Issue 1151968: (XMLRPC) multitude of sockets ending up in TIME_WAIT (original) (raw)

Issue1151968

Created on 2005-02-25 18:00 by jonsv322, last changed 2022-04-11 14:56 by admin. This issue is now closed.

Messages (4)
msg24365 - (view)	Author: Jonas Widén (jonsv322)	Date: 2005-02-25 18:00
A problem with a multitude of sockets ending up in TIME_WAIT when stressing a system with XML-RPC calls. This can cause a Windows network to missbehave. A solution on Windows platform is to set the sockopt NOLINGER with parameters (1, 0) for the socket before the socket is closed. This vill cause that Windows will release the socket direct without going into TIME_WAIT. The solution in code: In httplib.py@somewhere: NOLINGER = struct.pack('HH', 1, 0) # Release the resource back to the system if socket closed In httplib.py@632: self.sock.setsockopt(socket.SOL_SOCKET, socket.SO_LINGER, NOLINGER) self.sock.close() # close it manually... there may be other refs
msg24366 - (view)	Author: Jonas Widén (jonsv322)	Date: 2005-02-25 20:13
Logged In: YES user_id=352221 Read section, SO_LINGER at following address: http://msdn.microsoft.com/library/default.asp?url=/library/en-us/winsock/winsock/wspsetsockopt_2.asp
msg24367 - (view)	Author: Martin v. Löwis (loewis) *	Date: 2005-02-25 20:53
Logged In: YES user_id=21627 Why do you want to avoid the TIME_WAIT state, and why do you think it is a Python bug that it doesn't? See http://www.developerweb.net/sock-faq/detail.php?id=13 where Richard Stevens himself explains that SO_LINGER should *not* be used. It might be that SO_REUSEADDR is sufficient, see http://www.developerweb.net/sock-faq/detail.php?id=44 If you want xmlrpclib to use different socket options, you should subclass httplib.HTTP to redefine either connect() or close(), and you should subclass xmlrpclib.Transport to use your subclassed transport.
msg24368 - (view)	Author: Georg Brandl (georg.brandl) *	Date: 2006-03-17 19:29
Logged In: YES user_id=849994 Seems like Not a Bug.