Recover InnoDB dictionary

为什么我们需要恢复innodb的字典信息？当我们drop 一个表时，发现误操作，这时又没有备份，那么想恢复数据是非常困难的。所以我们想恢复被删除的表时，首先就需要恢复表结构，目前已经有了undrop-for-innodb这个工具，相比之前的工具percona-data-recovery-tool-for-innodb有了更大的进步，这两个工具都是同一个作者开发的。这个工具我以前也写了相关的博客--Percona Data Recovery Tool 单表恢复，现在undrop-for-innodb这个工具更加完美，少了很多限制。但是对于drop table以后表结构的恢复还是需要使用percona-data-recovery-tool-for-innodb。没有看见undrop-for-innodb可以恢复表结构，这也是比较遗憾的地方吧。

1.TwinDB恢复工具包编译安装

安装工具包依赖的软件

yum install gcc bison flex bzr -y

2.下载软件包并编译安装

[root@mysql-server- ~]# bzr branch lp:undrop-for-innodb
You have not informed bzr of your Launchpad ID, and you must do this to
write to Launchpad or access private data.  See "bzr help launchpad-login".
Branched  revision(s).
[root@mysql-server- ~]# 

[root@mysql-server- ~]# cd undrop-for-innodb/
[root@mysql-server- undrop-for-innodb]# make
cc -g -O3 -I./include -c stream_parser.c
cc -g -O3 -I./include  -pthread -lm stream_parser.o -o stream_parser
flex  sql_parser.l
bison  -o sql_parser.c sql_parser.y
sql_parser.y: conflicts:  shift/reduce
cc -g -O3 -I./include -c sql_parser.c
cc -g -O3 -I./include -c c_parser.c
cc -g -O3 -I./include -c tables_dict.c
cc -g -O3 -I./include -c print_data.c
cc -g -O3 -I./include -c check_data.c
cc -g -O3 -I./include  sql_parser.o c_parser.o tables_dict.o print_data.o check_data.o -o c_parser -pthread -lm
cc -g -O3 -I./include -o innochecksum_changer innochecksum.c
[root@mysql-server- undrop-for-innodb]# 

如果没有抛出错误，证明就ok了，会生成编译好的工具c_parser。

一. 解析 ibdata1

InnoDB的字典信息存储在ibdata1。因此，我们需要使用stream_parser工具对它进行解析并获得存储在字典中的记录页。该工具在make以后就自动生成了。

[root@mysql-server- undrop-for-innodb]# ./stream_parser -f /data/mysql/user_3306/data/ibdata1
Opening file: /data/mysql/user_3306/data/ibdata1
File information:
 
ID of device containing file:
inode number:
protection:                          (regular file)
number of hard links:
user ID of owner:
group ID of owner:
device ID (if special file):
blocksize for filesystem I/O:
number of blocks allocated:
time of last access:             Sun Aug   ::
time of last modification:       Sun Aug   ::
time of last status change:      Sun Aug   ::
total size, in bytes:              (26.000 MiB)
 
Size to process:                   (26.000 MiB)
All workers finished in  sec
[root@mysql-server- undrop-for-innodb]# 

stream_parser这个工具发现在ibdata1中，找到了存储页面类型（FIL_PAGE_INDEX或FIL_PAGE_TYPE_BLOB）由index_id的整理InnoDB的页面。

[root@mysql-server- pages-ibdata1]# pwd
/root/undrop-for-innodb/pages-ibdata1
[root@mysql-server- pages-ibdata1]# ll
total
drwxr-xr-x  root root  Aug   : FIL_PAGE_INDEX
drwxr-xr-x  root root  Aug   : FIL_PAGE_TYPE_BLOB
[root@mysql-server- pages-ibdata1]# 

SYS_TABLES

SYS_TABLES本身是一个表，该表用于规范innodb各种表定义和保存innodb中各种表的基本信息，结合SYS_COLUMNS，SYS_INDEXES和SYS_FOREIGN系统表定义了特定某个表的所有信息。

[root@mysql-server- undrop-for-innodb]# ll pages-ibdata1/FIL_PAGE_INDEX/.page
-rw-r--r--  root root  Aug   : pages-ibdata1/FIL_PAGE_INDEX/.page

