The goal of an alert is to let you know when there is a problem you need to attend to. The challenge is to set alert condition thresholds so that you are not alerted too much or too little. In general, the steps are as follows:

  1. Establish a baseline - Start with a basic system-defined alert profile applied and run APM for a long enough period (at least a couple of days) that you can see typical performance patterns appearing in the performance charts. The following system-defined alert profiles are a good starting point depending on what you are monitoring:
  2. Adjust alert profile if necessary - Depending on the baseline performance observed, you may want to customize an alert profile such that alert condition thresholds are violated when they are outside of the normal operating range and at a point where you would like to be alerted.
  3. Adjust if too many alerts - If you receive too may alerts, options include:
    • If possible, fix the issue causing the violation.
    • Increase an alert threshold.
    • Increase the length of time the metric must be in violation before an alert is triggered.
  4. Adjust if too few alerts - If you are not receiving alerts but users are reporting issues, options include:
    • Decrease an alert threshold.
    • Decrease the length of time the metric must be in violation before an alert is triggered.

However, be aware that it is possible that increasing an alert threshold can also increase the number of alerts generated. For example, consider the following where setting the threshold to 10 leads to 3 violations and increasing it to 20 leads to 7 violations. Note that increasing the time to clear would reduce the number of violations in both cases.

Graph showing a performance metric over time with thresholds set at 10 and 20. The graph is drawn such that there are three violations at 10 and 7 violations at 20.

Related Topics: