Re: very, very slow performance

On Sat, Feb 21, 2009 at 3:11 AM, Tena Sakai <tsakai@gallo.ucsf.edu> wrote:
> Hi Scott,
>
> Thanks for clustering suggestion.  I have never used
> cluster command and I need to read up before I can
> use it.
>
> I have adjusted postgres parameters per your recommen-
> dation.  Work_mem is now 8GB, checkpoint_segments is
> raised to 100.  Shared_buffers is still at 1GB because
> this is bound by SHMMAX of kernal, which I cannot alter
> for maybe a few days.  I am thinking of setting SHMMAX
> to 8GB.

Yeah, I'd reverse those two and set work_mem to something in the 512M
range max.  keep in mind work_mem is per query / per sort.  4 queries
with3 sorts each = max allocation of work_mem*3*4...

 I may or may not be able to move up to 8.3.6

> over this weekend.  If not, sometime during the week.

definitely a good idea.