Skip to main content
Version: 0.7 (Next)

Hardware & Provisioning

Alpha

Civo Bare Metal is alpha functionality. See the overview for pricing notes.

The Hardware view shows the physical servers (assets) at your connected sites and where each sits in its lifecycle — from first BMC contact to running a cluster node.

From racked server to ready asset​

Colony discovers and manages hardware out-of-band through each server's BMC (Redfish or IPMI) and provisions it over the network with PXE boot — no keyboard, no USB stick, no per-machine OS install.

A server moves through a defined lifecycle:

Pending → BMCDiscovering → Creating → Discovered → Provisioning → Ready
  • BMCDiscovering — Colony contacts the BMC with its credentials and reads MAC addresses, serial, and model
  • Discovered — the server is registered with the provisioning stack and ready to receive an OS
  • Provisioning — the server PXE-boots into a minimal environment (HookOS) and the OS install workflow runs (typically 5–15 minutes)
  • Ready — OS installed and operational

Full state machine, re-provisioning, and decommissioning: Hardware lifecycle.

How provisioning works under the hood​

Colony runs as a Kubernetes operator and delegates the low-level provisioning machinery to Tinkerbell: DHCP/PXE proxying, metadata service, workflow execution, and an artifact server for OS images. Everything is declared as Kubernetes resources:

  • ColonyHardware — registers a physical server (BMC address, credentials, MACs, static IP)
  • ProvisionPlan — declares the OS install intent for one or more servers (image, version, networking mode, disk wipe)
  • ProvisionWorkflow — tracks the execution against each individual server

Because these are plain Kubernetes manifests, the entire estate is GitOps-compatible: store hardware inventory and provisioning intent in Git and let Argo CD or Flux reconcile it. Applying the same manifest twice is safe.

Requirements at the site​

In brief — see Prerequisites for the full list:

  • Servers with a reachable BMC (Redfish or IPMI) and UEFI PXE boot enabled
  • Layer-2 connectivity on the provisioning VLAN (direct or via DHCP relay)
  • A management cluster for the Colony control plane (K3s recommended; 4 cores / 8 GB RAM minimum)

Air-gapped environments are fully supported — OS images and container images are served from a local artifact server and registry: Air-gap guide.

Learn more​