why is there now execption for windows? trying to listen twice tothe same port

[News123]

I'm having a small multiprocessing manager:

# ##########################
import socket,sys
from multiprocessing.managers import BaseManager

mngr = BaseManager(address=('127.0.0.1',8089),authkey='verysecret')
try:
srvr = mngr.get_server()
except socket.error as e:
print "probably address already in use"
sys.exit()
print "serving"
srvr.serve_forever()


Under linux this script can only be run once.
The second call will raise an exception, as the previous program is
already listening to pot 8089.


Under Windows however the program can be started twice.
and will print twice "serving". This surprises me

Howver only one of them will successfully listen and respond to
connections, but I don't get an exception.

Is this to be expected?

I'd like to be sure, that the manager is only started once and that it
can be started anytime it's not up.


N

 

[Irmen de Jong]

I'm having a small multiprocessing manager:

# ##########################
import socket,sys
from multiprocessing.managers import BaseManager

mngr = BaseManager(address=('127.0.0.1',8089),authkey='verysecret')
try:
srvr = mngr.get_server()
except socket.error as e:
print "probably address already in use"
sys.exit()
print "serving"
srvr.serve_forever()


Under linux this script can only be run once.
The second call will raise an exception, as the previous program is
already listening to pot 8089.


Under Windows however the program can be started twice.
and will print twice "serving". This surprises me

My guess is that somewhere in the multiprocessing package the SO_REUSEADDR option is
used on the socket. And that option has different semantics on Windows than it has on
other operating systems. At least one of them being the possibility of multiple bindings
on the same port without getting an error.
See http://bugs.python.org/issue2550.
Also see the code comments to bind_port in Lib/test/test_support.py, that suggests using
SO_EXCLUSIVEADDRUSE instead under Windows, but that is not much use to you unless you
monkeypatch the multiprocessing package code.

Hope this helps
-irmen

 

[News123]

Hi Irmen,

My guess is that somewhere in the multiprocessing package the
SO_REUSEADDR option is used on the socket. And that option has different
semantics on Windows than it has on other operating systems. At least
one of them being the possibility of multiple bindings on the same port
without getting an error.
See http://bugs.python.org/issue2550.
Also see the code comments to bind_port in Lib/test/test_support.py,
that suggests using SO_EXCLUSIVEADDRUSE instead under Windows, but that
is not much use to you unless you monkeypatch the multiprocessing
package code.

Yes this might be.


Then I think, that I should report it either as bug / enhancement
request. However I am not at all aware of the issue reporting process
concerning python. (special site / module developer, PEP . . . )

For the short term:
I wonder now how to solve my problem that only one manager is started
and the second one should just aborts.

Some days ago I started a discussion in the thread
"how to start a python script only once".
I hoped, that the listening socket of the manager would just take care
of the issue.

N

 

[Terry Reedy]

Hi Irmen,



Yes this might be.


Then I think, that I should report it either as bug / enhancement

I agree. I would guess that the difference is not intentional.

request. However I am not at all aware of the issue reporting process
concerning python. (special site / module developer, PEP . . . )

bugs.python.org

Terry Jan Reedy

 

[News123]

I agree. I would guess that the difference is not intentional.


bugs.python.org

Thanks,


I created an account and submitted the issue


N