Re: Mysql CPU usage above 100% with empty mail queues

Looks like the queue started filling up again. Just a few minutes ago I found that our queues were again up around 1200 messages. This time what I did is turned off bayesian filtering (use_bayes 0 in /etc/mail/spamassassin/local.cf), and restarted amavisd-new.

 

This seemed to really help things without me needing to touch mysql.

 

However, mysqld is still taking up at least 80-95% CPU constantly.

 

Any ideas? Do I need to wipe out my bayes_token table? Is it supposed to ever expire old tokens, or just continually grow until my server finally dies from having like 9,999,999,999,999 records in that table (which I suppose would eventually happen if it’s not cleaned out/recycled somehow)?

 

Again, thanks in advance for any ideas and expertise.

 

 

 

Hello,

 

I am concerned that there may be something wrong with my maia database or mysql server on my debian-etch maia system. Right now the mail queues are pretty much empty. I’m processing anywhere between 1-3 messages per second. This usually doesn̵7;t put much of a load on our server. However, right now mysql is taking anywhere between 80-110% CPU constantly (multi-core CPU).

 

This may have been related to a little issue we had about 30 minutes ago with the queues suddenly having about 1200 unprocessed messages in them. I couldnR17;t find a problem from any of the following logs:

 

/var/log/mail.log (or its variants mail.err, mail.warn, mail.info)

/var/log/clamav/clamav.log

/var/log/clamav/freshclam.log

/var/log/messages

 

I don’t know why /var/log/mysql.log & mysql.err are always empty. Logrotate is rotating them, but nothing gets written to them. In /etc/mysql/my.cnf it says “Error logging goes to syslog. This is a Debian improvement  ”, but I haven̵7;t seen it writing any mysql stuff to /var/log/mysql.log or mysql.err or to /var/log/messages.

 

Maybe this is because there have been no messages to write. Anyway, I’m kind of worried. I took the following steps to alleviate the issue with the 1200 messages in the queue:

 

·         Purged all records from awl table ; in mysql (using phpMyAdmin). It had over 3,000,000 records, and I thought this could be causing problems.

o   This didn’;t get things going though.

·         Restarted clamav-deamon, amavisd-new (maia), and postfix

o   This didn’;t seem to speed things up either.

·         Restarted mysql

o   This really got things going after a minute or two, it started processing messages pretty quickly again

 

Prior to each of these steps, messages were being processed, but just much more slowly than normal.

 

So now, for at least 20 minutes, it seems like things are going well. But I’m worried about why mysqld is taking so much CPU usage.

 

If you have any ideas, I would greatly appreciate your help!!  J

 

 

Also, here’;s some info about the existing mysqld tables in maia, which I’m getting from phpMyAdmin:

 

awl:762 records (80KB)

bayes_seen: 726,873 records (100MB)

bayes_token: 1,278,672 records (172MB)

maia_mail: 270,253 records (1.4GB)

maia_mail_recipients: 235,397 records (123.3MB)

maia_sa_rules: 1,131 records (192KB)

maia_sa_rules_triggered: 2,094,101 records (167.2MB)

maia_stats_history: 2,336 records (464KB)

 

All other tables are under 250 records, which I’m assuming is pretty small and shouldn217;t cause any negative impact.

 

Thanks again for your help in advance.

 

__

Douglas Mortensen

      Network Consultant

         Impala Networks

 

           ;   CCNA

           ; A+, Linux+, Network+