Message-ID: <f595ee7c-9e2d-e1c2-cdef-cab9c8197af2@joeconway.com>
Date: Wed, 17 May 2023 09:27:03 -0400
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.11.0
To: marc.morad@gmail.com, pgsql-bugs@lists.postgresql.org
References: <17937-c93266adaef384ff@postgresql.org>
Content-Language: en-US
From: Joe Conway <mail@joeconway.com>
Subject: Re: BUG #17937: Corrupt Indexes after reusing DRBD Device on new VM
In-Reply-To: <17937-c93266adaef384ff@postgresql.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Archived-At: 
 <https://www.postgresql.org/message-id/f595ee7c-9e2d-e1c2-cdef-cab9c8197af2%40joeconway.com>
Precedence: bulk

On 5/17/23 05:42, PG Bug reporting form wrote:
> After we migrated the VM from CentOS 7 to AlmaLinux 8, we had a few issues
> with broken/corrupt indexes. This problem occured a total of 3 times. Each

> The error messages all looked very similar to the following:
> Database error: [S]ERROR#012[V]ERROR#012[C]XX002#012[M]table tid from new
> index tuple (3346213,70) cannot find insert offset between offsets 2 and 4
> of block 5485 in index

> Have you ever heard of a similiar problem ?
> Any leads what might cause these problems would be very helpful.

CentOS 7 is glibc 2.17
AlmaLinux 8 is glibc 2.28

There are known issues with glibc upgrades (in particular 2.28, but also 
in general) causing collation changes which break all indexes which use 
glibc collations (basically text, varchar, char). You need to rebuild 
them, preferably before allowing *any* write activity (inserts or updates).

-- 
Joe Conway
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com