Skip to main content

Capact CI and CD

โš ๏ธ DEPRECATION: The gcr.io/projectvoltron registry is deprecated. We support this registry in the read-only mode until November 2021. At that time this registry will likely be garbage collected and no longer available. There still be an option to clone repository from a given version and build images locally.

New Docker images are available in ghcr.io/capactio.

This document describes jobs created to automate the process of testing, building, and deploying newly merged functionality.

Overview#

For all our CI/CD jobs, we use GitHub Actions. Our workflows are defined in the .github/workflows directory. All scripts used for the CI/CD purpose are defined in the /hack/ci/ directory. For example, the /hack/ci/setup-env.sh file has defined all environment variables used for every pipeline job.

Repository secrets#

All sensitive data is stored in GitHub secrets. As a result, we can access them in each workflow executed on our pipeline.

The following secrets are defined:

Secret nameDescription
CAPACT_GCS_CREDSHolds credentials which allow CI jobs to manage GCS bucket. Has the roles/storage.objectAdmin role.
CAPACT_GKE_CREDSHolds credentials which allow CI jobs to create and manage the GKE private cluster. Has the Compute Admin, Compute Network Admin, Kubernetes Engine Admin, Kubernetes Engine Cluster Admin, Service Account User, Storage Admin roles.
STAGE_CAPACT_GATEWAY_PASSWORDHolds the Gateway password for the long-running cluster.
GIT_CRYPT_KEYHolds a symmetric key used to decrypt files encrypted with git crypt.

Pipelines#

Pull request#

PR build

The job is defined in the pr-build.yaml file. It runs on pull requests created to the main branch.

Steps:

  1. Lint and test submitted code.
  2. Check documentation if the *.md files were modified.
  3. Run integration tests.
  4. Build Docker images for applications, tests and infra tools, and push them to ghcr.io/capactio using this pattern: ghcr.io/capactio/pr/{service_name}:PR-{pr_number}.

Main branch#

NOTE: To reduce the CI build time, we disable the entry-tests, build-tools and integration-tests jobs. They will be enabled when the project is open-sourced.

Default branch build

The job is defined in the .github/workflows/branch-build.yaml file. It runs on every new commit pushed to the main branch but skips execution for files which do not affect the building process, e.g. documentation.

Steps:

  1. Lint and test code.
  2. Build Docker images for applications, tests and infra tools, and push them to ghcr.io/capactio using this pattern: ghcr.io/capactio/{service_name}:{first_7_chars_of_commit_sha}.
  3. If Capact Helm Charts were changed:
    1. Change version in all Chart.yaml to {current_version}-{first_7_chars_of_commit_sha}.
    2. Package and push charts to the capactio-latest-charts GCS.
  4. Update the existing long-running cluster via CLI.
  5. Delete all Actions which are in the SUCCEEDED phase and whose names have the capact-upgrade- prefix.
  6. If any step failed, send a Slack notification.

Recreate a long-running cluster#

Recreate cluster

The job is defined in the .github/workflows/recreate_cluster.yaml file. It is executed on a manual trigger using the workflow_dispatch event. It uses already existing images available in the ghcr.io/capactio registry. As a result, you need to provide a git SHA from which the cluster should be recreated. Optionally, you can override the Docker image version used via the DOCKER_TAG parameter.

CAUTION: This job removes the old GKE cluster.

Let's encrypt certificates#

The recreate job checks if the certificate exists in the GCS bucket. If it does, it downloads it and checks if the certificate is still valid. If it's valid, it copies it to a long-running cluster, otherwise the job creates the Let's Encrypt certificates using Cert Manager and backs it up to the dedicated GCS bucket. By doing so, we ensure that we do not hit the quotas defined on the Let's Encrypt side.

Execute integration tests on a long-running cluster#

Integration tests

The job is defined in the .github/workflows/cluster_integration_tests.yaml file. It runs periodically according to cron defined in the job definition. It executes integration tests using the helm test command.

Accessing encrypted files on CI#

The sensitive data that needs to be accessed on a pipeline, such as overrides for passwords, certificates etc., must be stored in the hack/ci/sensitive-data directory. Files in that folder are encrypted using git crypt, which you should install and configure on your local machine. Currently, it works for *.txt files put in this directory, but this can be changed in the .gitattributes file.

The demo setup is as follows:

*.txt filter=git-crypt diff=git-crypt.gitattributes !filter !diff

It means that every *.txt file in the hack/ci/sensitive-data directory is encrypted before being push to a git repository. If you need to encrypt other files in a different directory, you have to create there a .gittatributes file with proper rules. Do not forget to add the .gitattributes !filter !diff statement as it prevents encryption for the .gitattributes file.

To decrypt the data locally, you must either use a symmetric key or add the GPG key. The procedure of decrypting files and working with it in a team is described here.

Currently, decrypt.yaml shows how to decrypt a file on CI.

Add a new pipeline#

To create a new pipeline you must follow the rules of the syntax used by GitHub Actions. The new workflow must be defined in the .github/workflows directory. All scripts for CI/CD purposes must be defined in the /hack/ci/ directory.

In the capactio organization, the GITHUB_TOKEN generated for GitHub Actions has the read-only scope by default. Additional scopes must be set for all jobs in a Workflow or a specific Job. See the authentication in a Workflow manual page for more info.

The following steps show how to checkout the code, set up the Go environment, and authorize to GHCR and GKE in case they are necessary.

    # Each job needs to request a given set of permissions.    permissions:      contents: read
    steps:          - name: Checkout code        uses: actions/[email protected]
      - name: Authorize to GHCR        run: echo "${{ secrets.GITHUB_TOKEN }}" | docker login ghcr.io -u ${{ github.actor }} --password-stdin
      - name: Authorize to GKE        uses: google-github-actions/setup-[email protected]        with:          service_account_key: ${{ secrets.CAPACT_GKE_CREDS }}          export_default_credentials: true
      - name: Setup env        run: |          . ./hack/ci/setup-env.sh
      - name: Setup Go        uses: actions/setup-[email protected]        with:          go-version: ${{env.GO_VERSION}}