Respect `environment_scope` with `only: kubernetes: active` strategy (gitlab-ci.yml)

Summary

When a job has only: kubernetes: active strategy, the job is only executed when the project has an enabled Kubernetes instance (Platform::Kubernetes or KubernetesService).

Here is the document.

---

See the example below. Job is going to be created only when pipeline has been scheduled or runs for a master branch, and only if kubernetes service is active in the project.

job:
  only:
    refs:
      - master
      - schedules
    kubernetes: active

---

And recently, we implemented multiple Kubernetes cluster per project. We can have multiple Kubernetes instances in a project. Each instance belongs to each environment.

But today kubernetes: active just checks if the project has any clusters. This could make a slight edge case.

For example,

---

1. A project has a cluster with review/* environment scope.

gitlab-ci.yml

deploy to production:
  environment:
    name: production
  only:
    kubernetes: active

---

The job will be executed, because the project has a cluster, but the corresponding cluster doesn't exist, thus this job will fail or cause unexpected behavior.

Solution

Check environment:name when kubernetes: active strategy exists.

e.g.

1. If kubernetes: active exists and environment:name exists, we check the activeness of the corresponding cluster.
2. If kubernetes: active exists and environment:name does not exist, we check if any clusters exist.

Possible workaround

For the workaround, users would need to create a cluster with a wildcard scope (*) for now. IIRC our gitlab-org/gitlab project does the same workaround for the same reason.

Related

• https://gitlab.com/gitlab-org/gitlab-ee/merge_requests/3603#note_50150076
• https://gitlab.com/gitlab-org/gitlab-ee/issues/3734

/cc @bikebilly @ayufan @grzesiek