Bare Metal Node Deployment¶

The 3 bare-metal nodes are deployed using appliance mode -- a two-stage ISO boot process that installs the OS, configures networking, wipes drives, and auto-registers each node with the Palette Management Appliance.

Deployment Mode: Appliance Mode¶

Appliance mode was selected over agent mode for the Toyota POC because:

OS upgrade demonstration is a POC success criterion -- appliance mode supports A/B partition upgrades
Consistent imaging -- each node gets an identical base OS with baked-in network config
Drive wipe -- the install ISO includes automated drive wiping to clear prior platform artifacts
Air-gapped -- no external network needed; all images pulled from PMA internal registry

Two-Stage Boot Process¶

graph LR
    ISO["Mount Install ISO<br/>(~1.8 GB)"] --> INSTALL["Auto-Install<br/>Wipe drives, install OS<br/>Configure LACP bond"]
    INSTALL --> OFF["Power Off"]
    OFF --> SITE["Mount Site<br/>User Data ISO"]
    SITE --> BOOT["Boot from Disk<br/>Apply hostname + IP"]
    BOOT --> REG["Auto-Register<br/>with Palette"]

    style ISO fill:#1F7A78,color:#fff
    style INSTALL fill:#005B5B,color:#fff
    style OFF fill:#043736,color:#fff
    style SITE fill:#005B5B,color:#fff
    style BOOT fill:#1F7A78,color:#fff
    style REG fill:#9EB277,color:#fff

Stage 1: Install ISO¶

A single install ISO is shared across all 3 nodes. It contains:

Ubuntu 24.04 base OS with Kairos
CanvOS v4.8.8 provider image
PXKe (kubeadm) 1.33.6 Kubernetes packages
LACP bond configuration for all 4 NICs
Drive wipe script -- runs sgdisk --zap-all on all 7TB drives
Legacy boot loader (UEFI not viable on this hardware)
Install target: /dev/sda (223.5GB boot drive)

The install ISO auto-runs without user interaction. After installation completes, the node powers off.

Stage 2: Site User Data ISO¶

Each node gets a unique site-user-data ISO containing:

Hostname (e.g., stg-wahvp004)
Static IP address (e.g., 10.25.233.4/24)
Gateway, DNS, NTP settings
Edge host registration token
Palette endpoint address

When the node boots from disk with the site-user-data ISO mounted, it reads the configuration on first boot only. Subsequent boots ignore the ISO. The node then registers with Palette automatically.

ISO Build: Production Install ISO¶

The production install ISO is built using CanvOS with customizations for the Toyota hardware.

Build Parameters¶

Parameter	Value
CanvOS Version	4.8.8
Base OS	Ubuntu 24.04
Kubernetes	kubeadm 1.33.6 (PXKe)
Boot Mode	Legacy (CSM)
Install Target	`/dev/sda`
Network	LACP bond (802.3ad) across 4 NICs
Drive Wipe	All drives > 5TB wiped on install and reset

CanvOS Build Configuration¶

.arg file (key parameters)

OS_DISTRIBUTION=ubuntu
OS_VERSION=24.04
K8S_DISTRIBUTION=kubeadm
ARCH=amd64
CUSTOM_TAG=toyota-tmna
IMAGE_REGISTRY=gcr.io/spectro-images-public/github/spectrocloud/CanvOS

user-data (baked into ISO)¶

user-data

#cloud-config
stylus:
  site:
    edgeHostToken: "OTVhZDVlMzc5NDI0YzVlYjgyYjg1YjQzNDg4NGQwZjc="
    paletteEndpoint: "10.25.232.155"
  registryCredentials:
    domain: "10.25.232.155:30003/spectro-content"
    username: "admin"
    password: "********"
    insecure: false

install:
  device: /dev/sda
  poweroff: true

users:
  - name: kairos
    passwd: kairos

stages:
  install:
    - name: "Wipe all large drives"
      commands:
        - |
          for disk in $(lsblk -dno NAME,SIZE | awk '$2 == "7T" {print "/dev/"$1}'); do
            sgdisk --zap-all "$disk"
            wipefs -a "$disk"
            dd if=/dev/zero of="$disk" bs=1M count=100
          done

  reset:
    - name: "Wipe all large drives on reset"
      commands:
        - |
          for disk in $(lsblk -dno NAME,SIZE | awk '$2 == "7T" {print "/dev/"$1}'); do
            sgdisk --zap-all "$disk"
            wipefs -a "$disk"
            dd if=/dev/zero of="$disk" bs=1M count=100
          done

