pid

PidFile class featuring:

• stale detection
• pidfile locking (fcntl)
• chmod (default is 0o644)
• chown
• custom exceptions

Context Manager, Daemons and Logging

PidFile can be used as a context manager:

from pid import PidFile
import os

with PidFile('foo') as p:
  print(p.pidname) # -> 'foo'
  print(p.piddir) # -> '/var/run' But you can modify it when initialize PidFile.
  print(os.listdir('/var/run')) # -> ['foo.pid']

# pid file will delete after 'with' literal.

Logging to file is also possible when using PidFile with a daemon context manager (e.g. python-daemon). This requires some care in handling the open files when the daemon starts to avoid closing them, which causes problems with the logging. In particular, the open handlers should be preserved:

import sys
import logging
import logging.config

import daemon
from pid impor PidFile

logging.config.fileConfig(fname="logging.conf", disable_existing_loggers=False)
log = logging.getLogger(__name__)

PIDNAME = "/tmp/mydaemon.pid"

def get_logging_handles(logger):
    handles = []
    for handler in logger.handlers:
        handles.append(handler.stream.fileno())
    if logger.parent:
        handles += get_logging_handles(logger.parent)
    return handles

def daemonize():
  file_preserve = get_logging_handles(logging.root)
  pid_file = PidFile(pidname=PIDNAME)

  with daemon.DaemonContext(stdout=sys.stdout,
                            stderr=sys.stderr,
                            stdin=sys.stdin,
                            pidfile=_pid_file,
                            files_preserve=files_preserve):

    run_daemon_job()
  print("DONE!")

if __name__ == "__main__":
  daemonize()

This assumes a logging.conf file has been created, see e.g. basic tutorial for logging.

Decorator

PidFile can also be used a a decorator:

from pid.decorator import pidfile

@pidfile()
def main():
  pass

if __name__ == "__main__":
  main()

Exception Order

In default mode PidFile will try to acquire a file lock before anything else. This means that normally you get a PidFileAlreadyLockedError instead of the PidFileAlreadyRunningError when running a program twice.

If you just want to know if a program is already running its easiest to catch just PidFileError since it will capture all possible PidFile exceptions.

Behaviour

Changes in version 2.0.0 and going forward:

• pid is now friendly with daemon context managers such as python-daemon where the PidFile context manager is passed as a parameter. The new corrected behaviour will ensure the process environment is determined at the time of acquiring/checking the lock. Prior behaviour would determine the process environment when instancing the class which may result in incorrect determination of the PID in the case of a process forking after instancing PidFile.

• Cleanup of pidfile on termination is done using atexit module. The default SIGTERM handler doesn't cleanly exit and therefore the atexit registered functions will not execute. A custom handler which triggers the atexit registered functions for cleanup will override the default SIGTERM handler. If a prior signal handler has been configured, then it will not be overridden.