On 2026-05-20 at 19:25:15 UTC-0400 (Wed, 20 May 2026 19:25:15 -0400)
FalconChristopher <[email protected]>
is rumored to have said:
What must you configure as to tell one instance of SpamAssassin to
only do minimal checking of the message.
Create a local.cf that turns off (score 0 )the rules you don't want to
check. You could also disable plugins that you do not want to run in the
.pre files, which effectively disables dependent rules. You'll still get
a score, but it is likely to be worthless. You should also make sure to
not add any SA headers, so as not to confuse the users.
Then moving that message to another version of SpamAssassin where it
gives the message a score ?
SpamAssassin has NOTHING to do with the routing of mail. There is
nothing SpamAssassin does other than giving a message a score, which may
indicate 'spam' or 'ham' depending on the threshold score, which
defaults to 5.
If you want to scan a message again after its first scan, you can either
pass it to a different MTA instance with its own SpamAssassin rig or you
can set up 2 instances of whatever your glue layer is (e.g. milter) and
have them use different SA configs. The transport of messages to
multiple filters is intrinsically a task for the MTA.
On 5/19/2026 12:17 AM, Dave Funk wrote:
Christopher,
running multiple instances of spamd/spamassassin is the easy part, if
they're on separate physical or virtual machines or on the same
machine in containers (Docker, Kuberneties, etc). If they're directly
on the same host then you need to be careful about their
configurations, make sure each one has its own config file, pid file,
rules files, working "HOME directory", Bayes db instance, etc.
In general spamassassin does not 'verify' messages, it just looks at
them and assigns a numerical score based on multiple "grading"
criteria. It is true that one of the criteria is if a message has a
DKIM/DMARK signature and the validity thereof but that is usually a
small part of the score.
Think of it like multiple teachers grading a student's term-paper.
The tricky part of scanning a message by more than one spamd is
making sure that your 'glue' architecture is giving each spamd
instance a pristine unaltered copy of the message, not one that has
been 'marked up' by preceding spamd instances.
I frequently run more than one spamd instance to test out new
versions of SA and test-drive trial rule sets.
I handle this by making the earlier instance(s) run in "silent" mode
(IE not add any classification or score headers) and only have the
last instance actually add the usual scoring headers and if needed
trigger a SMTP reject on high scoring spam.
The only way that I know the "opinions" of the earlier SA instances
is to look at their log entries, they don't show up in the messages.
Hope that helps.
Dave
On Mon, 18 May 2026, FalconChristopher wrote:
I want to get some understanding which I can run multiple `spamd` as
long as as the daemon is running by using the -host or -port flags.
Therefore if I'm running SpamAssassin on a remote server, then
running SpamAssassin on another remote server. I assume the emails
can pass though two layers of verification ?
Christopher
--
Bill Cole
[email protected] or [email protected]
(AKA @[email protected] and many *@billmail.scconsult.com
addresses)
Please keep discussion mailing list replies *on-list*
Not Currently Available For Hire
