基于GTID环境的数据恢复
下面说一个线上环境的数据恢复案例,线上环境误操作的事情,随时都有可能发生的,这时候运维DBA或者DBA会想,没人会傻到把数据库干掉的吧?又或者没有会闲得蛋痛删除了几条数据又想恢复这么无聊吧?适适这样的人才多着呢,不过,人非圣贤孰能无过,当这事情发生的时候,我们更多的是想办法去解决,以及多给开发或者新人DBA一些相关安全操作的培训。好了,废话不多说,我们来模拟一下线上数据被误操作的情形,以及是怎么把恢复的。
实验环境:
1、开启了GTID对应的选项
2、Binlog格式是row格式
针对以下的情况进行恢复:
delete from xxx (不带任何条件)
下面是表test1的表结构以及表里有几条数据:
<test>(root@localhost) [xuanzhi]> show create table test1\G *************************** 1. row *************************** Table: test1 Create Table: CREATE TABLE `test1` ( `id` int(11) NOT NULL AUTO_INCREMENT, `age` int(11) DEFAULT NULL, `name` char(10) DEFAULT NULL, PRIMARY KEY (`id`) ) ENGINE=InnoDB AUTO_INCREMENT=9 DEFAULT CHARSET=utf8 1 row in set (0.00 sec) <test>(root@localhost) [xuanzhi]> select * from test1; +----+------+------+ | id | age | name | +----+------+------+ | 5 | 20 | aa | | 6 | 22 | bb | | 7 | 23 | cc | | 8 | 24 | dd | +----+------+------+ 4 rows in set (0.00 sec) <test>(root@localhost) [xuanzhi]>
现在我们对这个库做一次完整的备份(备份的参数,这里不多说了,至少为什么要--master-data选项,后面会说,那为什么-q选项呢,可以参考博客:http://17173ops.com/2015/03/21/mysql-faq-why-turn-on-quick-option.shtml)
[root@localhost mysql-5.6.10]# mysqldump -uroot -p123456 -S /data/mysql-5.6.10/mysql.sock --single-transaction --master-data=2 -q xuanzhi > xuanzhi.sql Warning: Using a password on the command line interface can be insecure. Warning: A partial dump from a server that has GTIDs will by default include the GTIDs of all transactions, even those that changed suppressed parts of the database. If you don't want to restore GTIDs, pass --set-gtid-purged=OFF. To make a complete dump, pass --all-databases --triggers --routines --events. [root@localhost mysql-5.6.10]#
这个时候又有新数据写入了
<test>(root@localhost) [xuanzhi]> insert into test1 (age,name) values (30,'MySQL'); Query OK, 1 row affected (0.01 sec) <test>(root@localhost) [xuanzhi]> insert into test1 (age,name) values (40,'PYthon'); Query OK, 1 row affected (0.00 sec) <test>(root@localhost) [xuanzhi]> select * from test1; +----+------+--------+ | id | age | name | +----+------+--------+ | 1 | 20 | aa | | 2 | 22 | bb | | 3 | 23 | cc | | 4 | 24 | dd | | 5 | 30 | MySQL | | 6 | 40 | PYthon | +----+------+--------+ 6 rows in set (0.00 sec) <test>(root@localhost) [xuanzhi]>
可以看到多了id为5和6的这两条记录,这两条记录是在完备后产生的记录!这个时候,有人开始误操作了,delete的时候忘记带where条件了
<test>(root@localhost) [xuanzhi]> delete from test1; Query OK, 6 rows affected (0.01 sec) <test>(root@localhost) [xuanzhi]> select * from test1; Empty set (0.00 sec) <test>(root@localhost) [xuanzhi]>
可以看到很悲剧的事情发生了,表数据全没了,这时小心脏要顶住啊!!!!这时只能通过完整备份+Binlog去把数据恢复了
恢复思路 :
1、要清楚知道完备后的pos是从那里开始的,--master-data在这个时候起到了很重要的作用
2、在binlog找出delete from test1操作的上一个pos位置,预防在恢复时,又执行了delete操作,所以必须要找到delete前的pos位置
3、把之前做好的全备恢复回去
4、通过binlog基于position位置恢复
下面我们跟着思路来操作一下吧:
(1)找出备份后的position位置(在备份的时候没有使用--master-data是不会有以下信息的,--master-data的说明请参考官网):
