Tinkerbell

What it does

BMC management

Rufio manages baseboard management controllers and can execute one-off actions to bring physical machines to a desired state.checked 5 Oct 2026

Cluster API integration

Tinkerbell provides a Cluster API infrastructure provider that lets it serve as a bare-metal provider for Cluster API.checked 10 Oct 2026

Community support

The project directs users to its documentation, GitHub organization, and the #tinkerbell channel in the CNCF Slack workspace for help and participation.checked 5 Oct 2026

Core stack

The core stack consists of Smee, Tootles, Hook, and Tink.checked 10 Oct 2026

Custom actions

Actions can use any container image available to the worker.checked 5 Oct 2026

Deployment options

The installation guide documents Kubernetes, a standalone binary, and Docker Compose; Kubernetes is recommended for production use.checked 10 Oct 2026

Kubernetes integration

The Cluster API provider lets Tinkerbell act as a bare metal provider for Kubernetes cluster creation and management.checked 5 Oct 2026

Metadata

Tootles exposes hardware metadata over HTTP and can provide it in an AWS EC2 compatible format.checked 5 Oct 2026

Minimum resources

The architecture page lists minimum stack resources of 2 vCPUs, 4 GB RAM, and 300 MB disk.checked 10 Oct 2026

Network boot

Smee provides DHCP and TFTP services and serves iPXE and the initial installation image.checked 5 Oct 2026

Network requirement

The stack requires Layer 2 networking for its DHCP server, or a DHCP relay configured to forward traffic to the stack.checked 10 Oct 2026

Optional hardware management

Optional PBnJ and Rufio services communicate with baseboard management controllers.checked 10 Oct 2026

Production setup

The documentation recommends Kubernetes for production and specifies Kubernetes 1.26 or later as a prerequisite.checked 5 Oct 2026

Provisioning model

It uses an API-centric, declarative configuration and automation approach inspired by the Kubernetes community.checked 10 Oct 2026

Purpose

Tinkerbell is a bare-metal provisioning engine maintained by the Tinkerbell Community.checked 10 Oct 2026

Resource requirements

The documented minimum stack resources are 2 vCPUs, 4 GB RAM, and 300 MB disk, and target worker machines must support iPXE network boot.checked 5 Oct 2026

Setup limitation

The quick-start guide says its Vagrant and Libvirt path requires public internet and does not support air-gapped or proxied environments.checked 10 Oct 2026

Stack components

The core stack includes Tink, Smee, Tootles, and Hook, with optional power and boot services and a Cluster API provider.checked 5 Oct 2026

Support

The community page directs users to documentation, GitHub for bugs and feature ideas, and the CNCF Slack #tinkerbell channel for discussion.checked 10 Oct 2026

Target hardware

A target worker must support iPXE network boot and needs about 2 GB RAM to load HookOS and run a workflow; actual memory needs depend on action container images.checked 10 Oct 2026

Templates

Templates contain sequential tasks and actions, including actions such as streaming an image to disk, writing files, and partitioning disks.checked 5 Oct 2026

Workflow execution environment

HookOS is a lightweight in-memory operating system used to run Tink Worker and execute workflows.checked 10 Oct 2026

Workflow management

Workflows combine hardware and a template, are defined as Kubernetes custom resources, and are managed by tink-controller.checked 10 Oct 2026

Workflow model

A workflow combines hardware and a template for execution and is managed as a Kubernetes custom resource.checked 5 Oct 2026