Many threads to write log file at same time in Python

Question

I am writing a script to retrieve WMI info from many computers at the same time then write this info in a text file:

f = open("results.txt", 'w+') ## to clean the results file before the start


def filesize(asset):  
    f = open("results.txt", 'a+')  
    c = wmi.WMI(asset)  
    wql = 'SELECT FileSize,Name FROM CIM_DataFile where (Drive="D:" OR Drive="E:") and Caption like "%file%"'  
    for item in c.query(wql):  
        print >> f, item.Name.split("\\")[2].strip().upper(), str(item.FileSize)  




class myThread (threading.Thread):  
    def __init__(self,name):  
        threading.Thread.__init__(self)  
        self.name = name  
    def run(self):  
        pythoncom.CoInitialize ()  
        print "Starting " + self.name       
        filesize(self.name)  
        print "Exiting " + self.name  



thread1 = myThread('10.24.2.31')  
thread2 = myThread('10.24.2.32')  
thread3 = myThread('10.24.2.33')  
thread4 = myThread('10.24.2.34')  
thread1.start()  
thread2.start()  
thread3.start()  
thread4.start()

The problem is that all threads writing at the same time.

Answer 1

You can simply create your own locking mechanism to ensure that only one thread is ever writing to a file.

import threading
lock = threading.Lock()

def write_to_file(f, text, file_size):
    lock.acquire() # thread blocks at this line until it can obtain lock

    # in this section, only one thread can be present at a time.
    print >> f, text, file_size

    lock.release()

def filesize(asset):  
    f = open("results.txt", 'a+')  
    c = wmi.WMI(asset)  
    wql = 'SELECT FileSize,Name FROM CIM_DataFile where (Drive="D:" OR Drive="E:") and Caption like "%file%"'  
    for item in c.query(wql):  
        write_to_file(f, item.Name.split("\\")[2].strip().upper(), str(item.FileSize))

You may want to consider placing the lock around the entire for loop for item in c.query(wql): to allow each thread to do a larger chunk of work before releasing the lock.

Answer 2

print is not thread safe. Use the logging module instead (which is):

import logging
import threading
import time


FORMAT = '[%(levelname)s] (%(threadName)-10s) %(message)s'

logging.basicConfig(level=logging.DEBUG,
                    format=FORMAT)

file_handler = logging.FileHandler('results.log')
file_handler.setFormatter(logging.Formatter(FORMAT))
logging.getLogger().addHandler(file_handler)


def worker():
    logging.info('Starting')
    time.sleep(2)
    logging.info('Exiting')


t1 = threading.Thread(target=worker)
t2 = threading.Thread(target=worker)

t1.start()
t2.start()

Output (and contents of results.log):

[INFO] (Thread-1  ) Starting
[INFO] (Thread-2  ) Starting
[INFO] (Thread-1  ) Exiting
[INFO] (Thread-2  ) Exiting

Instead of using the default name (Thread-n), you can set your own name using the name keyword argument, which the %(threadName) formatting directive then will then use:

t = threading.Thread(name="My worker thread", target=worker)

---

(This example was adapted from an example from Doug Hellmann's excellent article about the threading module)

Answer 3

For another solution, use a Pool to calculate data, returning it to the parent process. This parent then writes all data to a file. Since there's only one proc writing to the file at a time, there's no need for additional locking.

Note the following uses a pool of processes, not threads. This makes the code much simpler and easier than putting something together using the threading module. (There is a ThreadPool object, but it's not documented.)

source

import glob, os, time
from multiprocessing import Pool

def filesize(path):
    time.sleep(0.1)
    return (path, os.path.getsize(path))

paths = glob.glob('*.py')
pool = Pool()                   # default: proc per CPU

with open("results.txt", 'w+') as dataf:
    for (apath, asize) in pool.imap_unordered(
            filesize, paths,
    ):
        print >>dataf, apath,asize

output in results.txt

zwrap.py 122
usercustomize.py 38
tpending.py 2345
msimple4.py 385
parse2.py 499