DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com;
	h=Message-ID:X-YMail-OSG:Received:X-Mailer:Date:From:Subject:To:Cc:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding;
	b=Mbk9Zox89Kt1Y29rZpeTc6FT8O13+sQQymX8U+k/yFGr8BQjsMsh5LZFDvl3FHoZUT9lrPHN0sQ9fdsSTBTkelh4QjhU1NyJzqwoRR8tAYd4QD+yAT9t7vP8d9qMy5/zlMWNbB3KBpE6Qj2k9TJ988jyjthnO7ikRGn6AqvnhbM=;
Message-ID: <584872.74177.qm@web39702.mail.mud.yahoo.com>
Date: Thu, 10 Jun 2010 23:25:47 -0700 (PDT)
From: Bob Lunney <bob_lunney@yahoo.com>
Subject: Re: requested shared memory size overflows size_t
To: Tom Wilcox <hungrytom@gmail.com>
Cc: pgsql-performance@postgresql.org
In-Reply-To: <4C10FA6C.6060600@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable

Tom,=0A=0AFirst off, I wouldn't use a VM if I could help it, however, somet=
imes you have to make compromises.  With a 16 Gb machine running 64-bit Ubu=
ntu and only PostgreSQL, I'd start by allocating 4 Gb to shared_buffers.  T=
hat should leave more than enough room for the OS and file system cache.  T=
hen I'd begin testing by measuring response times of representative queries=
 with significant amounts of data.=0A=0AAlso, what is the disk setup for th=
e box?  Filesystem?  Can WAL files have their own disk?  Is the workload OL=
TP or OLAP, or a mixture of both?  There is more that goes into tuning a PG=
 server for good performance than simply installing the software, setting a=
 couple of GUCs and running it.=0A=0ABob=0A=0A--- On Thu, 6/10/10, Tom Wilc=
ox <hungrytom@gmail.com> wrote:=0A=0A> From: Tom Wilcox <hungrytom@gmail.co=
m>=0A> Subject: Re: [PERFORM] requested shared memory size overflows size_t=
=0A> To: "Bob Lunney" <bob_lunney@yahoo.com>=0A> Cc: "Robert Haas" <robertm=
haas@gmail.com>, pgsql-performance@postgresql.org=0A> Date: Thursday, June =
10, 2010, 10:45 AM=0A> Thanks guys. I am currently=0A> installing Pg64 onto=
 a Ubuntu Server 64-bit installation=0A> running as a VM in VirtualBox with=
 16GB of RAM accessible.=0A> If what you say is true then what do you sugge=
st I do to=0A> configure my new setup to best use the available 16GB (96GB=
=0A> and native install eventually if the test goes well) of RAM=0A> on Lin=
ux.=0A> =0A> I was considering starting by using Enterprise DBs tuner to=0A=
> see if that optimises things to a better quality..=0A> =0A> Tom=0A> =0A> =
On 10/06/2010 15:41, Bob Lunney wrote:=0A> > True, plus there are the other=
 issues of increased=0A> checkpoint times and I/O, bgwriter tuning, etc.=A0=
 It may=0A> be better to let the OS cache the files and size=0A> shared_buf=
fers to a smaller value.=0A> > =0A> > Bob Lunney=0A> > =0A> > --- On Wed, 6=
/9/10, Robert Haas<robertmhaas@gmail.com>=A0=0A> wrote:=0A> > =0A> >=A0 =A0=
 =0A> >> From: Robert Haas<robertmhaas@gmail.com>=0A> >> Subject: Re: [PERF=
ORM] requested shared memory=0A> size overflows size_t=0A> >> To: "Bob Lunn=
ey"<bob_lunney@yahoo.com>=0A> >> Cc: pgsql-performance@postgresql.org,=0A> =
"Tom Wilcox"<hungrytom@googlemail.com>=0A> >> Date: Wednesday, June 9, 2010=
, 9:49 PM=0A> >> On Wed, Jun 2, 2010 at 9:26 PM, Bob=0A> >> Lunney<bob_lunn=
ey@yahoo.com>=0A> >> wrote:=0A> >>=A0 =A0 =A0 =0A> >>> Your other option, o=
f course, is a nice 64-bit=0A> linux=0A> >>>=A0 =A0 =A0 =A0 =0A> >> variant=
, which won't have this problem at all.=0A> >> =0A> >> Although, even there=
, I think I've heard that=0A> after 10GB=0A> >> you don't get=0A> >> much b=
enefit from raising it further.=A0 Not=0A> sure if=0A> >> that's accurate o=
r=0A> >> not...=0A> >> =0A> >> -- Robert Haas=0A> >> EnterpriseDB: http://w=
ww.enterprisedb.com=0A> >> The Enterprise Postgres Company=0A> >> =0A> >>=
=A0 =A0 =A0 =0A> > =0A> > =0A> >=A0 =A0 =0A> =0A> =0A=0A=0A