PI21393: OUT OF MEMORY OR CONNECTION ISSUE WHEN DEPLOYING APPLICATIONS TO WORKLIGHT SERVER / WORKLIGHT CONSOLE

Download for IBM Worklight V6.2.0 Fix Pack 1

APAR status

• Closed as program error.

Error description

• When deploying large mobile applications to the Worklight
  server, either through the wladm tool or the Worklight Console,
  OutOfMemory errors may occur. In Websphere Application Server
  Liberty Profile, it may also cause a connection error with the
  following message:
  
  FWLSE3000E: FWLSE3012E: JMX configuration error. Unable to
  obtain MBeans. Reason: "Read timed out".
  
  When a couple of large mobile applications are deployed and the
  server is restarted, the restart may fail due to the same
  reasons.
  The connection error may also occur in Worklight Studio which
  uses an embedded instance of Websphere Application Server
  Liberty Profile.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * Administrators of a Worklight 6.2.0 server who deploy large  *
  * mobile applications                                          *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * The problem can occur in two situations:                     *
  * 1) When the server starts, an OutOfMemory error may occur.   *
  * The server log will then contain the message "FWLSE0081E:    *
  * Failed synchronizing application from the administration     *
  * service. This prevents the application from running." The    *
  * server is not functioning any more in this state.            *
  * 2) When a large application is deployed, its binary is       *
  * transferred to the runtime and may also cause an OutOfMemory *
  * error. The deployment will fail. The server may need to be   *
  * restarted, in order to recover from the error.               *
  *                                                              *
  * In Websphere Application Server (including Liberty profile), *
  * instead of the OutOfMemory error, a connection timeout may   *
  * occur.                                                       *
  * If the OutOfMemory error occurs during data base operations, *
  * the server log may also contain                              *
  * "java.sql.SQLNonTransientConnectionException: No current     *
  * connection" as a side effect of the OutOfMemory error.       *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * -                                                            *
  ****************************************************************
  The Worklight server consists of two parts: the Worklight
  Administration Services and one or more Worklight Runtimes. When
  deploying an application, the Worklight Administration Services
  sends the application binary to the corresponding runtime. When
  starting the server, the Worklight Administration Services sends
  all previously deployed application binaries to the
  corresponding runtimes. The transfer is done via JMX
  communication. Unfortunately, the application server needs a lot
  of memory when sending binary data through JMX. Measurements
  showed a peak memory requirement of a factor 10 of the size of
  the transferred binaries. During server start, the peak memory
  is a factor 10 of the sum of sizes of all applications.
  

Problem conclusion

• Previously, when the server started, all previously deployed
  applications were transferred from the Worklight Administration
  Services to the runtime in one large block.
  
  The implementation was changed in the following way: the meta
  data is still transferred through JMX, but the meta data size is
  small. The application binary is transferred through a
  specialized REST service using a One-Time-URL mechanism. The URL
  of the application binary is only valid as long as the
  deployment is performed. The new mechanism has a peak memory
  requirement that is now approximately only 3 times the size of
  the largest application. This is much lower than in the original
  implementation. During the server start, the applications are no
  longer sent as large block but are sent sequentially, so that
  the peak memory depends only on the largest application, not on
  the sum of application sizes. Additionally, on WebSphere
  Application Server Liberty Profile, a larger JMX read timeout is
  set in the configuration.
  

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

  WORKLIGHT ENTER

• Fixed component ID

  5725I4300

Applicable component levels

• R620 PSY

     UP