How to Find and Stop High-CPU Processes

A runaway process consuming 100% of your CPU can cause websites to load slowly, game servers to lag, and SSH connections to freeze. Knowing how to quickly locate and handle high-CPU processes is a critical skill for every Linux server administrator.

In this guide, we show you how to inspect, isolate, and safely terminate or throttle high-CPU processes on a RivinCloud VPS.

1. Finding High-CPU Processes

Method A: Using ps (Fast & Scriptable)

To display the top 10 processes consuming the most CPU power:

ps aux --sort=-%cpu | head -n 11

This output prints the USER, PID (Process ID), %CPU, %MEM, TIME, and the exact COMMAND executed.

Method B: Using top

Launch top:

top

Press Shift + P inside top to sort processes by CPU utilization in descending order. Note the PID of the suspicious process.

Method C: Using pidof or pgrep

If you suspect a specific program (e.g., a node script or python bot) is causing problems:

pgrep -l node

2. Stopping High-CPU Processes Gracefully

Once you have identified the PID, never immediately kill it forcefully unless necessary.

Step 1: Send a Graceful Termination Signal (SIGTERM)

kill -15 <PID>
# or simply
kill <PID>

This instructs the process to save open files, close database connections, and exit cleanly.

Step 2: Forceful Kill (SIGKILL)

If the application is completely frozen and does not respond after 15-30 seconds:

kill -9 <PID>

[!WARNING] Using kill -9 terminates the process instantly at the kernel level. This can corrupt open database files or state caches if used indiscriminately.

Step 3: Terminating by Name with killall or pkill

sudo pkill -f problematic-script.py

3. Alternative: Throttling Instead of Killing

If a process is legitimate (such as a database backup or video rendering task) and you want it to run without hogging the whole CPU:

Using renice to Lower Priority

Linux scheduling priorities range from -20 (highest priority) to 19 (lowest priority).

sudo renice 15 -p <PID>

Using cpulimit

Install cpulimit to restrict a process to a specific maximum percentage (e.g., 50% CPU limit):

sudo apt install cpulimit -y
sudo cpulimit -p <PID> -l 50

Common Mistakes

• Killing system processes: Accidentally terminating systemd, sshd, or kworker processes.
• Using kill -9 by default: Always attempt standard kill (SIGTERM) first.
• Ignoring the root cause: Killing a process that immediately gets restarted by a supervisor loop without diagnosing why it crashed or looped.

Troubleshooting

• Process doesn't die even with kill -9: The process is likely in an uninterruptible sleep state (D state in top), usually blocked waiting for hanging disk I/O or an unmounted NFS share. You must resolve the underlying storage freeze or reboot.
• Zombie processes (Z state): Zombie processes are already dead and consume no CPU/RAM; they remain until their parent process acknowledges them. Restart the parent process to clear zombies.

Best Practices

• Use process managers like pm2 or systemd with resource limits configured (CPUQuota=80% in systemd unit files).
• Keep system packages updated to avoid known memory leaks and infinite CPU execution bugs.

Frequently Asked Questions

What does %CPU over 100% mean in top?

Linux counts 100% per CPU core/thread. On a 4-core VPS, a multi-threaded application can show up to 400% CPU usage.

How do I identify which user launched a process?

The first column in ps aux indicates the user account that spawned the process.

Conclusion

Quickly identifying and resolving high-CPU bottlenecks keeps your RivinCloud VPS responsive and stable. Bookmark these commands for immediate troubleshooting during unexpected server spikes!