Purgatory¶

Purgatory is a Python library for robust failure management, built to prevent repetitive errors in synchronous and asynchronous systems through an easy-to-use circuit breaker pattern.

Circuit breakers are essential in Python failure management, as they help maintain system safety by preventing cascading failures when dependencies or external services are temporarily unavailable, ensuring that your application remains stable even under unpredictable conditions.

Note

Circuit breakers detect failures and encapsulates the logic of preventing a failure from constantly recurring, during maintenance, temporary external system failure or unexpected system difficulties.

Source: https://en.wikipedia.org/wiki/Circuit_breaker_design_pattern

Features¶

Purgatory supports the creation of many circuit breakers easily, that can be used as context manager or decorator.

Circuit breaker can be asynchronous or synchronous.

Purgatory allows you to store circuit states in multiple backends, like in-memory or Redis, or even customize the storage backend to suit specific needs. This flexibility is useful for distributed systems, where keeping state in a shared location like Redis can improve synchronization.

Purgatory supports monitoring via event hooks, allowing developers to track circuit state changes and receive notifications on important events (like opening, closing, or half-opening a circuit). This feature provides better insight into the system’s health and makes it easier to react to circuit changes in real time.

Purgatory is fully typed and rigorously tested, which can make it easier to debug and integrate into larger, type-safe codebases.

Usage¶

Example with a context manager for an async API¶

from purgatory import AsyncCircuitBreakerFactory

circuitbreaker = AsyncCircuitBreakerFactory()
async with await circuitbreaker.get_breaker("my_circuit"):
   ...

Example with a decorator¶

from purgatory import AsyncCircuitBreakerFactory

circuitbreaker = AsyncCircuitBreakerFactory()

@circuitbreaker("another circuit")
async def function_that_may_fail():
   ...

Example with a context manager for an synchronous API¶

from purgatory import SyncCircuitBreakerFactory

circuitbreaker = SyncCircuitBreakerFactory()
with circuitbreaker.get_breaker("my_circuit"):
   ...

Circuit breakers states and monitoring¶

The state of every circuits can be stored in memory, shared in redis, or be completly customized.

It also support monitoring, using event hook.

Purgatory is fully typed and fully tested.

Alternatives¶

Here is a list of alternatives, which may or may not support coroutines.

aiobreaker - https://pypi.org/project/aiobreaker/

circuitbreaker - https://pypi.org/project/circuitbreaker/

pycircuitbreaker - https://pypi.org/project/pycircuitbreaker/

pybreaker - https://pypi.org/project/pybreaker/

lasier - https://pypi.org/project/lasier/

breakers - https://pypi.org/project/breakers/

pybreaker - https://pypi.org/project/pybreaker/

python-circuit - https://pypi.org/project/python-circuit/

Why another Circuit Breaker implementation ?¶

Purgatory has been develop to be used in Blacksmith where the library aiobreaker was used but I encountered limitation so, I decide to build my own implementation that feet well with Blacksmith.

Developper Manual: