[AMaViS-user] amavisd-new not using SA bayes?

[David N. <dn...@ne...>]

We've been running amavisd-new/spamassassin/postfix for around a month. SA
is set to autolearn, and amavisd-new is set to use SA.

While the Bayes database is growing steadily, the amount of mail tagged as
spam has not. Users' mailboxes are only slightly less clogged than before.

Are we missing something to get amavisd-new to use SA Bayes learning?

The setup:

FreeBSD 4.9
postfix 2.0.13
amavisd-new-20030616-p8
spamassassin 2.63

excerpt from amavisd.conf:

$sa_local_tests_only = 0;   # (default: false)
$sa_auto_whitelist = 1;    # turn on AWL (default: false)
$sa_mail_body_size_limit = 64*1024;

excerpt from local.cf:

use_bayes 1
# This turns Bayesean Learning on. 0 turns it off.

bayes_path /var/amavis/.spamassassin/bayes
# Bayesean database location.

auto_learn 1
# Auto learning lets SpamAssassin classify spam as SPAM.

skip_rbl_checks 0
# By default SpamAssassin runs the Realtime Blackhole List checks.
# It's better to turn this option off.

use_razor2 1
# Tells SA that we want to use Razor version 2

use_dcc 1
# In case you want DCC.

use_pyzor 0
# Tells SA that we don't want to use Pyzor

dcc_add_header 1
# DCC header in case you want it.

## Optional Score Increases
score DCC_CHECK 4.000
score RAZOR2_CHECK 2.500
score BAYES_99 4.300
score BAYES_90 3.500
score BAYES_80 3.000

Thansk in advance for any clues to get Bayes working

dn