Install a Hypervisor Host

A hypervisor host is a Talos Linux machine that runs virtual machines. To set one up, this page builds a Talos Linux image with the libvirtd system extension, installs it on a server with the configuration files that the extension’s services need, connects kubectl to the cluster, and checks that the host is ready.

This page installs Talos Linux on a server from an ISO. To use a cloud instance or a virtual machine on another hypervisor as the host, install Talos Linux with that platform’s guide, such as AWS, GCP, or Proxmox, using a Talos Linux image that includes the libvirtd extension. Add the files from Step 3 to the machine configuration with talosctl patch machineconfig --nodes <host-ip> --patch hypervisor-users.yaml, then continue from Step 7.

Before you begin

The steps on this page require:

A server that meets the requirements.
talosctl v1.15 or later.
kubectl, to access the cluster in Step 6.

Step 1: Build an image with the `libvirtd` extension

Create a Talos Linux ISO that includes the libvirtd extension.

Image Factory builds Talos Linux images with system extensions and identifies each combination of extensions by a schematic ID. The following steps use the Image Factory API to create a schematic with the libvirtd extension and download its Talos Linux ISO.

Run the following command to create schematic.yaml, a schematic that adds the libvirtd extension:

cat <<'EOF' > schematic.yaml
customization:
  systemExtensions:
    officialExtensions:
      - siderolabs/libvirtd
EOF

Submit the schematic to Image Factory:

curl -X POST --data-binary @schematic.yaml https://factory.talos.dev/schematics

Image Factory responds with the schematic ID in the id field:

{"id":"84feec1f502dbb1340c6a832821754ae5b4386d7349e86e8ae0b862bc9148197","schematic":"customization:\n    systemExtensions:\n        officialExtensions:\n            - siderolabs/libvirtd\n"}

The schematic ID is derived from the schematic’s contents, so this schematic always returns the same ID.

Save the schematic ID, the Talos Linux version, and the server’s architecture in variables. Replace <schematic-id> with the ID from the response, <talos-version> with v1.15.0-alpha.1 or later, and <arch> with amd64 or arm64:
export SCHEMATIC_ID=<schematic-id> export TALOS_VERSION=<talos-version> export ARCH=<arch>

Download the ISO that the server boots from:

curl -LO https://factory.talos.dev/image/${SCHEMATIC_ID}/${TALOS_VERSION}/metal-${ARCH}.iso

The following steps use the Image Factory UI to create a schematic with the libvirtd extension and download its Talos Linux ISO.

Open the Image Factory UI.
On the Hardware Type screen, select Bare-metal Machine.
On the Choose Talos Linux Version screen, select v1.15.0-alpha.1 or later.
On the Machine Architecture screen, select the server’s architecture, amd64 or arm64.
On the System Extensions screen, select siderolabs/libvirtd.
On the Customization screen, leave the defaults.
On the Schematic Ready screen, save the schematic ID and the Talos Linux version in variables. Replace <schematic-id> with the ID shown on the screen, and <talos-version> with the version you selected:
export SCHEMATIC_ID=<schematic-id> export TALOS_VERSION=<talos-version>
Download the ISO listed under First Boot. The server boots from it in Step 2.

Step 2: Boot the server from the ISO

The following steps boot the server into maintenance mode and prepare it for installation.

Boot the server from the ISO downloaded in Step 1. Talos Linux starts in maintenance mode, which means it runs from memory and waits for a machine configuration.
Save the IP address shown on the server’s console in a variable. If no display is connected, find the address in your DHCP server’s leases:
export HOST_IP=<host-ip>
Remove the installation media by unplugging the USB drive or unmounting the ISO, so that it cannot be selected as the install disk.

Step 3: Create the configuration files the extension needs

Two of the libvirtd extension’s services, ext-virtqemud and ext-virtstoraged, start only after the host has /etc/passwd and /etc/group files. Talos Linux does not create these files, so they are added through the machine configuration with EtcFileConfig documents. Create a configuration patch that defines both files:

cat <<'EOF' > hypervisor-users.yaml
apiVersion: v1alpha1
kind: EtcFileConfig
name: passwd
mode: 0o644
contents: |
  root:x:0:0:root:/root:/bin/sh
  qemu:x:107:107:qemu:/var/lib/libvirt:/bin/false
  nobody:x:65534:65534:nobody:/:/bin/false
---
apiVersion: v1alpha1
kind: EtcFileConfig
name: group
mode: 0o644
contents: |
  root:x:0:
  kvm:x:36:qemu
  qemu:x:107:
  nobody:x:65534:
EOF

For more about adding files under /etc, see Managing User-Owned Files. The next step adds this file to the generated machine configuration, so every host installed from it includes the files.

Step 4: Install Talos Linux

Applying a machine configuration to the server installs Talos Linux to its disk, and the configuration’s installer image determines what is installed. The following steps generate a machine configuration that installs Talos Linux with the libvirtd extension and apply it to the server.

