This chapter discusses setting up the individual stages that make up the reference CI/CD flow:

NOTE The container registry must be set up beforehand and the same CONTAINER_REGISTRY parameter must be used in all werf commands.

Building and publishing images

You can build and publish the images described in werf.yaml using the following command:

werf build --repo CONTAINER_REGISTRY

Fast commit testing

Fast tests (unit tests or linters) can be performed using the following werf command (it starts a one-shot container and runs the specified TEST_COMMAND command):

werf kube-run IMAGE_NAME -- TEST_COMMAND

The command above will run the TEST_COMMAND in the container in the Kubernetes cluster. Note that the IMAGE_NAME must be defined in the werf.yaml file — it can be an image built specifically for such tests or a regular application image.

Preparing release artifacts

At this step, you need to publish the werf bundle to the container registry. The werf bundle publish command will automatically make sure that all the necessary images are built and published to the container registry and will publish the bundle as an OCI Helm chart with the specified SEMVER tag:

werf bundle publish --repo CONTAINER_REGISTRY --tag SEMVER

The use of SEMVER tags with OCI Helm charts is mandatory at this point (static tags like main / latest are not yet supported). Depending on the CI/CD model used:

  • If the application is released under SEMVER Git tags, it is recommended to publish the bundle under the same SEMVER tag;
  • If the application is deployed to production from the main branch (either automatically from HEAD commit or manually from some other commit), it is recommended to publish the bundle under an artificial tag like 0.0.${CI_PIPELINE_ID}. In this case, CI_PIPELINE_ID is some built-in CI/CD system variable, which is a monotonically increasing number that identifies the entire pipeline of CI/CD steps for each Git commit.

Acceptance testing

The acceptance testing step requires deploying the application to a temporary production-like environment, running the test job, and waiting for the results. This task can be accomplished in a number of alternative ways.

Typically, acceptance testing should run automatically for new commits to the main branch. In addition, failed acceptance tests usually prevent the deployment of release artifacts to production.

Thus, you need to configure the CI/CD step to run the test command on new commits to the main integration branch.

  • The process can then be optimized so that tests are not run on every commit to the branch, but only on the latest commit at the moment;
  • It may be necessary to implements a queue to run only one testing environment (namespace and release) at a time, or run each commit in a separate unique environment.

The following sections will describe how to do this.

Running with werf converge

If the special test Job is defined as a Helm post-install hook in the test environment, run the following command in the project’s Git worktree:

werf converge --repo CONTAINER_REGISTRY --env test

This command will start testing by deploying the PROJECT_NAME-test release in a production-like Kubernetes environment in the PROJECT_NAME-test namespace and run the post-install test Job (or Jobs). Then the converge command will automatically collect the test Job (or Jobs) output and terminate the entire process if the test fails.

Use the dismiss command to terminate the test environment:

werf dismiss --env test

Running with release artifact and werf

In this case, the werf bundle apply command deploys the previously published release artifact.

If you have defined a special test Job running in the test environment as a post-install Helm hook, run the following command:

werf bundle apply --repo CONTAINER_REGISTRY --env test --namespace TEST_NAMESPACE --release TEST_RELEASE

Note that this command (unlike werf converge) does not require you to specify the Git worktree of the project, or its output will be the similar to the one of werf converge.

The TEST_NAMESPACE and TEST_RELEASE variables must be specified explicitly and cleaned up after use:

werf helm uninstall TEST_RELEASE -n TEST_NAMESPACE
kubectl delete ns TEST_NAMESPACE

Running with release artifact and Argo CD

First, you need to create a special Argo CD Application CRD:

    namespace: TEST_NAMESPACE
    chart: PROJECT_NAME
    targetRevision: SEMVER

Execute the following command in your CI/CD to run the tests:

argocd app sync TEST_APPNAME --revision SEMVER

— where SEMVER is the regular version or the artificial tag like 0.0.${CI_PIPELINE_ID} (see preparing release artifacts).

Deploying to production

Configuring Argo CD Application

When creating the Argo CD Application CRD, note the following fields:

    namespace: TEST_NAMESPACE
    chart: PROJECT_NAME
    targetRevision: SEMVER

NOTE The bundle must be published using the werf bundle publish --repo REGISTRY/PROJECT_NAME --tag SEMVER format to match the parameters specified in the CRD. REGISTRY/PROJECT_NAME is the repository address in the container registry that contains the published werf bundle as an OCI Helm chart.

Set targetRevision to some initial bootstrap version of your application. We will cover updating targetRevision when the new release comes out in the following sections.

Deploying by manual action

Add the command below to your CI/CD system. It will be invoked every time the user wants to manually deploy the application to production:

argocd app sync APPNAME --revision SEMVER

Note that the SEMVER revision should be generated in the same way as described in preparing release artifacts.

Continuous Deployment

Continuous deployment means that the latest release artifact for, e.g., the main branch will be automatically deployed to the Kubernetes production cluster.

To configure continuous deployment of release artifacts, add the following annotations to the Argo CD Application CRD:

kind: Application
  annotations: ~ 0.0 pullsecret:NAMESPACE/SECRET

The value of"~ 0.0" means that the operator must automatically roll out the chart to the latest patch version in the SEMVER range of 0.0.*.

You must also create a NAMESPACE/SECRET secret (it is recommended to use the type: secret) to access the CONTAINER_REGISTRY with the published release artifact.