IT37878: WebSphere Application Server activation specifications do not pause when an MDB calls setRollbackOnly()

APAR status

• Closed as program error.

Error description

• An IBM MQ messaging provider activation specification has been
  defined within WebSphere Application Server 9.0.5.3 (or later).
  The activation specification has been configured with:
  
  - The "Stop endpoint if message delivery fails" checkbox
  selected
  - The "Number of sequential delivery failures before suspending
  endpoint" property set to value 0.
  
  so that it will pause the first time that a message is rolled
  back.
  
  A message-driven bean (MDB) application is deployed into the
  application server, and configured to use this activation
  specification. The MDB contains logic to determine if a
  transaction has failed and needs to be rolled back - in this
  situation, the MDB calls the method:
  
    javax.ejb.EJBContext.setRollbackOnly()
  
  When the MDB detects that a transaction has failed, the
  transaction is successfully rolled back and the message that the
  MDB was processing at the time is made available for consumption
  on the queue.
  
  However, the activation specification does not pause. As a
  result,  it detects the same message again and redelivers it to
  the MDB for processing.
  

Local fix

Problem summary

• ****************************************************************
  USERS AFFECTED:
  This issue affects all users of the WebSphere Application Server
  9.0.5.3 (and later) MQ messaging provider who:
  
  - Have activation specifications defined with the "Stop endpoint
  if message delivery fails" checkbox selected, and the "Number of
  sequential delivery failures before suspending endpoint"
  property set.
  - And have deployed a message-driven bean (MDB) application into
  the application server which is configured to use the activation
  specification, and contains logic to call the method
  EJBContext.setRollbackOnly() to roll back transactions that have
  failed.
  
  
  Platforms affected:
  MultiPlatform
  
  ****************************************************************
  PROBLEM DESCRIPTION:
  APAR IT27950 was raised to address an issue which prevented
  managed connections associated with transactions that had been
  rolled back from being reused within JBoss Application Server.
  
  As part of these changes, the way in which the MQ resource
  adapter handled transactions which had been rolled back was
  modified so that the XA resource associated with those
  transactions was moved straight into an internal NON_EXISTENT
  state rather than a ROLLBACK_ONLY state. Although this allowed
  any managed connections associated with the XA resource and
  transaction to be reused within JBoss Application Server, it
  meant that the MQ resource adapter no longer notified WebSphere
  Application Server that a delivery failure had occurred if an
  message-driven bean application had called
  EJBContext.setRollbackOnly().
  
  As a result, the application server did not increment its
  internal delivery failure counter, and so would never pause an
  activation specification when the specified number of
  consecutive delivery failures (as defined by the activation
  specification property "Number of sequential delivery failures
  before suspending endpoint") had occurred.
  
  The code change associated with APAR IT27950 was first shipped
  in the MQ 9.1 LTS release at MQ version 9.1.0.4.  This was first
  included within the WebSphere Application Server installation at
  version 9.0.5.3.  Subsequently this is the first version of the
  WebSphere Application Server which is affected by this issue.
  

Problem conclusion

• To resolve this issue, the MQ 9.1 resource adapter has been
  updated so that if a transaction is rolled back, then the XA
  resource associated with a transaction will first be moved into
  an internal ROLLBACK_ONLY state, followed by an internal
  NON_EXISTENT state. This ensure that the MQ resource adapter:
  
  - Initially notifies WebSphere Application Server that a
  delivery failure has occurred, which allows the application
  server to increment its delivery failure counter and pause the
  activation specification if required,
  - And the clears some internal state within the managed
  connection associated with the XA resource and transaction,
  which allows the managed connection to be reused.
  
  ---------------------------------------------------------------
  The fix is targeted for delivery in the following PTFs:
  
  Version    Maintenance Level
  v9.1 LTS   9.1.0.11
  v9.2 LTS   9.2.0.5
  v9.x CD    9.2.5
  
  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