PH35226: MAKING LOG PART ROTATION CONFIGURABLE BASED ON FILE SIZE OR NUMBER OF LINES IN COMPUTE GRID

APAR status

• Closed as program error.

Error description

• When apps log to their job log, we write 1000 lines per file
  until it rotates to a new log part (i.e. part.1.log part.2.log,
  etc..).
  On the Job Scheduler side, for each job there?s a
  process that wakes up periodically and asks for new job logs to
  send back to sgrid. It doesn?t know how large the files are,
  just how many there are to read. The Scheduler puts lines from
  the log it received into a buffer and sends a message back to
  wsgrid with that info. This is repeated until all the lines
  from the log have been processed, and all the log parts are
  processed.
  The log part sizes can vary and potentially get very
  large if the application creates very long lines. This could
  cause an OOM if the heap is not large enough to hold it all.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:  All users of IBM WebSphere Application      *
  *                  Server                                      *
  *                  Java Batch                                  *
  ****************************************************************
  * PROBLEM DESCRIPTION: Java batch job logs with long lines     *
  *                      cause OutOfMemoryError with WSGRID if   *
  *                      the                                     *
  *                      heap is not large enough to hold it     *
  *                      all.                                    *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  When WebSphere Java Batch applications write logs, there are
  1000 lines written to a log "part" file at which point a new log
  "part" file is started (part.1.log part.2.log, and so on).
  If very long lines are written to these logs, it can cause an
  OutOfMemoryError in the Job Scheduler when log files are
  retrieved.  This is due to the fact that when WebSphere Java
  Batch
  job is submitted with the WSGRID utility, job logs are retrieved
  periodically by line.
  

Problem conclusion

• A new WebSphere property has been added to allow Java batch job
  log part rotation based on the size of the log file, rather than
  rotating at 1000 lines:
  
  GRID_ENDPOINT_MAX_JOB_LOG_SIZE
  
  If a valid maximum size is specified, job logs will use the
  current default behavior (1000 line count before rolling over to
  new part) unless the maximum size is hit before then, at which
  point it will roll over to a new log part.  A value of 10000
  bytes or larger can be specified for this variable.
  
  The fix for this APAR is targeted for inclusion in fix pack
  9.0.5.9 and 8.5.5.21. For more information, see 'Recommended
  Updates for WebSphere Application Server':
  https://www.ibm.com/support/pages/node/715553
  

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:

Fix information

• Fixed component name

  WEBSPHERE FOR Z

• Fixed component ID

  5655I3500

Applicable component levels