IV76685: DEADLOCKS TRIGGERED IN SOME CASES WHEN WORKFLOW_INSTANCE_SAVING=ALWAYS

APAR status

• Closed as fixed if next.

Error description

• It is possible to trigger a deadlock when saving workflow
  instances.  This was noticed when using TRIRIGA 3.3.2.2 with
  Microsoft SQL as a database, but this may happen with other
  vversions of TRIRIGA, as well.
  
  Symptoms are that after setting WF_INSTANCE_SAVE to 'ALWAYS'
  instead of 'ERRORS_ONLY' workflows were hanging at the same time
  that SQL and/or Java thread deadlock errors were being written
  to the server.log.  The following is an example of one of the
  types of errors being observed in the server.log when this
  problem happened:
  
     Transaction (Process ID 67) was deadlocked on lock |
     communication buffer resources with another process and has
     been chosen as the deadlock victim. Rerun the transaction.;
     nested exception is java.sql.SQLException: Transaction
     (Process ID 67) was deadlocked on lock | communication buffer
     resources with another process and has been chosen as the
     deadlock victim. Rerun the transaction.. Workflow
     Information: [...]
  

Local fix

• 1. Try to keep WF_INSTANCE_SAVE set to ERRORS_ONLY,
     particularly in PODUCTION environements
  
  2. On SQL Server, ensure MAXDOP is set to to a percentage of CPU
     cores on the server.  For eaxmple, if a 16 core DB server, is
     present set MAXDOP to 4.  This value should never be set to
     0.
  

Problem summary

• Added a warning if a deadlock occurs when saving workflow
  instances, stating the process could not save workflow instance
  step list. You may be saving too many instances. Do not set
  Workflow Instance Saving to ALWAYS in a produciton environment.
   This warning is targeted to go into the 2H2015 release.
  

Problem conclusion

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:

Modules/Macros

• 999
  

Fix information

Applicable component levels

• R342 PSY

     UP