httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <dgau...@arctic.org>
Subject Re: beta3 problems.. (fwd)
Date Sat, 14 Jun 1997 11:23:26 GMT

---------- Forwarded message ----------
Date: Sat, 14 Jun 1997 12:32:09 +0200
From: Eric.Schenk@dna.lth.se
To: Dean Gaudet <dgaudet-list-linux-net@arctic.org>
Cc: Eric.Schenk@dna.lth.se, Jared Mauch <jared@puck.nether.net>,
    linux-net@vger.rutgers.edu
Subject: Re: beta3 problems.. (fwd) 



Dean Gaudet <dgaudet-list-linux-net@arctic.org> writes:
>>>From a 1 million hit per day, 2.0.30 server, over one day:
>
>     40 accept: (client socket): Broken pipe
>      3 accept: (client socket): Connection refused
>   6231 accept: (client socket): Connection reset by peer
>     97 accept: (client socket): Network is unreachable
>    403 accept: (client socket): No route to host
>
>Yet a 300k hit per day, solaris 2.5.1 server on the same ethernet shows
>none of those errors.  The same for a BSDI 2.1 server on the same drop.
>
>We've had linux admins reporting this to apache for a while, but never
>thought to bring it up here.

This issue came up recently in a bug report filed with me.
What is happening is that Linux accept() will report failed
attempts to connect (e.g. a SYN packet arrived but we never made it
to the ESTABLISHED state). It appears that BSD derived stacks
silently ignore such cases. I have been meaning to track down if any
relevant standards have anything to say about this, but in any case
this behavior is easily changed and I am seriously considering this
for the 2.0.31 kernel release.

Cheers,

-- 
Eric Schenk                               www: http://www.dna.lth.se/~erics
Dept. of Comp. Sci., Lund University          email: Eric.Schenk@dna.lth.se
Box 118, S-221 00 LUND, Sweden   fax: +46-46 13 10 21  ph: +46-46 222 96 38




Mime
View raw message