没有备份怎么恢复被drop的表(利用undrop-for-innodb)

介绍:

    也许大家都难以理解,这么重要的数据为啥不备份(或者备份不可用)?而且还任性的drop table了。显然有备份是最好的，但是它们并不总是可用的。这种情况令人恐惧，但并非毫无希望。在许多情况下，可以恢复数据库或表中的几乎所有数据。恢复计划取决于InnoDB是否将所有数据保存在一个ibdata1中，还是每个表都有自己的表空间。本文将考虑innodb_file_per_table=OFF的情况。

drop 表恢复其他方法:

• 利用master同步
• 伪master+Binlog+同步
• 利用binlog改为realy-log在临时slave更新

本文重要部分:

• 表结构恢复
• 表数据恢复
• 未解决的问题

环境:

• 时间:#Sat Aug 4 19:37:24 CST 2018
• CentOS Linux release 7.2.1511 (Core)
• MySQL:5.7.23-log
• 使用工具:undrop-for-innodb

1.工具安装

• 依赖包安装

yum install -y make gcc flex bison

• 下载工具包(github)

#cd /opt/
[root@db13_19:55:25 /opt]  
#git clone https://github.com/twindb/undrop-for-innodb.git
Cloning into 'undrop-for-innodb'...
Resolving deltas: 100% (77/77), done.

• make

#cd /opt/undrop-for-innodb
#make
....
 
....
[root@db13_20:39:43 /opt/undrop-for-innodb]  
#ll
total 2920
-rw-r--r-- 1 root root    6271 Aug  4 19:55 check_data.c
-rw-r--r-- 1 root root   66128 Aug  4 20:39 check_data.o
-rwxr-xr-x 1 root root  727801 Aug  4 20:39 c_parser
-rw-r--r-- 1 root root   28587 Aug  4 19:55 c_parser.c
-rw-r--r-- 1 root root 1030296 Aug  4 20:39 c_parser.o
drwxr-xr-x 2 root root      92 Aug  4 19:55 dictionary
-rw-r--r-- 1 root root    1978 Aug  4 19:55 fetch_data.sh
drwxr-xr-x 2 root root    4096 Aug  4 19:55 include
-rw-r--r-- 1 root root    8936 Aug  4 19:55 innochecksum.c
-rwxr-xr-x 1 root root   36343 Aug  4 20:39 innochecksum_changer
-rw-r--r-- 1 root root  154459 Aug  4 20:39 lex.yy.c
-rw-r--r-- 1 root root   18047 Aug  4 19:55 LICENSE
-rw-r--r-- 1 root root    1942 Aug  4 19:55 Makefile
-rw-r--r-- 1 root root   16585 Aug  4 19:55 print_data.c
-rw-r--r-- 1 root root  127176 Aug  4 20:39 print_data.o
-rw-r--r-- 1 root root    3464 Aug  4 19:55 README.md
-rwxr-xr-x 1 root root    1536 Aug  4 19:55 recover_dictionary.sh
drwxr-xr-x 2 root root    4096 Aug  4 19:55 sakila
-rw-r--r-- 1 root root  103506 Aug  4 20:39 sql_parser.c
-rw-r--r-- 1 root root    8462 Aug  4 19:55 sql_parser.l
-rw-r--r-- 1 root root  296840 Aug  4 20:39 sql_parser.o
-rw-r--r-- 1 root root   26355 Aug  4 19:55 sql_parser.y
-rwxr-xr-x 1 root root   61725 Aug  4 20:39 stream_parser
-rw-r--r-- 1 root root   23103 Aug  4 19:55 stream_parser.c
-rw-r--r-- 1 root root  109304 Aug  4 20:39 stream_parser.o
-rw-r--r-- 1 root root   14764 Aug  4 19:55 sys_parser.c
-rw-r--r-- 1 root root    2182 Aug  4 19:55 tables_dict.c
-rw-r--r-- 1 root root   40264 Aug  4 20:39 tables_dict.o
-rwxr-xr-x 1 root root    6629 Aug  4 19:55 test.sh
drwxr-xr-x 3 root root      42 Aug  4 19:55 vagrant
[root@db13_20:39:57 /opt/undrop-for-innodb]  

• 增加用于恢复表结构的工具sys_parser(官方文档未使用):

