Dan,
What happens if you manually run a mail through with spamc? We saw this
sort of problem when spamd was taking too long and spamc timed out and
returned the mail unprocessed. We resolved it by using the -t flag
with spamc.
-Original Message-
From: Dan O'Brien [mailto:[EMAIL PROTECTED
We are using archmbox:
http://adc-archmbox.sourceforge.net
> -Original Message-
> From: Kris Deugau [mailto:[EMAIL PROTECTED]
> Sent: Friday, January 02, 2004 5:10 PM
> To: [EMAIL PROTECTED]
> Subject: Re: [SAtalk] A cron job to delete old spam?
>
>
> Liu Shuai wrote:
> > I want to set
I am unfamiliar with QMAIL, we use Postfix. Our Postfix content filter,
the one that sends mail through spamd, examines the returned mail for
'x-spam: yes' and delivers it to a local mail account. If the flag is
not found, it delivers it to the original recipient. Perhaps somethng
like this is poss
If it times out, does spamd decline to markup the header? I would think
that it would mark up with the score that it has, without the timed-out
tests.
-Original Message-
From: Jack Gostl [mailto:[EMAIL PROTECTED]
Sent: Wednesday, October 15, 2003 2:17 PM
To: Rich Puhek
Cc: Martin
I have been seeing this also, with 2.55 and 2.6. I had been working with
the theory that it had to do with Postfix's content filter system. but
since you are using procmail, perhaps that is not the case. My guess was
that spamd was taking too long, and the message was getting forwarded
anyway. The
Has anyone else seen anything like this, where spamd appears to be
truncating a message? I have a message that appears to go into spamd
with size 67762 and gets returned to postfix with size 4755. I received
the message with most of it missing. As far as I know from user reports,
it is only happeni
AM
To: Martin, Jeffrey
Cc: [EMAIL PROTECTED]
Subject: Re: [SAtalk] why is this learned as ham?
On Thu, Oct 02, 2003 at 09:35:44AM -0400, Martin, Jeffrey wrote:
> But the example scored 0.4, so it still shouldn't be autolearned,
> right?
run with -D and read the Mail::SpamAssassin::Conf d
Oops, I was on the wrong system. The real settings on this one are;
10_misc.cf:# learning system automatically, to train the Bayesian
scanner.
10_misc.cf:bayes_auto_learn_threshold_nonspam 0.1
10_misc.cf:bayes_auto_learn_threshold_spam 12.0
10_misc.cf:# Set this to 0 to turn off aut
I just upgraded a test system to 2.6, and it is doing something strange.
The header indicates that messages are being autolearned that shouldn't
be. The auto_learn settings in /usr/share/spamassassin are:
10_misc.cf:auto_learn_threshold_nonspam -2
10_misc.cf:auto_learn_threshold_spam15
This i
Even if it were whitelisted, there should still be the SA header tags
indicating scores, etc. It doesn't *look* like it could be over the max
message size. I've seen this myself quite a few times, where a message
gets to a user without being processed by SA, although it does pass
through the SA sys
We had this problem, I think it was a memory leak in Perl. We upgraded to RH 8.0 and
all was fine.
I ran vmstat every 30 seconds, and graphed the numbers over time. I saw a steadily
decreasing free memory number until it was all gone and the system became totally
unresponsive just like describ
11 matches
Mail list logo