coordinating 2 amavisd/spamassassin scans -- 1 PRE-queue, 1 POST-queue ?
PGNd
dev at pgnd.us
Sat Jun 13 23:04:04 CEST 2015
I run
postfix 3.0.1
amavisd-new-2.10.1
I've setup a postfix to use both pre- and post-queue filters.
I' configured amavisd to provide two spamassassin scans, one in each of two policybanks.
The 1st to do 'lighter' pre-queue SA (headers, no bayes) (and other) scans, the 2nd to do 'heavier' SA-only (body + bayes) post-queue work.
The two policy banks use different SA configs
$sa_configpath = '/var/lib/spamassassin';
$interface_policy{'10001'} = 'PREQ';
$policy_bank{'PREQ'} = {
protocol => 'SMTP',
originating => 0,
enable_dkim_verification => 1,
enable_dkim_signing => 0,
bypass_decode_parts => 0,
bypass_spam_checks_maps => [0],
bypass_banned_checks_maps => [0],
bypass_header_checks_maps => [0],
bypass_virus_checks_maps => [0],
...
sa_siteconfigpath => '/etc/spamassassin/local/prequeue
sa_spam_subject_tag => ...,
sa_tag_level_deflt => ...,
sa_tag2_level_deflt => ...,
...
}
$interface_policy{'10002'} = 'POSTQ';
$policy_bank{'POSTQUEUE'} = {
protocol => 'SMTP',
originating => 0,
enable_dkim_verification => 0,
enable_dkim_signing => 0,
bypass_decode_parts => 0,
bypass_spam_checks_maps => [0],
bypass_banned_checks_maps => [1],
bypass_header_checks_maps => [1],
bypass_virus_checks_maps => [1],
...
...
sa_siteconfigpath => '/etc/spamassassin/local/postqueue
sa_spam_subject_tag => ...,
sa_tag_level_deflt => ...,
sa_tag2_level_deflt => ...,
...
}
Question:
With those two policybanks, is
$sa_num_instances = 1;
@spam_scanners = (['SpamAssassin','Amavis::SpamControl::SpamAssassin'],);
still sufficient config?
Or must I increase/modify, e.g., eith either/or
$sa_num_instances = 2;
@spam_scanners = (['SpamAssassin1','Amavis::SpamControl::SpamAssassin'],
['SpamAssassin2','Amavis::SpamControl::SpamAssassin'],);
?
Question:
Also each SA scan -- prequeue & postqueue -- can add SA headers.
Should both do so, using different header names? If 2 sets of headers, is there any 'header cooperation' (making the 2nd scan aware of the 1st scans headers) that needs to be orchestrated between then 2 scans, for either functional or performance reasons?
More information about the amavis-users
mailing list