This article was previously published under Q150934
This article has been archived. It is offered "as is" and will no longer be updated.
For a Microsoft Windows 2000 version of this article, see 248345.
This article explains how to create a Performance Monitor log so thatengineers can use it to troubleshoot performance problems on acomputer running Windows NT.
Creating the Performance Monitor log:
On the View menu, click Log.
On the Edit menu, click Add To Log.
In the Computer field, specify the name of the local computer or a remote computer you want to get a log of. In Objects, add the counters that you want Performance Monitor to log. These counters will change depending on the issue that you are troubleshooting.
Click to highlight the objects that you wish to monitor and then click Add.
Under Options click Log.
Under File Name, name the log.
In Update Time, set the interval according to how long you want to run the Performance Monitor log. If you want to run the log for an extended period of time, you will want to set this update time to an interval much higher than 15 seconds or the log will be very large. If you are going to run the log for only an hour, then 15 seconds will be fine.
Click Start Log. The icon will change to a Stop Log icon once the log is started.
When you are ready to stop monitoring the selected objects and examine your log file, perform the following steps:
Under Options click Log and then click Stop Log.
In the Options menu click Data from, click Log File, and then click to select the radio button beside the perfmon.log field and locate your log file.
NOTE: You cannot open a log file simply by clicking File and then clicking Open.
After you open the log file, add the objects and counters that the log has monitored. Do this for each view needed, otherwise the log file's data will not be available.
If you are troubleshooting a performance issue or an issue that looks like a memory leak, the objects that Performance Monitor should log include but are not limited to the following items.
Memory resource issues:
Cache Memory Objects Paging file Process Processor System Terminal Services (if a Terminal Server)
For all other resource issues, add additional counters:
Logical disk NBT Connections Network interface Physical disk Redirector Server Server work queues Thread (do NOT capture if a terminal server) All Terminal Server counters (if a Terminal Server) All Protocol counters bound to network adapters
Disk counters require that you run diskperf -y from the command console and then restart the computer.
We usually do not capture threads on a terminal server because this counter can consume 80% of a data file.
You have several options if the computer that is running Perfmon is restarted or goes down while the log is going. You can start a new log, or, if you specify the name of an existing log file, the new data is appended to the end of the log file. It is important to let the engineer know the log was restarted. However, if you are monitoring remotely and the target computer goes down, this does not apply.
If there are processes that start after the log was started, they will not show up in the beginning of the log. To view different periods in the log, click Edit, click Time Window, and then use the sliding scale.
If the user logs off, the performance monitor log will stop. You can set up Performance logging as a service, but it is easiest to run Performance Monitor remotely from another Windows NT system.
For additional information about troubleshooting performance issues, click the article number below to view the article in the Microsoft Knowledge Base:
175658 Gathering Information for Troubleshooting Performance
Microsoft Windows NT Workstation 3.5, Microsoft Windows NT Workstation 3.51, Microsoft Windows NT Workstation 4.0 Developer Edition, Microsoft Windows NT Server 3.51, Microsoft Windows NT Server 4.0 Standard Edition