dagster_cloud.yaml#

File location#

The dagster_cloud.yaml file should be placed in the root of your Dagster project. Below is an example of a file structure modified from the Dagster+ ETL quickstart.

quickstart_etl
├── README.md
├── quickstart_etl
│   ├── __init__.py
│   ├── assets
│   ├── docker_image
├── ml_project
│   ├── quickstart_ml
│     ├── __init__.py
│     ├── ml_assets
├── random_assets.py
├── quickstart_etl_tests
├── dagster_cloud.yaml
├── pyproject.toml
├── setup.cfg
└── setup.py

If your repository contains multiple Dagster projects in subdirectories - otherwise known as a monorepository - add the dagster_cloud.yaml file to the root of where the Dagster projects are stored.

File structure#

Settings are formatted using YAML. For example, using the file structure above as an example:

# dagster_cloud.yaml

locations:
  - location_name: data-eng-pipeline
    code_source:
      package_name: quickstart_etl
    build:
      directory: ./quickstart_etl
      registry: localhost:5000/docker_image
  - location_name: ml-pipeline
    code_source:
      package_name: quickstart_ml
    working_directory: ./ml_project
    executable_path: venvs/path/to/ml_tensorflow/bin/python
  - location_name: my_random_assets
    code_source:
      python_file: random_assets.py
    container_context:
      k8s:
        env_vars:
          - database_name
          - database_username=hooli_testing
        env_secrets:
          - database_password

Settings#

The dagster_cloud.yaml file contains a single top-level key, locations. This key accepts a list of code locations; for each code location, you can configure the following:

• Location name
• Code source
• Working directory
• Build
• Python executable
• Container context

Location name#

This key is required. The location_name key specifies the name of the code location. The location name will always be paired with a code source.

# dagster_cloud.yaml

locations:
  - location_name: data-eng-pipeline
    code_source:
      package_name: quickstart_etl

Code source#

This section is required. The code_source defines how a code location is sourced.

A code_source key must contain either a module_name, package_name, or file_name parameter that specifies where to find the definitions in the code location.

# dagster_cloud.yaml

locations:
  - location_name: data-eng-pipeline
    code_source:
      package_name: quickstart_etl

# dagster_cloud.yaml

locations:
  - location_name: data-eng-pipeline
    code_source:
      package_name: quickstart_etl
  - location_name: machine_learning
    code_source:
      python_file: ml/ml_model.py

Working directory#

Use the working_directory setting to load Dagster code from a different directory than the root of your code repository. This setting allows you to specify the directory you want to load your code from.

Consider the following project:

quickstart_etl
├── README.md
├── project_directory
│   ├── quickstart_etl
│     ├── __init__.py
│     ├── assets
│   ├── quickstart_etl_tests
├── dagster_cloud.yaml
├── pyproject.toml
├── setup.cfg
└── setup.py

To load from /project_directory, the dagster_cloud.yaml code location would look like this:

# dagster_cloud.yaml

locations:
  - location_name: data-eng-pipeline
    code_source:
      package_name: quickstart_etl
    working_directory: ./project_directory

Build#

The build section contains two parameters:

• directory - Setting a build directory is useful if your setup.py or requirements.txt is in a subdirectory instead of the project root. This is common if you have multiple Python modules within a single Dagster project.
• registry - Applicable only to Hybrid deployments. Specifies the Docker registry to push the code location to.

In the example below, the Docker image for the code location is in the root directory and the registry and image defined:

# dagster_cloud.yaml

locations:
  - location_name: data-eng-pipeline
    code_source:
      package_name: quickstart_etl
    build:
      directory: ./
      registry: your-docker-image-registry/image-name # e.g. localhost:5000/myimage

Python executable#

For Dagster+ Hybrid deployments, the Python executable that is installed globally in the image, or the default Python executable on the local system if you use the local agent, will be used. To use a different Python executable, specify it using the executable_path setting. It can be useful to have different Python executables for different code locations.

For Dagster+ Serverless deployments, you can specify a different Python version by following these instructions.

# dagster_cloud.yaml

locations:
  - location_name: data-eng-pipeline
    code_source:
      package_name: quickstart_etl
    executable_path: venvs/path/to/dataengineering_spark_team/bin/python
  - location_name: machine_learning
    code_source:
      python_file: ml_model.py
    executable_path: venvs/path/to/ml_tensorflow/bin/python

Container context#

If using Hybrid deployment, you can define additional configuration options for code locations using the container_context parameter. Depending on the Hybrid agent you're using, the configuration settings under container_context will vary.

Refer to the configuration reference for your agent for more info:

• Docker agent configuration reference
• Amazon ECS agent configuration reference
• Kubernetes agent configuration reference