Guidelines for Monitoring Exchange & Other Related Services
Here are some general guidelines to monitor Exchange Server 2003. Since each environment is different, you should consider these guidelines as a good starting point, monitor your Exchange Server 2003, and then configure the parameters accordingly. To configure Monitoring, locate the server, Properties, and go to the Monitoring tab.
Virtual Memory
Should not fall below 25% free.
CPU
Should not exceed 80% over 5 minutes.
Free disk space
Configure the monitor to issue a warning when you have less than 250MB available and issue a critical alert when the drive has less than 50MB of free space.
X.400 queue growth
Should be empty for any environment that uses SMTP to connect to other messaging systems.
Guidelines for Monitoring Cluster Resources
Use System Monitor to identify memory fragmentation for each node in the cluster by monitoring the following counters.
MSExchangeIS\VM Largest Block Size
Should not fall below 32MB. Exchange will log a warning in the event log (Event ID=9582) if this falls below 16MB.
MSExchangeIS\VM Total 16MB Free Blocks
Should not drop below 3. If it does, restart all the services on the node.
MSExchangeIS\VM Total Large Free Block Bytes
Should not fall below 32MB. If it does, restart all the services on the node (or restart the server) and then fail back the Exchange Virtual Servers.
Monitoring Exchange Store
Free disk space
Must be equal to or greater than 110% of the size of the largest database.
Baseline Counters
The following counters should be monitored over time. The required values are only realistic if they are observed over time. For example, you may notice that % Processor Time value peaks to over 80% temporarily on occasion. We are only interested if these values are sustained over a period of time. For example, if you are using Performance Logs and Alerts to collect data for the recommended counters, configure it to collect data at regular intervals, such as every 10-15 minutes.
You may also want to collect data every week during specific hours, such as Monday-Friday between 8:00am-10:00am for several weeks to obtain a baseline of network performance during peak hours. These are just some suggestions. Every network is different and your baseline may be very different than other networks. Use the following counters as a guideline.
