Template Variables

Variable Types

Template variables make dashboards reusable across services, environments, and clusters. Users select values from dropdowns, and all panels update to reflect the selection. Grafana supports several variable types for different use cases.

Query Variables

Query variables pull values dynamically from Prometheus label values. The query field uses label_values() or query_result() functions. The variable updates automatically as new labels appear in Prometheus.

# All job values
label_values(http_requests_total, job)

# All endpoints for the selected job
label_values(http_requests_total{job="$job"}, endpoint)

# All instances
label_values(up, instance)

Custom Variables

Custom variables let you define a fixed list of options. Use them for environment names, severity levels, or any predefined set. Custom variables are static and do not change with data.

{
  "name": "environment",
  "type": "custom",
  "query": "production,staging,development",
  "current": {
    "selected": true,
    "text": "production",
    "value": "production"
  }
}

Interval Variables

Interval variables represent time ranges. They are used in rate() and increase() functions to let users change the aggregation window. The auto option calculates intervals based on the dashboard time range.

# Rate with variable interval
rate(http_requests_total[$interval])

# Increase with variable interval
increase(http_requests_total[$interval])

Datasource Variables

Datasource variables let users switch between Prometheus instances. This is useful for comparing metrics across clusters, regions, or environments. The variable resolves to the selected datasource UID.

{
  "name": "datasource",
  "type": "datasource",
  "query": "prometheus",
  "current": {
    "text": "Prometheus",
    "value": "prometheus"
  }
}

Variable Chaining

Variables can reference other variables. A downstream variable filters its values based on the upstream selection. This creates cascading dropdowns where selecting a job filters the available endpoints.

# Variable: job = label_values(http_requests_total, job)
# Variable: endpoint = label_values(http_requests_total{job="$job"}, endpoint)
# Variable: instance = label_values(http_requests_total{job="$job", endpoint="$endpoint"}, instance)