Content-Type: text/plain;
	charset=us-ascii
Mime-Version: 1.0 (1.0)
Subject: Re: pg_dump bug?
From: Scott Mead <scottm@openscg.com>
In-Reply-To: 
 <CAJWczd9Z-3vHrv3nF1z-tRUEUbCSSW=ynsbrqt1mY5TWXBE26Q@mail.gmail.com>
Date: Tue, 11 Apr 2017 06:33:33 -0400
Cc: pgsql-admin@postgresql.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <9E1C43FE-68CD-4661-98E9-2E682723EB08@openscg.com>
References: 
 <CAJWczd9Z-3vHrv3nF1z-tRUEUbCSSW=ynsbrqt1mY5TWXBE26Q@mail.gmail.com>
To: Andy Nercessian <ahnercessian@gmail.com>
Precedence: bulk
Sender: pgsql-admin-owner@postgresql.org


> On Apr 11, 2017, at 05:20, Andy Nercessian <ahnercessian@gmail.com> wrote:
>=20
> I wonder if anyone has experienced the problem I have with bloated pg_dum=
p files. This problem exists in both standard pg_dumps and --insert-columns=
 type pg_dumps. I am using postgresql 9.5 on Windows 7.
>=20
> Each table appears to have a 'copy table_name .... from stdin..' in multi=
ple places (hundreds in some cases). The first copy statement appears genui=
ne, but the second and subsequent statements appear to be in the middle of =
another copy clause between records. The outer copy clause is sometimes fro=
m the same table, and sometimes from a different table. Furthermore, from t=
he existence of the number of backslashes \t, \\t, \\\\t which are variable=
, it would appear as though nesting can be several levels deep.
>=20
> I'm not copying the actual data because apart from anything else, it's a =
giant text file. But this is roughly what it looks like:
>=20
> ------------------------------------
> copy bills from ....=20
> Record 1 Column 1 \t Record 1 column 2 \t ....
> Record 2 Column 2 \t Record 2 column 2 \t ...
> ......
> copy adverts from .... < this is a nested copy statement unrelated to bil=
ls, although occasionally it's another bills copy statement>
>=20
> Record 88 Column 1 \t Record 88 Column 2 \t ...
>=20
> -------------------------------------
>=20
> So in the middle of one copy statement - in between the records being cop=
ied, another one mysteriously appears and the records are duplicated. Each =
record can therefore be found in each of the copy statements, and the total=
 number of such duplications is in the hundreds for some tables.
>=20
> What could be causing this? I've made only the following observations:
> 1. The nested loops always occur in the same spot within the outer copy s=
tatements (if records have not changed between dumps) and are preceded by t=
he exact same column of the same record in each dump.
> 2. It is clear that these copy statements are NOT part of any record.=20
> 3. The same behaviour exists for insert style dumps (where there is no co=
py command).
> 4. Each copy statement includes the entire table, NOT a selection of reco=
rds from it.
>=20
> In years of using pg_dump I have never come across this issue.=20

Is it possible that you have more than 1 pg_dump running simultaneously aga=
inst the same output file?


>=20
>=20
> --=20
> Andy=20


--=20
Sent via pgsql-admin mailing list (pgsql-admin@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-admin