MySQL 慢查询日报（线上 prod8591）2026-10-25

统计概览

Top10 明细（按单条耗时）

高频慢查询模式（指纹聚合，取样 20,357 条）

按条数 Top5

按累计耗时 Top5

备份模式全天 671 条 / 13.19h，未触发深度分析（如条数环比异常翻倍再人工介入）。

触发深度分析的模式（条数>100 或 累计>30min，非备份，最多 8 个）

1. SELECT c.* FROM t8_user_id_card_number_list AS c WHERE c.img_type = ? AND c.status = ? AND c.image != ? AND (c.ocr_result = ? or c.ocr_result is null) ORDER BY —— 3,333 次 × 3.60s = 3.33h

• 代表性 SQL：

SELECT c.* FROM t8_user_id_card_number_list AS c WHERE c.img_type = 2 AND c.status = 2 AND c.image != '' AND (c.ocr_result = '' or c.ocr_result is null) ORDER BY c.id ASC,`c`.`id` LIMIT 1;

• 库表：a8591.t8_user_id_card_number_list；实例 alias：prod-8591/8591-a
• explain SQL：

SELECT c.* FROM `a8591`.`t8_user_id_card_number_list` AS c WHERE c.img_type = 2 AND c.status = 2 AND c.image != '' AND (c.ocr_result = '' or c.ocr_result is null) ORDER BY c.id ASC,`c`.`id` LIMIT 1;

• 执行计划（EXPLAIN 表节点，按 JOIN 顺序）：

• 慢因：扫描行数远大于返回行数(828,611/0)、全索引扫描(c: key=PRIMARY, rows=1，缺更优过滤索引)
• 建议：补索引（按 WHERE 等值列在前、范围列收尾建复合索引）; 用 EXPLAIN 验证 key 命中; 扩复合索引: 把其余 WHERE 等值列追加到现有索引尾部, 避免 10 万+ 行回表
• 可执行 DDL（建议先在低峰期 EXPLAIN 验证再执行）：

ALTER TABLE `a8591`.`t8_user_id_card_number_list` ADD INDEX `idx_img_type_status_ocr_result` (`img_type`, `status`, `ocr_result`);

2. select wallet, sum(if(type = ?, amount, ?)) - sum(if(type = ?, amount, ?)) as balance from a8591.t8_money_trade_log where account = ? and created_at <= ? group —— 153 次 × 49.70s = 2.11h

• 代表性 SQL：

select wallet, sum(if(type = 2, amount, 0)) - sum(if(type = 1, amount, 0)) as balance from `a8591`.`t8_money_trade_log` where `account` = 'A0100' and `created_at` <= '2026-10-25 00:06:48' group by `wallet`;

• 库表：a8591.t8_money_trade_log；实例 alias：prod-8591/8591-a
• explain SQL：

select wallet, sum(if(type = 2, amount, 0)) - sum(if(type = 1, amount, 0)) as balance from `a8591`.`t8_money_trade_log` where `account` = 'A0100' and `created_at` <= '2026-10-25 00:06:48' group by `wallet`;

• 执行计划（EXPLAIN 表节点，按 JOIN 顺序）：

• 慢因：扫描行数远大于返回行数(21,490,015/2)、索引范围扫描量过大(t8_money_trade_log: key=idx_aw, rows=29,516,460)
• 建议：补索引（按 WHERE 等值列在前、范围列收尾建复合索引）; 用 EXPLAIN 验证 key 命中
• 可执行 DDL（建议先在低峰期 EXPLAIN 验证再执行）：

ALTER TABLE `a8591`.`t8_money_trade_log` ADD INDEX `idx_account_created_at` (`account`, `created_at`);

• DDL 注：t8_money_trade_log 与现有索引 idx_aw(account, wallet) 前缀重复(account), 新索引生效后可评估删除旧索引

