Deploy LookML to Looker with GitHub Actions

We use Looker (now a Google/Alphabet company) create embedded dashboards in our platform. The projects in Looker at synced with Git repositories. If you make changes to your LookML outside of the Looker UI, you have to tell Looker that a change has occurred and for it to pull in those changes to make them available in the application.

This extra step of calling an HTTP URL is easy to do (doesn’t even require authentication) but easy to forget. I created a Github Actions workflow to automatically deploy our LookML when a push is mode to the main branch. This includes when a push is made directly to main or (recommended) a pull request is merged into main.

Step 1: Get Your Deployment Secret from Looker

In Looker’s Project Settings -> Configuration screen, you’ll find the “Set Webhook Secret” button. Click it to retrieve your project’s secret, which we’ll add to Github in Step 2.

A screenshot of the Project Settings screen in Looker's development mode. The "Webhook Deploy Secret" section is visible. — Looker Project Settings Screen

Make sure you click on the “Save Project Configuration” after copying your deployment key.

A screenshot from Looker's Webhook Deploy Secret settings. It shows a visible secret key with a "Don't Reset" button and an warning stating that: You wont' be able to access this secret again, so make sure to copy it now." — Looker’s Webhook Deploy Secret Settings after clicking on the “Set Webhook Secret” button.

Step 2: Add Your Secret to GitHub

Next, we need to store the deployment secret in your repository’s Action Secrets. This is found in the Settings tab of your repository under “Secrets and variables” and then “Actions”.

Add a new secret called LOOKER_DEPLOYMENT_KEY as a secret, with the value being the key from Step 1 above.

GitHub Actions Secrets > New Secret screen showing the Name field as LOOKER_DEPLOYMENT_KEY and a secret value populated from Step 1" — Screenshot of the New Action Secret screen in GitHub

Step 3: Add a GitHub Action to your Repository

To do this, add a file called deploy_to_looker.yml to your .github/workflows folder in the LookML repo (create the folder if it doesn’t exist)

The contents of the file should look like this. You need to replace the deployment URL with your own, and the name of the GitHub Secret if you didn’t use the one in the example above.

on:
  push:
    branches:
      - main
name: Deploy to Looker
jobs:
  deployment:
    runs-on: ubuntu-latest

    steps:
      - name: Make HTTP Request to deploy
        id: requestToLooker
        uses: fjogeleit/http-request-action@v1
        with:
          url: "https://myinstance.looker.com/webhooks/projects/myprojectname/deploy"
          customHeaders: '{"X-Looker-Deploy-Secret": "${{ secrets.LOOKER_DEPLOYMENT_KEY }}"}'
      - name: Show Response from Looker
        run: |
          echo ${{ steps.requestToLooker.outputs.response }}
      - name: Check the deployment completed
        if: ${{ fromJson(steps.myRequest.outputs.response).operations.error }}
        run: |
          echo "LookML Deployment Failed";
          exit 1

Step 4: Test a Deployment

Push a change to main, or merge a pull request into main and you’ll see under the “Actions” tab that the deployment has run.

In the screenshot below, you see the “Show Response from Looker” step opened with the JSON response Looker returns if the deployment succeeds:

{operations:[{error:false}]}

GitHub Actions, with the last run Action open, and the 'Show Response from Looker' results open. — GitHub Actions screen showing a successful deployment to Looker

Caveat

In my testing, I couldn’t figure out how to make the deployment fail in a way that would get an error code back from Looker. I have to assume right now that the last step would correctly detect an error. If you happen to know a trick, let me know.

Linting LookML

Next I want to look at linting LookML using Look At Me Sideways (LAMS), a project in the looker-open-source GitHub repository.