[root@localhost ~]# cat xuanzhi.sql |grep -i "change" -- CHANGE MASTER TO MASTER_LOG_FILE='mysql-bin.000023', MASTER_LOG_POS=1393; [root@localhost ~]#
此时我们可以看到MASTER_LOG_POS=1393,说明从1393开始的binlog事件,在完备的数据里是没有的,所以我们可以会用--start-position=1393去解析binlog,这样可以减少更多的日志信息,方便找到对应的操作
[root@localhost ~]# mysqlbinlog -v --base64-output=DECODE-ROWS --start-position=1393 -d xuanzhi /data/mysql-5.6.10/mysql-bin.000023 >test1.sql
(2) 我们通过test1.sql文件找到删除操作前的position位置:
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
[root@localhost ~]# cat test1.sql /*!50530 SET @@SESSION.PSEUDO_SLAVE_MODE=1*/; /*!40019 SET @@session.max_insert_delayed_threads=0*/; /*!50003 SET @OLD_COMPLETION_TYPE=@@COMPLETION_TYPE,COMPLETION_TYPE=0*/; DELIMITER /*!*/; # at 1393 #150622 20:59:27 server id 1 end_log_pos 1441 CRC32 0x083aaf66 GTID [commit=yes] SET @@SESSION.GTID_NEXT= 'fc77e38e-c8c1-11e4-a54c-000c2914208d:6'/*!*/; # at 1441 #150622 20:59:27 server id 1 end_log_pos 1516 CRC32 0xfaaf6148 Query thread_id=1 exec_time=0 error_code=0 SET TIMESTAMP=1434977967/*!*/; SET @@session.pseudo_thread_id=1/*!*/; SET @@session.foreign_key_checks=1, @@session.sql_auto_is_null=0, @@session.unique_checks=1, @@session.autocommit=1/*!*/; SET @@session.sql_mode=1075838976/*!*/; SET @@session.auto_increment_increment=1, @@session.auto_increment_offset=1/*!*/; /*!\C utf8 *//*!*/; SET @@session.character_set_client=33,@@session.collation_connection=33,@@session.collation_server=8/*!*/; SET @@session.lc_time_names=0/*!*/; SET @@session.collation_database=DEFAULT/*!*/; BEGIN /*!*/; # at 1516 #150622 20:59:27 server id 1 end_log_pos 1571 CRC32 0x9feaa179 Table_map: `xuanzhi`.`test1` mapped to number 74 # at 1571 #150622 20:59:27 server id 1 end_log_pos 1621 CRC32 0xa761e858 Write_rows: table id 74 flags: STMT_END_F ### INSERT INTO `xuanzhi`.`test1` ### SET ### @1=5 ### @2=30 ### @3='MySQL' # at 1621 #150622 20:59:27 server id 1 end_log_pos 1652 CRC32 0x1669add4 Xid = 112 COMMIT/*!*/; # at 1652 #150622 20:59:29 server id 1 end_log_pos 1700 CRC32 0x0c99198b GTID [commit=yes] SET @@SESSION.GTID_NEXT= 'fc77e38e-c8c1-11e4-a54c-000c2914208d:7'/*!*/; # at 1700 #150622 20:59:29 server id 1 end_log_pos 1775 CRC32 0x1492f3ed Query thread_id=1 exec_time=0 error_code=0 SET TIMESTAMP=1434977969/*!*/; BEGIN /*!*/; # at 1775 #150622 20:59:29 server id 1 end_log_pos 1830 CRC32 0xea811740 Table_map: `xuanzhi`.`test1` mapped to number 74 # at 1830 #150622 20:59:29 server id 1 end_log_pos 1881 CRC32 0x8d528f67 Write_rows: table id 74 flags: STMT_END_F ### INSERT INTO `xuanzhi`.`test1` ### SET ### @1=6 ### @2=40 ### @3='PYthon' # at 1881 #150622 20:59:29 server id 1 end_log_pos 1912 CRC32 0x0d0fc5fa Xid = 113 COMMIT/*!*/; # at 1912 #150622 21:02:01 server id 1 end_log_pos 1960 CRC32 0xa68bea7e GTID [commit=yes] SET @@SESSION.GTID_NEXT= 'fc77e38e-c8c1-11e4-a54c-000c2914208d:8'/*!*/; # at 1960 #150622 21:02:01 server id 1 end_log_pos 2035 CRC32 0x4aeaf6e7 Query thread_id=1 exec_time=0 error_code=0 SET TIMESTAMP=1434978121/*!*/; BEGIN /*!*/; # at 2035 #150622 21:02:01 server id 1 end_log_pos 2090 CRC32 0x4bdb4429 Table_map: `xuanzhi`.`test1` mapped to number 74 # at 2090 #150622 21:02:01 server id 1 end_log_pos 2204 CRC32 0x32c01bf1 Delete_rows: table id 74 flags: STMT_END_F ### DELETE FROM `xuanzhi`.`test1` ### WHERE ### @1=1 ### @2=20 ### @3='aa' ### DELETE FROM `xuanzhi`.`test1` ### WHERE ### @1=2 ### @2=22 ### @3='bb' ### DELETE FROM `xuanzhi`.`test1` ### WHERE ### @1=3 ### @2=23 ### @3='cc' ### DELETE FROM `xuanzhi`.`test1` ### WHERE ### @1=4 ### @2=24 ### @3='dd' ### DELETE FROM `xuanzhi`.`test1` ### WHERE ### @1=5 ### @2=30 ### @3='MySQL' ### DELETE FROM `xuanzhi`.`test1` ### WHERE ### @1=6 ### @2=40 ### @3='PYthon' # at 2204 #150622 21:02:01 server id 1 end_log_pos 2235 CRC32 0x2d665ce6 Xid = 115 COMMIT/*!*/; # at 2235 #150622 21:38:29 server id 1 end_log_pos 2283 CRC32 0x8df83114 GTID [commit=yes] SET @@SESSION.GTID_NEXT= 'fc77e38e-c8c1-11e4-a54c-000c2914208d:9'/*!*/; # at 2283 #150622 21:38:29 server id 1 end_log_pos 2357 CRC32 0x06b2ac66 Query thread_id=5 exec_time=0 error_code=0 SET TIMESTAMP=1434980309/*!*/; BEGIN /*!*/; # at 2357 # at 2409 # at 2452 #150622 21:38:29 server id 1 end_log_pos 2483 CRC32 0x484fe32f Xid = 129 COMMIT/*!*/; # at 2483 #150622 21:38:34 server id 1 end_log_pos 2531 CRC32 0x14c0ff08 GTID [commit=yes] SET @@SESSION.GTID_NEXT= 'fc77e38e-c8c1-11e4-a54c-000c2914208d:10'/*!*/; # at 2531 #150622 21:38:34 server id 1 end_log_pos 2605 CRC32 0x47664df5 Query thread_id=5 exec_time=0 error_code=0 SET TIMESTAMP=1434980314/*!*/; BEGIN /*!*/; # at 2605 # at 2657 # at 2700 #150622 21:38:34 server id 1 end_log_pos 2731 CRC32 0x3eb63a54 Xid = 130 COMMIT/*!*/; DELIMITER ; # End of log file ROLLBACK /* added by mysqlbinlog */; /*!50003 SET COMPLETION_TYPE=@OLD_COMPLETION_TYPE*/; /*!50530 SET @@SESSION.PSEUDO_SLAVE_MODE=0*/;
可以看到删除操作前的position位置是1912:
### INSERT INTO `xuanzhi`.`test1` ### SET ### @1=6 ### @2=40 ### @3='PYthon' # at 1881 #150622 20:59:29 server id 1 end_log_pos 1912 CRC32 0x0d0fc5fa Xid = 113 COMMIT/*!*/; # at 1912 #150622 21:02:01 server id 1 end_log_pos 1960 CRC32 0xa68bea7e GTID [commit=yes] SET @@SESSION.GTID_NEXT= 'fc77e38e-c8c1-11e4-a54c-000c2914208d:8'/*!*/; # at 1960 #150622 21:02:01 server id 1 end_log_pos 2035 CRC32 0x4aeaf6e7 Query thread_id=1 exec_time=0 error_code=0 SET TIMESTAMP=1434978121/*!*/; BEGIN /*!*/; # at 2035 #150622 21:02:01 server id 1 end_log_pos 2090 CRC32 0x4bdb4429 Table_map: `xuanzhi`.`test1` mapped to number 74 # at 2090 #150622 21:02:01 server id 1 end_log_pos 2204 CRC32 0x32c01bf1 Delete_rows: table id 74 flags: STMT_END_F ### DELETE FROM `xuanzhi`.`test1`
我们再基于--start-position和--stop-position去解析binlog,确保里面已经没有delete操作了:
