[Ubuntu/Debian/CentOS/Fedora] How to limit CPU usage for process?

William

There can be cases when your server performance can slow down when even a single process uses all of your server CPU capability. Such usage will cause issues for other processes to work properly on your VPS instance. Since when one process is using all of the resources, other processes have to wain in the queue to start their tasks. This can be solved by limiting CPU resource usage with CPULimit software.

The main purpose of CPULimit is to set a restriction on the specific process on how much CPU it can use. When the limitation is implemented, there is no influence on the nice value or the settings on priorities, but on real CPU usage instead.

0. Requirements:
Root or sudo.
nano or another text editor.

1.1 Installation if using Ubuntu/Debian:

sudo apt-get install cpulimit

1.2 Installation if using CentOS/Fedora:

Firstly, it will be required to install the EPEL repository if it was not done previously:

sudo yum install epel-release

Afterward, the installation can be initiated:

sudo yum install cpulimit

or

sudo dnf install cpulimit

2. Usage of CPULimit:

To check the effectiveness of CPULimit, let's create a process (usecpu.sh) that will consume all of your server's CPU:

sudo nano usecpu.sh

Enter the following code into the usecpu.sh file:

#!/bin/bash
while :; do :; done;

Save the code on your file. This code will create a loop that will consume all of the server's CPU. When the file is created, set executable permission for it:

sudo chmod +x usecpu.sh

The created process can be started using:

./usecpu.sh &

After initiation, you will receive a result of process id, for example, 1887:

[1] 1887

If everything is correct, the new process should use all of the server's CPU. You can check the usage of your CPU by using the top command:

PID USER      PR  NI    VIRT    RES    SHR S %CPU %MEM     TIME+ COMMAND           
 2912 root      20   0   12512    896    792 R 99.9  0.0   0:18.38 usecpu.sh         
 2893 root      20   0   39688   3660   3092 R  0.3  0.2   0:00.02 top               
    1 root      20   0   37912   6000   4004 S  0.0  0.3   0:01.52 systemd           
    2 root      20   0       0      0      0 S  0.0  0.0   0:00.00 kthreadd          
    3 root      20   0       0      0      0 S  0.0  0.0   0:00.05 ksoftirqd/0       
    5 root       0 -20       0      0      0 S  0.0  0.0   0:00.00 kworker/0:0H      
    6 root      20   0       0      0      0 S  0.0  0.0   0:00.10 kworker/u2:0      
    7 root      20   0       0      0      0 S  0.0  0.0   0:00.49 rcu_sched         
    8 root      20   0       0      0      0 S  0.0  0.0   0:00.00 rcu_bh            
    9 root      rt   0       0      0      0 S  0.0  0.0   0:00.00 migration/0       
   10 root      rt   0       0      0      0 S  0.0  0.0   0:00.01 watchdog/0        
   11 root      20   0       0      0      0 S  0.0  0.0   0:00.00 kdevtmpfs         
   12 root       0 -20       0      0      0 S  0.0  0.0   0:00.00 netns             
   13 root       0 -20       0      0      0 S  0.0  0.0   0:00.00 perf              
   14 root      20   0       0      0      0 S  0.0  0.0   0:00.00 khungtaskd        
   15 root       0 -20       0      0      0 S  0.0  0.0   0:00.00 writeback         
   16 root      25   5       0      0      0 S  0.0  0.0   0:00.00 ksmd   

As you can see from the table, the usecpu.sh process uses 99.9% of server CPU. Such CPU usage will prevent other processes from working successfully and can finally result in server OS inactivity. This is where CPULimit comes to usage.

If you wish to limit the process to use only 30% of your server total CPU, initiate the following command:

cpulimit -l 30 -p 2912 &

• * -l 30  will set the limit to the provided number of %. In this case 30%.
• * -p 2892   is used to identify the process by it's PID, as seen in the first column from top command.

Now, when the process is limited by CPULimit let's check the results again by initiating top command:

PID USER      PR  NI    VIRT    RES    SHR S %CPU %MEM     TIME+ COMMAND           
 2912 root      20   0   12512    996    896 T 31.6  0.0   0:29.76 usecpu.sh         
 2915 root       9 -11    8612   1544   1428 S  0.3  0.1   0:00.02 cpulimit          
    1 root      20   0   37912   6000   4004 S  0.0  0.3   0:01.53 systemd           
    2 root      20   0       0      0      0 S  0.0  0.0   0:00.00 kthreadd          
    3 root      20   0       0      0      0 S  0.0  0.0   0:00.06 ksoftirqd/0       
    5 root       0 -20       0      0      0 S  0.0  0.0   0:00.00 kworker/0:0H      
    6 root      20   0       0      0      0 S  0.0  0.0   0:00.11 kworker/u2:0      
    7 root      20   0       0      0      0 S  0.0  0.0   0:00.52 rcu_sched         
    8 root      20   0       0      0      0 S  0.0  0.0   0:00.00 rcu_bh            
    9 root      rt   0       0      0      0 S  0.0  0.0   0:00.00 migration/0       
   10 root      rt   0       0      0      0 S  0.0  0.0   0:00.02 watchdog/0        
   11 root      20   0       0      0      0 S  0.0  0.0   0:00.00 kdevtmpfs         
   12 root       0 -20       0      0      0 S  0.0  0.0   0:00.00 netns             
   13 root       0 -20       0      0      0 S  0.0  0.0   0:00.00 perf              
   14 root      20   0       0      0      0 S  0.0  0.0   0:00.00 khungtaskd        
   15 root       0 -20       0      0      0 S  0.0  0.0   0:00.00 writeback         
   16 root      25   5       0      0      0 S  0.0  0.0   0:00.00 ksmd  

Now the usecpu.sh process CPU consumption will be limited to 31.6% of CPU (can be +/- 5% deflection).

If you wish to use process name instead of PID for the limitation, you can form your command from such example:

cpulimit -l 25 ./usecpu.sh &