Initiale Infrastruktur-Dokumentation pve1 und pve2.
Enthält Host-Doku, MQTT/HA, Git-Setup, Power-Monitoring und GPU-Idle (pve2). Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
@@ -0,0 +1,32 @@
|
||||
# GPU Idle & Power-Monitoring — pve2
|
||||
|
||||
Ausführliche Doku im Repo: [`/root/code/pve-power-mqtt/docs/gpu-idle-pve2.md`](/root/code/pve-power-mqtt/docs/gpu-idle-pve2.md)
|
||||
|
||||
## Kurzfassung
|
||||
|
||||
- **2× GTX 1080** am Host-Treiber (kein VM-Passthrough)
|
||||
- Idle-Ziel: **~8 W/GPU, P8** — erfordert **Persistence Mode**
|
||||
- **Frigate (CT 101):** nur Intel iGPU/VAAPI → keine NVIDIA-Mounts in LXC
|
||||
- **AIDEV (CT 110):** NVIDIA-Mounts für ML bei Bedarf
|
||||
- **Power-Agent:** `pve-power-mqtt` → MQTT/Home Assistant
|
||||
|
||||
## Schnellcheck
|
||||
|
||||
```bash
|
||||
nvidia-smi --query-gpu=index,power.draw,pstate,persistence_mode --format=csv
|
||||
systemctl status nvidia-persistenced pve-power-mqtt
|
||||
grep nvidia /etc/pve/lxc/*.conf
|
||||
```
|
||||
|
||||
## Persistenced installieren (aus Repo)
|
||||
|
||||
```bash
|
||||
cd /root/code/pve-power-mqtt
|
||||
./deploy/nvidia-persistenced/install.sh
|
||||
```
|
||||
|
||||
## Änderungen Juni 2026
|
||||
|
||||
- CT 101: NVIDIA device mounts auskommentiert
|
||||
- `nvidia-persistenced`: kaputter `--user-persistence-mode` Flag entfernt, Service läuft
|
||||
- Doku in `server-power` Repo unter `docs/`
|
||||
Reference in New Issue
Block a user