Distributed K->V store using RAFT

Some experimentations with zeromq, peer-to-peer systems and consensus protocol.

How does it work

Follow pattern described in ZMQ book. Each peer has a ROUTER socket to receive messages and a DEALER socket to send message to each other peer.

There are bootstrap/anchor nodes that are known beforehand. Other nodes will connect to the network by sending a join request to them.

Not exhaustive list of things I want to implement. Will be updated from time to time.

Better logging
Frame protocol (ZmqMessage)
Finish configuration with validation
Add dockerfile
Create EnvConfig to read from environment variables
Create wrapper for zmq::poll (ZmqLoop)
Add timer to ZmqLoop
Complete ZmqLoop (remove objects)
Send ping to all other peers
Add peer reaper: when timeout, should check if some peers are not alive and KILL them.
Begin RAFT
Use msgpack for message content (first frame is address, second is message ID, third is binary content)
Simple election + timeout in raft
Generation script -> Add from in message handler
Generation script -> create enum from messages
Generation script -> Create msg_handler class
Raft: Create Raft state (persistent)
Raft: Create Raft state (in memory)
Raft: Create small cluster client to get leader ID
- on server side, need another thread to wait for client request. Threads will communicate using zmq sockets so need another loop
- on client side, just connect with zmq (python for fun?)
Logging: Find a way to log differently with node, raft and client processes
VM + State machine