Run Database Observability on clustered Grafana Alloy

Database Observability now supports Grafana Alloy clustering for MySQL and PostgreSQL. When you run several Alloy nodes with the same configuration, Alloy assigns each database to exactly one node. Telemetry collection is distributed across the cluster and automatically moves to another node if one becomes unavailable, without duplicating data.

Clustering also works with a single database. If you run more than one Alloy node, you get a highly available setup with no extra configuration.

How it works

To run Database Observability in clustered mode, update your database_observability.mysql or database_observability.postgres component:

1. Define all your databases as database_instance blocks in one component, instead of using one component per database.

2. Add the clustering block with enabled = true.

3. Configure the component’s embedded exporter in order to replace the separate exporter component. Enable the collector that the Database Observability app uses in the prometheus_exporter block:

   • PostgreSQL: enabled_collectors = ["stat_statements"]
   • MySQL: enable_collectors = ["perf_schema.eventsstatements"]

With the next Alloy release (1.21), step 3 won’t be required.

Currently, all databases in a clustered component share the same tuning settings, and you can’t relabel telemetry per database yet.

Get started

Clustering for Database Observability is available now in Grafana Alloy 1.20 and later. To learn more, see the Alloy clustering documentation.

2026-08-26

Investigate database performance with Grafana Assistant

2026-06-03

Visualize and manage configuration pipelines

2026-04-23

Build custom Alloy pipelines without writing Alloy syntax