On 02/06/10 14:45, a.smith@??? wrote:
> So looking again, its actually looks like I just have something very
> broken. Here´s a log of a spam mail with no spam score:
>
> 2010-06-02 10:10:08 [70746] 1OJjxf-000IP4-Cm<= info101@???
> H=(1x1.cz) [87.236.199.164]:37410 I=[89.21.28.37]:25 P=esmtp S=3670
> T="NOTIFICATION FOR RELEASE OF $10.5M" from<info101@???>
> for a.smith@???
> 2010-06-02 10:10:08 [70746] 1OJjxf-000IP4-Cm no immediate delivery:
> load average 6.13
> 2010-06-02 10:20:13 [71971] 1OJjxf-000IP4-Cm => a.smith
> <a.smith@???> F=<info101@???>
> P=<info101@???> R=spamcheck T=spamcheck_pipe S=3741
> QT=10m6s DT=2s
> 2010-06-02 10:20:13 [71971] 1OJjxf-000IP4-Cm Completed QT=10m6s
>
> 2010-06-02 10:20:13 [71973] 1OJk7Q-000Iir-9t<= info101@???
> U=mailnull P=spam-scannedukgsa S=3927 T="NOTIFICATION FOR RELEASE OF
> $10.5M" from<info101@???> for a.smith@???
> 2010-06-02 10:20:13 [71973] 1OJk7Q-000Iir-9t no immediate delivery:
> load average 6.60
> 2010-06-02 10:31:55 [73581] 1OJk7Q-000Iir-9t => a.smith
> <a.smith@???> F=<info101@???>
> P=<info101@???> R=virtual_localuser T=virtual_localdelivery
> S=4044 QT=11m43s DT=0s
> 2010-06-02 10:31:55 [73581] 1OJk7Q-000Iir-9t Completed QT=11m43s
>
>
> Its not listed anywhere in the spamassassin logs and has no spam score
> etc. However you can clearly see its being recieved via SMTP, passed
> to the spamcheck_pipe and then mailnull feeds it back in with the
> "P=spam-scannedukgsa" that was confusing me. So Im not sure how its
> being passed out via the spamcheck_pipe and back in again without
> appearing in the spamassassin logs.
>
Presumably spamd is not responding, or the response does not get back to
spamc in time, and Exim goes on to deliver the message. If spamc cannot
communicate with spamd it returns the message unaltered and unchecked.
The load is probably the problem, it could be that all the spamd
processes are occupied. It's worse if you are using Bayes, even more so
if you have autolearn enabled. This relies on stored information in
database files, and that can lead to locking problems when there are
several spamd working simultaneously. If they are blocked waiting for a
lock they may not run quickly enough to get the results back to spamc.
--
Nigel Wade, System Administrator, Space Plasma Physics Group,
University of Leicester, Leicester, LE1 7RH, UK
E-mail : nmw@???
Phone : +44 (0)116 2523548, Fax : +44 (0)116 2523555
