1

I'm seeing a production server which is peaking at over 100 threads and with occasional queue length spikes of 20-30. The underlying hardware has plenty of excess capacity.

I'm wondering if the self-tuning thread pool has a default max number of threads so it isn't increasing beyond this limit.

rmp
  • 199
  • 2
  • 9
  • What type of requests is your server handling? If you have multiple applications on the server it would be good to define separate workmanagers for each app. This way WebLogic could tune the workmanagers specifically for the application or type of traffic being handled. – Jeff West Nov 02 '11 at 17:38
  • Thanks Jeff. It's a single-app but a legacy one that can't easily be partitioned using work managers. We sorted out the problem -- well at least identified it. This installation is integrated with Site Minder and configured for authorization checks. It appears that the local Site Minder agent is occasionally blocking dozen's of service threads for hundreds of ms at a time. Weblogic is correctly inferring that increasing the thread pool won't improve throughput. – rmp Nov 21 '11 at 14:13

1 Answers1

0

There isn't a hard maximum defined for the pool but the underlying algorithm attempts to detect the processes/cores available and increase the threads without overloading the server.

To evaluate the effectiveness of the self-tuning pool, you could enable debug logging and observe the values listed for the Increment Advisor. If your physical architecture isn't the typical commodity hardware, the default work manager is probably underestimating the available resources.

Musannif Zahir
  • 3,001
  • 1
  • 21
  • 31