PI56807: WMQ QUEUE MANAGER HANG

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• Additional Symptom(s) Search Keyword(s):Queue Manager hangs
  during log processing
  Change Team finds that a channel needs to write a log record
  during MQPUT processing and CSQJW001 has detected that all of
  the log buffers are full so it must kick start the log-writer.
  However, the log RBA passed by CSQJW001 in LRB7RBA8 is
  calculated incorrectly due to the small size of the log buffer
  defined on the queue-manager (ie. OUTBUFF being set to the
  smallest value allowed). This results in CSQJW001 continually
  trying to kick start the log writer for an RBA that has already
  been written, hence no further log buffers are freed up.
  

Local fix

• Code OUTBUFF to at least 80
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of WebSphere MQ for z/OS Version 8 *
  *                 Release 0 Modification 0.                    *
  ****************************************************************
  * PROBLEM DESCRIPTION: When CSQ6LOGP is compiled with OUTBUF < *
  *                      68, the queue manager hangs if all      *
  *                      available log buffers are filled while  *
  *                      writing a log record to the queue       *
  *                      manager logs. High CPU usage can be     *
  *                      observed in the queue manager address   *
  *                      space, and in some cases the address    *
  *                      space of the application waiting for a  *
  *                      log record to be written.               *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  While attempting to write a log record to the queue manager
  recovery logs, CSQJW001 detected that no log buffers were
  available. CSQJW001 resumes the log writer task to write out the
  contents of the log buffers up to a specified RBA, freeing up
  those buffers for reuse. The RBA specified must be prior than
  the RBA of the last complete log record written to the logs and
  so to ensure this is always the case, an RBA 17 control
  intervals prior to the current RBA is passed. However when
  OUTBUF is less than 68, the passed RBA has already been written
  out and the log writer returns immediately - no log buffers are
  written out and so CSQJW001 detects that there are still no
  available log buffers and resumes the log writer task again,
  passing the same RBA, and the log writer again completes without
  writing out any buffers. This repeats indefinitely, leading to
  high cpu in the queue manager address space, and the address
  space writing out the log record.
  Other tasks/applications attempting to write log records while
  this is occurring will suspend on the log writer latch.
  

Problem conclusion

• The minimum value of OUTBUFF is increase to 80, requiring up to
  40K of additional private storage. This ensures the RBA required
  to be passed to the log writer task is always within the scope
  of the log buffer.
  To maintain compatibility, values of OUTBUFF less than 80 will
  be accepted, and will be treated internally as if a value of
  80 had been specified.
  
  The following page in the WebSphere MQ 8.0.0 Knowledge Center is
  updated:
  IBM MQ
   Configuring
    Configuring queue managers on z/OS
     Custominsing IBM MQ
      Task 17: Tailor your system parameter module
       Using CSQ6LOGP (q019310_.htm)
  
  The description of OUTBUFF is changed from
  "The parameter must be in the range 40 through 4000. The value
  specified is rounded up to a multiple of 4. "
  to
  "The parameter must be in the range 80 through 4000. The value
  specified is rounded up to a multiple of 4. Values between 40
  and 80 will be accepted for compatibility reasons, and will be
  treated as a value of 80."
  000Y
  CSQJS001
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

APAR Information

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

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

  UI36474

Modules/Macros

• CSQJS001
  

Fix information

• Fixed component name

  WMQ Z/OS 8

• Fixed component ID

  5655W9700

Applicable component levels

• R000 PSY UI36474

     UP16/05/04 P F605 ¢

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.