Skip to main content

Deploying Dagster+ to Kubernetes

In this tutorial, you'll deploy a Dagster+ project to Kubernetes using Dagster+ Hybrid. You'll set up three long-lived environments (dev, staging, and prod) and a GitHub Actions pipeline that automatically creates branch deployments for every pull request.

By the end, you'll have:

  • Environment-aware assets that behave differently in dev, staging, and prod
  • A multi-stage Docker image built with uv
  • Separate Helm-managed Kubernetes agents for each environment
  • A CI/CD workflow that maps branches to deployments and handles PR branch deploys automatically

Prerequisites​

To follow this guide, you'll need:

  • Python 3.10+ and uv installed (see the uv installation guide)
  • A Dagster+ account with three deployments named dev, staging, and prod
  • A Kubernetes cluster (GKE, EKS, AKS, or similar) with kubectl and helm configured
  • A container registry (GitHub Container Registry, Google Artifact Registry, etc.)
  • A GitHub repository for your code

Step 1: Bootstrap the project​

Scaffold a new Dagster project with dg:

uvx -U create-dagster project dagster-deploy-demo
cd dagster-deploy-demo

Add the dagster-cloud package so your code can interact with Dagster+ APIs:

uv add dagster-cloud

Project layout​

The project follows the standard dg project structure:

dagster-deploy-demo/
├── dagster_cloud.yaml          # Dagster+ code location config
├── Dockerfile                  # Multi-stage image built with uv
├── pyproject.toml
├── helm/
│   └── dagster-agent/
│       ├── values-dev.yaml
│       ├── values-staging.yaml
│       └── values-prod.yaml
├── .github/
│   └── workflows/
│       └── dagster-cloud-deploy.yml
└── src/
    └── project_dagster_plus_deployment/
        ├── definitions.py
        └── defs/
            ├── assets.py
            └── schedules.py

Step 2: Run locally​

Start the Dagster webserver to confirm the project loads correctly:

dg dev

Open http://127.0.0.1:3000 to see the empty asset graph.

Next steps​

Continue this example with defining environment-aware assets.