MySQL实时在线备份恢复方案：Replication+LVM Snapshot【上篇】
快照和复制技术的结合可以保证我们得到一个实时的在线MySQL备份解决方案
当主库发生误操作时，只需要恢复备库上的快照，然后再根据binlog执行point-in-time的恢复即可
下面假定一个场景：
主从架构，没有延迟，某DBA误操作：drop database
接下来我们按照以上场景进行备份恢复模拟测试
⑴ 主库准备测试数据
mysql> create database cnfol;
Query OK, 1 row affected (0.00 sec)
mysql> create table cnfol.t (id int primary key);
Query OK, 0 rows affected (0.02 sec)
mysql> insert into cnfol.t select 1;
Query OK, 1 row affected (0.01 sec)
Records: 1 Duplicates: 0 Warnings: 0
mysql> insert into cnfol.t select 2;
Query OK, 1 row affected (0.00 sec)
Records: 1 Duplicates: 0 Warnings: 0
到备库确认：
mysql> show databases;
+--------------------+
| Database |
+--------------------+
| information_schema |
| cnfol |
| mysql |
| test |
+--------------------+
4 rows in set (0.00 sec)
mysql> select * from cnfol.t;
+----+
| id |
+----+
| 1 |
| 2 |
+----+
2 rows in set (0.00 sec)
⑵ 加个全局读锁
在备库：
mysql> flush tables with read lock;
Query OK, 0 rows affected (0.00 sec)
⑶ 为备库所在分区创建快照
[root@localhost ~]# lvcreate --size 1G --snapshot --name backup_mysql /dev/vg/mysql
Logical volume "backup_mysql" created
[root@localhost ~]# lvs
LV VG Attr LSize Origin Snap% Move Log Copy% Convert
backup_mysql vg swi-a- 1.00G mysql 0.00
mysql vg owi-ao 2.00G
⑷ 获取二进制日志坐标
在备库：
mysql> show master status;
+------------------+----------+--------------+------------------+
| File | Position | Binlog_Do_DB | Binlog_Ignore_DB |
+------------------+----------+--------------+------------------+
| mysql-bin.000003 | 727 | | |
+------------------+----------+--------------+------------------+
1 row in set (0.00 sec)
⑸ 解锁
在备库：
mysql> unlock tables;
Query OK, 0 rows affected (0.00 sec)
⑹ 挂载快照
[root@localhost ~]# mount /dev/vg/backup_mysql /mnt/backup
[root@localhost ~]# cd /mnt/backup/mysql/data/cnfol/ && ls -alh
总计 32K
drwx------ 2 mysql dba 4.0K 10-14 09:57 .
drwx------ 5 mysql dba 4.0K 10-14 09:57 ..
-rw-rw---- 1 mysql dba 61 10-14 09:57 db.opt
-rw-rw---- 1 mysql dba 8.4K 10-14 09:57 t.frm
-rw-rw---- 1 mysql dba 14 10-14 09:57 t.MYD
-rw-rw---- 1 mysql dba 2.0K 10-14 10:06 t.MYI
⑺ 主库某无经验DBA误操作
mysql> drop database cnfol;
Query OK, 1 row affected (0.05 sec)
记录下此时时间：2013-10-14 10:17:10
备库确认是否存在库cnfol:
mysql> show databases;
+--------------------+
| Database |
+--------------------+
| information_schema |
| mysql |
| test |
+--------------------+
3 rows in set (0.01 sec)
⑻ 备份快照
[root@localhost backup]# pwd
/mnt/backup
[root@localhost backup]# tar -jcv -f /mnt/snapshot/mysql.tar.bz2 *
这里做备份的原因有2点
其一，昂贵的IO，因为磁头要在快照区和系统区来回跑
其二，快照区空间不足，因为是COW原理
⑼ 删除快照
[root@localhost ~]# umount /mnt/backup
[root@localhost ~]# lvremove --force /dev/vg/backup_mysql
Logical volume "backup_mysql" successfully removed
⑽ 格式化备库所在分区
[mysql@localhost ~]$ mysqladmin -uroot -poracle shutdown
131014 10:32:40 mysqld_safe mysqld from pid file /mnt/lvm/mysql/data/localhost.localdomain.pid ended
[1]+ Done mysqld_safe
[root@localhost ~]# umount /mnt/lvm
[root@localhost ~]# mkfs -t ext3 /dev/vg/mysql
[root@localhost ~]# mount /dev/vg/mysql /mnt/lvm
[root@localhost ~]# lvs
LV VG Attr LSize Origin Snap% Move Log Copy% Convert
mysql vg -wi-ao 2.00G
[root@localhost ~]# vgs
VG #PV #LV #SN Attr VSize VFree
vg 4 1 0 wz--n- 3.81G 1.81G
⑾ 解压缩快照到备库所在分区
# tar -jxv -f /mnt/snapshot/mysql.tar.bz2 -C /mnt/lvm/
[root@localhost lvm]# pwd
/mnt/lvm
[root@localhost lvm]# ls
lost+found mysql
⑿ 启动MySQL
⒀ 利用binlog执行point-in-time恢复
[mysql@localhost ~]$ mysqlbinlog --stop-datetime="2013-10-14 10:17:10" /mnt/lvm/mysql/data/mysql-bin.000003 | mysql -uroot -poracle
⒁ 确认数据
mysql> show databases;
+--------------------+
| Database |
+--------------------+
| information_schema |
| cnfol |
| mysql |
| test |
+--------------------+
4 rows in set (0.00 sec)
mysql> select * from cnfol.t;
+----+
| id |
+----+
| 1 |
| 2 |
+----+
2 rows in set (0.00 sec)
参考资料：
http://vbird.dic.ksu.edu.tw/linux_basic/0420quota.php#lvm_snapshot
http://www.mysqlperformanceblog.com/2006/08/21/using-lvm-for-mysql-backup-and-replication-setup/
By 迦夜
2013-10-14
Good Luck
