Home Server — Setup Summary & Storage Plan

Built 16 August 2026, updated 30 August 2026 — the always-on storage anchor for the whole multi-machine system

Hardware

Software

• OS: Ubuntu Server 26.04 LTS (clean install, fully updated)
• Access: SSH enabled, password auth working — ssh r@192.168.2.42
• Static IP: 192.168.2.42/24, gateway 192.168.2.1 (set via netplan, file: /etc/netplan/00-installer-config.yaml)
• Docker: Docker Engine + Compose plugin installed, running rootless-friendly (user r added to docker group — no sudo needed for docker commands)
• File sharing: Samba installed and configured
  • Share storage1 → /mnt/storage1 (the 2.7 TB drive)
  • Share storage2 → /mnt/storage2 (the 931 GB drive)
  • Connect from Mac via Finder → Cmd+K → smb://192.168.2.42

Storage Mounts

Both are in /etc/fstab (auto-mount on boot) using UUIDs, so they survive reboots and drive-letter reshuffles.

Wake-on-LAN (WoL) Configured 30 Aug

The server runs on-demand via Wake-on-LAN rather than 24/7, to reduce electricity use and wear on the older FX-4320 hardware.

Network card

Interface:        enp2s0
MAC address:      2c:56:dc:74:c4:ff
Supports Wake-on: pumbg  (g = magic packet supported)
Wake-on:          g      (enabled)

BIOS setting

Motherboard: ASUS M5A78L-M LE/USB3

BIOS → Power → APM Configuration → Power On By PCI-E/PCI → Enabled

systemd service (persistent across reboots)

File: /etc/systemd/system/wol.service

[Unit]
Description=Enable Wake-on-LAN
After=network.target

[Service]
Type=oneshot
ExecStart=/usr/sbin/ethtool -s enp2s0 wol g
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target

sudo systemctl enable wol.service
sudo systemctl start wol.service

Waking the server

From any machine on the local network:

wakeonlan 2c:56:dc:74:c4:ff

Alias set on all machines as wakeserver (see the ML Desktop setup). Server is reachable via SSH ~60–90 seconds after sending the magic packet.

Tested and confirmed

• Cold poweroff → magic packet → SSH connected
• Working from Mac (M4)
• Working from ML Desktop (Ubuntu 26.04)

Proposed Storage Organization Draft

Rather than a strict "active drive vs. cold drive" split, organize by what the data is and how it behaves:

storage2 (931 GB) — small, stable, frequently touched

storage2/
├── github/              # active repos, cloned from GitHub
│   ├── active/          # things currently being worked on
│   └── misc/            # small utility repos, one-offs
├── docs/                # project docs, notes, references
└── vs-projects/         # non-repo scripts

storage1 (2.7 TB) — large, growing, less frequently touched

storage1/
├── datasets/
│   └── alloy_lab/       # NdFeB SEM/XRD/VSM data (migrating from Mac)
│       ├── sem/
│       ├── mh/
│       └── xrd/
├── training/
│   ├── lerobot/         # robot arm training runs, checkpoints
│   └── other-ml/
├── to be sorted/        # old PPMS data from previous lab (sorting in progress)
│   └── PPMS1_old/
│       └── Data/
│           └── Andreas/ # historical MnGa/CoTi VSM measurements
└── archive/             # old backups, superseded project versions

PostgreSQL Migration Complete

PostgreSQL (alloy_lab DB) now runs in Docker on this server, not the Mac — migrated in one session, no data loss. All machines (Mac, ML Desktop) connect to one canonical DB here. No diverging copies, no data sitting on the Mac Desktop.

Known Follow-ups