Parallel requests block infinitely after exactly 100 requests using asyncio

Question

I've tried using both httpx and aiohttp, and both have this hard-coded limit.

import asyncio

import aiohttp
import httpx


async def main():
    client = aiohttp.ClientSession() 
    # client = httpx.AsyncClient(timeout=None)

    coros = [
        client.get(
            "https://query1.finance.yahoo.com/v8/finance/chart/",
            params={"symbol": "ADANIENT.NS", "interval": "2m", "range": "60d",},
        )
        for _ in range(500)
    ]

    for i, coro in enumerate(asyncio.as_completed(coros)):
        await coro
        print(i, end=", ")


asyncio.run(main())

Output -

0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99

And its just stuck at 99 with both libraries

This doesn't happen if a new Session is used for every request though.

What am I doing wrong? Isn't the whole point of asyncio to make things like this easy?

---

I tried re-writing this with threads, zmq and requests and it works great -

import zmq

N_WORKERS = 100
N_ITERS = 500

ctx = zmq.Context.instance()


def worker():
    client = requests.Session()

    pull = ctx.socket(zmq.PULL)
    pull.connect("inproc://#1")

    push = ctx.socket(zmq.PUSH)
    push.connect("inproc://#2")

    while True:
        if not pull.recv_pyobj():
            return

        r = client.get(
            "https://query1.finance.yahoo.com/v8/finance/chart/",
            params={"symbol": "ADANIENT.NS", "interval": "2m", "range": "60d",},
        )
        push.send_pyobj(r.content)


def ventilator():
    push = ctx.socket(zmq.PUSH)
    push.bind("inproc://#1")

    # distribute tasks to all workers
    for _ in range(N_ITERS):
        push.send_pyobj(True)

    # close down workers
    for _ in range(N_WORKERS):
        push.send_pyobj(False)



# start workers & ventilator
threads = [Thread(target=worker) for _ in range(N_WORKERS)]
threads.append(Thread(target=ventilator))
for t in threads:
    t.start()

# pull results from workers
pull = ctx.socket(zmq.PULL)
pull.bind("inproc://#2")

for i in range(N_ITERS):
    pull.recv_pyobj()
    print(i, end=", ")

# wait for workers to exit
for t in threads:
    t.join()

Answer 1

The problem is that you client.get(...) returns a request object with live handle to OS-level socket. Failing to close that object results in aiohttp running out of sockets, i.e. hitting the connector limit, which is 100 by default.

To fix the problem you need to close the object returned by client.get(), or use async with which will ensure that the object gets closed as soon as the with block is done. For example:

async def get(client):
    async with client.get(
            "https://query1.finance.yahoo.com/v8/finance/chart/",
            params={"symbol": "ADANIENT.NS", "interval": "2m", "range": "60d",}) as resp:
        pass

async def main():
    async with aiohttp.ClientSession() as client:
        coros = [get(client) for _ in range(500)]
        for i, coro in enumerate(asyncio.as_completed(coros)):
            await coro
            print(i, end=", ", flush=True)

asyncio.run(main())

Additionally, the aiohttp.ClientSession object should also be closed, which can also be accomplished using async with, as shown above.