This is an answer from our local sendmail guru in regard to how he has
his setup. His lists are not as large as yours, but you might be able
to glean some information in regard to the setup he has for sendmail.
Specifically, the multiple daemons.
-------- Original Message --------
Subject: Re: [oclug] Why is this list so slow? The answer
Date: Thu, 27 Apr 2000 12:24:11 -0700
From: Christopher McCrory <chrismcc@netus.com>
Reply-To: oclug@oclug.org
Organization: Netus Internetworking
To: oclug@oclug.org
References:
<Pine.LNX.3.95.1000426223712.10291B-100000@scsoftware.sc-software.com>
Hello...
When majordomo list emails come in they go through a "speed bump".
Specifically, they are queued instead of immediately being processed.
The mail queue runs every 7 minutes. So after 7 minutes the message is
sent through the majordomo program. The majordomo program then
processes the message to make sure it is "OK" and other stuff. then it
is queued again. On the next queue run, the majordomo program then
sends it through a program called 'bulk_mailer'. The bulk_mailer
program sorts the recipients based on the domainname, and breaks up the
one message into multiple messages of about 20 receipents. The multiple
messages are immediately sent. FYI, there are currently 152 email
addresses on the oclug list, so bulk_mailer creates about 8 separate
outgoing messages. Some end up going out faster than others. The speed
at which you get your copy is dependent on where you end up in the
order. Since the list is ordered based on your domainname when you
signed up makes no difference. So... there is at least a 14 minute lag
time, plus some processing overhead. This process prevents mail storms
from happening since mail list messages can only be processed one at a
time. I've been through ONE mail storm loop. It wasn't pretty. Now I
do it this way.
Yesterday, one of our clients mail servers went down (one guess what
OS). They receive a lot of email. We, by default, secondary email for
any clients running their own mail server. So my mail queue had lots of
messages that were taking up lots of queue time. So instead of the
mailing list messages being processed in 7 minutes, they were being
processed _much_later. That is why the list was slow yesterday.
For those that want even more technical stuff: I also run four sendmail
daemons. one accepts incoming connections and forks off a child process
for each incoming connection. The second daemon runs the primary mail
queue every 7 minutes. There is a cronjob that looks at the primary
mail queue and moves any messages over one hour old into the secondary
queue. The third sendmail daemon processes the secondary queue every 65
minutes. The same cronjob also looks at the secondary mail queue and
moves anything over four hours old into the terserary (sp?) queue. The
forth sendmail daemon processes this queue every four hours. Very Cool
stuff!
[orator@/root]$ mailqhowmany
Mail Queue (3 requests)
Mail Queue (8 requests)
Mail Queue (37 requests)
John Heil wrote:
>
> Hi Chris...
>
> Might it be possible to increase the majordomo
> distribution frequency... its seems to take
> a couple of hours or more to see a post hit the list...
>
> Thnx much
>
> -----------------------------------------------------------------
> John Heil
> South Coast Software
> Custom systems software for UNIX and IBM MVS mainframes
> 1-714-774-6952
> johnhscs@sc-software.com
> http://www.sc-software.com
> -----------------------------------------------------------------
>
> -
> Our web site: http://www.oclug.org
> to unsubscribe: mailto:oclug-request@oclug.org?body=unsubscribe
> subscribe to postonly: mailto:oclug-postonly-request@oclug.org?body=subscribe
--
Christopher McCrory
Lead Bithead, Netus Inc.
chrismcc@netus.com
admin@netus.com
"Linux: Because rebooting is for adding new hardware"
-
Our web site: http://www.oclug.org
to unsubscribe: mailto:oclug-request@oclug.org?body=unsubscribe
subscribe to postonly:
mailto:oclug-postonly-request@oclug.org?body=subscribe
|
|
