2

We are using Quart (Flask+asyncio) Python web framework. Every time the request is processed and the response is sent to a client, this (or similar) message is logged:

WARNING:asyncio:Executing <Task pending name='Task-11' coro=<ASGIHTTPConnection.handle_request() running at /usr/local/lib/python3.8/site-packages/quart/asgi.py:102> wait_for=<Future pending cb=[<TaskWakeupMethWrapper object at 0x7f742ba41ee0>()] created at /usr/local/lib/python3.8/asyncio/base_events.py:422> cb=[_wait.._on_completion() at /usr/local/lib/python3.8/asyncio/tasks.py:518] created at /usr/local/lib/python3.8/site-packages/quart/asgi.py:46> took 2.700 seconds

Since it is WARNING, we are kind of worried about what this could be. Does anyone have any idea why a log like this appears?

Also, I have seen more logs starting <Task pending name... before. Does anyone know what these logs are?

To replicate a similar log message, it is enough just to do this:

import time

from quart import Quart


app = Quart(__name__)


@app.route('/', methods=['POST'])
async def endpoint():
    time.sleep(0.5)
    return '', 200

If I set sleep() to a lower value (e.g. 0.05), the log message is not printed out.

davidism
  • 121,510
  • 29
  • 395
  • 339
Michal Moravik
  • 1,213
  • 1
  • 12
  • 23

1 Answers1

3

asyncio and other event loops require the tasks to yield control back to the event loop periodically so that it can switch to another task and execute tasks concurrently. This warning is indicating that a task is taking a long time between yields, thereby 'blocking' the event loop.

It is likely this is happening as your code is either doing something CPU intensive, or more likely is using non-asyncio IO e.g. using requests. You should investigate this as it will degrade your servers ability to serve multiple requests concurrently.

pgjones
  • 6,044
  • 1
  • 14
  • 12
  • 1
    Thanks a lot for your answer and the framework. So the message is related to the async behavior. Now it makes sense since our service has both CPU-intensive tasks as well as non-asyncio libraries such as requests and OpenCV. When we implemented the code, we realized that we don't leverage the advantages of using Quart. – Michal Moravik Dec 26 '21 at 21:13
  • 2
    This feature, https://pgjones.gitlab.io/quart/how_to_guides/sync_code.html, may help you with the CPU intensive tasks whilst allowing you to use the async advantages for request IO. – pgjones Dec 27 '21 at 18:11
  • Thanks for the link! :) Looks like something that can solve the issue. – Michal Moravik Dec 29 '21 at 08:52