RE: Socket Question - Host Refusing Connection

[mgarton@xxxxxxxxxxxxxxx]

Chris,

I essentially use Scott Klement's spawning server model.  I have a listener
job that receives the connections and hands it off to a server job.  At
peak volumes there are 3000 connections an hour.  I would expect the
backlog to be exceeded based on that volume.  That is why I wanted to know
if there is a way to see what is happening on the host .  Thanks.

Mark Garton
O'Reilly Auto Parts




------------------------------

message: 3
date: Thu, 4 Nov 2004 08:10:30 -0800
from: Chris Bipes <chris.bipes@xxxxxxxxxxxxxxx>
subject: RE: Socket Question - Host Refusing Connection

Without know more about your server configuration all I can do is take a
shot in the dark.  Are you accepting multiple connections or
accept/read/write/disconnect where you only process one client connection
at
a time?  If you handle multiple concurrent persistent connections, I would
spawn off the jobs and only handle a 100 or so per job.  This is the way
OS400 Telnet server works.  One listening program spawning off connections
to one of several programs handling a limited number of concurrent
persistent connections.  With a 1000 clients, you may need to re-design
your
server.

Chris Bipes

-----Original Message-----

I am occasionally getting  the following error "A remote host refused an
attempted connect operation" when my client program tries to connect. From
what I have read this is caused by exceeding the backlog on the listen()
api.  Is there anything else that can cause this error?  I have the
backlog parameter set to 400 with 1000 clients running this application.
It doesn't seem likely that it would exceed a backlog of 400.  Is there
any way I can see what is happening on the host besides adding logging to
my listener program?  Thanks.