IT25666: MASTER MAINTENANCE PREPARE PHASE CODE UPDATE TO PREVENT POSSIBLE MISSING SUBSCRIPTIONS AND CACHE INCONSISTENCIES BETWEEN CPUS.

WebSphere MQ for HP NonStop Server V5.3.1 fix pack 5.3.1.15
WebSphere MQ for HP NonStop Server V5.3.1 fix pack 5.3.1.16

APAR status

• Closed as program error.

Error description

• Errors AMQ9511 and AMQ9448 are logged along with FFST
  regarding ProbeID: RM220005 and Probe Type: MSGAMQ9511
  given by one or more slave repmans due to inconsistencies
  between CPUs regarding subscriptions in number and
  referenced object as well as expiration time.
  When multiple slaves are affected, in consequence the
  problem described in IT24388 may occur.
  This is a generic description of a multi root cause error
  scenario, please also check APARs about other problems
  identified: IT25665, IT25666, IT25669, IT25671.
  This APAR is about a general problem with the consistency
  of cluster cache maintenance, due to relying for multiple
  decisions on current usage state of object records
  in the cache.
  

Local fix

• N/A
  

Problem summary

• The information if a record is currently used by the
  queue manager is kept in shared memory, which is local
  for each CPU on NonStop.
  The maintenance operation on master were relying such
  decisions on the local cache state, not taking into account
  usage of entries on other CPUs. In consequence there have
  been problems due to missing subscriptions and cache
  inconsistencies between CPUs.
  

Problem conclusion

• -
  Code was added to perform a maintenance prepare phase
  querying slave instances, to lock their cache and provide
  information about records currently in use.
  Maintenance was changed to also respect usage information
  from slaves, when ever decisions based on object usage are
  taken.
  Additionally each decision is recorded in the metadata.
  which is provided to the slaves along with the request
  to process, to ensure decisions being consistent for all
  cache instances. To guarantee correctness, slaves will
  hold lock of cluster cache until completing maintenance.
  

Temporary fix

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

  WEBS MQ NSS ITA

• Fixed component ID

  5724A3902

Applicable component levels

• R531 PSY

     UP