#gcc `/usr/local/mysql57/bin/mysql_config --cflags` `/usr/local/mysql57/bin/mysql_config --libs` -o sys_parser sys_parser.c
 注:mysql_basedir: /usr/local/mysql57/bin/ 

2.表数据生成和drop表

• 初始化一个新实例:

[root@db212_20:58:44 /data/57mysql]  
#mkdir mysql3507/{data,logs,tmp} -p
#chown -R mysql:mysql mysql3507
[root@db212_21:07:34 /3507]  
\\复制my3507.cnf到mysql3507下
#ln -s /data/57mysql/mysql3507/ /3507
#/usr/local/mysql57/bin/mysqld --defaults-file=/3507/my3507.cnf --initialize-insecure
#/usr/local/mysql57/bin/mysqld --defaults-file=/3507/my3507.cnf&
[1] 11669
//取初始密码并登录:
/usr/local/mysql57/bin/mysql -S /tmp/mysql3507.sock -uroot 

(unknown)@localhost [(none)]>alter user user() identified by '*****';
root@localhost [(none)]>CREATE DATABASE wenyz;
root@localhost [(none)]>use wenyz;
Database changed
//创建表
root@localhost [wenyz]>CREATE TABLE `t2` (
    ->   `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
    ->   `ti` varchar(100) NOT NULL,
    ->   `date` date DEFAULT NULL,
    ->   PRIMARY KEY (`id`)
    -> ) ENGINE=InnoDB AUTO_INCREMENT=4079859 DEFAULT CHARSET=utf8;
Query OK, 0 rows affected (0.01 sec)
//造数据
root@localhost [wenyz]>insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
Records: 448  Duplicates: 0  Warnings: 0
 
//查看数据行数和checksum值
root@localhost [wenyz]>select count(*) from t2;
+----------+
| count(*) |
+----------+
|      896 |
+----------+
root@localhost [wenyz]>checksum table t2;
+----------+------------+
| Table    | Checksum   |
+----------+------------+
| wenyz.t2 | 3458542072 |
+----------+------------+
 
//DROP 表
root@localhost [wenyz]>drop table t2;
Query OK, 0 rows affected (0.01 sec)

3.利用stream_parser将ibdata1文件导出成page文件

#cd /opt/undrop-for-innodb/
[root@db212_21:25:52 /opt/undrop-for-innodb]  
#./stream_parser -f /3507/data/ibdata1
Opening file: /3507/data/ibdata1
File information:
....
Size to process:                 104857600 (100.000 MiB)
Size to process:                 104857600 (100.000 MiB)
time of last access:            1533388916 Sat Aug  4 21:21:56 2018
time of last modification:      1533388917 Sat Aug  4 21:21:57 2018
time of last status change:     1533388917 Sat Aug  4 21:21:57 2018
total size, in bytes:            104857600 (100.000 MiB)
 
Size to process:                 104857600 (100.000 MiB)
All workers finished in 0 sec

4.恢复表结构 [ Top ]

这里引入官方的一段描述:

InnoDB stores all data in B+tree indexes. A table has one clustered index PRIMARY, all fields are stored there. Thus, if the table has secondary keys, each key has an index. Each index is identified by index_id.

Consequently, if we want to recover a table, we have to find all pages that belong to a particular index_id.

• 接下来我们先来看看怎么找到table id和INDEX_ID(page文件编号)的，理解原理后再用程序跑一次，导入到临时数据库中

• 手工查找table id
  观察以下结果中,wenyz/t2后的40,即为table id

[root@db212_21:25:55 /opt/undrop-for-innodb]  
#./c_parser -4Df pages-ibdata1/FIL_PAGE_INDEX/0000000000000001.page -t dictionary/SYS_TABLES.sql | grep 'wenyz/t2' 
000000000521    3B00000149047E  SYS_TABLES  "wenyz/t2"  40  3   33  0   64  ""  0
SET FOREIGN_KEY_CHECKS=0;
LOAD DATA LOCAL INFILE '000000000521    3B00000149047E  SYS_TABLES  "wenyz/t2"  40  3   33  0   64  ""  0

• 通过table id查看page文件编号
  观察以下结果中40后的41,即为INDEX_ID(page文件编号)

[root@db212_21:29:54 /opt/undrop-for-innodb]  
#./c_parser -4Df pages-ibdata1/FIL_PAGE_INDEX/0000000000000003.page -t dictionary/SYS_INDEXES.sql | grep '40'
000000000521    3B0000014903A2  SYS_INDEXES 40  41  "PRIMARY"   1   3   0   4294967295
SET FOREIGN_KEY_CHECKS=0;
000000000521    3B0000014903A2  SYS_INDEXES 40  41  "PRIMARY"   1   3   0   4294967295

• 用程序恢复字典信息:编辑mysql登录信息.

