GCP E2E CI Job (e2e-gke)

What is it

The e2e-gke job is a presubmit CI job in the openshift/hypershift repo that validates GCP platform changes end-to-end. It creates two ephemeral GCP projects (control plane + hosted cluster), provisions a GKE Autopilot cluster, installs the HyperShift operator, creates a hosted cluster with WIF and PSC, validates it (TestCreateCluster), and tears everything down.

When does it trigger

The job triggers on PRs to openshift/hypershift when files matching GCP-related code paths are modified:

api/hypershift/v1beta1/gcp.*
hypershift-operator/controllers/.*/gcp.*
control-plane-operator/controllers/.*/gcp.*
cmd/cluster/gcp/.*
cmd/nodepool/gcp/.*

It can also be triggered manually with /test e2e-gke.

Current flags: always_run: false, optional: true, skip_report: true — meaning it won't block PRs and results aren't posted as GitHub status checks yet.

What happens if it fails

• The job result is not reported on the PR (skip_report: true), so failures don't block merging
• Post steps always run, including deprovision — GCP projects are cleaned up even if the test or the job is aborted
• Artifacts (logs, junit, hypershift-dump) are uploaded to GCS for debugging
• Concurrency is limited to 10 parallel runs via Boskos leases

CI Workflow

The job uses the hypershift-gcp-gke-e2e workflow defined in openshift/release:

Pre phase:

1. ipi-install-rbac — Grant image-puller permissions
2. hypershift-gcp-gke-provision — Create GCP projects, VPC, PSC subnet, GKE Autopilot cluster
3. hypershift-gcp-gke-prerequisites — Install CRDs and cert-manager
4. hypershift-install — Install HyperShift operator with GCP support
5. hypershift-gcp-control-plane-setup — Configure operator WIF for PSC and ExternalDNS
6. hypershift-gcp-hosted-cluster-setup — Create RSA keypair, WIF pool/SAs, HC network

Test phase:

1. hypershift-gcp-run-e2e — Run TestCreateCluster

Post phase:

1. hypershift-dump — Collect logs and artifacts
2. hypershift-gcp-gke-deprovision — Delete GCP projects, GKE cluster, DNS records