On Thu, Apr 22, 2010 at 10:38:51PM +0200, Mark Martinec wrote:
>
> > How about putting the bdb on a noatime device? (Or just changing the
> > file indivdidually?)
>
> bdb files in amavisd are opened rarely, so noatime may not gain anything
> there. It may be worth setting noatime on a temporary wo
On 4/22/2010 4:10 PM, Michael Scheidell wrote:
> On 4/22/10 5:03 PM, Noel Jones wrote:
>> With clamav (and likely other virus scanners), it's necessary
>> for the scanner to see the whole message for some signatures
>> to match. Normally one would just set $bypass_decode_parts =
>> 1 for this.
>>
On 4/22/10 5:03 PM, Noel Jones wrote:
> With clamav (and likely other virus scanners), it's necessary
> for the scanner to see the whole message for some signatures
> to match. Normally one would just set $bypass_decode_parts =
> 1 for this.
>
actually, there is a way to do this.
I use this,
With clamav (and likely other virus scanners), it's necessary
for the scanner to see the whole message for some signatures
to match. Normally one would just set $bypass_decode_parts =
1 for this.
But banned attachment blocking, bounce killer, and maybe other
features require decoding, so that
Alex,
> > There is also a more recent presentation:
> > http://www.ijs.si/software/amavisd/amavis-berlin-20090703.pdf
>
> Very cool. How about in plain-paper PDF or white background so we can
> print it? If it doesn't exist, then don't spend the time.
Sorry, no white bg around right now, but thi
Hi,
> There is also a more recent presentation:
>
> http://www.ijs.si/software/amavisd/amavis-berlin-20090703.pdf
Very cool. How about in plain-paper PDF or white background so we can
print it? If it doesn't exist, then don't spend the time.
>> How can I enable bayes in bdb with a global cache,
Michael,
> so in theory, if local.cf had required_score 5.0, the call to the SA pm
> would be able to compare the score it just computed vs the 5.0 and give
> an reasonable answer.
Exactly. Even in pure SpamAssassin installations (no amavis) it might
make sense to compare the score to a fixed val
On 4/22/10 11:24 AM, Mark Martinec wrote:
>
> If this serves only as an approximate decision, I believe it would
> not make much difference if kill level and required_hits are close
> but not exactly of the same value.
>
>
so in theory, if local.cf had required_score 5.0, the call to the SA pm
On Apr 22, 2010, at 11:33 AM, Mark Martinec wrote:
>> Does the same advice apply to MySQL-based quarantine tables?
>
> amavisd needs transactional model for its r/w databases,
> so you are practically required to use InnoDB engine with MySQL
> (or use PostgreSQL).
Understood. I've been using M
Clay,
> So to get full SpamAssassin related debug a command of amavisd -d 3,all
> (or a specific deubg area) i.e. 'amavisd -d 3,rules' would suffice given
> the documented and included chart as well.
Yes.
> Which can be pushed up to
> level 5 for additional amavisd-new related debug. Correct?
R
> Does the same advice apply to MySQL-based quarantine tables?
amavisd needs transactional model for its r/w databases,
so you are practically required to use InnoDB engine with MySQL
(or use PostgreSQL).
Mark
--
_
Dave,
> > Move Bayes and AWL databases to SQL, if you haven't already done so.
> > See sql/bayes_mysql.sql and sql/README.bayes in the SpamAssassin
> > package.
> > Using InnoDB and Mail::SpamAssassin::BayesStore::MySQL is a must
> > (i.e. do not use Mail::SpamAssassin::BayesStore::SQL as a
> > ba
Michael,
> I am assuming that since amavisd calls SA once per email (even with
> multiple recipients), that SA would be ignoring 'get_required_score' value.
> since that is different, potentially on all recipients.
Yes, the 'required_hits' and the get_required_score() are only used
by SpamAssassi
On 4/22/2010 9:53 AM, Mark Martinec wrote:
> Clay,
>
>> Since the change from using 'amavisd sa-debug' has been removed I was
>> beginning to use 'amavisd -d 5' to invoke to look at additional debug
>> output for SpamAssassin. Unfortunately, this does not appear to give as
>> much detail to the tes
On Apr 22, 2010, at 10:29 AM, Mark Martinec wrote:
>> I'm using an older version of amavisd-new now, but am currently in
>> the
>> process of upgrading. In the interim, I'd like to make some
>> performance improvements to ease the load until I can upgrade.
>
> Move Bayes and AWL databases to SQL,
Clay,
> Since the change from using 'amavisd sa-debug' has been removed I was
> beginning to use 'amavisd -d 5' to invoke to look at additional debug
> output for SpamAssassin. Unfortunately, this does not appear to give as
> much detail to the tests being performed that amavisd sa-debug did.
>
>
Stefan,
> On outgoing mails, I noticed that the DKIM headers inserted by
> amavisd-new look like that:
>
> DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=incertum.net; h=
> user-agent:in-reply-to:content-transfer-encoding
> :content-disposition:content-type:content-type:mi
osmcruzl,
> is there a way to rescue an incoming email that was blocked
> automatically from amavis, its seems that email was no delievered for
> policy "spam content" but this mails doesnt bring virus or spsm contents
> that way my internal user can check it ..
>
> this is the directoy of
Alex,
> I have a rather large bayes database with spamassassin now, and would
> like to investigate some performance tuning options to optimize memory
> usage and disk access.
>
> I came across this PDF from Marc from a few years ago, and although I
> think it's a bit dated, it looks helpful:
>
Hi,
I have a rather large bayes database with spamassassin now, and would
like to investigate some performance tuning options to optimize memory
usage and disk access.
I came across this PDF from Marc from a few years ago, and although I
think it's a bit dated, it looks helpful:
http://www.ijs.s
20 matches
Mail list logo
