How to Implement Multiprocessing-Aware Logging in Python

Multiprocessing in Python allows the creation of multiple processes that run independently. However, accessing shared resources like log files can become complex as multiple processes may attempt to write to them simultaneously.

To avoid this issue, the Python multiprocessing module provides module-level multiprocessing-aware logging capabilities. This enables the logger to prevent garbling of log messages by ensuring that only one process writes to a specific file descriptor at a time.

However, existing modules within the framework may not be multiprocessing-aware, leading to the need for alternative solutions. One approach involves creating a custom log handler that sends logging messages to the parent process via a pipe.

An implementation of this approach is provided below:

from logging.handlers import RotatingFileHandler
import multiprocessing, threading, logging, sys, traceback

class MultiProcessingLog(logging.Handler):
    def __init__(self, name, mode, maxsize, rotate):
        logging.Handler.__init__(self)

        # Set up the file handler for the parent process
        self._handler = RotatingFileHandler(name, mode, maxsize, rotate)
        
        # Create a queue to receive log messages from child processes
        self.queue = multiprocessing.Queue(-1)
        
        # Start a thread in the parent process to receive and log messages
        t = threading.Thread(target=self.receive)
        t.daemon = True
        t.start()

    def receive(self):
        while True:
            try:
                # Get a log record from the queue
                record = self.queue.get()
                
                # Log the record using the parent process's file handler
                self._handler.emit(record)
            # Exit the thread if an exception is raised
            except (KeyboardInterrupt, SystemExit):
                raise
            except EOFError:
                break
            except:
                traceback.print_exc(file=sys.stderr)

    def send(self, s):
        # Put the log record into the queue for the receiving thread
        self.queue.put_nowait(s)

    def _format_record(self, record):
        # Stringify any objects in the record to ensure that they can be sent over the pipe
        if record.args:
            record.msg = record.msg % record.args
            record.args = None
        if record.exc_info:
            dummy = self.format(record)
            record.exc_info = None
            
        return record

    def emit(self, record):
        try:
            # Format and send the log record through the pipe
            s = self._format_record(record)
            self.send(s)
        except (KeyboardInterrupt, SystemExit):
            raise
        except:
            self.handleError(record)

    def close(self):
        # Close the file handler and the handler itself
        self._handler.close()
        logging.Handler.close(self)

This custom log handler allows modules within the framework to use standard logging practices without having to be multiprocessing-aware themselves. The log messages are sent from the child processes to the parent process via a pipe, ensuring that they are not garbled and written correctly to the log file.