Scaleway

​

Prerequisites

• Enable block storage on your Scaleway account (Scaleway will only allow snapshots from their block storage, not URLs)
• Configure the scw CLI to access your account (optional - you can use the console instead)
• Have qemu-img and wget installed for image conversion

​

Image preparation

1. Download the image disk of the Talos version you wish to run:

You can create your own brew on Talos Factory if you need a custom image. A working extra kernel command line arguments is as follow : -console talos.platform=scaleway console=ttyS0,115200 talos.dashboard.disabled=0

2. Decompress and convert the image:
   zstd --decompress scaleway-amd64.raw.zst
   qemu-img convert -O qcow2 scaleway-amd64.raw scaleway-amd64.qcow2
   
3. Create an S3 bucket (if it doesn’t exist): Go to Object Storage in the Scaleway console and create a new bucket.

4. Upload to S3-compatible object storage: Use the Scaleway console Object Storage interface to upload the QCOW2 file directly.

​

Snapshot creation

1. Go to the Scaleway Web console.
2. Go to the QCOW2 object in your S3 bucket.
3. Click on the extra action Import object as snapshot.

4. Name the snapshot scaleway-amd64-v, and use a Local Storage snapshot type.

​

Instance deployment

1. Create instance using the snapshot/image via GUI, CLI, or Infrastructure as Code tools.
2. Open the Security Group in order to open TCP/50000 and TCP/6443.
3. The Talos disk image used to boot is 4.5GB, and is not enough for the EPHEMERAL partition.
4. Add an extra Disk of 25GB (Local Storage) for Talos EPHEMERAL partition. The Talos OS (4.5GB) will be on /dev/vda and the extra disk for EPHEMERAL partition will be on /dev/vdb. At this point, you can debug and communicate with the instance on —insecure mode :

VM_IP=<instance_public_ip>
talosctl -n $VM_IP version --insecure
talosctl -n $VM_IP get disks --insecure
talosctl -n $VM_IP get links --insecure

​

Talos configuration

# controlplane.patch.yaml
machine:
    install:
        disk: /dev/vdb
        wipe: true
    network:
        interfaces:
            - interface: eth0
              dhcp: true

cluster:
  allowSchedulingOnControlPlanes: true
  apiServer:
    extraArgs:
      "advertise-address": "0.0.0.0"
    apiServer:
        certSANs:
            - 127.0.0.1
            - <public_ip>

talosctl gen config talos-cluster https://$VM_IP:6443 --config-patch=@controlplane.patch.yaml --output-dir _out/ --with-docs=false --with-examples=false

---
apiVersion: v1alpha1
kind: VolumeConfig
name: EPHEMERAL
provisioning:
  diskSelector:
    match: disk.dev_path == '/dev/vdb'
  minSize: 2GB
  maxSize: 40GB
  grow: false

talosctl --nodes $VM_IP apply-config --file _out/controlplane.yaml --insecure

talosctl --talosconfig=./_out/talosconfig --nodes $VM_IP -e $VM_IP version
talosctl --talosconfig=./_out/talosconfig --nodes $VM_IP -e $VM_IP dashboard

​

Talos cluster bootstrap

talosctl --talosconfig=./_out/talosconfig -e $VM_IP -n $VM_IP bootstrap

​

Notes

• Because the STATE partition will be written to /dev/vda during the installation, the instance works correctly with Scaleway’s reboot functionality
• talosctl reset performs the reset operation but don’t start the maintenance mode automatically
• If you give a bigger disk (let’s say 25GB) for the Talos disk image for the Scaleway Instance, unfortunately Talos won’t be able to expand its volume to 25GB and still see 4.5GB which is not enough for the EPHEMERAL partition