PH37758: MQ Z/OS: REPOSITORY QUEUE IN LARGE CLUSTER INDEXING BY CORRELID SKIPS MESSAGES

APAR status

• Closed as duplicate of another APAR.

Error description

• When the repository queue is deep (such as in a large cluster,
  between checkpoints), updates to objects locate the message
  representing the old/changing definition by messageid, but the
  queue is indexed by correlid (as is required for other
  processing). This results in a scan of messages on the page
  chain for the queue, skipping messages until the required
  message is located.
  .
  Additional keywords/symptoms:
  CSQI004I SYSTEM.CLUSTER.REPOSITORY.QUEUE SCRQ indxtype
  

Local fix

Problem summary

Problem conclusion

Temporary fix

Comments

• Users Affected:
  
  All users of IBM MQ for z/OS Version 9
  Release 2 Continuous Delivery (CD) Release.
  
  Problem summary:
  
  Increased CPU usage in a clustering environment when updating
  records on SYSTEM.CLUSTER.REPOSITORY.QUEUE.
  SMF 116 Class 3 accounting records for the queue show a large
  number of 'skipped messages' (DWQ.GETSMSG) during MQGET
  processing.
  
  
  
  Problem resolution:
  
  The repository task will now locate records for updated objects
  using both MsgId and CorrelId.
  Additional optimisation has been added when matching both
  MsgId and CorrelId on indexed queues.
  
  
  The fix is targeted for delivery in APAR PH43968.
  

APAR Information

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

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

Fix information

Applicable component levels