From 390a3b3af1d6578d9d2011c67e4d8ed66c322699 Mon Sep 17 00:00:00 2001 From: sumingZero <469434916@qq.com> Date: Sat, 11 Oct 2025 16:44:12 +0800 Subject: [PATCH] [Fix] Each request in the decode instance encounters a load failure --- ucm/integration/vllm/uc_connector.py | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-) diff --git a/ucm/integration/vllm/uc_connector.py b/ucm/integration/vllm/uc_connector.py index 421011ca..074ddf4b 100644 --- a/ucm/integration/vllm/uc_connector.py +++ b/ucm/integration/vllm/uc_connector.py @@ -492,14 +492,23 @@ def get_finished(self, finished_req_ids: set[str]) -> tuple[set[str], set[str]]: continue unfinished_tasks = [] for task in tasks: - ret = self.connector.check(task) - if ret == -1: + ret, finish = self.connector.check(task) + if ret != 0: + logger.error( + f"Task {task} failed, check return {ret} for request {req_id}" + ) + self._load_failed_reqs.add(req_id) + break + if not finish: unfinished_tasks.append(task) continue - elif ret == 0 and self.connector.wait(task) == 0: - continue - self._load_failed_reqs.add(req_id) - break + wret = self.connector.wait(task) + if wret != 0: + logger.error( + f"Task {task} failed, wait return {wret} for request {req_id}" + ) + self._load_failed_reqs.add(req_id) + break if not unfinished_tasks: done_recving.add(req_id) self._need_load_reqs[req_id] = unfinished_tasks