What Is Bare Metal Provisioning? Tools, Architecture, and Automation Explained

Bare Metal Provisioning: How It Works, Which Tools Run It, and Why Speed Matters

Every bare metal server starts as unformatted hardware. Bare metal provisioning is the automated process that takes it from first power-on to production-ready state: network boot, OS install, firmware configuration, and post-install automation, no manual intervention required.

This guide defines bare metal provisioning, maps the four-stage architecture, covers the key open source tools (OpenStack Ironic, Foreman, Ansible, Terraform, Razor, MAAS, CloudStack), explains hybrid cloud bare metal provisioning, and addresses why provisioning speed is an operational constraint, not a convenience.

What Is Bare Metal Provisioning?

Bare metal provisioning is the automated process of preparing a physical server for production use: PXE network boot, OS installation directly onto hardware, firmware and BIOS configuration, and post-install scripts that bring the server to a known, repeatable state without a technician touching the machine.

Manual provisioning does not scale past one server. Automated bare metal provisioning replaces every manual step with API-driven operations that can bring a full rack from bare hardware to a configured production cluster in under 30 minutes.

Bare metal provisioning is distinct from bare metal imaging. Imaging clones a golden OS snapshot onto identical hardware. Provisioning builds server state from a defined configuration, handling variable hardware profiles, OS versions, and post-install requirements in the same pipeline.

Need instant bare metal provisioning across 45+ global PoPs?
‍‍‍
Explore NetActuate Bare Metal Servers

What Are the Four Stages of Bare Metal Provisioning Architecture?

Bare metal server provisioning runs in four stages: hardware discovery via BMC/IPMI/Redfish, PXE network boot to load a provisioning agent, automated OS installation with disk partitioning, and post-install configuration via Ansible or cloud-init to reach production-ready state.

Stage 1: Hardware Discovery. The provisioning platform queries the Baseboard Management Controller (BMC) via IPMI or Redfish to inventory CPU, RAM, storage, and NIC configuration. Hardware health is validated before OS installation begins.

Stage 2: PXE Network Boot. The server boots from the network via PXE or iPXE. DHCP directs it to a TFTP server hosting the boot image. A lightweight provisioning kernel loads into memory, handing control to the provisioning platform. A dedicated management network isolated from production traffic is required.

Stage 3: OS Installation. The platform deploys the target OS: Linux (RHEL, Ubuntu, Rocky, Debian) or Windows Server. Disk partitioning, bootloader config, and initial network setup run from a pre-defined server profile via Kickstart, preseed, or Autounattend.xml answer files.

Stage 4: Post-Install Configuration. Configuration management tools apply the final state. Ansible connects via SSH with no agent and runs idempotent playbooks that configure DNS, install packages, create users, set up monitoring agents, and run validation tests. The server is production-ready when all tasks pass.

What Are the Main Bare Metal Provisioning Tools?

The main open source bare metal provisioning tools are OpenStack Ironic, Foreman, Ansible, Terraform, Razor, MAAS, and CloudStack. Each addresses a different layer: discovery, OS deployment, post-install config, or declarative infrastructure as code.
‍

Tool

Type

Primary Use

OpenStack Ironic

Open source platform

Cloud-scale provisioning via Nova API; IPMI/Redfish driver backends

Foreman

Open source lifecycle mgmt

Full server lifecycle: discover, provision, configure, monitor

Ansible

Agentless config management

Post-boot OS configuration via SSH playbooks; integrates with Ironic and Foreman

Terraform

Infrastructure as Code

Declarative bare metal provisioning via provider API (netactuate Terraform provider)

Razor

Policy-based discovery tool

Auto-discovers unprovisioned hardware; assigns OS by policy rule

MAAS (Canonical)

Open source platform

Ubuntu-native PXE provisioning; fast network-based deploy

CloudStack

Open source cloud platform

Multi-hypervisor environments; bare metal zones alongside VM compute

OpenStack Ironic bare metal provisioning exposes physical servers through the same Nova API used for VMs, so operators manage physical and virtual compute identically. Ansible bare metal provisioning handles post-boot state via agentless SSH playbooks. Terraform bare metal provisioning uses the netactuate provider to manage physical nodes declaratively in HCL, the same workflow used for cloud resources.