  network:
    - name: "Configure LACP bond"
      commands:
        - |
          mkdir -p /etc/systemd/network
          # Bond NetDev
          cat > /etc/systemd/network/10-bond0.netdev << 'EOF'
          [NetDev]
          Name=bond0
          Kind=bond
          [Bond]
          Mode=802.3ad
          LACPTransmitRate=fast
          TransmitHashPolicy=layer3+4
          MIIMonitorSec=100ms
          EOF
          # NIC members
          for nic in enp134s0f0np0 enp134s0f0np1 enp175s0f0np0 enp175s0f0np1; do
            cat > /etc/systemd/network/10-${nic}.network << NICEOF
          [Match]
          Name=${nic}
          [Network]
          Bond=bond0
          NICEOF
          done
          # Bridge
          cat > /etc/systemd/network/20-br0.netdev << 'EOF'
          [NetDev]
          Name=br0
          Kind=bridge
          [Bridge]
          STP=false
          ForwardDelaySec=0
          EOF
          cat > /etc/systemd/network/20-bond0.network << 'EOF'
          [Match]
          Name=bond0
          [Network]
          Bridge=br0
          EOF

Site User Data ISOs¶

Three separate site-user-data ISOs are created, one per node:

STG-WAHVP004STG-WAHVP005STG-WAHVP006

site-user-data-004.yaml

#cloud-config
stylus:
  site:
    name: "stg-wahvp004"
    network:
      interfaces:
        br0:
          type: bridge
          ipAddress: "10.25.233.4/24"
          gateway: "10.25.233.254"
          dns:
            - "10.213.129.245"
          ntp:
            - "10.53.12.18"

site-user-data-005.yaml

#cloud-config
stylus:
  site:
    name: "stg-wahvp005"
    network:
      interfaces:
        br0:
          type: bridge
          ipAddress: "10.25.233.5/24"
          gateway: "10.25.233.254"
          dns:
            - "10.213.129.245"
          ntp:
            - "10.53.12.18"

site-user-data-006.yaml

#cloud-config
stylus:
  site:
    name: "stg-wahvp006"
    network:
      interfaces:
        br0:
          type: bridge
          ipAddress: "10.25.233.6/24"
          gateway: "10.25.233.254"
          dns:
            - "10.213.129.245"
          ntp:
            - "10.53.12.18"

Node Imaging Procedure¶

Repeat for each of the 3 bare-metal nodes:

Step 1: Mount Install ISO¶

Access the node's IPMI/BMC console
Mount the production install ISO via IPMI virtual media
Set boot order to CD/DVD first (Legacy mode)
Power on or reboot the node

Legacy Boot Only

The BIOS must be set to Legacy mode with CSM enabled. UEFI boot does not work on the NX-8150-G7 Supermicro hardware with the Palette ISO. See Troubleshooting for details.

Step 2: Wait for Auto-Install¶

The ISO will:

Boot the Kairos installer (Legacy/CSM mode)
Wipe all 7TB drives (SSD and NVMe) using sgdisk --zap-all
Install the OS to /dev/sda (223.5GB)
Configure the LACP bond and bridge
Power off the node when complete

Installation takes approximately 10-15 minutes depending on drive wipe duration.

Step 3: Mount Site User Data ISO¶

Unmount the install ISO from virtual media
Mount the node-specific site-user-data ISO (e.g., site-user-data-004.iso for STG-WAHVP004)
Boot the node from the local disk (not the ISO)

Step 4: First Boot and Registration¶

On first boot from disk:

The node reads the site-user-data ISO and applies the hostname and IP configuration
The node connects to the Palette Management Appliance at 10.25.232.155:443
The edge host registers using the token OTVhZDVlMzc5NDI0YzVlYjgyYjg1YjQzNDg4NGQwZjc=
The node appears in Palette under Edge Hosts with status Registered

Step 5: Verify Registration¶

In the Palette console:

Navigate to Edge Hosts
Verify all 3 nodes appear with their correct hostnames and IPs
Confirm status is Ready for each node

Post-Imaging Validation¶

Check	Expected	How to Verify
Node registered in Palette	3 nodes visible	Palette --> Edge Hosts
Hostname correct	stg-wahvp004, 005, 006	Palette --> Edge Host details
IP address correct	10.25.233.4, .5, .6	Palette --> Edge Host details
Bond active	4 NICs in LACP bond	SSH: `cat /proc/net/bonding/bond0`
Drives wiped	No prior partitions	SSH: `lsblk` shows clean drives
OS version	Ubuntu 24.04	SSH: `cat /etc/os-release`
Palette agent running	stylus service active	SSH: `systemctl status stylus`

Subsequent Updates

The site-user-data ISO is only read on first boot. All subsequent configuration changes are made through Palette cluster profiles and user-data management. The ISO can be unmounted after successful registration.