你好,游客 登录 注册 搜索
背景:
阅读新闻

MySQL表索引损坏致Crash及修复过程实例

[日期:2017-09-01] 来源:Linux社区  作者:神谕丶 [字体: ]

监控到一台MySQL实例在早上发生过Crash,上去看了一下,已经被mysqld_safe成功拉起。
上去检查一下错误日志,发现错误日志如下(已对表名,库名,路径做脱敏处理):

  1. ……………………………………(大量相同的报错)…………………………………………
  2. 2017-08-31T11:11:04.291424Z 32394522 [ERROR] InnoDB: Record in index `t_idx` of table `$db_name`.`$tb_name` was not found on update: TUPLE (info_bits=0, 9 fields): {[12]121098369601(0x010201000908030609060001),[9] (0x000000000000010000),[4]KOWA(0x0B0F0701),[4]AYNA(0x01090E01),[6]STAT44(0x030401040404),[4]AYNA(0x01090E01),[1]0(0x00),[1]0(0x00),[32]8f2a39b44fe74cd781527d856342d834(0x0806020103090204040605070403040708010502070408050603040204080304)} at: COMPACT RECORD(info_bits=0, 9 fields): {[12]121098369601(0x010201000908030609060001),[9] (0x000000000000010000),[4]KOWA(0x0B0F0701),[4]AYNA(0x01090E01),NULL,NULL,[1]0(0x00),[1]0(0x00),[32]8f2a39b44fe74cd781527d856342d834(0x0806020103090204040605070403040708010502070408050603040204080304)}
  3. 2017-08-31T03:11:04.291454Z 32394522 [Note] InnoDB: GIS MBR INFO: 1.31506e-47 and 1.02964e-71, 2.8816e-306, 1.93059e+53
  4. 2017-08-31 03:11:04 0x7fcaf04be700 InnoDB: Assertion failure in thread 140509591627520 in file row0ins.cc line 282
  5. InnoDB: Failing assertion: !cursor->index->is_committed()
  6. InnoDB: We intentionally generate a memory trap.
  7. InnoDB: Submit a detailed bug report to http://bugs.mysql.com.
  8. InnoDB: If you get repeated assertion failures or crashes, even
  9. InnoDB: immediately after the mysqld startup, there may be
  10. InnoDB: corruption in the InnoDB tablespace. Please refer to
  11. InnoDB: http://dev.mysql.com/doc/refman/5.7/en/forcing-innodb-recovery.html
  12. InnoDB: about forcing recovery.
  13. 03:11:04 UTC - mysqld got signal 6 ;
  14. This could be because you hit a bug. It is also possible that this binary
  15. or one of the libraries it was linked against is corrupt, improperly built,
  16. or misconfigured. This error can also be caused by malfunctioning hardware.
  17. Attempting to collect some information that could help diagnose the problem.
  18. As this is a crash and something is definitely wrong, the information
  19. collection process might fail.
  20. …………………………………………………………………………………………………………
  21. Trying to get some variables.
  22. Some pointers may be invalid and cause the dump to abort.
  23. Query (7fca7c0dbaa0): is an invalid pointer
  24. Connection ID (thread ID): 32394522
  25. Status: NOT_KILLED
  26. The manual page at http://dev.mysql.com/doc/mysql/en/crashing.html contains
  27. information that should help you find out what is causing the crash.
  28. …………………………………………………………………………………………………………
  29. (重启中)
  30. …………………………………………………………………………………………………………
  31. 2017-08-31T03:11:08.925622Z 0 [Note] $basedir/bin/mysqld: ready for connections.
  32. Version: '5.7.12-log' socket: '$datadir/mysqld.sock' port: 3306 Source distribution
  33. 2017-08-31T03:31:10.232145Z 1704 [ERROR] InnoDB: Record in index `t_idx` of table `$db_name`.`$tb_name` was not found on update: TUPLE (info_bits=0, 9 fields): {[12]198051077411(0x010908000501000707040101),[9] 7 (0x000000000000020700),[4]AOGA(0x010F0701),[4]AQGA(0x01010701),[6]STAT44(0x030401040404),NULL,[1]0(0x00),[1]0(0x00),[32]c6f98a358ace4897a11a27d689bb6884(0x0306060908010305080103050408090701010101020704060809020206080804)} at: COMPACT RECORD(info_bits=0, 9 fields): {[12]198051077411(0x010908000501000707040101),[9] 7 (0x000000000000020700),[4]AOGA(0x010F0701),[4]AQGA(0x01010701),NULL,NULL,[1]0(0x00),[1]0(0x00),[32]c6f98a358ace4897a11a27d689bb6884(0x0306060908010305080103050408090701010101020704060809020206080804)}
  34. 2017-08-31T03:31:10.232168Z 1704 [Note] InnoDB: GIS MBR INFO: 7.26084e-43 and 1.08604e-42, 2.8823e-306, 132832
  35. 2017-08-31T03:35:51.201716Z 2208 [ERROR] InnoDB: Flagged corruption of `t_idx` in table `$db_name`.`$tb_name` in CHECK TABLE; Wrong count


初步确定为因为名为t_idx的索引损坏导致的大量报错,并在处理update语句时导致crash。
检查binlog发现的确有很多对该表的update操作。
执行一下check table,发现的确有问题:

  1. mysql> CHECK TABLE `$db_name`.`$tb_name`;
  2. +--------------------+-------+----------+-------------------------------------------------------+
  3. | Table              | Op    | Msg_type | Msg_text                                               |
  4. +--------------------+-------+----------+-------------------------------------------------------+
  5. | $db_name.$tb_name  | check | Warning  | InnoDB: Index t_idx is marked as corrupted            |
  6. | $db_name.$tb_name  | check | error    | Corrupt                                               |
  7. +--------------------+-------+----------+-------------------------------------------------------+
  8. 2 rows in set (0.83 sec)



因该库为高可用主库,检查到备库状态正常,准备先手动做failover,再对该表进行修复。

因为表小,也比较幸运,修复过程十分顺利:

  1. mysql> OPTIMIZE TABLE `$db_name`.`$tb_name`;
  2. +--------------------+----------+----------+-------------------------------------------------------------------+
  3. | Table              | Op       | Msg_type | Msg_text                                                          |
  4. +--------------------+----------+----------+-------------------------------------------------------------------+
  5. | $db_name.$tb_name  | optimize | note     | Table does not support optimize, doing recreate + analyze instead |
  6. | $db_name.$tb_name  | optimize | status   | OK                                                                |
  7. +--------------------+----------+----------+-------------------------------------------------------------------+
  8. 2 rows in set (3.42 sec)
  9. mysql> ALTER TABLE `$db_name`.`$tb_name` ENGINE=INNODB;
  10. Query OK, 0 rows affected (3.09 sec)
  11. Records: 0 Duplicates: 0 Warnings: 0
  12. mysql> ANALYZE TABLE `$db_name`.`$tb_name`;
  13. +--------------------+---------+----------+----------+
  14. | Table              | Op      | Msg_type | Msg_text |
  15. +--------------------+---------+----------+----------+
  16. | $db_name.$tb_name  | analyze | status   | OK      |
  17. +--------------------+---------+----------+----------+
  18. 1 row in set (0.00 sec)
  19. mysql> CHECK TABLE `$db_name`.`$tb_name`;
  20. +--------------------+-------+----------+----------+
  21. | Table              | Op    | Msg_type | Msg_text |
  22. +--------------------+-------+----------+----------+
  23. | $db_name.$tb_name  | check | status   | OK      |
  24. +--------------------+-------+----------+----------+
  25. 1 row in set (0.98 sec)




〇 参考文档:
关于mysqlcheck与check/analyze/optimize table等命令:
http://www.linuxidc.com/Linux/2017-08/146647.htm

比较类似的一个case被提到了bug库:
https://bugs.mysql.com/bug.php?id=82997

问题描述节选:

  1. With some random DML running I managed to hit a problem on 5.7.13.
  2. Next step for me is to test current version and make a suitable testcase.
  3. Version: '5.7.13' socket: '' port: 3306 MySQL Community Server (GPL)
  4. [ERROR] InnoDB: Record in index `ed` of table `test`.`users` was n
  5. [Note] InnoDB: GIS MBR INFO: 1.20768e-153 and 4.76881e-038, 7.0436
  6. InnoDB: Assertion failure in thread 2384 in file row0ins.cc line 282
  7. InnoDB: Failing assertion: !cursor->index->is_committed()
本文永久更新链接地址http://www.linuxidc.com/Linux/2017-09/146646.htm
linux
相关资讯       MySQL表索引  MySQL Crash 
本文评论   查看全部评论 (0)
表情: 表情 姓名: 字数

       

评论声明
  • 尊重网上道德,遵守中华人民共和国的各项有关法律法规
  • 承担一切因您的行为而直接或间接导致的民事或刑事法律责任
  • 本站管理人员有权保留或删除其管辖留言中的任意内容
  • 本站有权在网站内转载或引用您的评论
  • 参与本评论即表明您已经阅读并接受上述条款