WebLogic Server always taking a quick dip in service threads

we have a HP BAC probe attached on one of our WebLogic servers and we have noticed that the server is of开发者_JAVA百科ten always taking a sudden and deep "dip" in the number of service threads available.

Does anyone has any encounter or anything to share how i may track this? Currently my thread dump capturing process is unable to capture this because often it is too late. Or is there any continuous thread dump capturing process i may consider?

You can define a JMX Notification via the Admin Console

which will be picked up by WebLogic Diagnostics Framework (WLDF)

This creates a diagnostic snapshot which you can analyze later.

See how to create notification for watches

http://download.oracle.com/docs/cd/E21764_01/apirefs.1111/e13952/taskhelp/diagnostics/CreateNotificationsForWatches.html

and specifically in admin console

http://download.oracle.com/docs/cd/E21764_01/apirefs.1111/e13952/taskhelp/diagnostics/CreateWatchesForADiagnosticModule.html

Select a watch type from the Watch Type list:

•Select Collected Metrics to set a watch based on metrics collected from MBean attributes. •Select Server Log to set a watch based on data written to server logs. •Select Event Data to set a watch based on data generated from a specified instrumentation event.