标签：执行 repeat tree 开始时间 简单 point 利用 对比 weight
cursor 与 insert ...select 对比：
cursor：安全，不会造成死锁，可以在服务运行阶段跑,比较稳定。
insert...select :速度快，但是可能造成死锁，相比cursor能够成倍提升，在服务停止的情况下迁移,速度快
数据迁移案例：
首先数据的迁移绝对不是一朝一夕能够快速迁移完成的 ，如果可以很快完成的 dump便可以搞定，没必要大费周折了。
既然不是一朝一夕能完成的，那么有关键的日志记录表能够良好的反应数据迁移的过程
迁移日志表脚本：
DROP TABLE IF EXISTS `cx_delete_log`; CREATE TABLE `cx_delete_log` ( `id` int(30) NOT NULL, `table_name` varchar(30) DEFAULT NULL, `start_tm` datetime DEFAULT NULL, `end_tm` datetime DEFAULT NULL, `status` int(10) DEFAULT NULL, `pro_create_time` datetime DEFAULT NULL, `pro_end_time` datetime DEFAULT NULL, PRIMARY KEY (`id`) ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4; -- ---------------------------- -- Records of cx_delete_log -- ---------------------------- INSERT INTO `cx_delete_log` VALUES (‘1‘, ‘cx_waybill‘, ‘2017-01-01 00:00:00‘, ‘2017-01-01 00:10:00‘, ‘1‘, ‘2017-03-21 14:06:21‘, null); ALTER TABLE `cx_delete_log` ADD INDEX `index_end_date` (`end_tm`) USING BTREE ;
首先迁移的增量字段确定。这里存在一个最优解，即你每分钟的数据量，与你mysql的性能，我们可以在此之间寻找到一个平衡点（cursor与insert...select并不相同），设置每次迁移多次时间段的数据，这样能，最大效率的利用我们数据库的性能。
我们插入了一条日志数据，有这个来控制每次数据迁移的间隔时间段，与开始时间
如：2017-01-01 00:00:00 start_tm数据迁移增量字段的开始时间，2017-01-01 00:10:00 end_tm数据增量字段迁移的结束时间，时间间隔10分钟。
状态为1代表成功，2017-03-21 14:06:21 代表执行这段时间数据迁移的执行时间，pro_end_time代表执行这段时间数据迁移的执行完毕时间时间。
存储过程：
存储过程分为两个部分，1.日志的写入2.迁移操作的进行
日志写入：
delimiter | drop procedure if exists batch_move_cx_waybill_partition_by_minute; create procedure batch_move_cx_waybill_partition_by_minute() begin DECLARE proNo INTEGER; DECLARE startDate TIMESTAMP; DECLARE endDate TIMESTAMP; DECLARE nextDate TIMESTAMP; DECLARE pro_endDate TIMESTAMP; SET pro_endDate=‘2017-03-20 00:00:00‘; REPEAT SELECT ID, end_tm,DATE_ADD(end_tm,INTERVAL ‘00:10:00‘ hour_second) INTO proNo,startDate ,endDate from cx_delete_log ORDER BY end_tm desc limit 1; IF curtime()>‘06:00:00‘ THEN SET startDate =pro_endDate; END IF; IF startDate <pro_endDate THEN INSERT into cx_delete_log(id,table_name,start_tm,end_tm,status,pro_create_time)VALUES(proNo+1,‘cx_waybill‘,startDate,endDate,‘0‘,SYSDATE()); CALL cursor_move_cx_waybill_partition(startDate); update cx_delete_log set pro_end_time =SYSDATE() , `status`=1 where id=proNo+1; COMMIT; END IF; UNTIL startDate >=pro_endDate END REPEAT; end | delimiter ;
迁移操作的进行：
实例为cursor，insert..select较为简单，直接带入就行，就不做示例了
