Re: os error thing again..

Richard Scranton (
Sat, 20 Sep 1997 23:37:48 -0400

Message-Id: <>
Date: Sat, 20 Sep 1997 23:37:48 -0400
From: Richard Scranton <>
Subject: Re: os error thing again..

What is the listen queue depth of the host operating system?  It
sounds very much like you are exhausting the host OS's resources.
When that happens, things can snowball until you are sitting with
most of your available socket connections in TIME_WAIT.  Listen
queue depth is usually a tunable kernel parameter, and you'll want
it to be fairly large.  Old versions of FreeBSD saw problems with
this if the default kernel was used because they were set up with
a listen queue depth of "5" as shipped in the 4.4BSD source.  Recent
versions have gone up to "32" and the current shipping version is
set to "128" be default.  HP and DEC give similar suggestions when
tuning their products for use as a web server host.  Good luck.  -rs wrote:
> Hello,
> I have checked on this a little more..
> First of all, -process 10 -listen-queue-depth 50 should mean in my
> oppinion that the os error happends as soon as 511 people try to get
> to the program. Meaning, if 10 are beeing processed and 500 already
> are in the queues (50 for each process), the next person will be
> rejected.
> I find this a little strange. What happends is this.
> Suddenly, OS Errors starting occureing.. for anywhere between 10
> seconds and 10 minutes.. Every connection is rejected, and there is
> one connecion every 1-2 seconds... Now, with a -processes 10 and a
> -listen-queue-depth 50, this sounds a little hard to beleive, no?
> I am useing the fcgi 1.4.x (not sure which one.. I think the
> newest non-beta one) and the newest fcgi devel lib..
> - Fabian Thylmann <aka NaTHaN>
> STATSnet - free statistics for your WebPage!

Richard Scranton - LDA Systems - Information Management Consulting  Columbus Cincinnati Cleveland Toledo Atlanta