IZ90144: A JMS LISTENER PORT STOPS PROCESSING MESSAGES AFTER WAS UPGRADE FROM V6 TO V7

WebSphere MQ V7.0 Fix Pack 7.0.1.6
WebSphere MQ V7.0.1 for i5/OS Fix Pack 7.0.1.6

APAR status

• Closed as program error.

Error description

• A JMS listener port stops processing messages after a WebSphere
  Application Server (WAS) upgrade from V6 to V7, leading to a
  build up of messages on the queue. Listener port appears as
  'Running' in the WAS Admin Console, and there are no exceptions
  reported in SystemOut.log.
  

Local fix

• 1. Restart listener or a WAS application server
  2. Disable ReadAhead feature in WAS V7.
     Please set a ReadAhead flag to 'as queue definition' for a
     destination in WAS Admin Console and restart the application
     server.
  

Problem summary

• ****************************************************************
  USERS AFFECTED:
  This issue affects users of the WebSphere MQ classes for
  Java/JMS who are using the Read Ahead functionality provided
  within version WebSphere MQ v7.
  
  This includes users of the WebSphere MQ Resource Adapter v7,
  such as the WebSphere Application Server v7, or WebSphere
  Application Server v6 which is configured to use the v7
  WebSphere MQ classes for JMS.
  
  Platforms affected:
  All Distributed (iSeries, all Unix and Windows) +Java
  ****************************************************************
  PROBLEM SUMMARY:
  WebSphere MQ v7 introduced new functionality known as 'Read
  Ahead'.  When working with non-persistent messages, this
  function permits the client classes to download messages from
  the queue before the client application requests the message.
  This can result in a more efficient system, where the message
  has already been retrieved over the client (TCP) connection by
  the time the client is ready to consume it.
  
  The Read Ahead functionality can also be used with
  non-destructive gets with persistent messages, such as when
  browsing a queue.  In the application server environment, this
  can be used by the activation specification or listener port
  browsing threads, when using the default Application Server
  Facilities (ASF) functionality of the WebSphere MQ Resource
  Adapter.
  
  The use of Read Ahead is configured as a client side queue
  definition setting.  For example, when using the WebSphere
  Application Server, in the Administration Console on the
  queue definition advanced properties:
  
  Resources -> JMS -> Queues -> [queue] > Advanced properties
  
  the setting is configured by the option titled:
  
  "Read ahead, and cache, non-persistent messages for consumers"
  
  The default setting is "As per queue definition", which means
  the value set on the queue on the queue manager is used, for
  the queue property "DEFREADA", which defaults to "NO".
  
  Note that when migrating WebSphere Application Server from v6
  to v7, the default client side value is set to "YES", which
  results in systems using Read Ahead during the process of
  consumption of Message Driven Beans (MDBs) which are being
  triggered by messages from a v7 WebSphere MQ queue.  This
  migration issue is discussed further in the following Technote:
  
  http://www.ibm.com/support/docview.wss?rs=180&uid=swg21448387
  
  
  The problem described by this APAR concerns the use of Read
  Ahead when using a WebSphere MQ classes for Java/JMS based
  client, including that used by an application server when using
  the WebSphere MQ Resource Adapter.
  
  When such a system was using Read Ahead, a defect in the
  threading model used by the WebSphere MQ Java classes resulted
  in the client losing track of the amount of data it had
  requested, received and processed from the Read Ahead local
  caching mechanism.  The problem was typically observed when
  the system was put under load with many applicable messages on
  the WebSphere MQ queue, which resulted in the maximum number
  of server sessions being driven simultaneously in the
  application server environment.
  
  When the problem was encountered, the only external observable
  event was that the MDB would stop consuming messages, despite
  there being applicable messages available for consumption on
  the queue.  In a system where messages continued to flow onto
  the source queue, this would result in a build up of messages.
  
  No messages were output into application server logs, and the
  Administration Console showed that the application/Listener
  Port were still running.
  
  Restarting the application (in the case where an Activation
  Specification was being used to drive the MDB), or restarting
  the Listener Port would result in messages once again being
  consumed from the queue.
  
  The frequency of the problem depended on the system load and
  message throughput.  Generally customers observed the issue
  after the system had been running for a few days.  In a test
  environment, the problem could be recreated within a few
  minutes of starting the application server.
  
  A work around for the problem is to disable Read Ahead.  This
  work around has no impact on the consuming application, other
  than a potential drop in performance in heavily loaded
  environments where the time taken to transfer message data from
  the queue manager to the client is the limiting factor.
  

Problem conclusion

• The WebSphere MQ classes for JMS have been modified to address
  the defect in the threading model, such that
  thread locking now prevents multiple threads from simultaneously
  attempting to update the key variables used to control the
  requests for data from the queue manager.
  
  ---------------------------------------------------------------
  The fix is targeted for delivery in the following PTFs:
  
  Version    Maintenance Level
  v7.0       7.0.1.6
  
  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:

  PM42810

Fix information

• Fixed component name

  WMQ AIX V7

• Fixed component ID

  5724H7221

Applicable component levels

• R700 PSY

     UP