PK97291: A DEADLOCK MAY OCCUR DURING SERVANT STARTUP WHEN WRITING AN FFDCDUE TO THE CLASSLOADER'S INSPECTION OF JAR AFTER PK72252

APAR status

• Closed as program error.

Error description

• The servant hangs while starting up due to a deadlock in the
  startup threads in the websphere servant JVM. A javacore of the
  servant shows the following:
  
  1LKDEADLOCK    Deadlock detected !!!
                 ---------------------
  2LKDEADLOCKTHR  Thread "server.startup : 2"
  3LKDEADLOCKWTR    is waiting for:
  4LKDEADLOCKMON      sys_mon_t:0x1ACD1DE8 infl_mon_t:
             com/ibm/ws/bootstrap/WsLogManager@1EC5FC50/1EC5FC5C:
  3LKDEADLOCKOWN    which is owned by:
  2LKDEADLOCKTHR  Thread "Thread-18" (0x53989800)
  3LKDEADLOCKWTR    which is waiting for:
  4LKDEADLOCKMON      sys_mon_t:0x532C7A70 infl_mon_t: 0x532C7AB0:
  4LKDEADLOCKOBJ      java/util/HashMap@30668500/3066850C:
  3LKDEADLOCKOWN    which is owned by:
  2LKDEADLOCKTHR  Thread "server.startup : 2" (0x53778500)
  
  The partial traceback of "server.startup: 2" is:
  at java/util/logging/LogManager.access$900
  at java/util/logging/LogManager$RootLogger.getHandlers
  at java/util/logging/Logger.log
  at com/ibm/ws/logging/WsLogger.deliverOrBuffer
  at com/ibm/ws/logging/WsLogger.log
  at com/ibm/ws/logging/WsLogger.logp
  at com/ibm/ws/ffdc/IncidentStreamImpl.initialize
  at com/ibm/ws/ffdc/IncidentStreamManager.getIncidentStream
  at com/ibm/ws/ffdc/DiagnosticEngine.processIncident
  at com/ibm/ws/ffdc/FFDCFilter.filterEngine
  at com/ibm/ws/ffdc/FFDCFilter.processException
  atcom/ibm/ws/classloader/SinglePathClassProvider.
  buildContentsCache
  .
  .
  at com/ibm/ws/classloader/ClassLoaderManager.initialize
  at com/ibm/ws/classloader/ClassLoaderManager.<init>
  at com/ibm/ws/runtime/component/DeployedApplicationImpl.start
  .
  .
  com/ibm/wsspi/runtime/component/
  WsComponentImpl$_AsynchInitializer.run
  com/ibm/ws/util/ThreadPool$Worker.run
  
  The partial traceback for "Thread-18" is
  
  at com/ibm/ws/classloader/SinglePathClassProvider.openZipFile
  at com/ibm/ws/classloader/SinglePathClassProvider.openZipFile
  at com/ibm/ws/classloader/SinglePathClassProvider.access$200
  at com/ibm/ws/classloader/SinglePathClassProvider$3.run
  at com/ibm/ws/security/util/AccessController.doPrivileged
  at com/ibm/ws/classloader/SinglePathClassProvider.getResource
  at com/ibm/ws/classloader/CompoundClassLoader.findResources
  at java/lang/ClassLoader.getResources
  at com/ibm/ws/logging/WsLoggerConfigurator$3.run
  at com/ibm/ws/security/util/AccessController.doPrivileged
  at com/ibm/ws/logging/WsLoggerConfigurator.getResources
  at com/ibm/ws/logging/WsLoggerConfigurator.
  getCurrentLoggingConfiguration
  at com/ibm/ws/logging/WsLoggerConfigurator.
  getLoggerConfiguration
  at com/ibm/ws/logging/WsLoggerConfigurator.configureLogger
  .
  .
  at com/ibm/ws/wswebcontainer/webapp/WebGroup.addWebApplication
  at com/ibm/ws/wswebcontainer/VirtualHost.addWebApplication
  at com/ibm/ws/wswebcontainer/WebContainer.addWebApp
  at com/ibm/ws/wswebcontainer/WebContainer.run
  at java/lang/Thread.run
  
  PK72252 that shipped in 6.1.0.23, 7.0.0.3, and 6.0.2.33
  introduced a zipfile cache in the websphere runtime code to
  improve performance.  The possible deadlock occurs in this new
  code.  This apar will address this problem.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:  All users of IBM WebSphere Application      *
  *                  Server                                      *
  ****************************************************************
  * PROBLEM DESCRIPTION: The server may hang on                  *
  *                      server or application startup.          *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  On Application Server startup, application jars are loaded and
  put in a global cache to reduce the frequency of opening and
  closing jar files. To maintain the consistency of the global
  cache, access is restricted to one thread at a time. Attempts
  to log messages and errors while the global cache has been
  locked may cause a deadlock.
  

Problem conclusion

• The class loading code has been changed to delay the logging of
  messages and errors until the global cache has been unlocked.
  
  APAR PK90343 is currently targeted for inclusion in Service
  Level (Fix Pack) 6.1.0.29 of WebSphere Application Server V6.1
  of the WebSphere Plugin.
  
  Please refer to URL:
  //www.ibm.com/support/docview.wss?rs=404&uid=swg27006970
  for Fix Pack availability.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

  PK90343

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  WEBSPHERE APP S

• Fixed component ID

  5724J0800

Applicable component levels

• R61A PSY

     UP

• R61H PSY

     UP

• R61I PSY

     UP

• R61P PSY

     UP

• R61S PSY

     UP

• R61W PSY

     UP

• R61Z PSY

     UP

• R700 PSY

     UP