Know how much an individual process or system-wide consume CPU or memory.
As a sysadmin, you often have to deal with an incident where the application is slow or unresponsive due to high CPU/memory/network utilization. If the server host just one process, then it’s easy to find out when the process consumes all the resources. However, imagine a shared server where multiple services are running, and you need to find which one is eating all the resources.
There are many monitoring software which does this out of the box. But if you don’t have one or looking for a command-based solution, then here you go. They are all FREE!
top
You may want to start by looking into top or htop result to see the processes overview.
As you can see below, it gives an excellent idea about what all processes are utilizing. If you look at the first one, which is MySQL is taking 11.9% of CPU and 2.5% of CPU.
top - 11:57:33 up 0 min, 1 user, load average: 3.69, 0.96, 0.32
Tasks: 165 total, 2 running, 113 sleeping, 0 stopped, 0 zombie
%Cpu(s): 21.0 us, 5.5 sy, 0.0 ni, 70.5 id, 1.7 wa, 0.0 hi, 1.3 si, 0.0 st
KiB Mem : 7637308 total, 5802888 free, 849512 used, 984908 buff/cache
KiB Swap: 0 total, 0 free, 0 used. 6495648 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1986 netdata 20 0 1738856 191560 22948 S 11.6 2.5 0:02.30 mysqld
3021 www-data 20 0 255288 78420 55484 S 6.6 1.0 0:01.55 php-fpm
3138 www-data 20 0 253096 79780 59228 S 6.6 1.0 0:00.92 php-fpm
3153 www-data 20 0 255116 79088 56472 S 5.0 1.0 0:00.70 php-fpm
3037 www-data 20 0 257200 81088 56216 S 4.3 1.1 0:01.50 php-fpm
3048 www-data 20 0 257088 78740 55380 S 4.3 1.0 0:01.46 php-fpm
3054 www-data 20 0 254160 72168 52108 S 3.7 0.9 0:01.32 php-fpm
3135 www-data 20 0 255084 75912 54836 S 3.7 1.0 0:00.91 php-fpm
3051 www-data 20 0 254096 73804 51964 S 3.0 1.0 0:01.38 php-fpm
2962 www-data 20 0 45280 7284 3488 R 2.0 0.1 0:00.22 openresty
1062 netdata 20 0 338748 76144 6720 S 1.0 1.0 0:01.31 netdata
1702 netdata 20 0 21852 4232 2352 S 1.0 0.1 0:00.34 apps.plugin
1729 netdata 20 0 18636 3280 2764 S 0.7 0.0 0:00.05 bash
1980 netdata 20 0 62008 12896 5796 S 0.7 0.2 0:00.14 redis-server
11 root 20 0 0 0 0 I 0.3 0.0 0:00.14 rcu_sched
1007 root 20 0 1347424 74524 38872 S 0.3 1.0 0:00.92 dockerd
1857 root 20 0 10600 5564 4276 S 0.3 0.1 0:00.03 containerd-shim
2045 root 20 0 9948 6028 5016 S 0.3 0.1 0:00.14 forego
2934 root 20 0 13616 8760 5928 S 0.3 0.1 0:00.07 docker-gen
2966 systemd+ 20 0 25784 7924 2340 S 0.3 0.1 0:00.06 nginxThe top is installed on almost all Linux distribution.
Once you identify the suspect, then you may want to focus on that process instead of everything like you saw above. You can still use top command but with some argument.
