monitor : temp GPU, conso watts, espace disque, heartbeat dispatcher

- nvidia-smi : +temperature.gpu + power.draw
- UI : tags °C / W / espace disque libre
- Dispatcher heartbeat toutes les 4s → point vert/rouge en haut du monitor
- Fix Docker SSH : copie + chmod 600 au démarrage (Bad owner)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app/main.py

@@ -213,7 +213,25 @@ async def partial_jobs(request: Request):
 @app.get("/partials/monitor", response_class=HTMLResponse)
 async def partial_monitor(request: Request):
     stats = await asyncio.gather(*[_worker_stats(w) for w in WORKERS])
-    return templates.TemplateResponse("_monitor.html", {"request": request, "workers": stats})
+    return templates.TemplateResponse("_monitor.html", {
+        "request": request,
+        "workers": stats,
+        "dispatcher": _dispatcher_status(),
+    })
+
+
+def _dispatcher_status() -> dict:
+    hb = DB_PATH.parent / "dispatcher.heartbeat"
+    try:
+        ts = _parse_ts(hb.read_text().strip())
+        if ts:
+            if ts.tzinfo is None:
+                ts = ts.replace(tzinfo=timezone.utc)
+            age = int((datetime.now(timezone.utc) - ts).total_seconds())
+            return {"alive": age < 30, "age_s": age}
+    except Exception:
+        pass
+    return {"alive": False, "age_s": None}
 
 
 async def _worker_stats(worker: dict) -> dict:
@@ -221,20 +239,30 @@ async def _worker_stats(worker: dict) -> dict:
     try:
         proc = await asyncio.create_subprocess_exec(
             "ssh", "-o", "ConnectTimeout=3", "-o", "BatchMode=yes", alias,
-            "nvidia-smi --query-gpu=memory.used,memory.total,utilization.gpu --format=csv,noheader,nounits && df -h / | tail -1",
+            "nvidia-smi --query-gpu=memory.used,memory.total,utilization.gpu,temperature.gpu,power.draw --format=csv,noheader,nounits && df -h / | tail -1",
             stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE,
         )
-        out, _ = await asyncio.wait_for(proc.communicate(), timeout=4)
+        out, _ = await asyncio.wait_for(proc.communicate(), timeout=5)
         text = out.decode().strip().splitlines()
-        gpu_line = text[0].split(",") if text else ["?", "?", "?"]
+        g = [x.strip() for x in text[0].split(",")] if text else ["?"] * 5
         disk = text[1].split() if len(text) > 1 else ["?"] * 6
+
+        def _int(v: str):
+            try:
+                return int(float(v))
+            except Exception:
+                return None
+
         return {
             **worker,
             "online": True,
-            "vram_used_mib": int(gpu_line[0].strip()) if gpu_line[0].strip().isdigit() else None,
-            "vram_total_mib": int(gpu_line[1].strip()) if gpu_line[1].strip().isdigit() else None,
-            "gpu_util_pct": int(gpu_line[2].strip()) if gpu_line[2].strip().isdigit() else None,
+            "vram_used_mib": _int(g[0]) if len(g) > 0 else None,
+            "vram_total_mib": _int(g[1]) if len(g) > 1 else None,
+            "gpu_util_pct": _int(g[2]) if len(g) > 2 else None,
+            "gpu_temp_c": _int(g[3]) if len(g) > 3 else None,
+            "gpu_power_w": _int(g[4]) if len(g) > 4 else None,
             "disk_used_pct": disk[4] if len(disk) > 4 else "?",
+            "disk_avail": disk[3] if len(disk) > 3 else "?",
         }
     except Exception as e:
         return {**worker, "online": False, "error": str(e)[:80]}