There are high end server monitoring packages out there that will track almost any type of perfomance counter that you can think of and most of them cost a pretty penny and have some learning curve. I’m not ready to commit to any of those packages at work yet but I still wanted to know when my servers (web servers in particular) started going haywire. Here is a quick, down and dirty way to monitor your Windows 2000 servers(probably works on Windows Server 2003 too but I haven’t tested it – note: I know that this method doesn’t scale well).

This requires 2 little apps. One is a command shell script and the other is a vbs script (in retrospect, I could have made a console C# app to do the same thing, but remember, I wanted, cheap, down and dirty. All of this code can be done in notepad).

All this code does it call the vbs script and pass command line parameters to it. I save this file in c:scripts and call it ActivatePerformanceAlert.cmd:

• Login to your server at the console
• Start the performance monitor
• Expand Performance Logs and Alerts in the MMC
• Right click on Alerts and select New Alert Settings
• Give the new alert a meaningful name like “ASP Requests Queued”, “Web Service Anonymous Connections”, etc. In this instance, I will make one for “Web Service Anonymous Connections”.
• Click the Add button and select a counter from the Performance Object. I’ll select “Web Service”, the site that I want to monitor (this only exists if you have multiple sites on the web server), and then select “Current Anonymous Users”.
• Usually you want to know when Anonymous users have grown past an expected limit so select “Over” in the drop down menu listed “Alert when the value is”
• In the “Limit” box enter the measured value that you want a notification if that number is exceeded. For example, if you want to be notified when your Anonymous Web Users are over 90, place the number 90 in the box
• Select a sample rate interval. You can select in seconds, minutes, hours, days. Select a realistic limit. Too low of a number (2 seconds per se) and your inbox will be flooded with these emails as long as the alert persists (it also may degrade server performance further compounding the issue at hand). For my example, I will select 5 minutes.
• Click on the Action tab
• Select the following, “Log an entry in the application event log”. I do this just in case I want historical information and then select “Run this program”.
• Enter in c:scriptsActivatePerformanceAlert.cmd in the box
• Click the command line arguments button and remove (and this is important) the check mark from “Single Argument String”
• All other boxes are checked except “Text Message”
• Click ok twice and you will be back at the alerts menu
• If the alert has a red icon next to it, right click on the alert and select “Start”