diff --git a/vllm_ascend/distributed/mooncake_connector.py b/vllm_ascend/distributed/mooncake_connector.py index 08b2e2b2f69..930324b622f 100644 --- a/vllm_ascend/distributed/mooncake_connector.py +++ b/vllm_ascend/distributed/mooncake_connector.py @@ -435,6 +435,8 @@ def _handle_request(self, req_meta: dict[str, Any]): f"{request_id}: {e}", exc_info=True) finally: + self._send_done_signal_to_free_remote_port(request_id, remote_host, + remote_port_send_num) if all_task_done: self.task_tracker.update_done_task_count(request_id) if request_id in self.proc_not_transfer_request: @@ -446,8 +448,6 @@ def _handle_request(self, req_meta: dict[str, Any]): self._send_done_recv_signal(request_id, remote_host, remote_handshake_port, remote_port_send_num) - self._send_done_signal_to_free_remote_port(request_id, remote_host, - remote_port_send_num) def _send_done_signal_to_free_remote_port(self, request_id, remote_host, remote_port_send_num):