The script is a very helpful to show how much RAM is used by single process. This script is written by P\u00e1draig Brady, He can be contacted at P@draigBrady.com. The script itself follows:<\/p>\n
#!\/usr\/bin\/env python\r\n\r\n# Try to determine how much RAM is currently being used per program.\r\n# Note the per program, not per process. So for example this script\r\n# will report mem used by all httpd process together. In detail it reports:\r\n# sum(all RSS for process instances) + max(shared mem for any process instance)\r\n#\r\n# The shared calculation below will factor out shared text and\r\n# libs etc. within a program, but not between programs. So there\r\n# will always be some overestimation. This will be the same for\r\n# all processes that just use libc for e.g. but more for others\r\n# that use larger shared libs like gnome, kde etc.\r\n\r\n# Author: P@draigBrady.com\r\n\r\n# V1.0 06 Jul 2005 Initial release\r\n# V1.1 11 Aug 2006 root permission required for accuracy\r\n# V1.2 08 Nov 2006 Add total to output\r\n# Use KiB,MiB,... for units rather than K,M,...\r\n# V1.3 22 Nov 2006 Ignore shared col from \/proc\/$pid\/statm for\r\n# 2.6 kernels up to and including 2.6.9.\r\n# There it represented the total file backed extent\r\n# V1.4 23 Nov 2006 Remove total from output as it's meaningless\r\n# (the shared values overlap with other programs).\r\n# Display the shared column. This extra info is\r\n# useful, especially as it overlaps between programs.\r\n# V1.5 26 Mar 2007 Remove redundant recursion from human()\r\n# V1.6 05 Jun 2007 Also report number of processes with a given name.\r\n# Patch from riccardo.murri@gmail.com\r\n\r\n# Notes:\r\n#\r\n# All interpreted programs where the interpreter is started\r\n# by the shell or with env, will be merged to the interpreter\r\n# (as that's what's given to exec). For e.g. all python programs\r\n# starting with \"#!\/usr\/bin\/env python\" will be grouped under python.\r\n# You can change this by changing comm= to args= below but that will\r\n# have the undesirable affect of splitting up programs started with\r\n# differing parameters (for e.g. mingetty tty[1-6]).\r\n#\r\n# For 2.6 kernels up to and including 2.6.13 and later 2.4 redhat kernels\r\n# (rmap vm without smaps) it can not be accurately determined how many pages\r\n# are shared between processes in general or within a program in our case:\r\n# http:\/\/lkml.org\/lkml\/2005\/7\/6\/250\r\n# A warning is printed if overestimation is possible.\r\n# In addition for 2.6 kernels up to 2.6.9 inclusive, the shared\r\n# value in \/proc\/$pid\/statm is the total file-backed extent of a process.\r\n# We ignore that, introducing more overestimation, again printing a warning.\r\n#\r\n# I don't take account of memory allocated for a program\r\n# by other programs. For e.g. memory used in the X server for\r\n# a program could be determined, but is not.\r\n#\r\n# This script assumes threads are already merged by ps\r\n\r\n# TODO:\r\n#\r\n# use ps just to enumerate the pids and names\r\n# so as to remove the race between reading rss and shared values\r\n\r\nimport sys, os, string\r\n\r\nif os.geteuid() != 0:\r\n sys.stderr.write(\"Sorry, root permission required.