SYS_INDEXES

SYS_INDEXES用于保存innodb中每个表定义的每个索引对象。

[root@mysql-server- undrop-for-innodb]# ll pages-ibdata1/FIL_PAGE_INDEX/.page
-rw-r--r--  root root  Aug   : pages-ibdata1/FIL_PAGE_INDEX/.page

SYS_COLUMNS

SYS_COLUMNS用于保存innodb引擎每个表定义的列，与SYS_TABLES相似。

[root@mysql-server- undrop-for-innodb]# ll pages-ibdata1/FIL_PAGE_INDEX/.page
-rw-r--r--  root root  Aug   : pages-ibdata1/FIL_PAGE_INDEX/.page

SYS_FIELDS

SYS_FIELDS用于保存innodb中每个索引的每个列对象。

[root@mysql-server- undrop-for-innodb]# ll pages-ibdata1/FIL_PAGE_INDEX/.page
-rw-r--r--  root root  Aug   : pages-ibdata1/FIL_PAGE_INDEX/.page

从SYS_TABLES和SYS_INDEXES导出记录 

获取记录了需要使用c_parser工具。但首先我们要创建转存的数据目录

[root@mysql-server- undrop-for-innodb]# mkdir -p dumps/default
[root@mysql-server- undrop-for-innodb]# 

InnoDB的字典总是在冗余格式，因此选择-4是必需的：

[root@mysql-server- undrop-for-innodb]# ./c_parser -4f pages-ibdata1/FIL_PAGE_INDEX/.page -t dictionary/SYS_TABLES.sql > dumps/default/SYS_TABLES > dumps/default/SYS_TABLES.sql
[root@mysql-server- undrop-for-innodb]# 

现在找到我们关心的数据库，我之前有导入MySQL官方的示例数据库，sakila

[root@mysql-server- undrop-for-innodb]# grep sakila dumps/default/SYS_TABLES
      SYS_TABLES      "sakila/actor"                                    ""
      SYS_TABLES      "sakila/address"                                          ""
      SYS_TABLES      "sakila/category"                                         ""
    960000013A0110  SYS_TABLES      "sakila/city"                                     ""
    970000013D0110  SYS_TABLES      "sakila/country"                                          ""
      SYS_TABLES      "sakila/customer"                                         ""
    9A000001460110  SYS_TABLES      "sakila/film"                                    ""
00000000031B    9C0000014C0110  SYS_TABLES      "sakila/film\_actor"                                      ""
00000000031D    9E000001500110  SYS_TABLES      "sakila/film\_category"                                   ""
00000000031F    A0000001540110  SYS_TABLES      "sakila/inventory"                                        ""
    A2000001590110  SYS_TABLES      "sakila/language"                                         ""
    A30000015B0110  SYS_TABLES      "sakila/payment"                                          ""
    A6000001620110  SYS_TABLES      "sakila/rental"                                   ""
    A90000016A0110  SYS_TABLES      "sakila/staff"                                   ""
00000000032A    AB0000016F0110  SYS_TABLES      "sakila/store"                                    ""
[root@mysql-server- undrop-for-innodb]# 

可见该库下的所有表都看见了（用过sakila示例库的同学都知道^_^）。

dumps/default/SYS_TABLES使使用LOAD DATA INFILE命令生成的。具体的命令c_parsers打印到标准错误输出。把改SQL保存在了dumps/default/SYS_TABLES.sql

[root@mysql-server- undrop-for-innodb]# cat dumps/default/SYS_TABLES.sql
SET FOREIGN_KEY_CHECKS=;
LOAD DATA LOCAL INFILE '/root/undrop-for-innodb/dumps/default/SYS_TABLES' REPLACE INTO TABLE `SYS_TABLES` FIELDS TERMINATED BY '\t' OPTIONALLY ENCLOSED BY '"' LINES STARTING BY 'SYS_TABLES\t' (`NAME`, `ID`, `N_COLS`, `TYPE`, `MIX_ID`, `MIX_LEN`, `CLUSTER_NAME`, `SPACE`);

用同样的方法导出SYS_INDEXES：

[root@mysql-server- undrop-for-innodb]# ./c_parser -4f pages-ibdata1/FIL_PAGE_INDEX/.page -t dictionary/SYS_INDEXES.sql > dumps/default/SYS_INDEXES > dumps/default/SYS_INDEXES.sql
[root@mysql-server- undrop-for-innodb]# 

