MySQL is quite popular among open source web apps, but it is prone to performance issues, if not maintained properly.
Performance issues happen primarily through table fragmentation, unoptimized memory settings, and more. All this can result in MySQL high CPU usage, and application errors.
Here at Bobcares, our Dedicated Support Engineers maintain thousands of web and MySQL servers of web hosts, eCommerce companies and digital marketers all over the world.
We function as the tech team of these companies and keep their servers stable through proactive maintenance and 24/7 monitoring.
Our process model for stable MySQL servers
Here are the 3 key operations that we perform to keep MySQL servers stable:
1. Initial server optimization
When a customer signs up, our engineers audit the server performance and optimize the MySQL server settings to fix any pre-existing resource bottlenecks.
After analyzing the website type and traffic, we tweak the key settings in MySQL servers that help to ensure optimal resource usage. This includes connection limits, buffer size and query cache size, and more.
2. 24/7 monitoring & emergency rescue
Comment spamming, brute forcing, bot attack or even a legitimate increase in traffic can cause high MySQL load. Which is why it is important to monitor server metrics 24/7.
Server admins from Bobcares monitor MySQL health 24/7. It helps us to detect a spike in the CPU usage or abusive program immediately and fix it before it crashes the entire server.
3. Follow-up audit & optimization
Databases grow over time, and the load on it changes when website traffic changes. That is why follow-up audit and re-optimization is critical in maintaining MySQL server stability.
A core job of our Dedicated Server Administrators is to audit customer servers regularly and find out performance bottlenecks before they happen and fix it before any customers are affected.
Today, we’ll take a look at how our Server Admins react to a MySQL high CPU situation, and what we do to prevent its recurrence.
[ You don’t have to be a MySQL expert to keep your sites fast and stable. Let us help you. Our experts will keep your servers fast, secure and rock solid. ]
How to detect MySQL high CPU usage
Very often people link high server load to high CPU usage. That is not always true. The server load can go high due to bottlenecks in any resource. This can include memory, disk I/O, network or CPU.
The top reason for MySQL induced server load is due to memory or I/O exhaustion. If it is CPU bottleneck, the output of the top command would look like this:
top - 8:31:15 up 23 days, 3:33, 1 user, load average: 22.31, 19.72, 25.02
Tasks: 81 total, 5 running, 76 sleeping, 0 stopped, 0 zombie
Cpu(s): 87.8%us, 0.3%sy, 0.0%ni, 11.2%id, 0.0%wa, 0.2%hi, 0.5%si, 0.0%st
Mem: 2975920k total, 1018219k used, 1957701k free, 287408k buffers
If it is I/O induced bottleneck, the %wa (called wait average) will have the highest CPU%. In contrast, if it is a memory induced load, the “free” memory limits to just a few MBs.
Click here to know how to fix a MySQL high memory usage situation.
Fixing MySQL high CPU
If the server load is indeed related to high CPU usage, we’ve found the following fixes to be useful:
- Enable InnoDB to handle a high number of concurrent connections – Check MySQL “PROCESSLIST”, and if you see a lot of queries in “LOCK” status, it means a lot of queries are put on hold because MyISAM tables are handling other transactions. To fix this convert those tables into the InnoDB engine which supports row-level locking.
- Enable persistent connections – If you have only a single application that receives thousands of connections per hour, enabling persistent MySQL connections can improve performance. If the server has multiple applications (like a shared web hosting server) this may not work.
- Block abusive processes – When a website is under attack (like DoS, comment spamming, etc.), it ends up in an abnormally high number of established connections in a short time. Use the “PROCESSLIST” in MySQL to identify the top users, and block access to the abusive connections.
- Optimize database queries – Some web applications use complex queries to display site information. These queries can take a long time to execute, and cause CPU load. Get the list of such queries from the “slow query log” and reduce the number of joins and other table manipulations under a single query.
- Check for “leap second bug” – On July 1st, 2015, there was an addition of a leap second to standard UTC time. In servers running old Linux kernel versions, and which uses time servers, this is seen to cause MySQL high load. If you have an old Linux kernel, try resetting the time using the command date -s “$(date)”.
[ Don’t lose your customers to a slow server. Let us help you. Our experts will monitor & maintain your servers 24/7 so that it remains lightning fast.]
How to prevent MySQL high CPU usage
Server traffic changes and databases grow over time. MySQL high CPU issues can be prevented to a large extent if the database server is audited and tuned for performance.
- MySQL performance tuning – MySQL uses various buffers and cache systems to execute queries. As the volume and complexity of database queries change, the server settings need to be adjusted for optimum performance. There are various tools such as mysqltuner to identify any settings that need adjustment.
- Security audit and hardening – Spamming and DoS attacks can easily overwhelm a database server. Implement web application firewalls such as ModSecurity, and DoS firewalls such as ModEvasive to prevent attackers from affecting server uptime.
- Implementing load balancing – As the server traffic grows it might be required to split the load into multiple servers. MySQL can be configured for master-master and master-slave replication which allows queries to be served from any server in a cluster.
- Optimizing database queries – If web applications are poorly coded, no amount of database optimization will fix the server load. Monitor MySQL’s “slow query log” and reduce the number of JOINs to make the database faster.
- Using high-performance alternatives such as Percona – MySQL has many memory/CPU bottlenecks due to the way it processes queries. MySQL spin-offs such as MariaDB and Percona resolves these issues and helps achieve better stability.
How to start
MySQL can be a bit thick to understand. It takes patience and a lot of experience to know what solution works where.
You can either jump headfirst into MySQL docs and become an expert (in about 2-3 years), or you can get your server fixed (and keep it fixed) by hiring an administrator.
And no, it doesn’t have to be costly. You can hire an admin now.
Or if you want one-time quick assistance, click here to submit a support request.
Thank you *very much* for your advice. Came home today, the first day of January 2017, to find my mariadb process taking up about 40% CPU … with no database queries, should have been totally idle. As my server (Centos 7) is running on an Amazon ‘T2’ instance which relies on never running out of ‘CPU credit’ – which for my machine means only using an average of 20% CPU – this was a concern. Another half day of this and I would have been sunk.
I read your article, not thinking I was going to find anything to help … saw your item #5 … ran the ‘date’ command … and CPU has dropped to zero.
THANK YOU!
mysqld was consumming the entire cpu up to 1377% as my server has 14 cores, I resolved this by indexing all keywords used in WHERE clause and increasing query_cache_limit to 1024M &
query_cache_size to 2048M
where to change that please
Hello Bao,
Make these changes in the MySQL configuration file, usually found at my.cnf.
date -s “$(date)” helped me. Mysql was running at 400% before.
Hello, please help.. How can i do this ?? I am new to this and had been facing this problem for 2 months.
We’ll be happy to talk to you on chat (click on the icon at right-bottom).