Provision bare metal via Terraform and Ansible on NetActuate's global platform.
‍‍‍
View automation and API documentation

How Does Hybrid Cloud Bare Metal Provisioning Work?

Hybrid cloud bare metal provisioning combines dedicated physical servers and cloud VMs under one control plane. Bare metal handles performance-critical and compliance-sensitive workloads. Cloud VMs handle elastic and burst capacity. The same provisioning APIs manage both layers.

Performance-critical workloads run on bare metal: database primaries, AI training, and media encoding nodes that need consistent CPU cycles and direct NIC access. Stateless tiers, management tooling, and burst capacity run as virtual machines or containers on the managed cloud platform.

Teams evaluating cloud repatriation commonly land on this model: I/O-intensive workloads move back to bare metal for cost and performance while cloud-native tooling continues managing deployment pipelines. NetActuate runs both tiers across the same PoP footprint with BGP Anycast, manageable through a single portal and API.

Why Does Bare Metal Provisioning Speed Matter?

Provisioning speed determines how fast a bare metal fleet responds to capacity events, hardware failures, or workload migrations. A platform that provisions in minutes rather than hours eliminates bare metal's historical disadvantage versus cloud VMs on deployment velocity.

Legacy provisioning required a technician to rack a server, insert installation media, and configure the machine manually, measured in days. Automated provisioning collapses that to under 10 minutes for a pre-racked server with a working BMC connection.

For disaster recovery architectures, provisioning speed is a direct RTO input. A bare metal fleet that provisions in 8 minutes versus 48 minutes changes the recovery design. Fast provisioning at a distributed PoP footprint means workloads rebuild at a secondary location without pre-provisioned standby hardware sitting idle.

Deploy bare metal globally in minutes across 45+ PoPs.
‍‍‍
Talk to a NetActuate infrastructure specialist

Frequently Asked Questions

What is the difference between bare metal provisioning and bare metal imaging?

Bare metal imaging clones a golden OS snapshot onto identical hardware. Bare metal provisioning builds server state from a defined configuration, handling variable hardware and OS requirements. Provisioning is the correct approach for heterogeneous fleets; imaging breaks when hardware differs across nodes.

What is the role of PXE in bare metal server provisioning?

PXE (Preboot Execution Environment) is the network boot protocol that allows a server with no OS to boot from a TFTP server over the network. It is the entry point of every automated bare metal provisioning pipeline. iPXE extends PXE with HTTP booting, scripting, and UEFI support for modern firmware environments.

Can bare metal provisioning run on Windows Server?

Yes. Windows bare metal OS provisioning uses Autounattend.xml answer files for unattended installation. Foreman supports Windows provisioning via WinRM. Ansible configures post-boot Windows state using the windows_core role collection. OpenStack Ironic has limited Windows support depending on the driver backend in use.

What is the difference between Ironic and Foreman for bare metal provisioning?

OpenStack Ironic is purpose-built for cloud-scale bare metal provisioning within the OpenStack ecosystem, provisioning physical servers through the Nova API. Foreman is a standalone lifecycle management tool that handles provisioning, configuration, and monitoring without requiring OpenStack, making it the better choice for non-OpenStack environments.

What is Razor bare metal provisioning?

Razor is a policy-based auto-discovery tool that boots a lightweight microkernel on unprovisioned servers to inventory hardware, then matches each server against a policy to automatically assign an OS and configuration. It eliminates the manual step of assigning an OS to each new server and is well-suited for large fleets where hardware arrives at variable times.

Conclusion

Bare metal provisioning is the automation layer that makes physical servers competitive with cloud VMs on deployment velocity without giving up the performance advantage of dedicated hardware. PXE boot, OS installation, and Ansible post-install configuration can bring a server from power-on to production in under 10 minutes.

OpenStack Ironic, Foreman, Terraform, and Ansible each address a different layer of the provisioning stack. Hybrid cloud architectures route workloads to bare metal or virtual compute based on performance and cost constraints from the same control plane.

NetActuate's edge infrastructure delivers automated bare metal provisioning across 45+ global PoPs, with BGP Anycast routing, Terraform and Ansible automation support, and a single API for physical and virtual compute management.