Re: Massive increase in InnoDB corruption issues since upgrade to 8.0.29

[Ralf Meyer]

Dear Daniel,


Thank you for your reply.

Actually, this seems to be exactly the issue we are having. On some systems, this issue only leads to crashes once the table is accessed (as you have described). When we look at the table contents, we can see what you mentioned ("Values are randomly spread out across columns"). I guess the reason why we have some servers where MySQL does not start anymore sometime after the issue is that the table somehow ends up in the crash recovery data. In this case, there are actions to be performed on that table on every start of the server and therefore it then immediately crashes, as it cannot successfully complete the crash recovery.

Interestingly enough, there seems to be some connection (at least in our case) to Wordpress and especially the Plugin "Woocommerce". The table that is corrupted is, in most cases, called wc_admin_note_actions and part of the Woocommerce plugin for Wordpress. That being said, we have also seen this happen to other tables, but I guess the underlying issue could be the same. Even more interesting, I have actually just found this description of exactly our problem: https://serverfault.com/questions/1101865/mysql-errors-corruption-after-updating-wordpress

So indeed, there seems to be some kind of bigger issue in MySQL 8.0.29 related to InnoDB corruption.

So far, we have tried quite a bit to be able to reproduce it, with no luck though. On upgrade it seems that Woocommerce does a "DROP COLUMN":
https://github.com/woocommerce/woocommerce-admin/blob/main/includes/wc-admin-update-functions.php#L364

That could be related to your description and the problem. We however, could not reproduce it so far, with the datasets that we have.