Since warning appears on some time ago Zabbix, reported Escalators process pressure, increased Escalators parameters, such as changes to 80-90 just do not alarm, I felt wrong.
To the device, I found that mysql top 95% of the CPU are eaten
On mysql to show processlist;
We found a large sleep process, show status like 'Thread%'; see the connection number only 138.
Modify the my.cnf file, max_connections will be reduced to 300,
We found no improvement, Baidu to read and write timeout time may be no global optimization problems
Modify the my.cnf mysql configuration:
Add in the [mysqld]
wait_timeout=120
interactive_timeout=120
Restart mysql, CPU occupancy rate is really decreased, currently occupies about 15%.
The same high CPU may be due to other causes, the next person to share some of the following rule of thumb:
* 1, excessive alarm is not restored, it will also cause this problem needs to be emptied zabbix the following table in order to reduce CPU usage mysql, and this approach a temporary solution, the need for timely dispose of alarm.
Alerts；problem；problemtag；events；eventrecovery；escalations。
2, the alarm will trigger a set of multiple single, mysql can effectively reduce the occupation of the CPU, there is a reasonable set alarm triggers, regular cleaning zabbix alarm table.
3. Check occupy high CPU SQL, top -H -p <mysql process id>, then use the table to locate the problem inside the sql mysql:
SELECTa.THREADOSID,b.user,b.host,b.db,b.command,b.time,b.state,http://b.info
FROMperformanceschema.threads a,informationschema.processlist b
WHEREb.id = a.processlistid;
From here you can analyze which of sql taking up more CPU, so do the next step optimization. *