Private preview Grafana Cloud

Configure dataset for Host Monitoring

Configure how the knowledge graph processes Host Monitoring metrics from Node Exporters and process metrics. This configuration enables entity discovery for non-Kubernetes hosts and processes running on physical or virtual machines.

For background on dataset configurations, environment labels, and cloud region labels, refer to Manage dataset configurations.

Note

Activating the knowledge graph using Host Monitoring is in private preview. To request access, contact your Grafana account representative or Grafana Support.

Before you begin

Host Monitoring discovers entities from the following exporters and metric sources:

  • Node Exporter metrics (node_uname_info, node_exporter_build_info)
  • Windows Exporter metrics (windows_os_info)
  • Process Exporter metrics for process discovery (namedprocess_.+)
  • OpenTelemetry process metrics using semantic conventions (process_.+)
  • Windows process metrics (windows_process_.+)

The Host Monitoring dataset is considered healthy when either node_exporter_build_info or windows_os_info metrics are present in your stack.

Complete the following steps to edit a dataset configuration for Host Monitoring:

  1. Sign into Grafana Cloud and click Observability > Configuration.

  2. Click Connect knowledge graph.

  3. Next to Host Monitoring, click Edit.

  4. Complete the following fields:

    FieldDescription
    Environment labelSelect the source environment you want to monitor with the knowledge graph.
    EnvironmentsDeselect the environments you want to exclude from the knowledge graph.
    Cloud region labelOptional. If you selected a different environment label, choose cluster as the cloud region label. You can select a different site label, as required.
    Cloud regionsDeselect the regions you want to exclude from the knowledge graph.
  5. If you want to add additional filters, complete the following steps:

    1. Click + Add filter.

    2. Populate the fields with filter values and an operator.

      You can add multiple filters.

  6. To manage framework selection, toggle Advanced options and deselect any frameworks you want to exclude from the knowledge graph.

  7. Optionally, add more environments.

  8. Click Save Host Monitoring settings.

Required metric labels

The knowledge graph requires the following metrics and labels for entity discovery:

MetricRequired labelsPurpose
node_uname_infoinstanceDiscover Linux and Unix hosts
windows_os_infoinstanceDiscover Windows hosts
process_.+VariesDiscover processes as services
windows_process_.+VariesDiscover Windows process metrics

If the required labels are missing, the knowledge graph can’t discover entities from those metrics. Refer to Troubleshooting for details on resolving missing label issues.

Frameworks and vendors

Host Monitoring includes the following frameworks for entity discovery:

  • Node Exporter: Discovers non-Kubernetes nodes from node_uname_info metrics
  • Windows Exporter: Discovers non-Kubernetes Windows hosts from windows_os_info metrics
  • Process Exporter: Discovers processes as Service and ServiceInstance entities from Prometheus exporter metrics
  • OpenTelemetry Process (SemConv): Discovers processes as Service and ServiceInstance entities from OpenTelemetry semantic conventions
  • Windows Process: Discovers processes from Windows Exporter process collector metrics

Node Exporter and Windows Exporter frameworks are core to Host Monitoring. You can manage framework selection in Advanced options during configuration.

Next steps