Request hangs in unreader until timeout is reached

[xcu]

Hi all,

I'm having a strange issue where, a few times every day, a request hangs until the worker is restarted. According to the traceback it is hanging in the chunk function of unreader.py.

The request is a POST that contains an image, with content type header being something like
multipart/form-data; boundary=XXXXYYYYYYZZZZ

Gunicorn version 20.1.0, using only one worker in the pod.

File /home/cesar/.venv/lib/python3.9/site-packages/flask_restx/reqparse.py, line 216, in parse
File /home/cesar/.venv/lib/python3.9/site-packages/flask_restx/reqparse.py, line 141, in source
File /home/cesar/.venv/lib/python3.9/site-packages/werkzeug/utils.py, line 106, in __get__
File /home/cesar/.venv/lib/python3.9/site-packages/werkzeug/wrappers/request.py, line 498, in files
File /home/cesar/.venv/lib/python3.9/site-packages/flask/wrappers.py, line 114, in _load_form_data
File /home/cesar/.venv/lib/python3.9/site-packages/werkzeug/wrappers/request.py, line 270, in _load_form_data
File /home/cesar/.venv/lib/python3.9/site-packages/werkzeug/formparser.py, line 328, in parse
File /home/cesar/.venv/lib/python3.9/site-packages/werkzeug/formparser.py, line 357, in _parse_multipart
File /home/cesar/.venv/lib/python3.9/site-packages/werkzeug/formparser.py, line 495, in parse
File /home/cesar/.venv/lib/python3.9/site-packages/werkzeug/formparser.py, line 540, in _chunk_iter
File /home/cesar/.venv/lib/python3.9/site-packages/werkzeug/wsgi.py, line 804, in readinto
File /home/cesar/.venv/lib/python3.9/site-packages/newrelic/api/wsgi_application.py, line 119, in read
File /home/cesar/.venv/lib/python3.9/site-packages/gunicorn/http/body.py, line 215, in read
File /home/cesar/.venv/lib/python3.9/site-packages/gunicorn/http/body.py, line 130, in read
File /home/cesar/.venv/lib/python3.9/site-packages/gunicorn/http/unreader.py, line 37, in read
File /home/cesar/.venv/lib/python3.9/site-packages/gunicorn/http/unreader.py, line 64, in chunk
File /home/cesar/.venv/lib/python3.9/site-packages/gunicorn/workers/base.py, line 203, in handle_abort

Any ideas of what could be causing it? I can see that the line in question of unreader.py is doing
self.sock.recv(self.mxchunk)

So it seems like it hangs while waiting for the data?

[m-aamirmumtaz]

Can you please share how much data size in request body?