Requests Hang Indefinitely (Hours+) Until Manual Gunicorn Restart, pre_requestLogs Not Printed Gunicorn 20.0.4

[wl779754345-netizen]

The requester sends a request but never receives a response packet. If the requester does not set a timeout, the request remains stuck and unresponsive for several hours or even longer, until the Gunicorn service is manually restarted (the restart likely breaks the connection, causing the client to finally receive an exception).
This issue is not reproducible on demand. When a request hangs without a response, the service logs indicate that other requests are being processed normally both immediately before and after the problematic event.
A pre_requesthook implementation was added for logging, but no log entries from it are printed when this hang occurs.

this is a Gunicorn Django service with no proxy involved; the requester connects directly to the Gunicorn service. What could be the possible cause?

it my pre_request hook:

def pre_request(worker, req):

    Args:
        worker: Gunicorn worker
        req: req
    """
    try:
        request_id = get_header(req.headers, 'REQUEST-ID') or str(uuid.uuid4())
        start_time = time.time()
        
        req.start_time = start_time
        req.request_id = request_id

        formatted_time = datetime.fromtimestamp(start_time).strftime('%Y-%m-%d %H:%M:%S.%f')[:-3]

        logger.info(
            "[%s] [%s] pre request: %s %s, Worker PID: %s",
            request_id,
            formatted_time,
            req.method,
            req.path,
            worker.pid
        )
        
        logger.debug(
            "[%s] Request details - Headers: %s, Query: %s, Remote: %s",
            request_id,
            dict(req.headers) if req.headers else {},
            req.query if req.query else {},
            req.remote_addr if hasattr(req, 'remote_addr') else 'Unknown'
        )
    except Exception as e:
        logger.error("pre_request hook error: %s", str(e))

it is my configuration

import os

bind = "0.0.0.0:{}".format(os.environ.get("MMDCAUTH_PORT", "8000"))

_cpu_cores =12
_run_mode = os.environ.get("RUN_MODE", "")

if _run_mode != "product":
    workers = 9
else:
    workers = _cpu_cores * 2 + 1

worker_class = "gthread"

threads = 64

timeout = 120

graceful_timeout = 30

max_requests = 500000

max_requests_jitter = 5000

backlog = 40960

keepalive = 5

limit_request_line = 0

limit_request_field_size = 0

limit_request_fields = 200


_log_dir = "/home/xx/xx/var/log"

accesslog = "{}/gunicorn_access.log".format(_log_dir)

errorlog = "{}/gunicorn_error.log".format(_log_dir)

loglevel = os.environ.get("GUNICORN_LOG_LEVEL", "info")

access_log_format = '%(h)s %(l)s %(u)s %(t)s "%(r)s" %(s)s %(b)s "%(f)s" "%(a)s" %(D)s %(p)s'

capture_output = True

syslog = False


daemon = False

pidfile = "/home/xx/xx/var/run/gunicorn.pid"


# umask
umask = 0o022

chdir = "/home/xx/xx"

tmp_upload_dir = None


forwarded_allow_ips = "*"

proxy_protocol = False


preload_app = False

reuse_port = False


from settings.gunicorn_hooks import pre_request, post_request


def on_starting(server):
    pass


def on_reload(server):
    pass


def when_ready(server):
    pass


def worker_int(worker):
    pass


def worker_abort(worker):
    pass


def pre_fork(server, worker):
    pass


def post_fork(server, worker):
    pass


def post_worker_init(worker):
    pass


def worker_exit(server, worker):
    pass


def nworkers_changed(server, new_value, old_value):
    pass


def on_exit(server):
    pass

[benoitc]

Do you reproduce it with latest version of Gunicorn?

[wl779754345-netizen]

We have not updated the Gunicorn version. Our leader hopes to identify the root cause and fix it, so we have spent considerable time adjusting parameters, implementing asynchronous logging, optimizing request latency, and so on. Currently, this issue occurs once every few weeks. Do you recommend upgrading the version?

[benoitc]

Yes. Latest version of gthread worker has considerablyben improved since the version 24. Le ven. 13 févr. 2026 à 09:08, wl779754345-netizen ***@***.***> a écrit :

…

We have not updated the Gunicorn version. Our leader hopes to identify the root cause and fix it, so we have spent considerable time adjusting parameters, implementing asynchronous logging, optimizing request latency, and so on. Currently, this issue occurs once every few weeks. Do you recommend upgrading the version? — Reply to this email directly, view it on GitHub <#3506 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAADRIWUJENM5HNJQIIVMI34LWA7HAVCNFSM6AAAAACU74HGFWVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTKNZZGI3DINQ> . You are receiving this because you commented.Message ID: ***@***.***>