How to log CPU load to a file in order to investigate a problem?

3

5 Answers

This works very well:

 while true; do uptime >> uptime.log; sleep 1; done

• This will log your cpu load every second and append it to a file uptime.log.

  You can then import this file into Gnumeric or the OpenOffice spreadsheet to create a nice graph (select 'separated by spaces' on import).

As Scaine noticed, this won't be enough to diagnose the problem. So, additionally, run this (or use his answer for this part):

while true; do (echo "%CPU %MEM ARGS $(date)" && ps -e -o pcpu,pmem,args --sort=pcpu | cut -d" " -f1-5 | tail) >> ps.log; sleep 5; done

• This will append the Top 10 most CPU hungry processes to a file ps.log every five seconds.

  Note that this is not the full boat-load of information top would give you. This is just the top 10, and just their CPU Usage, Memory Usage and the first argument (i.e. their command without further arguments, as in /usr/bin/firefox)

After you've used a Spreadsheet to create a graph to see when your CPU load went through the roof, you can then search this file for the nearest time to see what process has caused it.

This is what those files will look like:

uptime.log

~$ cat uptime.log 22:57:42 up 1 day, 4:38, 4 users, load average: 1.00, 1.26, 1.21 22:57:43 up 1 day, 4:38, 4 users, load average: 0.92, 1.24, 1.21 22:57:44 up 1 day, 4:38, 4 users, load average: 0.92, 1.24, 1.21 22:57:45 up 1 day, 4:38, 4 users, load average: 0.92, 1.24, 1.21 ...

ps.log

%CPU %MEM ARGS Mo 17. Jan 23:09:47 CET 2011 0.7 0.9 /usr/bin/compiz 0.8 0.5 /usr/lib/gnome-panel/clock-applet 1.1 1.7 /opt/google/chrome/chrome 1.2 0.3 /usr/bin/pulseaudio 1.8 4.0 /opt/google/chrome/chrome 2.6 1.5 /opt/google/chrome/chrome 2.6 3.2 /usr/bin/google-chrome 3.6 2.6 /opt/google/chrome/chrome 4.9 1.5 /usr/bin/X 5.7 1.6 /opt/google/chrome/chrome
%CPU %MEM ARGS Mo 17. Jan 23:09:48 CET 2011 0.7 0.9 /usr/bin/compiz 0.8 0.5 /usr/lib/gnome-panel/clock-applet 1.0 1.7 /opt/google/chrome/chrome 1.2 0.3 /usr/bin/pulseaudio 1.8 4.0 /opt/google/chrome/chrome 2.6 1.5 /opt/google/chrome/chrome 2.6 3.2 /usr/bin/google-chrome 3.6 2.6 /opt/google/chrome/chrome 4.9 1.5 /usr/bin/X 5.7 1.6 /opt/google/chrome/chrome ...

4

You can run the top command in batch mode by using the -b option, then dump that to a file.

On start up of your PC, open a terminal, run

top -b > ~/cpu.txt

Then when your PC freezes, simply open the (probably huge) text file and check the last entry for some detail on what was running just before the crash. In the fact the file will be so stupidly large that you're better off running a tail -250 ~/cpu.txt instead.

Also check your /var/log/kern.log in case your issue is hardware related (unlikely if this is only happening after an upgrade, but worth checking nonetheless).

I found a great answer by Christopher to this question on Unix and Linux that uses top:

top -n 1 -b > top.out

This will give you 1 iteration of top then stop, and then push it to a file.

For those who need to run this command after your putty ( SSH client ) session end. you can use command screen ( or install it using apt-get )

I`ve got an script for get the process load, with rotating logs:

#!/bin/bash
MaxFileSize=204800
DaysToKeep=7
echo -e "\n Fecha:"`date` >> /var/log/ps.log
echo -e "\n Uptime: "`uptime` >> /var/log/ps.log
ps -e -o pcpu,pmem,args --sort=pcpu | tail >> /var/log/ps.log
#Get size in bytes**
file_size=`du -b /var/log/ps.log | tr -s '\t' ' ' | cut -d' ' -f1`
if [ $file_size -gt $MaxFileSize ];then timestamp=`date +%s` mv /var/log/ps.log /var/log/ps.log.$timestamp gzip /var/log/ps.log.$timestamp touch /var/log/ps.log # remove old files find /var/log -name "ps.log.*" -type f -mtime +$DaysToKeep -delete
fi

My original source can be browsed here

Your Answer

Sign up or log in

Sign up using Google Sign up using Facebook Sign up using Email and Password Submit

Post as a guest

Post Your Answer Discard

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy