Date: Thu, 26 Jan 2023 13:39:44 +0100 (CET)
From: Erik Wienhold <ewie@ewie.name>
To: Massimo Ortensi <mortensi@unimaticaspa.it>,
	pgsql-admin@lists.postgresql.org
Message-ID: <2137860386.717969.1674736784938@office.mailbox.org>
In-Reply-To: <1pL0ni-00CnZb-3m@unimaticaspa.it>
References: 
 <CAK+uD7heWzrAjdRcSYq-z=crnb5mEs8JgsfGYtrK+vb0yrjeKg@mail.gmail.com>
 <b977effa32ce6e3088469871e294bea5d4d513b1.camel@cybertec.at>
 <1pKzDp-00CgfG-JH@unimaticaspa.it>
 <1968679708.705665.1674730579540@office.mailbox.org>
 <1pL0ni-00CnZb-3m@unimaticaspa.it>
Subject: Re: Problems with unique constraints
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Importance: Normal
Archived-At: 
 <https://www.postgresql.org/message-id/2137860386.717969.1674736784938%40office.mailbox.org>
Precedence: bulk

> On 26/01/2023 12:52 CET Massimo Ortensi <mortensi@unimaticaspa.it> wrote:
>
> we had OS upgrades, but with the same locale.

Which OS?  Locale may still be the same but collation information may have
changed, resulting in a different sort order.

https://postgresql.verite.pro/blog/2018/08/27/glibc-upgrade.html

Has libc version changed with the OS upgrades?

> Is there any chance to avoid rebuilding of all indexes/constraints by che=
cking
> the correctness of indexes ?

Extension amcheck (functions bt_index_check and bt_index_parent_check) can
verify that btree entries for text are in lexical order.

https://www.postgresql.org/docs/current/amcheck.html

> Il 26/01/2023 11:56, Erik Wienhold ha scritto:
>
> > On 26/01/2023 11:11 CET Massimo Ortensi <mortensi@unimaticaspa.it> wrot=
e:
> >
> > I had troubles with some unique constraints recently.
> >
> > They simply didn't work, not seeing some records (select found only 47
> > records with specific values instead of 65, insert could create
> > duplicate records).
> >
> > The constraint were on multiple columns. Drop constraint and create
> > constraint fixed the problem.
> >
> > This happened on two servers :
> >
> > - a version 10 server,=C2=A0 where an abnormal poweroff happened 6 mont=
hs
> > ago, found 3 tables with the problem throughout last 6 months.
> >
> > - a test server, found the problem on 1 table after upgrading version 1=
0
> > to 14
> >
> > Anythng similar happened to anybody ?
> > How can I find out if similar problems are present ? constraint and
> > index looked ok and were used
> >
> Are the indexes on string types?  Changes to the locale data is one cause=
:
> https://www.postgresql.org/message-id/328626.1647362495@sss.pgh.pa.us

--
Erik