Monitoring Alerts

AnalyticOps allows you to monitor the performance of models and model versions against specific key metrics. In addition to metric gathering and tracking, the model monitoring involves alerting based on these metric values. The purpose of alerting is to detect important deviations in the data used to train a model versus the data currently being used in scoring. This detection provides the ability to set alert thresholds on key metrics related to the deviation of data. Using AnalyticOps UI, you can ability to define alerts by setting thresholds on performance metrics.

AnalyticOps is using Prometheus alert manager to provide the support of alerting in UI. Prometheus is a very easy-to-use yet powerful tool that allows you to make the alerting rules very sophisticated and keep the alerting rule complexity encapsulated for the end-user.

The Alerts module lets you configure alert rules for a selected project and view the details of all active alerts on a single screen. This chapter covers the following details:

For details of enabling and configuring alerts, see Model Monitoring Alerts.

View List of Alerts

To view alerts added to a project:

  1. Click the Alerts from the Navigation bar of a selected project.
    The list of active alerts for the project displays in the Work area and the Alerts option highlights in the Navigation bar. The notification icon shows the number of active alerts in the project.

    The list of Alerts displays the following details for each alert.

    Property

    Description

    Alert Name

    Specifies the name of the alert.

    Model Version ID

    Specifies the model version ID for which the alert is generated.

    Starts At

    Specifies the date and time when the alert is generated.

    Status

    Specifies the status of alert as Active or Suppressed. For details of status, see Snooze and Unsnooze Alerts.

    View Details

    Lets you view the details of a selected alert. For more information, see View Alert Details.

Search Alerts List

You can search an alert in the list by its name using the Search bar shown on the top of the list.

To search an alert in the list:

  1. Enter the name of the alert in the Search bar.
    The list filters itself and displays only the related alerts.

To remove search and go back to the complete alerts list:

  1. Click the Close icon on the right side of the Search bar.
    The search results remove and the list displays all the active alerts.

View Alert Details

You can view the details of an alert including all the generated instances for different metrics along with description and expression for each metric. You can also go to the Prometheus graphs view to see the details of each metric in a visual form.

To view details of an alert:

  1. In the Alerts list, click the View Details icon for an alert.

    The Alert details page display.

    The Alert details page lists all the instances of the alert generated for different metrics defined in the alert configuration. For details of alert configuration, see Configure Alerts.

    Each alert instance displays the following details.

    Property

    Description

    Metric

    Specifies the name of the metric for which the alert is generated.

    Starts At

    Specifies the date and time when the alert is generated.

    Status

    Specifies the status of alert as Active or Suppressed. For details of status, see Snooze and Unsnooze Alerts.

    Selecting an alert instance row in the grid expands the right side pane.

    For each selected instance, the below details display in the right side pane:

    Property

    Description

    Description

    Specifies the description of the alert instance as defined in the alert configuration. For details of alert configuration, see Configure Alerts.

    Expression

    Displays the expression defined for the metric in the alert configuration. For details of alert configuration, see Configure Alerts.

    View Data Graph for Alert

    Lets you move to the Prometheus workspace where you can see the visual graph of the metric value.

Snooze and Unsnooze Alerts

Snoozing is a way to temporarily stops notifications of an active alert for a certain amount of time. You can snooze an alert via the Alerts detail page.

When an alert is snoozed, the alert’s status becomes suppressed. All notifications related to that alerts are stopped from that moment till the end of the specified time. When the snooze ends, the alert’s status becomes active again. AnalyticOps allows you to unsnooze an alert explicitly and make it active before the snooze time ends.

Snooze Alert

You can snooze a selected alert or snooze all alerts at once using the action buttons on the top.

To snooze an alert:

  1. Select the alert from the Alerts list.
    The Snooze For button enables.

  2. Click the Snooze For button.
    The Snooze For menu displays. You can snooze an alert for 30 mins, 1 hour, 2 hours, 4 hours or a custom time period.

    https://docs.tdaoa.com/images/ug_alerts_snooze_for.png
  3. Select an option from the menu or click Custom to specify a time period.
    For the Custom option, the Snooze Alert For dialog displays.

    https://docs.tdaoa.com/images/ug_alerts_anooze_for.png
  4. Specify the time in mins, hours or days to snooze the alert. Click Snooze.
    A confirmation dialog displays. Comment: the dialog image will be added when implemented.

  5. Click Snooze. The alert snoozes for the specified time and its status changes to Suppressed.

To snooze all alerts:

  1. Click the Snooze All For button.
    The Snooze All For menu displays. You can snooze all alerts for 30 mins, 1 hour, 2 hours, 4 hours or a custom time period.

    https://docs.tdaoa.com/images/ug_alerts_snooze_all.png
  2. Select an option from the menu or click Custom to specify a time period.
    For the Custom option, the Snooze Alert For dialog displays.

    https://docs.tdaoa.com/images/ug_alerts_snooze_for_all.png
  3. Specify the time in mins, hours or days to snooze all alerts. Click Snooze.
    A confirmation dialog displays. Comment: the dialog image will be added when implemented.

  4. Click Snooze.
    All alerts snooze for the specified time and their status change to Suppressed.

Unsnooze Alert

Comment: The section will be added when available in the build.