Capture processes with significant CPU usage (Bash shell script)

find-cpu-hogs.sh

#!/bin/bash

# Output CSV-formatted statistics about each process using > ${threshold}% CPU,
# sampling every ${delay} seconds. Assume we'll find them in the top ${max}
# processes returned by `top` sorting by CPU usage.

threshold=${1:-5}
delay=${2:-5}
max=${3:-25}

echo "Threshold: ${threshold}% CPU. Sampling every ${delay} seconds." > /dev/stderr

echo "Timestamp,Process,CPU"
while True
do
    ps aux -c -r | tail -n +2 | head -n ${max} | while read -r line
    do
        cpu=$(echo "${line}" | awk '{print $3}')
        process=$(echo "${line}" | awk '{print $11}')
        timestamp=$(date +"%Y-%m-%d %H:%M:%S")

        if (( $(echo "${cpu} >= ${threshold}" | bc -l) ))
        then
            echo "${timestamp},${process},${cpu}"
        fi
    done
    echo "." > /dev/stderr
    sleep ${delay}
done

Save it as e.g. find-cpu-hogs.sh and run chmod +x find-cpu-hogs.sh in a terminal to make it executable.

Run as:

$ ./find-cpu-hogs.sh | tee cpu-hogs.csv

This will output all processes using more than 5% CPU, sampling every 5 seconds, to a CSV file, which you can then analyse in Excel, say. It will write a . to the console (stderr) but not the file (assuming you use tee as above) each time it samples so you know it’s doing something even if no processes go over the threshold.

Use CTRL+C to stop it.

If you run it again, it will overwrite the CSV file. You can use the -a argument to tee to append instead (though you’ll get the header twice so you’ll need to clean up the file a bit later).

You can change the threshold and frequency by passing one or two command line arguments, e.g.:

$ ./find-cpu-hogs.sh 10 20 | tee -cpu-hogs.csv

This will put the threshold at 10% and sample every 20 seconds.