socket closedown problem on DU4.0D patch 2

From: Wigg, David <David.Wigg_at_wang.com>
Date: Wed, 04 Nov 1998 16:49:49 +0100

Hi,
        I have an apllication that has a server component on an Alpha server
running DU4.0D with consolidated patch 2. The app opens a socket, listens
for connections, forks a process to deal with the incoming data, (the
process will be very short lived and will exit shortly after reading the
data and squirreling it away somewhere) and continues listening.

There is a strangeness happening. The app works fine in a live environment,
but under test, when the number of incoming connections is high I get a lot
of connections left behind in TIME_WAIT in netstat. Eventually there are
enough of them to prevent further connections.

The question is this:
Why are the connections in TIME_WAIT? I have an analyser on the network and
the whole transaction only takes 9 packets. Each side sends a FIN and ACK
and *looks* to have shut down correctly, so I would expect the connections
to have cleared down. Does anyone know what the TIME_WAIT is for, and why I
don't see a clean socket shutdown?

Thanks


Best Regards,

Dave

David Wigg
Senior Consultant
Solutions Integration, Wang Global,
Phone: +44(0)161 848 4146
Email : david.wigg_at_wang.com
Received on Wed Nov 04 1998 - 15:52:58 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:38 NZDT