[root@localhost ~]# mysqlbinlog -v --base64-output=DECODE-ROWS --start-position=1393 --stop-position=1912 -d xuanzhi /data/mysql-5.6.10/mysql-bin.000023 >test1.sql
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
[root@localhost ~]# cat test1.sql /*!50530 SET @@SESSION.PSEUDO_SLAVE_MODE=1*/; /*!40019 SET @@session.max_insert_delayed_threads=0*/; /*!50003 SET @OLD_COMPLETION_TYPE=@@COMPLETION_TYPE,COMPLETION_TYPE=0*/; DELIMITER /*!*/; # at 1393 #150622 20:59:27 server id 1 end_log_pos 1441 CRC32 0x083aaf66 GTID [commit=yes] SET @@SESSION.GTID_NEXT= 'fc77e38e-c8c1-11e4-a54c-000c2914208d:6'/*!*/; # at 1441 #150622 20:59:27 server id 1 end_log_pos 1516 CRC32 0xfaaf6148 Query thread_id=1 exec_time=0 error_code=0 SET TIMESTAMP=1434977967/*!*/; SET @@session.pseudo_thread_id=1/*!*/; SET @@session.foreign_key_checks=1, @@session.sql_auto_is_null=0, @@session.unique_checks=1, @@session.autocommit=1/*!*/; SET @@session.sql_mode=1075838976/*!*/; SET @@session.auto_increment_increment=1, @@session.auto_increment_offset=1/*!*/; /*!\C utf8 *//*!*/; SET @@session.character_set_client=33,@@session.collation_connection=33,@@session.collation_server=8/*!*/; SET @@session.lc_time_names=0/*!*/; SET @@session.collation_database=DEFAULT/*!*/; BEGIN /*!*/; # at 1516 #150622 20:59:27 server id 1 end_log_pos 1571 CRC32 0x9feaa179 Table_map: `xuanzhi`.`test1` mapped to number 74 # at 1571 #150622 20:59:27 server id 1 end_log_pos 1621 CRC32 0xa761e858 Write_rows: table id 74 flags: STMT_END_F ### INSERT INTO `xuanzhi`.`test1` ### SET ### @1=5 ### @2=30 ### @3='MySQL' # at 1621 #150622 20:59:27 server id 1 end_log_pos 1652 CRC32 0x1669add4 Xid = 112 COMMIT/*!*/; # at 1652 #150622 20:59:29 server id 1 end_log_pos 1700 CRC32 0x0c99198b GTID [commit=yes] SET @@SESSION.GTID_NEXT= 'fc77e38e-c8c1-11e4-a54c-000c2914208d:7'/*!*/; # at 1700 #150622 20:59:29 server id 1 end_log_pos 1775 CRC32 0x1492f3ed Query thread_id=1 exec_time=0 error_code=0 SET TIMESTAMP=1434977969/*!*/; BEGIN /*!*/; # at 1775 #150622 20:59:29 server id 1 end_log_pos 1830 CRC32 0xea811740 Table_map: `xuanzhi`.`test1` mapped to number 74 # at 1830 #150622 20:59:29 server id 1 end_log_pos 1881 CRC32 0x8d528f67 Write_rows: table id 74 flags: STMT_END_F ### INSERT INTO `xuanzhi`.`test1` ### SET ### @1=6 ### @2=40 ### @3='PYthon' # at 1881 #150622 20:59:29 server id 1 end_log_pos 1912 CRC32 0x0d0fc5fa Xid = 113 COMMIT/*!*/; DELIMITER ; # End of log file ROLLBACK /* added by mysqlbinlog */; /*!50003 SET COMPLETION_TYPE=@OLD_COMPLETION_TYPE*/; /*!50530 SET @@SESSION.PSEUDO_SLAVE_MODE=0*/; [root@localhost ~]#
可以看到没有DELETE操作了!
