Skip to content

Spinoff is a framework for writing distributed, fault tolerant and scalable applications in Python. Spinoff is based on the Actor Model and borrows from Akka (location transparency, actor references, etc) and Erlang (processes, nodename@hosthost style node references (not implemented yet)).

License

magicbill/spinoff

 
 

Repository files navigation

Overview

Spinoff is a framework for writing distributed, fault tolerant and scalable applications in Python. <build status indicator on travis-ci.org/eallik/spinoff>

Spinoff is based on the Actor Model and borrows from Akka (location transparency, actor references, etc) and Erlang (processes, nodename@hosthost style node references (not implemented yet)).

Spinoff has been built using Gevent and ZeroMQ for remoting.

Spinoff is currently under continuous development but is nevertheless usable for writing real applications—it can also be fault tolerant towards some of the bugs in its own code.

Hype

"anyway, spinoff is really super nice to play with all this stuff because it's cheap to write code in it it really shines at making quick prototypes"

...so let's hope one day there will be a testimonial here saying "it really shines at making production software".

Example

The following is only a very small "peek preview" style example of what the framework can do. More examples and full documentation will follow soon.

# spinoff/examples/example1.py
from gevent import sleep, with_timeout
from spinoff.actor import Actor
from spinoff.util.logging import dbg


class ExampleProcess(Actor):
    def run(self):
        child = self.spawn(ExampleActor)

        while True:
            dbg("sending greeting to %r" % (child,))
            child << 'hello!'

            dbg("waiting for ack from %r" % (child,))
            with_timeout(5.0, self.get, 'ack')

            dbg("got 'ack' from %r; now sleeping a bit..." % (child,))
            sleep(1.0)


class ExampleActor(Actor):
    def pre_start(self):
        dbg("starting")

    def receive(self, msg):
        dbg("%r from %r" % (content, self.sender))
        self.sender << 'ack'

    def post_stop(self):
        dbg("stopping")

The example can be run using the following command:

$ spin spinoff.examples.example1.ExampleProcess

or

$ spin spinoff.examples.example1.ExampleProcess

Distributed Example (with Remoting)

# spinoff/examples/example2.py
from gevent import sleep, with_timeout

from spinoff.actor import Actor
from spinoff.util.logging import dbg


class ExampleProcess(Actor):
    def run(self, other_actor):
        if isinstance(other_actor, str):
            other_actor = self.node.lookup_str(other_actor)
        while True:
            dbg("sending greeting to %r" % (other_actor,))
            other_actor << 'hello!'

            dbg("waiting for ack from %r" % (other_actor,))
            with_timeout(5.0, self.get, 'ack')

            dbg("got 'ack' from %r; now sleeping a bit..." % (other_actor,))
            sleep(1.0)


class ExampleActor(Actor):
    def pre_start(self):
        dbg("starting")

    def receive(self, msg):
        dbg("%r from %r" % (msg, self.sender))
        self.sender << 'ack'

    def post_stop(self):
        dbg("stopping")

The example can be run using the following commands:

$ spin -pid node1.pid -nid localhost:9700 spinoff.examples.example2.ExampleActor -name other
$ spin -pid node2.pid -nid localhost:9701 spinoff.examples.example2.ExampleProcess -using "other_actor='localhost:9700/other'"

Same Distributed Code without Remoting

The following example demonstrates how it's possible to run the same code, unmodified, in a single thread with no network/remoting involved whatsoever. It's an illustration of how actors (components) written using Spinoff are xagnostic of how they are wired to work with other actors and can thus be viewed as abstract components containing only pure domain logic and no low level transportation and topology details.

# spinoff/examples/example2_local.py
from spinoff.actor import Actor
from spinoff.util.logging import dbg

from .example2 import ExampleProcess, ExampleActor


class LocalApp(Actor):
    def run(self):
        dbg("spawning ExampleActor")
        actor1 = self.spawn(ExampleActor)

        dbg("spawning ExampleProcess")
        self.spawn(ExampleProcess.using(other_actor=actor1))

        self.get()  # so that the entire app wouldn't exit immediately

The example can be run using the following commands:

bash $ spin spinoff.examples.example2_local.LocalApp One might be tempted to ask, then, what is the difference between remoting frameworks such as CORBA and Spinoff. The difference is that actors define clear boundaries where remoting could ever be used, as opposed to splitting a flow of tightly coupled logic into two nodes on the network, which, still providing valid output, can degrade in performance significantly. This is not to say that actors with location transparency suffer none of such issues but the extent to which the problem exists is, arguably, an order of magnitude lower.

About

Spinoff is a framework for writing distributed, fault tolerant and scalable applications in Python. Spinoff is based on the Actor Model and borrows from Akka (location transparency, actor references, etc) and Erlang (processes, nodename@hosthost style node references (not implemented yet)).

Resources

License

Stars

Watchers

Forks

Packages

No packages published