[debian-mysql] Bug#996028: Bug#996028: #996028 InnoDB: corrupted TRX_NO after upgrading to 10.3.31

[Ondrej Zary]

On Friday 22 October 2021, Yves-Alexis Perez wrote:
> On Thu, 2021-10-14 at 19:38 +0200, Ondrej Zary wrote:
> >     MDEV-15912: Remove traces of insert_undo
> >
> >     Let us simply refuse an upgrade from earlier versions if the
> >     upgrade procedure was not followed. This simplifies the purge,
> >     commit, and rollback of transactions.
> >
> >     Before upgrading to MariaDB 10.3 or later, a clean shutdown
> >     of the server (with innodb_fast_shutdown=1 or 0) is necessary,
> >     to ensure that any incomplete transactions are rolled back.
> >     The undo log format was changed in MDEV-12288. There is only
> >     one persistent undo log for each transaction.
> 
> 
> Hi Ondrej,
> it might be worth trying with that patch reverted, but I'm a bit confused at
> the commit message. That seems to imply my server didn't have a clean shutdown
> before upgrading to 10.3, which is possible but that upgrade is quite old by
> now (afaict it's between Stretch and Buster), so I'm unsure what to do now.
> 
> innodb_fast_shutdown=0 doesn't terminate on 10.3.25, and I'm not sure if there
> are other ways to ensure a “clean shutdown”.
> 
> Regards,

With 10.3.29 running, I've dumped all databases (mysqldump --all-databases -p >mysql.dump), then dropped all databases, stopped mariadb and deleted /var/lib/mysql/ib*.
Then restarted mariadb, restored databases (source mysql.dump) and finally upgraded to 10.3.31.

So my problem is solved. I've also found the patch that's causing the problem. I'm not going to waste any more time on this.

During many years of running mysql, upgrade just worked all the time. Looks like there are some quality problems with the developement process now.

-- 
Ondrej Zary