\\n\");\r\n sys.exit(1)\r\n\r\nPAGESIZE=os.sysconf(\"SC_PAGE_SIZE\")\/1024 #KiB\r\nour_pid=os.getpid()\r\n\r\n#(major,minor,release)\r\ndef kernel_ver():\r\n kv=open(\"\/proc\/sys\/kernel\/osrelease\").readline().split(\".\")[:3]\r\n for char in \"-_\":\r\n kv[2]=kv[2].split(char)[0]\r\n return (int(kv[0]), int(kv[1]), int(kv[2]))\r\n\r\nkv=kernel_ver()\r\n\r\ndef getShared(pid):\r\n if os.path.exists(\"\/proc\/\"+str(pid)+\"\/smaps\"):\r\n shared_lines=[line\r\n for line in open(\"\/proc\/\"+str(pid)+\"\/smaps\").readlines()\r\n if line.find(\"Shared\")!=-1]\r\n return sum([int(line.split()[1]) for line in shared_lines])\r\n elif (2,6,1) <= kv <= (2,6,9):\r\n return 0 #lots of overestimation, but what can we do?\r\n else:\r\n return int(open(\"\/proc\/\"+str(pid)+\"\/statm\").readline().split()[2])*PAGESIZE\r\n\r\ncmds={}\r\nshareds={}\r\ncount={}\r\nfor line in os.popen(\"ps -e -o rss=,pid=,comm=\").readlines():\r\n size, pid, cmd = map(string.strip,line.strip().split(None,2))\r\n if int(pid) == our_pid:\r\n continue #no point counting this process\r\n try:\r\n shared=getShared(pid)\r\n except:\r\n continue #ps gone away\r\n if shareds.get(cmd):\r\n if shareds[cmd] < shared:\r\n shareds[cmd]=shared\r\n else:\r\n shareds[cmd]=shared\r\n #Note shared is always a subset of rss (trs is not always)\r\n cmds[cmd]=cmds.setdefault(cmd,0)+int(size)-shared\r\n if count.has_key(cmd):\r\n count[cmd] += 1\r\n else:\r\n count[cmd] = 1\r\n\r\n#Add max shared mem for each program\r\nfor cmd in cmds.keys():\r\n cmds[cmd]=cmds[cmd]+shareds[cmd]\r\n\r\nsort_list = cmds.items()\r\nsort_list.sort(lambda x,y:cmp(x[1],y[1]))\r\nsort_list=filter(lambda x:x[1],sort_list) #get rid of zero sized processes (kernel threads)\r\n\r\n#The following matches \"du -h\" output\r\n#see also human.py\r\ndef human(num, power=\"Ki\"):\r\n powers=[\"Ki\",\"Mi\",\"Gi\",\"Ti\"]\r\n while num >= 1000: #4 digits\r\n num \/= 1024.0\r\n power=powers[powers.index(power)+1]\r\n return \"%.1f %s\" % (num,power)\r\n\r\ndef cmd_with_count(cmd, count):\r\n if count>1:\r\n return \"%s (%u)\" % (cmd, count)\r\n else:\r\n return cmd\r\nprint \" Private + Shared = RAM used\\tProgram \\n\"\r\nfor cmd in sort_list:\r\n print \"%8sB + %8sB = %8sB\\t%s\" % (human(cmd[1]-shareds[cmd[0]]), human(shareds[cmd[0]]), human(cmd[1]),\r\n cmd_with_count(cmd[0], count[cmd[0]]))\r\nprint \"\\n Private + Shared = RAM used\\tProgram \\n\"\r\n\r\n#Warn of possible inaccuracies\r\n#1 = accurate\r\n#0 = some shared mem not reported\r\n#-1= all shared mem not reported\r\ndef shared_val_accurate():\r\n \"\"\"http:\/\/wiki.apache.org\/spamassassin\/TopSharedMemoryBug\"\"\"\r\n if kv[:2] == (2,4):\r\n if open(\"\/proc\/meminfo\").read().find(\"Inact_\") == -1:\r\n return 1\r\n return 0\r\n elif kv[:2] == (2,6):\r\n if os.path.exists(\"\/proc\/\"+str(os.getpid())+\"\/smaps\"):\r\n return 1\r\n if (2,6,1) <= kv <= (2,6,9):\r\n return -1\r\n return 0\r\n else:\r\n return 1\r\n\r\nvm_accuracy = shared_val_accurate()\r\nif vm_accuracy == -1:\r\n sys.stderr.write(\"Warning: Shared memory is not reported by this system.\\n\")\r\n sys.stderr.write(\"Values reported will be too large.\\n\")\r\nelif vm_accuracy == 0:\r\n sys.stderr.write(\"Warning: Shared memory is not reported accurately by this system.\\n\")\r\n sys.stderr.write(\"Values reported could be too large.\\n\")<\/pre>\nShare this:<\/h3>