MySQL慢查询导致CPU飙高的完整指南

top - 08:30:00 up 30 days, load average: 2.50, 2.30, 2.10
MiB Swap: 4096.0 total, 3964.7 free, 131.3 used. 6434.6 avail Mem

PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
1878724 lxd    20   0 2809352 560612  34364 S 102.3   6.9   4:45.44 mysqld
1788 root   20   0 1007448  38620   8908 T   9.6   0.5  7053:32 oneav

SHOW ENGINE INNODB STATUS\G

LOG
---
Log capacity                 104857600
Log capacity used            104857600   ← 100% 满！
...
BUFFER POOL AND MEMORY
---
Buffer pool size   8192
Buffer pool hit rate 948 / 1000          ← 仅 94.8%（正常 >99%）
...
ROW OPERATIONS
---
175526.47 reads/s                        ← 异常高

SHOW VARIABLES LIKE 'innodb_log_file_size';
+----------------------+----------+
| Variable_name        | Value    |
+----------------------+----------+
| innodb_log_file_size | 50331648 |   -- 48MB
+----------------------+----------+

SHOW VARIABLES LIKE 'innodb_log_files_in_group';
+---------------------------+-------+
| Variable_name             | Value |
+---------------------------+-------+
| innodb_log_files_in_group | 2     |
+---------------------------+-------+

[mysqld]
# Redo Log（核心问题）
innodb_log_file_size = 256M
innodb_log_files_in_group = 3     # 总容量 768MB
# Buffer Pool
innodb_buffer_pool_size = 2G
# 其他优化
innodb_flush_log_at_trx_commit = 2
innodb_flush_method = O_DIRECT

SET GLOBAL slow_query_log = ON;
SET GLOBAL long_query_time = 2;
SET GLOBAL log_queries_not_using_indexes = ON;

-- 验证配置
SHOW VARIABLES LIKE 'slow_query_log%';
+---------------------+----------------------------------+
| Variable_name       | Value                            |
+---------------------+----------------------------------+
| slow_query_log      | ON                               |
| slow_query_log_file | /var/lib/mysql/slow.log          |
| long_query_time     | 2.000000                         |
+---------------------+----------------------------------+

# Time: 2026-06-03T12:09:41.527275Z
# User@Host: swust[swust] @ [172.18.0.3] Id: 398
# Query_time: 4.782904  Lock_time: 0.000003 Rows_sent: 1  Rows_examined: 310516
SET timestamp=1780488576;
SELECT count(*) AS count_1 
FROM (SELECT ... 31个字段 ... FROM datasets WHERE ...) AS anon_1;

# Time: 2026-06-03T12:09:46.161278Z
... 相同 SQL，每 5 秒执行一次

USE qa_gen;
CREATE INDEX idx_user_stage_time ON datasets(user_id, current_stage, created_at);

-- 改造前（慢）：先查全部字段再计数
SELECT COUNT(*) FROM (
    SELECT * FROM datasets 
    WHERE user_id = 11 
      AND current_stage = 'question_generate' 
      AND created_at >= '2026-06-01 11:46:14'
) AS t;

-- 改造后（快）：直接 COUNT
SELECT COUNT(*) FROM datasets 
WHERE user_id = 11 
  AND current_stage = 'question_generate' 
  AND created_at >= '2026-06-01 11:46:14';

# 不要这样做（慢）
count = session.query(func.count()).select_from(
    session.query(Dataset).filter(...).subquery()
).scalar()

# 应该这样做（快）
count = session.query(func.count(Dataset.id)).filter(
    Dataset.user_id == 11,
    Dataset.current_stage == 'question_generate',
    Dataset.created_at >= '2026-06-01 11:46:14'
).scalar()

EXPLAIN SELECT COUNT(*) FROM datasets 
WHERE user_id = 11 
  AND current_stage = 'question_generate' 
  AND created_at >= '2026-06-01 11:46:14'\G

-- 优化前：
-- type: ALL (全表扫描)
-- rows: 310516
-- Extra: Using where

-- 优化后：
-- type: ref (索引查找)
-- key: idx_user_stage_time (使用索引)
-- rows: 1847
-- Extra: Using index (覆盖索引)