Let’s say you know the process id (PID); you can use the below command.
top -p $PID
Below an example of top -p 3102
top - 11:59:56 up 3 min, 1 user, load average: 0.72, 0.70, 0.31
Tasks: 1 total, 0 running, 1 sleeping, 0 stopped, 0 zombie
%Cpu(s): 7.1 us, 2.9 sy, 0.0 ni, 89.1 id, 0.3 wa, 0.0 hi, 0.7 si, 0.0 st
KiB Mem : 7637308 total, 5802024 free, 783672 used, 1051612 buff/cache
KiB Swap: 0 total, 0 free, 0 used. 6555636 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3102 www-data 20 0 329500 82376 60640 S 0.0 1.1 0:03.35 php-fpmYou may also use grep with top. Below an example of checking Docker utilization.
root@geekflare-com:~# top | grep docker
1007 root 20 0 1347424 74524 38872 S 0.3 1.0 0:01.38 dockerd
2934 root 20 0 14676 9652 5928 S 0.3 0.1 0:00.54 docker-gen
1007 root 20 0 1347424 74524 38872 S 0.3 1.0 0:01.39 dockerd
1007 root 20 0 1347424 74524 38872 S 1.0 1.0 0:01.42 dockerd
2934 root 20 0 14740 9652 5928 S 0.3 0.1 0:00.55 docker-gen
2934 root 20 0 14740 9652 5928 S 0.3 0.1 0:00.56 docker-genhtop
Similar to the top but with more information. As you can, it got the command column, which is handy to identify the process path. And also it is colorful.
htop may not be installed by default, but you can always do it as below.
Install htop on Ubuntu
apt-get install htopInstall htop on CentOS/RHEL 8.x
dnf install https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm
dnf update
dnf install htopglances
As the name says, you get a system utilization view on a single screen. Running processes are sorted by their CPU utilization.
You can install glances on CentOS 8 using DNF as below.
dnf install glancesfor CentOS7, you can use YUM
yum install glancesatop
A similar to the above listed but with a brilliant feature to record the output in a file so you can view them later. Imagine there is a pattern of having an issue at a specific time window. You can schedule to write the output in a file through crontab or other, and later you can playback.
To record the output in a file:
atop -w filenameand, to playback:
atop -r filenameIt supports multiple arguments like interval, samples, etc. and I would strongly recommend taking a look at the man page.
If you are just interested in real-time troubleshooting, then just execute atop and you should see like below.
You can install atop as below.
dnf install atopps
Let’s check ps command now.
You can use the ps command with PID to print their CPU and memory utilization.
ps -p $PID -o %cpu,%memThe output should look like this.
root@sr-master-us:~# ps -p 1048 -o %cpu,%mem
%CPU %MEM
0.2 3.0
root@sr-master-us:~#nmon
Interactive command-line monitoring tool for CPU, memory, disks, network, NFS, and virtual memory utilization. To view the top process (by utilization), you can execute nmon and press t button.
You can install nmon as below.
dnf install nmonMonit
Monit is a web-based and command-line open-source solution to monitor server resources, daemons, files, directory, file systems, etc.
Monit also got a cool widget.
Its light-weight monitoring software. But, there is more to explore here.
Monitorix
A lightweight open-source utility to monitor the Linux server. Monitorix got in-built HTTP so you can check the utilization and other stuff on the web. Some of the other usage reports include:
- Kernal/temperature
- Filesystem and I/O
- Network traffic
- Apache/Mail/FTP/Nginx
- MySQL/Varnish/Memcached
Monitorix also offers alert configuration so you can get notified when things are not right. It will be a good choice when you are managing cloud-based servers and looking for a proactive monitoring solution.
Netdata
Netdata is a real-time performance monitoring for system resources, applications, web servers, databases, DNS, mail, hardware sensors, and a lot more. It is open-source and getting started is easy. All the data is collected, stored, and streamed for you to visualize interactively. Data is collected every second, so you never miss anything.
Loved by many industry leaders.
So what you are waiting for, try and take control of your Linux servers.
btop
btop is a handy resource monitor fully interactive with a beautiful UI helping you manage the Linux servers.
You can easily visualize the processes in a tree view, filter from the process list, and manage the resource hogs. btop also comes powered with an auto-scaling graph showing network usage.
Besides, you can also check disk speed and complete I/O activity.
There is more to this that you can experience on Linux, FreeBSD, and macOS.
Conclusion
I hope the above tools help you to visualize the server utilization in real-time so you can take necessary action. If you just started as a system administrator and looking to get hands-on training, then check out this Udemy course.