Setting Alerts to Monitor System Events
Applies To: Windows Server 2003, Windows Server 2003 R2, Windows Server 2003 with SP1, Windows Server 2003 with SP2
An alert is a notification that is generated as a result of an event. You can use alerts for ongoing system monitoring; for example, you can set an alert to notify you if CPU usage reaches 100 percent.
To monitor your MIIS 2003 implementation, it is recommended that you use alerts for the following events:
Errors affecting service. Whenever possible, ensure that each known critical error has an alert associated with it. If you cannot associate the error with an alert, consider writing a query to test for the error state, and then run the query on an ongoing basis, configuring an alert to be generated when the query finds the error.
Deadlocks. Ensure that when the value of any SQLServer:Locks\Number of Deadlocks counter exceeds a given threshold, it generates an alert. If an application does not typically cause deadlocks, set this threshold to 1 or 2. If you run concurrent management agents, set this threshold from 10 to 20.
CPU usage. If your application typically freezes or degrades at a certain level of CPU usage, set an alert to be generated when the CPU usage exceeds that level.
Disk usage. If your application freezes or degrades at a certain disk usage level (or queue level), set an alert to be generated when the disk usage exceeds that level.
Table scans. Ensure that your application database generates an alert if excessive table scans occur. First, establish the baseline value for excessive table scans in your environment. To determine your baseline, monitor the SQLServer:Access Methods\Full Scans/sec counter for a period of time; then set an alert. If you have many small tables that do not require indexes, a large number of acceptable table scans can occur before the number becomes excessive.
For information about setting and managing alerts, see “Performance Logs and Alerts” in Help and Support Center for Windows Server 2003.