asyncactor 0.16.2

Theory of operation

Assume that you have a non-lossy network with a bounded latency (let’s assume one second). Assume further that you’d like to know within ten seconds whether your node is still online.

AsyncActor sends one beacon message every seven to nine seconds. The message includes a list of N previous hosts who have transmitted the beacon; the host that’s last in this list will be the next transmitter.

The time slot starting at the seven-second mark is used for random hosts which would like to enter the beacon sending business. This is somewhat likely if the list of hosts is currently smaller than N. The slot at eight seconds is used for the hosts at the end of the list; the last host will send first, but if its beacon is not seen then the next-to-last will send its message, and so on.

The time slot at nine seconds is used for last-resort messages, i.e. any participating host can and will send its beacon message.

Collisions are resolved at the ten-second mark, i.e. the list of messages is ordered deterministically: the winner will announce to its clients that a new slot has started and whether all N host slots are filled.

It uses anyio <https://github.com/agronholm/anyio> as its underlying async framework.

$ pip install asyncactor

$ easy_install asyncactor

$ python setup.py install

from asyncactor import client as actor
from somewhere import some_transport

async with some_transport.connect('localhost') as t:
    async with actor(t, prefix=('actor','test')) as client:
        async for client.events as m:
            print(m)

$ serf agent & # start serf agent
$ mosquitto
$ python3 -mpytest tests