[root@db212_21:57:04 /opt/undrop-for-innodb]  
 
//将文件中三处mysql替换成/usr/local/mysql57/bin/mysql --login-path=p3507
#vi recover_dictionary.sh 
 43 /usr/local/mysql57/bin/mysql --login-path=p3507 -e "CREATE DATABASE IF NOT EXISTS test"
...
 50 /usr/local/mysql57/bin/mysql --login-path=p3507 test < dictionary/$t.sql
...
 58 /usr/local/mysql57/bin/mysql --login-path=p3507 test < dumps/default/$t.sql 

• 执行/recover_dictionary.sh,恢复字典信息

[root@db212_22:13:57 /opt/undrop-for-innodb] 
#./recover_dictionary.sh 
Generating dictionary tables dumps... OK
Creating test database ... OK
Creating dictionary tables in database test:
SYS_TABLES ... OK
SYS_COLUMNS ... OK
SYS_INDEXES ... OK
SYS_FIELDS ... OK
All OK
Loading dictionary tables data:
SYS_TABLES ... 52 recs OK
SYS_COLUMNS ... 284 recs OK
SYS_INDEXES ... 68 recs OK
SYS_FIELDS ... 90 recs OK
All OK
[root@db212_22:14:02 /opt/undrop-for-innodb]  

• 登录mysql查看信息字典信息:

mysql --login-path=p3507
root@localhost [(none)]>use test;
Database changed
root@localhost [test]>select * from SYS_TABLES where name like 'wenyz/t2%';
+----------+----+--------+------+--------+---------+--------------+-------+
| NAME     | ID | N_COLS | TYPE | MIX_ID | MIX_LEN | CLUSTER_NAME | SPACE |
+----------+----+--------+------+--------+---------+--------------+-------+
| wenyz/t2 | 40 |      3 |   33 |      0 |      64 |              |     0 |
+----------+----+--------+------+--------+---------+--------------+-------+
1 row in set (0.00 sec)
 
root@localhost [test]>select * from SYS_INDEXES where table_id=40;
+----------+----+---------+----------+------+-------+------------+
| TABLE_ID | ID | NAME    | N_FIELDS | TYPE | SPACE | PAGE_NO    |
+----------+----+---------+----------+------+-------+------------+
|       40 | 41 | PRIMARY |        1 |    3 |     0 | 4294967295 |
+----------+----+---------+----------+------+-------+------------+
//注意记录上表中id,此ID为INDEX_ID(page文件编号)等会表数据恢复要使用
1 row in set (0.00 sec)
 
root@localhost [test]>

• 表恢复t2表结构 [ Top ]

./sys_parser -h 127.0.0.1 -u root -p xxxx -d test wenyz/t2
./sys_parser: error while loading shared libraries: libmysqlclient.so.20: cannot open shared object file: No such file or directory
#ln -s /opt/mysql-5.7.23-linux-glibc2.12-x86_64/lib/libmysqlclient.so.20 /usr/lib64/libmysqlclient.so.20 
[root@db212_22:23:25 /opt/undrop-for-innodb]  
#./sys_parser -h 127.0.0.1 -u root -p zstzst -d test wenyz/t2
CREATE TABLE `t2`(
    `id` INT UNSIGNED NOT NULL,
    `ti` VARCHAR(100) CHARACTER SET 'utf8' COLLATE 'utf8_general_ci' NOT NULL,
    `date` DATE,
    PRIMARY KEY (`id`)
) ENGINE=InnoDB;
[root@db212_22:23:30 /opt/undrop-for-innodb]  

• 将恢复的表结构存到/tmp/t2.sql

#cat /tmp/t2.sql
CREATE TABLE `t2`(
    `id` INT UNSIGNED NOT NULL,
    `ti` VARCHAR(100) CHARACTER SET 'utf8' COLLATE 'utf8_general_ci' NOT NULL,
    `date` DATE,
    PRIMARY KEY (`id`)
) ENGINE=InnoDB;
[root@db212_22:26:20 /opt/undrop-for-innodb]  

5.表数据恢复 [ Top ]

• 查看数据是否存在

以下命令中使用的0000000000000041.page为上面提到的INDEX_ID(page文件编号),/tmp/t2.sql为上面恢复的表结构.

