Change local job monitoring to poll database instead of TCP

- Use get_queue_by_job_id to check job status
- More reliable than TCP polling for local jobs

vidai/cluster_master.py

@@ -628,35 +628,41 @@ class ClusterMaster:
             return None
 
     async def _monitor_job_result(self, job_id: str, process_type: str) -> None:
-        """Monitor for job result from local backend."""
+        """Monitor for job result by polling database."""
         try:
-            from .comm import SocketCommunicator
-            from .config import get_backend_web_port
+            from .database import get_queue_by_job_id
+            import json
 
             # Poll for result
             for _ in range(300):  # Poll for up to 5 minutes (300 * 1s)
                 try:
-                    backend_comm = SocketCommunicator(host='localhost', port=get_backend_web_port(), comm_type='tcp')
-                    backend_comm.connect()
-
-                    # Send get_result request
-                    result_request = Message(
-                        msg_type='get_result',
-                        msg_id=f'poll_{job_id}',
-                        data={'request_id': job_id}
-                    )
-                    backend_comm.send_message(result_request)
-
-                    # Try to receive response
-                    response = backend_comm.receive_message()
-                    if response and response.msg_type in ['analyze_response', 'train_response']:
-                        result_data = response.data
+                    queue_entry = get_queue_by_job_id(job_id)
+                    if queue_entry:
+                        status = queue_entry['status']
+                        if status == 'completed':
+                            result = json.loads(queue_entry['result']) if queue_entry['result'] else {}
                             print(f"Received result for job {job_id}")
 
                             # Handle result
                             await self._handle_job_result({
                                 'job_id': job_id,
-                            'result': result_data
+                                'result': result
+                            })
+
+                            # Clean up
+                            if job_id in self.pending_jobs:
+                                del self.pending_jobs[job_id]
+                            return
+
+                        elif status == 'failed':
+                            error = queue_entry.get('error', 'Unknown error')
+                            result = {'status': 'failed', 'error': error}
+                            print(f"Job {job_id} failed: {error}")
+
+                            # Handle failed result
+                            await self._handle_job_result({
+                                'job_id': job_id,
+                                'result': result
                             })
 
                             # Clean up
@@ -664,13 +670,13 @@ class ClusterMaster:
                                 del self.pending_jobs[job_id]
                             return
 
-                    elif response and response.msg_type == 'result_pending':
-                        # Result not ready yet, wait and try again
+                        elif status == 'processing':
+                            # Still processing, wait and try again
                             await asyncio.sleep(1)
                             continue
 
                 except Exception as e:
-                    print(f"Error polling for job {job_id} result: {e}")
+                    print(f"Error polling for job {job_id} status: {e}")
                     await asyncio.sleep(1)
 
             # Timeout - job took too long