole classpath of my application being used by the Weblogic Admin console would be helpful too.
WLST Script:
connect("merlin", "merlin595")
# check to see if the apb data source exists and if so, make sure it's targeting the AdminServer
try:
     # Either deploy or redeploy application
     try:
          print 'Checking for the existence of the APB application.'
          cd('/AppDeployments/apbEar')
          print 'The APB application will be redeployed.'
          progress=redeploy('apbEar')
          print 'Status of redeployment: ' + progress.getState()
     except WLSTException:
          print 'The APB application has not yet been deployed to this server, so application will be deployed for the first time now.'
          progress=deploy(appName='apbEar', path='/opt/weblogic/merlinpb/merlinpbEar/apbEar', targets='AdminServer')
          print 'Status of initial deployment: ' + progress.getState()
     # Start application
     try:
          print 'Starting APB application'
          startApplication(appName='apbEar')
     except WLSTException:
          pass
     disconnect()
     print 'End of script, exiting WLST...'
     exit()
except WLSTException:
     print 'Could not deploy apb. Please log into the admin console and attempt the deployment manually.'
Oh, one final thing I've noticed when I use this script. The last thing the script says WHEN it actually runs successfully is:
<Nov 13, 2008 10:57:13 AM EST> <Warning> <JNDI> <BEA-050001> <WLContext.close() was called in a different thread than the one in which it was created.>
Why does it say that? What does that mean? Is my script mot structure correctly above?
Any help is greatly appreciated. Thanks again!
Rob

Answer 1

Hey Mike,
The problem I was experiencing was that we had code that was applying an XSL stylesheet against XML we are getting from the database. In order for this stylesheet to work, we had to set code that does the following:
System.setProperty("javax.xml.transform.TransformerFactory", "net.sf.saxon.TransformerFactoryImpl");
Now the first time this code was called, and someone subsequently went to the Weblogic admin console, the screens would be completely unreachable...we were experiencing loads of compilation errors. So I did some digging and tried various other ways to set the TransformerFactory class that wouldn't affect the system as a whole. I found, by sheer luck, the fact that you can set this parameter in the weblogic-application.xml file located in the META-INF directory of your root EAR folder. The very first element in the XML of my weblogic-application.xml file is now:
<wls:xml>
     <wls:parser-factory>
          <wls:transformer-factory>net.sf.saxon.TransformerFactoryImpl</wls:transformer-factory>
     </wls:parser-factory>
</wls:xml>
I now deploy the application with this and my reports are transformed properly and the admin console and command-line WLST scripts are all working uninterrupted as well without any adverse side-effects. Hope that helps. Just an aside, in preparation for the upgrade to WLS10, I restructured our application to a split development directory structure...this structure in combination with Workshop 10 really has made development and deployment much easier.
Best of luck. Cheers!
Rob