Hi All,
This post is purely to document my setup and configurations to create, manage, update setup my home lab setting across multiple machines. This is not any guide nor explanation for the reasoning.
Mostly I REALLY do not have any idea what I'm doing, mostly following useless chatgpt or gemini. But since everything is changing or I lose history here I will document some of my setup and commands to configure things. This post is just for me and I hope I will keep it updated so you might reference, comments and advises are welcomed but I do not guarantee I will implement them.
1.NAS setup
Hardware:
Base POS5000 - Point of sale decommissioned computer
Intel XEON E3-1225 3.4GHz (4c/4t)
RAM 8GB DDR3
Disks: 128GB system SSD, 4x 2,5" HDD WD5000LPLX-60ZNTT1 500GB - you cannot beat 500GB for $14 joined in RAID5 encrypted ~~1,5TB
mpcie 2x sata controller from Aliexpress
Software stack:
OMV as host operating system
Tailscale
Immich
NAS
1.1 Host operating setup:
1.1.1 CPU Power:
Setting max cpu freq to 2,5GHz and governor to powersave. Resulting in idle power draw less than 35W
Remember to log as a root
Check active variables
omv-env get
Set variables:
omv-env set OMV_CPUFREQUTILS_MAXSPEED 2500MHz
omv-env set OMV_CPUFREQUTILS_GOVERNOR=powersave
Run with those variables:
omv-salt stage run prepare
omv-salt deploy run cpufrequtils
Based on https://forum.openmediavault.org/index.php?thread/43404-how-to-lower-cpu-frequency-and-disable-cores/
1.1.2 Immich setup:
Tried manual installation via OMV or docker myself - updates always resulted in immich not working. So I just use Install script.
https://docs.immich.app/install/script
Installation might sometimes stuck or take unreasonable time well above 30mins.
1.1.3 Immich update
created simple bash script:
#!/bin/sh
cd [YourPath]/Compose/immich-app
docker compose down
docker compose pull
docker compose up --remove-orphans -d
This solution works without issues since Sept 2025. I'm not updating it as soon as new release comes as it is hidden behind my LAN and Tailscale VPN
1.1.4 Tailscale
Install directly via terminal as a root - no issues there
2.Home PC
2.1 Running local AI models (not LLM's)
For AMD GPU Ubuntu 24.04 is a must - nothing else worked for me no bazzite nor ubuntu 26.04 or any later.
secret souse is PyENV to manage Python versions - Who the hell can work with python on day to day with this pure madness!
2.1.1 setup host system
sudo usermod -a -G render,video $LOGNAME
#download AMD drivers manually
sudo dpkg -i amdgpu-install_7.2.70200-1_all.deb
sudo amdgpu-install -y --usecase=graphics,rocm
Validate:
dkms status
rocminfo
Results:
dkms status
amdgpu/6.16.13-2278356.24.04, 6.17.0-14-generic, x86_64: installed
amdgpu/6.16.13-2278356.24.04, 6.17.0-19-generic, x86_64: installed
2.1.2 Git clone comfy UI
2.1.3 PyENV
follow the guide here https://github.com/pyenv/pyenv?tab=readme-ov-file#linuxunix or here:
git clone https://github.com/pyenv/pyenv.git ~/.pyenv
cd ~/.pyenv && src/configure && make -C src
cd ComfyUI
pyenv install 3.10.14
pyenv local 3.10.14 # this should set 3.10.14 to this comfyui folder only (not sure)
2.1.4 set the Python virtual enviroment venv
cd ComfyUI
python -m venv .python310venv
2.1.5 install dependencies
2.1.6 run
#!/bin/bash
cd ComfyUI/
source 310venv/bin/activate
export HSA_OVERRIDE_GFX_VERSION=10.3.0 #for RX6750XT please adjust for your GPU
python main.py
Profit
- Always ON NAS for just Immich
Hardware:
Tiny HP Pro Desk 600 G3 6100T 8GB 120GB SSD
WIFI upgraded to AX210 Wifi from Ali
Updates to come