Re: Auto-tuning work_mem and maintenance_work_mem

On Mon, Feb 17, 2014 at 07:39:47PM +0100, Andres Freund wrote:
> On 2014-02-17 13:33:17 -0500, Robert Haas wrote:
> > On Mon, Feb 17, 2014 at 11:33 AM, Andres Freund <andres(at)2ndquadrant(dot)com> wrote:
> > >> And I still disagree with this- even in those cases. Those same untuned
> > >> servers are running dirt-simple queries 90% of the time and they won't
> > >> use any more memory from this, while the 10% of the queries which are
> > >> more complicated will greatly improve.
> > >
> > > Uh. Paging.
> >
> > What about it?
>
> It's often the source of a good portion of the queries and load in web
> applications. Multiple joins and more than one row... I have several
> time seen stats changes or bad to-be-sorted columns cause large amounts
> of memory to be used.

Perhaps we should have said there was "general agreement" to increase
work_mem and maintenence_work_mem by 4x, not that there was 100%
agreement. It would be nice to have 100% agreement, but if we _require_
that then defaults would probably never be changed.

> Anyway, I've stated my opinion that I do not think it's a good idea to
> raise that particular default (while agreeing with all the others) and I
> know I am in the minority, so I don't think we need to argue this out...

OK, good. If you did feel there was need for more discussion, we would
need to push this change to PG 9.5.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ Everyone has their own god. +