better redis error handling

services/aggregator/main.py

@@ -133,13 +133,16 @@ class MetricsServicer(metrics_pb2_grpc.MetricsServiceServicer):
                 key = f"{metric_type}:{','.join(f'{k}={v}' for k, v in labels.items())}"
             metrics_dict[key] = value
 
-        # Update Redis (current state)
-        await self.redis.update_machine_state(
-            machine_id=machine_id,
-            hostname=hostname,
-            metrics=metrics_dict,
-            timestamp_ms=timestamp_ms,
-        )
+        # Update Redis (current state) - don't fail stream if Redis is down
+        try:
+            await self.redis.update_machine_state(
+                machine_id=machine_id,
+                hostname=hostname,
+                metrics=metrics_dict,
+                timestamp_ms=timestamp_ms,
+            )
+        except Exception as e:
+            self.logger.warning("redis_update_failed", error=str(e))
 
         # Insert into TimescaleDB (historical)
         try:
@@ -161,16 +164,19 @@ class MetricsServicer(metrics_pb2_grpc.MetricsServiceServicer):
         except Exception as e:
             self.logger.warning("machine_registry_update_failed", error=str(e))
 
-        # Publish event for subscribers (alerts, gateway)
-        await self.publisher.publish(
-            topic="metrics.raw",
-            payload={
-                "machine_id": machine_id,
-                "hostname": hostname,
-                "timestamp_ms": timestamp_ms,
-                "metrics": metrics_dict,
-            },
-        )
+        # Publish event for subscribers (alerts, gateway) - don't fail stream
+        try:
+            await self.publisher.publish(
+                topic="metrics.raw",
+                payload={
+                    "machine_id": machine_id,
+                    "hostname": hostname,
+                    "timestamp_ms": timestamp_ms,
+                    "metrics": metrics_dict,
+                },
+            )
+        except Exception as e:
+            self.logger.warning("event_publish_failed", error=str(e))
 
         self.logger.debug(
             "batch_flushed",