[root@db212_22:33:44 /opt/undrop-for-innodb]  
#./c_parser -6f pages-ibdata1/FIL_PAGE_INDEX/0000000000000041.page -t /tmp/t2.sql |head -2
-- Page id: 459, Format: COMPACT, Records list: Valid, Expected records: (180 180)
000000000507    A7000001210110  t2  4079859 "d553635af1a3b" "2018-08-04"
000000000508    A8000001230110  t2  4079860 "44d64b99fc30d1b"   "2018-08-04"
 
[root@db212_22:33:44 /opt/undrop-for-innodb]  

• 利用c_parser将0000000000000041.page导出成可执行sql

//注意:此处几个文件名程序把导出的两个数据文件的文件名关系是写死的,以下dumps/default/t2中的t2是需要和表名一致,在t2_load.sql中会引用此文件路经.
./c_parser -6f pages-ibdata1/FIL_PAGE_INDEX/0000000000000041.page -t /tmp/t2.sql > dumps/default/t2 2> dumps/default/t2_load.sql  
#cd dumps/default/
[root@db212_22:41:01 /opt/undrop-for-innodb/dumps/default]  
#ll
total 132
-rw-r--r-- 1 root root 21232 Aug  4 22:14 SYS_COLUMNS
...
-rw-r--r-- 1 root root 62923 Aug  4 22:40 t2
-rw-r--r-- 1 root root   308 Aug  4 22:40 t2_load.sql
[root@db212_22:41:04 /opt/undrop-for-innodb/dumps/default]  
#cat t2_load.sql 
SET FOREIGN_KEY_CHECKS=0;
LOAD DATA LOCAL INFILE '/opt/undrop-for-innodb/dumps/default/t2' REPLACE INTO TABLE `t2` FIELDS TERMINATED BY '\t' OPTIONALLY ENCLOSED BY '"' LINES STARTING BY 't2\t' (`id`, `ti`, `date`);
-- STATUS {"records_expected": 896, "records_dumped": 896, "records_lost": false} STATUS END

• 将表结构t2.sql和表数据t2和t2_load.sql导入数据库

root@localhost [test]>source /tmp/t2.sql
Query OK, 0 rows affected (0.02 sec)
 
root@localhost [test]>source /opt/undrop-for-innodb/dumps/default/t2_load.sql
Query OK, 0 rows affected (0.00 sec)
 
Query OK, 896 rows affected (0.01 sec)
Records: 896  Deleted: 0  Skipped: 0  Warnings: 0
 
root@localhost [test]>select count(*) from t2;
+----------+
| count(*) |
+----------+
|      896 |
+----------+
1 row in set (0.00 sec)
 
root@localhost [test]>checksum table t2;
+---------+------------+
| Table   | Checksum   |
+---------+------------+
| test.t2 | 3458542072 |
+---------+------------+
1 row in set (0.00 sec)
 
root@localhost [test]>
 
//drop前数据信息对比:
root@localhost [wenyz]>select count(*) from t2;
+----------+
| count(*) |
+----------+
|      896 |
+----------+
1 row in set (0.00 sec)
 
root@localhost [wenyz]>checksum table t2;
+----------+------------+
| Table    | Checksum   |
+----------+------------+
| wenyz.t2 | 3458542072 |
+----------+------------+
1 row in set (0.00 sec)

• 至此数据已经完全恢复.

