Collect CPU and memory statistics for multiple process easily

Sunday 07, July 2024   |   Post link

Overview

This blog post discusses collecting process level CPU and memory statistics. This one is going to be quick - here is a tiny Python script which collects process level CPU and memory stats and outputs it on console or to a specified file.

Using the script

Pre-requisites

1. Python3
2. Tested on both Windows and Linux
3. Install psutil Python library

Install psutil

pip3 install psutil

Using the script

Syntax:
To view detailed help:
	python3 collect --help
To collect process metrics:
	python3 <list of process-names> <collection-interval-sec> <output-filepath>

Suppose I want to monitor the CPU and memory consumption of the firefox.exe process every 5 seconds at the console, I can invoke the script as:

python3 collect firefox,edge 5

I can also provide a file where the same information will be written:

python3 collect firefox,edge 5 metrics.log

You could of course redirect the console output to a file instead of specifiying a file as a parameter.

The code

# Author        : Siddharth Barman
# Date          : 2-JULY-2024
# Description   : Program to collect process statistics into a file or console. 

import psutil as ps
import sys, time

ARG_PROCESS_LIST = 1
ARG_INTERVAL_SEC = 2
ARG_OUTPUT_FILE  = 3

def default():
	print("Collect process metrics.")
	print("Syntax:")
	print("To view detailed help: run python3 collect --help")

def help():
	print("Collect process metrics.")
	print("Syntax:")
	print("To view detailed help:")
	print("  python3 collect --help")
	print("To collect process metrics: ")
	print("  python3 <list of process-names> <collection-interval-sec> <output-filepath>")
	print("  python3 collect firefox,edge 5 ./logs/metrics.log")

def print_or_write(file_handle, string):
	if file_handle == None:
		print(string)
	else:
		file_handle.write(string)
		file_handle.write("\n")

def is_process_in_list(process_name, process_name_list):
	process_name = process_name.lower()
	for pname in process_name_list:
		proc_name = pname.lower()        
		if process_name == proc_name:
			return True
	return False

def collect(process_names, interval_seconds, output_filepath):    
	if output_filepath != None:
		f = open(output_filepath, 'w')
	else:
		f = None

	try:
		process_info = {}
		while True:
			for pid in ps.pids():
				try:
					p = ps.Process(pid)
					if is_process_in_list(p.name(), process_names):                        
						line = "pid={0}, name={1}, CPU={2}, Mem={3}, args={4}".format(pid, p.name(), p.cpu_percent(interval=1) / ps.cpu_count(), 
																						p.memory_full_info().uss,  p.cmdline()[0])
						print_or_write(f, line)
				except:                
					pass            
			time.sleep(interval_seconds)            
	except KeyboardInterrupt as k:
		if output_filepath != None:
			f.close()
	except:        
		if output_filepath != None:
			f.close()
		raise

def main():
	if len(sys.argv) == 1:
		default()
		exit()

	if (sys.argv[1] == "--help"):
		help()
		exit()

	process_names = sys.argv[ARG_PROCESS_LIST].split(",")
	interval_seconds = int(sys.argv[ARG_INTERVAL_SEC])    
	output_filepath = None
	if len(sys.argv) > 3:
		output_filepath = sys.argv[ARG_OUTPUT_FILE]
	
	collect(process_names, interval_seconds, output_filepath)

main()

There are a number of additional features I think this script could have like - monitoring a set of PIDs instead of having to provide a process name. If you have other ideas, I would be interested to know about them. That's all for now, have a great day!

Categories: Python (1) Testing (1)

Tags: Logging(1) Processes(1)