> well, my current opinion is that we should spend some nonzero amount
> of thought into figuring out what to do.
I'd suggest to do it like this:
Do autovac_refresh_stats() once per autovacuum_naptime/2 and share the
result among all autovacuum workers.
This would guarantee that autovacuum is fired no later than
autovacuum_naptime after the condition for the run became true.
If it's not that easy to code then don't share it among the workers and
do it once per worker - typically there are not so many workers.
And for bigger installations document that it's highly recommend to put
the stats file on ramdisk.
Dne 17.2.2010 0:12, Alvaro Herrera napsal(a):
> Tom Lane wrote:
>> Alvaro Herrera<alvherre(at)commandprompt(dot)com> writes:
>>> Jakub Ouhrabka wrote:
>>>> Was autovacuum requesting to write this 20MB file 650x per minute?
>>> Yes, exactly.
>>> Ideally, autovacuum would only request a new copy of the file if the one
>>> it got was considerably out of date. Obviously a tenth of a second is
>>> not old enough.
>> Wasn't it you that insisted on a short staleness criterion for autovac
>> in the first place?
> well, my current opinion is that we should spend some nonzero amount of
> thought into figuring out what to do.
In response to
pgsql-hackers by date
|Next:||From: Jakub Ouhrabka||Date: 2010-02-17 07:42:05|
|Subject: Re: Problem with 8.4 stats collector high load|
|Previous:||From: Matthias Brantner||Date: 2010-02-17 07:23:45|
|Subject: Re: XQuery support|