mysql调优随笔

随着业务量增加，客户数增长，最近应用运行过程出现CPU超95%，此时慢SQL严重影响应用体验，为保障应用能够提供更好的性能，提升客户体验，增加客户黏性，我准备对使用频繁的慢SQL、执行时间过长的慢SQL进行调优，以下是一次调优案例。

以下为存在性能问题的SQL（处理过）：

SELECT 
    XXX
FROM
    ro
        LEFT JOIN
    o ON ro.main_id = o.main_id
        LEFT JOIN
    dark ON dark.main_id = o.main_id
        LEFT JOIN
    c ON ro.main_id = c.main_id
        LEFT JOIN
    (SELECT 
        main_id, COUNT(1) AS CA, create_time
    FROM
        tcr
    WHERE
        CC_TYPE <> 25
    GROUP BY main_id) all_c ON ro.main_id = all_c.main_id
        LEFT JOIN
    (SELECT 
        main_id, begin_time AS latestTime
    FROM
        tlootr
    WHERE
        ISNULL(end_time)
    GROUP BY main_id
    ORDER BY begin_time DESC) lor ON ro.main_id = lor.main_id
        LEFT JOIN
    (SELECT 
        main_id, COUNT(1) AS overdueCount
    FROM
        tlootr
    GROUP BY main_id) ooc ON ro.main_id = ooc.main_id
WHERE
    1 = 1 AND ro.collection_order_type = 1
        AND dark.order_id IS NULL
        AND c.user_name LIKE '%丁牛%'
GROUP BY ro.main_id
ORDER BY lor.latestTime DESC , ro.main_id DESC
LIMIT 20

经过查看执行计划，发现SQL存在索引未被使用情况，导致查询数据量超级大，返回时长超过10秒。进一步分析，主要由group by滥用造成。由于group by 索引字段（main_id），导致索引字段（main_id）无法使用，于是，将索引字段的group by去掉，将count算法，提到select部分进行，并将group by的字段，改为主表的字段，优化后，数据返回时长小于1秒。
总结：
group by会导致索引失效，要避免在索引字段使用group by。