Home > Database > Mysql Tutorial > body text

MySQL5.1 中Innodb事务完整性Bug

WBOY
Release: 2016-06-07 16:31:46
Original
1313 people have browsed it

今天和 51.com 的 MySQL DBA 景春同学一起遇到了个 MySQL 非常扯淡的Bug:在 5.1 版本中,Innodb 存储引擎如果使用autocommit=0的情况下,单条SQL在执行过程中如果异常中断的话,事务完整性可能无法保证,不论是STATEMENT还是MIXED的binlog_format,都存在相

今天和 51.com 的 MySQL DBA 景春同学一起遇到了个 MySQL 非常扯淡的Bug:在 5.1 版本中,Innodb 存储引擎如果使用autocommit=0的情况下,单条SQL在执行过程中如果异常中断的话,事务完整性可能无法保证,不论是STATEMENT还是MIXED的binlog_format,都存在相同的问题,可以重现,屡试不爽。
测试环境如下:
OS:SunOS 5.10 Generic_137138-09
DB:MySQL 5.1.31/32-log Source distribution
binlog_format:MIXED/STATEMENT
tx_isolation:REPEATABLE-READ

测试脚本如下:

[root@dc-5 /tmp]#cat deletetest.sh
#/bin/sh
mysql -uadmin -h127.0.0.1 -pxxx -e"set autocommit=0;delete from test.test;"
[root@dc-5 /tmp]#cat killtest.sh
#/bin/sh
mysqladmin -uroot processlist |grep "admin"|awk '{print $2}'|xargs mysqladmin -uroot kill

将删除的脚本放到后台执行,然后执行kill脚本:

[root@dc-5 /tmp]#time ./deletetest.sh &
[1] 6708
[root@dc-5 /tmp]#./killtest.sh
ERROR 1053 (08S01) at line 1: Server shutdown in progress

real    0m2.901s
user    0m0.007s
sys     0m0.007s
[root@dc-5 /tmp]#
[1]+  Exit 1                  time ./deletetest.sh

到Master 和 Slave 两端分别检查数据,看上去很正常:

root@dc-5 : (none)01:35:43> selectcount(*)fromtest.test;
+----------+
|
count(*) |
+----------+
|
1000000|
+----------+
1rowinset(1.71sec)
 
root@dc-6 : (none)01:36:01> selectcount(*)fromtest.test;
+----------+
|
count(*) |
+----------+
|
1000000|
+----------+
1rowinset(1.69sec)

我们再将set autocommit=0拿掉试试看,也就是允许系统自动提交:

[root@dc-5 /tmp]#cat deletetest.sh
#/bin/sh
mysql -uadmin -h127.0.0.1 -pxxx -e"delete from offer1.test;"
[root@dc-5 /tmp]#time ./deletetest.sh &
[1] 6722
[root@dc-5 /tmp]#./killtest.sh
[root@dc-5 /tmp]#ERROR 1053 (08S01) at line 1: Server shutdown in progress

real    0m2.462s
user    0m0.006s
sys     0m0.007s

[1]+  Exit 1                  time ./deletetest.sh

再检查 Master 和 Slave 两端的数据:

root@dc-5 : (none)01:40:30> selectcount(*)fromoffer1.test;
+----------+
|
count(*) |
+----------+
887377 |
+----------+
1rowinset(1.66sec)
root@dc-6 : offer101:44:05selectcount(*)fromoffer1.test;
+----------+
|
count(*) |
+----------+
|
1000000|
+----------+
1rowinset(1.70sec)

Master 和 Slave 两端数据居然出现不一致,在 Master 端已经删除掉了部分数据,在 Slave 端却没有任何变化。执行 deletetest.sh 脚本前后检查 Master 的 Binary Log(SHOW Master STATUS),没有任何变化。这个 Bug 简直是太扯淡了,数据完全没有事务完整性可言了啊。

Related labels:
source:php.cn
Statement of this Website
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn
Popular Tutorials
More>
Latest Downloads
More>
Web Effects
Website Source Code
Website Materials
Front End Template
About us Disclaimer Sitemap
php.cn:Public welfare online PHP training,Help PHP learners grow quickly!