Menu
DocumentationGrafana CloudAlerts and IRMMachine LearningSiftSift panel
Grafana Cloud

Sift panel

The Sift panel can be used to show Sift investigations and results directly on your dashboards.

Note

The Sift panel is only available on Grafana Cloud instances. It also requires the Grafana Machine Learning plugin to have been initialized.

Refer to Enable Grafana Machine Learning for instructions on how to do this.

To get started, add a new panel to your dashboard and select Sift from the panel type selector in the top right. You don’t need to add a query.

From the panel editor you can select what to display in the panel: investigations or analysis results.

Investigations

The Investigations display mode lists investigations run during the dashboard’s current time range.

Each investigation is displayed as a row along with its creation time and labels. An icon to the left indicates whether the investigation found interesting results.

Analysis Results

The Analysis Results display mode lists the individual analyses of investigations run during the dashboard’s current time range.

Each analysis is displayed as a row along with the name of the investigation and any labels associated with the investigation. An icon to the left indicates whether the investigation found interesting results.

In this mode you have two further options:

  • Checks: filters the results to only specific checks.
  • Only interesting checks: filters the results to only show analyses with interesting results.

Labels

Both modes allow you to add labels which can be used to filter the list of investigations or analyses. The labels are combined using ‘AND’ logic.

For example, if you only wished to show investigations where the labels matched cluster="prod" and namespace="default" you could add two label filters.

Template variables

The values of these labels can also refer to template variables. The values of any multi-valued template variables are combined using ‘OR’ logic.

For example, you may have a template variable named cluster with selected options dev and prod, and a template variable named namespace with selected options default and app.

In this case you could add two label filters to the panel: one with name cluster and value $cluster and one with name namespace and value $namespace. The panel would then show all investigations matching both:

  • cluster="dev" OR cluster="prod"; AND
  • namespace="default" OR namespace="app"

Template variables with ‘All’ values are also supported.