AMD CI

Continuous Integration for AMD GPU Operator on OpenShift.

OpenShift Cluster Provisioner

Deploy OpenShift clusters using kcli on local or remote libvirt hosts.

Quick Start

# 1. Copy the example config
cp cluster-config.yaml.example cluster-config.yaml

# 2. Edit with your settings
vim cluster-config.yaml

# 3. Deploy
make cluster-deploy CONFIG_FILE_PATH=cluster-config.yaml

Configuration File

Create a YAML config file with your cluster settings:

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

# Optional
cluster_name: ocp
ctlplanes: 1
workers: 0

Required Fields

Field	Description
ocp_version	OpenShift version (e.g., "4.20" or "4.20.6"). If only major.minor, latest patch is used.
pull_secret_path	Path to Red Hat pull secret. Get it from https://console.redhat.com/openshift/install/pull-secret

Optional Fields

Field	Default	Description
cluster_name	ocp	Name of the cluster
domain	example.com	Cluster domain
ctlplanes	1	Number of control plane nodes (1 = SNO)
workers	0	Number of worker nodes
ctlplane.numcpus	6	vCPUs per control plane
ctlplane.memory	18432	Memory (MB) per control plane
worker.numcpus	4	vCPUs per worker
worker.memory	16384	Memory (MB) per worker
disk_size	120	Disk size (GB) per node
network	default	Libvirt network name
api_ip	192.168.122.253	API VIP address
pci_devices	[]	PCI devices for GPU passthrough
wait_timeout	3600	Timeout (seconds) waiting for cluster ready
no_wait	false	Skip waiting for cluster ready
version_channel	stable	OCP release channel (stable, fast, candidate)

Local Deployment

Deploy on the local machine (requires libvirt/kcli installed):

# cluster-config.yaml
ocp_version: "4.20"
pull_secret_path: ~/keys/pull-secret.json
cluster_name: my-cluster

make cluster-deploy CONFIG_FILE_PATH=cluster-config.yaml

Remote Deployment

Deploy on a remote libvirt host via SSH:

# cluster-config.yaml
ocp_version: "4.20"
pull_secret_path: ~/keys/pull-secret.json
cluster_name: my-cluster

remote:
  host: myserver.example.com
  user: root
  ssh_key_path: ~/.ssh/id_rsa

make cluster-deploy CONFIG_FILE_PATH=cluster-config.yaml

GPU Passthrough

Pass PCI devices (GPUs) to cluster nodes:

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

pci_devices:
  - "0000:b3:00.0"
  - "0000:b3:00.1"

Multi-Node Cluster

Deploy HA cluster with multiple control planes and workers:

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

ctlplanes: 3
workers: 2

ctlplane:
  numcpus: 8
  memory: 32768

worker:
  numcpus: 16
  memory: 65536

Commands

# Deploy cluster
make cluster-deploy CONFIG_FILE_PATH=cluster-config.yaml

# Delete cluster
make cluster-delete CONFIG_FILE_PATH=cluster-config.yaml

# Dry run (show config without deploying)
make cluster-dry-run CONFIG_FILE_PATH=cluster-config.yaml

# Show help
make help

Requirements

• Python 3.10+
• kcli installed (local or remote)
• libvirt configured
• Red Hat pull secret