Use baseline alert conditions to define thresholds that adjust to your data's behavior. This feature is available for New Relic Alerts users who have apps monitored by New Relic APM, New Relic Browser, or use NRQL queries. Baseline alerting is useful to create alert conditions that:
- Dynamically adjust to changing data and trends, such as weekly or seasonal patterns.
- Establish adjustable values for new applications with as-yet-unknown patterns.
- Offer more hands-off thresholds that only notify you when data is identified as abnormal.
Understanding how baseline alerting works will help you create effective conditions for them.
- You choose the metric data for which you want to create an alert condition.
- New Relic Alerts uses the past values of that data to dynamically predict the data's upcoming behavior. This ongoing prediction is called a baseline and appears as a dotted black line in the UI.
- You use the slider bar to adjust the thresholds (the gray bands around the baseline in the chart) that control the alert condition's sensitivity to trigger more or fewer violations.
- You can set both a Critical threshold (the outer, light gray band in the chart) and an optional Warning threshold (the inner, dark gray band).
- When your data escapes the predicted "normal" behavior, you will receive an alert notification.
Create baseline alert policy
All entities in a single alert condition share the same threshold settings. If you want to set different threshold settings for your APM or Browser entities, create multiple conditions for your policy.
To assign a baseline alert condition to a policy and one or more entities in New Relic Alerts:
- Follow the basic workflow process to set up an alert policy.
- When creating a condition, select APM > Application metric baseline, Browser > Metric baseline, or NRQL as the product target, with one or more APM or Browser apps as the target entities.
- Adjust and set the baseline thresholds.
- Follow the UI prompts to complete and save the alert condition.
Adjust baseline alert thresholds
Creating an effective baseline condition threshold is a back-and-forth process of:
- Adjusting threshold settings
- Reviewing the violations those settings would theoretically create in your data
Here's an example of how to create baseline alert thresholds for your APM > Application metric baseline or Browser > Metric baseline condition:
- From the When any target application dropdown, select the metric you want to monitor.
- Select the time range (2 days or 7 days) for the preview chart using the dropdown selector above the chart.
- Set the Critical (red) threshold, which appears as a light gray band in the preview chart. Use the frequency and time settings above the slider bar to select what triggers a violation. Use the slider to adjust the threshold.
- Optional: Follow the same procedure to set the Warning (yellow) threshold, which is the darker gray band in the chart.
- Optional: If the alert condition applies to multiple apps, select your choice from the dropdown above the chart to see how your thresholds apply to different apps.
- When finished setting your thresholds, save your baseline alert condition.
The 2-day and 7-day preview charts are not the time period used to compute the baseline. They are simply a time range for the displayed data. The baseline is computed from up to several weeks of data, if available.
The algorithm for baseline conditions in New Relic Alerts is mathematically complex. Some of the major rules governing its predictive abilities include:
|Data trait||Baseline rules|
|Consistency of data||
For metric values that remain in a consistent range or that trend slowly and steadily, their more predictable behavior means that their thresholds will become tighter around the baseline. Data that is more varied and unpredictable will have looser (wider) thresholds.
For shorter-than-one-week cyclical fluctuations (such as weekly Wednesday 1 pm deployments or nightly reports), the baseline algorithm looks for these cyclical fluctuations and attempts to adjust to them.
|Age of data||
For data that has only existed for a short time, the baseline will likely fluctuate a good deal and not be very accurate. This is because there is not yet enough data to determine its usual values and behavior. The more history the data has, the more accurate the baseline and thresholds will become.