Skip to main content

Deploy with an agent

Copy agent prompt

Before you start

Connect your AWS account using the Spring Winter quickstart. You need postgreSQL for application state and an image that includes required plugins. Check the upstream deployment instructions and pin a release before you deploy. Source repository.

Create and connect the database

  1. Create a private SQL database with PostgreSQL for this deployment. If you already have a database for this deployment, verify that the engine and version match the pinned app release. Wait for the database to be ready.
  2. Open Connect on the database. Use the writer endpoint, port, database name and application credentials. Create the application database and role if needed. Confirm that the application and migration process can reach the private writer.
  3. Set GF_DATABASE_TYPE=postgres and the GF_DATABASE_HOST, GF_DATABASE_NAME, GF_DATABASE_USER, GF_DATABASE_PASSWORD and TLS settings. Keep analytics data source connections separate. Put secret values in runtime configuration or the migration process’s secret store.
  4. Start Grafana and wait for its startup migrations and /api/health before configuring data sources.

Configure the deployment

  1. Select a Spring Winter project in the AWS Region you want to use. Use a web server for the HTTP app and configure dependencies before you start it.
  2. Use grafana/grafana on port 3000 with /api/health.
  3. Configure GF_DATABASE_TYPE=postgres, the GF_DATABASE connection settings, and GF_SERVER_ROOT_URL.
  4. Bake plugins and provisioned configuration into the image if they would otherwise be lost with the task.
When you run the HTTP app on Spring Winter, set a health path that returns HTTP 200 without signing in. Use Runtime variables for runtime configuration. Task storage is temporary. Verify that every file you need after replacement goes to durable storage.

Verify the deployment

Check the deployment in History if you created Spring Winter compute. Wait for the workflow to succeed, then test the application over HTTPS. Sign in, connect a test data source, save a dashboard and verify that it survives replacement.