Get started

Five steps from a bare Ubuntu box to a running platform.

flowchart LR
    prereq["Prerequisites<br/>hardware, SSH, accounts"]
    env["Environment<br/>and secrets"]
    stage1["Stage 1<br/>provision the cluster"]
    stage2["Stage 2<br/>deploy the platform"]
    verify["Verify"]

    prereq --> env --> stage1 --> stage2 --> verify

Step	Roughly	What you end up with
Prerequisites	30 min	A reachable Ubuntu host with key-based SSH on a non-default port
Environment and secrets	30 min	A working tooling container with secrets injected from Bitwarden
Stage 1: provision the cluster	30–60 min	A hardened host running Kubernetes, with a local kubeconfig
Stage 2: deploy the platform	60+ min	GitLab, ArgoCD, monitoring, storage, ingress with TLS
Verify the install	15 min	Confidence that all of the above actually works

Everything runs in a container

You do not install kubectl, helm, terraform or ansible on your machine. task docker:build builds an Alpine image with all of them pinned, and task docker:exec drops you into it. The only things you need locally are Docker and Task.

Read this before Stage 1

Stage 1 hardens hosts: it enables UFW, removes snapd, disables swap and may reboot to enable memory cgroups. Run it against a machine you are willing to have reconfigured, not your daily driver.

Adding cloud machines later

Cloud (Stage 0) is a separate, optional Terraform root that creates machines in a cloud account and joins them to a tailnet. It produces workers rather than a control plane, so it runs against a cluster the five steps above already built, not ahead of them. The order of operations is the Oracle free tier worker runbook.