Implement proper shutdown of scan/monitor tasks

scanner/scanner/__init__.py

@@ -1,5 +1,5 @@
-import asyncio
 import logging
+from asyncio import CancelledError, TaskGroup
 from contextlib import asynccontextmanager
 
 from fastapi import FastAPI
@@ -29,19 +29,29 @@ async def lifespan(_):
 		# there's no way someone else used the same id, right?
 		is_master = await db.fetchval("select pg_try_advisory_lock(198347)")
 		if is_master:
-			await migrate();
+			await migrate()
 		# creating the processor makes it listen to requests event in pg
-		provider = tmdb #CompositeProvider(tmdb)
+		provider = tmdb  # CompositeProvider(tmdb)
 		async with (
 			RequestProcessor(db, client, provider) as processor,
 			get_db() as db,
 		):
-			_ = asyncio.create_task(processor.process_all())
+			# see https://github.com/python/cpython/issues/108951
+			try:
+				async with TaskGroup() as tg:
+					_ = tg.create_task(processor.process_all())
 					scanner = FsScanner(client, RequestCreator(db))
 					if is_master:
-				_ = asyncio.create_task(scanner.monitor())
+						_ = tg.create_task(scanner.monitor())
-				_ = asyncio.create_task(scanner.scan(remove_deleted=True))
+						_ = tg.create_task(scanner.scan(remove_deleted=True))
 					yield
+					_ = tg.create_task(cancel())
+			except CancelledError:
+				pass
+
+
+async def cancel():
+	raise CancelledError()
 
 
 app = FastAPI(