[MDEV-30422] Merge new release of InnoDB 5.7.41 to 10.3 Created: 2023-01-17  Updated: 2023-04-19  Resolved: 2023-01-17

Status: Closed
Project: MariaDB Server
Component/s: Storage Engine - InnoDB
Affects Version/s: 10.3, 10.4
Fix Version/s: 10.3.38, 10.4.28

Type: Bug Priority: Blocker
Reporter: Marko Mäkelä Assignee: Marko Mäkelä
Resolution: Fixed Votes: 0
Labels: corruption, not-10.5, race, upstream

Issue Links:
Blocks
blocks MDEV-31087 Merge new release of InnoDB 5.7.42 to... Closed
is blocked by MDEV-29766 Merge new release of InnoDB 5.7.40 to... Closed
Relates
relates to MDEV-12024 InnoDB startup fails to wait for recv... Closed
relates to MDEV-24449 Corruption of system tablespace or la... Closed

 Description   

MySQL 5.7.41 includes the following InnoDB changes:

Not applicable to MariaDB Server

Bug #33822493 : pb2 weekly-5.7 test failures due to InnoDB: Assertion failure
This race condition was fixed in MDEV-12024 in February 2017.

Applicable to MariaDB Server

Bug#34400586 Backport MySQL contribution id: 31497 to 5.7
The commit message appears to refer to this change in MySQL 8.0.
This aims to prevent a last-minute dirtying of the buffer pool due to a change buffer merge. In MariaDB Server 10.5, not only was the buf_flush_page_cleaner() thread refactored, but also MDEV-19514 should prevent any change buffer merge from occurring at this point.
I believe that the Oracle internal bug report was similar to MDEV-20481, which turns out to have been fixed by MDEV-24448 and MDEV-24449.

All that said, something like this fix should be applicable to MariaDB 10.3 and 10.4, to prevent a very rare condition where read-ahead of a secondary index leaf page had been submitted during or right before shutdown. As always (see MDEV-30009), it is very hard to reproduce any bugs in the change buffer.


Generated at Thu Feb 08 10:16:08 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.