Test Anything That Runs in Kubernetes
ork e2e is a Kubernetes behavioural verification tool. It spins up a real cluster, installs things, watches the API server, and asserts GVK state. Nothing in that pipeline requires the workload under test to be an Orkestra operator.
A Helm chart installs into the same API server. A kubectl apply produces the same objects. The assertion layer does not care about the installation mechanism — it watches GVKs. A Deployment/my-app is the same object whether it was created by an Orkestra operator, a Helm chart, a controller-runtime reconciler, or kubectl apply.
The insight
Today there is no way to look at a container image, a deployment manifest, or a Helm chart and say it works. Signatures prove ownership. They say nothing about behaviour. A signed chart with a broken init container is still signed. A signed operator that silently drops events under load is still signed.
ork e2e built the verification layer in the hardest context — operators with long-lived state, drift correction, complex status, and partial failure modes. If you can prove an operator works correctly before distribution, you can prove a Helm chart works. The machinery is the same. The e2e file is the contract. The cluster is the witness.
Contents
| Page | What it covers |
|---|---|
| How it works | custom.target: kubernetes, the e2e file structure, running it |
| Use cases | Helm charts, third-party operators, platform stacks |
| CI integration | GitHub Actions bridge — one step to add e2e to any workflow |
→ Start with How it works