cluster-config.yaml.example

# OpenShift Cluster Configuration
# All fields are required — missing keys will cause an error.

ocp_version: "4.20"
pull_secret_path: ~/keys/pull-secret.json

cluster_name: ocp
domain: example.com
ctlplanes: 1
workers: 0

ctlplane:
  numcpus: 6
  memory: 18432

worker:
  numcpus: 4
  memory: 16384

disk_size: 120
network: default
api_ip: "192.168.122.253"

remote:
  host: null
  user: root
  ssh_key_path: null
  libvirt_pool_path: null   # optional: force a specific libvirt storage pool directory. Leave null to auto-select the mount with the most free space
  min_free_space_gb: 100    # minimum free space (GB) required for a storage location to be considered acceptable

pci_devices: []
wait_timeout: 3600
version_channel: stable

# VM snapshot caching: restore a pre-built OCP cluster instead of deploying from scratch.
# When enabled, the first job for an OCP version does a full deploy + base operator install
# (NFD, KMM, MachineConfig), takes a VM snapshot, then continues. Subsequent jobs for the
# same OCP version restore the snapshot (~5 min instead of ~50 min).
snapshot:
  enabled: false          # set to true to enable snapshot caching
  max_cached: 3           # max number of OCP version snapshots to keep on the host

# Must-gather diagnostics
must_gather:
  artifact_dir: ./must-gather-output   # local directory to store collected diagnostics

# Post-install: install AMD GPU Operator and dependencies (NFD, KMM) per OLM doc
# https://instinct.docs.amd.com/projects/gpu-operator/en/release-v1.4.1/installation/openshift-olm.html
operators:
  install: false
  gpu_operator_version: "1.4"   # AMD GPU Operator minor version (e.g. "1.4"); resolves to latest patch (e.g. 1.4.1)
  machine_config_role: worker   # use "master" for Single Node OpenShift (SNO)
  driver_version: "31.30"
  enable_metrics: true