chris at techquility.net
Sun Jul 26 15:58:19 IST 2009
>On 23/07/2009 15:31, Jonas A. Larsen wrote:
>> I actualy traced it down to all the timeouts happening around 05:00.
>> That is when the mysql backup runs, so I think it must be some some
>> sort of mysql being unavailable while the backup runs.
>What other cron jobs happen on your system around then? Such as all the
>ones in /etc/cron.daily? (Your /etc/crontab will tell you when those
>run). It may be that is when the sa-rebuild happens, which is when it
>cleans up the Bayes database and removes all the old tokens from it.
>This makes Bayes unavailable for quite a while. Look for "Bayes" in
>MailScanner.conf and you will find there are some configuration options
>to alter this behaviour.
>> I use mysql for both mailwatch and bayes.
>> However I still believe a retry function for SA timeouts would be
>> to have in MS.
>> Julian: Please add it to wish list J
>> *From:* mailscanner-bounces at lists.mailscanner.info
>> [mailto:mailscanner-bounces at lists.mailscanner.info] *On Behalf Of
>> *Martin Hepworth
>> *Sent:* 23. juli 2009 15:31
>> *To:* MailScanner discussion
>> *Subject:* Re: Spamassassin timeouts
>> normally this is either due to
>> 1) DNS issues (lots of RBLs - have you got most of them turned off?),
>> 2) bayes (have you got this running on a faster SDBM database than
>>> default <snip>
I recently had a server that was having spamassassin timeouts and
showing mysql using a lot of cpu. Turned out on this server the cron job
for cleaning up bayes was missing so the bayes tables were growing
uncontrollably. Running sa-learn --force-expire, and then running an
optimize table on all bayes tables, cleaned this up. Then we added a
cron job to clean up bayes every night.
More information about the MailScanner