Raft replication in Python


Keywords
python, raft, replication, asynchronous
License
MIT
Install
pip install raftos==0.1

Documentation

raftos

Build Status PyPI version

Asynchronous replication framework based on Raft Algorithm for fault-tolerant distributed systems.

Install

pip install raftos

Register nodes on every server

import raftos


loop.create_task(
    raftos.register(
        # node running on this machine
        '127.0.0.1:8000',

        # other servers
        cluster=[
            '127.0.0.1:8001',
            '127.0.0.1:8002'
        ]
    )
)
loop.run_forever()

Data replication

counter = raftos.Replicated(name='counter')
data = raftos.ReplicatedDict(name='data')


# value on a leader gets replicated to all followers
await counter.set(42)
await data.update({
    'id': 337,
    'data': {
        'amount': 20000,
        'created_at': '7/11/16 18:45'
    }
})

In case you only need consensus algorithm with leader election

await raftos.wait_until_leader(current_node)

or

if raftos.get_leader() == current_node:
    # make request or respond to a client

or

raftos.configure({
    'on_leader': start,
    'on_follower': stop
})

Whenever the leader falls, someone takes its place.

Paper & Video