Common Mistakes Using Python3 asyncio

Table of Contents

1 Introduction

Python3 asyncio is a powerful asynchronous library. However, the complexity results in a very steep learning curve. 1 Compared to C# async/await, the interfaces of Python3 asyncio is verbose and difficult to use. And the document is somewhat difficult to understand. (Even Guido admited the document is not clear! 2) Here I summerize some of the common mistakes when using asyncio.

2 RuntimeWarning: coroutine foo was never awaited

This runtime warning can happen in many scenarios, but the cause are same: A coroutine object is created by the invocation of an async function, but is never inserted into an EventLoop.

Consider following async function foo():

async def foo():
    # a long async operation
    # no value is returned

If you want to call foo() as an asynchronous task, and doesn't care about the result:

# RuntimeWarning: coroutine foo was never awaited

This is because invoking foo() doesn't actually runs the function foo(), but created a "coroutine object" instead. This "coroutine object" will be executed when current EventLoop gets a chance: awaited/yield from is called or all previous tasks are finished.

To execute an asynchronous task without await, use loop.create_task() with loop.run_until_complete():

task = loop.create_task(foo())

If the coroutine object is created and inserted into an `EventLoop`, but was never finished, the next warning will appear.

3 Task was destroyed but it is pending!

The cause of this problem is that the EventLoop closed right after canceling pending tasks. Because the Task.cancel() "arranges for a CancelledError to be thrown into the wrapped coroutine on the next cycle through the event loop", and "the coroutine then has a chance to clean up or even deny the request using try/except/finally." 3

To correctly cancel all tasks and close EventLoop, the EventLoop should be given the last chance to run all the canceled, but unfinished tasks.

For example, this is the code to cancel all the tasks:

def cancel_tasks():
    # get all task in current loop
    tasks = Task.all_tasks()           
    for t in tasks:


Below code correctly handle task canceling and clean up. It starts the EventLoop by calling loop.run_forever(), and cleans up tasks after receiving loop.stop():

    # run_forever() returns after calling loop.stop()
    tasks = Task.all_tasks()
    for t in [t for t in tasks if not (t.done() or t.cancelled())]:
        # give canceled tasks the last chance to run

4 Task/Future is awaited in a different EventLoop than it is created

This error is especially surprising to people who are familiar with C# async/await. It is because most of asyncio is not thread-safe, nor is asyncio.Future or asyncio.Task. Also don't confuse asyncio.Future with concurrent.futures.Future because they are not compatible (at least until Python 3.6): the latter is thread-safe while the former is not.

In order to await an asyncio.Future in a different thread, asyncio.Future can be wrapped in a concurrent.Future:

def wrap_future(asyncio_future):
    def done_callback(af, cf):
        except Exception as e:

    concur_future = concurrent.Future()
        lambda f: done_callback(f, cf=concur_future))
    return concur_future

asyncio.Task is a subclass of asyncio.Future, so above code will also work.