Skip to main content

Harvester Cluster

Three NUCs (nuc-01, nuc-02, nuc-03) form the Harvester HCI cluster. They are installed via PXE boot using automated cloud-init-style Harvester config files.

How Harvester PXE Boot Works​

  1. NUC powers on → DHCP assigns IP + next-server + filename
  2. NUC downloads pxelinux.0 from TFTP on nuc-00
  3. PXE menu boots Harvester kernel + initrd from HTTP on nuc-00
  4. Harvester installer fetches its per-node config file from nuc-00
  5. Automated install proceeds; node reboots into Harvester

Per-Node Config Files​

Create one config file per node in /var/www/html/harvester/. Config files follow the naming convention used in the enclave repo: config-create-nuc-01.yaml, config-join-nuc-02.yaml, etc.

config-create-nuc-01.yaml (First Node — Creates Cluster)​

scheme_version: 1

server_url: ""  # empty = this is the first/create node

token: "KentuckyHarvester"

os:
  hostname: nuc-01
  password: "$6$rounds=4096$SALTSALT$HASHEDPASSWORD"  # openssl passwd -6
  ntp_servers:
    - 0.suse.pool.ntp.org
    - 1.suse.pool.ntp.org
    - 2.suse.pool.ntp.org
  dns_nameservers:
    - 10.10.12.9
    - 10.10.12.8
    - 8.8.8.8
  ssh_authorized_keys:
    - "ssh-ed25519 AAAAC3... enclave-admin"

install:
  mode: create
  management_interface:
    interfaces:
      - name: enp86s0
    method: static
    ip: 10.10.12.101
    subnet_mask: 255.255.252.0
    gateway: 10.10.12.1
    mtu: 1500
  device: /dev/sda
  vip: 10.10.12.100
  vip_mode: static

system_settings:
  auto-disk-provision-paths: "/dev/nvme0n1"  # secondary NVMe for Longhorn

config-join-nuc-02.yaml (Join Node)​

scheme_version: 1

server_url: "https://10.10.12.100"  # Harvester VIP of first node

token: "KentuckyHarvester"

os:
  hostname: nuc-02
  password: "$6$rounds=4096$SALTSALT$HASHEDPASSWORD"
  ntp_servers:
    - 0.suse.pool.ntp.org
    - 1.suse.pool.ntp.org
    - 2.suse.pool.ntp.org
  dns_nameservers:
    - 10.10.12.9
    - 10.10.12.8
    - 8.8.8.8
  ssh_authorized_keys:
    - "ssh-ed25519 AAAAC3... enclave-admin"

install:
  mode: join
  management_interface:
    interfaces:
      - name: enp86s0
    method: static
    ip: 10.10.12.102
    subnet_mask: 255.255.252.0
    gateway: 10.10.12.1
    mtu: 1500
  device: /dev/sda

system_settings:
  auto-disk-provision-paths: "/dev/nvme0n1"

config-join-nuc-03.yaml follows the same pattern as nuc-02 with hostname nuc-03 and IP 10.10.12.103.

Generate Hashed Password​

openssl passwd -6 'YourSecurePassword'

Installation Sequence​

Critical: Install nodes in order. nuc-01 must complete and the Harvester VIP (10.10.12.100) must be reachable before starting nuc-02 and nuc-03.

Step 1: Boot nuc-01​

Power on nuc-01. In the BIOS one-time boot menu (F7), select Network/PXE boot.

Monitor installation progress from nuc-00:

# Watch Apache access log
tail -f /var/log/httpd/access_log | grep nuc-01

Installation takes 10–15 minutes. The node reboots twice.

Step 2: Verify nuc-01 is up​

# Harvester API should respond at VIP
curl -k https://10.10.12.100/ping
# Expected: "pong"

# SSH to node
ssh rancher@10.10.12.101

# Check Harvester cluster health
kubectl get nodes --kubeconfig /etc/rancher/k3s/k3s.yaml
# Expected: nuc-01 Ready

Step 3: Boot nuc-02 and nuc-03​

Once nuc-01 is healthy, PXE boot nuc-02 and nuc-03 (can be done simultaneously).

Monitor join progress:

# From nuc-01
watch kubectl get nodes --kubeconfig /etc/rancher/k3s/k3s.yaml

Allow 15–20 minutes for both nodes to join.

Step 4: Retrieve kubeconfig​

# Copy Harvester kubeconfig to your workstation
scp rancher@10.10.12.100:/etc/rancher/k3s/k3s.yaml ~/.kube/harvester-config

# Update server address in kubeconfig (replace 127.0.0.1 with VIP)
sed -i 's|https://127.0.0.1:6443|https://10.10.12.100:6443|g' ~/.kube/harvester-config

# Verify
kubectl --kubeconfig ~/.kube/harvester-config get nodes

Cluster Verification​

# All nodes Ready
kubectl get nodes -o wide
# Expected:
# NAME    STATUS   ROLES                  AGE   VERSION
# nuc-01  Ready    control-plane,master   ...   v1.x.x
# nuc-02  Ready    control-plane,master   ...   v1.x.x
# nuc-03  Ready    control-plane,master   ...   v1.x.x

# Longhorn storage (may take a few minutes to appear)
kubectl get sc
# Expected: harvester-longhorn (default)

# Check Longhorn pods
kubectl get pods -n longhorn-system

# Harvester UI accessible
open https://10.10.12.100

Login to the Harvester UI with the password set in the config files.

Post-Install: Add Longhorn Disks​

If the secondary NVMe (/dev/nvme0n1) was not automatically provisioned:

  1. Navigate to Harvester UI → Host → select a node
  2. Click Edit → Storage
  3. Add /dev/nvme0n1 as a data disk
  4. Repeat for all nodes

Longhorn will format and add these disks to its storage pool.

Post Install Config​

Create Networking for Virtual Machines (optional - requires an additional NIC on each host)​

In the Harvester UI, expand Networks and click on "Cluster Network Configuration". Click "Create a Cluster Network" - I name mine: clstrnet-vms (12 char limit), Click "Create" Now you will see "Cluster Network: clstrnet-vms" - and click on "Create Network Configuration". I name mine: netconf-vms - select Uplink on the left and the appropriate NIC in the middle pane. (you can choose bond options, etc... here)

Next, click on "Virtual Machine Networks", which I name: vmnet-vms and select "UntaggedNetwork" (this may be different for your environment) - followed by selecting the Cluster Network we created earlier "clstrnet-vms". Click Create. You're done!

Uplaod Image(s) for VM Operating System​

In this case we will use the sl-micro.x86_64-6.1-base-qcow-gm.qcow2 image. (first grab the URL to the image which is hosted on your web server) In the Harvester UI, click on "Images" on the left and then "Create" on the right. I generally let the name autopopulate - just paste that URL in the "URL" field and the name will get updated.

ssh key​

Copy the ssh public key you wish to use to Harvester In the Harvester UI, expand "Advanced" on the left and click on "SSH Keys" (the key should start with "ssh-rsa" or "ssh-ecdsa" and may end with an email address or username

Cloud Configuration Templates (cloud-init)​

For naming these, I tend to name them after the OS I intend to use them with (SL-micro vs SLES, for example)

In the Harvester UI, expand "Advanced" on the left and click on "Cloud Configuration Templates". Click "Create" and provide a name, then paste your cloud-init in the box under "User Data" ProTip: There are 2 types here - User Data and Network Data. I tend to just put everything in User Data - but you figure out what is best for you.

Proceed to Rancher Manager.