Flatcar Linux + HAMI Development Environment

Reproduce and fix a HAMI issue on Flatcar Linux with its immutable (r/o) rootfs.

Overview

• VM: Flatcar Linux via Vagrant + VirtualBox (2 vCPUs, 8G RAM)
• Kubernetes: kubeadm on a single-node cluster
• Provisioning: Butane config compiled to Ignition, injected via config drive
• Cluster components: Cilium CNI + HAMI, deployed via atmos + helmfile

Project structure

flatcar_hami/
├── Vagrantfile                    # VM: 2 CPU, 8G RAM, host-only IP
├── config/
│   └── butane.yaml               # Ignition: binaries, systemd units, containerd
├── atmos/
│   ├── atmos.yaml                 # atmos CLI config
│   └── components/                # Stack components (Cilium, HAMI)
├── helmfile/
│   └── helmfile.yaml             # Cilium + HAMI releases
├── Gemfile                        # Bundler: Vagrant gem
├── Gemfile.lock                   # Bundler lock (committed)
├── Makefile                       # Vagrant lifecycle + status check + deploy trigger
├── AGENTS.md
├── README.md
└── .gitignore

Quick start

make up          # butane → ignition → vagrant up (kubeadm init auto-runs on boot)
make status      # wait for kubeadm-init to complete
make kubeconfig  # pull admin.conf → ./kubeconfig
make deploy      # atmos helmfile apply (Cilium + HAMI)

Setup

make box-add    # one-time: add Flatcar Vagrant box (select virtualbox)
make up         # butane -> ignition -> vagrant up

Prerequisites

• Ruby + Bundler (bundle install manages Vagrant via Gemfile)
• VirtualBox
• butane (Ignition config compiler)
• kubectl
• atmos CLI
• helmfile CLI