speed of bayes (mysql)

Jason Ede J.Ede at birchenallhowden.co.uk
Sat Jun 21 10:58:17 IST 2008

I've found a few things eventually that seem to make a huge difference

Firstly, after I run sa-learn -force-expire I run mysqlcheck -A -o to optimise and defrag all the tables.

Then I found http://rackerhacker.com/mysqltuner/

This made several suggestions and I've tried some of them and it seems to make a big difference to the time taken doing the mysql bayes checks, although I'll need to give it a bit longer to make sure the times are consistently down.

I've been wondering if putting the bayes in innodb format will help, although I've seen problems with the innodb data file growing and really need to have innodb_file_per_table option set in my.cnf so that can use mysqlcheck -o to keep the tables optimised and the file size down.


From: mailscanner-bounces at lists.mailscanner.info [mailto:mailscanner-bounces at lists.mailscanner.info] On Behalf Of Jason Ede
Sent: 19 June 2008 09:13
To: MailScanner discussion
Subject: speed of bayes (mysql)

I've bayes running in mysql and after upgrading I ran it in debug mode to see what the times were like... Generally it was fine apart from the bayes check which took ages! (batch size was 10 messages)

08:52:10 [7768] dbg: bayes: database connection established
08:52:10 [7768] dbg: bayes: found bayes db version 3
08:52:10 [7768] dbg: bayes: Using userid: 5
08:53:56 [7768] dbg: bayes: seen (c21a23d9453de70a43644e737a49e81c47640d54 at sa_generated) put
08:53:56 [7768] dbg: bayes: learned 'c21a23d9453de70a43644e737a49e81c47640d54 at sa_generated', atime: 1213858923
08:53:56 [7768] dbg: learn: initializing learner
08:53:56 [7768] dbg: check: is spam? score=-3.598 required=5
08:53:56 [7768] dbg: check: tests=BAYES_00,HTML_MESSAGE,RCVD_IN_DNSWL_LOW

Any ideas on things I can check/investigate to speed this up?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mailscanner.info/pipermail/mailscanner/attachments/20080621/ff463539/attachment.html

More information about the MailScanner mailing list