feat: monitoring — dashboards Grafana + alertes Prometheus

Dashboards (ConfigMaps, auto-importés par sidecar Grafana) :
- dashboard-kubernetes : cluster overview, pods, PVCs, restarts
- dashboard-infrastructure : compute/storage-01/gpu-01 CPU/RAM/disque/réseau + GPU ROCm
- dashboard-ai : llama-server GPU/CPU tokens/s, requêtes, VRAM, température GPU

Alertes (PrometheusRule) :
- alerts-node : CPU>90%, RAM>90%, disque>80%, storage-01/gpu-01 down, load élevé
- alerts-k8s : NodeNotReady, PodCrashLoop, PodPending, DeploymentUnavailable, PVCNotBound, ArgoCDOutOfSync
- alerts-ai : llama-server GPU/system/monitor down, GPU temp>80°C, VRAM>95%, queue élevée

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

k8s/infra/monitoring/alerts/alerts-ai.yaml

@@ -0,0 +1,80 @@
+apiVersion: monitoring.coreos.com/v1
+kind: PrometheusRule
+metadata:
+  name: funk-ai-alerts
+  namespace: monitoring
+  labels:
+    release: kube-prometheus-stack
+spec:
+  groups:
+    - name: funk.ai
+      interval: 1m
+      rules:
+
+        # --- llama-server ---
+        - alert: LlamaServerGPUDown
+          expr: up{job="llama-server-gpu"} == 0
+          for: 2m
+          labels:
+            severity: critical
+          annotations:
+            summary: "llama-server GPU (Qwen3-8B) inaccessible"
+            description: "Le serveur LLM GPU sur gpu-01:1234 ne répond plus depuis 2 min. Inférence GPU unavailable."
+
+        - alert: LlamaServerSystemDown
+          expr: up{job="llama-server-system"} == 0
+          for: 2m
+          labels:
+            severity: warning
+          annotations:
+            summary: "llama-server CPU system (Qwen3-1.7B) inaccessible"
+            description: "Le serveur LLM CPU system sur gpu-01:1236 ne répond plus. Profil Hermes system degraded."
+
+        - alert: LlamaServerMonitorDown
+          expr: up{job="llama-server-monitor"} == 0
+          for: 2m
+          labels:
+            severity: warning
+          annotations:
+            summary: "llama-server CPU monitor (Qwen3-1.7B) inaccessible"
+            description: "Le serveur LLM CPU monitor sur gpu-01:1237 ne répond plus. Profil Hermes monitor degraded."
+
+        # --- GPU temperature (ROCm via textfile_collector) ---
+        - alert: GPUTemperatureHigh
+          expr: rocm_gpu_temperature_celsius > 80
+          for: 5m
+          labels:
+            severity: warning
+          annotations:
+            summary: "GPU RX 6700XT température élevée : {{ $value | printf \"%.0f\" }}°C"
+            description: "Température GPU à {{ $value | printf \"%.1f\" }}°C depuis 5 min. Vérifier la ventilation."
+
+        - alert: GPUTemperatureCritical
+          expr: rocm_gpu_temperature_celsius > 90
+          for: 2m
+          labels:
+            severity: critical
+          annotations:
+            summary: "GPU RX 6700XT surchauffe : {{ $value | printf \"%.0f\" }}°C"
+            description: "Température GPU critique à {{ $value | printf \"%.1f\" }}°C. Risque de throttling ou shutdown thermique."
+
+        # --- VRAM ---
+        - alert: GPUVRAMAlmostFull
+          expr: |
+            (rocm_vram_used_bytes / rocm_vram_total_bytes) * 100 > 95
+          for: 5m
+          labels:
+            severity: warning
+          annotations:
+            summary: "VRAM GPU presque saturée"
+            description: "VRAM utilisée à {{ $value | printf \"%.1f\" }}% ({{ $labels.gpu }}). Risque d'OOM sur les prochaines requêtes."
+
+        # --- Requêtes en attente (backpressure) ---
+        - alert: LlamaServerHighQueueGPU
+          expr: llamacpp:requests_deferred{job="llama-server-gpu"} > 5
+          for: 2m
+          labels:
+            severity: warning
+          annotations:
+            summary: "File d'attente llama-server GPU élevée"
+            description: "{{ $value | printf \"%.0f\" }} requêtes en attente sur le serveur GPU depuis 2 min."