PI55397: WMQ EXCESSIVE POST CALLS WHEN INTERNAL QMGR TRACE IS ACTIVE ON SYSTEM WITH MANY CONNECTED APPLICATIONS

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• Excessive POST calls are observed when internal queue manager
  trace is active on a system with a large number of connected
  applications. Change Team determined that performance is
  impacted at Version 8 such that applications may experience
  slower response times. Transmission queue backlogs were also
  observed.
  .
  Additional keywords:
  CML LOCL lock
  

Local fix

• Increase the maximum number of trace blocks using for example
    /cpf RECOVER QMGR(TUNE TABMAXBLK 10240)
  where "cpf" is the command prefix for the queue manager.
  
  The blocks are 8KB in size, so a setting of 10240 will require
  80MB of storage. This is taken from the 64 bit address range
  above the bar, so it should not cause a storage problem.
  
  The TABMAXBLK value is not preserved over a queue manager
  restart, so you can include the RECOVER QMGR command in a
  CSQINP2 member minus the "/cpf" syntax.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of WebSphere MQ for z/OS Version 8 *
  *                 Release 0 Modification 0.                    *
  ****************************************************************
  * PROBLEM DESCRIPTION: When running with MQ MSTR trace active  *
  *                      on a busy system, a high volume of post *
  *                      calls can occur causing performance     *
  *                      degradation due to lock contention.     *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  If MQ trace is started with TRACE(GLOBAL) and DEST(RES), the
  queue manager trace is written to a 64 bit storage area in the
  MSTR address space. The storage area is split into blocks which
  are used by individual tasks to capture their trace.
  
  On a queue manager where there are sufficient connections to
  cause all the blocks (1024 by default) to be in use, some tasks
  will not have a trace block available. For these tasks, on each
  trace call a check is be made to see if a block is available.
  If not, the trace data is written to the shared trace storage in
  ECSA.
  
  During the check for available blocks, a post call is made to a
  worker task in the MSTR address space. The LOCAL/CML lock is
  required to complete the post. The frequency of post calls from
  multiple tasks can result in contention for the LOCAL/CML lock.
  This can cause slow response times and increased CPU usage for
  applications making MQ API calls.
  

Problem conclusion

• Trace processing has been modified to reduce the frequency of
  post calls and relieve the LOCAL/CML lock contention.
  000Y
  CSQMSNTF
  CSQWVGSG
  

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:

  PI55911 UI35204

Modules/Macros

• CSQMSNTF CSQWVGSG
  

Fix information

• Fixed component name

  WMQ Z/OS 8

• Fixed component ID

  5655W9700

Applicable component levels

• R000 PSY UI35204

     UP16/03/03 P F603 ¢

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.