本文件已封存並已停止維護。

Practices to Avoid

更新日期: 2010年4月

適用於: System Center Operations Manager 2007

Focus on the easy–to-monitor instead of the required

We recognize the natural tendency to start a health model design for an application by analyzing the monitoring information made available by the application through events and other easily obtained information. However, this approach assumes that the application is proactively identifying all issues and may leave a significant potential for an application issue to go undetected. If an application has a problem that does not generate an event or other information, then the issue will require a synthetic transaction or other way of identification.

One consistent strategy is to analyze the different features and potential issues for each classe defined in the service model. A process such as Failure Mode Analysis can help in this analysis. As soon as the monitoring requirements are identified, appropriate monitors and rules can be defined to address those requirements. This strategy encourages development of synthetic transactions and other proactive monitoring techniques in order to obtain information that the application does not make available on its own.

Running monitors and rules too frequently

Monitors and rules that use a script or WMI query that run too frequently can result in excess overhead on the agent. The overhead will vary, depending on the script or query and the number of other scripts and queries in the management pack. A general guideline is not to run a script or query at any interval more frequently than 5 minutes. Even with this guidance, the script or query should not be run any more frequently than is required for needed monitoring.

For collection rules, too frequent collection can result in significantly increased storage requirements and return minimal value. Although a single performance value is fairly small, storage requirements quickly accumulate for collection targeted at multiple instances across multiple agents. We recommend that you never have a collection frequency less than 5 minutes for counters with significant volatility. A frequency of 15 minutes is effective for counters with less variance. It also a best practice to use optimized collection for any counter that is not expected to consistently have a significantly different value between consecutive samples.

Multiple scripts running at the same time

If several monitors and rules that are running scripts are defined to run on the same interval, then there is a potential for them to consistently run at the same time. Every time that the interval arrives, all the scripts run, and a noticeable affect can be observed on the agent processor. One method to avoid this condition is to use a different synchronization time for each monitor or rule to make sure that they run at different times.

Targeting classes with multiple instances

As is discussed in the Targets section of this guide, a workflow targeted at a class that has multiple instances on an agent will have multiple copies of the workflow running on the same agent. Such workflows will typically require additional criteria by using a $Target variable to uniquely identify the particular instance. Failure to provide unique criteria can result in multiple alerts being created for a single event, or all instances of a particular class entering an error state instead of a single instance.

另請參閱

顯示: