MySQL Flashback 工具介绍
MySQL Flashback 工具介绍
-
DML Flashback
- 独立工具，通过伪装成slave拉取binlog来进行处理
- MyFlash 「大众点点评」
- binlog2sql 「大众点评(上海)」
- mysqlbinlog_flashback (更倾向于阿里RDS) 「58到家」
- patch形式集成到官方工具mysqlbinlog
- 简单脚本。先用mysqlbinlog解析出文本格式的binlog，再根据回滚原理用正则进行匹配并替换
- 独立工具，通过伪装成slave拉取binlog来进行处理
-
DDL Flashback
- Percona Data Recovery Tool for InnoDB --> code
- undrop-for-innodb --> code 「2017-01-01 已经闭源并收费」
DDL Falshback 要求离线，比 DML Flashback 要求更多，生产环境数据可能覆盖导致恢复不全
因为 5.6 开始，有单独的 purge 线程，数据恢复可能性更低了
部署位置
-
IP1
/data1/dbatools/mysqlflashback
- flashback_57 「mysqlbinlog」(因为系统原因，mysqlbinlog 放在 CentOS 7机器上)
- binlog2sql 「建议使用」
- MyFlash「建议使用」(测试暂时还不能在 CentOS 7上使用)
-
IP2
/data1/dbatools/mysqlflashback
- binlog2sql 「建议使用」
- MyFlash「建议使用」
MyFlash
github：MyFlash
MyFlash是由美团点评公司技术工程部开发维护的一个回滚DML操作的工具。该工具通过解析v4版本的binlog，完成回滚操作。相对已有的回滚工具，其增加了更多的过滤选项，让回滚更加容易。
版本兼容
- MySQL 5.6、5.7
限制
- binlog格式必须为row，且binlog_row_image=full
- 目前只支持单个 binlog 文件解析
- 只能回滚DML（增、删、改）
依赖
- glibc 版本
编译
git clone https://github.com/Meituan-Dianping/MyFlash.git
cd MyFlash
gcc -w `pkg-config --cflags --libs glib-2.0` source/binlogParseGlib.c -o binary/flashback
使用
基本用法
cd binary
# 回滚整个文件 && 应用
./flashback --binlogFileNames=mysql-bin.000005
mysqlbinlog binlog_output_base.flashback | mysql -h<host> -u<user> -p
# 回滚文件中 INSERT && 应用
./flashback --sqlTypes='INSERT'--binlogFileNames=mysql-bin.000005
mysqlbinlog binlog_output_base.flashback | mysql -h<host> -u<user> -p
# 回滚大文件 && 应用 (大于20M建议分割回滚文件)
./flashback --binlogFileNames=mysql-bin.000005
./flashback --maxSplitSize=1 --binlogFileNames=binlog_output_base.flashback
mysqlbinlog binlog_output_base.flashback.000001 | mysql -h<host> -u<user> -p
...
mysqlbinlog binlog_output_base.flashback.<N> | mysql -h<host> -u<user> -p
flashback 回滚的文件，还是mysqlbinlog格式的，如果想看 sql，还是需要原生 mysqlbinlog 进行解析
/usr/local/mysql-5.7.18/bin/mysqlbinlog binlog_output_base.flashback
开启 gtid MySQL 应用flashback报错
ERROR 1782 (HY000) at line 16: @@SESSION.GTID_NEXT cannot be set to ANONYMOUS when @@GLOBAL.GTID_MODE = ON.
在导入时加入--skip-gtids mysqlbinlog --skip-gtids | mysql -uxxx -pxxx
选项
./flashback --help
Usage:
flashback [OPTION...]
Help Options:
-?, --help Show help options
Application Options:
--databaseNames databaseName to apply. if multiple, seperate by comma(,)
--tableNames tableName to apply. if multiple, seperate by comma(,)
--start-position start position
--stop-position stop position
--start-datetime start time (format %Y-%m-%d %H:%M:%S)
--stop-datetime stop time (format %Y-%m-%d %H:%M:%S)
--sqlTypes sql type to filter . support INSERT, UPDATE ,DELETE. if multiple, seperate by comma(,)
--maxSplitSize max file size after split, the uint is M
--binlogFileNames binlog files to process. if multiple, seperate by comma(,)
--outBinlogFileNameBase output binlog file name base
--logLevel log level, available option is debug,warning,error
--include-gtids gtids to process
--exclude-gtids gtids to skip
| 参数 | 解释 | |
|---|---|---|
| 1 | databaseNames | 指定需要回滚的数据库名。多个用“,”隔开。不指定该参数，相当于指定了所有数据库 |
| 2 | tableNames | 指定需要回滚的表名。多个表可以用“,”隔开。如果不指定该参数，相当于指定了所有表 |
| 3 | start-position | 指定回滚开始的位置。如不指定，从文件的开始处回滚。请指定正确的有效的位置，否则无法回滚 |
| 4 | stop-position | 指定回滚结束的位置。如不指定，回滚到文件结尾。请指定正确的有效的位置，否则无法回滚 |
| 5 | start-datetime | 指定回滚的开始时间。格式必须是 %Y-%m-%d %H:%M:%S。 如不指定，则不限定时间 |
