From: Tom Lane <tgl@sss.pgh.pa.us>
To: fredmw7@gmail.com
cc: pgsql-bugs@lists.postgresql.org
Subject: Re: Server table rows permanently fixed (cannot be deleted or
 truncated)
In-reply-to: 
 <CAJ8mXMoKCyXhxo=d_qjxvSNSC8w1yDbw-WRgSKz40_bei8EzcA@mail.gmail.com>
References: 
 <CAJ8mXMoKCyXhxo=d_qjxvSNSC8w1yDbw-WRgSKz40_bei8EzcA@mail.gmail.com>
Comments: In-reply-to Fred Williams <fredmw7@gmail.com>
	message dated "Thu, 17 Nov 2022 07:49:34 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1834625.1668701055.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 17 Nov 2022 11:04:15 -0500
Message-ID: <1834626.1668701055@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/1834626.1668701055%40sss.pgh.pa.us>
Precedence: bulk

Fred Williams <fredmw7@gmail.com> writes:
> I am unable to delete or edit a specific range of table rows (I'll call
> them corrupted rows) for a specific primary key. If I truncate the entir=
e
> table, the corrupted rows remain. In the corrupted rows, I can successfu=
lly
> update non-primary key fields, but not primary-key fields.

> *Miscellaneous*

>    1. The problem does not occur on a test server using the same exact
>    table.
>    2. If I rename the table, I can remove the corrupted rows and/or upda=
te
>    the primary keys (such as changing the DateTime). However, when I ren=
ame it
>    back, the corrupted rows reappear!

TBH, I'm suspecting pilot error.  I wonder whether you have another
table by the same name in a different schema, which is the one
containing the "corrupted" rows, and after you rename this table
out of the way you're unintentionally accessing the other one.

That theory doesn't explain the rows surviving TRUNCATE, but
maybe inheritance could --- IIRC, TRUNCATE will not touch child
tables.  So maybe the alias table is also an inheritance child
of the one you are modifying?

This command in psql would clarify much:

postgres=3D# \d+ *."NCAASchedule"

			regards, tom lane