(3)进行之前的备份恢复
[root@localhost ~]# mysql -uroot -p123456 -S /data/mysql-5.6.10/mysql.sock xuanzhi <./xuanzhi.sql Warning: Using a password on the command line interface can be insecure. ERROR 1840 (HY000) at line 24: GTID_PURGED can only be set when GTID_EXECUTED is empty. [root@localhost ~]#
我们查看下数据是否已经恢复回去了
<test>(root@localhost) [xuanzhi]> select * from test1; Empty set (0.00 sec)
好悲剧的看到数据没有正常恢复回去,我们回过头来看看上面的报错,GTID_PURGED can only be set when GTID_EXECUTED is empty,意味着GTID_PURGED它已经有值了,那我们可以用-f选择强制进行数据恢复
再次查看备份的数据是否恢复回来了,可以看到,备份的数据已经恢复回来了
<test>(root@localhost) [xuanzhi]> select * from test1;
+----+------+------+
| id | age | name |
+----+------+------+
| 1 | 20 | aa |
| 2 | 22 | bb |
| 3 | 23 | cc |
| 4 | 24 | dd |
+----+------+------+
4 rows in set (0.00 sec)
但备份后的数据依然没有,所以下一步要利用binlog进行数据恢复:
[root@localhost ~]# mysqlbinlog --start-position=1393 --stop-position=1912 mysql-bin.000023 | /usr/local/mysql-5.6.10/bin/mysql -uroot -p123456 -S /data/mysql-5.6.10/mysql.sock xuanzhi Warning: Using a password on the command line interface can be insecure. [root@localhost ~]#
看到没报错,你可以开心一小会了,然而当你看到结果时,会让你哭笑不得,我们查看下数据是否恢复回来了:
<test>(root@localhost) [xuanzhi]> select * from test1; +----+------+------+ | id | age | name | +----+------+------+ | 1 | 20 | aa | | 2 | 22 | bb | | 3 | 23 | cc | | 4 | 24 | dd | +----+------+------+ 4 rows in set (0.00 sec) <test>(root@localhost) [xuanzhi]>
哭了吧,完备后的数据并没有恢复回来!这时就蛋痛了,思路这么清晰,操作也步骤也没问题,数据怎么没恢复回来呢?这时唯一让我觉得有可能是因为GTID的问题,因为恢复时它可能检查到对应的GTID已经有了,它就不执行操作了
(4)进行reset master操作(切记,这只能在测试环境或者虚拟环境执行,千万不能在线上环境执行,因为它会自动清掉所有binlog,会产生一个新的binlog),当你在一个全新环境进行恢复时,就不需要做以下的reset master操作
在reset master前,一定要做好binlog备份,要不你想通过binlog再去恢复你想要的数据,也找不到了,所以我们进行拷贝一份到/root下
[root@localhost ~]# cp -a /data/mysql-5.6.10/mysql-bin.000023 /root/
再进行reset master
<test>(root@localhost) [xuanzhi]> reset master; Query OK, 0 rows affected (0.00 sec)
再进行一次binlog恢复操作:
[root@localhost ~]# mysqlbinlog --start-position=1393 --stop-position=1912 mysql-bin.000023 | /usr/local/mysql-5.6.10/bin/mysql -uroot -p123456 -S /data/mysql-5.6.10/mysql.sock xuanzhi Warning: Using a password on the command line interface can be insecure. [root@localhost ~]#
再去查看数据,是否正常恢复了:
<test>(root@localhost) [xuanzhi]> select * from test1; +----+------+--------+ | id | age | name | +----+------+--------+ | 1 | 20 | aa | | 2 | 22 | bb | | 3 | 23 | cc | | 4 | 24 | dd | | 5 | 30 | MySQL | | 6 | 40 | PYthon | +----+------+--------+ 6 rows in set (0.00 sec) <test>(root@localhost) [xuanzhi]>
嘻嘻,这时候你就可以开心笑一会了!
那么,我们如何避免这个delete from tb_name不带条件的呢?其实是有办法的,但这只针对运维DBA或者DBA在操作时候有用,但对于PHP和JAVA程序,它的连接操作方式,就没办法避免了
set sql_safe_updates=on 或者 set global sql_safe_updates=on;
<test>(root@localhost) [xuanzhi]> set sql_safe_updates=on; Query OK, 0 rows affected (0.00 sec) <test>(root@localhost) [xuanzhi]> select * from test1; Empty set (0.00 sec) <test>(root@localhost) [xuanzhi]> delete from test1; ERROR 1175 (HY000): You are using safe update mode and you tried to update a table without a WHERE that uses a KEY column <test>(root@localhost) [xuanzhi]>
可以看到没带where条件的操作是不允许的。
总结:
一、一般恢复单个表的数据或者单个表的几条数据,都建议在测试环境进行,待数据都恢复完了,再导回线上的库
二、不要用低版本的binlog命令去解析高版本的binary log,可能会出错,这个值得注意的,特别是多实例且版本不同的服务器上
二、误操作时时会有,应多加小心,恢复数据的是比较蛋痛的一件事,一方面是时间问题,另一方面是数据是否完整恢复了
三、要制定良好的备份计划,在出现特殊情况时,这就是救命稻草了,这是DBA必须要做好的一件事。
作者:陆炫志 出处:xuanzhi的博客 http://www.cnblogs.com/xuanzhi201111 您的支持是对博主最大的鼓励,感谢您的认真阅读。本文版权归作者所有,欢迎转载,但请保留该声明。
|