Question

I am stress testing my production server which is 1-Core linux server. While stress testing the load averages was oscillating between 2 and 5 but the CPU utilization was between 70%-100%. I looked up the load average definition and i understood that if the laod average >1 for an extended period of time then the CPU utilization for that time must be 100%

Answer 1

Load is the number of processes waiting (usually for disk IO). It is not a percentage of CPU usage, disk, network or anything else.

You can see the disk wait processes with a command such as this:

watch 'ps -eo pid,user,state,command | awk '\''$3~/D/ { print $0 }'\'

Which I got from here: https://serverfault.com/questions/316922/how-to-detect-the-process-and-mysql-query-that-makes-high-load-on-server which didn’t work but I changed it to make it work.

This also works for me, but shows lots of lines having nothing to do with state “D”

ps aux | grep " D "

UPDATE: here is an alternative script using BSD style options:

ps auxw | awk 'index($8,"STAT") || $8~/D/ { print $0 }'