Date: Sat, 23 Jul 2005 14:36:50 -0600
From: Robert Creager <Robert_Creager@LogicalChaos.org>
Cc: Robert Creager <Robert_Creager@StorageTek.com>,
	Tom Lane <tgl@sss.pgh.pa.us>, pgsql-performance@postgresql.org,
	Andy Hewitt III <ralph_hewitt_iii@StorageTek.com>
Subject: Re: Huge performance problem between 7.4.1 and 8.0.3 - CS
Message-ID: <20050723143650.5678061a@thunder.logicalchaos.org>
In-Reply-To: <20050719224908.15415222@thunder.logicalchaos.org>
References: <20050718114952.00004ac5@C118181.stortek.com>
	<15472.1121709173@sss.pgh.pa.us>
	<20050719105014.00007635@C118181.stortek.com>
	<7322.1121792062@sss.pgh.pa.us>
	<20050719120951.00002520@C118181.stortek.com>
	<20050719224908.15415222@thunder.logicalchaos.org>
Organization: Starlight Vision, LLC.
Mime-Version: 1.0
Content-Type: multipart/signed;
	boundary="Signature_Sat__23_Jul_2005_14_36_50_-0600_4Cp=mBFyqM7DO1VE";
	protocol="application/pgp-signature"; micalg=pgp-sha1

--Signature_Sat__23_Jul_2005_14_36_50_-0600_4Cp=mBFyqM7DO1VE
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable


I've now backed off to version 7.4.1, which doesn't exhibit the problems th=
at
8.0.3 does.  I guess I'll wait 'till the next version and see if any progre=
ss
has occurred.

Rob

When grilled further on (Tue, 19 Jul 2005 22:49:08 -0600),
Robert Creager <Robert_Creager@logicalchaos.org> confessed:

> When grilled further on (Tue, 19 Jul 2005 12:09:51 -0600),
> Robert Creager <Robert_Creager@StorageTek.com> confessed:
>=20
> > On Tue, 19 Jul 2005 12:54:22 -0400
> > Tom Lane <tgl@sss.pgh.pa.us> wrote:
> >=20
> > > Hmm, I hadn't thought about the possible impact of multiple concurrent
> > > vacuums.  Is the problem caused by that, or has performance already g=
one
> > > into the tank by the time the cron-driven vacuums are taking long eno=
ugh
> > > to overlap?
> >=20
> >=20
> > I'll re-start the database, vacuum full analyze and restart the runs wi=
thout
> the
> > cron vacuum running.
> >=20
>=20
> It took a few hours, but the problem did finally occur with no vacuum run=
ning
on
> 803.  CS is averaging 72k.  I cannot quantitatively say it took longer to
> reproduce than with the vacuums running, but it seemed like it did.
>=20
> Can any information be gotten out of this?  Should I try CVS HEAD?
>=20
> Thoughts?
>=20
> Thanks,
> Rob
>=20
> --=20
>  22:41:36 up 6 days,  2:16,  6 users,  load average: 0.15, 0.21, 0.30
> Linux 2.6.5-02 #8 SMP Mon Jul 12 21:34:44 MDT 2004


--=20
 14:35:32 up 9 days, 18:10,  5 users,  load average: 2.17, 2.19, 2.15
Linux 2.6.5-02 #8 SMP Mon Jul 12 21:34:44 MDT 2004

--Signature_Sat__23_Jul_2005_14_36_50_-0600_4Cp=mBFyqM7DO1VE
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iEYEARECAAYFAkLiqmIACgkQLQ/DKuwDYzksAQCdGNv/wST8EwGMlSgbUY0jWmrB
GVgAoI/OWLXYULkLGMvKLzlB1+CeKSNB
=Y7SY
-----END PGP SIGNATURE-----

--Signature_Sat__23_Jul_2005_14_36_50_-0600_4Cp=mBFyqM7DO1VE--