紧急处理：
运行平稳的数据库，如果遇到CPU狂飙，到80%左右，那一定是开发写的烂SQL导致的，DBA首先要保证的是，数据库别跑挂了，所以我们要把那些运行慢的SQL杀死并记录到文件里，以便后面的排查。

pt-kill  --match-info "^(select|SELECT)"--busy-time 3--victim all --interval 1--kill --print --daemonize > /root/kill.txt

pt-kill详解
这条命令是只把select耗时3秒以上的SQL全部杀死，并打印出来

模拟测试：

也可以通过show processlist；命令详细排查：
1.通过top命令查询服务器cup占用情况

我们看到cup占用率已经达到199.7%（服务器是双核）
2.执行 mysql -u root -p 进入mysql，输入show full processlist; 查看正在连接mysql的线程
show processlist详解

我们看到这个sql是导致mysql数据库飙升的原因
3.杀死这些占用资源的烂sql
kill 504;
kill 509;

通过mysql慢日志分析：
1、查看mysql日志记录是否开启：show variables like ‘slow_query_log’;

没开启的话先开启日志记录： set global slow_query_log = on;

2、查看记录慢sql的时间：show global variables like ‘long_query_time’;

默认是10秒，可以根据需要修改，这里修改为执行时间超过3秒的sql被记录：

set global long_query_time =3;

3、查看日志存放位置：show variables like ‘slow_query_log_file’;

4、查看日志： cat /data/mysql/XXX/long.log

# Time: 170104 10:41:04
# User@Host: root[root] @  [127.0.0.1]  Id: 785264
# Query_time: 2.082531  Lock_time: 0.000132 Rows_sent: 0  Rows_examined: 437963
SET timestamp=1483497664;
SELECT id,update_time,update_end_time,ctime FROM `sg_point_log` WHERE pid = 105 AND date(ctime) = curdate() AND (update_time != 0 OR update_end_time != 0) ORDER BY id DESC limit 1;
# Time: 170104 10:42:06
# User@Host: root[root] @  [127.0.0.1]  Id: 785461
# Query_time: 2.037191  Lock_time: 0.000162 Rows_sent: 0  Rows_examined: 438013
SET timestamp=1483497726;
SELECT id,update_time,update_end_time,ctime FROM `sg_point_log` WHERE pid = 203 AND date(ctime) = curdate() AND (update_time != 0 OR update_end_time != 0) ORDER BY id DESC limit 1;
# Time: 170104 10:42:10
# User@Host: root[root] @  [127.0.0.1]  Id: 785549
# Query_time: 2.003415  Lock_time: 0.000071 Rows_sent: 0  Rows_examined: 438016
SET timestamp=1483497730;
SELECT id,update_time,update_end_time,ctime FROM `sg_point_log` WHERE pid = 281 AND date(ctime) = curdate() AND (update_time != 0 OR update_end_time != 0) ORDER BY id DESC limit 1;
# Time: 170104 10:42:14
# User@Host: root[root] @  [127.0.0.1]  Id: 785264
# Query_time: 2.152803  Lock_time: 0.000127 Rows_sent: 0  Rows_examined: 438018
SET timestamp=1483497734;
SELECT id,update_time,update_end_time,ctime FROM `sg_point_log` WHERE pid = 8 AND date(ctime) = curdate() AND (update_time != 0 OR update_end_time != 0) ORDER BY id DESC limit 1;
# Time: 170104 10:42:16
# User@Host: root[root] @  [127.0.0.1]  Id: 785264
# Query_time: 2.040670  Lock_time: 0.000077 Rows_sent: 0  Rows_examined: 438021
SET timestamp=1483497736;
SELECT id,upload_img_time,ctime FROM `sg_point_log` WHERE pid = 7 AND date(ctime) = curdate() AND upload_img_time != 0 ORDER BY id DESC limit 1;
# Time: 170104 10:42:18
# User@Host: root[root] @  [127.0.0.1]  Id: 785264
# Query_time: 2.139948  Lock_time: 0.000124 Rows_sent: 0  Rows_examined: 438025
SET timestamp=1483497738;
SELECT id,update_time,update_end_time,ctime FROM `sg_point_log` WHERE pid = 7 AND date(ctime) = curdate() AND (update_time != 0 OR update_end_time != 0) ORDER BY id DESC limit 1;

可以看出罪魁祸首就是这个sql