| 6 | stop-datetime | 定回滚的结束时间。格式必须是 %Y-%m-%d %H:%M:%S。 如不指定，则不限定时间 |
| 7 | sqlTypes | 指定回滚的sql类型。支持的类型是INSERT, UPDATE ,DELETE。多个类型可以用“,”隔开。 |
| 8 | maxSplitSize | 一旦指定该参数，对文件进行固定尺寸的分割（单位为M），过滤条件有效，但不进行回滚操作。该参数主要用来将大的binlog文件切割，防止单次应用的binlog尺寸过大，对线上造成压力 |
| 9 | binlogFileNames | 需要回滚的binlog文件，目前只支持单个文件，后续会增加多个文件支持 |
| 10 | outBinlogFileNameBase | 输出的binlog文件前缀，如不指定，则默认为binlog_output_base.flashback |
| 11 | logLevel | 仅供开发者使用，默认级别为error级别。在生产环境中不要修改这个级别，否则输出过多 |
| 12 | include-gtids | 指定需要回滚的gtid,支持gtid的单个和范围两种形式。 |
| 13 | exclude-gtids | 指定不需要回滚的gtid，用法同include-gtids |
binlog2sql
伪装成slave拉取binlog，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，建议表有主键。
github: binlog2sql
用途
-
数据快速回滚(闪回)
-
主从切换后新master丢数据的修复
-
从binlog生成标准SQL，带来的衍生功能
版本兼容
- Python 2.6、2.7
- MySQL 5.6 (简单测试在 5.5、5.7 gtid 均可用，不排除无 Bug，建议 5.6 以后版本使用)
依赖
- python-mysql-replication
- PyMySQL
- wheel
限制 (对比mysqlbinlog)
-
mysql server必须开启，离线模式下不能解析
-
参数 binlog_format 必须为ROW，参数 binlog_row_image 必须为FULL，暂不支持MINIMAL (5.6.2引入)
-
GEOMETRY 不能解析 (因为python-mysql-replication Limitations)
-
解析速度不如 mysqlbinlog
-
表有关联表，关联表并不会被回滚，需与业务方沟通清楚
-
不支持DDL
-
建议回滚的表有主键
优点 (对比mysqlbinlog)
- 纯Python开发，安装与使用都很简单
- 自带flashback、no-primary-key解析模式，无需再装补丁
- flashback模式下，更适合闪回实战
- 解析为标准SQL，方便理解、调试
- 代码容易改造，可以支持更多个性化解析
安装
git clone https://github.com/danfengcao/binlog2sql.git && cd binlog2sql
pip install -r requirements.txt
升级依赖模块
# cat requirements.txt
PyMySQL==0.7.11
wheel==0.29.0
mysql-replication==0.13
wget https://codeload.github.com/noplay/python-mysql-replication/tar.gz/0.15
tar zxf 0.15
cd python-mysql-replication-0.15
python2.7 setup.py install
使用
MySQL server必须设置以下参数
[mysqld]
server_id = 1
log_bin = /var/log/mysql/mysql-bin.log
max_binlog_size = 1G
binlog_format = row
binlog_row_image = full
user需要的最小权限集合
select, super/replication client, replication slave
建议授权
GRANT SELECT, REPLICATION SLAVE, REPLICATION CLIENT ON *.* TO
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
- 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
基本用法
解析标准 SQL (<u>会有 DDL SQL，回滚模式是没有 DDL的</u>)
python2.7 binlog2sql/binlog2sql.py -h127.0.0.1 -P6000 -umyadmin_common -p'password' -dtest -tt1 --start-file='mysql-bin.000192'
INSERT INTO `test`.`t1`(`id`, `name`) VALUES (5, 'name5'); #start 4 end 259 time 2017-09-14 19:59:31
DELETE FROM `test`.`t1` WHERE `id`=4 AND `name`='name4' LIMIT 1; #start 286 end 438 time 2017-09-14 19:59:42
UPDATE `test`.`t1` SET `id`=3, `name`='name300' WHERE `id`=3 AND `name`='name3' LIMIT 1; #start 465 end 631 time 2017-09-14 19:59:54
解析回滚SQL (在上面命令基础上加上 --flashback 或 -B)
python2.7 binlog2sql/binlog2sql.py -h127.0.0.1 -P6000 -umyadmin_common -p'ZTlh916@mxq' -dtest -tt1 --start-file='mysql-bin.000192' -B
UPDATE `test`.`t1` SET `id`=3, `name`='name3' WHERE `id`=3 AND `name`='name300' LIMIT 1; #start 465 end 631 time 2017-09-14 19:59:54
INSERT INTO `test`.`t1`(`id`, `name`) VALUES (4, 'name4'); #start 286 end 438 time 2017-09-14 19:59:42
DELETE FROM `test`.`t1` WHERE `id`=5 AND `name`='name5' LIMIT 1; #start 4 end 259 time 2017-09-14 19:59:31
选项
mysql连接配置
-h host; -P port; -u user; -p password
解析模式
--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SLEEP SELECT(1)。可选。与stop-never或no-primary-key不能同时添加。
解析范围控制
--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。
对象过滤
-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。
