Skip to content

Quick start — on the Proxmox host

The simplest way: run everything on the Proxmox host itself. No separate machine, no GitHub runner.

1. Get the code

ssh root@<proxmox-host>
git clone https://github.com/jwklijnsma/shift-mox /root/shiftmox
cd /root/shiftmox

Private repo

GitHub doesn't accept account passwords for git. Give the host a read-only deploy key so ocp.sh can git pull itself.

2. Describe your cluster

$EDITOR group_vars/all/cluster.yml

The minimum to change: cluster_name, base_domain, the network (machine_network, gateway, dns_servers, api_vip, ingress_vip, ip_start), and Proxmox (proxmox_nodes, proxmox_bridge, proxmox_disk_storage). See cluster.yml.

3. Deploy

./ocp.sh deploy

The first run:

  • installs git, tmux, nmstate and a separate Ansible venv in /opt/ansible-venv (the hypervisor's own Python stays untouched);
  • asks once for the pull secret and the Proxmox token secret (./ocp.sh token creates the token);
  • creates ~/.ssh/id_ed25519 if you don't have one;
  • runs the install inside a tmux session ocp, so a dropped SSH connection doesn't stop it. Logs go to .work/.

A full install takes about 1–2 hours. At the end you get:

./ocp.sh status     # nodes, version, unhealthy operators, console URL, kubeadmin password

4. Use the cluster

export PATH=/root/shiftmox/.work/bin/<version>:$PATH
export KUBECONFIG=/root/shiftmox/.work/<cluster>/kubeconfig
oc get nodes
cat .work/<cluster>/kubeadmin-password

5. Optional: worker autoscaling

./ocp.sh autoscale

Then scale in the console under Compute › MachineSets › Edit Machine count, or let the ClusterAutoscaler do it. See Autoscaling.