SELECT
SCHEMA_NAME as '数据库名',
LEFT(DIGEST_TEXT, 150) as 'SQL语句摘要',
COUNT_STAR as '执行次数',
ROUND(AVG_TIMER_WAIT/1000000000000, 4) as '平均执行时间(秒)',
ROUND(SUM_TIMER_WAIT/1000000000000, 4) as '总执行时间(秒)',
ROUND(MAX_TIMER_WAIT/1000000000000, 4) as '最大执行时间(秒)',
ROUND(MIN_TIMER_WAIT/1000000000000, 4) as '最小执行时间(秒)',
CASE
WHEN COUNT_STAR > 0 THEN ROUND(SUM_ROWS_EXAMINED/COUNT_STAR, 0)
ELSE 0
END as '平均扫描行数',
CASE
WHEN COUNT_STAR > 0 THEN ROUND(SUM_ROWS_SENT/COUNT_STAR, 0)
ELSE 0
END as '平均返回行数',
FIRST_SEEN as '首次出现',
LAST_SEEN as '最后出现'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
AND DIGEST_TEXT IS NOT NULL
AND COUNT_STAR > 0
AND AVG_TIMER_WAIT > 1000000000 -- 平均执行时间超过1毫秒
ORDER BY AVG_TIMER_WAIT DESC
LIMIT 20;
2. 按总执行时长排序的慢查询
sql复制代码
SELECT
SCHEMA_NAME as '数据库名',
LEFT(DIGEST_TEXT, 120) as 'SQL语句摘要',
COUNT_STAR as '执行次数',
ROUND(AVG_TIMER_WAIT/1000000000000, 4) as '平均执行时间(秒)',
ROUND(SUM_TIMER_WAIT/1000000000000, 4) as '总执行时间(秒)',
ROUND(MAX_TIMER_WAIT/1000000000000, 4) as '最大执行时间(秒)',
CASE
WHEN COUNT_STAR > 0 THEN ROUND(SUM_ROWS_EXAMINED/COUNT_STAR, 0)
ELSE 0
END as '平均扫描行数',
-- MySQL 5.7 不支持窗口函数，使用子查询计算比例
ROUND(
(SUM_TIMER_WAIT /
(SELECT SUM(SUM_TIMER_WAIT)
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
)
) * 100, 2
) as '占总时间比例(%)'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
AND DIGEST_TEXT IS NOT NULL
AND COUNT_STAR > 0
AND SUM_TIMER_WAIT > 5000000000000 -- 总执行时间超过5秒
ORDER BY SUM_TIMER_WAIT DESC
LIMIT 20;
3. MySQL 5.7 慢查询配置检查
sql复制代码
-- 查看慢查询相关配置
SHOW VARIABLES LIKE '%slow%';
-- 查看慢查询时间阈值
SHOW VARIABLES LIKE 'long_query_time';
-- 查看performance_schema配置状态
SHOW VARIABLES LIKE 'performance_schema';
-- 检查performance_schema相关表是否启用
SELECT TABLE_NAME, ENABLED
FROM performance_schema.setup_instruments
WHERE NAME LIKE '%statement%'
AND NAME LIKE '%sql%';
-- 检查consumers是否启用
SELECT NAME, ENABLED
FROM performance_schema.setup_consumers
WHERE NAME LIKE '%statements%';
4. 扫描行数分析（找出全表扫描）
sql复制代码
SELECT
SCHEMA_NAME as '数据库名',
LEFT(DIGEST_TEXT, 100) as 'SQL语句摘要',
COUNT_STAR as '执行次数',
ROUND(AVG_TIMER_WAIT/1000000000000, 4) as '平均执行时间(秒)',
SUM_ROWS_EXAMINED as '总扫描行数',
CASE
WHEN COUNT_STAR > 0 THEN ROUND(SUM_ROWS_EXAMINED/COUNT_STAR, 0)
ELSE 0
END as '平均扫描行数',
SUM_ROWS_SENT as '总返回行数',
CASE
WHEN SUM_ROWS_EXAMINED > 0 THEN ROUND(SUM_ROWS_SENT/SUM_ROWS_EXAMINED*100, 2)
ELSE 0
END as '扫描效率(%)',
CASE
WHEN SUM_ROWS_SENT > 0 THEN ROUND(SUM_ROWS_EXAMINED/SUM_ROWS_SENT, 0)
ELSE SUM_ROWS_EXAMINED
END as '扫描/返回比例'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
AND DIGEST_TEXT IS NOT NULL
AND COUNT_STAR > 0
AND SUM_ROWS_EXAMINED > 100000 -- 扫描行数超过10万
ORDER BY SUM_ROWS_EXAMINED DESC
LIMIT 15;
5. 高频执行的慢查询
sql复制代码
SELECT
SCHEMA_NAME as '数据库名',
LEFT(DIGEST_TEXT, 120) as 'SQL语句摘要',
COUNT_STAR as '执行次数',
ROUND(AVG_TIMER_WAIT/1000000000000, 4) as '平均执行时间(秒)',
ROUND(SUM_TIMER_WAIT/1000000000000, 4) as '总执行时间(秒)',
ROUND(AVG_TIMER_WAIT/1000000, 2) as '平均执行时间(毫秒)',
CASE
WHEN COUNT_STAR > 0 THEN ROUND(SUM_ROWS_EXAMINED/COUNT_STAR, 0)
ELSE 0
