Starburst Galaxy

Create a role for metrics scraping

Starburst recommends creating a dedicated role just for metrics collection. The role only needs the Monitor cluster privilege on the clusters you want to scrape.

In Galaxy, open Access > Roles and privileges, click Add role, and create a role such as metrics-scraper. Leave Grant to the creating role unchecked. Then open that role, go to Privileges, and add the Monitor cluster privilege for the cluster you want to observe.

Starburst notes that the Use cluster privilege is already inherited from the public role, so you only need to grant Monitor cluster here. If you plan to scrape multiple clusters, repeat the privilege step for each cluster.

Create a service account

Prometheus should authenticate with a service account instead of a personal user account. Starburst’s metrics guide uses that service account as the identity for scraping.

Open Access > Service accounts, click Create new service account, and name it something clear such as metrics-scraper. It is better to use a service-style name here instead of a human name. Set the default role to the role you created above, then generate a password and save it somewhere safe.

The password is shown only once, so save it before you leave the screen.

The username format looks like this:

metrics-scraper@<your-account>.galaxy.starburst.io

Get the cluster host

Prometheus needs the cluster host so it knows where to scrape metrics from.

In Galaxy, open Partner connect, select the Trino Python tile, choose the cluster you want to monitor, and copy the Host value.

Starburst documents the same flow for finding the cluster URL used by monitoring clients.

Configure Prometheus

Now wire the two sides together. Prometheus scrapes Starburst Galaxy over HTTPS, then forwards the scraped metrics to Parseable with remote write.

Create a prometheus.yml file like this:

global:
  scrape_interval: 15s

scrape_configs:
  - job_name: 'starburst-galaxy'
    metrics_path: /v1/metrics
    scheme: https
    basic_auth:
      username: 'metrics-scraper@<your-account>.galaxy.starburst.io'
      password: '<service-account-password>'
    static_configs:
      - targets: ['<cluster-host>']
        labels:
          cluster: 'starburst-galaxy'

remote_write:
  - url: "http://<parseable-ingestor>:8000/v1/prometheus/write"
    basic_auth:
      username: <parseable-username>
      password: <parseable-password>
    headers:
      X-P-Stream: starburst-metrics
      X-P-Log-Source: otel-metrics

The Starburst side of this configuration follows their documented scrape pattern for Galaxy metrics, including metrics_path: /v1/metrics, HTTPS, and basic authentication with the service account.

On the Parseable side, remote_write sends Prometheus metrics into Parseable, X-P-Stream: starburst-metrics decides the destination dataset, and X-P-Log-Source: otel-metrics marks the payload as metrics data.

Run Prometheus and verify the scrape

If you want a quick local setup, a minimal Docker Compose file is enough:

# docker-compose.yml
services:
  prometheus:
    image: prom/prometheus:latest
    ports:
      - "9090:9090"
    volumes:
      - ./prometheus.yml:/etc/prometheus/prometheus.yml
    restart: unless-stopped

Start it:

docker-compose up -d

Then open:

http://localhost:9090/targets

If the scrape is working, the starburst-galaxy target should show as UP. The Prometheus Targets page is the first place to validate the scrape.

Verify data in Parseable

Once Prometheus starts forwarding the metrics, open Parseable, select the starburst-metrics dataset, and run a PromQL query like this:

count by (__name__) ({cluster="starburst-galaxy"})

This gives you a quick view of which metric series are arriving for that cluster. If the integration is working, you should start seeing JVM, query, and cluster metrics from the Galaxy cluster.

In practice, this is the point where the integration becomes useful. You are no longer just checking whether Prometheus can scrape the endpoint. You can now keep those metrics in Parseable, query them with PromQL, and inspect them next to the rest of your observability data.