CPU usage shell script

D

dcol

I am trying to build a CPU Usage shell script and need a shell command that will give me the result of the usage percentage of lets say CPU 0. Seems sysctl kern.cp_time may be the way to go, but not sure of the results that generates.
Maybe a better command, not sure. I am not a guru by any stretch.

Ultimate purpose of this script is to alert me via email if the cpu usage hits 100%. Got every part of the script working except for the CPU Usage calculation.
Any help is greatly appreciated.
Thanks
Dan
 
Too generalized for a newbie writing a script. These are possible commands but they display too much info. I need to enter the actual usage number into a variable.
For example, I use the following to give me a number in the variable 'temp' that reflects the temperature of CPU 0
Code: 
temp=$( sysctl dev.cpu.0.temperature | sed -e 's|.*: \([0-9.]*\)C|\1|' )
I need the same for CPU 0 usage.
 
dcol said:
temp=$( sysctl dev.cpu.0.temperature | sed -e 's|.*: \([0-9.]*\)C|\1|' )
Click to expand...
Use
Code: 
temp=$(sysctl -n dev.cpu.0.temperature)
No need for the complicated sed(1) pipe. See sysctl(8).

dcol said:
These are possible commands but they display too much info. I need to enter the actual usage number into a variable.
Click to expand...
It's fairly easy to filter out only the data you need with some awk(1) or through a clever tr(1)/cut(1) manipulation.
 
The code I posted works and there is no need to reinvent the wheel on that, but thanks for that suggestion.
Looking for a free and easy solution to use in my PFsense firewall which uses FreeBSD which is why I want to just add a simple cron script to perform a simple task on a schedule.
The temperature script works perfectly, no need to change that. Now I want to do the same for the CPU usage, which is why I am asking if anyone knows of a command sequence that will allow me to put that info into a variable. Monitoring tools are overkill because I am only looking for hardware monitoring, not network.
Anyway not sure how to use them with PFsense anyway.
If this cannot be done, then let me know and I will move on.
 
sysctl dev.cpu.0.cx_usage is always at 100, even when system is at 5%, so that would not work. Thought of that earlier.
I looked at all the sysctl.cpu commands, none have what I need. I really don't want to add a packaged monitor system. Don't want the extra overhead.

Let me explain why I need a simple CPU Usage script. When I am running inline Suricata, it can sometimes ramp up the CPU depending on network usage. When the CPU usage gets above 90%, the system is unable to keep up. I want to be notified when this occurs so I can remedy the situation. This usually requires a reboot since just disabling Suricata never seems to lower the CPU usage. Right now I use the temp monitor script because I know that when the CPU starts peaking, the temps rise, so I monitor that. But I would rather monitor the usage since I don't want to keep overheating the system.

I just hoping someone had some coding around that would measure the CPU usage and pass it to a variable.
 
You could use sysutils/conky. Something like this:

conky.conf:
Code: 
out_to_x no
own_window no
out_to_console yes
background no
max_text_width 0
update_interval 2.0
short_units yes
if_up_strictness address
use_spacer left
override_utf8_locale no
cpu_avg_samples 2

TEXT = [[
${cpu cpu0}
]]

cpu_usage.sh:
Code: 
#!/usr/bin/env sh

cpu_usage() {

   cpu_0=$(conky -qdc /path/to/conky.conf)

   echo -n "$cpu_0"
}

echo $(cpu_usage)

It was not tested but should work with eventual necessary minimal fix (I wrote it now). :)

You could even include the cpu temp too.
 
Don't want to have to install any additional packages that are not included with PFsense.
 
You are correct, and I have, but no one seems to provide any answers there. Which is why I came here.
 
How about using ps -o %cpu to look at how "idle" the machine is, and trigger your monitoring point when it falls below a certain threshold:
Code: 
ps -o %cpu -p `pgrep -S idle`
The percentage seems to be based on 100% * # of cpus (which you can get from sysctl hw.ncpu).

EDIT: Beware that ps shows snapshot values, not averaged over time, so you'll probably want to do that as part of your monitoring.
 
You must log in or register to reply here.
Back
Top