dagster-uc 0.6.6

CLI for managing user code deployments of a Dagster instance that was deployed on kubernetes.

Introduction

This experimental CLI allows you to manage user code deployments for a Dagster instance deployed on Kubernetes. It packages your code branch into a Docker container, uploads it to your container registry, and updates your existing Dagster instance to enable your user code deployment.

Pre-requisites

• kubectl with a valid config
• Helm 3
• Podman
• Python 3.10+
• Azure CLI (if you are using Azure container registry and use_az_login)

Installation

• Install from PyPI:

  pip install dagster-uc
  

• Create a configuration file named .config_user_code_deployments.yaml in the root of your repository or in your home directory. You can also create one by running:

  dagster-uc init-config -f '.config_user_code_deployments.yaml'
  

Configuration (nested structure)

The configuration format is now nested and grouped by concerns (e.g. docker, kubernetes, helm). The top-level keys you will commonly use are:

• defaults — values applied to every environment unless overridden.
• Per-environment sections (e.g. dev, acc, prd) — environment-specific overrides.
• docker — Docker/build related settings (dockerfile, registry, image prefix, build env vars, etc).
• kubernetes — Kubernetes-specific settings (context, namespace, resource requests/limits, env and secret lists).
• helm — Helm-related settings (for example, skip schema validation).
• chart — chart-specific values you may want to supply into the user-deployments Helm chart.
• use_latest_chart_version, use_project_name, and project_name_override — deployment behavior flags.

Order of loading configuration:

1. defaults
2. environment-specific keys (e.g. dev)
3. environment variable overrides

Below is an example config that mirrors the new nested structure:

defaults:
  repository_root: "."
  code_path: example/repo.py
  dagster_version: 1.11.16
  cicd: false
  use_project_name: True
  use_latest_chart_version: True
  # Docker configuration (grouped under `docker`)
  docker:
    docker_root: "."
    docker_env_vars:
      - FOO
      - FOO='string'
    image_prefix: "example"
    use_az_login: True
    container_registry_chart_path: "helm/dagster/dagster-user-deployments"

  # Helm configuration (grouped under `helm`)
  helm:
    skip_schema_validation: True

  # Kubernetes configuration (grouped under `kubernetes`)
  kubernetes:
    namespace: dagster
    node: cpunode
    requests:
      cpu: 150m
      memory: 750Mi
    limits:
      cpu: 4000m
      memory: 2000Mi
    user_code_deployment_env_secrets:
      - name: dagster-storage-secret

  chart: {}

dev:
  environment: dev
  dagster_gui_url: "http://dagster.dev"

  docker:
    dockerfile: "docker/dev.Dockerfile"
    container_registry: dagster-uc.dev.acr.io

  kubernetes:
    context: "aks-dev"
    user_code_deployment_env:
      - name: ON_K8S
        value: '1'
      - name: ENVIRONMENT
        value: dev

acc:
  environment: acc
  dagster_gui_url: "http://dagster.acc"
  project_name_override: 'example-acc'

  docker:
    dockerfile: "docker/acc.Dockerfile"
    container_registry: dagster-uc.acc.acr.io

  kubernetes:
    context: "aks-acc"
    user_code_deployment_env:
      - name: ON_K8S
        value: '1'
      - name: ENVIRONMENT
        value: acc

Notes on common config keys (now nested):

• code_path — path to the Python module that starts your Dagster definitions (used to start the user-code gRPC server).
• docker.dockerfile — path to the Dockerfile to build the image.
• docker.container_registry — target container registry for the built image (per-environment override).
• docker.image_prefix — prefix used when naming images.
• docker.docker_env_vars — list of environment variables to pass into the build process.
• docker.use_az_login — set to True if you need to log in to Azure before pushing images.
• kubernetes.context — the kube context to use for deployments in that environment.
• kubernetes.namespace — the namespace where Dagster and user deployments live.
• kubernetes.requests / kubernetes.limits — resource requests and limits for the user code deployment pod.
• kubernetes.user_code_deployment_env — a list of name/value env objects to inject into the user-code deployment container.
• kubernetes.user_code_deployment_env_secrets — a list of secrets to be mounted/injected as environment variables.
• helm.skip_schema_validation — useful for older Helm chart setups or when schema validation causes issues.
• use_project_name — when True, the project name from pyproject.toml is prefixed to the deployment name.
• project_name_override - When set, the project name from pyproject.toml is overridden with this value

Overriding Config with Environment Variables

Environment variables can still be used to override configuration at load time but you need to scaffold them in the yaml using cicd: ${CICD}. You can export common flags (for example CICD=TRUE or VERBOSE=TRUE) to affect behavior — top-level boolean or string fields are typically overridden this way. If you need to override nested values in automation, set the environment variables your automation expects before running the CLI.

Usage

• Deploy the currently checked out Git branch:

  dagster-uc deployment deploy
  

• See all available commands:

  dagster-uc --help
  

Branch naming and deployments

• When cicd: true is set, the deployment name is derived from the environment value.

• When cicd: false, the deployment name is derived from the Git branch name. The branch name is normalized by replacing non-alphanumeric characters with hyphens and stripping leading/trailing hyphens. Example: feat: my amazing feature -> feat-my-amazing-feature

• You can deploy the same branch multiple times by supplying --deployment-name-suffix, which appends a suffix to the deployment name.

• When use_project_name is enabled, the internal deployment name will be prefixed by a project slug derived from your pyproject.toml. Internally the prefix separator is -- so an example name may be my-project--feat-a, which appears in the Dagster UI as project:branch.

Building and container behavior

• The build process passes a BRANCH_NAME build-arg so your code can behave differently per branch (e.g. selecting secrets, configuration).
• Images are versioned: the CLI will check the registry for existing tags and increment a version to avoid reusing tags that could break running jobs.
• Use a registry lifecycle/garbage-collection policy to keep old images from accumulating.

Example Dockerfile pattern:

FROM python:3.11-slim
ARG BRANCH_NAME
ARG DIR="APP"
WORKDIR $DIR
COPY my_project my_project
COPY pyproject.toml uv.lock README.md ./
RUN --mount=type=cache,target=/root/.cache/uv \
    uv sync --no-dev --link-mode=copy
ENV PATH="/$DIR/.venv/bin:$PATH"
ENV BRANCH_NAME=${BRANCH_NAME}

Kubernetes / Helm

• Make sure kubernetes.context can access the kubernetes.namespace.
• Configure kubernetes.requests and kubernetes.limits for the user-code deployment pod appropriately.
• Pass environment variables via kubernetes.user_code_deployment_env and secrets via kubernetes.user_code_deployment_env_secrets.
• Helm values and chart overrides can be supplied under chart in the config file — these are passed to the user-deployments Helm chart.

Tips

• Keep a defaults section in your config file to reduce duplication between environments.
• Use environment-specific overrides for registry, kube context, and secrets.
• If you use Azure, set docker.use_az_login: True and ensure your environment has access to az and the appropriate credentials.

Troubleshooting

• If Helm chart upgrades fail due to schema validation, try setting helm.skip_schema_validation: True in your defaults or environment override.
• Check that kubernetes.context points to the correct cluster and that your kube credentials have permission to modify the target namespace.
• Ensure the code_path points to an importable Python module that starts your Dagster definitions.

Example test config

A small, real example is checked into the repo under tests/config/.config_user_code_deployments.yaml and demonstrates the new nested structure used by the CLI tests.