6.未解决的问题 [ Top ]

问题1.
innodb(非独立表空间)情况,drop表后,用工具读ibdata1对应数据页文件,如果是这个表大于350行左右的数据,页文件是存在的,但小于350行左右,页文件就不存在(drop前页文件是存在的).这是个什么原因呢
脚本输出信息:

以下为精简信息,完整输出信息请点击下载文本(由于不能上传txt,所以在文件名后面加了sh,下载后请删除.sh):
320行,没有drop记录:
320行,drop
640行,drop

 
select count(*) from t2
640
Table   Checksum
wenyz.t2    1273189789
...
-rw-r--r-- 1 root root   16384 Aug  4 23:09 0000000000000040.page
-rw-r--r-- 1 root root   98304 Aug  4 23:09 0000000000000041.page /被drop表空间文件
-rw-r--r-- 1 root root   16384 Aug  4 23:09 18446744069414584320.page
 
count(*)320(行)
Table   Checksum
wenyz.t2    3018070873
....
-rw-r--r-- 1 root root   16384 Aug  4 23:12 0000000000000038.page
-rw-r--r-- 1 root root   32768 Aug  4 23:12 0000000000000039.page
-rw-r--r-- 1 root root   16384 Aug  4 23:12 0000000000000040.page \\这里缺41.page
-rw-r--r-- 1 root root   16384 Aug  4 23:12 18446744069414584320.page

把drop命令在脚本里注释后(还是320行)

以下为精简信息,完整输出信息附后:
#/tmp/init3507.sh
BrI?Zu>o=1uN
mysql: [Warning] Using a password on the command line interface can be insecure.
root@localhost [(none)]>alter user user() identified by 'xxxxx';
Query OK, 0 rows affected (0.00 sec)
 
root@localhost [(none)]>exit
Bye
mysql: [Warning] Using a password on the command line interface can be insecure.
count(*) ---------------------------------------------------------------------------------320行数据
320
Table   Checksum
wenyz.t2    2368041617
...
-rw-r--r-- 1 root root   32768 Aug  4 23:16 0000000000000039.page
-rw-r--r-- 1 root root   16384 Aug  4 23:16 0000000000000040.page
-rw-r--r-- 1 root root   32768 Aug  4 23:16 0000000000000041.page //320行时不drop是有此表空间文件的
-rw-r--r-- 1 root root   16384 Aug  4 23:16 18446744069414584320.page
[root@db211_23:16:21 /opt/undrop-for-innodb]  
#

• 为了快速测试数据恢复,使用的脚本
  init3507.sh
  /tmp/init3507.sh

pkill mysqld
rm -rf /3507/data/* /3507/logs/*
/usr/local/mysql57/bin/mysqld --defaults-file=/3507/my3507.cnf --initialize-insecure
/usr/local/mysql57/bin/mysqld --defaults-file=/3507/my3507.cnf&
sleep 2
/usr/local/mysql57/bin/mysql -S /tmp/mysql3507.sock -uroot  </tmp/create_t2.sql
rm -rf /opt/undrop-for-innodb/pages-ibdata1
#ls -l
cd /opt/undrop-for-innodb/
sleep 5
rm -rf /opt/undrop-for-innodb/pages-ibdata1
cd /opt/undrop-for-innodb/
/opt/undrop-for-innodb/stream_parser -f /3507/data/ibdata1
ls -l /opt/undrop-for-innodb/pages-ibdata1/FIL_PAGE_INDEX

/tmp/create_t2.sql (由于不能上传sql文件,所以只能改为sh)

alter user user() identified by 'xxxx';
CREATE DATABASE wenyz;
use wenyz;
CREATE TABLE `t2` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `ti` varchar(100) NOT NULL,
  `date` date DEFAULT NULL,
  PRIMARY KEY (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=4079859 DEFAULT CHARSET=utf8;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) values(substring(MD5(RAND()),floor(RAND()*26)+1,15),now()) ;
insert into t2(ti,date) select ti,date from t2;
insert into t2(ti,date) select ti,date from t2;
insert into t2(ti,date) select ti,date from t2;
insert into t2(ti,date) select ti,date from t2;
insert into t2(ti,date) select ti,date from t2;
select count(*) from t2;
checksum table t2;
drop table t2;

7.参考过的资料 [ Top ]

• 官方文档:twindb:
  https://github.com/twindb/undrop-for-innodb
  https://twindb.com/recover-after-drop-table-innodb_file_per_table-is-off/
• 叶老师博客:B+树
• 李真旭(Roger)博客:
• yunqishequ1的博客