top and CPU usage

milomak · 11-14-2018, 04:23 PM

i'm trying to understand how the individual app %CPU corresponds to the total CPU usage of the pc (which I assume is %Cpu(s) - us)

So looking at the below, the first few apps highlighted individually sum above 100%. Userspace Cpu usage seems at 22%

How should I be reading this?

scasey · 11-14-2018, 04:30 PM

Have you read the documentation? See man top
Granted, there's a lot there to absorb, but that's probably the best way to start. Come back with specific questions if you don't understand something.
Some food for thought: How many processors (CPUs) does your system have?

syg00 · 11-14-2018, 05:31 PM

"Lies, damned lies, and statistics".

Not only are the numbers different (overall versus per process) the latter are, as scasey hints, non-normalised.
And if everything is sleeping, how are they using CPU ?.
Back to my first sentence.

It doesn't pay to get too involved in the precise numbers, but use them as a general indication of what's happening - this is sampled data averaged over a time period after all.

berndbausch · 11-14-2018, 05:54 PM

On my system, just starting a virtual machine:

Code:

top - 08:50:18 up  1:24,  1 user,  load average: 1.41, 0.43, 0.15
Tasks: 189 total,   1 running, 133 sleeping,   0 stopped,   0 zombie
%Cpu(s): 48.8 us,  1.2 sy,  0.0 ni, 49.3 id,  0.1 wa,  0.5 hi,  0.2 si,  0.0 st
KiB Mem : 16304420 total,  7985040 free,  6007852 used,  2311528 buff/cache
KiB Swap:  8191996 total,  8191996 free,        0 used.  9876744 avail Mem

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
 1965 qemu      20   0 9360252 5.219g  23676 S 199.0 33.6   2:14.72 qemu-syste+

It has an Intel i3 CPU, which looks like four CPUs to Linux. Notice the individual CPU usage of the qemu process. CPU summary is scaled and always adds up to 100% (or close).

chrism01 · 11-14-2018, 10:51 PM

@OP ; if you run top, then tap the '1' key whilst its running, it'll show you all the cpus individually - it may help

scasey · 11-15-2018, 12:21 AM

top may be useful in determining the cause of a problem, but I was taught to use the first line to see if there even is a problem:

Code:

top - 23:16:11 up 110 days, 23:38,  3 users,  load average: 0.01, 0.05, 0.05

Once can get that same information with the uptime command, which is probably what top uses

Code:

# uptime
 23:19:15 up 110 days, 23:41,  3 users,  load average: 0.32, 0.10, 0.07

Read up on what load average is telling you...man uptime has a paragraph

milomak · 11-15-2018, 04:10 PM

there was a point where the 2 top were 400% and 200%

you can imagine the system hung and i could not get those numbers

dc.901 · 11-16-2018, 08:15 AM

+1 on chrism01's advise. Your machine does not have just 1 core; due to multiple cores and multithreads, top result can be over 100%.
top is not the only way to look at utilization. Look at sar data also.

milomak · 11-22-2018, 05:42 PM

what is sar data?

dc.901 · 11-22-2018, 05:52 PM

Quote:

Originally Posted by milomak

what is sar data?

Don't know if that was sarcastic question?

If not, good info here about sar:

https://en.wikipedia.org/wiki/Sar_(Unix)

https://www.thegeekstuff.com/2011/03/sar-examples/

KenJackson · 11-26-2018, 04:05 PM

Quote:

Originally Posted by scasey

Have you read the documentation? ... Come back with specific questions if you don't understand something.

I always appreciate it when people give helpful, friendly advice like this instead of just passing on by.