my14_mysql指定时间恢复之模拟从库

2022-09-19 10:33:28

场景

*********************************
线上库数据误删除，存在几天前的一份全备数据，现需要恢复这些误删除的数据
本例方案：在另外一台服务器上，恢复全备，搭建binlog server复制从全备时的binlog日志，然后模拟从库使用SQL_THEAD追加数据
方案特点：可恢复到全备到当前之间的任何时刻；使用模拟从库的方式，不用担心需要追加的binlog太多会引发异常

常用使用binlog恢复的异常有

Max_allowed_packet问题

Blob/Binary/text字段问题

特殊字符的转义问题

没有”断点恢复”：执行出错后，没有足够的报错，也很难从失败的地方继续恢复

操作简记
**********************************
现库环境（非GTID环境）
---------------------------------------------
mysqld_safe --defaults-file=/etc/my_bakdb.cnf --user=mysql &
mysql -uautomng -p********* -P3319 -h127.0.0.1
mysql> select count(*) from sbtest5;
+----------+
| count(*) |
+----------+
| 300000 |
+----------+
1 row in set (1.30 sec)

全备数据准备
-----------------------------------------------
vim xbakdb.sh
#!/usr/bin/env bash
rm -rf /data0/backup/dbbak_20180813/*
innobackupex --defaults-file=/etc/my_bakdb.cnf --no-timestamp --user automng --host=192.168.56.82 --port=3319 --password ******** /data0/backup/dbbak_20180813

chmod +x xbakdb.sh

使用screen记录日志输出，默认输出到/tmp/screenlog_bakdb.log
[root@red4 mysql]# screen -L -t bakdb -dmS backup /opt/scripts/backup/mysql/xbakdb.sh
[root@red4 mysql]# tailf /tmp/screenlog_bakdb.log

备份完成后应用日志
innobackupex --user automng --host=192.168.56.82 --port=3319 --password ********* --apply-log /data0/backup/dbbak_20180813

模拟事务
----------------------------------------------
备份过程执行两次删除
mysql> delete from sbtest5 limit 10;
Query OK, 10 rows affected (0.45 sec)

备份完成后再执行两次删除
mysql> delete from sbtest5 limit 10;
Query OK, 10 rows affected (0.11 sec)

此时库中该表删除了40条记录，预计全备恢复时为删除20记录
mysql> select count(*) from sbtest5;
+----------+
| count(*) |
+----------+
| 299960 |
+----------+
1 row in set (0.04 sec)

使用sysbench清除现在的表，再插入80张表，每张表5万条记录，意思就是在恢复的过程中，原库一直很繁忙，会产生大量的binlog日志

全备恢复
----------------------------------------------------
将备份数据复制到另外一个服务器上
scp /etc/my_bakdb.cnf mysql01:/etc/
scp -r dbbak_20180813/ mysql01:/data0/backup

在mysql01服务器上创建bakdb使用的目录与目录
mkdir -p /data0/mysql/log/bakdb
mkdir -p /data0/mysql/data/bakdb
touch /data0/mysql/log/bakdb/error.log
chown -R mysql.mysql /data0/mysql/log/bakdb

注意：由于是在新服务器上恢复，xtrabackup不备份日志相关文件，所以需要手工创建一个日志文件并对日志目录授权

直接将应用日志后的备份文件复制到库的位置
cp -r dbbak_20180813/* /data0/mysql/data/bakdb/
chown -R mysql.mysql /data0/mysql/data/bakdb/
mysqld_safe --defaults-file=/etc/my_bakdb.cnf --user=mysql &
mysql -uautomng -p********* -P3319 -h127.0.0.1

查看数据发现少了20条记录，与前面的预计相符
mysql> select count(*) from sbtest5;
+----------+
| count(*) |
+----------+
| 299980 |
+----------+
1 row in set (1.96 sec)

搭建binlog server
-------------------------------------------------------
注意磁盘空间，要寻找一个足够放得下全备至恢复点日志的目录
mkdir -p /data0/backup/bakdb_binlog/
mysqlbinlog --no-defaults -h192.168.56.82 -P3319 -uautomng -p******** -R --raw --stop-never mysql-bin.000043 -r /data0/backup/bakdb_binlog/ --stop-never-slave-server-id=33119 &

确认恢复位置
-----------------------------------------------------
模拟的业务是创建80张表，每张表5万数据，将位置定在第31张表，即恢复到第31张表创建之后但插入数据之前的位置，通过几次尝试确定语句在mysql-bin.000048 中
mysqlbinlog --no-defaults -v -v --base64-output=decode-rows /data0/backup/bakdb_binlog/mysql-bin.000048 > ./tmp/mysql_48.sql
mysql_48.sql 1.7G太大，先拆分成小文件
split -b 300m mysql_48.sql

/*!*/;
# at 411781429
#180815 11:11:40 server id 3319 end_log_pos 411781490 Anonymous_GTID last_committed=409 sequence_number=410 rbr_only=yes
/*!50718 SET TRANSACTION ISOLATION LEVEL READ COMMITTED*//*!*/;
SET @@SESSION.GTID_NEXT= 'ANONYMOUS'/*!*/;
# at 411781490
#180815 11:11:40 server id 3319 end_log_pos 411781558 Query thread_id=18 exec_time=0 error_code=0
SET TIMESTAMP=1534302700/*!*/;
BEGIN
/*!*/;
# at 411781558
#180815 11:11:40 server id 3319 end_log_pos 412305775 Rows_query
# INSERT INTO sbtest31(k, c, pad) VALUES(247741, '46161912814-94771992086-29465110879-96852384793-12688454296-19046139779-19533481261-64495786388-42440185851-37409777004',
'88689078905-06893547748-69501204574-82481850912-44422599985'),

411781429标记时，设置事务级别为RC（相比默认的RR级别，RC级别可以减少并发事务--此处为并发insert 所造成的锁等待，sysbench支持多线程），后面开始插入sbtest31的数据，本次恢复的位置就定为411781429

恢复前的数据备份
------------------------------------------------------
开始操作数据了，操作前先对原数据做一次备份，操作失败了还有一份备份
mysqladmin shutdown -uautomng -p******* -S /data0/mysql/log/bakdb/mysql_bakdb.sock
cp -r /data0/mysql/data/bakdb/ /data0/backup/

手工制作 realy_log
------------------------------------------
mysql> show variables like 'relay_log_basename';
+--------------------+-------------------------------------------+
| Variable_name | Value |
+--------------------+-------------------------------------------+
| relay_log_basename | /data0/mysql/data/bakdb/mysql02-relay-bin |
+--------------------+-------------------------------------------+
1 row in set (1.22 sec)

mysql> show variables like 'relay_log_index';
+-----------------+-------------------------------------------------+
| Variable_name | Value |
+-----------------+-------------------------------------------------+
| relay_log_index | /data0/mysql/data/bakdb/mysql02-relay-bin.index |
+-----------------+-------------------------------------------------+
1 row in set (0.00 sec)

# cat relay_log.sh
#!/usr/bin/env bash

for i in $(ls mysql-bin.00*)
do
nn=$(echo $i|cut -d'.' -f2);
cp $i mysql02-relay-bin.$nn;
echo "./mysql02-relay-bin.$nn" >> mysql02-relay-bin.index;
done

执行脚本前，先停止一下binlog server，因为最后一个文件还在不断地增加，或者恢复时不使用最后一个文件
# cat mysql02-relay-bin.index
./mysql02-relay-bin.000043
./mysql02-relay-bin.000044
./mysql02-relay-bin.000045
./mysql02-relay-bin.000046
./mysql02-relay-bin.000047
./mysql02-relay-bin.000048
./mysql02-relay-bin.000049
./mysql02-relay-bin.000050
./mysql02-relay-bin.000051
./mysql02-relay-bin.000052
./mysql02-relay-bin.000053
./mysql02-relay-bin.000054
./mysql02-relay-bin.000055
./mysql02-relay-bin.000056
./mysql02-relay-bin.000057
./mysql02-relay-bin.000058
./mysql02-relay-bin.000059
./mysql02-relay-bin.000060
# ls -ltrh | grep mysql02
-rw-r-----. 1 root root 1.1G Aug 16 11:27 mysql02-relay-bin.000043
-rw-r-----. 1 root root 1.1G Aug 16 11:28 mysql02-relay-bin.000044
-rw-r-----. 1 root root 1.1G Aug 16 11:30 mysql02-relay-bin.000045
-rw-r-----. 1 root root 1.1G Aug 16 11:31 mysql02-relay-bin.000046
-rw-r-----. 1 root root 1.1G Aug 16 11:32 mysql02-relay-bin.000047
-rw-r-----. 1 root root 1.1G Aug 16 11:33 mysql02-relay-bin.000048
-rw-r-----. 1 root root 1.1G Aug 16 11:35 mysql02-relay-bin.000049
-rw-r-----. 1 root root 1.1G Aug 16 11:36 mysql02-relay-bin.000050
-rw-r-----. 1 root root 1.1G Aug 16 11:37 mysql02-relay-bin.000051
-rw-r-----. 1 root root 1.1G Aug 16 11:39 mysql02-relay-bin.000052
-rw-r-----. 1 root root 1.1G Aug 16 11:40 mysql02-relay-bin.000053
-rw-r-----. 1 root root 1.1G Aug 16 11:41 mysql02-relay-bin.000054
-rw-r-----. 1 root root 1.1G Aug 16 11:42 mysql02-relay-bin.000055
-rw-r-----. 1 root root 1.1G Aug 16 11:44 mysql02-relay-bin.000056
-rw-r-----. 1 root root 362M Aug 16 11:44 mysql02-relay-bin.000057
-rw-r-----. 1 root root 193 Aug 16 11:44 mysql02-relay-bin.000058
-rw-r-----. 1 root root 169 Aug 16 11:44 mysql02-relay-bin.000059
-rw-r-----. 1 root root 150 Aug 16 11:44 mysql02-relay-bin.000060
-rw-r--r--. 1 root root 486 Aug 16 11:44 mysql02-relay-bin.index

复制制作的relay_log到指定目录，默认在数据目录下，此处单独创建一个目录作为relay log使用
mkdir -p /data0/mysql/log/bakdb_relay_log
cp mysql02-relay-bin.* /data0/mysql/log/bakdb_relay_log
chown -R mysql.mysql /data0/mysql/log/bakdb_relay_log/

修改配置文件relay相关参数
-------------------------------------------------------------
修改server-id，使从库与主库的server-id不同

配置relay_log 指定relay_log地址

配置文件中定义，relay_log_basename是自动计算的，不需要在配置文件中定义
relay_log = /data0/mysql/log/bakdb_relay_log/mysql02-relay-bin
relay_log_index = /data0/mysql/log/bakdb_relay_log/mysql02-relay-bin.index
relay_log_info_file = /data0/mysql/log/bakdb_relay_log/relay-log.info
relay_log_info_repository = FILE
relay_log_purge = 0
skip-slave-start

重启服务器查看参数定义，此处再检查一次relay log的权限是否为mysql，否则就授权

启动SQL_THREAD恢复数据
---------------------------------------------------------------
# cat xtrabackup_binlog_info
mysql-bin.000043 4526

mysql -uroot -p********* -S /data0/mysql/log/bakdb/mysql_bakdb.sock

mysql> ? change master to
mysql> ? start slave

CHANGE MASTER TO
RELAY_LOG_FILE='mysql02-relay-bin.000043',
RELAY_LOG_POS=4526;

预计恢复完毕后，sbtest31已创建但无数据
start slave sql_thread until RELAY_LOG_FILE = 'mysql02-relay-bin.000048', RELAY_LOG_POS = 411781429;

mysql> select count(*) from sbtest31;
+----------+
| count(*) |
+----------+
| 0 |
+----------+
1 row in set (0.00 sec)

至此，恢复成功。

码农公寓

相关文章