Quickstart

A first run on a local kind cluster. You need Docker running, plus go, task, kind, kubectl, and helm on your machine. On a machine with Homebrew: brew install go task kind kubectl helm.

git clone https://github.com/ydixken/operator-testbench
cd operator-testbench
task e2e:kind OPERATOR=valkey

The task creates the kind cluster testbench (one control plane, two workers), switches your kubectl context to kind-testbench, preloads the Valkey images into the cluster, installs valkey-operator at the version the pack pins, and runs every scenario that is not labelled chaos across two Ginkgo processes. A full run takes 15 to 30 minutes on a laptop; most of it is waiting for Valkey clusters to form and for slots to migrate.

Each scenario shows up as one Ginkgo container, each step as one spec:

cluster-ready 01 apply a 3x1 cluster
cluster-ready 02 the cluster reports Ready for the applied generation
...
Ran 19 of 19 Specs in 66.820 seconds
SUCCESS! -- 19 Passed | 0 Failed | 0 Pending | 0 Skipped

The Ginkgo report goes to reports/report.json and reports/junit.xml. A failed step prints the namespace's pods and events and the operator's log tail as report entries.

task e2e:kind OPERATOR=valkey FOCUS='cluster-ready'   # one scenario
task e2e:kind OPERATOR=valkey LABELS=chaos            # the chaos scenarios
task kind:down                                        # delete the cluster

To run against the cluster your kubectl context already points at, use task e2e OPERATOR=valkey. It prints the context and asks before touching anything; Environments explains the rules that apply to a real cluster.

Next: Concepts for the vocabulary, Writing scenarios to add a scenario, Operator packs to add an operator.