mysql分区表设计（一）
数据库中有多张表要做分区表（按日期分区），但是这些表用于分区的日期字段记录形式并不统一，有如下三种类型：
PARTITION par_20151225 VALUES LESS THAN (736323) ENGINE = InnoDB PARTITION par_20151010 VALUES LESS THAN (1444492800) ENGINE = InnoDB PARTITION par_20151006 VALUES LESS THAN (1452480) ENGINE = InnoDB
首先创建一个元数据信息表，用于记录要分区的表以及分区类型，如下：
mysql> select * from op_tb_partition; +-------------------------------+----------+ | tb_name | timetype | +-------------------------------+----------+ | NL_APP_ACTION_ERROR_TRACE | unixtime | | NL_APP_ACTION_SQL_TRACE | unixtime | | NL_APP_ACTION_TRACE | unixtime | | NL_MOB_APP_CRASH_DATA | usertime | | NL_MOB_APP_INTERACTION_TRACE | unixtime | | NL_NET_MOB_PAGE_DATA | unixtime | | NL_NET_MOB_PING_DATA | unixtime | | NL_NET_PAGE_DATA | unixtime | | NL_NET_PING_DATA | unixtime | | NL_MOB_APP_ERROR_TRACE | datetime | | NL_MOB_APP_SOCKET_ERROR_TRACE | datetime | +-------------------------------+----------+ 11 rows in set (0.00 sec)
此外，还需创建一张日志表，用于记录存储过程执行的详细信息，如下：
mysql> desc change_partition_log; +------------+---------------+------+-----+-------------------+-----------------------------+ | Field | Type | Null | Key | Default | Extra | +------------+---------------+------+-----+-------------------+-----------------------------+ | tm_base | timestamp | NO | | CURRENT_TIMESTAMP | on update CURRENT_TIMESTAMP | | op_type | varchar(20) | YES | | NULL | | | change_sql | varchar(2000) | YES | | NULL | | +------------+---------------+------+-----+-------------------+-----------------------------+ 3 rows in set (0.00 sec)
为了是维护工作变的简单一些，希望能用一个存储过程来实现所有类型表的分区维护，因此编写了如下的存储过程：
DELIMITER $$
DROP PROCEDURE IF EXISTS lens_mobapp_trace.add_partition_original;
CREATE PROCEDURE `add_partition_original`()
BEGIN
DECLARE table_name varchar(255) DEFAULT "";
DECLARE tmp_sql varchar(255) DEFAULT "";
DECLARE max_day_will int DEFAULT 0;
DECLARE max_day_now int DEFAULT 0;
DECLARE difference int DEFAULT 0;
DECLARE i int DEFAULT 1;
DECLARE tmp_day_value int DEFAULT 0;
DECLARE tmp_day int DEFAULT 1;
DECLARE alter_sql varchar(255) DEFAULT "";
DECLARE par_name varchar(255) DEFAULT "";
DECLARE done int DEFAULT - 1;
DECLARE myCursor CURSOR FOR
SELECT
tb_name
FROM op_tb_partition;
DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = 1;
SET @tmp_sql = CONCAT("select to_days(date_add(now(),interval 6 day)) into @max_day_will");
PREPARE stmt FROM @tmp_sql;
EXECUTE stmt;
OPEN myCursor;
myLoop:
LOOP
FETCH myCursor INTO table_name;
IF done = 1 THEN
LEAVE myLoop;
END IF;
set @tmp_sql = concat("SELECT timetype FROM op_tb_partition WHERE tb_name = '", table_name, "' into @timetype");
prepare stmt from @tmp_sql;
execute stmt;
case
when @timetype="datetime" then
set @tmp_sql = concat("SELECT max(substring(partition_name,5)) FROM INFORMATION_SCHEMA.partitions WHERE TABLE_NAME = '", table_name, "' and TABLE_SCHEMA ='lens_mobapp_trace' into @max_day_now");
prepare stmt from @tmp_sql;
execute stmt;
set @tmp_sql = concat("select @max_day_will - to_days('",@max_day_now,"') into @difference");
prepare stmt from @tmp_sql;
execute stmt;
while i <= @difference do
set @tmp_sql = concat("select date_format(date_add('",@max_day_now,"' ,interval ",i," day), '%Y%m%d') into @tmp_day");
prepare stmt from @tmp_sql;
execute stmt;
set @tmp_sql = concat("select to_days(date_format(date_add('",@max_day_now,"' ,interval 1+",i," day), '%Y%m%d')) into @tmp_day_value");
prepare stmt from @tmp_sql;
execute stmt;
set @par_name = concat("par_",@tmp_day);
SET @alter_sql = CONCAT("alter table ", table_name, " add partition(partition ", @par_name, " values less than (", @tmp_day_value, "))");
insert into change_partition_log values( now(), "add_trace", @alter_sql);
prepare stmt from @alter_sql;
execute stmt;
set i = i+1;
end while;
set i = 1;
when @timetype="usertime" then
set @tmp_sql = concat("SELECT max(substring(partition_name,5)) FROM INFORMATION_SCHEMA.partitions WHERE TABLE_NAME = '", table_name, "' and TABLE_SCHEMA ='lens_mobapp_trace' into @max_day_now");
