Example: bare metal deployment¶

This example walks through deploying the full harvester profile (3-node Harvester HCI + Rancher Prime) on a bare metal Linux host. The same steps apply to large cloud VMs with nested virtualization enabled.

Host requirements¶

Resource	Minimum
OS	SLES 16 or Leap 16 (other Linux only with `RODEO_ALLOW_ANY_KVM_HOST=1`, unsupported)
RAM	64 GiB (3 × 16 GiB Harvester + 8 GiB Rancher + host overhead)
CPU	28 vCPU to spare
Disk	900 GiB free in `/var/lib/libvirt/images` (or a dedicated second disk)
KVM	`/dev/kvm` present; no nested virt needed on bare metal

Step 1: install rodeo-cli¶

On a clean SLES 16 / Leap 16 host:

curl -fsSL https://raw.githubusercontent.com/avaleror/rodeo-cli/main/install.sh | bash

This clones the repo, sets up a Python environment internally, and links rodeo as a system command. No venv to activate, no PATH to set.

To install KVM packages, libvirt, ansible-core, and kubectl separately:

rodeo install-deps

rodeo up (Step 3) runs this automatically if deps are missing.

Step 2: check the host¶

rodeo doctor

Expected output for a ready host:

Host facts
  ✓  root
  ✓  /dev/kvm
  ✓  nested virt        (not required on bare metal, shown as warning if absent)
  ✓  RAM     120 GiB total, 100 GiB available
  ✓  disk    900 GiB free in /var/lib/libvirt/images
  ✓  ansible-playbook
  ✓  ansible-galaxy
  ✓  kubectl
  ✓  python: libvirt
  ✓  python: lxml

Recommended profile: harvester (60 GiB needed, 100 GiB available)

Step 3: deploy¶

rodeo up --profile harvester

rodeo up will: 1. Start a tmux session named rodeo-harvester so the deploy survives disconnects 2. Confirm the host checks pass 3. Generate ~/.rodeo/secrets.yaml with random credentials 4. Seed the lab directory into ~/rodeo-labs/harvester/ 5. Self-escalate with sudo and start the deploy pipeline 6. Print login info when done

If your SSH session drops, re-attach to the running deploy:

tmux attach -t rodeo-harvester

Detach without stopping the deploy: Ctrl+b d.

Open a second tmux window (Ctrl+b c) to watch progress while the deploy runs:

rodeo watch         # split-panel TUI: phases + serial logs
rodeo logs harvester1   # or tail a single node

Total time: 90–150 minutes (most of it is Harvester iPXE install on nested KVM; bare metal is significantly faster).

Step 4: log in¶

After rodeo up finishes:

Harvester UI: https://<host-ip>:8443
Rancher UI: https://<host-ip>:30002
Username: admin
Password: shown in the success output, or check ~/.rodeo/secrets.yaml

Using a dedicated second disk¶

If your host has a separate data disk (check with lsblk), configure it in rodeo-plan.yaml:

storage:
  device: /dev/nvme1n1    # your data disk: verify with lsblk first
  image_dir: /var/lib/libvirt/images

The kvm_host Ansible role will prepare and mount it before provisioning VMs.

Redeploying after a clean¶

rodeo clean --all --yes --secrets   # destroy everything, wipe credentials
rodeo up --profile harvester        # fresh deploy with new credentials

Multi-disk host with separate OS and data disks¶

Typical bare metal server: OS on /dev/nvme0n1, data on /dev/nvme1n1.

# rodeo-plan.yaml
storage:
  device: /dev/nvme1n1
  mount_point: /var/lib/libvirt/images
  image_dir: /var/lib/libvirt/images
  fs_type: xfs

The kvm_host role partitions the device, creates an XFS filesystem, and mounts it at image_dir if it is not already mounted.

Day-2 operations¶

Task	Command
Health check	`rodeo status`
SSH into a node	`rodeo ssh harvester1`
Tail serial log	`rodeo logs harvester2`
Graceful stop	`rodeo stop --all --yes`
Start after stop	`rodeo start --all --yes`
Destroy the lab	`rodeo clean --yes`
Full host reset	`rodeo clean --all --yes --secrets`