MySQL自己的 mysqlbinlog | mysql 回滚不好用，自己写个简单脚本试试：
想法是用mysqlbinlog把需要回滚的事务区域从mysql-bin.file中找到，然后通过脚本再插入DB。
## INSERT 需要将新增数据删除 对应DELETE
## DELETE 需要将删除数据恢复 对应INSERT
## UPDATE 需要将修改数据恢复 对应UPDATE
## 手动读取BINLOG，并找到对应位置和对应事务
## 手动删除除事务外的其他说明语句
INSERT回滚最简单，其次是DELETE的，UPDATE操作比较麻烦。
#!/bin/bash
## INSERT 需要将新增数据删除 对应DELETE
## DELETE 需要将删除数据恢复 对应INSERT
## UPDATE 需要将修改数据恢复 对应UPDATE
## 手动读取BINLOG，并找到对应位置和对应事务
## 手动删除除事务外的其他说明语句 path=`pwd`
tmp_file=$path/tmp.file
cache_file=$path/cache.file
binlog_file=$path/$
sql_file=$path/rollback.sql
columns='$2' display_err()
{
echo -e "\033[45;36m$1\033[0m"
} display_ok()
{
echo -e "\033[40;32m$1\033[0m"
} fun_INSERT()
{
echo -e "\033[40;32m$1\033[0m"
} fun_DELETE()
{ ##### GET ONE TRANSACTION ROWS #####
local row_count=`awk 'BEGIN{print 12 + 2}'` ##### GET COUNtS OF TRANSACTION #####
n=`cat $binlog_file | grep DELETE | wc -l` ##### GET TABLE NAME #####
table=`sed -n '1p' $binlog_file | awk '{print $3}'` for ((i=;i<=n;i++))
do
##### MAKE SQL #####
sed -n '3,'$row_count'p' $binlog_file | awk '{print $1}'| awk -F '=' '{print $2}' | tr -t '\n' ',' | sed 's/,$//' | sed 's/^/(/' | sed 's/$/)/' >> $tmp_file ##### GET POS FOR CUT #####
local pos=`awk 'BEGIN{print '$row_count' + 1 }'` ##### CLEAN cache_file FOR TRANSACTION HAVE DONE #####
sed -n ''$pos',$w '$cache_file'' $binlog_file ##### COPY TO binlog_file #####
/bin/cp -f $cache_file $binlog_file ##### CHECK THE NUMBLE OF TRANSACTION ##### done
##### \n to , #####
sed 's/)(/),(/g' $tmp_file | sed 's/^/insert into t1 values /' | sed 's/)$/);/' > $sql_file
rm -rf $tmp_file
echo done! ##### CHECK THE NUMBLE OF TRANSACTION TO MAKE SQL ##### } case $ in
insert ) fun_INSERT
;;
delete ) fun_DELETE
;;
update ) fun_update
;;
* ) display_err "ERROR!!!! Only insert or delete or update could be input!"
;;
esac