3. SELECT * FROM t8_mall_stat ORDER BY created_at DESC; —— 1 次 × 6692.43s = 1.86h

• 代表性 SQL：

SELECT *
FROM t8_mall_stat
ORDER BY created_at DESC;

• 库表：a8591.t8_mall_stat；实例 alias：prod-8591/8591-a
• explain SQL：

SELECT *
FROM `a8591`.`t8_mall_stat`
ORDER BY created_at DESC;

• 执行计划（EXPLAIN 表节点，按 JOIN 顺序）：

• 计划标志：⚠️ filesort 排序
• 慢因：EXPLAIN 全表扫描/未用索引、排序未走索引(filesort)、大结果集传输 SELECT * (Bytes_sent=2,791,286,072)
• 建议：补索引（按 WHERE 等值列在前、范围列收尾建复合索引）; 用 EXPLAIN 验证 key 命中; 把 ORDER BY 列纳入复合索引, 或减小排序前扫描范围; 改分页/按需列, 避免 SELECT * 整表捞取; 大导出走离线通道
• 可执行 DDL（建议先在低峰期 EXPLAIN 验证再执行）：

ALTER TABLE `a8591`.`t8_mall_stat` ADD INDEX `idx_created_at` (`created_at`);

4. SELECT a.user_id AS buyerId, IFNULL(c.sellUser,IFNULL(d.sellUser,NULL)) AS sellerId, a.order_number AS orderNo, IFNULL(c.id,IFNULL(d.id,NULL)) AS txNo, a.order_ —— 1 次 × 5140.91s = 1.43h

• 代表性 SQL：

SELECT
	a.user_id AS buyerId,
	IFNULL(c.sellUser,IFNULL(d.sellUser,NULL)) AS sellerId,
	a.order_number AS orderNo,
	IFNULL(c.id,IFNULL(d.id,NULL)) AS txNo,
	a.order_status AS orderStatus,
	IFNULL(c.status,IFNULL(d.status,NULL)) AS txStatus,
  IFNULL(e.type,IFNULL(f.type,NULL)) AS gameType,
  CONVERT(CAST(IFNULL(e.name,IFNULL(f.name,NULL)) AS BINARY) USING utf8) AS gameName,
  CONVERT(CAST(IFNULL(g.type_name,IFNULL(h.type_name,NULL)) AS BINARY) USING utf8) AS goodsType,
  CONVERT(CAST(IFNULL(c.title,IFNULL(d.title,NULL)) AS BINARY) USING utf8) AS goodsTitle,
	a.pay_by AS payMethod,
	a.user_agent AS orderClient,
	IFNULL(c.price,IFNULL(d.price,NULL)) AS price,
	a.voucher_money AS couponAmount,
	i.active_id AS activityName,
	i.coupon_id AS couponId,
	IF(IFNULL(j.set_type,IFNULL(k.set_type,NULL)) = 'specifyBuyer','yes','no') AS IF_special,
	IF(a.order_createtime NOT LIKE '0000-00-00 00:00:00', a.order_createtime, NULL) AS createTime,
	IF(a.order_paytime NOT LIKE '0000-00-00 00:00:00', a.order_paytime, NULL) AS payTime,
	FROM_UNIXTIME( IFNULL(IF(c.actionTime=0,NULL,c.actionTime),IFNULL(IF(d.actionTime=0,NULL,d.actionTime),NULL)), '%Y-%m-%d %H:%i:%s' ) AS actionTime,
	IF(a.order_canceltime NOT LIKE '0000-00-00 00:00:00', a.order_canceltime, NULL) AS cancelTime
FROM
	t8_ware_order a
	LEFT JOIN t8_ware_order_item b ON a.id = b.order_id
	LEFT JOIN ware_item c ON b.ware_id = c.id
	LEFT JOIN ware_item_history d ON b.ware_id = d.id
	LEFT JOIN gameList e ON c.gameName = e.id
	LEFT JOIN gameList f ON d.gameName = f.id
	LEFT JOIN t8_ware_type g ON c.wareType = g.id
	LEFT JOIN t8_ware_type h ON d.wareType = h.id
	LEFT JOIN t8_user_voucher i ON a.voucher_id = i.id
	LEFT JOIN ( SELECT * FROM ware_set WHERE set_type = 'specifyBuyer' ) j ON c.id = j.ware_id
	LEFT JOIN ( SELECT * FROM ware_set WHERE set_type = 'specifyBuyer' ) k ON d.id = k.ware_id
WHERE 
	a.order_createtime >= '2026-01-01' AND a.order_createtime < '2027-01-01'
ORDER BY 
	a.order_createtime DESC;

