fix: await takeover polling jobs

digital-avatar-app/backend/main.py

@@ -4,7 +4,7 @@ from fastapi.middleware.cors import CORSMiddleware
 import os
 import logging
 
-from apscheduler.schedulers.background import BackgroundScheduler
+from apscheduler.schedulers.asyncio import AsyncIOScheduler
 from apscheduler.triggers.interval import IntervalTrigger
 import redis as redis_lib
 
@@ -23,6 +23,9 @@ from responses import ok
 
 logger = logging.getLogger(__name__)
 
+takeover_scheduler = None
+takeover_db = None
+
 app = FastAPI(title="会会数字分身 API", version="1.0.0")
 
 app.add_middleware(
@@ -110,9 +113,14 @@ def seed():
 
 @app.on_event("startup")
 def on_startup():
+    global takeover_scheduler, takeover_db
+
     init_db()
     seed()
 
+    # Release stale resources when startup is invoked again by a reload/test.
+    stop_takeover_scheduler()
+
     # --- Takeover scheduler ---
     try:
         # Initialize Redis (optional)
@@ -137,16 +145,42 @@ def on_startup():
 
         # Initialize takeover service
         from services.takeover_service import TakeoverService
-        takeover_service = TakeoverService(SessionLocal(), boxim_client, redis_client)
+        takeover_db = SessionLocal()
+        takeover_service = TakeoverService(takeover_db, boxim_client, redis_client)
 
-        # Start periodic polling job
+        # AsyncIOScheduler awaits the service coroutine instead of dropping it.
-        scheduler = BackgroundScheduler()
+        takeover_scheduler = AsyncIOScheduler()
-        scheduler.add_job(
+        takeover_scheduler.add_job(
             takeover_service.poll_and_process_messages,
             trigger=IntervalTrigger(seconds=10),
             id="takeover_message_poll",
         )
-        scheduler.start()
+        takeover_scheduler.start()
         logger.info("Takeover message polling scheduler started (interval=10s)")
     except Exception as e:
+        stop_takeover_scheduler()
         logger.warning(f"Failed to initialize takeover scheduler, app will continue without it: {e}")
+
+
+def stop_takeover_scheduler():
+    global takeover_scheduler, takeover_db
+
+    if takeover_scheduler is not None:
+        try:
+            if takeover_scheduler.running:
+                takeover_scheduler.shutdown(wait=False)
+        except Exception as e:
+            logger.warning(f"Failed to stop takeover scheduler cleanly: {e}")
+        finally:
+            takeover_scheduler = None
+
+    if takeover_db is not None:
+        try:
+            takeover_db.close()
+        finally:
+            takeover_db = None
+
+
+@app.on_event("shutdown")
+def on_shutdown():
+    stop_takeover_scheduler()

digital-avatar-app/backend/tests/test_takeover_scheduler.py

@@ -14,7 +14,7 @@ def test_app_has_startup_event():
 @patch("services.takeover_service.TakeoverService")
 @patch("services.boxim_client.BoxIMClient")
 @patch("main.redis_lib.from_url")
-@patch("main.BackgroundScheduler")
+@patch("main.AsyncIOScheduler")
 def test_scheduler_initialized_with_redis(mock_scheduler_class, mock_redis_from_url, mock_boxim_cls, mock_takeover_cls):
     """Verify scheduler is initialized when Redis is available."""
     mock_redis = MagicMock()
@@ -25,6 +25,7 @@ def test_scheduler_initialized_with_redis(mock_scheduler_class, mock_redis_from_
     mock_boxim_cls.return_value = mock_boxim
 
     mock_takeover = MagicMock()
+    mock_takeover.poll_and_process_messages = AsyncMock()
     mock_takeover_cls.return_value = mock_takeover
 
     with patch("main.init_db"), patch("main.seed"), patch.dict("os.environ", {"REDIS_URL": "redis://localhost:6379"}):
@@ -32,19 +33,23 @@ def test_scheduler_initialized_with_redis(mock_scheduler_class, mock_redis_from_
         on_startup()
 
     mock_scheduler_class.return_value.add_job.assert_called_once()
+    scheduled_callable = mock_scheduler_class.return_value.add_job.call_args.args[0]
     call_kwargs = mock_scheduler_class.return_value.add_job.call_args[1]
+    assert scheduled_callable is mock_takeover.poll_and_process_messages
     assert call_kwargs["id"] == "takeover_message_poll"
+    mock_scheduler_class.return_value.start.assert_called_once_with()
 
 
 @patch("services.takeover_service.TakeoverService")
 @patch("services.boxim_client.BoxIMClient")
-@patch("main.BackgroundScheduler")
+@patch("main.AsyncIOScheduler")
 def test_scheduler_starts_without_redis(mock_scheduler_class, mock_boxim_cls, mock_takeover_cls):
     """App should start even when REDIS_URL is not set."""
     mock_boxim = MagicMock()
     mock_boxim_cls.return_value = mock_boxim
 
     mock_takeover = MagicMock()
+    mock_takeover.poll_and_process_messages = AsyncMock()
     mock_takeover_cls.return_value = mock_takeover
 
     with patch("main.init_db"), patch("main.seed"), patch.dict("os.environ", {"REDIS_URL": ""}, clear=False):
@@ -57,7 +62,7 @@ def test_scheduler_starts_without_redis(mock_scheduler_class, mock_boxim_cls, mo
 @patch("services.takeover_service.TakeoverService")
 @patch("services.boxim_client.BoxIMClient")
 @patch("main.redis_lib.from_url")
-@patch("main.BackgroundScheduler")
+@patch("main.AsyncIOScheduler")
 def test_scheduler_starts_when_redis_fails(mock_scheduler_class, mock_redis_from_url, mock_boxim_cls, mock_takeover_cls):
     """App should start even when Redis ping fails."""
     mock_redis_from_url.side_effect = ConnectionError("Connection refused")
@@ -66,6 +71,7 @@ def test_scheduler_starts_when_redis_fails(mock_scheduler_class, mock_redis_from
     mock_boxim_cls.return_value = mock_boxim
 
     mock_takeover = MagicMock()
+    mock_takeover.poll_and_process_messages = AsyncMock()
     mock_takeover_cls.return_value = mock_takeover
 
     with patch("main.init_db"), patch("main.seed"), patch.dict("os.environ", {"REDIS_URL": "redis://badhost:6379"}):
@@ -75,7 +81,7 @@ def test_scheduler_starts_when_redis_fails(mock_scheduler_class, mock_redis_from
     mock_scheduler_class.return_value.add_job.assert_called_once()
 
 
-@patch("main.BackgroundScheduler")
+@patch("main.AsyncIOScheduler")
 def test_scheduler_fails_gracefully(mock_scheduler_class):
     """If scheduler init raises, the app should still start (exception caught)."""
     mock_scheduler_class.side_effect = RuntimeError("Scheduler crash")
@@ -87,6 +93,24 @@ def test_scheduler_fails_gracefully(mock_scheduler_class):
     # No exception should propagate
 
 
+def test_scheduler_shutdown_releases_resources():
+    """Shutdown should stop polling and close its dedicated database session."""
+    import main
+
+    mock_scheduler = MagicMock()
+    mock_scheduler.running = True
+    mock_db = MagicMock()
+    main.takeover_scheduler = mock_scheduler
+    main.takeover_db = mock_db
+
+    main.on_shutdown()
+
+    mock_scheduler.shutdown.assert_called_once_with(wait=False)
+    mock_db.close.assert_called_once_with()
+    assert main.takeover_scheduler is None
+    assert main.takeover_db is None
+
+
 # --- poll_and_process_messages ---