WAS traditional Hung Thread Detection Recipe

1. Decide on your hung thread detection interval and threshold. In general, the interval is performant even at 1 second. In general, the threshold should be set to your maximum expected response time plus 20%.
2. WAS Administrative Console } Servers } Server Types } Websphere application servers } $SERVER } Server Infrastructure } Administration } Custom Properties
3. Click New...
   1. Name = com.ibm.websphere.threadmonitor.interval
   2. Value = $SECONDS_INTERVAL
   3. Click OK
4. Click New...
   1. Name = com.ibm.websphere.threadmonitor.threshold
   2. Value = $SECONDS_THRESHOLD
   3. Click OK
5. Save configuration
6. Synchronize nodes
7. Restart
8. Monitor SystemOut*log for WSVR0605W messages and review the stack to understand what the threads were doing at the time the threshold was exceeded. Ensure you review garbage collection and operating system statistics as well.

Notes:

1. If many requests exceed the threshold at the same time, this can introduce its own overhead in acquiring and printing the stack traces. By default, com.ibm.websphere.threadmonitor.false.alarm.threshold is the threshold for an exponential backoff that occurs every 100 breaches, but if there is a concern about such a potential overhead, then reduce com.ibm.websphere.threadmonitor.false.alarm.threshold.

Previous Section (WAS traditional Diagnostic Trace from Startup Recipe) | Next Section (WAS traditional HTTP Access Log Recipe) | Back to Table of Contents