GitHub Action for running the getdbt.com dbt CLI with BigQuery

README.md

Using GitHub Actions to run dbt

This example shows you how to use GitHub Actions to run dbt against BigQuery.

1. Follow the instructions on getdbt.com for installing and initializing a dbt project.

2. Copy this action (dbt.yml) into the workflows directory.

    mkdir .github
    mkdir .github/workflows
    cp ~/Downloads/dbt.yml .github/workflows/
   

3. Follow the instructions on getdbt.com for creating a BigQuery service account, download the json key file, and copy it into a GitHub Secret named DBT_GOOGLE_BIGQUERY_KEYFILE.

4. In the GitHub Action dbt.yml file, replace the Google Project and BigQuery dataset environment variables with your project's variables.

5. Push to GitHub and watch it run.

dbt.yml

name: dbt

on:
  push:
    branches:
      - master

env:
  DBT_PROFILES_DIR: ./
  DBT_GOOGLE_PROJECT: your-gcp-project-id
  DBT_GOOGLE_BIGQUERY_DATASET: bigquery-dataset-id
  DBT_GOOGLE_BIGQUERY_KEYFILE: ./.gcloud/dbt-service-account.json

jobs:
  dbt:
    name: dbt
    runs-on: ubuntu-latest

    steps:
      - uses: actions/checkout@master
      - uses: actions/setup-python@v1
        with:
          python-version: "3.7.x"
      - run: pip3 install dbt
      - run: dbt --version
      - run: 'echo "$KEYFILE" > ./.gcloud/dbt-service-account.json'
        shell: bash
        env:
          KEYFILE: ${{secrets.DBT_GOOGLE_BIGQUERY_KEYFILE}}
      - run: dbt run
      - run: dbt test

profiles.yml

# Add this file to the root of your dbt project and use environment variables
# to configure dev vs production. I recommend using direnv locally with .envrc files.

default:
  target: bigquery
  outputs:
    bigquery:
      type: bigquery
      method: service-account
      keyfile: "{{ env_var('DBT_GOOGLE_BIGQUERY_KEYFILE') }}"
      project: "{{ env_var('DBT_GOOGLE_PROJECT') }}"
      dataset: "{{ env_var('DBT_GOOGLE_BIGQUERY_DATASET') }}"
      threads: 10
      timeout_seconds: 300
      location: US
      priority: interactive

I recommend using dbt build instead of separate dbt run and dbt test. This means you have the choice of not running models downstream from models that have failed tests.