[SOLVED] is there a way to show top summarized by user?

Skaperen · 07-12-2020, 07:39 PM

i would like to see a display like top that shows a summary by user with one line per user and the heavier users on top. for each user this would have the sum of all processes by that user. is there some kind of configuration (i didn't a simple direct option in the man page) for top or some alternate software to do this?

eight.bit.al · 07-12-2020, 09:52 PM

A quick search found:

run either top -u {userName} or htop -u {userName} commands in a terminal.
Not exactly what you wanted.

8bit

frankbell · 07-12-2020, 10:12 PM

I also looked into this and found what eight.bit.al found, a way to show output per user. I could not find a way to sort output by user.

It occurs to me that there might be a way to pipe the output of top to a script that in turn would then sort the output by user and store it in a file, but that would be only a snapshot, not a dynamic representation.

Sadly, I would be woefully unqualified to produce such a script.

berndbausch · 07-12-2020, 11:26 PM

top has an excellent manual page. It also has a nice (in my opinion) help facility; just type h while it's running.

Section 3b of the man page tells you how to display any field you can dream for, including the user ID, and how to sort by any field.

I haven't checked whether you can then summarize by user.

Skaperen · 07-13-2020, 05:55 PM

just to be clear, i don't want to pick one user and show just that user's processes. i want to see which user is using the most resource.

a program that does this would get all the same data that top does, then collect a list of all the users. it would make an array with rows just like it did for processes. the it would loop through the array of processes, find the row for its real user in the user array, and add the resources to that row. the PID field would just be incremented by one to count how many active processes that user has. once all that is done, then sort the user array as appropriate and display it instead of the process array.

yeah, maybe i should get the source for top, add this code, add an option to engage this, and release the patch.

berndbausch · 07-13-2020, 06:39 PM

Quote:

Originally Posted by Skaperen

a program that does this would get all the same data that top does, then collect a list of all the users. it would make an array with rows just like it did for processes. the it would loop through the array of processes, find the row for its real user in the user array, and add the resources to that row. the PID field would just be incremented by one to count how many active processes that user has. once all that is done, then sort the user array as appropriate and display it instead of the process array.

awk is great at summarizing data like this. Since the structure of top's output lines is quite regular, it could serve as convenient input for a simple awk script.

Even more convenient might be the original data source under /proc, in particular values like /proc/PID/stat or /proc/PID/io. See the manual page.

MadeInGermany · 07-14-2020, 11:34 AM

Here is a part of a script that I once wrote:

Code:

#!/bin/sh

PATH=/usr/xpg4/bin:/usr/bin:/bin:/usr/sbin:/usr/ucb
export PATH
set -f
# secure GNU grep
unset GREP_OPTIONS

if [ -f /proc/cpuinfo ]; then
 scaleps=`grep -c '^processor' /proc/cpuinfo` # ps pcpu is per vcpu
else
 scaleps=1 # ps pcpu is per total
fi

#all in one variable:
allprocs=`ps -e -o pcpu= -o vsz= -o rss= -o user=`
#user= must be last in ps to not be truncated

#assume 900 processes maximum:
echo "#procs per user, top 3:"
echo "$allprocs" |
 awk '{s[$4]++} END {for(i in s) printf " %-10s %-4d , %2d %%\n",i,s[i],int(s[i]*100/900)}' |
 sort -nr -k 2,2 | head -3
#assume 6000 threads maximum:
echo "#threads per user, top 3:"
ps -Le -o user= |
 awk '{s[$1]++} END {for(i in s) printf " %-9s %-4d , %2d %%\n",i,s[i],int(s[i]*100/6000)}' |
 sort -nr -k 2,2 | head -3
echo "#cpu% per user, top 3:"
echo "$allprocs" |
 awk '{s[$4]+=$1} END {for(i in s) printf " %-9s %2.1f\n",i,s[i]/c}' c=$scaleps |
 sort -nr -k 2,2 | head -3
echo "#vmemoryMB per user, top 3:"
echo "$allprocs" |
 awk '{s[$4]+=$2} END {for(i in s) printf " %-9s %4.1f\n",i,s[i]/1024}' |
 sort -nr -k 2,2 | head -3
echo "#residentMB per user, top 3:"
echo "$allprocs" |
 awk '{s[$4]+=$3} END {for(i in s) printf " %-9s %4.1f\n",i,s[i]/1024}' |
 sort -nr -k 2,2 | head -3

Skaperen · 07-15-2020, 10:06 PM

i still want to have it in a display just like top itself does.

shruggy · 07-16-2020, 04:00 AM

Try atop. The key u gives summaries by user.

Skaperen · 07-18-2020, 12:36 AM

atop is just what i was looking for. i am using the -u command line option.