Ow Mun Heng <Ow.Mun.Heng at wdc.com> writes:
> On Tue, 2008-08-12 at 16:08 +0200, Cyril SCETBON wrote:
>> 
>> Cyril SCETBON wrote:
>>  As we have a lot of 
>> statement, I should use logswitch_start manually if I want it to happen 
>> more often. But what means "every so often" in the documentation ? (see 
>> above)
>> 
>
> I'm forcing the switch every 12 hours on mine to get the log file sizes
> down. once it hits ~1-2Gm then I get the dreaded fetch 100 from log.

In v1.2, the slon initiates running this on a weekly basis; take a
look in the table "sl_registry"; there should be a tuple with "reg_key
= 'logswitch_weekly.dow'"; that controls which day of the week it
runs, and there may also be a tuple with reg_key =
'logswitch_weekly.time'.

But basically, the initial design was to do so weekly.

If you want it to take place more often, then run logswitch_start()
more often, and twice a day doesn't seem outrageous.

In v2.0, this is revisited; we *never* DELETE data from the log
tables, exclusively using TRUNCATE, and use the slon parameter
"cleanup_interval" to indicate how much old data gets kept around.  If
that interval is 2 hours, then you can expect to see a TRUNCATE about
every 2 hours.
-- 
(format nil "~S@~S" "cbbrowne" "acm.org")
http://www3.sympatico.ca/cbbrowne/linux.html
"The only completely consistent people are the dead."
-- Aldous Huxley