Windows Server 2003 System Monitor Tips for Exchange Server Administrators
Topic Last Modified: 2006-02-09
By Nino Bilic
System Monitor, which was formerly known as Performance Monitor (Perfmon), contains features that can save time when setting up to monitor servers running Microsoft® Exchange. This article covers some of the time-saving features from previous versions and new features introduced in the Microsoft Windows Server™ 2003 and Microsoft Windows® XP operating systems. This article does not cover searching for information in the logs.
This article provides only an overview of the setup of the System Monitor snap-in. For more information about setting up System Monitor, see the For More Information section.
Windows NT® Server 4.0 Performance Monitor (Perfmon) allowed you to add all objects to the monitor at once. This feature was not available in the Windows 2000 Server operating system, although a workaround existed by using the Perfmon4.exe file from the Windows Server 2000 Resource Kit. The Windows Server 2003 operating system offers the feature. In Windows Server 2003, to set up a log, click the Add Objects button.
On this screen, you choose the objects. Any object that you add from this screen will be added with all counters that go under it. You can select multiple objects by using the SHIFT and CTRL keys.
Alerts are a useful feature that lets you set up System Monitor to check when certain counts reach a predefined value, and an action that you specified is executed. This is useful, for example, if you want to start monitoring the server after the CPU utilization goes over 90 percent.To set up alerts after you start System Monitor
Double-click Performance Logs and Alerts, and then click Alerts. Note that any existing alerts are listed in the details pane. A green icon indicates that the logs are running, and a red icon indicates alerts that have stopped.
Right-click a blank area in the details pane, and then click New Alert Settings.
In the Name box, type the name of the trace log you want to create, and then click OK.
To define a comment for your alert, add counters, add alert thresholds, and add the sample interval, use the General tab. The following figure shows an alert set to a value where CPU utilization goes over 90 percent. CPU utilization will be checked every 5 seconds, which is a default value. At this point, this alert is not being logged.
To define actions that should occur when the counter data triggers an alert, use the Action tab. In the following figure, the alert is set to start the performance log called CPU, MEMORY, DISC. This is an example of a custom performance log created earlier.
In the preceding figure, notice that one of the options is to Run the Program when the alert is triggered. An example where this feature can be useful is if you want to run USERDUMP or another tool when CPU utilization spikes to 100 percent.
To define when the service should begin checking for what you have set up in this alert, use the Schedule tab. You can set up both the starting and ending points for alerts.
After you find what you are looking for in the performance log, you can save the essential information as an HTML page for viewing on other computers. To do this, right-click the graph window and choose the Save As option.
This allows you to save the current view as an HTML page. You can now send this page to a different computer to be viewed. To open this file properly, you need to have the actual performance log that you received your data from, and that log has to be saved in exactly the same location and with the same name as you had when you saved the HTML page. This HTML page works through a Microsoft ActiveX® control to display the information from the actual log file. The location and name of the Perfmon log file can be viewed and modified in the HTML page. For example, if you open the page in Notepad, you can search for LogFileName001, if you are viewing only one log file. You can modify the path and name of the log file if needed.
After the location matches, you can open the HTML page, and it will appear in your browser. Note that you can add performance counters as if System Monitor was actually running.
System Monitor by default provides maximum, minimum, and average values of sampled data represented by the vertical lines in logs with more than 100 data points. This can sometimes lead to graphs being hard to read, because vertical lines are displayed. The following figure shows the clutter of the vertical lines.
To delete the vertical lines, you can change a registry entry on computers running Windows Server 2003 or Windows XP, to improve the readability of graphs that contain a large amount of information. For more information, see Microsoft Knowledge Base article 283110, "Vertical lines are displayed in the Sysmon tool that obscure the graph view." The following figure displays an example of the same graph with the registry key change implemented.
Windows Server 2003 and Windows XP provide additional logging options. For example, you can now log into an SQL database. You can also log into a circular file, which means that you can predefine the log file size. Then, as monitoring is performed, if the file is filled up, it starts to overwrite from the beginning of the file again. If you monitor the server for a long time period, you can leave System Monitor running for days or weeks, and then stop it after your problem recurs.
To set up circular logging, go to the Log Files tab.
For additional information, see the following Microsoft Knowledge Base articles, Exchange Server resources, and Web sites: