Operation and Maintenance
Linux Operation and Maintenance
How to troubleshoot high CPU usage in Linux
How to troubleshoot high CPU usage in Linux
Start by identifying high-CPU processes using top, htop, or ps aux --sort=-%cpu; analyze system-wide behavior with vmstat, sar, and iostat to detect I/O or kernel issues; inspect problematic processes via top -H, strace, or perf; then resolve by killing, limiting, or renicing processes, and monitor for improvements.

High CPU usage in Linux can slow down your system and affect performance. Identifying and resolving the root cause requires a systematic approach using built-in tools and analysis. Here’s how to troubleshoot it effectively.
Identify the Process Using High CPU
Start by pinpointing which process or processes are consuming excessive CPU resources.
- Run top – This gives a real-time view of system processes sorted by CPU usage. Look at the %CPU column to spot offenders.
- Use htop (if installed) – A more user-friendly version of top with color coding and better navigation. Install it via your package manager if needed.
- Check with ps aux --sort=-%cpu – Lists all running processes sorted by CPU usage, useful for quick snapshots.
Analyze System-Wide CPU Behavior
Determine whether the load is due to user processes, system interrupts, or I/O wait.
- Run vmstat 1 5 – Shows CPU breakdown into user, system, idle, and I/O wait time every second for 5 iterations. High "sy" (system) time may indicate kernel or driver issues.
- Use sar -u 1 5 (from sysstat) – Reports detailed CPU utilization over time. Helps spot trends beyond real-time observation.
- Check for I/O bottlenecks with iostat -x 1 – If %util is near 100% and await is high, disk I/O could be causing CPU to wait.
Inspect Specific Processes in Detail
Once you’ve identified a problematic process, dig deeper into its behavior.
- Use top -H -p
– Breaks down CPU usage per thread within a process. Useful when one thread is spiking. - Attach strace -p
– Traces system calls made by the process. Look for repeated or failing calls that waste CPU. - Analyze with perf top -p
– Shows which functions inside a process use the most CPU (requires perf installation).
Take Corrective Action
After diagnosing the issue, decide on the appropriate fix.
- Kill or restart misbehaving processes: kill -9
as a last resort, but prefer graceful termination. - Limit CPU usage with cpulimit --pid=
--limit=50 to temporarily reduce impact. - Adjust process priority using renice to lower its scheduling weight.
- Update or patch software if the high CPU is due to known bugs.
- Monitor after changes to confirm improvement using the same tools.
Basically, start with visibility, drill into the details, then act. Most cases stem from runaway scripts, inefficient applications, or resource starvation. Regular monitoring helps catch issues before they escalate.
The above is the detailed content of How to troubleshoot high CPU usage in Linux. For more information, please follow other related articles on the PHP Chinese website!
Hot AI Tools
Undress AI Tool
Undress images for free
AI Clothes Remover
Online AI tool for removing clothes from photos.
Undresser.AI Undress
AI-powered app for creating realistic nude photos
ArtGPT
AI image generator for creative art from text prompts.
Stock Market GPT
AI powered investment research for smarter decisions
Hot Article
Popular tool
Notepad++7.3.1
Easy-to-use and free code editor
SublimeText3 Chinese version
Chinese version, very easy to use
Zend Studio 13.0.1
Powerful PHP integrated development environment
Dreamweaver CS6
Visual web development tools
SublimeText3 Mac version
God-level code editing software (SublimeText3)
Hot Topics
20516
7
13630
4
How to manage software packages using Dnf and Rpm in Linux?
Mar 07, 2026 am 01:50 AM
dnfinstall prompts packagenotfound because it queries the enabled warehouse rather than the local installation status; rpm-q queries the local RPM database, and the scopes of the two are different.
How to monitor system performance and resources in Linux? (Top & Htop)
Mar 06, 2026 am 01:04 AM
The total top CPU usage is not 100% because it is calculated based on a single core, and the maximum is 800% for an 8-core system; the htop terminal size error needs to be fixed with eval$(resize); the real memory pressure is available rather than used; top command line truncation can be solved with the -c parameter.
How to setup file sharing using Samba on Linux? (SMB Protocol)
Mar 13, 2026 am 12:33 AM
The main reason why Windows cannot see the Samba share is that the firewall blocks UDP137–139/TCP445 or NetBIOS name resolution fails; it is necessary to confirm that the workgroup is consistent, the interfaces are configured correctly, the file permissions match forceuser/forcegroup, and set doscharset=UTF-8 to solve Chinese garbled characters.
How to extend a Logical Volume (LVM) in Linux without downtime?
Mar 13, 2026 am 12:53 AM
Logical volumes and file systems can be expanded online. You need to expand the LV first and then the file system. It is recommended to use lvextend-r for automatic synchronization adjustment, but you must ensure that the LVM and file system tool versions are compatible.
How to check open ports and listening services in Linux? (Netstat & SS)
Mar 10, 2026 am 01:08 AM
Netstat displays fewer LISTEN ports than ss because it does not display process information by default that non-root users do not have access to; ss can read all listening sockets by default without process names, and sudonetstat-tulpn is required to display them completely.
How to set up SSH key authentication on Linux? (Passwordless Login)
Mar 11, 2026 am 12:46 AM
It is recommended to use ssh-keygen-ted25519 to generate a key pair, because it is faster, more secure, and has a shorter key than the default RSA; it is necessary to strictly set the ~/.ssh directory permissions to 700 and authorized_keys to 600, and use ssh-v to confirm whether the client is Offering public key and whether the server rejects it.
How to format disk partitions using the command line in Linux?
Mar 15, 2026 am 12:01 AM
When fdisk is stuck at the Command prompt, it is normally waiting for input. Enter q to exit safely; you must umount before mkfs, otherwise it may fail silently; partedmkpart does not support the specified file system type, and mkfs needs to be executed separately.
How to configure a static IP address on Linux? (Netplan & NetworkManager)
Mar 14, 2026 am 12:02 AM
Netplan reports "InvalidYAML" when configuring a static IP due to indentation errors, missing spaces after colons, or mixed tabs; gateway4 has been deprecated and routes to:default must be used instead; NetworkManager needs to be modified before down/up takes effect; the renderer field is used to determine the backend during coexistence; incorrect DNS configuration will cause ping to succeed but curl to fail.





