Gunicorn 24.0 has been released!

[benoitc]

Highlights

ASGI Worker (Beta)

Gunicorn now includes native asyncio-based ASGI support for running async Python frameworks like FastAPI, Starlette, and Quart — without external dependencies.

gunicorn -k asgi myapp:app

Features:

• HTTP/1.1 with keepalive connections
• WebSocket support
• Lifespan protocol for startup/shutdown hooks
• Optional uvloop for improved performance

New settings: --asgi-loop, --asgi-lifespan, --root-path

uWSGI Binary Protocol

Support for receiving requests from nginx via uwsgi_pass directive:

 gunicorn --protocol uwsgi myapp:app

New settings: --protocol uwsgi, --uwsgi-allow-from

Documentation

Migrated to MkDocs with Material theme for improved navigation and mobile experience at https://gunicorn.org

Security Updates

Updated minimum versions for async workers to address known vulnerabilities:

• eventlet >= 0.40.3 (CVE-2021-21419, CVE-2025-58068)
• gevent >= 24.10.1 (CVE-2023-41419, CVE-2024-3219)
• tornado >= 6.5.0 (CVE-2025-47287)

Install

pip install --upgrade gunicorn

Links

• https://docs.gunicorn.org/en/latest/news.html
• https://docs.gunicorn.org/en/latest/asgi.html
• https://docs.gunicorn.org/en/latest/uwsgi.html

[NZHYDROGEN]

#3447

[LincolnPuzey]

Great news! The new docs look great.

I noticed the "Documentation" and "Changelog" links on the PyPI sidebar still point to
https://docs.gunicorn.org, which isn't loading.

[benoitc]

thanks! I am working on it. It should be redirected but for some reason this doesn't work on https...

[benoitc]

this should be fixed now. Thank you for the feedback :)

[sky-cake]

Hello, are the lifespan hooks designed to run once per worker? I am running into the same situation with Hypercorn + Quart right now.

Following https://quart.palletsprojects.com/en/latest/how_to_guides/startup_shutdown/ ,

from asyncio import sleep
from quart import Quart

app = Quart(__name__)

@app.route('/')
async def hello():
    return 'hello'

@app.before_serving
async def on_start():
    await sleep(0.01)
    print('Wish to only print once, on start')

@app.after_serving
async def on_end():
    await sleep(0.01)
    print('Wish to only print once, on end')

gunicorn test:app --worker-class asgi --asgi-loop uvloop --workers 4 --asgi-lifespan on

[2026-01-24 07:18:24] [43784] [INFO] Starting gunicorn 24.1.1
[2026-01-24 07:18:24] [43784] [INFO] Listening at: http://127.0.0.1:8000 (43784)
[2026-01-24 07:18:24] [43784] [INFO] Using worker: asgi
[2026-01-24 07:18:24] [43785] [INFO] Booting worker with pid: 43785
[2026-01-24 07:18:24] [43787] [INFO] Booting worker with pid: 43787
[2026-01-24 07:18:24] [43789] [INFO] Booting worker with pid: 43789
Wish to only print once, on start
[2026-01-24 07:18:24] [43785] [INFO] ASGI server listening on http://127.0.0.1:8000
[2026-01-24 07:18:24] [43791] [INFO] Booting worker with pid: 43791
Wish to only print once, on start
[2026-01-24 07:18:24] [43787] [INFO] ASGI server listening on http://127.0.0.1:8000
Wish to only print once, on start
[2026-01-24 07:18:24] [43789] [INFO] ASGI server listening on http://127.0.0.1:8000
Wish to only print once, on start
[2026-01-24 07:18:24] [43791] [INFO] ASGI server listening on http://127.0.0.1:8000

[benoitc]

The current behavior is correct per the ASGI Lifespan Protocol specification:

"Lifespans should be executed once per event loop that will be processing requests. In a multi-process environment there will be lifespan events in each process."

With 4 workers, you get 4 separate processes, each with its own event loop and application instance. Running lifespan once per worker ensures resources like database connections are properly initialized in each process.

For operations that should run only once globally, use gunicorn's server-level hooks:

# gunicorn.conf.py
def when_ready(server):
    print("Runs once in master, before workers spawn")

[benoitc]

To add more context, gunicorn provides hooks at different levels:

Global (master process): on_starting, when_ready — runs once before workers spawn

Per-worker: ASGI lifespan startup/shutdown — runs once per worker process

Per-request: pre_request, post_request — runs before/after each HTTP request

Example:

# gunicorn.conf.py

def when_ready(server):
    # Runs once in master process
    print("Server ready!")

def pre_request(worker, req):
    # Runs before each request
    worker.log.debug("%s %s", req.method, req.path)

def post_request(worker, req, environ, resp):
    # Runs after each request
    pass

The per-request hooks work with all worker types including the ASGI worker.

[sky-cake]

Awesome, thank you so much, this is great!

https://gunicorn.org/configure/