locket.py

Locket implements a lock that can be used by multiple processes provided they use the same path.

import locket

# Wait for lock
with locket.lock_file("path/to/lock/file"):
    perform_action()

# Raise error if lock cannot be acquired immediately
with locket.lock_file("path/to/lock/file", timeout=0):
    perform_action()

# Raise error if lock cannot be acquired after thirty seconds
with locket.lock_file("path/to/lock/file", timeout=30):
    perform_action()

# Without context managers:
lock = locket.lock_file("path/to/lock/file")
try:
    lock.acquire()
    perform_action()
finally:
    lock.release()

Locks largely behave as (non-reentrant) Lock instances from the threading module in the standard library. Specifically, their behaviour is:

Locks are uniquely identified by the file being locked, both in the same process and across different processes.
Locks are either in a locked or unlocked state.
When the lock is unlocked, calling acquire() returns immediately and changes the lock state to locked.
When the lock is locked, calling acquire() will block until the lock state changes to unlocked, or until the timeout expires.
If a process holds a lock, any thread in that process can call release() to change the state to unlocked.
Behaviour of locks after fork is undefined.

pitrou/locket.py

locket.py