IT33751: WHEN MQ HAS TO HANDLE A LOT OF PROCESS STARTS AND TERMINATIONS IN A SHORT TIME, SOME PROCESS TERMINATIONS MAY BE MISSED

IBM MQ for HP NonStop Server V8.1 Fix Pack 8.1.0.8

APAR status

• Closed as program error.

Error description

• When MQ has to handle a lot of process starts and terminations
  in a short time, due to a defect some process terminations may
  be missed.
  
  Unfortunately this does not lead to an FFST file, an error
  message in the error logs or EMS.
  The most likely observation is a blocking RUNMQSC process.
  The RUNMQSC process might have been accepted a command
  "ping channel..." but does not return to the prompt.
  The problem might show up with other process types like
  LISTENER or SERVICE as well.
  

Local fix

• It's not a risk to stop the process managers as they will be
  restarted by MQ.
  If a blocking or not startable process can be observed and
  there is no related error log or FFST file, stopping the
  process managers should be considered.
  If there is more than one process manager running on the system,
  they should be stopped one by one until the block is dissolved.
  The process manager is amqzmgr0. This command lists the running
  process managers:
  $ ps -l | grep amqzmgr0
  The kill command on OSS and stop command on TACL can be
  used to stop that process.
  

Problem summary

• The root cause is an implementation error in the MQ internal
  process table.
  Due to a process table slot reuse, MQ does not handle process
  terminations properly and treats some already terminated
  processes as still being alive.
  This inconsistency affects new process starts and leads to
  RUNMQSC blocks and service start problems.
  

Problem conclusion

• The product has been changed to fix this
  defect.
  ---------------------------------------------------------------
  The fix is targeted for delivery in the following PTFs:
  
  Version    Maintenance Level
  v8.1       8.1.0.8
  The latest available maintenance can be obtained from
  'WebSphere MQ Recommended Fixes'
  http://www-1.ibm.com/support/docview.wss?rs=171&uid=swg27006037
  If the maintenance level is not yet available information on
  its planned availability can be found in 'WebSphere MQ
  Planned Maintenance Release Dates'
  http://www-1.ibm.com/support/docview.wss?rs=171&uid=swg27006309
  ---------------------------------------------------------------
  

Temporary fix

• The problem has been observed in scenarios where MQSC
  commands have been automatically created in parallel.
  When each MQSC command starts a process (ping channel
  for instance), the process manager may have to start lots
  of processes and handle their process termination in a
  short time.
  Such scenarios should be avoided, once the issue showed up.
  

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

  MQ FOR HPE NS O

• Fixed component ID

  5724A3903

Applicable component levels

• R810 PSY

     UP