Automatic Application Hang Detection

 

• Automatic Application Hang Detection is an InSight featurewhich, when enabled, attempts to detect certain application hang conditions.  Once detected, a ‘possible hang’ snapshot dump is generated.  The ‘possible hang’ snapshot dump must be confirmed or unconfirmed.  If confirmed, it is renamed to an actual snapshot dump.  If unconfirmed the temporary ‘ possbilehang’ snapshot dump is deleted.
• Hang detection is based on monitoring calls from the application to the system.  Normally, the application is consistently making calls to the operating system on a very regular basis.  Even when idle and waiting for input, nearly all applications wake up at least once per second and perform some work before going back into a wait.   The hang detection logic simply says that, if no system calls are made by the application within some amount of time (3 secsby default), there is potentially a hang.  If the hang lasts 10 times this detection limit, an actual hang is very likely.
• The feature needed to understand a difference between ‘possible’ and ‘likely’ because operators can be quick to restart the terminal if they think itshung by powering it off then on.
• Application hang detection was initially designed to detect hangs caused by hardware or hardware driver issues.  Applications were making calls to the printer and hanging indefinitely.  Snapshot dump data was used to understand specifically what happened just prior to  thehang.
• Confirmation occurs if the terminal is restarted and, during the next IPL, the terminal agent finds an existing unconfirmed snapshot dump file.
• Confirmation also occurs if the application remains hung for 10 timeout periods.   Unconfirmationhappens if the application resumes making calls before 10 timeout periods.
• The setting of the timeout parameter is, by its nature, a balancing act between the possibility of generating meaningless snapshot dumps and having a good chance of having the data needed to resolve whatever the issue is.
• The snapshot dump feature has a related and kind-of built-in problem automatic problem detection capability.  This is because sequential clear keys are used to force a snapshot dump and because clear is often hit repeatedly be operators when something is wrong.
• The option is enabled via CDI parameters and is disabled by default.  
• Required CDI keyword : tagentApplHangDetectList.  Note that full documentation for cdikeywords is available in c:\qsa\cdi\insight.all as part of controller installation.
• Optional CDI keyword :   tagentApplHangDetectTime.   ontroller installation.
• Related CDI keyword: tagentSnapDumpClearKeys