恢复数据
误操作大概时间范围
python binlog2sql/binlog2sql.py -h127.0.0.1 -P3306 -uadmin -p'admin' -dtest -ttbl --start-file='mysql-bin.000052' --start-datetime='2016-12-13 20:25:00' --stop-datetime='2016-12-13 20:30:00'
找到具体位点进行恢复
python binlog2sql/binlog2sql.py -h127.0.0.1 -P3306 -uadmin -p'admin' -dtest -ttbl --start-file='mysql-bin.000052' --start-position=3346 --stop-position=3556 -B > rollback.sql | cat
<u>回滚前，一定找业务确认回滚数据</u>
TIPS
-
闪回的目标：快速筛选出真正需要回滚的数据。
-
先根据库、表、时间做一次过滤，再根据位置做更准确的过滤。
-
由于数据一直在写入，要确保回滚sql中不包含其他数据。可根据是否是同一事务、误操作行数、字段值的特征等等来帮助判断。
-
执行回滚sql时如有报错，需要查实具体原因，一般是因为对应的数据已发生变化。由于是严格的行模式，只要有唯一键(包括主键)存在，就只会报某条数据不存在的错，不必担心会更新不该操作的数据。业务如果有特殊逻辑，数据回滚可能会带来影响。
-
如果只回滚某张表，并且该表有关联表，关联表并不会被回滚，需与业务方沟通清楚。
mysqlbinlog_flashback
跟 binlog2sql 一样，伪装成slave拉取binlog，从MySQL binlog解析出你要的SQL，目前已经稳定运行在阿里RDS、db字符集为 utf8 的产生环境，<u>其他环境要谨慎使用</u>。
github: mysqlbinlog_flashback
版本兼容
- 阿里RDS
- 默认 utf8 DB (gbk等字符集的 DB 理论上支持)
限制
- mysql server必须开启，离线模式下不能解析
- 参数 binlog_format 必须为ROW，参数 binlog_row_image 必须为FULL，暂不支持MINIMAL (5.6.2引入)
- GEOMETRY 不能解析 (因为python-mysql-replication Limitations)
- 解析速度不如 mysqlbinlog
- 表有关联表，关联表并不会被回滚，需与业务方沟通清
- 不支持DDL
- <u>表必须有主键</u> (只限制在 mysqlbinlog_flashback，虽然 binlog2sql 不强制，但建议表有主键)
-
<u>阿里 RDS、db 字符集为 utf8 的环境稳定使用，其他环境要谨慎使用</u> (只限制在 mysqlbinlog_flashback)
-
<u>修改 python-mysql-replication</u> (只限制在 mysqlbinlog_flashback)
- 某些数据被封装，没有暴露出来
- Ali-rds导致的bug
-
<u>字段类型限制</u> (只限制在 mysqlbinlog_flashback)
ALLOW_TYPE={ "varchar":True, "char":True, "datetime":True, "date":True, "time":True, "timestamp":True, "bigint":True, "mediumint":True, "smallint":True, "tinyint":True, "int":True, "smallint":True, "decimal":True, "float":True, "double":True, "longtext":True, "tinytext":True, "text":True, "mediumtext":True }
-
mysql 导入一定要指定 字符集 --default-character-set=utf8
安装
git clone https://github.com/58daojia-dba/mysqlbinlog_flashback.git
cd mysqlbinlog_flashback
使用
MySQL server必须设置以下参数
[mysqld]
server_id = 1
log_bin = /var/log/mysql/mysql-bin.log
max_binlog_size = 1G
binlog_format = row
binlog_row_image = full
基本用法
python2.7 mysqlbinlog_back.py --host="127.0.0.1" --port=7000 --user="mysql_user" --password='password' --schema=test --table=t2
ls -l log/*
-rw-r--r-- 1 root root 1.9K Sep 19 17:40 save_data_dml_test_20170919_174001.sql
-rw-r--r-- 1 root root 179 Sep 19 17:40 save_data_create_table_test_20170919_174001.sql
-rw-r--r-- 1 root root 1.1K Sep 19 17:40 flashback_test_20170919_174001.sql <--- 反向sq文件
mysql -uroot -p -h -P --default-character-set=utf8 < flashback_test_20170919_174001.sql
选项
-s SCHEMA, --schema=SCHEMA 必须的，指定 DB
-t TABLES, --tables=TABLES 必须的，指定 Table，用逗号分隔
-N BINLOG_END_TIME, --binlog_end_time=BINLOG_END_TIME
format yyyy-mm-dd hh24:mi:ss,default is current time
-S BINLOG_START_FILE_NAME, --binlog_start_file_name=BINLOG_START_FILE_NAME
binlog start file name,default is current logfile of db
-L BINLOG_START_FILE_POSITION, --binlog_start_file_position=BINLOG_START_FILE_POSITION
binlog start file name
-E BINLOG_START_TIME, --binlog_start_time=BINLOG_START_TIME
binlog start time,format yyyy-mm-dd hh24:mi:ss
-l OUTPUT_FILE_PATH, --output_file_path=OUTPUT_FILE_PATH
file path that sql generated,,default ./log
-I, --skip_insert skip insert(WriteRowsEvent) event
-U, --skip_update skip update(UpdateRowsEvent) event
-D, --skip_delete skip delete(DeleteRowsEvent) event
-a, --add_schema_name add schema name for flashback sql
mysqlbinlog
以patch形式集成到官方工具mysqlbinlog中。以彭立勋提交的patch为代表 MySQL下实现闪回的设计思路 (MySQL Flashback Feature)
mysqlbinlog 闪回工具
-
彭立勋 patch （只针对 MySQL 5.5) --> 具体步骤参考 尝试mysqlbinlog的flashback功能
-
姜承尧 - 关于MySQL Flashback工具的使用及介绍
优点
-
上手成本低，原 mysqlbinlog 选项可以直接利用，只是多一个闪回选项
-
支持离线解析
-
支持事务闪回，语句逆转、顺序也颠倒
缺点
- 兼容性差、项目活跃度不高，mysql5.5，5.6，5.7 分别开发了patch
- 难以添加新功能，实战效果欠佳，例如过滤一张表 (姜承尧版本也是支持的)
- 安装稍显麻烦。需要对mysql源码打补丁再编译生成
限制
-
参数 binlog_format 必须为ROW，参数 binlog_row_image 必须为FULL，暂不支持MINIMAL
-
不支持DDL
使用
MySQL server必须设置以下参数
[mysqld]
server_id = 1
log_bin = /var/log/mysql/mysql-bin.log
max_binlog_size = 1G
binlog_format = row
binlog_row_image = full
基本用法
解析 mysqlbinlog
./mysqlbinlog -vv mysql-bin.000067 --start-datetime="2017-09-19 15:00:00" --stop-datetime="2017-09-19 15:38:10"
回滚
./mysqlbinlog -vv mysql-bin.000067 --start-position=1427 --stop-position=2588 -B | mysql -umysqlha_common -p'password' -h10.13.2.29 -P 7000 test
position 位点找需要回滚事务 BEGIN 前面那个 position.
选项
-B, --flashback Flashback data to start_postition or start_datetime.
-A, --skip-database=name 解析BinLog时过滤掉该数据库 (local log only)
-a, --skip-table=name 解析BinLog时过滤掉该表，一般与skip_datebase配套使用 (local log only)
--skip-database 可以单独使用，--skip-table需要与 skip-database配置使用，<u>多个db或table不生效</u>
--skip-database=test --skip-table=t1
-O, --split-size-interval=# 将BinLog文件按照指定的大小拆分为多个段，解析结果为打印每个段的起始offset位置
注意，当进行flashback时，flashback的内容先保存在内存中。若你的binlog大小为10G，那么需要额外的10G内存先暂时保存这部分信息。在某些情况下，如云环境、或服务器内存较小，会导致无法输出flashback的日志。这时可以通过此参数来设置内存保存文件的大小，例如将此值设置为100M，那么每100M就会刷新到一个文件
-D, --datetime-to-pos=name 基于输入的时间信息，解析出该时间对应的第一个BinLog event偏移位置，格式参照start-datetime
flashback时要先找到起始的偏移量，DBA可以先通过此参数定位到具体位置，然后再进行flashback操作
「不是很好用啊，少了一个」
-T, --table=name 仅解析该表，一般与database配套使用 (local log only)
-E, --fb-event=name 仅解析该类型的Log event，一般与database、table选项配套使用。(local log only)
可选的值有:
- DELETE
- INSERT
- UPDATE
DDL Flashbackup
DDL Falshback 要求离线，比 DML Flashback 要求更多，生产环境数据可能覆盖导致恢复不全。
MySQL 对于空间的重用机制与Oracle 有很大区别，对于Oracle 而言，如果是delete的数据，还是很难被覆盖掉的，对于drop 和truncate 则领导别论。然而MySQL则有所不同，MySQL 默认会启动一些purge 进程来进行空间重用，这是MySQL 5.6的情况
localhost.(none)>show variables like '%innodb_purge_threads';
+----------------------+-------+
| Variable_name | Value |
+----------------------+-------+
| innodb_purge_threads | 1 |
+----------------------+-------+
1 row in set (0.00 sec)
在MySQL 5.7 版本中更为坑爹，MySQL 默认会启动4个purge 线程，因此很容易就会导致空间被重用，最终导致数据无法恢复，如下是MySQL 5.7的purge相关参数
localhost.(none)>show variables like '%innodb_purge_threads';
+----------------------+-------+
| Variable_name | Value |
+----------------------+-------+
| innodb_purge_threads | 4 |
+----------------------+-------+
1 row in set (0.00 sec)
Percona Data Recovery Tool for InnoDB
限制
-
只对InnoDB/XtraDB表有效，而无法恢复MyISAM表
-
离线恢复，MySQL Server 是关闭的
-
不能保证数据总一定可被恢复。
被重写的数据不能被恢复，这种情况下可能需要针对系统或物理的方式来恢复，不属于本工具的范畴。
-
恢复的最好时机是当你发现数据丢失时，尽快备份MySQL数据文件
-
innodb_file_per_table = 0
安装
现在可以下载还是 0.5，2011-08-28，0.6 还未发布
yum install glibc glibc-static -y
wget https://launchpad.net/percona-data-recovery-tool-for-innodb/trunk/release-0.5/+download/percona-data-recovery-tool-for-innodb-0.5.tar.gz
tar -xvf percona-data-recovery-tool-for-innodb-0.5.tar.gz
cd percona-data-recovery-tool-for-innodb-0/mysql-source/
./configure
cd ..
