Latency and fault tolerance library inspired by Hystrix.
Pycopine is a latency and fault tolerance library designed to isolate points of access to remote systems, services and 3rd party libraries, stop cascading failure and enable resilience in complex distributed systems where failure is inevitable.
As this copy-pasted text suggests, pycopine is heavily inspired by Hystrix.
Pycopine requires Python 3.2+, but may be backportet to 2.7 in the future.
Pycopine is not released yet.
- Detect and report failing services.
- Short-circuit services that fail on high load to help them recover.
- Monitor failure rates and performance metrics to detect bottlenecks.
- Manage thread pool and queue sizes on demand, at runtime, from everywhere.
- … (more to come)
Lets say we want to speak to a remote service that is slow, unreliable or both:
import time import random def crappy_service(input): ''' The most useless piece of code ever.''' time.sleep(5) if 'OK' != random.choice(['OK', 'OK', 'f**ck']): raise RuntimeError('We broke something.') return input
You could throw lots of threads and try/except clauses at the problem and hope to not break the internet. Or you could use pycopine:
from pycopine import Command class MyCommand(Command): ''' Does nothing with the input, but with style. ''' def run(self, input): return crappy_service(input) def fallback(self, input): return input # Run and wait for the result result = MyCommand('input').result() # Give up after 2 seconds result = MyCommand('input').result(timeout=2) # Fire and forget MyCommand('input').submit() # Do stuff in parallel foo = MyCommand('input_a').submit() bar = MyCommand('input_b').submit() results = [foo.result(), bar.result()] # Change your mind midway through foobar = MyCommand('input').submit() if foobar.wait(timeout=2): result = foobar.reault() else: foobar.cancel(RuntimeError('No time for this sh**t'))