docker

Docker for dbt

This docker file is suitable for building dbt Docker images locally or using with CI/CD to automate populating a container registry.

Building an image:

This Dockerfile can create images for the following targets, each named after the database they support:

• dbt-core (no db-adapter support)
• dbt-postgres
• dbt-redshift
• dbt-bigquery
• dbt-snowflake
• dbt-spark
• dbt-third-party (requires additional build-arg)
• dbt-all (installs all of the above in a single image)

In order to build a new image, run the following docker command.

docker build --tag <your_image_name> --target <target_name> <path/to/dockerfile>

---

Note: Docker must be configured to use BuildKit in order for images to build properly!

---

By default the images will be populated with the most recent release of dbt-core and whatever database adapter you select. If you need to use a different version you can specify it by git ref using the --build-arg flag:

docker build --tag <your_image_name> \
  --target <target_name> \
  --build-arg <arg_name>=<git_ref> \
  <path/to/dockerfile>

valid arg names for versioning are:

• dbt_core_ref
• dbt_postgres_ref
• dbt_redshift_ref
• dbt_bigquery_ref
• dbt_snowflake_ref
• dbt_spark_ref

---

NOTE: Only override a single build arg for each build. Using multiple overrides may lead to a non-functioning image.

---

If you wish to build an image with a third-party adapter you can use the dbt-third-party target. This target requires you provide a path to the adapter that can be processed by pip by using the dbt_third_party build arg:

docker build --tag <your_image_name> \
  --target dbt-third-party \
  --build-arg dbt_third_party=<pip_parsable_install_string> \
  <path/to/dockerfile>

Examples:

To build an image named "my-dbt" that supports redshift using the latest releases:

cd dbt-core/docker
docker build --tag my-dbt --target dbt-redshift .

To build an image named "my-other-dbt" that supports bigquery using dbt-core version 0.21.latest and the bigquery adapter version 1.0.0b1:

cd dbt-core/docker
docker build \
  --tag my-other-dbt \
  --target dbt-bigquery \
  --build-arg [email protected] \
  --build-arg [email protected] \
 .

To build an image named "my-third-party-dbt" that uses Materilize third party adapter and the latest release of dbt-core:

cd dbt-core/docker
docker build --tag my-third-party-dbt \
  --target dbt-third-party \
  --build-arg dbt_third_party=dbt-materialize \
  .

Special cases

There are a few special cases worth noting:

• The dbt-spark database adapter comes in three different versions named PyHive, ODBC, and the default all. If you wish to overide this you can use the --build-arg flag with the value of dbt_spark_version=<version_name>. See the docs for more information.

docker build --tag my_dbt \
  --target dbt-postgres \
  --build-arg [email protected] \
  <path/to/dockerfile>

• If you need to build against another architecture (linux/arm64 in this example) you can overide the build_for build arg:

docker build --tag my_dbt \
  --target dbt-postgres \
  --build-arg build_for=linux/arm64 \
  <path/to/dockerfile>

Supported architectures can be found in the python docker dockerhub page.

Running an image in a container:

The ENTRYPOINT for this Dockerfile is the command dbt so you can bind-mount your project to /usr/app and use dbt as normal:

docker run \
  --network=host \
  --mount type=bind,source=path/to/project,target=/usr/app \
  --mount type=bind,source=path/to/profiles.yml,target=/root/.dbt/profiles.yml \
  my-dbt \
  ls

---

Notes:

• Bind-mount sources must be an absolute path
• You may need to make adjustments to the docker networking setting depending on the specifics of your data warehouse/database host.

---