Aug 10, 2017; 10:15pm Scottonline Scott Re: Bayes auto-learn - not
happening, tentative success
Well, here's a development...
About the only difference in my old, functioning box and this new "clean"
install was the location of the bayes files.
Old box:
/var/spool/amavisd/.spamassassin/
N
Well, here's a development...
About the only difference in my old, functioning box and this new "clean"
install was the location of the bayes files.
Old box:
/var/spool/amavisd/.spamassassin/
New box:
/etc/mail/bayes
The other details that caught my attention were that on the old box, the
ONL
Trying to check for any locking issues I ran sa-update in debug moed
su amavis -c 'sa-learn -D --spam --showdots --mbox /home/mail/onespam'
Appears to be creating and dropping lock files. Nothing left over after
running..
Aug 10 16:48:39.109 [7524] dbg: bayes: expiry starting
Aug 10 16:48:39.11
Scouring the differences between this and my old server I see this:
Old server:
-rw--- 1 amavis amavis 83472 Aug 10 15:51 bayes_journal
-rw--- 1 amavis amavis 1986 Aug 10 15:51 bayes.mutex
-rw--- 1 amavis amavis 328491008 Aug 10 15:51 bayes_seen
-rw--- 1 amavis ama
> Imho You need 100 ham and 100 spam to auto learning working. Do manual
learning
See earlier post today. I've got it loaded up, right?:
[root@tn2 mail]# su amavis -c 'sa-learn --dump magic'
0.000 0 3 0 non-token data: bayes db version
0.000 0349
Imho You need 100 ham and 100 spam to auto learning working. Do manual
learning
08.08.2017 8:20 PM "Scott Techlist" napisał(a):
> Centos7
> Postfix 3.2.2
> Amavisd-new 2.11.0
> Spamassassin 3.4.0
> Site-wide configuration
>
> This is a new box and I've configured some conservative values for
> a
OK, so I don't think auto-learn works on spam. What about HAM?
I've raised the floor to auto-learn HAM to 1. Before anyone gives me any
grief, it's just for testing. I'll rebuild the bayes db from a corpus when
I get it working.
So SPAM takes the 3-way patch, 3 from the header, 3 from the body
OK, so I don't think auto-learn works on spam. What about HAM?
I've raised the floor to auto-learn HAM to 1. Before anyone gives me any
grief, it's just for testing. I'll rebuild the bayes db from a corpus when
I get it working.
So SPAM takes the 3-way patch, 3 from the header, 3 from the body
>surely, it makes no sense blow up the database with already 100%
>classified samples - you even don't do that uncnditional with a
>hand-trained database (at least not forever, at the begin it makes sense
>to get additional tokens)
I think you misunderstood my question. I meant that as I look at
>surely, it makes no sense blow up the database with already 100%
>classified samples - you even don't do that uncnditional with a
>hand-trained database (at least not forever, at the begin it makes sense
>to get additional tokens)
I think you misunderstood my question. I meant that as I look at
If any particular message has a
* 3.5 BAYES_99 BODY: Bayes spam probability is 99 to 100%
* [score: 1.]
Is it safe to assume that spam or one close to it has been learned and so it
would not be a candidate for auto-learn?
Maybe I'm not being patient enough.
--
View this mess
FYI, here's the verbose headers for that same one that flowed above:
X-Spam-Flag: YES
X-Spam-Score: 23.904
X-Spam-Level: ***
X-Spam-Status: Yes, score=23.904 tag=- tag2=5 kill=6.4
tests=[BAYES_999=0.2, BAYES_99=3.5, DCC_CHECK=3.2,
DIGEST_MULTIPLE=0.293, HEAD
Here is a debug log for one that just flowed. I don't see anything about why
auto-learn was unavailable. But it shows it's talking to the db anyway I
think.
Is there a way to set auto_learn_force to yes? The log format makes one
thing it's a global setting but all I can find it looks like a per
Here's a verbose log of amavis/spamassassin processing another high score
that just came through. I don't see a peep about auto-learn. But it was
unavailable too.
(posting via nabble, apologies if it wraps)
Aug 10 11:03:39 mail2 amavis[377]: (00377-01) LMTP :10024
/var/spool/amavisd/tmp/amavis
>why this?
>When you run from amavisd, you only need permission for amavis user, not
for
>anyone.
To be sure that is not the problem. I can tighten it up once working. I
understand thisis what one woudl normally use if they had a multi-user
enviroment. But it can't hurt the problem for testing,
On 08/10/2017 10:06 AM, techlist06 wrote:
Update: Still NOT working, but I'm giving it hell trying to figure out why :)
First a couple of answers to other's questions:
- John, others, not an ISP, high is relative I'm sure but the volume is much
higher than I can duplicate and review every flag
On 10.08.17 10:06, techlist06 wrote:
Update: Still NOT working, but I'm giving it hell trying to figure out why :)
- Matus: re:" autolearn=unavailable apparently due to not accessible bayes
database [due to permissions]". I hope you are right. That would make
sense to me. See below please
Felix Defrance skrev den 2017-08-10 17:09:
Ok but in many other cases, spamassassin not failed at this point and
validate the signing.. why it fail sometimes, and not all time ?
verify when it fails what size the failed mail have, if that its
truncated with amavisd, then extend mailtruncateing
Le 10/08/2017 à 16:33, RW a écrit :
> On Thu, 10 Aug 2017 16:12:11 +0200
> Felix Defrance wrote:
>
>
>> In the first lines on log, you could see opendkim results are success.
>>
>> Aug 9 10:25:42 vmail opendkim[21923]: 0D81A778B1D: DKIM verification
>> successful
>> Aug 9 10:25:43 vmail opendmarc
Update: Still NOT working, but I'm giving it hell trying to figure out why :)
First a couple of answers to other's questions:
- John, others, not an ISP, high is relative I'm sure but the volume is much
higher than I can duplicate and review every flagged message. Right now
running at about 10
On Thu, 10 Aug 2017 15:12:11 +0100, Felix Defrance
wrote:
In the first lines on log, you could see opendkim results are success.
Aug 9 10:25:42 vmail opendkim[21923]: 0D81A778B1D: DKIM verification
successful
Aug 9 10:25:43 vmail opendmarc[7879]: 0D81A778B1D: groupeastek.fr none
That why
On Thu, 10 Aug 2017 16:12:11 +0200
Felix Defrance wrote:
> In the first lines on log, you could see opendkim results are success.
>
> Aug 9 10:25:42 vmail opendkim[21923]: 0D81A778B1D: DKIM verification
> successful
> Aug 9 10:25:43 vmail opendmarc[7879]: 0D81A778B1D: groupeastek.fr
> none
>
Le 09/08/2017 à 18:53, David Jones a écrit :
> On 08/09/2017 10:19 AM, Felix Defrance wrote:
>> Do you have any idea why the body has been altered sometimes ? I
>> don't have any log about amavis alterate body message.
>>
>
> This happens when any server in the path modify some of the headers or
Le 10/08/2017 à 14:46, Benny Pedersen a écrit :
> RW skrev den 2017-08-10 02:06:
>
>> If amavis only passes part of the email to SA, it isn't going to pass
>> DKIM.
>
> i think there is more underlaying problems there, amavisd have its own
> dkim verify and signer, even if spamassassin does not s
RW skrev den 2017-08-10 02:06:
If amavis only passes part of the email to SA, it isn't going to pass
DKIM.
i think there is more underlaying problems there, amavisd have its own
dkim verify and signer, even if spamassassin does not see the fullmail,
amavisd can verify and sign it still, so
25 matches
Mail list logo