QEMU/QCOW2 Installation Guide

Installation Guide | QEMU/QCOW2

This guide provides step-by-step instructions for deploying RLC Pro AI using QEMU and QCOW2 images. It covers prerequisites, downloading and verifying images, setting up a KVM hypervisor, customizing the initial instance with cloud-init, and verifying the deployment.

Prerequisites

CIQ Portal access

This guide assumes you have followed the instructions in the Getting Started guide to perform the following tasks:

Log in to CIQ Portal and download the installer files (QCOW2 image and checksum file).
QCOW2 image downloaded and checksum verified.

Hardware requirements

This guide assumes your hardware meets the requirements.

KVM RLC Pro AI deployment

Virtual machine (VM) images are files that contain a complete operating system and its applications. They are commonly stored in disk formats such as *.qcow2, *.img, or *.raw.

Some common methods for deploying operating systems using VM images include:

Using a local hypervisor: Kernel-based Virtual Machine (KVM) is the standard hypervisor on Linux distributions. Applications such as QEMU, virt-install, and virsh are used for running and managing virtual machines on KVM. Proxmox VE and other KVM-based platforms consume the same QCOW2 images.
Using a cloud platform: Cloud platforms such as AWS, Microsoft Azure, and Google Cloud provide services for creating and managing virtual machines, and RLC Pro AI is available directly from their marketplaces.

The following section provides instructions for deploying an RLC Pro AI image on a KVM hypervisor.

Deploying RLC Pro AI on a KVM hypervisor

Initial tool installation and hypervisor subsystem setup

Install the necessary tools:


sudo dnf install qemu-kvm libvirt virt-manager virt-install

Start the libvirtd service and enable it for automatic startup:
```
sudo systemctl enable --now libvirtd
```

Initial RLC Pro AI instance customization (cloud-init)

Cloud-init automates the initialization of cloud instances during their first boot. It lets you define user accounts, set passwords, configure SSH access, and perform other bootstrap tasks automatically.

To automatically create a user named testuser with the password ciq_4_YOU, follow these steps:

Use a HereDoc to generate the cloudinit-userdata.yaml file:
```
cat <<EOF > cloudinit-userdata.yaml
#cloud-config
users:
  - name: testuser
    password: ciq_4_YOU
    chpasswd: { expire: False }
    groups: wheel
    shell: /bin/bash
    sudo: ["ALL=(ALL) NOPASSWD:ALL"]

ssh_pwauth: true
EOF
```
Explanation of cloud-init directives:

users:
- Creates a new user named testuser.
- Sets the password to ciq_4_YOU.
- Adds the user to the wheel group for sudo privileges.
- Assigns /bin/bash as the default shell.
- Grants passwordless sudo access.
ssh_pwauth: true
- Enables SSH password authentication.
Confirm that the file was created correctly:
```
cat cloudinit-userdata.yaml
```
Note
For more information about cloud-init see the full cloud-init documentation

Launch the RLC Pro AI virtual machine

Use the virt-install tool to launch the QCOW2 image:
```
sudo virt-install \
  --name rlc-pro-ai \
  --memory 4096 \
  --vcpus 4 \
  --disk path=/path/to/RLC-PRO-AI-<release-date>-GenericCloud-x86_64.qcow2 \
  --cloud-init user-data="/path/to/cloudinit-userdata.yaml" \
  --import \
  --os-variant rhel9.0 \
  --network network=default \
  --graphics none \
  --console pty,target_type=serial
```
Explanation of virt-install options:
- --name: Sets the virtual machine's name.
- --memory: Specifies the amount of RAM.
- --vcpus: Defines the number of virtual CPUs.
- --disk: Specifies the path and format of the VM disk.
- --import: Directly imports the specified disk image.
- --os-variant: Optimizes the VM configuration for Rocky Linux 9.
- --network: Connects the VM to the default virtual network.
- --graphics none: Disables the graphical interface.
- --console: Enables a serial console interface for management.
Note
To give the VM direct access to an NVIDIA GPU, configure PCI passthrough (VFIO) for the GPU device and add it to the VM. Without a GPU, RLC Pro AI runs AI workloads on the CPU with no configuration changes.

Once the VM is launched, it can be accessed and managed using virt-manager or the command line.

Verifying the deployment


hostnamectl

Expected output (abbreviated):


Operating System: Rocky Linux from CIQ Pro AI 9.7 (Blue Onyx)
     CPE OS Name: cpe:2.3:o:ciq:rocky_linux_from_ciq_pro_ai:9.7
          Kernel: Linux 6.12.74-1.1.0.0.el9_clk.x86_64
    Architecture: x86-64

The el9_clk kernel suffix confirms the CIQ Linux Kernel is active. You can also confirm the AI stack is ready:


python3 -c "import torch; print(torch.__version__)"


2.8.0.post12

Post-Install

Once installation is complete, proceed to the Post-Installation Guide for enrollment and updates.

Getting Started with RLC Pro AI