Minimum hardware capacity for 35k e-mail scans/day

[Denis Beauchemin]

Julian Field a écrit :
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Also, is it worth me changing the default spam.assassin.prefs.conf file 
> to set this on all new installations?
>
> Or should I specifically *not* change to this setting for some good reason?
>
> Thanks for your opinions,
> Jules.
>
> Julian Field wrote:
>   
>> * PGP Signed: 11/15/07 at 18:13:21
>>
>> Can someone add that to the Wiki please?
>> This is worth knowing!
>>
>> Steve Freegard wrote:
>>     
>>> Hi Edward,
>>>
>>> Edward Prendergast wrote:
>>>       
>>>> The queue seems to go up with the load. By the time the load is 
>>>> around 10
>>>> the queue starts to climb without any sign of going down. It usually 
>>>> reaches
>>>> the 300 mark within half an hour to an hour, but has gone up as far as
>>>> 3,000. When I turn bayes off the load goes down and mail queues start
>>>> becoming manageable again. 
>>>>         
>>> Ok - sounds like Bayes is a factor that is causing the slowness.  Do 
>>> you have Bayes in MySQL or DB_File?
>>>
>>> If DB_File then I would suggest setting 'bayes_learn_to_journal 1' in 
>>> spam.assassin.prefs.conf as from experience it will reduce the lock 
>>> contention and speed-up batch processing considerably.
>>>
>>> Kind regards,
>>> Steve.
>>>
>>> -- 
>>> Steve Freegard
>>> Fort Systems Ltd.
>>>       
>> Jules
>>
>>     
>
>
>   
Julian,

According to the man page:
           bayes_learn_to_journal   (default: 0)
               If this option is set, whenever SpamAssassin does Bayes 
learning, it
               will put the information into the journal instead of 
directly into the
               database.  This lowers contention for locking the 
database to execute
               an update, but will also cause more access to the journal 
and cause a
               delay before the updates are actually committed to the 
Bayes database.

After reading this, I was not under the impression I should turn it on 
on my servers...

Denis

-- 
   _
  °v°   Denis Beauchemin, analyste
 /(_)\  Université de Sherbrooke, S.T.I.
  ^ ^   T: 819.821.8000x62252 F: 819.821.8045