make
# ll -rth
total 3.0M
-rw-r--r-- 1 510 wheel 2.0K Aug 28 2011 tables_dict.c
-rwxr-xr-x 1 510 wheel 302 Aug 28 2011 split_dump.pl
-rw-r--r-- 1 510 wheel 11K Aug 28 2011 print_data.c
-rw-r--r-- 1 510 wheel 15K Aug 28 2011 page_parser.c
-rw-r--r-- 1 510 wheel 2.7K Aug 28 2011 Makefile
-rw-r--r-- 1 510 wheel 74 Aug 28 2011 INSTALL
-rw-r--r-- 1 510 wheel 9.0K Aug 28 2011 innochecksum.c
-rw-r--r-- 1 510 wheel 8.1K Aug 28 2011 incrementalupdate.c
-rw-r--r-- 1 510 wheel 12K Aug 28 2011 ibdconnect.c
-rwxr-xr-x 1 510 wheel 2.0K Aug 28 2011 fetch_data.sh
-rwxr-xr-x 1 510 wheel 12K Aug 28 2011 create_defs.pl
-rw-r--r-- 1 510 wheel 22K Aug 28 2011 constraints_parser.c
-rw-r--r-- 1 510 wheel 6.2K Aug 28 2011 check_data.c
drwxr-xr-x 2 510 wheel 4.0K Aug 28 2011 include
drwxr-xr-x 2 510 wheel 4.0K Aug 28 2011 docs
drwxr-xr-x 40 510 wheel 4.0K Sep 26 17:53 mysql-source
drwxr-xr-x 2 root root 4.0K Sep 26 17:55 lib
-rwxr-xr-x 1 root root 725K Sep 26 17:55 constraints_parser
-rwxr-xr-x 1 root root 1.2M Sep 26 17:55 page_parser
-rwxr-xr-x 1 root root 15K Sep 26 17:55 innochecksum
-rwxr-xr-x 1 root root 973K Sep 26 17:55 ibdconnect
page_parser工具将根据InnoDB的底层实现原理，解析表的页和行结构。constraints_parser工具暂时不使用，后续还需要在定义表结构之后，重新编译生成它。
如果MySQL是5.0之前的版本，InnoDB采取的是REDUNDANT格式，运行以下命令：
./page_parser -4 -f /path/to/ibdata1
如果MySQL是5.0以后的版本，InnoDB采取的是COMPACT格式，运行以下命令：
./page_parser -5 -f /path/to/ibdata1
运行后，page_parser工具会创建一个pages-<TIMESTAMP>的目录，其中TIMESTAMP是UNIX系统时间戳。在这个目录下，为每个index ID，以页的index ID创建一个子目录
...
...
...
这个工具限制很多，没有测试，具体步骤可以参考 "其他参考"
其他参考：
Percona Data Recovery Tool 单表恢复
使用Percona Data Recovery Tool for InnoDB恢复数据
Undrop MySQL InnoDB 中恢复被drop的表，当 innodb_file_per_table=off时
undrop-for-innodb
「2017-01-01 已经闭源并收费」，来源wiki
MySQL · 数据恢复 · undrop-for-innodb 「数据库内核月报 － 2017 / 11」
如果 innodb_file_per_table = on 恢复可能性更低
- idb 文件被脏写
- 有单独的 purge 线程
限制
- innodb_purge_threads = 0
innodb_purge_threads <=5.6.1，默认为0，范围 [0-1]
= 5.6.2, <= 5.6.4 默认为0，范围 [0-32]
= 5.6.5 默认为1，范围 [1-32]
= 5.7.8 默认为4，范围 [4-32]
5.6 以后，DDL 恢复还是比较困难的
安装
git clone https://github.com/chhabhaiya/undrop-for-innodb
cd undrop-for-innodb
make
使用 「MySQL 5.6.24」
1、truncate table
localhost.test>select * from t_recover;
+----+--------+
| id | name |
+----+--------+
| 1 | test1 |
| 2 | test2 |
| 3 | test3 |
| 4 | test4 |
| 5 | test5 |
| 6 | test6 |
| 7 | test7 |
| 8 | test8 |
| 9 | test9 |
| 10 | test10 |
+----+--------+
10 rows in set (0.00 sec)
mysql> truncate table t_recover;
2、获取数据字典
# ./stream_parser -f /data1/mysql7000/ibdata1
Opening file: /data1/mysql7000/ibdata1
File information:
ID of device containing file: 2055
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
Opening file: /data1/mysql7000/ibdata1
File information:
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
ID of device containing file: 2055
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
Opening file: /data1/mysql7000/ibdata1
total size, in bytes: 104857600 (100.000 MiB)
number of blocks allocated: 205008
File information:
Size to process: 104857600 (100.000 MiB)
ID of device containing file: 2055
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
Opening file: /data1/mysql7000/ibdata1
File information:
ID of device containing file: 2055
inode number: 7012359
protection: 100640 (regular file)
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
total size, in bytes: 104857600 (100.000 MiB)
Opening file: /data1/mysql7000/ibdata1
total size, in bytes: 104857600 (100.000 MiB)
File information:
Size to process: 104857600 (100.000 MiB)
ID of device containing file: 2055
Size to process: 104857600 (100.000 MiB)
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
total size, in bytes: 104857600 (100.000 MiB)
Opening file: /data1/mysql7000/ibdata1
File information:
Size to process: 104857600 (100.000 MiB)
ID of device containing file: 2055
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
Opening file: /data1/mysql7000/ibdata1
user ID of owner: 647
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
File information:
group ID of owner: 553
device ID (if special file): 0
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
blocksize for filesystem I/O: 4096
ID of device containing file: 2055
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
number of blocks allocated: 205008
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
total size, in bytes: 104857600 (100.000 MiB)
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
Size to process: 104857600 (100.000 MiB)
Opening file: /data1/mysql7000/ibdata1
File information:
ID of device containing file: 2055
inode number: 7012359
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
protection: 100640 (regular file)
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
number of hard links: 1
user ID of owner: 647
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
group ID of owner: 553
device ID (if special file): 0
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
total size, in bytes: 104857600 (100.000 MiB)
Size to process: 104857600 (100.000 MiB)
Opening file: /data1/mysql7000/ibdata1
File information:
ID of device containing file: 2055
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
total size, in bytes: 104857600 (100.000 MiB)
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
Size to process: 104857600 (100.000 MiB)
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
total size, in bytes: 104857600 (100.000 MiB)
Opening file: /data1/mysql7000/ibdata1
File information:
Size to process: 104857600 (100.000 MiB)
ID of device containing file: 2055
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
total size, in bytes: 104857600 (100.000 MiB)
Size to process: 104857600 (100.000 MiB)
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
total size, in bytes: 104857600 (100.000 MiB)
Size to process: 104857600 (100.000 MiB)
Opening file: /data1/mysql7000/ibdata1
File information:
ID of device containing file: 2055
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
total size, in bytes: 104857600 (100.000 MiB)
Size to process: 104857600 (100.000 MiB)
Opening file: /data1/mysql7000/ibdata1
File information:
ID of device containing file: 2055
inode number: 7012359
protection: 100640 (regular file)
number of hard links: 1
user ID of owner: 647
group ID of owner: 553
device ID (if special file): 0
blocksize for filesystem I/O: 4096
number of blocks allocated: 205008
time of last access: 1506651425 Fri Sep 29 10:17:05 2017
time of last modification: 1506651465 Fri Sep 29 10:17:45 2017
time of last status change: 1506651465 Fri Sep 29 10:17:45 2017
total size, in bytes: 104857600 (100.000 MiB)
Size to process: 104857600 (100.000 MiB)
All workers finished in 0 sec
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
# cd pages-ibdata1/
# ll -rth
total 8.0K
drwxr-xr-x 2 root root 4.0K Sep 29 10:18 FIL_PAGE_TYPE_BLOB
drwxr-xr-x 2 root root 4.0K Sep 29 10:18 FIL_PAGE_INDEX
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。
undrop-for-innodb]# ll pages-ibdata1/FIL_PAGE_INDEX/0000000000000001.page
-rw-r--r-- 1 root root 32768 Sep 29 10:18 pages-ibdata1/FIL_PAGE_INDEX/0000000000000001.page
SYS_INDEXES
SYS_INDEXES用于保存innodb中每个表定义的每个索引对象
undrop-for-innodb]# ll pages-ibdata1/FIL_PAGE_INDEX/0000000000000003.page
-rw-r--r-- 1 root root 16384 Sep 29 10:18 pages-ibdata1/FIL_PAGE_INDEX/0000000000000003.page
SYS_COLUMNS
SYS_COLUMNS用于保存innodb引擎每个表定义的列，与SYS_TABLES相似。
undrop-for-innodb]# ll pages-ibdata1/FIL_PAGE_INDEX/0000000000000002.page
-rw-r--r-- 1 root root 32768 Sep 29 10:18 pages-ibdata1/FIL_PAGE_INDEX/0000000000000002.page
SYS_FIELDS
SYS_FIELDS用于保存innodb中每个索引的每个列对象
undrop-for-innodb]# ll pages-ibdata1/FIL_PAGE_INDEX/0000000000000004.page
-rw-r--r-- 1 root root 32768 Sep 29 10:18 pages-ibdata1/FIL_PAGE_INDEX/0000000000000004.page
3、扫描逻辑卷
./stream_parser -f /dev/sda7 -t 105000000k
Opening file: /dev/sda7
File information:
ID of device containing file: 5
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
Opening file: /dev/sda7
File information:
ID of device containing file: 5
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
total size, in bytes: 0 (0.000 exp(+0))
Opening file: /dev/sda7
File information:
Size to process: 107520000000 (100.136 GiB)
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
ID of device containing file: 5
inode number: 6047
protection: 60660 total size, in bytes: 0 (0.000 exp(+0))
(block device)
Opening file: /dev/sda7
number of hard links: 1
Size to process: 107520000000 (100.136 GiB)
File information:
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
ID of device containing file: 5
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
Opening file: /dev/sda7
File information:
ID of device containing file: 5
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
Opening file: /dev/sda7
user ID of owner: 0
group ID of owner: 6
File information:
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
ID of device containing file: 5
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
user ID of owner: 0
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
group ID of owner: 6
total size, in bytes: 0 (0.000 exp(+0))
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
total size, in bytes: 0 (0.000 exp(+0))
Size to process: 107520000000 (100.136 GiB)
Size to process: 107520000000 (100.136 GiB)
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
Opening file: /dev/sda7
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
File information:
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
ID of device containing file: 5
total size, in bytes: 0 (0.000 exp(+0))
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
Size to process: 107520000000 (100.136 GiB)
number of blocks allocated: 0
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
Opening file: /dev/sda7
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
File information:
total size, in bytes: 0 (0.000 exp(+0))
ID of device containing file: 5
inode number: 6047
Size to process: 107520000000 (100.136 GiB)
protection: 60660 (block device)
number of hard links: 1
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
Opening file: /dev/sda7
total size, in bytes: 0 (0.000 exp(+0))
File information:
ID of device containing file: 5
Size to process: 107520000000 (100.136 GiB)
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
total size, in bytes: 0 (0.000 exp(+0))
Size to process: 107520000000 (100.136 GiB)
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
total size, in bytes: 0 (0.000 exp(+0))
Size to process: 107520000000 (100.136 GiB)
Opening file: /dev/sda7
File information:
Opening file: /dev/sda7
File information:
ID of device containing file: 5
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
ID of device containing file: 5
user ID of owner: 0
inode number: 6047
group ID of owner: 6
protection: 60660 device ID (if special file): 2055
(block device)
blocksize for filesystem I/O: 4096
number of hard links: 1
number of blocks allocated: 0
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
total size, in bytes: 0 (0.000 exp(+0))
total size, in bytes: 0 (0.000 exp(+0))
Size to process: 107520000000 (100.136 GiB)
Size to process: 107520000000 (100.136 GiB)
Opening file: /dev/sda7
File information:
ID of device containing file: 5
inode number: 6047
protection: 60660 (block device)
number of hard links: 1
user ID of owner: 0
group ID of owner: 6
device ID (if special file): 2055
blocksize for filesystem I/O: 4096
number of blocks allocated: 0
time of last access: 1490933835 Fri Mar 31 12:17:15 2017
time of last modification: 1477372695 Tue Oct 25 13:18:15 2016
time of last status change: 1477372695 Tue Oct 25 13:18:15 2016
total size, in bytes: 0 (0.000 exp(+0))
Size to process: 107520000000 (100.136 GiB)
Worker(2): 1.12% done. 2017-09-29 10:22:56 ETA(in 00:03:12). Processing speed: 44.000 MiB/sec
Worker(3): 1.12% done. 2017-09-29 10:22:56 ETA(in 00:03:12). Processing speed: 44.000 MiB/sec
....
....
....
Worker(11): 97.77% done. 2017-09-29 10:23:21 ETA(in 00:00:02). Processing speed: 87.828 MiB/sec
Worker(0): 97.74% done. 2017-09-29 10:23:21 ETA(in 00:00:02). Processing speed: 87.844 MiB/sec
Worker(9): 97.74% done. 2017-09-29 10:23:21 ETA(in 00:00:02). Processing speed: 87.844 MiB/sec
Worker(8): 99.80% done. 2017-09-29 10:23:20 ETA(in 00:00:00). Processing speed: 87.828 MiB/sec
Worker(11): 98.80% done. 2017-09-29 10:23:21 ETA(in 00:00:01). Processing speed: 87.828 MiB/sec
Worker(0): 98.77% done. 2017-09-29 10:23:21 ETA(in 00:00:01). Processing speed: 87.828 MiB/sec
Worker(9): 98.77% done. 2017-09-29 10:23:21 ETA(in 00:00:01). Processing speed: 87.828 MiB/sec
Worker(0): 99.80% done. 2017-09-29 10:23:21 ETA(in 00:00:00). Processing speed: 87.828 MiB/sec
Worker(11): 99.83% done. 2017-09-29 10:23:21 ETA(in 00:00:00). Processing speed: 87.828 MiB/sec
Worker(9): 99.80% done. 2017-09-29 10:23:21 ETA(in 00:00:00). Processing speed: 87.844 MiB/sec
All workers finished in 219 sec
[root@idaross29 undrop-for-innodb]# ./stream_parser -f /dev/sda7 -t 105000000k
