Author: Troy Engel Date: To: exim-users Subject: Re: [exim] random user filters being skipped - possible causes?
Anyone have *any* ideas?
thx,
-te
Troy Engel wrote: > I have a mail server with exim 4.42 running on a rather ancient Redhat
> 7.3 system, with SpamAssassin 3.01 and DSPAM 2.10.6. Randomly (as far as
> I can tell) a user filter is getting skipped when returning from an
> external call to SpamAssassin. Not always, but sometimes.
>
> - the user's forward file is 100% correct
> - sending the missed email through the filter my hand reports that it
> will be filtered correctly (exim -d-all+filter -bf <filter> < email)
> - the same user is working sometimes, not working others
>
> Here's a log entry of a SPAM being tagged (subject) and filtered for
> this user:
>
> 2004-11-08 14:51:11 1CRILo-0001Gw-Cj <= queasiest@??? U=mail
> P=spam-scanned S=4251 id=110001c4c5e4$435ba43c$48749e1b@???
> T="[*SPAM*] Re: Cum Hungry Teen exclusive inretnal cumosht videos"
> 2004-11-08 14:51:11 1CRILo-0001Gw-Cj => /home/b/bXXXXX/Maildir/.Junk/
> <bXXXXX@???> R=userforward T=forward_delivery S=4366
> 2004-11-08 14:51:11 1CRILo-0001Gw-Cj Completed
>
> then later, a failure:
>
> 2004-11-08 15:00:55 1CRIVC-0001Md-8q <= <> U=mail P=spam-scanned S=12932
> id=200411082259.OAB13500@??? T="[*SPAM*] Returned
> mail: Too many hops 27 (25 max): from <<bXXXXX@???>> via
> portal.alchemy.co"
> 2004-11-08 15:00:55 1CRIVC-0001Md-8q => bXXXXX <bXXXXX@???>
> R=localuser T=local_delivery S=13027
> 2004-11-08 15:00:55 1CRIVC-0001Md-8q Completed
>
> What I can't seem to figure out is why the random skipping of the user's
> forward file filtering in the second instance; the filter *does* catch
> this second email if I run it by hand with -bf, so it's as if the
> forward file is not even consulted.
>
> Anyone have any ideas where to look next? I'm wondering if there are too
> many mails in the queue Exim misbehaves, or if the volume of mail is too
> high? It's so random, I would think a hardware problem under normal
> circumstances.
>
> thanks,
> -te
>
