CLOSE_WAIT issue (Software fails to close connections)

I've got a server application which has a bad habit of failing to
close socket connections after doing web services requests to other
servers. This leaves behind many sockets in the CLOSE_WAIT state.
Eventually (after 3-4 days) the number of sockets in this condition
becomes so large that it exceeds the maximum open files limit.

Unfortunately, the problem is in a third-party library which I cannot
immediately fix.

Is it possible to configure Linux to abandon CLOSE_WAIT sockets after
a timeout period? Is there a reason why the default behavior is no
timeout?

Frank

flarosa <frank@franklarosa.com> wrote:
> Unfortunately, the problem is in a third-party library which I
> cannot immediately fix.

> Is it possible to configure Linux to abandon CLOSE_WAIT sockets
> after a timeout period? Is there a reason why the default behavior
> is no timeout?

Because absent other information, it is impossible for TCP to tell the
difference between a simplex connection (send only relative to the
CLOSE_WAIT end) and an incomplete connection tear-down.

Anything in the stack that arbitrarily nuked such a connection from
the CLOSE_WAIT end is at best guessing.

Even TCP Keepalives cannot really deal with this since it is likely
there is still a TCP endpoint in FIN_WAIT_2 at the other end.

This will be an excellent test of the third party's responsiveness to
bugreports.

rick jones
--
The glass is neither half-empty nor half-full. The glass has a leak.
The real question is "Can it be patched?"
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...