Tuning Postgresql on FreeBSD 5.1

[Paul Pathiakis]

Hi,

	I'm running 5.1-RELEASE.  No, I didn't enable it in my kernel config.  Top 
shows only cpu 0 and cpu 2 running.  However, as the machine boots it tells 
me that it enables cpu 1, 2, and 3.  The only thing in my kernel that I 
enabled is SMP and APIC_IO.

	Thanks!

	P.


On Tuesday 26 August 2003 03:19 pm, Marc G. Fournier wrote:
> You are running -STABLE?  Did you enable HTT in your kernel config file?
>
> On Tue, 26 Aug 2003, Paul Pathiakis wrote:
> > Marc,
> >
> > 	I've tried disabling HTT in the BIOS (it's an Intel board).  I've
> > disable HTT, saved the changes and the kernel is still seeing 4 CPUs when
> > it boots. Any ideas?
> >
> > 	Thanks,
> >
> > 	P.
> >
> > On Tuesday 26 August 2003 12:44 am, Marc G. Fournier wrote:
> > > > > Again, the machine is a twin 2.8 Xeon HTT machine.  HTT is turned
> > > > > on and the machine sees 4 cpus.
> > > >
> > > > Have you played with toggling the sysctl machdep.cpu_idle_hlt?
> > >
> > > In fact, disabled HTT altogether ... I have a machine with pretty much
> > > the same specs (2.4 vs 2.8 Xeon's) and I found performance noticeably
> > > improved with HTT disabled ... not just with processes, but with
> > > interactive sessions as well ...
> > >
> > > > >From your postgresql.conf:
> > > > >
> > > > > shared_buffers = 48000          # min max_connections*2 or 16, 8KB
> > > > > each
> > > >
> > > > WHOA!  This is too high by a factor of about 10.  You probably want a
> > > > shared buffers set to 4096.
> > >
> > > Why?  If you have the memory and all that ... All my production servers
> > > run:
> > >
> > > /usr/local/bin/postmaster -B 40960 -N 512 -i -p 5432
> > > -D/usr/local/pgsql/5432 -S (postgres)
> > >
> > > > > sort_mem = 32768                # min 64, size in KB
> > > >
> > > > This also seems high, divide by 8 and you're at a more reasonable
> > > > level.
> > >
> > > Again, depends on alot of things here ... if he only has the one
> > > connection to the DB, allowing for 32M of RAM to be used for sorting
> > > isn't a bad thing, since it keeps the sorts off of the hard drive ...
> > > that is one stat that I wish we kept somehow ... "max sort size" ...