SA Ignoring Custom Scores

[XN-Matt]

We've got about 25 custom scores set via the GUI with pmg-smtp-filter reloaded via the GUI and also on the command line.

Is there anyway to debug why SA is ignoring the custom scores?

 

[Stoiko Ivanov]

for each mail the system should log which SA-rules are matching (pmg-smtp-filter writes those logs).

if your custom rules are not among the hits my guess would be that they are not matching the mail...

if you have a mail, which should match - you can use /usr/bin/spamassassin and input that mail - to get some debug output:
https://cwiki.apache.org/confluence/display/SPAMASSASSIN/HowToDebug
https://spamassassin.apache.org/full/3.4.x/doc/spamassassin-run.html

I hope this helps!

 

[XN-Matt]

They are matching the mail as they appear in the mail headers but the score is not the custom one set.

Have run the message through to check but no errors were seen.

Are all custom scores matchable or will there be some just ignored?

 

[Stoiko Ivanov]

hmm one potential explanation is that the custom-score file '/etc/mail/spamassassin/pmg-scores.cf' does not get read/included ...
This could happen if you've overridden the spamassassin configuration for the templateing system:
https://pmg.proxmox.com/pmg-docs/pmg-admin-guide.html#pmgconfig_template_engine

(sometimes this happens inadvertently by copying all files from /var/lib/pmg/templates...)

does

spamassassin -D all --lint 2>&1 |grep pmg-scores

produce any output?

 

[XN-Matt]

I think this was actually due to a cluster sync issue. All the non-master nodes were stuck in "syncing". No matter what I did I could not get that to change so rebooted them all and now things are fine.

So the issue appears to be that the changes on one node (master), were not filtering through to the others and giving the impression that they were being ignored.

Might be a good idea to write up more about the cluster page and the statuses and what the expected output should be (i.e active) and what syncing means and how long that should be shown (i.e not days/hours continually)

 

[Stoiko Ivanov]

think this was actually due to a cluster sync issue. All the non-master nodes were stuck in "syncing". No matter what I did I could not get that to change so rebooted them all and now things are fine.

that would explain the issue - glad that the cause was found.

Might be a good idea to write up more about the cluster page and the statuses and what the expected output should be (i.e active) and what syncing means and how long that should be shown (i.e not days/hours continually)

The cluster documentation could indeed profit from an expansion - would you please open an enhancement request over at https://bugzilla.proxmox.com?
(of course should you have time and wish to do so - a patch would be even better )