Apache and the FastCGI manager losing track of UNIX sockets

infinity@gamersinn.com
Sat, 6 Sep 1997 12:35:38 -0400 (EDT)

From: infinity@gamersinn.com
Date: Sat, 6 Sep 1997 12:35:38 -0400 (EDT)
To: fastcgi-developers@OpenMarket.com
Subject: Apache and the FastCGI manager losing track of UNIX sockets
Message-Id: <Pine.LNX.3.93.970906122713.31407A-100000@infinity.gamersinn.com>


I'm running apache 1.2.1 and the first beta of mod_fastcgi 2 (I think, there's
no version listed in the actual file :P) on a Linux 2.0.25 machine, and after
a days or so's uptime,  so many requests, so many times of exiting, or
something, the reqests coming in for a FastCGI either tries to connect to the
wrong UNIX socket, or the FastCGI application is listening on the wrong UNIX
socket.

I do a system strace on a FastCGI application that seem to be hanging, and it
is sitting on an accept(), I try to pull the page up again, still sitting on
the same accept(). I kill that process, start the system trace on the new one
that is spawned, hits the first accept() and never gets past that point. The
only way to actually rectify this situation is to restart apache.

All of my FastCGI processes are in perl, BTW. Has anyone else had something
similar happen? If so, how the heck do I fix it? :)

	Thanks,
	Matt Burgoon