Re: [Exim] Mass Mailings and Sendmail(exim) process's...

Top Page
Delete this message
Reply to this message
Author: Nico Erfurth
Date:  
To: Shaun.Murphy
CC: exim-users
Subject: Re: [Exim] Mass Mailings and Sendmail(exim) process's...
Shaun.Murphy@??? wrote:
> This is a multi-part message in MIME format. -- [ Picked text/plain
> from multipart/alternative ]


include "rant_about_linewrapping.h"

EximVersion?
OS?

> We have a client here that sends out thousands and thousands of
> emails a day. Sometimes when i'm in her server i can see 800,000+
> emails in queue. She was using qmail and i just bumped up the
> concurrentprocesses so that 500 remote process's would run. But not
> we have moved her to a bigger server and this server now runs exim.
> Her scripts that do the mailings are perl based and they call
> /usr/sbin/sendmail. I've notifed that /usr/sbin/sendmail is a
> symbolic link to /usr/sbin/exim. The other day i was logged into her
> server and she had 500+ /usr/sbin/sendmail process's in the tree.
> Her load adverage was running up around 400+ and the server was
> crawling. From what i can tell, to me it looks like it's trying to
> send the messages right when their sent to /usr/sbin/sendmail rather
> than putting them in queue and waiting for exim -bd -q1m to run them.
> Anybody who has any information about what i can do to speed up the
> mailserver and help reduce load on it would be much appriciated.
> This server is a p4 1.8 with 1gb DDR and 3 38gb SCSI3 Drives (raid
> 5). Her old server was a p3 1ghz with 1gb ram running on a 80gb ide
> drive (she killed a drive about once every 2 months). Whats sucks
> about all this is that the old server running qmail was sending out
> mail faster than the new one is. My AIM is OCTekShaun and my email
> is Shaun.Murphy@???. Thanks.



Have you tested the usual things?
setting split_spool_directory?
setting queue_smtp_domains = * and running exim -q1m or something
similar?
Installing a local (caching) dnsserver?
playing around with the max_queue_runners-setting
what about remote_max_parallel?

On some OSes it's useful to place the retry databases in a ram-fs.

Do you use any heavy lookups inside of your config?

Sorry, but you don't give enough informations to analyze the problem.

ciao