查看一下结果，确保没有异常。

[root@mysql-server- undrop-for-innodb]# head - dumps/default/SYS_INDEXES
-- Page id: , Format: REDUNDANT, Records list: Valid, Expected records: ( )
    810000012D0177  SYS_INDEXES                 "ID\_IND"
    810000012D01A5  SYS_INDEXES                 "FOR\_IND"
    810000012D01D3  SYS_INDEXES                 "REF\_IND"
    810000012D026D  SYS_INDEXES                 "ID\_IND"
[root@mysql-server- undrop-for-innodb]# head - dumps/default/SYS_INDEXES.sql
SET FOREIGN_KEY_CHECKS=;
LOAD DATA LOCAL INFILE '/root/undrop-for-innodb/dumps/default/SYS_INDEXES' REPLACE INTO TABLE `SYS_INDEXES` FIELDS TERMINATED BY '\t' OPTIONALLY ENCLOSED BY '"' LINES STARTING BY 'SYS_INDEXES\t' (`TABLE_ID`, `ID`, `NAME`, `N_FIELDS`, `TYPE`, `SPACE`, `PAGE_NO`);
[root@mysql-server- undrop-for-innodb]# 

现在我们可以使用字典了，如果表在数据库中，那么会更加方便。
加载字典表到MySQL

通过grep可以从SYS_TABLES和SYS_INDEXES的主要用途是得到为了表名index_id。如果SYS_TABLES和SYS_INDEXES在MySQL中，那么操作将变得更简单。

root@localhost : (none) 18:10:31> create database yayun;
Query OK, 1 row affected (0.00 sec)
 
root@localhost : (none) 18:10:42> 

上面创建的库主要用来导入两张表而已。

[root@mysql-server- undrop-for-innodb]# mysql -uroot -p123456 -S /data/mysql/user_3306/mysql.sock yayun < dictionary/SYS_TABLES.sql
[root@mysql-server- undrop-for-innodb]# mysql -uroot -p123456 -S /data/mysql/user_3306/mysql.sock yayun < dictionary/SYS_INDEXES.sql
[root@mysql-server- undrop-for-innodb]# 

继续导入数据

[root@mysql-server- undrop-for-innodb]# mysql -uroot -p123456 -S /data/mysql/user_3306/mysql.sock yayun < dumps/default/SYS_TABLES.sql
[root@mysql-server- undrop-for-innodb]# mysql -uroot -p123456 -S /data/mysql/user_3306/mysql.sock yayun < dumps/default/SYS_INDEXES.sql
[root@mysql-server- undrop-for-innodb]# 

现在我们的InnoDB字典信息已经存在在MySQL中，我们可以像普通的表进行查询

root@localhost : (none) 20:56:44> use yayun
Database changed
root@localhost : yayun 20:56:48> SELECT * FROM SYS_TABLES WHERE NAME = 'sakila/actor';
+--------------+----+--------+------+--------+---------+--------------+-------+
| NAME         | ID | N_COLS | TYPE | MIX_ID | MIX_LEN | CLUSTER_NAME | SPACE |
+--------------+----+--------+------+--------+---------+--------------+-------+
| sakila/actor | 13 |      4 |    1 |      0 |       0 |              |     0 |
+--------------+----+--------+------+--------+---------+--------------+-------+
1 row in set (0.08 sec)
 
root@localhost : yayun 20:56:51> SELECT * FROM SYS_INDEXES WHERE TABLE_ID = 13;
+----------+----+---------------------+----------+------+-------+---------+
| TABLE_ID | ID | NAME                | N_FIELDS | TYPE | SPACE | PAGE_NO |
+----------+----+---------------------+----------+------+-------+---------+
|       13 | 15 | PRIMARY             |        1 |    3 |     0 |     307 |
|       13 | 16 | idx_actor_last_name |        1 |    0 |     0 |     308 |
+----------+----+---------------------+----------+------+-------+---------+
2 rows in set (0.05 sec)
 
root@localhost : yayun 20:58:14> 

通过上述的查询，我们可以看到sakila.actor表有两个索引，PRIMARY和idx_actor_last_name，分别对应15,16,。如果我们不小心误操作，drop table actor，那么恢复数据就需要知道相应的id。后续文章将介绍drop table后如何恢复数据。

参考资料

https://twindb.com/how-to-recover-innodb-dictionary/

https://twindb.com/innodb-dictionary/