Prometheus Integration

Dataiku exposes Unified Monitoring health statuses as a Prometheus-compatible API. This allows you to integrate Unified Monitoring statuses into your existing monitoring infrastructure.

Configuration

To scrape this data, you need to add a scrape configuration to your Prometheus configuration file (typically named prometheus.yml).

For more details, please refer to the Prometheus official documentation.

Below is a sample configuration you can reuse and adjust to your environment and needs.

global:
  scrape_interval: 15s
  evaluation_interval: 15s

scrape_configs:
  - job_name: 'dataiku-unified-monitoring'
    scheme: https
    metrics_path: '/public/api/unified-monitoring/deployer/metrics-prometheus'
    static_configs:
      - targets: ['<your-dss-host>:<port>'] # e.g., dataiku.example.com:443

    # Authentication using a Dataiku API Key
    bearer_token: '<YOUR_DATAIKU_API_KEY>'

Available Statuses

The /public/api/unified-monitoring/deployer/metrics-prometheus endpoint exposes data representing the health status of your deployed projects and endpoints. All statuses are exposed as Prometheus metrics (gauges) with the following values:

  • -1: NO_STATUS (Data is not available or not applicable)

  • 0: HEALTHY

  • 1: WARNING

  • 2: ERROR

Project Statuses

The following statuses are available for Projects. For more details, see Dataiku Projects.

  • global_status: The aggregated overall health status.

  • deployment_status: The health status of the deployment, as reported by the Deployer.

  • models_status: The worst model status aggregated from all models in the project. See Understanding Model Status.

  • execution_status: Aggregated status of scenario executions and webapps.

  • data_status: Status of Data Quality.

  • governance_status: Status related to governance policies and sign-offs (requires Dataiku Govern).

Endpoint Statuses

The following statuses are available for Endpoints. For more details, see API Endpoints.

  • global_status: The aggregated overall health status.

  • deployment_status: The health status of the endpoint, as reported by the Deployer.

  • model_status: The worst model status of all models matched to the API endpoint. See Understanding Model Status.

  • governance_status: Status related to governance policies and sign-offs (requires Dataiku Govern).

Labels

Statuses are enriched with Prometheus labels to allow for filtering and aggregation. Labels common to all object types include:

  • type: For projects, either MULTI_AUTOMATION_NODE or AUTOMATION_NODE. For endpoints, either MANAGED_API_ENDPOINT, or EXTERNAL_API_ENDPOINT.

  • displayName: The user-friendly name of the project or endpoint.

  • deploymentId: The unique identifier of the deployment.

  • infrastructureId: The ID of the target infrastructure (or the external endpoint scope name).

Other labels are available depending on the object type (bundleName, deployedProjectKey, endpointName, …)