Hardware
Bill of Materials
NOTE: for the most accurate/up-to-date BOM, review the Hardware Overvew
| Qty | Item | Role |
|---|---|---|
| 1 | Intel NUC 13 Pro (NUC13ANHi3) | Admin host (nuc-00) |
| 3 | Intel NUC 13 Pro (NUC13ANHi7) | Harvester compute nodes (nuc-01/02/03) |
| 4 | 64 GB DDR4 SO-DIMM (2× 32 GB) | RAM per node |
| 4 | 1 TB NVMe SSD (primary) | OS/boot per node |
| 3 | 2 TB NVMe SSD (secondary) | Longhorn/Harvester data (nuc-01/02/03) |
| 1 | 16-port gigabit switch (managed) | LAN fabric |
| Several | Cat6 patch cables | Node-to-switch + uplink |
| 1 | USB keyboard + HDMI display (or KVM) | Initial console access |
| 4 | USB flash drives (≥ 8 GB) | Bootable OS installers |
NUC Specifications
nuc-00 (Admin Host) — NUC13ANHi3
| Component | Spec |
|---|---|
| Model | Intel NUC 13 Pro (NUC13ANHi3) |
| RAM | 64 GB DDR4 |
| Storage | 1 TB NVMe (OS + VMs) |
| Network | Onboard GbE |
| Role | KVM host, PXE server, infra VMs |
The admin host runs KVM and hosts three VMs (nuc-00-01, nuc-00-02, nuc-00-03) providing DHCP, DNS (primary and secondary), HAProxy, and Keepalived. It also serves PXE/HTTP for Harvester installation.
An LVM volume group is created from the NVMe drive to provide flexible VM disk allocation:
/dev/nvme0n1p1 → /boot/efi (512 MB)
/dev/nvme0n1p2 → /boot (1 GB)
/dev/nvme0n1p3 → / (ext4) (100 GB)
/dev/nvme0n1p4 → LVM PV (remainder)
└── vg-infra
├── lv-nuc-00-01 (40 GB) → nuc-00-01 VM disk
├── lv-nuc-00-02 (40 GB) → nuc-00-02 VM disk
└── lv-nuc-00-03 (40 GB) → nuc-00-03 VM disk
nuc-01, nuc-02, nuc-03 (Harvester Cluster) — NUC13ANHi7
| Component | Spec |
|---|---|
| Model | Intel NUC 13 Pro (NUC13ANHi7) |
| RAM | 64 GB DDR4 |
| Storage (primary) | 1 TB SSD (/dev/sda — Harvester OS) |
| Storage (secondary) | 2 TB NVMe (/dev/nvme0n1 — Longhorn data volumes) |
| Network | Onboard GbE (enp86s0) |
| Role | Harvester HCI nodes |
Harvester uses the primary SSD for its OS and etcd, and the secondary NVMe is added to the Longhorn storage pool for persistent volume claims.
Physical Layout
The enclave uses a 16-port managed switch. Port assignments:
┌─────────────────────────────────────────┐
│ 16-port Switch │
│ Port 1: nuc-00 Port 9: nuc-02-kvm │
│ Port 2: nuc-01 Port 10: nuc-03-kvm │
│ Port 3: nuc-02 Port 11: (unused) │
│ Port 4: nuc-03 Port 12: (unused) │
│ Port 5: nuc-01-vms Port 13: (unused) │
│ Port 6: nuc-02-vms Port 14: (unused) │
│ Port 7: nuc-03-vms Port 15: spark-e │
│ Port 8: nuc-01-kvm Port 16: uplink │
└─────────────────────────────────────────┘
BIOS Configuration
Apply the following BIOS settings to each NUC before OS installation:
- Boot order: Network (PXE) first, SSD/NVMe second
- Secure Boot: Disabled (required for Harvester PXE boot)
- VT-x / VT-d: Enabled (required for KVM and KubeVirt on the nucs)
- Wake-on-LAN: Enabled (optional, useful for remote power-on)
- Auto power-on: Enabled after power loss (optional, for recovery)
Access the BIOS via F2 during POST, or F7 for the one-time boot menu, F12 for Network Boot.
Cabling Notes
- Label each cable at both ends (e.g.,
nuc-00 ↔ sw-p1) - Use different cable colors if possible: blue for management, yellow for uplink
- Leave slack for future additions — don't cable-tie too tight
Power Consumption
| State | Per NUC | 4× NUCs |
|---|---|---|
| Idle | ~15 W | ~60 W |
| Moderate load | ~45 W | ~180 W |
| Peak load | ~65 W | ~260 W |
The full cluster draws roughly 2–3 amps at 120V under normal load — a standard outlet circuit handles this comfortably.