[BugFix] Don't cancel asyncio tasks directly from destructors (#22476)

Signed-off-by: Nick Hill <nhill@redhat.com>

vllm/utils/__init__.py

@@ -687,20 +687,31 @@ class AsyncMicrobatchTokenizer:
         max_length = kwargs.get("max_length")
 
         if not truncation:
-            return ("encode", add_special_tokens, False, None)
+            return "encode", add_special_tokens, False, None
 
         model_max = getattr(self.tokenizer, "model_max_length", None)
         if max_length is None or (model_max is not None
                                   and max_length == model_max):
-            return ("encode", add_special_tokens, True, "model_max")
+            return "encode", add_special_tokens, True, "model_max"
 
-        return ("encode", "other")
+        return "encode", "other"
 
     def __del__(self):
-        for task in self._batcher_tasks:
+        if ((tasks := getattr(self, "_batcher_tasks", None))
-            if not task.done():
+                and (loop := getattr(self, "_loop", None))
+                and not loop.is_closed()):
+
+            def cancel_tasks():
+                for task in tasks:
                     task.cancel()
 
+            loop.call_soon_threadsafe(cancel_tasks)
+
+
+def cancel_task_threadsafe(task: Task):
+    if task and not task.done() and not (loop := task.get_loop()).is_closed():
+        loop.call_soon_threadsafe(task.cancel)
+
 
 def make_async(
     func: Callable[P, T],

vllm/v1/engine/async_llm.py

@@ -27,7 +27,7 @@ from vllm.transformers_utils.config import (
 from vllm.transformers_utils.tokenizer import AnyTokenizer
 from vllm.transformers_utils.tokenizer_group import init_tokenizer_from_configs
 from vllm.usage.usage_lib import UsageContext
-from vllm.utils import Device, cdiv, deprecate_kwargs
+from vllm.utils import Device, cancel_task_threadsafe, cdiv, deprecate_kwargs
 from vllm.v1.engine import EngineCoreRequest
 from vllm.v1.engine.core_client import EngineCoreClient
 from vllm.v1.engine.exceptions import EngineDeadError, EngineGenerateError
@@ -219,8 +219,7 @@ class AsyncLLM(EngineClient):
         if engine_core := getattr(self, "engine_core", None):
             engine_core.shutdown()
 
-        if handler := getattr(self, "output_handler", None):
+        cancel_task_threadsafe(getattr(self, "output_handler", None))
-            handler.cancel()
 
     async def get_supported_tasks(self) -> tuple[SupportedTask, ...]:
         return await self.engine_core.get_supported_tasks_async()

vllm/v1/engine/core_client.py

@@ -23,7 +23,8 @@ from vllm.config import VllmConfig
 from vllm.logger import init_logger
 from vllm.lora.request import LoRARequest
 from vllm.tasks import SupportedTask
-from vllm.utils import get_open_port, get_open_zmq_inproc_path, make_zmq_socket
+from vllm.utils import (cancel_task_threadsafe, get_open_port,
+                        get_open_zmq_inproc_path, make_zmq_socket)
 from vllm.v1.engine import (EngineCoreOutputs, EngineCoreRequest,
                             EngineCoreRequestType,
                             ReconfigureDistributedRequest, ReconfigureRankType,
@@ -342,10 +343,8 @@ class BackgroundResources:
         if self.coordinator is not None:
             self.coordinator.close()
 
-        if self.output_queue_task is not None:
+        cancel_task_threadsafe(self.output_queue_task)
-            self.output_queue_task.cancel()
+        cancel_task_threadsafe(self.stats_update_task)
-        if self.stats_update_task is not None:
-            self.stats_update_task.cancel()
 
         # ZMQ context termination can hang if the sockets
         # aren't explicitly closed first.