Skip to content

Commit 78e5616

Browse files
bugfix
Signed-off-by: baxingpiaochong <771405853@qq.com>
1 parent fc2bcbe commit 78e5616

File tree

1 file changed

+9
-2
lines changed

1 file changed

+9
-2
lines changed

vllm_ascend/distributed/mooncake_connector.py

Lines changed: 9 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -67,12 +67,16 @@ def __init__(self):
6767
# intentionally delayed. Each entry is a tuple of (request_id,
6868
# timestamp). If a request remains in this queue for too long, it will
6969
# be force-freed.
70+
self.record_finished_requests: set[str] = set()
7071
self.delayed_free_requests: deque[Tuple[str, float]] = deque()
7172

7273
def update_done_task_count(self, request_id: str):
7374
with self.done_task_lock:
7475
self.finished_requests.add(request_id)
75-
self._remove_delayed_requests(request_id)
76+
if any(item[0] == request_id for item in self.delayed_free_requests):
77+
self._remove_delayed_requests(request_id)
78+
else:
79+
self.record_finished_requests.add(request_id)
7680

7781
def get_and_clear_finished_requests(self) -> set[str]:
7882
"""
@@ -90,7 +94,10 @@ def get_and_clear_finished_requests(self) -> set[str]:
9094
def add_delayed_request(self, request_id: str, delay_start_time: float):
9195
"""Add a delayed free request."""
9296
with self.done_task_lock:
93-
self.delayed_free_requests.append((request_id, delay_start_time))
97+
if request_id not in self.record_finished_requests:
98+
self.delayed_free_requests.append((request_id, delay_start_time))
99+
else:
100+
self.record_finished_requests.discard(request_id)
94101

95102
def _retrieve_expired_requests(self):
96103
"""Retrieve all expired delayed requests."""

0 commit comments

Comments
 (0)