PH40880: CICS COMMIT processing can leave Runaway timer active

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• A CICS region suffered the following abend which caused a
  problem with the shared JVM class cache and which caused all
  CICS regions to hang that used this class cache.
  
  +DFHSJ0001 An abend (code ---/AKEA) has occurred at offset
  X'64BC' in module DFHSJTH.
  
  ---------------------------------------------------------------
  
  The dump showed that a purge was issued against a task running
  on a T8 TCB leading to an abendU0999 after the following
  sequence of events:
  
  - It appeared that some activity occurred (probably EJB related)
    that exposes a CICS defect which causes the Runaway detection
    to still be active for a TCB/Thread in the CICS liberty
    threadpool after application activity has completed.
  
  - Later on the same thread is used for a 'container' operation
    that is not directly related to application activity and
    causes a scan of classes in files for annotations.
  
  - because this annotation scanning is intensive (both in CPU
    and I/O), the Runaway timer, that was invalidly left on, is
    tripped and an abendU0999 is issued on this TCB/pthread.
  
  This APAR was taken to address the root cause defect, where
  commit processing in CICS can leave the runaway timer active
  for threads that are returned to the CICS Liberty threadpool.
  

Local fix

• Not available
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All CICS users.                              *
  ****************************************************************
  * PROBLEM DESCRIPTION: DFHSJ0001 AKEA ABEND in DFHSJTH due to  *
  *                      Runaway timer expiring.                 *
  ****************************************************************
  CICS module DFHRMOT has functions that support commit and
  rollback.
  
  Prior to the call to Phase 2 processing in these functions
  there are calls that stop and start runaway processing and
  force purge protection.
  
  The intention is that:
  - runaway processing should be stopped before, and started
  after, the Phase 2 call.
  - force purge protection should be started before, and stopped
  after, the Phase 2 call.
  
  These calls were not correctly balanced for both the commit and
  rollback cases, so that following the Phase 2 call:
  
  - runaway processing was started having not been stopped
  - force purge protection was not stopped having been started
  
  As a result of this defect a problem can occur when EJB
  applications run in a CICS Liberty JVM server.
  
  In this case, following the rollback or commit for the EJB, the
  runaway processing is still active when the thread is returned
  to the CICS Liberty threadpool, when it should be off.
  
  Subsequently, if this thread is used for a CPU intensive
  operation (for example, installing an application that uses CDI)
  the runaway timer can trip, an abendU0999 is issued and message
  
  +DFHSJ0001 An abend (code ---/AKEA) has occurred at offset
  X'64BC' in module DFHSJTH.
  
  is issued.
  
  Keywords: ABENDU0999 RUNAWAY EJB LIBERTY
  

Problem conclusion

• Module DFHRMOT was updated to correctly balance the setting of
  runaway detection and force purge protection over the Phase 2
  processing, for both commit and rollback.
  

Temporary fix

Comments

APAR Information

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

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

  UI79329 UI79330

Modules/Macros

• DFHRMOT
  

Fix information

• Fixed component name

  CICS TS Z/OS V5

• Fixed component ID

  5655Y0400

Applicable component levels

• R200 PSY UI79330

     UP22/02/16 P F202

• R300 PSY UI79329

     UP22/02/16 P F202

Fix is available

• Select the PTF appropriate for your component level. You will be required to sign in. Distribution on physical media is not available in all countries.