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

[Alex Broens]

On 11/20/2007 8:22 PM, Julian Field wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> 
> 
> Alex Broens wrote:
>> On 11/20/2007 7:08 PM, Julian Field wrote:
>>> Another criticism of SDBM: I just tried converting my Bayes db files 
>>> (670Mb) to SDBM and they grew by more than a factor of 3 ! (2048Mb)
>> imho, a 670MB Bayes DB is a massive perfomance hog.
>> Using autolearn and expiring every night, I doubt you'd really need 
>> more than the default settings. You'd probably be surprised how fast 
>> SA will process your mail with a less than 50 MB token file.
>>
>> If you are counting the Bayes_seen file, this irrelevant and can be 
>> deleted unless you need sa-learn to "forget" msgs.
> Yes, I was. Good to know I can just delete this. Can you just confirm 
> that deleting the bayes_seen* files is safe, before I do it?
> 

yep its safe... 100% *unless" you want to run sa-learn --forget, for 
example via Mailwatcatch - even then, you'd hardly need more than 1 
month worth's of data

best it to cron a rm -f to avoid surprises.
I've come across 50 GB seen files

automagically pruning the bayes_seen file is a RFE on SA Bugzilla

Alex