List the disks available on a server by running the following command against its IP address:
talosctl get disks --insecure --nodes $HOST_IP
Save the ID of the disk to install to, such as sda or nvme0n1, and a name for the cluster in variables:
export DISK_NAME=<disk-id> export CLUSTER_NAME=<cluster-name>
Generate the machine configuration. The --install-image flag points to the installer image built from your schematic, so the installed system includes the libvirtd extension, and the --config-patch flag adds the files from Step 3:
talosctl gen config $CLUSTER_NAME https://$HOST_IP:6443 \ --install-disk /dev/$DISK_NAME \ --install-image factory.talos.dev/metal-installer/${SCHEMATIC_ID}:${TALOS_VERSION} \ --config-patch hypervisor-users.yaml
The command writes the machine configuration files controlplane.yaml and worker.yaml, and a talosconfig file for talosctl.
Apply the control plane configuration to the server, which makes it the cluster’s control plane. Talos Linux installs to the disk and reboots:
talosctl apply-config --insecure --nodes $HOST_IP --file controlplane.yaml
Point talosctl at the generated talosconfig file, and set the server as its endpoint:
export TALOSCONFIG=$(pwd)/talosconfig talosctl config endpoints $HOST_IP
After the server finishes rebooting, bootstrap the cluster. Run this command once:
talosctl bootstrap --nodes $HOST_IP

Step 5: Add worker hosts (optional)

To run virtual machines on more than one server, add servers to the cluster as workers. Each worker is a hypervisor host with its own virtual machines.

Boot each server from the ISO downloaded in Step 1. After it boots, remove its installation media.
Save the servers’ IP addresses in a variable:
WORKER_IP=("<worker-ip-1>" "<worker-ip-2>")
Check each server’s disks. The worker configuration installs Talos Linux to /dev/$DISK_NAME, the same disk as on the first server:
for ip in "${WORKER_IP[@]}"; do talosctl get disks --insecure --nodes "$ip" done
If a server has no disk with that ID, create a patch that installs Talos Linux to a different disk. Replace <disk-id> with the ID of the disk to install to:
cat <<'EOF' > worker-disk.yaml apiVersion: v1alpha1 kind: UnattendedInstallConfig provisioning: diskSelector: match: disk.dev_path == "/dev/<disk-id>" EOF

Apply the worker configuration to each server that installs to the same disk as the first server:

for ip in "${WORKER_IP[@]}"; do
  talosctl apply-config --insecure --nodes "$ip" --file worker.yaml
done

For a server that needs the patch from the previous step, apply the worker configuration with the patch instead:

talosctl apply-config --insecure --nodes <worker-ip> --file worker.yaml --config-patch worker-disk.yaml

Step 6: Access the cluster with kubectl

After the cluster is bootstrapped, Kubernetes starts on the hosts. Download a kubeconfig file for the cluster and point kubectl at it:

talosctl kubeconfig ./kubeconfig --nodes $HOST_IP
export KUBECONFIG=$(pwd)/kubeconfig

To list the cluster’s nodes, run:

kubectl get nodes

The output lists each host as a node. Its STATUS column shows Ready once the node has finished starting, which can take a few minutes after bootstrap.

Step 7: Check that the hosts are ready

To confirm that the libvirtd extension is installed, run:

talosctl get extensions --nodes $HOST_IP

The NAME column lists libvirtd. The extension runs the ext-virtqemud service, which Talos uses to start virtual machines. To confirm that the service is running, run:

talosctl services --nodes $HOST_IP

The output lists ext-virtqemud with the state Running when the host is ready to run virtual machines. To check a worker host, run the same commands with its IP address.

Next steps

To create a virtual machine on the host, see Run your first virtual machine.

Overview

Getting Started

Storage

Virtual Machines

Before you begin

Step 1: Build an image with the `libvirtd` extension

Step 2: Boot the server from the ISO

Step 3: Create the configuration files the extension needs

Step 4: Install Talos Linux

Step 5: Add worker hosts (optional)

Step 6: Access the cluster with kubectl

Step 7: Check that the hosts are ready

Next steps

​Before you begin

​Step 1: Build an image with the libvirtd extension

​Step 2: Boot the server from the ISO

​Step 3: Create the configuration files the extension needs

​Step 4: Install Talos Linux

​Step 5: Add worker hosts (optional)

​Step 6: Access the cluster with kubectl

​Step 7: Check that the hosts are ready

​Next steps

Before you begin

Step 1: Build an image with the `libvirtd` extension

Step 2: Boot the server from the ISO

Step 3: Create the configuration files the extension needs

Step 4: Install Talos Linux

Step 5: Add worker hosts (optional)

Step 6: Access the cluster with kubectl

Step 7: Check that the hosts are ready

Next steps