2019獨(dú)角獸企業(yè)重金招聘Python工程師標(biāo)準(zhǔn)>>>

版本?5.7.20-19-log

mysql 官網(wǎng)bug?https://bugs.mysql.com/bug.php?id=69721 ?

主庫和從庫配置一樣，但是執(zhí)行計(jì)劃不一樣。

sql

SELECT ORD.*, INFO.NAME AS SHOPNAME, INFO.CONTACT_PHONE AS SHOPPHONE, IF(SS.ID IS NULL, 0, 0)AS REFUNDSTATUS
FROM SHOP_ORDER_INFO ORD
LEFT JOIN SHOP_INFO INFO ON INFO.ID=ORD.SHOP_ID
LEFT JOIN(SELECT ID, ORDER_ID FROM SHOP_ORDER_AFTER_SALES
WHERE STATUS='REFUND_SUCCESS' GROUP BY ORDER_ID)SS ON ORD.ID=SS.ORDER_ID
WHERE ORD.DEL_FLAG='0'
ORDER BY ORD.CREATE_DATE DESC
limit 0,10

主庫執(zhí)行計(jì)劃

1 PRIMARY ORD ALL 36547 10 Using where; Using temporary; Using filesort
1 PRIMARY INFO eq_ref PRIMARY PRIMARY 194 fengdai_shop.ORD.shop_id 1 100
1 PRIMARY <derived2> ALL 3 100 Using where; Using join buffer (Block Nested Loop)
2 DERIVED SHOP_ORDER_AFTER_SALES index idx_order_id idx_order_id 195 36 10 Using where

從庫執(zhí)行計(jì)劃

1 PRIMARY ORD ALL 43970 10 Using where; Using filesort
1 PRIMARY INFO eq_ref PRIMARY PRIMARY 194 fengdai_shop.ORD.shop_id 1 100
1 PRIMARY <derived2> ref <auto_key0> <auto_key0> 195 fengdai_shop.ORD.id 3 100
2 DERIVED SHOP_ORDER_AFTER_SALES index idx_order_id idx_order_id 195 368 10 Using where

主庫比從庫慢了5倍

執(zhí)行器的選擇有問題 明明派生表生成索引速度更快，但是主庫選擇了Block Nested Loop算法。

解決方法：

1 sql換一種方式

2 當(dāng)前會話把這種算法關(guān)掉

SET SESSION optimizer_switch ='block_nested_loop=off';

select @@optimizer_switch;

?

?

?

轉(zhuǎn)載于:https://my.oschina.net/zhuzhenyu/blog/2873684

總結(jié)

以上是生活随笔為你收集整理的mysql的一个bug Block Nested Loop的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。