cybozu-go / moco

MySQL operator on Kubernetes using GTID-based semi-synchronous replication.

Home Page:https://cybozu-go.github.io/moco/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Undo log truncate always fails on replica instances

masa213f opened this issue · comments

Describe the bug

The undo log truncate always fails on replica instances which are set super_read_only = 1.
This is a known issue in MySQL. It has not been fixed yet.

https://bugs.mysql.com/bug.php?id=104573

In addition, temporary files (/var/lib/mysql/data/undo_{1,2}_trunc.log) are left when this error occurs.
And, at the startup of MySQL, an unintended Undo Tablespace Truncation Recovery is executed.

2023-05-15T10:40:53.406807Z 1 [Note] [MY-012902] [InnoDB] Undo tablespace number 1 was being truncated when mysqld quit.
...
2023-05-15T10:40:54.134268Z 0 [ERROR] [MY-013266] [InnoDB] Thread# 0 -  Failed to finish truncating Undo Tablespace 'innodb_undo_001'

Environments

  • Version: MOCO v0.16.0