END as '平均扫描行数',
DATE(FIRST_SEEN) as '首次出现日期',
DATE(LAST_SEEN) as '最后出现日期',
TIMESTAMPDIFF(DAY, FIRST_SEEN, LAST_SEEN) as '持续天数'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
AND DIGEST_TEXT IS NOT NULL
AND COUNT_STAR > 1000 -- 执行次数超过1000次
AND AVG_TIMER_WAIT > 100000000 -- 平均执行时间超过100毫秒
ORDER BY COUNT_STAR DESC, AVG_TIMER_WAIT DESC
LIMIT 15;
6. 当前正在执行的查询
sql复制代码
SELECT
p.ID as '进程ID',
p.USER as '用户',
p.HOST as '主机',
p.DB as '数据库',
p.COMMAND as '命令类型',
p.TIME as '执行时间(秒)',
p.STATE as '状态',
LEFT(IFNULL(p.INFO, ''), 200) as 'SQL语句',
CASE
WHEN p.TIME > 60 THEN '极慢'
WHEN p.TIME > 10 THEN '慢'
WHEN p.TIME > 1 THEN '一般'
ELSE '正常'
END as '性能等级'
FROM information_schema.PROCESSLIST p
WHERE p.COMMAND != 'Sleep'
AND p.TIME > 1 -- 执行时间超过1秒
AND p.ID != CONNECTION_ID() -- 排除当前连接
ORDER BY p.TIME DESC;
7. 慢查询统计汇总
sql复制代码
SELECT
'指标类型' as metric_type,
'数值' as metric_value,
'单位' as unit
FROM (SELECT 1 as dummy) t
WHERE 1=0 -- 创建表头
UNION ALL
SELECT
'总查询类型数',
CAST(COUNT(*) as CHAR),
'个'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
UNION ALL
SELECT
'慢查询类型数(>1秒)',
CAST(COUNT(*) as CHAR),
'个'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
AND AVG_TIMER_WAIT > 1000000000000
UNION ALL
SELECT
'极慢查询类型数(>10秒)',
CAST(COUNT(*) as CHAR),
'个'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
AND AVG_TIMER_WAIT > 10000000000000
UNION ALL
SELECT
'总执行次数',
CAST(SUM(COUNT_STAR) as CHAR),
'次'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
UNION ALL
SELECT
'总执行时间',
CAST(ROUND(SUM(SUM_TIMER_WAIT)/1000000000000/3600, 2) as CHAR),
'小时'
FROM performance_schema.events_statements_summary_by_digest
WHERE SCHEMA_NAME IS NOT NULL
AND SCHEMA_NAME NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys');
8. 表结构和索引分析
8.1 表索引详情查询
sql复制代码
SELECT
s.TABLE_SCHEMA as '数据库',
s.TABLE_NAME as '表名',
s.INDEX_NAME as '索引名',
s.COLUMN_NAME as '列名',
s.SEQ_IN_INDEX as '索引位置',
s.CARDINALITY as '基数',
s.NULLABLE as '可为空',
CASE s.INDEX_TYPE
WHEN 'BTREE' THEN 'B树索引'
WHEN 'HASH' THEN '哈希索引'
WHEN 'FULLTEXT' THEN '全文索引'
ELSE s.INDEX_TYPE
END as '索引类型',
CASE
WHEN s.INDEX_NAME = 'PRIMARY' THEN '主键'
WHEN s.NON_UNIQUE = 0 THEN '唯一索引'
ELSE '普通索引'
END as '索引分类'
FROM information_schema.STATISTICS s
WHERE s.TABLE_SCHEMA NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
ORDER BY s.TABLE_SCHEMA, s.TABLE_NAME, s.INDEX_NAME, s.SEQ_IN_INDEX;
8.2 表大小统计
sql复制代码
SELECT
t.TABLE_SCHEMA as '数据库',
t.TABLE_NAME as '表名',
t.ENGINE as '存储引擎',
IFNULL(t.TABLE_ROWS, 0) as '估算行数',
ROUND(IFNULL(t.DATA_LENGTH, 0)/1024/1024, 2) as '数据大小(MB)',
ROUND(IFNULL(t.INDEX_LENGTH, 0)/1024/1024, 2) as '索引大小(MB)',
ROUND((IFNULL(t.DATA_LENGTH, 0) + IFNULL(t.INDEX_LENGTH, 0))/1024/1024, 2) as '总大小(MB)',
t.AUTO_INCREMENT as '自增值',
t.CREATE_TIME as '创建时间',
t.UPDATE_TIME as '更新时间'
FROM information_schema.TABLES t
WHERE t.TABLE_SCHEMA NOT IN ('information_schema', 'performance_schema', 'mysql', 'sys')
AND t.TABLE_TYPE = 'BASE TABLE'
ORDER BY (IFNULL(t.DATA_LENGTH, 0) + IFNULL(t.INDEX_LENGTH, 0)) DESC
LIMIT 20;