• 库表：a8591.t8_ware_order；实例 alias：prod-8591/8591-a
• explain SQL：

SELECT * FROM `a8591`.`t8_ware_order` WHERE order_createtime >= '2026-01-01' LIMIT 10

• 执行计划（EXPLAIN 表节点，按 JOIN 顺序）：

• 慢因：扫描行数远大于返回行数(38,553,243/3,355,215)、索引范围扫描量过大(t8_ware_order: key=order_createtime, rows=7,017,602)
• 建议：补索引（按 WHERE 等值列在前、范围列收尾建复合索引）; 用 EXPLAIN 验证 key 命中

5. select t1_design_img.id, t1_design_img.views, t1_design_img.collection, t1_design_img.favor_count, t1_design_img.comment_count, t1_design_img.pin_share_count + —— 2,765 次 × 1.39s = 1.07h

• 代表性 SQL：

select `t1_design_img`.`id`, `t1_design_img`.`views`, `t1_design_img`.`collection`, `t1_design_img`.`favor_count`, `t1_design_img`.`comment_count`, t1_design_img.pin_share_count + t1_design_img.fb_share_count + t1_design_img.line_share_count as share_count, t1_design_img.downs + t1_design_img.save_count + t1_design_img.screenshot_count as download_count from `t1_design_img` inner join `t1_design_work` on `t1_design_img`.`w_id` = `t1_design_work`.`id` where `t1_design_work`.`check_status` = 1 and `t1_design_work`.`is_show` = 1 and `t1_design_work`.`is_hide` = 0 and `t1_design_work`.`publish_time` <= '2026-10-25 00:04:16' and `t1_design_img`.`id` > 21384 order by `t1_design_img`.`id` asc limit 100;

• 库表：t591.t1_design_img；实例 alias：prod-8591/100
• explain SQL：

select `t1_design_img`.`id`, `t1_design_img`.`views`, `t1_design_img`.`collection`, `t1_design_img`.`favor_count`, `t1_design_img`.`comment_count`, t1_design_img.pin_share_count + t1_design_img.fb_share_count + t1_design_img.line_share_count as share_count, t1_design_img.downs + t1_design_img.save_count + t1_design_img.screenshot_count as download_count from `t591`.`t1_design_img` inner join `t591`.`t1_design_work` on `t1_design_img`.`w_id` = `t1_design_work`.`id` where `t1_design_work`.`check_status` = 1 and `t1_design_work`.`is_show` = 1 and `t1_design_work`.`is_hide` = 0 and `t1_design_work`.`publish_time` <= '2026-10-25 00:04:16' and `t1_design_img`.`id` > 21384 order by `t1_design_img`.`id` asc limit 100;

• 执行计划（EXPLAIN 表节点，按 JOIN 顺序）：

• 计划标志：⚠️ filesort 排序、⚠️ 产生临时表
• 慢因：扫描行数远大于返回行数(597,932/100)、DISTINCT/GROUP BY 产生临时表(大范围去重/聚合)、排序未走索引(filesort)
• 建议：补索引（按 WHERE 等值列在前、范围列收尾建复合索引）; 用 EXPLAIN 验证 key 命中; 缩小聚合/去重范围(补时间等过滤条件), 或预聚合到中间表/定时任务; 把 ORDER BY 列纳入复合索引, 或减小排序前扫描范围
• 可执行 DDL（建议先在低峰期 EXPLAIN 验证再执行）：

