一、概述
binlog2sql是一个开源项目,应用于大众点评线上环境。类似于ORACLE中的闪回功能,binlog2sql可以基于时间点或者位置偏移量进行数据恢复。从MySQL binlog解析出你要的SQL。根据不同选项,你可以得到原始SQL、回滚SQL、去除主键的INSERT SQL等。也就是对于insert操作会生成对应的delete语句,反之delete操作会生出对应的insert语句,update操作会生成相反的语句。
关于binlog2sql的闪回详细介绍可参考闪回原理与实战。我也是偶然间看到一个大神关于这个神器的介绍,猛然心动,决心要动手演练一把。
我的测试环境介绍
l Python 2.6
l MySQL 5.1.73
二、binlog2sql安装
binlog2sql工具可以自己下载 https://github.com/danfengcao/binlog2sql
下面这些包都要装全,否则执行脚本会报错
python-pip
PyMySQL
python-mysql-replication
argparse
Linux机器下载并安装binlog2sql
[[email protected] binlog2sql-master]# wget https://codeload.github.com/danfengcao/binlog2sql/zip/master [[email protected] install_page]# unzip binlog2sql-master.zip Archive: binlog2sql-master.zip bb09b8f9079ca4d3cacd0186f35ddf4b3e1cfa7e creating: binlog2sql-master/ inflating: binlog2sql-master/.gitignore inflating: binlog2sql-master/LICENSE inflating: binlog2sql-master/README.md creating: binlog2sql-master/binlog2sql/ inflating: binlog2sql-master/binlog2sql/__init__.py inflating: binlog2sql-master/binlog2sql/binlog2sql.py inflating: binlog2sql-master/binlog2sql/binlog2sql_util.py creating: binlog2sql-master/example/ inflating: binlog2sql-master/example/mysql-flashback-priciple-and-practice.md inflating: binlog2sql-master/requirements.txt creating: binlog2sql-master/tests/ inflating: binlog2sql-master/tests/test_binlog2sql_util.py [[email protected] install_page]# ls binlog2sql-master binlog2sql-master.zip [[email protected] binlog2sql-master]# cd binlog2sql-master #下面脚本执行的时候也要在这么路径下 [[email protected] binlog2sql-master]# pip install -r requirements.txt Downloading/unpacking PyMySQL==0.7.11 (from -r requirements.txt (line 1)) Downloading PyMySQL-0.7.11.tar.gz (71kB): 71kB downloaded Running setup.py egg_info for package PyMySQL Downloading/unpacking wheel==0.29.0 (from -r requirements.txt (line 2)) Downloading wheel-0.29.0.tar.gz (54kB): 54kB downloaded Running setup.py egg_info for package wheel no previously-included directories found matching 'wheel/test/*/dist' no previously-included directories found matching 'wheel/test/*/build' Downloading/unpacking mysql-replication==0.13 (from -r requirements.txt (line 3)) Downloading mysql-replication-0.13.tar.gz Running setup.py egg_info for package mysql-replication Installing collected packages: PyMySQL, wheel, mysql-replication Running setup.py install for PyMySQL Running setup.py install for wheel no previously-included directories found matching 'wheel/test/*/dist' no previously-included directories found matching 'wheel/test/*/build' Installing wheel script to /usr/bin Running setup.py install for mysql-replication Successfully installed PyMySQL wheel mysql-replication Cleaning up...
