Configuring the Application Server to Handle Cache Files and Replay Files

When an application server instance crashes, cache files and replay files are generated automatically. Over time, the size of these files can consume a large amount of disk space if there are recurring crashes in a domain. To minimize the buildup of cache files and replay files, you can modify the psappsrv.cfg file based on the following rules:

• When a crash occurs, the system creates a directory in the domain's LOGS directory.

• The dump file is saved in a directory within the domain's LOGS directory.

• The DumpMemoryImageAtCrash setting in the Trace section of the psappsrv.cfg file saves the memory image of the failed process in Microsoft Windows.

  This functionality is only available for Windows. If the value of DumpMemoryImageAtCrash is MINI, a miniature memory image (with a size less than or equal to 64K) is generated. If the value is FULL, then a full memory image is created. Depending on how much memory is consumed by the application, this full memory image can be quite large. The location of the memory image is the same as the replay file.

• If DumpManagerObjectsAtCrash is set to Y, then the application server instance:

  1. Generates the replay file.

  2. Dumps the customized objects being used by the current service request into the special cache directory.

     The cache directory resides in the same location as the replay file.

  3. If the value of DumpMemoryImageAtCrash is NONE and the platform is set to MS Windows, a miniature memory image is created.

• The settings for DumpManagerObjectsAtCrash and DumpMemoryImageAtCrash are dynamic.

  That is, the application server doesn't need to be restarted for these settings to be effective.

• There is no separate setting for generating the replay file.

  This file is generated as mentioned previously.

• Regardless of the setting in DumpManagerObjectsAtCrash, a summary report of objects in each managed type for which at least one object is loaded in memory is written to the dump file or application log file.

  The summary report resembles the following example:

  PDM Definitions: Total=36   Customized=0   In-Use=10  
  RDM Definitions: Total=53   Customized=52   In-Use=50  
  MDM Definitions: Total=1   Customized=0   In-Use=0  
  PCM Definitions: Total=199  Customized=0   In-Use=3  
  PGM Definitions: Total=1   Customized=0   In-Use=1  
  CRM Definitions: Total=67   Customized=0   In-Use=0  
  SSM Definitions: Total=1   Customized=0   In-Use=1  
  CLM Definitions: Total=1   Customized=0   In-Use=0  
  UPM Definitions: Total=1   Customized=0   In-Use=0  
  

  Total indicates the total number of in-memory definitions being used by the current service. Customized indicates how many of those objects are customized, and In-Use indicates how many of those objects were being used at the time of the crash.

If DumpManagerObjectsAtCrash is set to Y, the summary for each managed object type follows the list of configured objects that are being dumped as part of the crash information gathering. If a configured object is in use, its name is prefixed with an asterisk.

A sample report for a managed object type follows:

  RDM(PSOPTIONS/ENG)
  *RDM(PSTREEDEFNLABLS/ENG)
  RDM Definitions: Total=10   Customized=2   In-Use=1