Implementing Dynamic Filtering with Grafana Dashboard Variables

Static dashboards fail as soon as your infrastructure scales. Creating a separate dashboard for every server, region, or microservice is unmaintainable. To solve this, Grafana uses Dashboard Variables—interactive inputs like dropdown menus at the top of a dashboard that allow users to filter all panels simultaneously.

The goal of a well-implemented variable strategy is to move from hardcoded values in your queries to dynamic references that update automatically as new data is added to your data sources.

Prerequisites

• Access to a Grafana instance (v7.0+ recommended).
• At least one configured data source (e.g., Prometheus, PostgreSQL, or InfluxDB).
• Edit permissions for the dashboard in question.

Creating a Query Variable

The most powerful variable type is the 'Query' variable, which fetches its options directly from your data source. This ensures that if a new host is spun up, it appears in your dashboard dropdown without manual configuration.

Navigate to Dashboard Settings (the gear icon in the top right) and select Variables. Click Add variable.

Example: Prometheus Host Label Selection

To create a list of all hosts from a Prometheus environment, use the following configuration:

Field	Value
Name	host
Type	Query
Data source	Prometheus
Query	label_values(up, instance)
Multi-value	Checked
Include All option	Checked

Injecting Variables into Queries

Once the variable is defined, you must reference it in your panel queries. The syntax varies based on the data source's query language.

Prometheus (PromQL) Syntax

When using multi-value variables in Prometheus, you must use regex matching to handle the multiple strings passed by Grafana:

rate(http_requests_total{instance=~"${host}"}[5m])

Note the =~ operator is critical here because Grafana replaces ${host} with a pipe-separated list of values (e.g., host1|host2).

SQL (PostgreSQL/MySQL) Syntax

For SQL databases, use the IN clause to handle multiple selections. Use the :csv formatter to ensure the variable is passed as a comma-separated list of quoted strings:

SELECT * FROM logs WHERE server_name IN (${host:csv})

Implementing Chained Variables (Dependencies)

You can chain variables so that selecting one filters the options available in another. For example, selecting a 'Region' should only show 'Hosts' located in that region.

1. Create the first variable (e.g., region) with a query like label_values(region).
2. Create the second variable (e.g., host).
3. In the host query, reference the first variable: label_values(instance, region="${region}").

When the region changes, the host dropdown will automatically refresh to show only relevant instances.

Verification and Diagnostics

To ensure your variables are functioning correctly, use the following checks:

• Query Inspection: Click on a panel, select Inspect > Query. Verify that the ${host} placeholder has been replaced by the actual selected value before being sent to the data source.
• Multi-value Validation: Select multiple values in the dropdown. If the panel returns no data or a syntax error, verify that you are using =~ (Prometheus) or IN (SQL) instead of a strict equality operator.
• Latency Check: If the dashboard takes several seconds to load upon changing a variable, check the indexing of the columns used in the variable query.

Rollback and Recovery

If a variable configuration breaks the dashboard panels:

• Revert Variable: Return to Dashboard Settings > Variables and delete the problematic variable or change its type back to 'Custom' to hardcode values temporarily.
• Query Reset: Remove the variable reference from the panel query (e.g., replace ${host} with a literal string) to restore panel visibility while debugging the variable query.