OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
Subject: performance issues
From: Adam Levin (alevinaudible.com)
Date: Wed May 17 2000 - 09:05:16 CDT

So I come in this morning and find out that almost no mail is coming
through, because apparently the powers-what-be here have sent out a mass
mailing (solicited -- they signed up for it) of 45,000 messages through
our regular office mail server.

I see about 40,000 messages sitting in the incoming directory, and they
are *slowly* getting processed. My master.cf looks like this:

smtp inet n - y - - smtpd
pickup fifo n n y 60 1 pickup
cleanup unix - - y - 0 cleanup
qmgr fifo n - y 300 1 qmgr
rewrite unix - - y - - trivial-rewrite
bounce unix - - y - 0 bounce
defer unix - - y - 0 bounce
smtp unix - - y - - smtp
showq unix n - y - - showq
error unix - - y - - error
local unix - n n - - local
cyrus unix - n n - - pipe
    flags=R user=cyrus argv=/cyrus/bin/deliver -e -q -m ${extension}
${user}
uucp unix - n n - - pipe
    flags=F user=uucp argv=uux -r -n -z -a$sender - $nexthop!rmail
($recipient)
ifmail unix - n n - - pipe
    flags=F user=ftn argv=/usr/lib/ifmail/ifmail -r $nexthop ($recipient)
bsmtp unix - n n - - pipe
    flags=F user=foo argv=/usr/local/sbin/bsmtp -f $sender $nexthop
$recipient

So I'm looking over the postfix documentation, and I realize that our mail
server is a pretty big hub, so I should probably change those (I believe
they're the defaults from postfix-19990906-pl9, which is what I'm
running).

So, I upped the smtp and smtpd to 1000, making sure that my
default concurrency limit is set to 10 (local is still 2). I also upped
the qmgr to 200, and changed the rest of the processes to the default
instead of '0' (Is that bad?).

Almost immediately, 12,000 messages were cleared out (in about 6 minutes
-- 33 messages per second). This is a Sun Enterprise 250 with 2x400MHz
CPUs and 2GB of memory. It's also our primary DNS server, but it's never
highly loaded as far as CPU or run queue. We've got a bunch of separate
disks, and postfix has its own spindle for the queues and a separate
spindle for the mailboxes.

However, after we got down to about 15,000 messages remaining, it started
slowing down again.

The questions are: any thoughts as to why it slowed down again? Anything
bad about having 200 qmgr, 50 defer, 50 bounce and 50 cleanups, assuming
the system isn't going into anaphylactic shock? Anything else I should be
considering on a server that once a week will be hit with an outgoing
mailing in the tens and possibly hundreds of thousands, as well as having
to handle regular office mail of a few thousand per day? We have about
100 users who use MSOutlook to POP mail off.

Thanks,
-Adam

Adam Levin, Senior Unix Systems Administrator | http://www.audible.com/
Audible, Inc.
Wayne, NJ, 07470 There are two kinds of people in this world,
973-890-4070 x297 and I'm one of them.