Raft Consensus

This is an implementation of the Raft consensus algorithm in Python. You can read more about the algorithm [here] (https://raft.github.io/raft.pdf) or opt for a visual representation of how it works here.

Tech Used:

• Python
• Protocol Buffers
• gRPC

Usage:

Install and activate the virtual environment then compile the .proto file using:

pip install virtualenv

virtualenv venv

source venv/bin/activate

pip install -r requirements.txt

python3 -m grpc_tools.protoc raft.proto --proto_path=. --python_out=. --grpc_python_out=.

Then, create a config.conf file in the same directory with the desired server addresses and port numbers. An example of config.conf file:

    0 127.0.0.1 50000
    1 127.0.0.1 50001
    2 127.0.0.1 50002

The application consists of 2 main parts:

• Server - server.py:

  These are the main nodes that essentially implement the algorithm and go through the cycles of Leader, Candidate, Follower and perform operations such as Leader Election and Log Replication.

  The server nodes output simple updates of when they change their state (e.g. leader -> follower), make or receive votes, and answer queries received from the client.

  To run a server node, simply run an instance of server.py and pass the corresponding number associated with this server from config.conf file. Here are two examples of a server node's output:

  Example 1:

    >python3 server.py 0
    The server starts at 127.0.0.1:50000
    I am a follower. Term: 0
    The leader is dead
    I am a candidate. Term: 1
    Voted for node 0
    Votes received
    I am a leader. Term: 1
    Command from client: getleader
    0 127.0.0.1:50000
  

  Example 2:

    >python3 server.py 0
    The server starts at 127.0.0.1:50000
    I am a follower. Term: 0
    The leader is dead
    I am a candidate. Term: 1
    Voted for node 0
    Votes received
    I am a leader. Term: 1
    Command from client: getleader
    0 127.0.0.1:50000
    Command from client: suspend 2
    Sleeping for 2 seconds
    I am a follower. Term: 2
    Command from client: getleader
    2 127.0.0.1:50002
  

• Client - client.py:

  Client nodes are essentially "users" of the system. They can perform the following operations:

  • connect SERVER_ADDRESS SERVER_PORT_NUMBER: Connects to the corresponding server node and issues all subsequent commands to that node until either the node dies or connect is called on another server node.

  • getleader: Sends a request to the currently connected server to identify the current leader of the raft and prints the output.

  • setval KEY VAL: Asks the currently connected server to save VAL to the distributed system under KEY.

  • getval KEY: Asks the currently connected server to return the currently saved VAL under KEY in the distributed system.

  • suspend PERIOD: Commands the currently connected server to sleep for PERIOD seconds, during which that server node basically exists the system and re-enters after PERIOD seconds.

  • quit: Terminates the Client.

  To run a client node, simply run an instance of client.py as such:

  Example 1:

    >python3 client.py
    The client starts
    > connect 127.0.0.1 50001
    > getleader
    0 127.0.0.1:50000
    > getval key1
    None
    > setval key1 100
    > getval key1
    100
  

  Example 2:

    >python3 client.py
    The client starts
    >connect 127.0.0.1 50000
    >getleader
    0 127.0.0.1:50000
    >suspend 2
    >getleader
    2 127.0.0.1:50002
    >quit
    The client ends
  

Author Contact Info:

Jaffar Totanji
Email: jaafarti@gmail.com
Telegram: @KuroHata7

Ahmad Haidar
Email: togoyany@gmail.com
Telegram: @XmtosX

P.S: This work was done as a solution to an assignment given to us as part of the Distributed and Network Programming course at IU.