ALTER TABLE `t591`.`t1_design_img` ADD INDEX `idx_w_id_id` (`w_id`, `id`);
ALTER TABLE `t591`.`t1_design_work` ADD INDEX `idx_check_stat_is_show_is_hide_publish_ti` (`check_status`, `is_show`, `is_hide`, `publish_time`);

• DDL 注：t1_design_img 与现有索引 w_id(w_id) 前缀重复(w_id), 新索引生效后可评估删除旧索引
• DDL 注：t1_design_work 与现有索引 check_status(check_status, is_hide, is_show, publish_time) 前缀重复(check_status), 新索引生效后可评估删除旧索引

6. SELECT * FROM t8_files WHERE (path_s3 = ? OR path_s3 IS NULL) ORDER BY id ASC LIMIT ?; —— 2,836 次 × 1.30s = 1.03h

• 代表性 SQL：

SELECT * FROM `t8_files` WHERE (path_s3 = '' OR path_s3 IS NULL) ORDER BY id ASC LIMIT 10;

• 库表：t8_game.t8_files；实例 alias：prod-8591/8591-b
• explain SQL：

SELECT * FROM `t8_game`.`t8_files` WHERE (path_s3 = '' OR path_s3 IS NULL) ORDER BY id ASC LIMIT 10;

• 执行计划（EXPLAIN 表节点，按 JOIN 顺序）：

• 计划标志：⚠️ filesort 排序
• 慢因：扫描行数远大于返回行数(1,773,457/10)、索引范围扫描量过大(t8_files: key=idx_path_s3_id, rows=2,197,833)、排序未走索引(filesort)
• 建议：补索引（按 WHERE 等值列在前、范围列收尾建复合索引）; 用 EXPLAIN 验证 key 命中; 把 ORDER BY 列纳入复合索引, 或减小排序前扫描范围
• 可执行 DDL（建议先在低峰期 EXPLAIN 验证再执行）：

ALTER TABLE `t8_game`.`t8_files` ADD INDEX `idx_path_s3_id` (`path_s3`, `id`);

• DDL 注：t8_files 与现有索引 idx_path_s3_id(path_s3, 1) 前缀重复(path_s3), 新索引生效后可评估删除旧索引

7. SELECT order_id, ware_id FROM t8_ware_order_item; —— 1 次 × 3403.27s = 56.7min

• 代表性 SQL：

SELECT 
	order_id,
    ware_id
FROM 
	t8_ware_order_item;

• 库表：a8591.t8_ware_order_item；实例 alias：prod-8591/8591-a
• explain SQL：

SELECT 
	order_id,
    ware_id
FROM 
	`a8591`.`t8_ware_order_item`;

• 执行计划（EXPLAIN 表节点，按 JOIN 顺序）：

• 慢因：全索引扫描(t8_ware_order_item: key=user_id, rows=22,988,793，缺更优过滤索引)
• 建议：扩复合索引: 把其余 WHERE 等值列追加到现有索引尾部, 避免 10 万+ 行回表

8. ( SELECT FROM_UNIXTIME( a.recTime, ? ) AS markTime, b.sellUser AS sellerId, b.buyUser AS buyerId, b.id AS wareId, CONVERT ( CAST( b.title AS BINARY ) USING utf8 —— 1 次 × 3011.01s = 50.2min