prepare stmt from @tmp_sql;
execute stmt;
set @tmp_sql = concat("select @max_day_will - to_days('",@max_day_now,"') into @difference");
prepare stmt from @tmp_sql;
execute stmt;
while i <= @difference do
set @tmp_sql = concat("select date_format(date_add('",@max_day_now,"' ,interval ",i," day), '%Y%m%d') into @tmp_day");
prepare stmt from @tmp_sql;
execute stmt;
set @tmp_sql = concat("select nl_to_timestamp(date_format(date_add('",@max_day_now,"' ,interval 1+",i," day), '%Y%m%d')) into @tmp_day_value");
prepare stmt from @tmp_sql;
execute stmt;
set @par_name = concat("par_",@tmp_day);
SET @alter_sql = CONCAT("alter table ", table_name, " add partition(partition ", @par_name, " values less than (", @tmp_day_value, "))");
insert into change_partition_log values( now(),"add_trace" , @alter_sql);
prepare stmt from @alter_sql;
execute stmt;
set i = i+1;
end while;
set i = 1;
when @timetype="unixtime" then
set @tmp_sql = concat("SELECT max(substring(partition_name,5)) FROM INFORMATION_SCHEMA.partitions WHERE TABLE_NAME = '", table_name, "' and TABLE_SCHEMA ='lens_mobapp_trace' into @max_day_now");
prepare stmt from @tmp_sql;
execute stmt;
set @tmp_sql = concat("select @max_day_will - to_days('",@max_day_now,"') into @difference");
prepare stmt from @tmp_sql;
execute stmt;
while i <= @difference do
set @tmp_sql = concat("select date_format(date_add('",@max_day_now,"' ,interval ",i," day), '%Y%m%d') into @tmp_day");
prepare stmt from @tmp_sql;
execute stmt;
set @tmp_sql = concat("select unix_timestamp(date_format(date_add('",@max_day_now,"' ,interval 1+",i," day), '%Y%m%d')) into @tmp_day_value");
prepare stmt from @tmp_sql;
execute stmt;
set @par_name = concat("par_",@tmp_day);
SET @alter_sql = CONCAT("alter table ", table_name, " add partition(partition ", @par_name, " values less than (", @tmp_day_value, "))");
insert into change_partition_log values( now(),"add_trace" , @alter_sql);
prepare stmt from @alter_sql;
execute stmt;
set i = i+1;
end while;
set i = 1;
end case;
END LOOP myLoop;
CLOSE myCursor;
END$$
DELIMITER ;
由于业务需求，这些表的数据只需要保留3个月，再次编写删除分区的存储过程，如下：
DELIMITER $$
DROP PROCEDURE IF EXISTS lens_mobapp_trace.del_partition_original; CREATE PROCEDURE `del_partition_original`()
BEGIN
declare table_name varchar(255) default 0;
declare tmp_sql varchar(255) default "";
declare max_par_now varchar(255) default "";
declare par_del_num varchar(255) default "";
declare par_del varchar(255) default "";
declare alter_sql varchar(255) default "";
declare i int default 0;
declare done int default -1;
declare myCursor cursor for select tb_name from op_tb_partition;
declare continue handler for not found set done = 1;
open myCursor;
myLoop:LOOP
fetch myCursor into table_name;
if done = 1 then
leave myLoop;
end if;
set @tmp_sql = concat("select count(*) FROM INFORMATION_SCHEMA.partitions WHERE TABLE_NAME = '",table_name,"' and TABLE_SCHEMA ='lens_mobapp_trace' into @max_par_now");
prepare stmt from @tmp_sql;
execute stmt;
set @par_del_num = @max_par_now-90-7;
while i < @par_del_num do
set @tmp_sql = concat("select partition_name from INFORMATION_SCHEMA.partitions where TABLE_SCHEMA ='lens_mobapp_trace' and table_name='",table_name,"' order by partition_name limit 1 into @par_del");
prepare stmt from @tmp_sql;
execute stmt;
set @alter_sql = concat("alter table ",table_name," drop partition ",@par_del);
prepare stmt from @alter_sql;
execute stmt;
insert into change_partition_log values(now(),"drop_trace", concat(i,"--",@par_del,"--",@alter_sql));
set i = i+1;
end while;
set i = 0;
end loop myLoop;
close myCursor;
END$$
DELIMITER ;
最后编写定期执行上述存储过程的调度event，如下：
DELIMITER ;; CREATE EVENT lens_mobapp_trace.`manager_partition_original` ON SCHEDULE EVERY 1 DAY STARTS '2013-12-24 00:30:00' ON COMPLETION PRESERVE ENABLE DO BEGIN call add_partition_original; call del_partition_original; END ;; DELIMITER ;