Background task in fastapi making blocking requests

Question

I have a resource intensive async method that i want to run as a background task. Example code for it looks like this:

@staticmethod
async def trigger_task(id: str, run_e2e: bool = False):
    try:
        add_status_for_task(id)
        result1, result2 = await task(id)
        update_status_for_task(id, result1, result2)
    except Exception:
        update_status_for_task(id, 'FAIL')


@router.post("/task")
async def trigger_task(background_tasks: BackgroundTasks):
    background_tasks.add_task(EventsTrigger.trigger_task)
    return {'msg': 'Task submitted!'}

When i trigger this endpoint, I expect an instant output: {'msg': 'Task submitted!'}. But instead the api output is awaited till the task completes. I am following this documentation from fastapi.

fastapi: v0.70.0 python: v3.8.10

I believe the issue is similar to what is described here. Request help in making this a non-blocking call.

Does this answer your question? [fastapi asynchronous background tasks blocks other requests?](https://stackoverflow.com/questions/67599119/fastapi-asynchronous-background-tasks-blocks-other-requests) — mihi, Dec 02 '21 at 11:15
Does this answer your question? [FastAPI runs api-calls in serial instead of parallel fashion](https://stackoverflow.com/questions/71516140/fastapi-runs-api-calls-in-serial-instead-of-parallel-fashion) — Chris, Mar 10 '23 at 17:57

score 0 · Answer 1 · answered Nov 28 '21 at 08:07

0

What I have learned from the github issues,

You can't use async def for task functions (Which will run in background.)
As in background process you can't access the coroutine, So, your async/await will not work.
You can still try without async/await. If that also doesn't work then you should go for alternative.

Alternative Background Solution

Celery is production ready task scheduler. So, you can easily configure and run the background task using your_task_function.delay(*args, **kwargs)
Note that, Celery also doesn't support async in background task. So, whatever you need to write is sync code to run in background.

Good Luck :)

answered Nov 28 '21 at 08:07

Mohammad Ashraful Islam

1,879
15
29

1

from fast api [docs](https://fastapi.tiangolo.com/tutorial/background-tasks/), we can actually use both async and normal methods. – udit Nov 29 '21 at 04:32
Yes. But from the bug report/issues I proposed a way. It's may not correct. Whatever!! – Mohammad Ashraful Islam Nov 29 '21 at 08:24

score 0 · Answer 2 · answered Nov 29 '22 at 02:01

Unfortunately you seem to have oversimplified your example so it is a little hard to tell what is going wrong.

But the important question is: are add_status_for_task() or update_status_for_task() blocking? Because if they are (and it seems like that is the case), then obviously you're going to have issues. When you run code with async/await all the code inside of it needs to be async as well.

This would make your code look more like:

async def trigger_task(id: str, run_e2e: bool = False):
    try:
        await add_status_for_task(id)
        result1, result2 = await task(id)
        await update_status_for_task(id, result1, result2)
    except Exception:
        await update_status_for_task(id, 'FAIL')


@router.post("/task/{task_id}")
async def trigger_task(task_id: str, background_tasks: BackgroundTasks):
    background_tasks.add_task(EventsTrigger.trigger_task, task_id)
    return {'msg': 'Task submitted!'}

score -2 · Answer 3 · answered Nov 28 '21 at 20:51

-2

How are you running your app?

According to the uvicorn docs its running with 1 worker by default, which means only one process will be issued simultaneously. Try configuring your uvicorn to run with more workers. https://www.uvicorn.org/deployment/

$ uvicorn example:app --port 5000 --workers THE_AMOUNT_OF_WORKERS
or
uvicorn.run("example:app", host="127.0.0.1", port=5000, workers=THE_AMOUNT_OF_WORKERS)

answered Nov 28 '21 at 20:51

Yair Siman Tov

1
2

This does not seems to be a correct approach because with this, if i run 4 workers and have 20 users, then at a time only 4 of them will be able to run their task because this was a blocking call. – udit Nov 29 '21 at 04:27
It should handle it right if you are using async await in python right. Check that the thing u do support async work. I would recommend u to think about using python celery, my team use it to run async tasks for a large number of users. Depends on your amount of users and on the amount of time required to complete the task i would recommend scaling your service. – Yair Siman Tov Dec 01 '21 at 15:23

Background task in fastapi making blocking requests

3 Answers3