Skip to content

Deployment

task onboard brings the entire stack up in a single command: the colima VM, the minikube cluster, every app under platform/, and the local Python and Node toolchain. This page assumes the tools from Prerequisites are already installed.

Quick start

Pick a name for the kube context (any string; retail-lakehouse is the convention used throughout this repo) and run:

task onboard KUBE_CONTEXT=retail-lakehouse DEPLOY_METHOD=scripts

Expect 10 to 20 minutes on first run. When it returns, the cluster is up, every platform/ app is healthy, and the local dev environment is wired.

What task onboard does

task onboard is a thin orchestration over three steps:

  1. Provision the cluster. Boots a colima VM and starts a minikube cluster on top of it.
  2. Install the platform. Walks every directory under platform/ in dependency order and applies it to the cluster.
  3. Set up the local toolchain. Installs Python deps with uv, Node deps with npm, and the project's pre-commit hooks.

DEPLOY_METHOD controls how step 2 is performed:

Method What it does When to pick it
scripts Runs each platform/*/bootstrap.sh directly in dependency order Recommended for first-time exploration; failures surface step by step in the terminal
argocd Bootstraps Argo CD itself, then applies the app-of-apps manifest under charts/app-of-apps/ Mirrors a real GitOps workflow and gives you a UI to inspect drift and health

You can switch methods later by tearing down (task offboard) and re-running task onboard with the other value.

Validate

Once task onboard returns, re-run validation any time you want a quick health check:

task apps:validate KUBE_CONTEXT=retail-lakehouse

It runs every platform/*/validate.sh in parallel. A non-zero exit means at least one app failed its readiness probe.

Tear down

To free the resources entirely:

task offboard KUBE_CONTEXT=retail-lakehouse

This deletes the minikube cluster and the colima VM. Both deletes are no-ops if the targets don't exist, so it's safe to re-run.