increase the logging levels to verbose or diagnostic - this will also show
positive checking results
Thomas
Von:Michael Thomas
An: ASSP development mailing list
Datum: 26.04.2012 07:58
Betreff:Re: [Assp-test] Antwort: Scored SPF
Thomas,
Okay. I am trying to track down
Thomas,
Okay. I am trying to track down an issue I have seen. Message gets
passed. Log file shows that there were no RBL checks. I have seen two or
three variations. In all cases, at least two or three messages from the
same IP before the issue message were blocked. Recipient of issue
message
>Does the extra debugging on rebuildspamdb give any more information
No, the population of the spamdb and HMM_DB's is a DB import - which is a
series of INSERT statements - each contains up to 2000 records (1000 is
default)
Thomas
Von:Colin
An: assp-test@lists.sourceforge.net
Dat
>&SenderBaseOK($fh,$this->{ip})
returns 1 if in scoring mode
Thomas
Von:Michael Thomas
An: ASSP development mailing list
Datum: 25.04.2012 20:10
Betreff:[Assp-test] Scored SenderBase
Thomas,
Scored Country Code (DoSenderBase) when failed, exits getheader() as
absolute f
I think the speed is limited by the DB engine not by the network speed.
Thomas
Von:Colin
An: assp-test@lists.sourceforge.net
Datum: 25.04.2012 20:10
Betreff:Re: [Assp-test] Antwort: Re: Antwort: Slow processing
rebuild output
2012-04-25 11:56:39 Start populating Hidden M
>SPFok($fh)
returns 1 if it only scores
Thomas
Von:Michael Thomas
An: ASSP development mailing list
Datum: 25.04.2012 20:04
Betreff:[Assp-test] Scored SPF
Thomas,
ASSP version 2.1.2(12116)
Scored SPF that fails in getheader() exits function as absolute fail,
but witho
I've set everything up on a private interface now and no change.
Network activity is minimal during populating of the HMM chains. MySQL
is using 5-10% CPU and 12.8% of the available memory and there's no
excessive load.
The status page has the thread listed as "checkdbcon"
Does the extra debug
Thomas,
Scored Country Code (DoSenderBase) when failed, exits getheader() as
absolute fail with insufficient score to block.
In getheader() See:
if (! &SenderBaseOK($fh,$this->{ip}))
--
Michael Thomas
Mathbox
978-687-3300
Toll Free: 1-877-MATHBOX (1-877-628-4269)
---
2012-04-25 11:56:39 Start populating Hidden Markov Model. HMM-check is
disabled for this time!
2012-04-25 11:56:40 start populating Hidden Markov Model ham chains with
653223 records!
2012-04-25 12:00:33 Finished populating Hidden Markov Model ham chains
with 653223 records!
2012-04-25 12:00:33
Thomas,
ASSP version 2.1.2(12116)
Scored SPF that fails in getheader() exits function as absolute fail,
but without sufficient score to fail. It should check
&MsgScoreTooHigh($fh,$done).
This
==
if(! SPFok($fh) ) {
if ($this->{error}) {$this->{skipnotspam} = 0;return;}
}
===
>An hour of that was the HMM generation
This is really too long.
You can't see the generation of the HMM in the rebuild log - you can only
see the populating time, which is a simple DB import.
My (more slow than fast) mysql server stores ~ 500.000 - 700.000 records
per minute.
Apr-25-12 04:1
On 25/04/2012 18:03, Thomas Eckardt wrote:
>> however there are some cases where one may score
>> and the other doesn't
>
> This is the dance on the 'wire rope' !
> But in most of this cases the HMM will be right. So an option is to
> disable the Bayes check. HMM has only a problem with very very s
>however there are some cases where one may score
>and the other doesn't
This is the dance on the 'wire rope' !
But in most of this cases the HMM will be right. So an option is to
disable the Bayes check. HMM has only a problem with very very small mails
(less than 10 words).
>only so many th
Thanks,
After correcting the ramdisk problem my rebuildspamdb went from taking
9+ hours to an hour and a half. An hour of that was the HMM generation,
is that always slower than bayes or do you think there might be anything
I can do to speed that up a bit more?
I didn't get any slow files repo
>What then follows is a list of the full 490 files, not just the 10
longest.
this will be fixed - meanwhile increase the first value of
'RebuildFileTimeLimit' .
Thomas
Von:Colin
An: ASSP development mailing list
Datum: 25.04.2012 12:52
Betreff:[Assp-test] Slow processing r
I did think of that, however there are some cases where one may score
and the other doesn't. This setup would let to a false negative as the
score would be halved. I suspect that will happen less than the false
positives though.
As to correctness, I used to keep an eye on it way back when I use
If you use both checks - think about - and set
'baysValencePB'/'bayslocalValencePB' to the half (eg 19 in your case).
Check your corpus - the HMM check will be nearly 100% correct, if the
corpus is OK.
Thomas
Von:Colin
An: ASSP development mailing list
Datum: 25.04.2012 14:35
Bet
Hello
sometime I need to activate the local bayesian check only for some email or
only for some domain,
for example t...@test.com .
To do that I need to enter all my local email in
Skip Bayesian for this local senders* (noBayesian_local)
except of t...@test.com
However this solution is not very
Hi folks,
I have had a few errors lately so I have been looking to change the
settings around a bit.
I run the system with the lower limit at 40 and the higher limit at 50.
the valencePB is set to 39.
This normally works well because any spam message trip additional checks
whilst bayes errors
Hi Thomas,
My install has stopped sending me emails when a new version has been
released. Was this feature disabled, or have I got something else going on?
At 01:10 AM 4/25/2012, Thomas Eckardt wrote:
>Hi all,
>
>fixed in assp 2.1.2 build 12116:
>
>- on some systems the rebuild spamdb failed wi
That's correct, 5.14.2.
At 12:57 AM 4/25/2012, Thomas Eckardt wrote:
>Hi Scott,
>
>only for me to know - I assume that you run perl 5.14.x - because only
>perl 5.14.x could run in to the 'wide character' error - I'm right?
>
>Thomas
>
>
>
>
>Von:Scott MacLean
>An: ASSP development mailin
Hi there,
My rebuildspamdb has been taking a very long time lately, I realised that I
hadn't set up a ramdisk for the tmpDB folder on my secondary since doing mysql
so I have done that and the difference is amazing. Hopefully that's solved it!
Because the rebuild has been running slow, I have n
22 matches
Mail list logo
