infrastructure/docs/wip/UPGRADE-2026-HARDWARE.md

# Hardware Upgrade Plan 2026

**Status:** 📋 PLANNED
**Target:** Q1 2026
**Author:** Kaloyan

---

## Overview

Migration from single-server architecture to dual-server with survival capabilities.

## Device Naming Convention

| Name | Hardware | Role | Notes |
|------|----------|------|-------|
| **XTRM-U** | Current NAS chassis | Current production | Will be decommissioned |
| **XTRM-N5** | Minisforum N5 Air | New production NAS | Inherits XTRM-U storage, OS, services + new NVMe pool |
| **XTRM-N1** | N100 ITX (XTRM-U board) | Survival node | Same 4x2.5GbE NIC, 1U 10" case, no HDDs |

---

## Hardware Specifications

### XTRM-N5 (Minisforum N5 Air) - Production Server

| Component | Specification |
|-----------|---------------|
| **CPU** | Intel N95/N100 or similar |
| **RAM** | TBD |
| **Storage (Fast)** | 3x Samsung 990 EVO Plus 1TB (NVMe ZFS RAID-Z1) |
| **Storage (Capacity)** | Existing 3.5" HDD Array (from XTRM-U) |
| **Network** | 10GbE Primary + 5GbE Secondary |
| **OS** | Unraid 7 (internal boot from NVMe pool) |

### XTRM-N1 (N100 ITX) - Survival Node

| Component | Specification |
|-----------|---------------|
| **Board** | XTRM-U motherboard (4x 2.5GbE Intel NICs) |
| **Chassis** | 1U 10" Rackmount (modular/3D printed) |
| **Storage** | 2x Fikwot FX501Pro 512GB (NVMe mirror) |
| **Power** | PicoPSU + 12V External Brick |
| **OS** | Proxmox VE or Debian |

---

## Migration Flow

```
XTRM-U (Current State)
├── Hardware (board, 4x2.5GbE) ──────────────► XTRM-N1 (1U 10" case)
│                                              └── New role: Survival node
│                                              └── New storage: 2x Fikwot 512GB
│
├── Storage (3.5" HDDs) ─────────────────────► XTRM-N5
├── OS (Unraid + config) ────────────────────► XTRM-N5
├── Services (containers) ───────────────────► XTRM-N5
│                                              └── New: Samsung NVMe pool
│
└── Chassis ─────────────────────────────────► Retired
```

---

## Post-Migration Architecture

### XTRM-N1 (Survival Node)

**Purpose:** Keep network alive if production server is offline

| Service | Purpose | Priority |
|---------|---------|----------|
| OPNsense | Firewall/Router (replaces HAP1 routing) | P0 |
| AdGuard Home | DNS | P0 |
| Vaultwarden | Password manager | P0 |
| Authentik (replica) | SSO backup | P1 |

### XTRM-N5 (Production Server)

**Purpose:** Primary services and storage

| Service | Purpose | Priority |
|---------|---------|----------|
| Nextcloud | File sync/storage | P1 |
| Plex | Media streaming | P2 |
| Authentik (primary) | SSO | P1 |
| Gitea | Git hosting | P2 |
| Woodpecker CI | CI/CD | P2 |
| NetBox | Network documentation | P2 |
| All other services | Various | P3 |

---

## Rack Layout (Post-Migration)

### 10" Rack (9U)

| U | Device | Notes |
|---|--------|-------|
| U9 | Shelf + ISP Gateway | Unchanged |
| U8 | PP2 (12-port) | Unchanged |
| U7 | Shelf + ZX1 (or CRS310) | Future: upgrade to CRS310 |
| U6 | **XTRM-N1** | NEW: Survival node |
| U5 | (empty) | |
| U4 | (XTRM-N5 continued) | |
| U3 | (XTRM-N5 continued) | |
| U2 | (XTRM-N5 continued) | |
| U1 | **XTRM-N5** | NEW: Production server |

---

## Software Configuration

### XTRM-N5: Unraid 7 Features

1. **Internal Boot Pool**
   - Use "Add Bootable Pool" feature
   - ZFS RAID-Z1 with 3x Samsung 990 EVO Plus
   - Partition 1: 8GB (OS boot)
   - Partition 2: ~1.8TB usable (data/containers)
   - Remove USB boot stick

2. **Storage Pools**
   - **nvme_pool**: 3x Samsung 990 EVO Plus (~1.8TB usable)
   - **hdd_pool**: Existing 3.5" HDDs (capacity tier)

### XTRM-N1: Proxmox Configuration

1. **Storage**
   - ZFS mirror: 2x Fikwot 512GB
   - Boot + VMs + Containers

2. **VMs/Containers**
   - OPNsense VM (if routing moves here)
   - AdGuard LXC
   - Vaultwarden LXC
   - Authentik LXC (replica)

---

## Network Changes

### Current (HAP1 as router)

```
Internet → HAP1 (Router) → ZX1 → XTRM-U
                        → CSS1 → Room outlets
```

### Option A: Keep HAP1 as Router

```
Internet → HAP1 (Router) → ZX1 → XTRM-N5
         ↓              → XTRM-N1 (services only)
         → CSS1 → Room outlets
```

### Option B: Move Routing to XTRM-N1

```
Internet → HAP1 (Bridge) → XTRM-N1 (OPNsense) → ZX1 → XTRM-N5
                                              → CSS1 → Rooms
```

**Decision:** TBD based on performance testing

---

## Thermal Management

| Strategy | Implementation |
|----------|----------------|
| Samsung NVMe in N5 | "Thermal Champions" - run cooler than Fikwot |
| Physical spacing | 1U gap between XTRM-N1 and XTRM-N5 |
| Active cooling | Exhaust fans tied to drive temp sensors |
| HDD hibernation | Aggressive spindown, NVMe handles active I/O |

---

## Future Milestone: CRS310 Upgrade

When budget allows, replace ZX1 with MikroTik CRS310-8G+2S+IN:

| Feature | Benefit |
|---------|---------|
| 10G DAC to CSS326 | Proper backbone |
| L3 offloading | Inter-VLAN routing off HAP1 |
| 8x 2.5G ports | All servers at full speed |

---

## Migration Checklist

### Pre-Migration
- [ ] Order XTRM-N5 hardware
- [ ] Order 3x Samsung 990 EVO Plus 1TB
- [ ] Order 2x Fikwot FX501Pro 512GB
- [ ] Design/print 1U 10" chassis for N1
- [ ] Order PicoPSU + 12V brick

### XTRM-N1 Setup (First)
- [ ] Extract XTRM-U motherboard
- [ ] Install in 1U chassis
- [ ] Install Fikwot NVMe drives
- [ ] Install Proxmox VE
- [ ] Deploy survival services (AdGuard, Vaultwarden)
- [ ] Test network survival (unplug XTRM-U)

### XTRM-N5 Setup
- [ ] Unbox and install Samsung NVMe drives
- [ ] Create ZFS boot pool
- [ ] Migrate Unraid config from USB
- [ ] Transfer HDD array
- [ ] Verify all services running
- [ ] Update DNS/IPs as needed

### Post-Migration
- [ ] Update NetBox with new devices
- [ ] Update documentation
- [ ] Decommission old XTRM-U chassis
- [ ] Performance testing
- [ ] Thermal monitoring

---

## Related Documents

- `GITOPS-CONTAINERS.md` - Phase 2: Container configuration in Git