diff --git a/ansible/roles/gateway/templates/nftables.conf.j2 b/ansible/roles/gateway/templates/nftables.conf.j2
index 465f497..bb25344 100644
--- a/ansible/roles/gateway/templates/nftables.conf.j2
+++ b/ansible/roles/gateway/templates/nftables.conf.j2
@@ -32,6 +32,9 @@ table inet filter {
         # Services données (accès cluster uniquement)
         tcp dport { 5432, 6333, 6334, 9000, 9001 } ip saddr 192.168.10.0/24 accept
 
+        # Monitoring — node_exporter + alertmanager webhook (accès cluster)
+        tcp dport { 9100, 9093 } ip saddr 192.168.10.0/24 accept
+
         # Hermes gateway (accès LAN domestique + cluster)
         tcp dport 8080 ip saddr { 192.168.1.0/24, 192.168.10.0/24 } accept
 
diff --git a/ansible/roles/llama_server/templates/llama-server.service.j2 b/ansible/roles/llama_server/templates/llama-server.service.j2
index 375e1f2..ea3f860 100644
--- a/ansible/roles/llama_server/templates/llama-server.service.j2
+++ b/ansible/roles/llama_server/templates/llama-server.service.j2
@@ -20,6 +20,7 @@ ExecStart=/opt/llama.cpp/build/bin/llama-server \
     --embeddings \
     --pooling {{ llama_pooling }} \
 {% endif %}
+    --metrics \
     --log-disable
 Restart=on-failure
 RestartSec=10
diff --git a/k8s/infra/monitoring/values.yaml b/k8s/infra/monitoring/values.yaml
index 40a2e30..6a9a748 100644
--- a/k8s/infra/monitoring/values.yaml
+++ b/k8s/infra/monitoring/values.yaml
@@ -63,13 +63,6 @@ prometheus:
               instance: gpu-01
               role: gpu
 
-      - job_name: gpu-01-rocm
-        static_configs:
-          - targets: ['192.168.10.20:9101']
-            labels:
-              instance: gpu-01
-              role: gpu
-
       - job_name: llama-server-gpu
         static_configs:
           - targets: ['192.168.10.20:1234']