使用asyncio.wait_for+shield仍触发超时的问题排查

我用Python结合aioboto3写了异步下载S3文件的代码，运行时部分文件能成功下载解压，但突然触发asyncio.TimeoutError导致任务取消，而且耗时还不到一分钟——我试过调大超时值，完全没用。想搞懂问题出在哪，为什么asyncio.shield没能阻止任务超时？

代码实现 #

import boto3
import botocore.exceptions as bexc
import asyncio
import aioboto3
from aiobotocore.session import get_session
import gzip
import utils.processor as processor

async def download_s3_async(self, file):   
    session = aioboto3.Session(profile_name=self.current_profile)
            
    async with session.client("s3") as client:
        start_log = processor.cf_get_start_time_from_filename(file.split("/")[3])
        end_log = processor.cf_get_end_time_from_filename(file.split("/")[3])   

        s3_obj = await client.get_object(Bucket=self.cf_bucket, Key=file)
        task = s3_obj["Body"].read()

        try:
            cat = await asyncio.wait_for(asyncio.shield(task), timeout=60000000)
        except asyncio.TimeoutError:
            print('Gave up waiting, task canceled')

        res = gzip.decompress(cat).decode("utf-8")

        print("File with {}-{} timestamp was downloaded and decompressed".format(start_log, end_log))
        return res

async def iterate(self, files):
    result = await asyncio.gather(*map(self.download_s3_async, files))
    return result

def call_async_wrapper(self, s_time, e_time):
    files = self.get_s3_objects_list(s_time, e_time)
    loop = asyncio.get_event_loop()
    results = loop.run_until_complete(self.iterate(files))
    return results      

def get_waf_actors_info(self, s_date, s_time, e_time):
    self.current_dist = "somepath/path/{}/".format(str(s_date).replace("-", ""))
    self.call_async_wrapper(s_date, s_time, e_time)

运行日志 #

...
File with 00:01:29-00:01:59 timestamp was downlaoded and decompressed
File with 00:00:24-00:00:54 timestamp was downlaoded and decompressed
File with 00:03:33-00:04:03 timestamp was downlaoded and decompressed
File with 00:07:01-00:07:31 timestamp was downlaoded and decompressed
File with 00:01:01-00:01:31 timestamp was downlaoded and decompressed
File with 00:06:14-00:06:44 timestamp was downlaoded and decompressed
File with 00:03:59-00:04:29 timestamp was downlaoded and decompressed
File with 00:03:02-00:03:32 timestamp was downlaoded and decompressed
File with 00:09:06-00:09:36 timestamp was downlaoded and decompressed
File with 00:04:12-00:04:42 timestamp was downlaoded and decompressed
File with 00:02:32-00:03:02 timestamp was downlaoded and decompressed
File with 00:02:23-00:02:53 timestamp was downlaoded and decompressed
File with 00:00:08-00:00:38 timestamp was downlaoded and decompressed
File with 00:09:31-00:10:01 timestamp was downlaoded and decompressed
File with 00:00:23-00:00:53 timestamp was downlaoded and decompressed
Gave up waiting, task canceled

1. 你误解了asyncio.shield的作用 #

asyncio.shield的功能是防止被外部任务取消，但它不会改变asyncio.wait_for本身的超时逻辑——wait_for到了设定时间依然会抛出TimeoutError，只是不会把取消信号传递给被shield的任务。但你这里设置的timeout=60000000（约16666小时），显然不是这个超时触发的，所以问题不在这。

2. 真正的元凶：aiobotocore内置的HTTP超时 #

aioboto3的底层依赖aiobotocore，它的客户端默认有自己的HTTP超时设置（connect_timeout默认60秒，read_timeout默认60秒）。当下载大文件或者网络较慢时，底层HTTP客户端会先触发超时，这个异常会让s3_obj["Body"].read()抛出错误，而这个错误被wait_for捕获后，表现出来像是asyncio.TimeoutError，和你设置的wait_for超时完全无关。

3. 代码里的额外问题：任务创建方式错误 #

你写的task = s3_obj["Body"].read()只是把一个协程对象赋值给了task，并没有真正创建异步任务。asyncio.shield需要包裹的是asyncio.Task对象，否则起不到任何保护作用。正确的写法应该是：

task = asyncio.create_task(s3_obj["Body"].read())

4. 解决步骤 #

• 调大aiobotocore的超时配置：创建S3客户端时，显式设置超时参数，覆盖默认值：

  import botocore.config
  
  async with session.client(
      "s3",
      config=botocore.config.Config(
          read_timeout=300,  # 5分钟，根据文件大小调整
          connect_timeout=60
      )
  ) as client:
  

• 修正任务创建方式：用asyncio.create_task创建任务后再用shield包裹。
• 区分异常类型：捕获botocore的超时异常，避免和asyncio的TimeoutError混淆：

  try:
      cat = await asyncio.wait_for(asyncio.shield(task), timeout=60000000)
  except asyncio.TimeoutError:
      print('Asyncio等待超时')
  except bexc.ReadTimeoutError:
      print('S3下载读取超时')
  

• 复用session：不要每次下载都创建新的aioboto3.Session，可以把session作为类属性复用，提升性能。

内容的提问来源于stack exchange，提问作者ShHolmes