Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 29 Mar 2006 11:27:47 -0500
From:      Chuck Swiger <cswiger@mac.com>
To:        scuba@centroin.com.br
Cc:        freebsd-questions@freebsd.org
Subject:   Re: Network tunning
Message-ID:  <442AB583.7090700@mac.com>
In-Reply-To: <20060329125019.V4053@trex.centroin.com.br>
References:  <20060329125019.V4053@trex.centroin.com.br>

next in thread | previous in thread | raw e-mail | index | archive | help
scuba@centroin.com.br wrote:
>     I need some help to fine tune a mail server.
>     The machine is running (Fbsd 5.4) for a long time, but I'm seeing
> too many connections droped in sendmail log, and some users complains
> about timeout connecting to smtp port.
> 
>     From the handbook I increased the kern.ipc.somaxconn to 1024 as
> sugested. Using "ps axH" shows about 350 processes.
> 
>     With "netstat -s -p tcp" worrys me the value of "listen queue
> overflows", is it normal?

No, it implies that the system couldn't process the incoming connections rapidly
enough to drain the queue of requests, which is why your users are seeing timeouts.

Basicly, your mail server has a certain maximum throughput before it bottlenecks
and extra sendmail processes get stuck because they can't get enough resources,
and the thing falls over.  Running out of RAM and starting to swap would be a
sure sign of this happening, check your vmstat output...

What kind of mail volume are we talking about here?  You should be able to do on
the order of 1-10 messages a second on reasonable hardware (aka ~1 million/day).

There's a book called "Sendmail Performance Tuning", hmm, by Nick Chistianson or
something like that, which you should try to get ahold of.  You could start by
limiting the maximum number of child processes that sendmail will spawn to a
more reasonable number which will fit into the available RAM and leave room for
any other anti-spam or virus-scanning things, too.

-- 
-Chuck



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?442AB583.7090700>