Parse ps output to show top CPU, memory and process count, summed over identically named processes

readme.md

download and install

git clone git://gist.github.com/3435564.git sum_top_gist
chmod +x sum_top_gist/sum_top.sh 
ln -s `readlink -f sum_top_gist/sum_top.sh`  ~/bin/sum_top
sum_top

pretty.py

import string

def pretty_size(size):
    "size in bytes"
    size = int(size)
    units = ' KMGTPE'+string.lowercase[::-1]
    bytes = 'B'

    i = 0
    while size>1024:
        i += 1
        size /= 1024.0
    return '% 2.1f %s' %(size, units[i]+bytes)

def pretty_time(t):
    "time t in secs"
    t = float(t)
    unit = 'secs'
    weeks_in_year = 365.25/7
    d = ((60,'mins'),(60,'hours'),(24,'days'),(7,'weeks'),(weeks_in_year,'year'),
         (10,'decade'),(10,'century'),(10,'millenia'))
    for mx, name in d:
        if t>mx:
            t /= mx
            unit = name
        else:
            break
    return '%2.1f %s' %(t, unit)

def pretty_percent(n, tot=None):
    if tot:
        n = 100.0*n/tot
    return '%2.1f %%' % n

sum_top.py

'''
Basic usage:
ps -eo args,%cpu,rss --sort %mem | tail -n+2 | python sum_top.py

cpu
  106.0 % /usr/libexec/mysqld --based
   95.8 % /usr/sbin/httpd
    8.7 % python manage.py garbage_co
    0.2 % [kipmi0]
    0.1 % memcached -d -l 127.0.0.1 -
mem
 991.7 MB /usr/sbin/httpd
 358.8 MB /usr/libexec/mysqld --based
  42.0 MB memcached -d -l 127.0.0.1 -
  23.1 MB /opt/dell/srvadmin/sbin/dsm
  18.7 MB python manage.py garbage_co
count
    25  x /usr/sbin/httpd
     7  x smtp -t unix -u
     7  x /opt/dell/srvadmin/sbin/dsm
     5  x /usr/sbin/saslauthd -m /var
     4  x [kjournald]


You want to record some server stats while you hammer it with ab?
Try this:
nohup python -c 'import time,os,sys;[time.sleep(1) or os.system("sum_top.sh") for _ in range(60*60)]' &
'''

import sys
import pprint
from collections import defaultdict

import pretty

ls = sys.stdin.readlines()

c = defaultdict(lambda: 0)
m = defaultdict(lambda: 0)
n = defaultdict(lambda: 0)

for l in ls:
    cmd, cpu, mem = l.rsplit(None, 2)
    c[cmd] += float(cpu)
    m[cmd] += float(mem)
    n[cmd] += 1

for title, dd, fmter in (
        ('cpu', c, pretty.pretty_percent),
        ('mem', m, lambda i: pretty.pretty_size(i*1024)),
        ('count', n, lambda i: '%d  x' % i),
    ):
    print title

    print '\n'.join('%s %s' % (fmter(v).rjust(9), k) for (k,v) in sorted(dd.items(), key=lambda kv: kv[1], reverse=1)[:5])

sum_top.sh

# eg
# ln -s ~/sum_top_gist/sum_top.sh ~/bin/sum_top
DIR="$(dirname "$(readlink -f "$0")")"

echo -e "\n\n===============================\n"
echo -e "$(uptime)\n"
echo -e "$(free -m)\n"
echo -e "$(ps -eo args,%cpu,rss --sort %mem | tail -n+2 | python $DIR/sum_top.py)\n"
echo -e "\n===============================\n\n"