• 代表性 SQL：

( 
SELECT
	FROM_UNIXTIME( a.recTime, '%Y-%m-%d %H:%i:%s' ) AS markTime,
	b.sellUser AS sellerId,
	b.buyUser AS buyerId,
	b.id AS wareId,
	CONVERT ( CAST( b.title AS BINARY ) USING utf8 ) AS wareTitle,
	b.price AS warePrice,
	b.STATUS AS wareStatus,
	b.wareType AS wareType,
	c.type AS gameType,
	CONVERT ( CAST( c.NAME AS BINARY ) USING utf8 ) AS gameName,
	CONVERT ( CAST( d.NAME AS BINARY ) USING utf8 ) AS service,
	CONVERT ( CAST( e.cname AS BINARY ) USING utf8 ) AS disputeType,
	f.c_status AS progress,
  	f.current_p_status AS disputeStatus,
	FROM_UNIXTIME( b.actionTime, '%Y-%m-%d %H:%i:%s' ) AS actionTime,
	FROM_UNIXTIME( f.lasttime, '%Y-%m-%d %H:%i:%s' ) AS updateTime 
FROM
	serviceRecord a
	LEFT JOIN ware_item AS b ON a.theId = b.id
	LEFT JOIN gameList AS c ON b.gameName = c.id
	LEFT JOIN gameList AS d ON b.serverName = d.id
	LEFT JOIN t8_trouble_ware_type AS e ON a.trouble_type = e.cid
	LEFT JOIN t8_problem_ware AS f ON b.id = f.wareId 
WHERE
	a.type = 'Deal' 
	AND a.tag = '9' 
  AND b.id is not null
)			
UNION
( 
SELECT
	FROM_UNIXTIME( a.recTime, '%Y-%m-%d %H:%i:%s' ) AS markTime,
	b.sellUser AS sellerId,
	b.buyUser AS buyerId,
	b.id AS wareId,
	CONVERT ( CAST( b.title AS BINARY ) USING utf8 ) AS wareTitle,
	b.price AS warePrice,
	b.STATUS AS wareStatus,
	b.wareType AS wareType,
	c.type AS gameType,
	CONVERT ( CAST( c.NAME AS BINARY ) USING utf8 ) AS gameName,
	CONVERT ( CAST( d.NAME AS BINARY ) USING utf8 ) AS service,
	CONVERT ( CAST( e.cname AS BINARY ) USING utf8 ) AS disputeType,
	f.c_status AS progress,
  	f.current_p_status AS disputeStatus,
	FROM_UNIXTIME( b.actionTime, '%Y-%m-%d %H:%i:%s' ) AS actionTime,
	FROM_UNIXTIME( f.lasttime, '%Y-%m-%d %H:%i:%s' ) AS updateTime 
FROM
	serviceRecord a
	LEFT JOIN ware_item_history AS b ON a.theId = b.id
	LEFT JOIN gameList AS c ON b.gameName = c.id
	LEFT JOIN gameList AS d ON b.serverName = d.id
	LEFT JOIN t8_trouble_ware_type AS e ON a.trouble_type = e.cid
	LEFT JOIN t8_problem_ware AS f ON b.id = f.wareId 
WHERE
	a.type = 'Deal' 
	AND a.tag = '9' 
	AND b.id is not null
);

• 库表：a8591.serviceRecord；实例 alias：prod-8591/8591-a
• explain SQL：

SELECT * FROM `a8591`.`serviceRecord` WHERE type = 'Deal' LIMIT 10

• 执行计划（EXPLAIN 表节点，按 JOIN 顺序）：

• 慢因：索引范围扫描量过大(serviceRecord: key=type_theId_recTime, rows=2,725,769)
• 建议：补索引（按 WHERE 等值列在前、范围列收尾建复合索引）; 用 EXPLAIN 验证 key 命中

逐条优化建议汇总

P 行 = 深析模式（指纹聚合，按全天累计耗时降序 P1 最大，是优化主战场）；T 行 = 单条耗时 Top3 的一次性慢 SQL（多为 bi/导出查询，治理手段通常是限流/降频/移到从库，而非建索引）。

*采集脚本 .agents/scripts/mysql_slow_prod_report.py，中间数据 outputs/slow_report_prod8591_20261025.json；全程只读。*