IT34804: Subscription messages are delivered out of sequence when using clustered pub/sub with multiple receiver channels

APAR status

• Closed as program error.

Error description

• When a cluster queue manager is configured to be part of
  multiple overlapping clusters by defining separate channels
  (CLUSRCVR) for each cluster, (and it is also part of a
  publish/subscribe cluster topology), publications using the
  publish/subscribe cluster can be unexpectedly delivered out of
  sequence.
  
  This is because a sending (publishing) queue manager is able to
  route messages using both automatically defined channels based
  upon the target queue managers' CLUSRCVR definitions.
  
  For example, a queue manager called HUB has the following
  definitions ...
  DEFINE CHL(TO.CLUSTER1.PAYMENT) CHLTYPE(CLUSRCVR)
  CLUSTER(CLUSTER1) CONNAME ...
  DEFINE CHL(TO.CLUSTER2.STOCK) CHLTYPE(CLUSRCVR)
  CLUSTER(CLUSTER2) CONNAME ...
  
  A remote queue manager publishing messages could alternate
  between theses two channels with no guarantee of arrival
  sequencing (message ordering)
  Queue Manager "Name Resolution" simply attempts to find a route
  to the remote queue manager, however, publications where at all
  possible, should use channels which are shared in the same
  cluster as the TOPIC / TopicString combination.
  

Local fix

Problem summary

• ****************************************************************
  USERS AFFECTED:
  Systems of Publish Subscribe Clusters with multiple auto defined
  cluster sender channels to a destination queue manager which
  happen to be shared in different clusters.
  
  
  Platforms affected:
  MultiPlatform
  
  ****************************************************************
  PROBLEM DESCRIPTION:
  The problem was caused by the passing of all eligible
  destination channels to the workload management algorithm
  without consideration of the cluster that the topic object is a
  member of.
  As such, when there was more than one destination, the workload
  management algorithm could workload balance the publications and
  arrival of messages at the destination could be in a different
  order than when they were put.
  

Problem conclusion

• The problem is resolved by initially employing a filter so that
  channels are chosen that are in the same cluster as the
  clustered topic as expected.
  If a channel cannot be found after filtering, then an
  alternative route will be used if available.
  
  ---------------------------------------------------------------
  The fix is targeted for delivery in the following PTFs:
  
  Version    Maintenance Level
  v9.1 LTS   9.1.0.20
  v9.2 LTS   9.2.0.20
  v9.3 LTS   9.3.0.10
  v9.x CD    9.3.4
  
  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

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

  IBM MQ BASE MP

• Fixed component ID

  5724H7271

Applicable component levels