supervisord is running on CentOS server. If I do
ps -e -o %mem,%cpu,cmd | grep supervisord | awk '{memory+=$1;cpu+=$2} END {print memory,cpu}'
I get 0 0 just because supervisord is just an initialization daemon. It runs four child processes on my server:
# pgrep -P $(pgrep supervisord) | wc -l
4
How can I find summarized CPU and memory usage of these child processes in one-line-command?
The code from happyraul's answer,
pgrep -P $(pgrep supervisord) | xargs ps -o %mem,%cpu,cmd -p | awk '{memory+=$1;cpu+=$2} END {print memory,cpu}'
will get only one child layer.
If you want to search for all processes that were derived from a main pid, use this code:
ps -o pid,ppid,pgid,comm,%cpu,%mem -u {user name} | {grep PID_PRINCIPAL}
The pid of main process is the PGID of child processes.
Given a pid,
pid=24535
pstree -p $pid | grep -o '([0-9]\+)' | grep -o '[0-9]\+' |\
xargs ps -o %mem,%cpu,cmd -p | awk '{memory+=$1;cpu+=$2} END {print memory,cpu}'
# 15.5 905.2
I had no luck getting the pids of all child processes from pgrep.
CPU time used divided by the time the process has been running - that is, historical average of the %CPU usage, not its current value.
Try using xargs:
pgrep -P $(pgrep supervisord) | xargs ps -o %mem,%cpu,cmd -p | awk '{memory+=$1;cpu+=$2} END {print memory,cpu}'
