xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-19 21:34:27 +08:00
Code Issues Packages Projects Releases Wiki Activity

[V1] Avoid socket errors during shutdown when requests are in in-flight (#16807)

Browse Source 
Signed-off-by: Nick Hill <nhill@redhat.com>
This commit is contained in:
Nick Hill 2025-04-22 21:36:29 -07:00 committed by GitHub
parent 047797ef90
commit b2f195c429
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 7 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/v1/engine/core.py
View File

@ -384,7 +384,7 @@ class EngineCoreProc(EngineCore):
except SystemExit: except SystemExit:
logger.debug("EngineCore exiting.") logger.debug("EngineCore exiting.")
raise
except Exception as e: except Exception as e:
if engine_core is None: if engine_core is None:
logger.exception("EngineCore failed to start.") logger.exception("EngineCore failed to start.")

10
vllm/v1/engine/core_client.py
View File

@ -312,6 +312,7 @@ class BackgroundResources:
def __call__(self): def __call__(self):
"""Clean up background resources.""" """Clean up background resources."""
self.engine_dead = True
for core_engine in self.core_engines: for core_engine in self.core_engines:
core_engine.close() core_engine.close()
@ -564,7 +565,7 @@ class SyncMPClient(MPClient):
self._send_input(EngineCoreRequestType.ADD, request) self._send_input(EngineCoreRequestType.ADD, request)
def abort_requests(self, request_ids: list[str]) -> None: def abort_requests(self, request_ids: list[str]) -> None:
if len(request_ids) > 0: if request_ids and not self.resources.engine_dead:
self._send_input(EngineCoreRequestType.ABORT, request_ids) self._send_input(EngineCoreRequestType.ABORT, request_ids)
def profile(self, is_start: bool = True) -> None: def profile(self, is_start: bool = True) -> None:
@ -735,7 +736,7 @@ class AsyncMPClient(MPClient):
self._ensure_output_queue_task() self._ensure_output_queue_task()
async def abort_requests_async(self, request_ids: list[str]) -> None: async def abort_requests_async(self, request_ids: list[str]) -> None:
if len(request_ids) > 0: if request_ids and not self.resources.engine_dead:
await self._send_input(EngineCoreRequestType.ABORT, request_ids) await self._send_input(EngineCoreRequestType.ABORT, request_ids)
async def profile_async(self, is_start: bool = True) -> None: async def profile_async(self, is_start: bool = True) -> None:
@ -902,5 +903,6 @@ class DPAsyncMPClient(AsyncMPClient):
async def _abort_requests(self, request_ids: list[str], async def _abort_requests(self, request_ids: list[str],
engine: CoreEngine) -> None: engine: CoreEngine) -> None:
await self._send_input(EngineCoreRequestType.ABORT, request_ids, if not self.resources.engine_dead:
engine) await self._send_input(EngineCoreRequestType.ABORT, request_ids,
engine)