Date: Sun, 27 May 2007 11:06:48 -0500
From: "Jim C. Nasby" <decibel@decibel.org>
To: Leandro Guimar?es dos Santos <lgsantos@contax.com.br>
Cc: pgsql-performance@postgresql.org
Subject: Re: Memory allocation and Vacuum abends
Message-ID: <20070527160648.GP92628@nasby.net>
References: 
 <6DF4B01FFBC80140B4FA64BCA9824E8C02F778ED@MESCTX01RJORP.contax-br.contax.root>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha1;
	protocol="application/pgp-signature"; boundary="ibq+fG+Ci5ONsaof"
Content-Disposition: inline
In-Reply-To: 
 <6DF4B01FFBC80140B4FA64BCA9824E8C02F778ED@MESCTX01RJORP.contax-br.contax.root>
User-Agent: Mutt/1.5.13 (2006-08-11)


--ibq+fG+Ci5ONsaof
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

What does top report as using the most memory?

On Wed, May 23, 2007 at 11:01:24PM -0300, Leandro Guimar?es dos Santos wrot=
e:
> Hi all,
>=20
> =20
>=20
> I have a 4 CPU, 4GB Ram memory box running PostgreSql 8.2.3 under Win 200=
3 in a very high IO intensive insert application.
>=20
> =20
>=20
> The application inserts about 570 rows per minute or 9 rows per second.
>=20
> =20
>=20
> We have been facing some memory problem that we cannot understand.
>=20
> =20
>=20
> From time to time memory allocation goes high and even after we stop post=
gresql service the memory continues allocated and if were restart the servi=
ce the Postgres crash over.
>=20
> =20
>=20
> It's a 5 GB database size already that was born 1 and a half month ago. W=
e have 2 principal tables partitioned.
>=20
> =20
>=20
> Above is the log file. Do anyone have any idea what could the problem be.=
=2E....
>=20
> =20
>=20
> Thanks in advance.
>=20
> =20
>=20
> =20
>=20
> 2007-05-23 13:21:00 LOG:  CreateProcess call failed: A blocking operation=
 was interrupted by a call to WSACancelBlockingCall.
>=20
>              (error code 1450)
>=20
> 2007-05-23 13:21:00 LOG:  could not fork new process for connection: A bl=
ocking operation was interrupted by a call to WSACancelBlockingCall.
>=20
>            =20
>=20
> 2007-05-23 13:21:06 LOG:  could not receive data from client: An operatio=
n on a socket could not be performed because the system lacked sufficient b=
uffer space or because a queue was full.
>=20
> 2007-05-23 13:21:17 LOG:  server process (PID 256868) exited with exit co=
de 128
>=20
> 2007-05-23 13:21:17 LOG:  terminating any other active server processes
>=20
> 2007-05-23 13:21:17 WARNING:  terminating connection because of crash of =
another server process
>=20
> 2007-05-23 13:21:17 DETAIL:  The postmaster has commanded this server pro=
cess to roll back the current transaction and exit, because another server =
process exited abnormally and possibly corrupted shared memory.
>=20
> 2007-05-23 13:21:17 HINT:  In a moment you should be able to reconnect to=
 the database and repeat your command.
>=20
> 2007-05-23 13:21:17 WARNING:  terminating connection because of crash of =
another server process
>=20
> 2007-05-23 13:21:17 DETAIL:  The postmaster has commanded this server pro=
cess to roll back the current transaction and exit, because another server =
process exited abnormally and possibly corrupted shared memory.
>=20
> 2007-05-23 13:21:17 WARNING:  terminating connection because of crash of =
another server process
>=20
> 2007-05-23 13:21:17 DETAIL:  The postmaster has commanded this server pro=
cess to roll back the current transaction and exit, because another server =
process exited abnormally and possibly corrupted shared memory.
>=20
> =20
>=20

--=20
Jim Nasby                                      decibel@decibel.org
EnterpriseDB      http://enterprisedb.com      512.569.9461 (cell)

--ibq+fG+Ci5ONsaof
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.3 (FreeBSD)

iD8DBQFGWayYdO30qud8SkgRAs5IAJwPWAoMtjnurqLyqbtTxwszvKpRGQCfX+rt
ergzuyQxJKL4MkV7gYwPjRQ=
=gA/B
-----END PGP SIGNATURE-----

--ibq+fG+Ci5ONsaof--