This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

classification
Title: Multiprocessing module hangs on os.read() on Linux
Type: behavior Stage:
Components: IO Versions: Python 3.9
process
Status: open Resolution:
Dependencies: Superseder:
Assigned To: Nosy List: to7m
Priority: normal Keywords:

Created on 2020-12-13 23:37 by to7m, last changed 2022-04-11 14:59 by admin.

Messages (1)
msg382945 - (view) Author: (to7m) Date: 2020-12-13 23:37
As best I can tell, sometimes when a Listener or Connection is not properly initialised, the Client fails to communicate properly with it. Instead of raising an exception, the Client hangs.


receiver.py:


from multiprocessing.connection import Listener

while True:
    with Listener(("localhost", 10000), authkey=b"test") as listener:
        with listener.accept() as connection:
            print(connection.recv())


client.py (intended as a stress test):


from multiprocessing.connection import Client

for i in range(1000):
    successfully_sent = False
    while not successfully_sent:
        try:
            with Client(("localhost", 10000), authkey=b"test") as client:
                client.send(i)
        except (ConnectionRefusedError, ConnectionResetError):
            continue
        successfully_sent = True


Also noteworthy: I posted on StackExchange (https://stackoverflow.com/questions/65276145/multiprocessing-receive-all-messages-from-multiple-runtimes) and it seems that the code there (only 1000 messages) took around an hour to run for a Windows user, whereas it would take less than a second to successfully run on Linux.
History
Date User Action Args
2022-04-11 14:59:39adminsetgithub: 86797
2020-12-14 18:26:27to7msetfiles: - receive.py
2020-12-13 23:37:58to7mcreate