Create a monitor manually
There are a number of configuration options that you can use to customize your monitor. You access these options from the Add Monitor flow and you can enter this flow two ways:
- On the Monitors List, select Add Monitor.
- From Explorer, when you click + Add you get a short form that lets you add a monitor with no configuration options. At the top of that form you'll see a blue link that says Manually configure monitor. This takes you into the Add Monitor flow with the metric name pre-filled, and you'll be able to perform all monitor configuration.
Once you enter the Add Monitor flow you will see the screen below. The top of the screen shows three tabs, Define, Train, and Preview.
Define
To create your monitor, in the Define pane:
- Select the monitor's metric or metric slice.
- Name the monitor.
- Select the Symptom to detect. For definitions of the options, see Monitor configuration Define settings.
- If you're creating an anomaly monitor, you can enable Account for trend to allow the monitor to include trends in threshold calculations, and enable Advanced forecasting to have your monitor create metric forecasts.
- By default, a monitor starts generating incidents as of the moment it goes live. If you want the monitor to identify incidents in historical data, set the Backfill Incidents Starting date to have the monitor look for incidents in metric data starting from that date.
- Optionally, choose an alerting channel and muting schedule. See Schedules for details on creating a schedule. Select Mute notifications to mute notifications just for the monitor.
- When you're done, select Next at the top right corner.
Train
On the Train pane, provide values for any relevant settings. The available settings depend on which Symptom to detect you chose in the preceding step. For definitions of the options, see Monitor configuration Train settings.
Manual threshold monitors
- If you chose Manual threshold, set your thresholds here. You can set one threshold, or both upper and lower thresholds, and can use either Values or Percent Change as thresholds. If you choose Percent Change, you'll also need to set the Period— the amount of time to compare against other periods when calculating the percent change. Click the button to the left of Upper Threshold Value or Lower Threshold Value to enable or disable that threshold.
- To adjust the detection settings, click the pencil to the right of Detection Settings and then use the options in the modal that opens. Different settings apply to different monitor types— see Common monitor settings and Manual-threshold monitor settings for details.
- Click Save and train at the top right corner.
- After your monitor trains, click Offline to make the monitor live. The monitor will begin logging new incidents.
Your monitor is now running. View the metric in Explorer and open the Monitors menu to see your new monitor in the dropdown. If it detects metric values outside of your thresholds, it generates an incident.
Anomaly detection monitors
- Add a training period by clicking + to the right of Training. A calendar opens where you can select a start date and an end date. An anomaly monitor needs at least two weeks of data to train. You can add multiple training periods. Training periods should only cover time ranges where your metric has known-good behavior. To identify these periods, review your metric's behavior in Explorer or in preview.
- To adjust the detection settings, click the pencil to the right of Detection Settings and then use the options in the modal that opens. Different settings apply to different monitor types— see Common monitor settings and Anomaly monitor settings for details.
- Click Save and train to train your monitor.
- After your monitor trains, click Offline to make the monitor live. The monitor will begin logging new incidents. To also check historical data for incidents, click ... to the right of Offline, enter a Backfill Incidents Starting date, and then click Make live to make your monitor live. The monitor will check historical data for incidents as well as logging new incidents. All incidents, historical or otherwise, can be opened from the metric chart in Explorer.
After training has completed and your monitor is live, it starts monitoring. To see it in Explorer, on the metric's chart open the Monitors menu: your new monitor appears in the dropdown. If Lightup's anomaly detection engine detects the metric is outside of bounds, an incident will be generated.
Updated about 1 month ago