PH03166: MQ: THE DELETION OF A SHARED QUEUE DOESN'T CORRECTLY UPDATE THE CLUSTER CACHE CAUSING HIGH CPU.

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• The problem is as follows:
  .
  Two queue managers (QMA and QMB) are members
  of a QSG and are also members of a cluster.
  The QSG has a shared queue (SQ1) which is defined
  as being in the cluster. This results in both
  queue managers advertising an instance of that
  queue to other members of the cluster.
  .
  SQ1 is then deleted. This should cause both
  queue managers to send an update to the cluster
  to notify other members that the queue manager
  no longer hosts an instance of that clustered
  queue. However, for shared queues this update
  does not happen (at least, not straight away).
  .
  The result of this is that the cluster cache
  on each qmgr has two records for the queue
  (one for each qmgr), but neither has an instance
  of the queue to put messages to.
  .
  When a message is put with a queue name
  SQ1 on QMA, it detects that there isn't a local
  queue instance, so it uses the cluster cache to
  resolve the location of the queue name.
  As no local instance exists, it selects the only
  other entry for the queue (QMB) and puts the
  message to the SYSTEM.CLUSTER.TRANSMIT.QUEUE to
  be sent to QMB.
  .
  When the message is sent over the channel,
  QMB also detects that there is no local instance
  of the queue, so goes to the cluster cache and
  determines that QMA is the only available instance.
  .
  The message loops between the two qmgrs. This
  causes high CPU, and if the message is persistent
  then it also causes the high logging volume
  seen by the customer.
  .
  Additional Symptom(s) Search Keyword(s):
  

Local fix

• Restart the QMQRs. The cache did get updated after the queue
  managers were restarted.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of IBM MQ for z/OS Version 9       *
  *                 Release 1 Modification 0.                    *
  ****************************************************************
  * PROBLEM DESCRIPTION: Deleting a shared cluster queue may     *
  *                      result in the cluster definitions for   *
  *                      the shared queue remaining in the       *
  *                      cluster after a successful shared queue *
  *                      delete.                                 *
  ****************************************************************
  If multiple members of a QSG are also members of the same
  cluster, when a shared cluster queue is deleted, the cluster
  records for the queue may continue to exist in the cluster. This
  can result in the cluster hosting records for queues which no
  longer are valid. If messages are put to one of these queues,
  cluster resolution will attempt to put the message to another
  QMGR in the cluster where the queue was previously hosted, which
  results in further cluster resolution and subsequent puts to
  other cluster QMGRs, which can result in infinite loop of
  cluster resolution and puts to other QMGRs. This is due to
  shared queue deletes not correctly broadcasting the delete of
  the cluster queue in this case.
  
  The looping between QMGRs can result in high CPU usage on all
  the QMGRs involved. If the message put was persistent, this will
  also result in high logging volumes. When this scenario is
  encountered, a cancel may be required to stop the QMGR.
  

Problem conclusion

• Shared queue delete broadcast for cluster queues has been
  corrected to ensure cluster records are correctly deleted when a
  delete shared queue command is issued.
  

Temporary fix

Comments

APAR Information

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

  PI79259

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

  UI60586

Modules/Macros

• CSQMUQLC
  

Fix information

• Fixed component name

  IBM MQ Z/OS V9

• Fixed component ID

  5655MQ900

Applicable component levels

• R100 PSY UI60586

     UP19/01/26 P F901 ¢

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.