python3 xmlrpc socket timeout in docker

Question

I'm trying to debug a very simple xmlrpc client/server. The server is running locally on a MacOS running sonoma. Here's the server code for it:

from xmlrpc.server import SimpleXMLRPCServer

def endpoint():
    import time
    time.sleep(0.02)

with SimpleXMLRPCServer(("0.0.0.0", 8886), allow_none=True) as server:
    server.register_function(endpoint)
    server.serve_forever()

The client is running within a docker container with image Ubuntu:22.04 on the same host as the server (same behaviour was noted using python:3.9 image.)

import xmlrpc.client
import cProfile

SERVER_PORT = 8886
SERVER_IP="<IP>"

transp = xmlrpc.client.Transport()
transp.timeout = 1

proxy = xmlrpc.client.ServerProxy(f"http://{SERVER_IP}:{SERVER_PORT}/", allow_none=True, transport=transp)

def read(proxy):
    import time
    times_stuck = []
    deadline = time.monotonic() + 60 * 15

    while(time.monotonic() < deadline):
        profiler = cProfile.Profile()
        s_time = time.monotonic()
        profiler.enable()
        proxy.endpoint()
        profiler.disable()
        e_time = time.monotonic()

        print("[STUCK: {}] [{:.2f}s] endpoint".format(len(times_stuck) ,e_time-s_time))

        if e_time - s_time > 1.0:
            print("\n>> GOT STUCK: {:.2f}s\n".format(e_time - s_time))
            profiler.print_stats(sort='cumulative')
            times_stuck.append(e_time - s_time)

        time.sleep(0.5)

    return times_stuck

If I run the exact same code using a regular virtual environment I cannot reproduce the code getting stuck

The profiler shows an unusual long waiting period on socket connect

[STUCK: 5] [0.03s] endpoint
[STUCK: 5] [64.32s] endpoint

>> GOT STUCK: 64.32s

         450 function calls in 64.316 seconds

   Ordered by: cumulative time

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
        1    0.000    0.000   64.316   64.316 client.py:1111(__call__)
        1    0.000    0.000   64.316   64.316 client.py:1442(__request)
        1    0.000    0.000   64.316   64.316 client.py:1150(request)
        1    0.000    0.000   64.316   64.316 client.py:1163(single_request)
        1    0.000    0.000   64.289   64.289 client.py:1266(send_request)
        1    0.000    0.000   64.289   64.289 client.py:1300(send_content)
        1    0.000    0.000   64.289   64.289 client.py:1265(endheaders)
        1    0.000    0.000   64.289   64.289 client.py:1027(_send_output)
        2    0.000    0.000   64.289   32.144 client.py:968(send)
        1    0.000    0.000   64.289   64.289 client.py:945(connect)
        1    0.000    0.000   64.289   64.289 socket.py:691(create_connection)
        1   64.289   64.289   64.289   64.289 {method 'connect' of '_socket.socket' objects}
        1    0.000    0.000    0.027    0.027 client.py:1329(getresponse)
        1    0.000    0.000    0.027    0.027 client.py:312(begin)
        1    0.000    0.000    0.026    0.026 client.py:279(_read_status)

I noticed that sometimes it get's stuck for over 60 seconds, sometimes 30 and other times 15 seconds

the fin_timeout in the container is the default one, and is set to net.ipv4.tcp_fin_timeout=60, so I think something is causing it to hit this timeout.

Tried updating the docker engine, but no luck. Also tried starting the container with different networking modes, but still nothing.

Anyone has any ideas as to what might be causing this?