How does Python Eventlet concurrently request?

Question

The client uses eventlet with the following code.

import eventlet
import urllib.request

urls = [
    "http://localhost:5000/",
    "http://localhost:5000/",
    "http://localhost:5000/",
]


def fetch(url: str) -> str:
    return urllib.request.urlopen(url).read()


pool = eventlet.GreenPool(1000)
for body in pool.imap(fetch, urls):
    print("got body", len(body), body)

The server side is built using fastapi, so why can we distinguish if the client is really concurrently requested, we add a delay

from fastapi import FastAPI
import uvicorn
import time

app = FastAPI()


@app.get('/')
def root():
    time.sleep(3)
    return {"message": "Hello World"}


if __name__ == '__main__':
    uvicorn.run("api:app", host="0.0.0.0", port=5000)

Start the server first, then run the client code, using Linux's time command to see how long it takes.

got body 25 b'{"message": "Hello World"}'
got body 25 b'{"message": "Hello World"}'
got body 25 b'{"message": "Hello World"}'
python -u "013.py" 0.25s user 0.01s system 2% cpu 9.276 total

As you can see, it took 9 seconds instead of 3 seconds, so this ``eventlet` is not parallel, am I using it wrong? What is the correct way to open it?

Translated with www.DeepL.com/Translator (free version)