MQTT and nodes update

drb-c2-core/app/internal/mqtt_handler.py

@@ -208,13 +208,14 @@ class MQTTHandler:
     # Outbound — send a command to a specific node
     # ------------------------------------------------------------------
 
-    def send_command(self, node_id: str, payload: dict):
+    def send_command(self, node_id: str, payload: dict) -> bool:
         topic = f"nodes/{node_id}/commands"
         if self._client and self._connected:
             self._client.publish(topic, json.dumps(payload), qos=1)
             logger.info(f"Command sent to {node_id}: {payload.get('action')}")
-        else:
-            logger.warning(f"MQTT not connected — could not send command to {node_id}")
+            return True
+        logger.warning(f"MQTT not connected — could not send command to {node_id}")
+        return False
 
     def push_config(self, node_id: str, system_config: dict):
         topic = f"nodes/{node_id}/config"
@@ -240,12 +241,22 @@ class MQTTHandler:
     async def connect(self):
         self._loop = asyncio.get_event_loop()
         self._client = self._build_client()
-        try:
-            self._client.connect(settings.mqtt_broker, settings.mqtt_port, keepalive=60)
-            self._client.loop_start()
-            logger.info(f"MQTT connecting to {settings.mqtt_broker}:{settings.mqtt_port}")
-        except Exception as e:
-            logger.error(f"MQTT connection error: {e}")
+        # Start the paho network loop first so it drives reconnects automatically,
+        # then keep attempting the initial TCP connect until it succeeds.
+        self._client.loop_start()
+        asyncio.create_task(self._connect_with_retry())
+
+    async def _connect_with_retry(self):
+        delay = 5
+        logger.info(f"MQTT connecting to {settings.mqtt_broker}:{settings.mqtt_port}")
+        while True:
+            try:
+                self._client.connect(settings.mqtt_broker, settings.mqtt_port, keepalive=60)
+                return  # paho loop_start + reconnect_delay_set handles the rest
+            except Exception as e:
+                logger.warning(f"MQTT connect failed ({e}) — retrying in {delay}s")
+                await asyncio.sleep(delay)
+                delay = min(delay * 2, 60)
 
     async def disconnect(self):
         if self._client: