MySQL NDB 批量更新100万行数据

桃扇骨 2024-04-18 00:20 29阅读 0赞

生产NDB数据库中的一张800多万行的大表需要更新部分字段，为减少对数据库性能的影响，编写了一个存储过程来实现：  
1. 导出这张表的主键字段，数据量有100多万条  
select MSISDN from TEST where LastAccessTimeStamp=0  
 into outfile '/tmp/TEST\_out.txt' fields terminated by ',' ;  
  
2. 在其中一个SQL节点，创建两张临时表  
\--创建第一张临时表，用于和生产数据库的源表主键进行关联，以更新部分字段  
create table tmp\_Subscribers\_01(id int not null auto\_increment primary key,MSISDN char(20)) engine=innodb;  
\--向第一张临时表导入之前的导出数据  
load data infile '/tmp/Subscribers\_out.txt' into table tmp\_Subscribers\_01 fields terminated by ',' (MSISDN);  
\--创建第二张临时表，用于记录数据更新的进度  
create table tmp\_Subscribers\_02(id int, MSISDN char(20),cdate datetime) engine=innodb;  
  
3. 编写数据更新的存储过程  
drop procedure proc\_Subscribers\_update;  
 delimiter $$  
 create procedure proc\_Subscribers\_update(IN v\_fetch\_cnt INT, IN v\_sleep\_secs INT)  
 begin  
 DECLARE v\_count INT;  
 DECLARE v\_times INT DEFAULT 1;  
 DECLARE v\_max\_value INT;  
   /\*compute the times that the loop runs\*/  
   select ceil(count(MSISDN)/v\_fetch\_cnt) into v\_count from tmp\_Subscribers\_01;  
   /\*compute the maximum rows that have been already updated\*/  
   WHILE v\_times < v\_count DO  
     select ifnull(max(id),0) into v\_max\_value from tmp\_Subscribers\_02;  
     if v\_max\_value < v\_fetch\_cnt \* v\_count then  
       SET v\_times = 1 + floor(v\_max\_value/v\_fetch\_cnt);  
       update TEST s,tmp\_Subscribers\_01 t set s.LastAccessTimeStamp=1420066800  
       where s.MSISDN=t.MSISDN and t.id > v\_max\_value and t.id <= v\_fetch\_cnt \* v\_times;  
       /\*record the processing rows\*/  
       insert into tmp\_Subscribers\_02 select id, MSISDN, now() from tmp\_Subscribers\_01 where id = v\_fetch\_cnt \* v\_times;  
       select concat('The job',' has already updated ', v\_fetch\_cnt \* v\_times, ' rows..') as Info;  
       select sleep(v\_sleep\_secs);  
     end if;  
     commit;  
   END WHILE;  
   select concat('The job',' is ','finished!') as Info;  
   commit;  
 end$$  
 delimiter ;  
  
\--执行存储过程  
\--其中第一个传入参数为每次更新的行数，第二个参数为每次更新后的休眠时间  
call proc\_Subscribers\_update(10000, 2);

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/26506993/viewspace-2123040/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/26506993/viewspace-2123040/