Date: Tue, 25 Nov 2003 19:18:42 -0500 From: Bill Vermillion <bv@wjv.com> To: freebsd-isp@freebsd.org Subject: Re: start multiple sendmail daemons from rc.conf? Message-ID: <20031126001842.GA8884@wjv.com> In-Reply-To: <20031126070221.R17841@hewey.af.speednet.com.au> References: <3FC395B3.C59E4AC3@tcworks.net> <20031125175601.GA4988@wjv.com> <67773D2B-1F75-11D8-B94F-003065ABFD92@mac.com> <20031126055846.X17841@hewey.af.speednet.com.au> <C09FA907-1F84-11D8-BEAA-000393681B06@lafn.org> <20031125204822.GD6772@wjv.com> <20031126070221.R17841@hewey.af.speednet.com.au>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Nov 26, 2003 at 07:23 , while impersonating an expert on the internet, Andy Farkas sent this to stdout: > Bill Vermillion wrote: > > > >So even though sendmail says it is rejecting connections > > > >(twice!), it then goes and accepts two more! Grrr. [lots deleted - wjv] > > Until he gets more memory he can add swap space - man 8 swapon - > > but to blame sendmail for an under powered or mis-configured system > > is wrong. > I know exactly what happened to my system. I was there watching > it self destruct. But the beauty of FreeBSD is its stability; > the box is still running and hasn't even been rebooted. > Yep, it needs more RAM and more swap. But that aint gonna happen :) Why? Read the vnconfig and swapon man pages and you can see you can make a special file and use that for swap. That means no reconfiguring of the devices. Of course the presumes you have enough space on the HD to do so. I personally think it's better to have enough swap to keep problems from occuring. A friend is running a web server on OS/X and it builds swap files dynamically. When it runs out of swap it creates another 80MB swap file. I just checked and there are 12 of them right now. Before he started moving some load to another machine I saw as high as 27 80MB files in use as swap. Not as good as dedicated HD space but good enough to keep you going? I'm just curious as to why you don't want to add swap - as it will keep at least one problem from happening. > Some background on my setup: sendmail, procmail, bogofilter, and > amavis/uvscan. sendmail receives a message, sends it to amavis > which spawns a uvscan, then procmail bogofilters it. Works well, > but is rather resource intensive (and as you've noticed, this > box is under-resourced). Boy do I know that feeling. After a nearby lightning hit weaked my mobo causing it to finally fail a month later I was on a system that if I got a load of mail things would creep along. Now I have a much faster system. > If I disconnect from the 'Net for a few hours, mail gets queued > upstream, and when I reconnect, vast quantities of email come in > all at once spawning hundreds of processes. (the poo hits the > fan at around 300 procs) Oh yes. I would use fetchmail to go to the server, but I'm on a DSL so only when the DSL wedged did I see that problem. I can really relate to that. > My sendmail.cf has this in it: > O QueueLA=3 > O RefuseLA=4 > O DelayLA=2 > O MaxDaemonChildren=16 > O ConnectionRateThrottle=3 > but the situation can still occur :) That's pretty low. You are saying even with those you run into problems. Wow. I was thinking you might be running with the default 8 and 12 and could lower those. Mabye we can take up a collection :-) -- Bill Vermillion - bv @ wjv . com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031126001842.GA8884>