delimiter || drop procedure if exists cursor_move_cx_waybill_partition; create procedure cursor_move_cx_waybill_partition(IN startDate_tmp TIMESTAMP) begin DECLARE isEXist INTEGER; DECLARE WAYBILLNO_TMP VARCHAR(32); DECLARE RESOURCECODE_TMP VARCHAR(10); DECLARE APPOINTMENTDDELIVERYTIME_TMP TIMESTAMP; DECLARE APPOINTMENTDDELIVERYENDTIME_TMP TIMESTAMP; DECLARE EXPECTEDDELIVERYTIME_TMP TIMESTAMP; DECLARE ORIGINATEID_TMP VARCHAR(32); DECLARE DESTINATIONID_TMP VARCHAR(32); DECLARE WAYBILLSTATUS_TMP VARCHAR(4); DECLARE VERSIONNO_TMP INTEGER; DECLARE STATUS_TMP CHAR(1); DECLARE FAILSTARTUS_TMP VARCHAR(10); DECLARE FAILCAUSEDESC_TMP VARCHAR(500); DECLARE MEMBERID_TMP VARCHAR(32); DECLARE HIDEFLAG_TMP VARCHAR(4); DECLARE CREATEUSERID_TMP VARCHAR(32); DECLARE CREATETIME_TMP TIMESTAMP; DECLARE UPDATEUSERID_TMP VARCHAR(32); DECLARE UPDATETIME_TMP TIMESTAMP; DECLARE APPOINTMENTNO_TMP VARCHAR(32); DECLARE SOURCECITY_TMP VARCHAR(32); DECLARE DESTCITY_TMP VARCHAR(32); DECLARE SOURCECITYCODE_TMP VARCHAR(20); DECLARE DESTCITYCODE_TMP VARCHAR(20); DECLARE EMPCODE_TMP VARCHAR(10); DECLARE RECEMPCODE_TMP VARCHAR(10); DECLARE RECMEMBERID_TMP VARCHAR(32); DECLARE WAYBILLFEE_TMP DOUBLE(10,0); DECLARE MOBILE_TMP VARCHAR(20); DECLARE RECMOBILE_TMP VARCHAR(20); DECLARE RECTIME_TMP TIMESTAMP; DECLARE ENDTIME_TMP TIMESTAMP; DECLARE EXPECTMEMBERID_TMP VARCHAR(32); DECLARE ORIGINATE_TMP VARCHAR(500); DECLARE DESTINATION_TMP VARCHAR(500); DECLARE PAYTYPE_TMP VARCHAR(4); DECLARE SECRECYTYPE_TMP VARCHAR(32); DECLARE SEND_EVALUATE_TMP VARCHAR(10); DECLARE PRODUCT_TYPE_TMP VARCHAR(32); DECLARE SERVICES_PROD_CODE_TMP VARCHAR(200); DECLARE CHANGERECORD_TMP VARCHAR(32); DECLARE IS_ISSUED_PREFRENCE_TMP VARCHAR(10); DECLARE REC_EVALUATE_TMP VARCHAR(10); DECLARE done INTEGER; DECLARE bill_cursor CURSOR for SELECT WAYBILLNO, RESOURCECODE , APPOINTMENTDDELIVERYTIME , APPOINTMENTDDELIVERYENDTIME , EXPECTEDDELIVERYTIME ,ORIGINATEID , DESTINATIONID , WAYBILLSTATUS , VERSIONNO , STATUS, FAILSTARTUS, FAILCAUSEDESC, MEMBERID, HIDEFLAG, CREATEUSERID, CREATETIME, UPDATEUSERID, UPDATETIME, APPOINTMENTNO, SOURCECITY, DESTCITY, SOURCECITYCODE, DESTCITYCODE, EMPCODE, RECEMPCODE, RECMEMBERID,WAYBILLFEE, MOBILE, RECMOBILE, RECTIME, ENDTIME, EXPECTMEMBERID, ORIGINATE, DESTINATION, PAYTYPE, SECRECYTYPE, SEND_EVALUATE, PRODUCT_TYPE, SERVICES_PROD_CODE, CHANGERECORD, IS_ISSUED_PREFRENCE, REC_EVALUATE FROM cx_waybill WHERE CREATETIME > startDate_tmp AND CREATETIME <= DATE_ADD(startDate_tmp,INTERVAL ‘00:10:00‘ hour_second); DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = 1; open bill_cursor; loop_bill:LOOP IF done = 1 THEN LEAVE loop_bill; END IF; FETCH bill_cursor INTO