django-pglock
performs advisory locks, table locks, and helps manage blocking locks.
Here's some of the functionality at a glance:
- pglock.advisory for application-level locking, for example, ensuring that tasks don't overlap.
- pglock.model for locking an entire model.
- pglock.timeout for dynamically setting the timeout to acquire a lock.
- pglock.prioritize to kill blocking locks for critical code, such as migrations.
- The PGLock and BlockedPGLock models for querying active and blocked locks.
- The
pglock
management command that wraps the models and provides other utilities.
Use pglock.advisory to acquire a Postgres advisory lock:
import pglock
with pglock.advisory("my_lock_id"):
# This code blocks until the "my_lock_id" lock is available
Above our code will block until the lock is available, meaning
no instances of the function will run simultaneously. Use
the timeout
argument to configure how long to wait for
the lock. A timeout of zero will return immediately:
with pglock.advisory("my_lock_id", timeout=0) as acquired:
if acquired:
# The lock is acquired
Use side_effect=pglock.Raise
to raise a django.db.utils.OperationalError
if
the lock can't be acquired. When using the decorator, you can also use
side_effect=pglock.Skip
to skip the function if the lock can't be acquired:
@pglock.advisory(timeout=0, side_effect=pglock.Skip)
def non_overlapping_func():
# This function will not run if there's another one already running.
# The decorator lock ID defaults to <module_name>.<function_name>
pglock.model can take a lock on an entire model during a transaction. For example:
from django.db import transaction
import pglock
with transaction.atomic():
pglock.model("auth.User")
# Any operations on auth.User will be exclusive here. Even read access
# for other transactions is blocked
pglock.model uses Postgres's LOCK statement, and it accepts the lock mode as a argument. See the Postgres docs for more information.
Note
pglock.model is similar to pglock.advisory. Use the timeout
argument
to avoid waiting for locks, and supply the appropriate side_effect
to adjust runtime behavior.
pglock.prioritize will terminate any locks blocking the wrapped code:
import pglock
@pglock.prioritize()
def my_func():
# Any other statements that have conflicting locks will be killed on a
# periodic interval.
MyModel.objects.update(val="value")
pglock.prioritize is useful for prioritizing code, such as migrations, to avoid situations where locks are held for too long.
Use pglock.timeout to dynamically set Postgres's lock_timeout runtime setting:
import pglock
@pglock.timeout(1)
def do_stuff():
# This function will throw an exception if any code takes longer than
# one second to acquire a lock
Use pglock.models.PGLock to query active locks. It wraps Postgres's pg_locks view. Use pglock.models.BlockedPGLock to query locks and join the activity that's blocking them.
Use python manage.py pglock
to view and kill locks from the command line. It has
several options for dynamic filters and re-usable configuration.
django-pglock
is compatible with Python 3.7 - 3.10, Django 2.2 - 4.1, and Postgres 10 - 15.
We recommend everyone first read:
- :ref:`installation` for how to install the library.
After this, there are several usage guides:
- :ref:`advisory` for using advisory locks.
- :ref:`model` for locking models.
- :ref:`timeout` for setting dynamic lock timeouts.
- :ref:`prioritize` for prioritizing code that may be blocked.
- :ref:`proxy` for an overview of the proxy models and custom queryset methods.
- :ref:`command` for using and configuring the management command.
Core API information exists in these sections:
- :ref:`settings` for all available Django settings.
- :ref:`module` for documentation of the
pglock
module and models. - :ref:`release_notes` for information about every release.
- :ref:`contributing` for details on contributing to the codebase.