FastAPI 的接口中使用 asyncio.Lock 效率怎么这么低？获取和释放都要几秒

app = FastAPI() lock: Optional[asyncio.Lock] = None spider = Spider() @app.get('/xxx') async def xxx(s: str): t1 = time() await lock.acquire() t2 = time() r = await spider.check(s) t3 = time() lock.release() t4 = time() logging.info(f"等待耗时{t2-t1}s,check 耗时{t3-t2}s,归还耗时{t4-t3}s") return r async def main(): log_cOnfig= generate_logging_dict() loop = asyncio.get_event_loop() global lock lock = asyncio.Lock() cOnfig= Config(app=app, loop=loop, log_cOnfig=log_config, workers=1, host="0.0.0.0", port=8000) server = Server(config) tasks = [server.serve()] await asyncio.gather(*tasks) await spider.close() if __name__ == '__main__': config_root_logger() asyncio.run(main()) 

获取 lock 和归还 lock 的耗时居然在 5~8s 吗？这也太奇葩了吧。 选取几次代表性的日志如下：

等待耗时 5.4836273193359375e-06s,check 耗时 1.3153741359710693s,归还耗时 8.58306884765625e-06s

等待耗时 5.9604644775390625e-06s,check 耗时 0.6164107322692871s,归还耗时 1.1205673217773438e-05s

等待耗时 6.67572021484375e-06s,check 耗时 0.5971970558166504s,归还耗时 8.344650268554688e-06s