Debugging and resolving high loadavg when running MailScanner
Richard Curtis
richard at LEGENDJSM.CO.UK
Wed Feb 16 14:39:10 GMT 2005
[ The following text is in the "windows-1250" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
I am not claiming to be a master admin or anything - I am just trying to
solve what I believe to be a problem using what I know to the best of my
abilities.
The reason that I percieve there to be a problem is that when the load
goes high, web page serving and even ssh logins can take a long period of
time. That is what led me to beleive there is an issue.
I will go and look into the tools such as SAR that you mentioned and see
what I can establish from that.
Thanks for the advice... Im still learning and just trying to work out
issues as I find them...
Richard
Excuse me, but why are you "hung up" on loadfactor as a
performance meter?
It is just the average run queue size (averaged over 1, 5 and 15
minutes), and as such is meant
to vary.
You say you have a 2 CPU system, in which case (with a "SMP
capable" kernel) you'd be fine
with anything up to 2 in a "sustained load" (meaning more than in
transient peaks).
Add to this that processes in "non-interruptible wait" (usually for
IO resources) add an unconditional
1 (one) each, it is not at all surprising that the load goes a bit
beyond your "CPU count" when you
have a larger batch of messages coming in.
Unless you perceive true slowdowns, you don't really have a problem
with that.
Now, the swap bit _could_ be an indicator of true problems... You
say you use it for other things than
as a mailhub... So you might be in a slightly memory starved state,
easily determined via vmstat or sar,
if you have a fair amount of ongoing swapping.
But using swap on a system that runs for an extended period of time
is usually not an indicator of
problems in and of itself, it is the actual moving to/from swap of
frequently used memory slices that'll
be a killer.
If you can determine that you have this situation, you have only
two venues: reduce the RAM usage
(by reducing what the machine does), or add more RAM.
Nothing you've said so far indicate that you do have a problem...
Does the machine occasionally
slow down in a perceptible and measurable way? Does MailScanner
report huge batches?
I'd like to recommend the excellent "swordfish book" from o'rielly
(Systems performance tuning).
Even though it is geared toward Solaris primarily and linux (2.4)
secondarily, it has a lot of general
good tips (and tool usage tips that are rather generic for most
unices).
-- Glenn
-----Original Message-----
From: MailScanner mailing list
[mailto:MAILSCANNER at JISCMAIL.AC.UK] On Behalf Of Richard
Curtis
Sent: den 16 februari 2005 13:04
To: MAILSCANNER at JISCMAIL.AC.UK
Subject: Re: Debugging and resolving high loadavg when
running MailScanner
I am not using either... running
spamassassin -D --lint -p
/etc/MailScanner/spam.assassin.prefs.conf says there are two
problems - but the two problems appear to be that razor and
pyzor are not installed..
The load is sitting at around 0.5 at the moment - but when I
was looking about 5 minutes ago it shot to 5.0 when a large
batch of email came in.
I am wondering if I need to do something with the number of
children. I currently have it set to 10 - based on the
reccomendation of 5 per cpu.
The other thing I think might be related to my problems are
the fact that the system is using a fair amount of swap -
would that contribute to the load ?
Rich
Are you using razor or pyzor in that setup and does
spamassassin -D --lint -p
/etc/MailScanner/spam.assassin.prefs.conf
show any problems with either of these?
If so,
razor-admin -discover
pyzor discover
might be useful (I've put them in a cron job here).
Cheers,
Phil
----
Phil Randal
Network Engineer
Herefordshire Council
Hereford, UK
________________________________________________________________________________
From: MailScanner mailing list
[mailto:MAILSCANNER at JISCMAIL.AC.UK] On Behalf Of
Richard Curtis
Sent: 16 February 2005 11:33
To: MAILSCANNER at JISCMAIL.AC.UK
Subject: Debugging and resolving high loadavg
when running MailScanner
Hi All,
I have been running MailScanner (Version
4.35.11) for some time, but over the last 2 -3
weeks, my loadavg has gone from ~0.10-0.20 to as
high as 4. This may not seem high to some, but
as I have other services on this box I want to
try to reduce the load.
The box runs httpd, sendmail/MailScanner with
clamav, radius and pop. There are multiple
hosted domains on the server.
The box is a dual AMD Athlon MP 1200 with 512MB
RAM.
All mail is scanned through MailScanner for
virus's, but only a small number of the domains
(low traffic ones) have SpamAssassin enabled in
MailScanners configs.
I have to admit that I am a bit confused by some
of the MailScanner config options - for example,
the effects of "Deliver In Background = yes" and
"Delivery Method = batch".
By setting the delivery method to queue, what
kind of impact on delivery time can I expect ?
Will this have a significant effect of lowering
the load ?
One other question, I note from looking through
the logs, I see entries like:
Feb 16 11:25:32 lancelot MailScanner[7544]: New
Batch: Scanning 1 messages, 1720 bytes
Feb 16 11:25:32 lancelot MailScanner[7544]: MCP
Checks completed at 1720 bytes per second
Feb 16 11:25:32 lancelot MailScanner[7544]: Spam
Checks completed at 1720 bytes per second
Now what confuses me here is that I have MCP
checks and Spam checks set to off for this mail
domain - so why are the checks happening ?
Any suggestions on performance tuning of
MailScanner, or possible sendmail config tweaks
would be greatly appreciated.
Regards
Richard
--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.8 -
Release Date: 14/02/05
------------------------ MailScanner list
------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk
with the words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ
(http://www.mailscanner.biz/maq/)
and the archives
(http://www.jiscmail.ac.uk/lists/mailscanner.html).
Support MailScanner development - buy the book
off the website!
------------------------ MailScanner list
------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the
words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ
(http://www.mailscanner.biz/maq/)
and the archives
(http://www.jiscmail.ac.uk/lists/mailscanner.html).
Support MailScanner development - buy the book off the
website!
--
No virus found in this incoming message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.8 - Release
Date: 14/02/05
--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.8 - Release Date:
14/02/05
------------------------ MailScanner list
------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ
(http://www.mailscanner.biz/maq/)
and the archives
(http://www.jiscmail.ac.uk/lists/mailscanner.html).
Support MailScanner development - buy the book off the
website!
------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ (http://www.mailscanner.biz/maq/)
and the archives
(http://www.jiscmail.ac.uk/lists/mailscanner.html).
Support MailScanner development - buy the book off the website!
--
No virus found in this incoming message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.8 - Release Date: 14/02/05
--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.8 - Release Date: 14/02/05
------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ (http://www.mailscanner.biz/maq/)
and the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).
Support MailScanner development - buy the book off the website!
More information about the MailScanner
mailing list