Understanding clusters#
A cluster is a Trino coordinator plus its workers, running on AWS. If you only run queries you don't manage clusters, but you do choose which one runs your queries, so it helps to understand their state.
The clusters view#
The Clusters view is a searchable, filterable table showing each cluster's name, status, instance type, region, worker count, autoscaling, auto-suspend setting, and owner. Click a row to open its detail page (metrics, utilization, and recent scaling events).
What the statuses mean#
| Status | Meaning |
|---|---|
| Not enabled | Defined but never started; no AWS resources running. |
| Creating | Initial AWS resources are being provisioned. |
| Starting | Nodes are booting (see cold start below). |
| Running | Ready to accept queries. |
| Scaling | Adding or removing workers automatically. |
| Suspending | Shutting down idle nodes. |
| Suspended | Idle; no nodes running. A query will resume it first. |
| Updating | Applying configuration changes. |
| Failed | Something went wrong; an admin needs to look. |
| Deleting | All tracked AWS resources are being torn down. |
Picking a cluster#
To run a query the cluster must be Running. In the SQL editor and notebooks, the cluster selector lists the clusters your roles grant you access to.
Cold starts#
To save money, clusters can auto-suspend when idle. The first query against a suspended cluster automatically resumes it. Because each node re-downloads Java and Trino on first boot, this "cold start" typically takes 3–5 minutes. The UI tells you when a cluster is resuming — just wait for it to reach Running, then run your query.
If a cluster is autoscaling, it adds workers when queries queue up and removes them when things go quiet, so performance adjusts to demand automatically.
Connecting other tools#
Besides the built-in SQL editor and notebooks, you can point an external client — a BI tool, driver, or the Trino CLI — at a cluster. Click the info button (ⓘ) on a cluster row to open its Connection info popup, then copy the ready-made JDBC URL, ODBC string, or CLI command. Each string is scoped to your username. If the fields aren't available yet, an admin may still need to set up a cluster domain or start the cluster.