mysql優化 - mysql慢查詢問題
問題描述
表結構(表的數據量在100萬左右)CREATE TABLE `t_user_notification` ( `notify_id` bigint(20) NOT NULL, `user_id` bigint(20) NOT NULL, `notify` varchar(1024) COLLATE utf8mb4_unicode_ci DEFAULT NULL, `priority` tinyint(1) NOT NULL DEFAULT ’0’, `insert_time` datetime DEFAULT NULL, PRIMARY KEY (`notify_id`), KEY `idx_user_notification__priority_user_id` (`user_id`,`priority`), KEY `idx_userid_notifyid` (`user_id`,`notify_id`)) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_cisql語句一:
select notify_id, notify from t_user_notification where user_id = 1 and notify_id > 1 and priority = 1 order by notify_id asc limit 20G
這條語句執行大概花了10s,explain結果如下:
*************************** 1. row *************************** id: 1 select_type: SIMPLEtable: t_user_notification_0399 type: index_mergepossible_keys: PRIMARY,idx_user_notification__priority_user_id,idx_userid_notifyid key: idx_user_notification__priority_user_id,PRIMARY key_len: 17,8 ref: NULL rows: 22629Extra: Using intersect(idx_user_notification__priority_user_id,PRIMARY); Using where; Using filesort語句二:
SELECT notify_id, notify, priority FROM t_user_notificationWHERE user_id = 1AND ((priority = 1 AND notify_id > 123) OR (priority = 0 AND notify_id > 345))ORDER BY notify_id ASC LIMIT 20G
這條語句執行時間不定,偶爾出現超時
問題如何優化索引使sql執行速度加快
問題解答
回答1:第一個語句:從sql語句本身來說,是單表查詢,已經使用了索引,且索引中選擇性較高的字段userid也放到了前面,感覺沒什么可以優化的地方。從業務邏輯來說,可以考慮是否能增加insert_time的條件,并創建(userid, insert_time)的索引,這樣通過索引能夠過濾掉的記錄更多。還有一個思路就是拿空間換時間,創建一個包括查詢結果和條件所有字段的索引,如(user_id, priority, notify_id, notify),這樣只用查詢索引就能得到結果,避免了回表的查詢成本。另外就是看業務上是否允許不排序,這樣可以減少mysql做排序的成本。
第二個語句:除了第一個語句的優化思路外,可以考慮重寫sql,讓mysql使用索引idx_userid_notifyid
SELECT notify_id, notify, priority FROM t_user_notificationWHERE user_id = 1AND notify_id > 123AND (priority = 1 OR (priority = 0 AND notify_id > 345))ORDER BY notify_id ASC LIMIT 20回答2:
語句1:
select notify_id, notify from t_user_notification where user_id = 1 and priority = 1 and notify_id > 1 order by notify_id asc limit 20G
語句2:不懂怎么優化,但應該盡量避免用OR。
回答3:比較奇怪,為什么不用自增主鍵?然后看樣子應該用了分區吧,按user_id分嗎?
這個索引:
KEY `idx_user_notification__priority_user_id` (`user_id`,`priority`)
改成這樣
KEY `idx_user_notification__priority_user_id` (`user_id`,`priority`,`notify_id`)
select t1.notify_id, t1.notify from t_user_notification t1, (select notify_id from t_user_notification where user_id = 1 and notify_id > 1 and priority = 1 order by notify_id asc limit 20) t2where t1.notify_id = t2.notify_id;回答4:
語句2改成這樣試試
SELECT notify_id, notify, priorityFROM t_user_notificationJOIN(SELECT notify_id FROM t_user_notificationWHERE user_id = 1AND notify_id > 123AND (priority = 1 OR (priority = 0 AND notify_id > 345))ORDER BY notify_id ASC LIMIT 20)as tmp using(notify_id)回答5:
強制使用索引吧,你這里涉及到索引合并 select * from xxx FORCE INDEX(primary_key)
相關文章:
1. 關docker hub上有些鏡像的tag被標記““This image has vulnerabilities””2. Span標簽3. css - 求推薦適用于vue2的框架 像bootstrap這種類型的4. docker-machine添加一個已有的docker主機問題5. android新手一枚,android使用httclient獲取服務器端數據失敗,但是用java工程運行就可以成功獲取。6. css - 關于div自適應問題,大家看圖吧,說不清7. angular.js使用$resource服務把數據存入mongodb的問題。8. java - Collections類里的swap函數,源碼為什么要新定義一個final的List型變量l指向傳入的list?9. python - django如何每次調用標簽的時候都取隨機數據10. SessionNotFoundException:會話ID為null。調用quit()后使用WebDriver嗎?(硒)
