| name | e2e-tests |
| description | Run end-to-end tests for CAPOA using Ansible playbooks against a remote bare-metal test host. Use when running e2e tests, setting up test environments, or troubleshooting e2e failures. |
E2E Tests
End-to-end tests run as Ansible playbooks against a remote bare-metal host. They provision a Kind management cluster, deploy CAPI + CAPOA + CAPM3, emulate bare-metal nodes via libvirt, and install an OpenShift cluster.
Prerequisites
- distrobox container
capoa-dev (see the dev-env skill)
.venv activated inside distrobox with ansible installed- SSH access (key-based) to the remote test host
- A base64-encoded pull secret for container registries
Setup
Inside the distrobox container:
distrobox enter capoa-dev
cd <project-root>
source .venv/bin/activate
If .venv does not exist or ansible is missing, create it:
python3 -m venv .venv
source .venv/bin/activate
pip install ansible ansible-lint jinja2-cli kubernetes
Install Ansible Galaxy collections (done automatically by make e2e-test, or manually):
ansible-galaxy collection install -r test/ansible-requirements.yaml
Required Environment Variables
Set these before running:
| Variable | Description |
|---|
REMOTE_HOST | Hostname or IP of the remote test machine |
SSH_KEY_FILE | Path to SSH private key for the remote host |
SSH_AUTHORIZED_KEY | SSH public key content (e.g. $(cat ~/.ssh/id_rsa.pub)) |
PULLSECRET | Base64-encoded container registry pull secret |
DIST_DIR | Directory for build artifacts (e.g. /tmp/dist) |
CONTAINER_TAG | Image tag for built provider images (default: local) |
Check ~/.cursor/skills/ for a personal capoa-local-env skill that may contain preferred defaults for these variables.
Running
make e2e-test
This installs Ansible Galaxy collections and runs:
ansible-playbook test/playbooks/run_test.yaml -i test/playbooks/inventories/remote_host.yaml
The full run takes 30+ minutes.
On immutable distributions (Fedora Silverblue/CoreOS), override Ansible cache paths:
ANSIBLE_HOME=/tmp/.ansible ANSIBLE_LOCAL_TEMP=/tmp/.ansible.tmp XDG_CACHE_HOME=/tmp/.cache ANSIBLE_CACHE_PLUGIN_CONNECTION=/tmp/.ansible-cache make e2e-test
Optional Environment Variables
| Variable | Default | Description |
|---|
CLUSTER_TOPOLOGY | multinode | multinode, sno, multinode-okd, or sno-okd |
NUMBER_OF_NODES | 7 | Number of emulated BMH nodes |
CAPI_VERSION | v1.11.3 | Cluster API version |
CAPM3_VERSION | v1.11.2 | CAPM3 version |
SKIP_BUILD | false | Skip building provider images |
SKIP_CAPI_INSTALL | false | Skip CAPI/MCE component installation |
SKIP_CLUSTER | false | Skip cluster provisioning and assertions |
APPLY_EXAMPLE_ONLY | false | Only render and apply the example template |
USE_MCE_CHART | false | Install via MCE Helm chart instead of clusterctl |
INSTALLATION_MODE | downstream | downstream or upstream (MCE mode) |
GITHUB_TOKEN | -- | Required for upstream MCE mode |
Common Patterns
SNO (Single Node OpenShift)
export CLUSTER_TOPOLOGY="sno"
export NUMBER_OF_NODES="1"
make e2e-test
Re-run without rebuilding images
SKIP_BUILD=true make e2e-test
Re-apply cluster manifest only
Against an already-provisioned environment:
APPLY_EXAMPLE_ONLY=true CLUSTER_TOPOLOGY=sno make e2e-test
Demo mode (infra only, no CAPI, no cluster)
SKIP_BUILD=true SKIP_CAPI_INSTALL=true SKIP_CLUSTER=true make e2e-test
Playbook Structure
The playbook runs these roles in order: system_dependencies -> sourcecode_setup -> network_setup -> baremetal_emulation -> kind_setup -> build_images -> components_install -> bmh_setup -> cluster_install -> assert_install -> assert_upgrade (conditional). See test/playbooks/README.md for details on each role.
