SpamAssassin and Virus delays

[Matt Kettler]

Some general speed tuning recommendations for SA:

Certainly one of the most time consuming parts of most SpamAssassin checks
is the DNS blacklist lookups. If you don't have very fast local DNS,
disable as many of the blacklists as you can while still getting an
acceptable level of spam catching. Do this by setting their score to 0 in
your /etc/mail/spamassassin/local.cf or MailScanner's
spam.assassin.user.prefs file.

I have reasonably fast DNS but I still turned a couple off, there are
several others, but this might be a starting point. Your criteria for
selecting which ones to disable may vary. I picked all the ones with scores
<0.5, and the multihop/unconfirmed lists that I think aren't really suited
for general-purpose use.

score RCVD_IN_OSIRUSOFT_COM          0
score X_OSIRU_DUL_FH                 0
score X_OSIRU_SPAMWARE_SITE          0
score RCVD_IN_MULTIHOP_DSBL          0
score RCVD_IN_UNCONFIRMED_DSBL       0


also I'd recommend shortening some timeouts:

check_mx_attempts 2
check_mx_delay  3
rbl_timeout 10

and I forced razor1 off and only use razor 2. You might consider disabling
both since razor servers seems to choke up sometimes. (this can be
shortened using the razor_timeout setting)
score RAZOR_CHECK                    0



At 02:23 PM 10/16/2002 -0400, Thomas DuVally wrote:
>Hi all,
>
>I've recently put in MailScanner V3.22-14.  We've been humming along
>until we see peak times.  While delivery is fast, scanning is slow.
>Then we start seeing large delays (30+min).
>
>We were receiving over 7500 messages per hour, and were able to scan
>only up to 5200 in that same hour.
>
>To copy we turned off SpamAssassin and scanning of text messages.  This
>helped out a lot!
>
>I want to be able to turn on at least SpamAssassin again, but are there
>any tweaks I can use to help speak things up?
>
>Oh, one other thing, I'm using Symantec Carrier Scanner on Unix to scan,
>which I had to hack support for.  It can scan about 3 file a second
>(theoretically 10800 an hour)
>
>Any help?
>
>Will v4 be able help?
>--
>Thomas DuVally <thomas_duvally at brown.edu>
>Brown University