Configuring Worker Process Pinging
Updated: August 22, 2005
Applies To: Windows Server 2003, Windows Server 2003 with SP1
Application pool health detection is the ability of the WWW service to detect that a worker process assigned to the application pool is in an unhealthy state by periodically pinging, or sending a ping message to, the worker process.
Optimizing Ping Intervals
You can specify a value for the ping interval by examining the potential amount of system resources that can be consumed by unhealthy worker processes over the interval (the default is 30 seconds). Then determine how long you can wait for system resources that unhealthy worker processes are using to be freed so that they are available to other worker processes.
If a worker process becomes unhealthy, the client does not receive a response until another worker process is started to replace the unhealthy worker process. This leaves the unhealthy worker process active without releasing the system resources used by the process. Over a period of time, the amount of system resources used by an unhealthy worker process can accumulate and affect the overall performance of the Web server.
For example, you might have an application that causes worker processes to become unstable. Over a five-minute period of time, the amount of system resources consumed by these unstable worker processes can consume Web server resources and affect the performance of other applications that are on the same Web server. You can configure IIS 6.0 to ping worker processes in that application pool to check the health of the worker processes every 30 seconds, which ensures that the unhealthy worker processes are shut down before system resources are consumed.
Set ping intervals to avoid false failure conditions caused when healthy worker processes do not respond within the specified response time. You can check this by testing to ensure that the WWW service does not shut down a worker process that is still running, but is too busy to return the ping in the specified response time. If your tests indicate that your worker processes are shutting down because of false failure conditions, increase the ping response time, or alter the load on the worker process.
By default, worker process pinging is enabled in IIS 6.0. The default ping interval is 30 seconds, and the default ping response time is 60 seconds.
For information about how to configure worker process pinging by using IIS Manager, see Configure Application Pool Health. For information about how to configure worker process pinging by using a command-line procedure or scripts, see the PingingEnabled Metabase Property, PingInterval Metabase Property, and PingResponseTime Metabase Property.