Re: Sudden instability

[Dan Nelson]

In the last episode (Jul 08), SG said:
> Dan Nelson wrote:
> > This could be due to spamassassin taking long enough to process the
> > message that sendmail timed out the milter connection.  Note that
> > the incorrect samples have no spamassassin headers at all.  If you
> > take a look at the maillog, check for a "sm-mta[xxx]: Milter: (xxx)
> > timeout" error followed a few seconds later by a "spamd[xxx]: clean
> > message in xxx seconds".  Raising the milter timeout values in your
> > sendmail config file or turning off DNS checks in spamassassin will
> > help here.
> 
> Thanks, I saw several 'clean message' messages in maillog, but these
> did not seem to follow timeouts (I didn't see any timeouts in the
> log), my values for the milter are:
> 
> F=, T=C:15m;S:4m;R:4m;E:10m
> 
> Which timeouts should I increase, by how much?

4 minutes is more than enough :) 
 
> By DNS in spamassassin do you mean the dns_available switch?

Yeah.  Your timeouts are long enough that what I described will
basically never happen.

If you don't see any milter timeout messages, then something is making
a decision to not run spamassasin on that mail.  The 2 most likely
explanations: those messages came in via a trusted sender (listed by
the -i flag to spamass-milter) and checking was skipped, or they were
over 250k and spamc short-circuited them right back without running
them through spamassassin.  Neither of those seem to be the case for
you though.  Try adding "-d misc" to your spamass-milter commandline
and see whether you see an "SA Score" maillog entry for the next
message that doesn't get marked.

-- 
        Dan Nelson
        address@hidden