WAYBILLNO_TMP, RESOURCECODE_TMP, APPOINTMENTDDELIVERYTIME_TMP, APPOINTMENTDDELIVERYENDTIME_TMP,EXPECTEDDELIVERYTIME_TMP,ORIGINATEID_TMP, DESTINATIONID_TMP, WAYBILLSTATUS_TMP, VERSIONNO_TMP, STATUS_TMP, FAILSTARTUS_TMP, FAILCAUSEDESC_TMP, MEMBERID_TMP, HIDEFLAG_TMP, CREATEUSERID_TMP, CREATETIME_TMP, UPDATEUSERID_TMP, UPDATETIME_TMP, APPOINTMENTNO_TMP, SOURCECITY_TMP, DESTCITY_TMP, SOURCECITYCODE_TMP, DESTCITYCODE_TMP, EMPCODE_TMP, RECEMPCODE_TMP, RECMEMBERID_TMP,WAYBILLFEE_TMP, MOBILE_TMP, RECMOBILE_TMP, RECTIME_TMP, ENDTIME_TMP, EXPECTMEMBERID_TMP, ORIGINATE_TMP, DESTINATION_TMP, PAYTYPE_TMP, SECRECYTYPE_TMP, SEND_EVALUATE_TMP, PRODUCT_TYPE_TMP, SERVICES_PROD_CODE_TMP, CHANGERECORD_TMP, IS_ISSUED_PREFRENCE_TMP, REC_EVALUATE_TMP; IF WAYBILLNO_TMP is NULL THEN LEAVE loop_bill; END IF; SELECT count(1) INTO isEXist from cx_waybill_partition w where w.WAYBILLNO=WAYBILLNO_TMP; IF isEXist >0 THEN LEAVE loop_bill; ELSE INSERT INTO cx_waybill_PARTITION (WAYBILLNO, RESOURCECODE, APPOINTMENTDDELIVERYTIME, APPOINTMENTDDELIVERYENDTIME,EXPECTEDDELIVERYTIME,ORIGINATEID, DESTINATIONID, WAYBILLSTATUS, VERSIONNO, STATUS, FAILSTARTUS, FAILCAUSEDESC, MEMBERID, HIDEFLAG, CREATEUSERID, CREATETIME, UPDATEUSERID, UPDATETIME, APPOINTMENTNO, SOURCECITY, DESTCITY, SOURCECITYCODE, DESTCITYCODE, EMPCODE, RECEMPCODE, RECMEMBERID,WAYBILLFEE, MOBILE, RECMOBILE, RECTIME, ENDTIME, EXPECTMEMBERID, ORIGINATE, DESTINATION, PAYTYPE, SECRECYTYPE, SEND_EVALUATE, PRODUCT_TYPE, SERVICES_PROD_CODE, CHANGERECORD, IS_ISSUED_PREFRENCE, REC_EVALUATE)VALUES(WAYBILLNO_TMP, RESOURCECODE_TMP, APPOINTMENTDDELIVERYTIME_TMP, APPOINTMENTDDELIVERYENDTIME_TMP,EXPECTEDDELIVERYTIME_TMP,ORIGINATEID_TMP, DESTINATIONID_TMP, WAYBILLSTATUS_TMP, VERSIONNO_TMP, STATUS_TMP, FAILSTARTUS_TMP, FAILCAUSEDESC_TMP, MEMBERID_TMP, HIDEFLAG_TMP, CREATEUSERID_TMP, CREATETIME_TMP, UPDATEUSERID_TMP, UPDATETIME_TMP, APPOINTMENTNO_TMP, SOURCECITY_TMP, DESTCITY_TMP, SOURCECITYCODE_TMP, DESTCITYCODE_TMP, EMPCODE_TMP, RECEMPCODE_TMP, RECMEMBERID_TMP,WAYBILLFEE_TMP, MOBILE_TMP, RECMOBILE_TMP, RECTIME_TMP, ENDTIME_TMP, EXPECTMEMBERID_TMP, ORIGINATE_TMP, DESTINATION_TMP, PAYTYPE_TMP, SECRECYTYPE_TMP, SEND_EVALUATE_TMP, PRODUCT_TYPE_TMP, SERVICES_PROD_CODE_TMP, CHANGERECORD_TMP, IS_ISSUED_PREFRENCE_TMP, REC_EVALUATE_TMP); END IF; end Loop; close bill_cursor; COMMIT; end || delimiter ;
定时器的创建：
每天凌晨一点执行：
---轮循删除定时器 drop event if exists waybill_move_event; create event waybill_move_event on schedule every 1 DAY STARTS ‘2017-01-17 01:00:00‘ on completion preserve ENABLE do call batch_move_cx_waybill_partition_by_minute(); alter event waybill_move_event on completion preserve enable;
其实，在中间踩过一些坑，但是很多都是百度都能解决的 。事实证明，这样的数据迁移，其速度并不慢，而且性能相对稳定，在后续的观察中，并未对数据库造成太多的压力，可看作种蚂蚁搬家。
标签：执行 repeat tree 开始时间 简单 point 利用 对比 weight
原文地址：http://www.cnblogs.com/lianshan/p/6724987.html
