VPS CPU at 100%? A Performance Optimization Troubleshooting Guide

ps aux --sort=-%cpu | head -20

This lists the top 20 processes by CPU usage. The meaning of each column in the output:

• USER - The user running the process
• PID - Process ID
• %CPU - CPU usage percentage
• %MEM - Memory usage percentage
• COMMAND - The process command

watch -n 1 'ps aux --sort=-%cpu | head -10'

Refreshes once per second so you can observe trends in CPU usage.

After finding the PID of a suspicious process, look at its details:

# 查看进程的完整命令行
cat /proc/<PID>/cmdline | tr '\0' ' '

# 查看进程的工作目录
ls -l /proc/<PID>/cwd

# 查看进程打开的文件
lsof -p <PID>

Infinite loops or inefficient queries in website code are the most common cause.

Characteristics: php-fpm or python processes continuously at 100% CPU

Troubleshooting:

# 查看 PHP-FPM 进程状态
ps aux | grep php-fpm

# 查看正在执行的 PHP 脚本
lsof -c php-fpm | grep '\.php'

Solution:

• Check recently modified code
• Check the PHP error log /var/log/php-fpm/error.log
• Temporarily restart PHP-FPM: systemctl restart php-fpm

Slow MySQL queries can bring down the entire server.

Characteristics: The mysqld process uses very high CPU

Troubleshooting:

# 登录 MySQL 查看当前查询
mysql -e "SHOW PROCESSLIST;"

# 查看慢查询日志（需要先开启）
tail -f /var/log/mysql/slow.log

Solution:

• Optimize slow queries and add indexes
• Kill stuck queries: mysql -e "KILL <process_id>;"
• Consider upgrading the VPS configuration or using a dedicated database server

This is a serious problem that needs immediate handling.

Characteristics:

• Unfamiliar process names, such as kdevtmpfsi, kinsing, or random strings
• Processes are hidden or cannot be killed normally
• CPU stays at 100% but you cannot find the cause

Troubleshooting:

# 检查定时任务
crontab -l
cat /etc/crontab
ls -la /etc/cron.d/

# 检查可疑的启动项
systemctl list-unit-files --type=service | grep enabled

# 检查异常的网络连接
netstat -antp | grep ESTABLISHED

Solution:

• Kill the suspicious processes immediately
• Clean up malicious entries in scheduled tasks
• Change all passwords and check SSH keys
• Update the system and all software
• In severe cases, reinstalling the system is recommended

Too many Nginx or Apache workers are configured, or the PHP-FPM process pool is set too large.

Troubleshooting:

# 查看 Nginx worker 数量
ps aux | grep nginx | wc -l

# 查看 PHP-FPM 进程数量
ps aux | grep php-fpm | wc -l

Solution: Adjust parameters according to the VPS configuration. For a 1-core VPS, recommended values:

• Nginx workers: 1-2
• PHP-FPM pm.max_children: 5-10

Background system tasks may temporarily saturate the CPU.

Characteristics: High CPU in processes such as apt, yum, rsync, gzip

Solution: Usually you just need to wait for the task to finish. If it is not urgent, you can lower the priority:

# 降低进程优先级（19 是最低）
renice 19 -p <PID>

# 正常终止
kill <PID>

# 强制终止（进程无响应时使用）
kill -9 <PID>

# 按进程名批量杀
pkill -f <进程名关键词>

Use the cpulimit tool to limit the CPU usage of a specific process:

# 安装
apt install cpulimit -y

# 限制某进程最多使用 50% CPU
cpulimit -p <PID> -l 50

# 启动时设置低优先级
nice -n 19 <命令>

# 修改运行中进程的优先级
renice 10 -p <PID>

# 安装 nmon
apt install nmon -y
nmon

# 或者安装 glances（功能更全）
pip3 install glances
glances

Write a simple monitoring script that sends a notification when CPU exceeds a threshold:

#!/bin/bash
CPU_THRESHOLD=90
CPU_USAGE=$(top -bn1 | grep 'Cpu(s)' | awk '{print int($2)}')

if [ $CPU_USAGE -gt $CPU_THRESHOLD ]; then
    echo "CPU 使用率 ${CPU_USAGE}% 超过阈值" | mail -s "VPS CPU 告警" [email protected]
fi

Add it to crontab to check every 5 minutes:

*/5 * * * * /path/to/cpu_monitor.sh