三、Mysql环境要求
1、 MySQL server必须设置以下参数:
[mysqld] server-id=160 log-bin=mysql-binlog max_binlog_size=1G binlog_format=row
2、 创建一个闪回用户
[email protected] test1 19:48:06> create user [email protected]'%' identified by '123456'; Query OK, 0 rows affected (0.00 sec) [email protected] test1 19:49:06>grant select,replication slave,replication client on *.* to t[email protected]'%'; Query OK, 0 rows affected (0.00 sec) [email protected] test1 19:49:50>flush privileges; Query OK, 0 rows affected (0.00 sec)
注:user需要的最小权限集合:
select, super/replication client, replication slave
权限说明
select:需要读取server端information_schema.COLUMNS表,获取表结构的元信息,拼接成可视化的sql语句 super/replication client:两个权限都可以,需要执行'SHOW MASTER STATUS', 获取server端的binlog列表 replication slave:通过BINLOG_DUMP协议获取binlog内容的权限
3、 模拟一次生产事故,误删数据
test1库tb1表原有数据
[email protected] test1 20:08:52>select * from tb1; +-------+------+ | name | age | +-------+------+ | kobe | 21 | | james | 22 | | jack | 23 | | mike | 24 | | bob | 25 | +-------+------+ 5 rows in set (0.01 sec) [email protected] test1 20:08:59>delete from tb1 where age <23; Query OK, 2 rows affected (0.00 sec) [email protected] test1 20:09:03>select * from tb1; +-------+------+ | name | age | +-------+------+ | jack | 23 | | mike | 24 | | bob | 25 | +-------+------+ 3 rows in set (0.01 sec)
四、恢复数据步骤
1、登录mysql,查看目前的binlog文件
[email protected] test1 20:09:59>show master status; +---------------------+----------+--------------+------------------+ | File | Position | Binlog_Do_DB | Binlog_Ignore_DB | +---------------------+----------+--------------+------------------+ | mysql-binlog.000002 | 341 | | | +---------------------+----------+--------------+------------------+ 1 row in set (0.00 sec)
最新的binlog文件是mysql-binlog.000002,我们再定位误操作SQL的binlog位置。误操作人只能知道大致的误操作时间,我们根据大致时间过滤数据。
2、 接下来就该这个神器登场了。
先来介绍一下binlog2sql参数
--stop-never 持续同步binlog。可选。不加则同步至执行命令时最新的binlog位置。 -K, --no-primary-key 对INSERT语句去除主键。可选。 -B, --flashback 生成回滚语句,可解析大文件,不受内存限制,每打印一千行加一句SELECT SLEEP(1)。可选。与stop-never或no-primary-key不能同时添加。 --start-file 起始解析文件。必须。 --start-position/--start-pos start-file的起始解析位置。可选。默认为start-file的起始位置。 --stop-file/--end-file 末尾解析文件。可选。默认为start-file同一个文件。若解析模式为stop-never,此选项失效。 --stop-position/--end-pos stop-file的末尾解析位置。可选。默认为stop-file的最末位置;若解析模式为stop-never,此选项失效。 --start-datetime 从哪个时间点的binlog开始解析,格式必须为datetime,如'2016-11-11 11:11:11'。可选。默认不过滤。 --stop-datetime 到哪个时间点的binlog停止解析,格式必须为datetime,如'2016-11-11 11:11:11'。可选。默认不过滤。 -d, --databases 只输出目标db的sql。可选。默认为空。 -t, --tables 只输出目标tables的sql。可选。默认为空。
3、 根据预估时间,执行下面命令找出对应的position
[[email protected] binlog2sql]# python binlog2sql.py -h 192.168.221.160 -utest -p123456 -dtest1 -ttb1 --start-file='mysql-binlog.000002' --start-datetime='2017-12-04 20:00:00' --stop-datetime='2017-12-04 20:10:00' DELETE FROM `test1`.`tb1` WHERE `age`=21 AND `name`='kobe' LIMIT 1; #start 4 end 271 time 2017-12-04 20:08:59 DELETE FROM `test1`.`tb1` WHERE `age`=22 AND `name`='james' LIMIT 1; #start 4 end 271 time 2017-12-04 20:08:59
我们得到了误操作sql的准确位置在4-271之间,再根据位置进一步过滤,使用flashback模式生成回滚sql,检查回滚sql是否正确(注:真实环境下,此步经常会进一步筛选出需要的sql。结合grep、编辑器等)
4、 使用flashback模式生成回滚sql
[[email protected] binlog2sql]# python binlog2sql.py -h 192.168.221.160 -utest -p123456 -dtest1 -ttb1 --start-file='mysql-binlog.000002' --start-position=4 --stop-position=271 -B > tb1_rollback.sql
查看闪回导出的文件
[[email protected] binlog2sql]# cat tb1_rollback.sql INSERT INTO `test1`.`tb1`(`age`, `name`) VALUES (22, 'james'); #start 4 end 271 time 2017-12-04 20:08:59 INSERT INTO `test1`.`tb1`(`age`, `name`) VALUES (21, 'kobe'); #start 4 end 271 time 2017-12-04 20:08:59
5、 确认回滚sql正确,执行回滚语句。登录mysql确认,数据回滚成功。
[[email protected] binlog2sql]# mysql -uroot test1 -p123456 <tb1_rollback.sql Enter password:
6、登录数据库检验数据完整性
[email protected] test1 20:18:04>select * from tb1; +-------+------+ | name | age | +-------+------+ | kobe | 21 | | james | 22 | | jack | 23 | | mike | 24 | | bob | 25 | +-------+------+ 5 rows in set (0.00 sec)
可以看到,之前删除的两条数据又回来了
五、结语
binlog2sql是属于纯Python开发,安装与使用都很简单,易于上手,可操作性强,解析为标准SQL,方便理解、调试。但仍存在一些缺点,比如只能在mysql开启的状态下运行,不能离线操作,数据量较大时会暴露出解析速度慢的问题。总体来说,仍不失成为一个很NICE的工具。