A threshold is a limit (high or low) placed on a specific monitored metric. When a limit is exceeded for more than a specified number of sampling intervals (its tolerance), that threshold is crossed.

For example, you could set a threshold of 5% maximum CPU time on system processes on all nodes, and give the threshold a tolerance of three. Then, if a node had more than 5% of its CPU time used for system processes for more than 3 consecutive sampling intervals, that threshold would be crossed.

You can set thresholds to notify you when they are crossed. The Threshold Notifications dialog box is the default method of notification and provides you with detailed information.

Caution

Executing resource-intensive commands when a threshold is crossed causes the system load to increase. The increased load can cause more frequent threshold crossings, and in some cases, the threshold crossings are due solely to command execution. This can result in an excessive and continually growing system load.

To avoid this situation, increase the tolerance for the expression being monitored. The command will not execute until the threshold is crossed the number of times specified by the tolerance level.

Some other examples of thresholds:

• A node's I/O Queue exceeds a dozen processes for more than 10 consecutive sampling intervals

• A node's Disk Transfers exceed 25/second for more than 5 consecutive sampling intervals

• A node's Total Bad IP Packets exceed zero in any sampling interval

When a threshold is crossed, the following occurs:

• The event is logged (written in the Performance Manager log file: /var/opt/pm/log/pmgr_gui.log).

• A command (if specified) is run. Performance Manager has a number of comm ands built in, but it is also extensible. You or your system administrator can c reate your own commands. This command can do anything from sending you mail about the problem, to taking steps to fix the problem.

Setting Thresholds

Threshold Sessions

Threshold Environment Variables