如何捕获仅含asyncio栈跟踪的OSError？（Uvicorn+FastAPI环境）

我在Windows系统上用FastAPI和Uvicorn搭建了运行异步函数的API服务器。服务器偶尔会因asyncio抛出网络错误导致无响应，但我的脚本没法捕获这类错误。我想通过try-catch捕获错误后关闭并重启服务器，但错误回溯里完全没有我的代码。

项目文件结构 #

1. API定义文件 app.py：

from fastapi import FastAPI

app = FastAPI()


@app.get("/check-status")
async def root():
    return {"Alive": True}

2. 服务器配置文件 config.json：

{
    "host": "0.0.0.0",
    "port": 80
}

3. 启动脚本 run_app.py（已包含错误捕获逻辑）：

import uvicorn
import json
from datetime import datetime

if __name__ == '__main__':
    with open('path/to/config.json', 'r') as f:
        config = uvicorn.Config(**json.load(f))
    server = uvicorn.Server(config)
    try:
        server.run()
    except Exception as e:
        err_timestamp = datetime.now().ctime()
        print(f'{err_timestamp}: Exception raised\n{e}\nRestarting...')
        server.shutdown()
        server.run()

触发的错误信息 #

服务器仍会因以下错误崩溃：

Task exception was never retrieved
future: <Task finished name='Task-33094' coro=<IocpProactor.accept.<locals>.accept_coro() done, defined at C:\Users\Administrateur\AppData\Local\Programs\Python\Python38\lib\asyncio\windows_events.py:563> exception=OSError(22, 'The specified network name is no longer available', None, 64, None)>
Traceback (most recent call last):
  File "C:\Users\Administrateur\AppData\Local\Programs\Python\Python38\lib\asyncio\windows_events.py", line 566, in accept_coro
    await future
  File "C:\Users\Administrateur\AppData\Local\Programs\Python\Python38\lib\asyncio\windows_events.py", line 812, in _poll
    value = callback(transferred, key, ov)
  File "C:\Users\Administrateur\AppData\Local\Programs\Python\Python38\lib\asyncio\windows_events.py", line 555, in finish_accept
    ov.getresult()
OSError: [WinError 64] The specified network name is no longer available

Accept failed on a socket
socket: <asyncio.TransportSocket fd=732, family=AddressFamily.AF_INET, type=SocketKind.SOCK_STREAM, proto=6, laddr=('0.0.0.0', 80)>
Traceback (most recent call last):
  File "C:\Users\Administrateur\AppData\Local\Programs\Python\Python38\lib\asyncio\proactor_events.py", line 818, in loop
    conn, addr = f.result()
  File "C:\Users\Administrateur\AppData\Local\Programs\Python\Python38\lib\asyncio\windows_events.py", line 566, in accept_coro
    await future
  File "C:\Users\Administrateur\AppData\Local\Programs\Python\Python38\lib\asyncio\windows_events.py", line 812, in _poll
    value = callback(transferred, key, ov)
  File "C:\Users\Administrateur\AppData\Local\Programs\Python\Python38\lib\asyncio\windows_events.py", line 555, in finish_accept
    ov.getresult()
OSError: [WinError 64] The specified network name is no longer available

这个错误完全发生在asyncio内部，没有经过我的代码抛出，也没法复现测试。已知这类WinError是网络问题导致的，需要实现错误捕获来处理这种情况。

环境信息 #

• Python 3.8.10
• Uvicorn 0.17.4
• FastAPI 0.73.0
• Windows 10

解决方案 #

1. 捕获asyncio未处理的任务异常 #

这类错误属于异步任务未捕获的异常，不会直接向上抛出到server.run()的调用栈，常规try-catch无法捕获。可以给asyncio事件循环设置全局异常处理器，捕获这类未处理异常：

修改run_app.py如下：

import uvicorn
import json
import asyncio
from datetime import datetime

def handle_uncaught_exception(loop, context):
    # 打印错误日志
    err_timestamp = datetime.now().ctime()
    print(f"{err_timestamp}: 捕获到未处理异常")
    print(f"异常详情: {context.get('exception', context['message'])}")
    # 异步触发服务器重启
    asyncio.create_task(restart_server())

async def restart_server():
    global server
    print("正在关闭服务器...")
    server.shutdown()
    print("正在重启服务器...")
    # 重新创建Server实例，避免复用旧实例的状态问题
    with open('path/to/config.json', 'r') as f:
        config = uvicorn.Config(**json.load(f))
    server = uvicorn.Server(config)
    await server.serve()

if __name__ == '__main__':
    # 获取或创建事件循环
    loop = asyncio.get_event_loop()
    # 设置全局未捕获异常处理器
    loop.set_exception_handler(handle_uncaught_exception)
    
    with open('path/to/config.json', 'r') as f:
        config = uvicorn.Config(**json.load(f))
    server = uvicorn.Server(config)
    
    try:
        loop.run_until_complete(server.serve())
    except Exception as e:
        err_timestamp = datetime.now().ctime()
        print(f'{err_timestamp}: 捕获到异常\n{e}\n正在重启...')
        loop.run_until_complete(restart_server())

2. 升级依赖版本 #

当前使用的Uvicorn版本（0.17.4）较旧，后续版本对Windows下的asyncio异常处理有优化。可尝试升级到稳定新版本：

pip install --upgrade uvicorn fastapi

3. 线程池监控启动（备选方案） #

如果全局异常处理仍不生效，可以用线程启动服务器，监控线程状态，当线程退出时自动重启：

import uvicorn
import json
import threading
from datetime import datetime
import time

def start_server():
    with open('path/to/config.json', 'r') as f:
        config = uvicorn.Config(**json.load(f))
    server = uvicorn.Server(config)
    server.run()

if __name__ == '__main__':
    while True:
        try:
            err_timestamp = datetime.now().ctime()
            print(f"{err_timestamp}: 启动服务器...")
            server_thread = threading.Thread(target=start_server)
            server_thread.start()
            server_thread.join()
        except Exception as e:
            err_timestamp = datetime.now().ctime()
            print(f'{err_timestamp}: 服务器异常退出\n{e}\n10秒后重启...')
            time.sleep(10)

内容的提问来源于stack exchange，提问作者Aidanpm