cube
Home server. Primary compute and storage node. Runs Incus with NVIDIA GPU passthrough for AI and photo workloads.
Hardware
- Form factor: ZimaCube Pro (IceWhale Technology)
- CPU: 12th Gen Intel Core i5-1235U (12 cores)
- RAM: 64 GB
- GPU: NVIDIA RTX 2000 Ada Generation (16 GB)
- OS drive: 1× 953.9 GB NVMe (ext4)
- Storage pool: 3× 16.4 TB HDD in RAIDZ1 (
storagepool), 2× 931.5 GB NVMe as L2ARC cache (ZFS)
System
- OS: Debian 13 (trixie)
- Kernel: 6.12.73+deb13-amd64
- Shell: fish
- Containerization: Incus 6.0.4
- Key services: Ollama (native), Tailscale
Access
- Local IP: 10.10.10.10
- Tailscale IP: 100.121.19.125 (
cube.little-lenok.ts.net) - Out-of-band: JetKVM
Containers
| Name | Service | URL |
|---|---|---|
| openwebui | Open WebUI | ai.fismen.no (proxied via fismen) |
| photos | Immich | photos.fismen.no (proxied via fismen) |
Notes
- Ollama runs natively (not in a container) with CUDA on the RTX 2000 Ada
- Incus version behind fismen.no (6.0.4 vs 6.21) — needs update
- ZFS pool name:
storage