[Triage] max requests leads to hung workers

[ewdurbin]

Type

Bug Report

Description

When setting --max-requests/max_requests and running under Linux, when the worker reaches its auto-restart, it crashes/hangs.

Steps to Reproduce (for bugs)

cat > app.py <<EOF
def application(environ, start_response):
    http_status = "200 OK"
    response_headers = [("Content-Type", "text/plain")]
    response_text = "Hello, World!"

    start_response(http_status, response_headers)
    return [response_text.encode("utf-8")]
EOF

cat > Dockerfile <<EOF
FROM ghcr.io/astral-sh/uv:python3.14-trixie-slim
COPY app.py /
EXPOSE 8000
CMD ["uvx", "gunicorn@24.0.0", "--max-requests", "1", "--bind", "0.0.0.0:8000", "app:application"]
EOF

docker run --rm -it -p 8000:8000 $(docker build -q .)

in another terminal:

curl localhost:8000

Observe output from running container

[2026-01-23 17:24:24 +0000] [28] [INFO] Starting gunicorn 24.0.0
[2026-01-23 17:24:24 +0000] [28] [INFO] Listening at: http://0.0.0.0:8000 (28)
[2026-01-23 17:24:24 +0000] [28] [INFO] Using worker: sync
[2026-01-23 17:24:24 +0000] [31] [INFO] Booting worker with pid: 31
[2026-01-23 17:24:31 +0000] [31] [INFO] Autorestarting worker after current request.
[2026-01-23 17:24:31 +0000] [31] [INFO] Worker exiting (pid: 31)
[2026-01-23 17:24:31 +0000] [28] [ERROR] Unhandled signal: cld
[2026-01-23 17:25:01 +0000] [28] [CRITICAL] WORKER TIMEOUT (pid:31)

Note that the worker does not recover and is not restarted.

Configuration

gunicorn --max-requests 1 app:application

Logs / Error Output

No response

Gunicorn Version

gunicorn 24.0.0

Python Version

Python 3.14

Worker Class

sync (default)

Operating System

Debian Trixie

Additional Context

No response

Checklist

• I have searched existing discussions and issues for duplicates
• I have checked the documentation and FAQ

[miketheman]

I took a gander at this, and added a --max-requests to this test, with this diff:

diff --git a/tests/test_signal_integration.py b/tests/test_signal_integration.py
index 868e1564..452cc17d 100644
--- a/tests/test_signal_integration.py
+++ b/tests/test_signal_integration.py
@@ -93,6 +93,7 @@ def gunicorn_server(app_module):
         '--access-logfile', '-',
         '--error-logfile', '-',
         '--log-level', 'info',
+        '--max-requests', '1',
         app_name
     ]

(or two, etc)

On a Debian system, test failure:

tests/test_signal_integration.py::TestSignalHandlingIntegration::test_multiple_requests_under_load - TimeoutError: timed out

The same on 23.0.0 passes.

git bisect points to b650332 - which makes a lot of sense.

[miketheman]

Looks like we're not the only ones: #3453

[benoitc]

this is fixed in 3179789 Thanks for the report! New release is landing in a couple of hours.