Could not create directory pages-sda7
mkdir(): File exists
4、创建数据字典表
recover_dictionary.sh
修改 mysql 登陆相关命令 mysql -u -p -S
# ./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 ... 24 recs OK
SYS_COLUMNS ... 148 recs OK
SYS_INDEXES ... 14 recs OK
SYS_FIELDS ... 34 recs OK
All OK
localhost.test>show tables;
+----------------+
| Tables_in_test |
+----------------+
| SYS_COLUMNS |
| SYS_FIELDS |
| SYS_INDEXES |
| SYS_TABLES |
5、查询需要恢复的表的index_id信息
localhost.test>select * from SYS_TABLES where name like 'test/t_recover';
+----------------+----+--------+------+--------+---------+--------------+-------+
| NAME | ID | N_COLS | TYPE | MIX_ID | MIX_LEN | CLUSTER_NAME | SPACE |
+----------------+----+--------+------+--------+---------+--------------+-------+
| test/t_recover | 25 | 2 | 1 | 0 | 80 | | 11 |
+----------------+----+--------+------+--------+---------+--------------+-------+
1 row in set (0.02 sec)
localhost.test>select * from SYS_INDEXES where table_id = 25;
+----------+----+---------+----------+------+-------+---------+
| TABLE_ID | ID | NAME | N_FIELDS | TYPE | SPACE | PAGE_NO |
+----------+----+---------+----------+------+-------+---------+
| 25 | 28 | PRIMARY | 1 | 3 | 11 | 3 |
+----------+----+---------+----------+------+-------+---------+
1 row in set (0.01 sec)
可以看到被truncate的表的index_id 为28，我们应该进一步从28 的page中获取数据。
6、拿到表结构，并检查数据是否可以恢复
- 表结构
cat t_recover.sql
create table t_recover(id int primary key,name varchar(20));
- 确认数据是否存在
undrop-for-innodb]# ./c_parser -6f pages-ibdata1/FIL_PAGE_INDEX/0000000000000028.page -t t_recover.sql | head -5
SET FOREIGN_KEY_CHECKS=0;
-- Page id: 3, Format: COMPACT, Records list: Valid, Expected records: (0 0)
LOAD DATA LOCAL INFILE '-- Page id: 3, Found records: 0, Lost records: NO, Leaf page: YES
-- Page id: 3, Format: COMPACT, Records list: Valid, Expected records: (0 0)
/data1/jinlin3/undrop-for-innodb/dumps/default/t_recover' REPLACE INTO TABLE `t_recover` FIELDS TERMINATED BY '\t' OPTIONALLY ENCLOSED BY '"' LINES STARTING BY 't_recover\t' (`id`, `name`);
7、抽取 Page 中的数据
undrop-for-innodb]# ./c_parser -6f pages-ibdata1/FIL_PAGE_INDEX/0000000000000028.page -t t_recover.sql > dumps/default/t_recover 2> dumps/default/t_recover_load.sql
8、加载数据
source dumps/default/t_recover_load.sql
但是，我测试是 5.6 ，数据被 purge 掉了
undrop-for-innodb]# cat dumps/default/t_recover
-- Page id: 3, Format: COMPACT, Records list: Valid, Expected records: (0 0)
-- Page id: 3, Found records: 0, Lost records: NO, Leaf page: YES
-- Page id: 3, Format: COMPACT, Records list: Valid, Expected records: (0 0)
-- Page id: 3, Found records: 0, Lost records: NO, Leaf page: YES
undrop-for-innodb]# cat dumps/default/t_recover_load.sql
SET FOREIGN_KEY_CHECKS=0;
LOAD DATA LOCAL INFILE '/data1/jinlin3/undrop-for-innodb/dumps/default/t_recover' REPLACE INTO TABLE `t_recover` FIELDS TERMINATED BY '\t' OPTIONALLY ENCLOSED BY '"' LINES STARTING BY 't_recover\t' (`id`, `name`);
若 truncate table 后，立刻 shutdown MySQL，文件是有内容，但不对啊！
undrop-for-innodb]# cat dumps/default/t_truncate
-- Page id: 3, Format: COMPACT, Records list: Valid, Expected records: (17 17)
0000000B0000 00000000000025 t_truncate -2147483648 ""
0000000C0000 00000000000025 t_truncate -2147483648 "\0\0@�"
0000000D0000 00000000000025 t_truncate -2147483648 "\0\0@\b"
0000000E0000 00000000000025 t_truncate -2147483648 ""
0000000E0000 00010000000025 t_truncate -2147483648 "\0"
0000000F0000 00000000000025 t_truncate -2147483648 "\0\0"
000000100000 00000000000025 t_truncate -2147483648 "\0\0"
000000120000 00030000000025 t_truncate -2147483648 "\0\0@ps"
000000130000 00000000000025 t_truncate -2147483648 ""
000000140000 00000000000025 t_truncate -2147483648 "\0\0"
000000140000 00010000000025 t_truncate -2147483648 "\0\0"
000000150000 00000000000025 t_truncate -2147483648 ""
000000160000 00000000000025 t_truncate -2147483648 ""
000000170000 00000000000025 t_truncate -2147483648 "\0\0"
000000180000 00000000000025 t_truncate -2147483648 "\0\0@ka"
0000001A0000 00000000000025 t_truncate -2147483648 ""
0000001C0000 00000000000025 t_truncate -2147483648 ""
-- Page id: 3, Found records: 17, Lost records: NO, Leaf page: YES
<u>这是一个悲伤的故事！</u>
其他参考：
undrop for innodb c_parser 源码分析
MySQL 如何对InnoDB使用Undrop来恢复InnoDB数据
希望小伙伴永远不需要用到此方法，备份才是王道！
作者：saup007
链接：https://www.jianshu.com/p/31818e1727be
來源：简书
简书著作权归作者所有，任何形式的转载都请联系作者获得授权并注明出处。
