Cluster API Provider for Managed Bare Metal Hardware

This repository contains a Machine actuator implementation for the Kubernetes Cluster API.

For more information about this actuator and related repositories, see metal3.io.

Development Environment

API

See the API Documentation for details about the providerSpec API used with this cluster-api provider.

MachineSet Scaling

If you would like a MachineSet to be automatically scaled to the number of matching BareMetalHosts, annotate that MachineSet with key metal3.io/autoscale-to-hosts and any value.

When reconciling a MachineSet, the controller will count all of the BareMetalHosts that either:

  • match the MachineSet's Spec.Template.Spec.ProviderSpec.HostSelector and have a ConsumerRef that is nil
  • has a ConsumerRef that references a Machine that is part of the MachineSet

This ensures that in case a BareMetalHost has previously been consumed by a Machine, but either labels or selectors have since been changed, it will continue to get counted with the MachineSet that its Machine belongs to.

Machine Remediation

MachineHealthCheck Controller in Machine-API operator is checking Node's health. If you would like to remediate unhealthy Machines you should add the following to MachineHealthCheck CR:

annotations:
   machine.openshift.io/remediation-strategy: external-baremetal

Remediation is done by power-cycling the Host associated with the unhealthy Machine, using BMO reboot API

Remediation steps (triggered by the annotation mentioned above):

  1. Power off the host
  2. Add poweredOffForRemediation annotation to the unhealthy Machine
  3. Delete the node
  4. Power on the host
  5. Wait for the node the come up (by waiting for the node to be registered in the cluster)
  6. Remove poweredOffForRemediation annotation and the MAO's machine unhealthy annotation