sajmaru / raft-distributed-systems Goto Github PK

#Phase 3 : Raft Leader Elections The following readme file has been provided to explain the project structure and how to structure/organize your code.

You must use only a single image for your RAFT implementation and scale it up to atleast a 5 node cluster.

1. Node : Your raft implementation image should be built from this directory. Add your Dockerfile and relevant code for Raft implementation here

2. Controller : Controller is how we will interact with your cluster, it will send out JSON requests to your cluster and your implementation should be able to handle it.

   • Test_file.py - This file will send a JSON request to the mentioned node.

3. docker-compose.yml: configuration to scale your image and create your cluster.

4. Message.json: Message structure which your Raft Implementation will use for message passing between containers.
   Note - Please do not remove the controller container. You have been provided a simple test case. Utilize this to interact with your cluster, we will test your implementation exhaustively so make sure to create more tests and do your own testing.

We will be using sockets to communicate between different containers. Examples will be provided(Python) to get you up and running with the same, please make sure you go through that.

We will use JSON requests to communicate between the nodes. Your message should contain the following fields -

• sender_name : node sending the message
• request : type of Message/Request
• term : current term of the sender
• key : message key
• value : message value

Different fields need to be populated for different messages.

1. Node Requests - Sender name will be the name of the sender

• APPEND_RPC - Leader must send APPEND_RPC messages to let other nodes know of its presence
• VOTE_REQUEST - Candidate should send VOTE_REQUEST messages to ask for votes from other nodes
• VOTE_ACK - VOTE_ACK must be sent back if the vote is granted to a requested vote, if not then drop/ignore.

2. Controller Requests

• CONVERT_FOLLOWER - convert the node to the follower state
• TIMEOUT - timeout the node immediately
• SHUTDOWN - shutdown all threads running on the node, no errors should be thrown
• LEADER_INFO - return leader info with key=LEADER and value=Node? which is the current leader

Sender name for node requests will be the node that sent the message, and "Controller" for all requests sent by the controller

Your JSON requests must be serialized and encoded to utf-8 before transmission. "Controller/convert_follower_node1.py" example shows how to do that in python. Post that the msg will be sent as UDP packets i.e. between nodes as well as the controller.

Similarly, at the receiving end you must decode the JSON request on the receiving end before beginning your processing. Make sure you're able to read the JSON request sent out by the "Controller/convert_follower_node1.py" example

You can use docker assigned IPs or container names on bridged networks in conjunction with 5555 as the port for all communication.

Controller will be used to send different controller requests mentioned in section IV to interact with your system and test out different configurations. A sample test request has been provided for your reference.

HOW TO USE THE TEST CASE?

• Option1 - run convert_follower_node1.py from the DockerFile.
• Option2 - Open Controller CLI and run the following command to send a JSON Controller request

python convert_follower_node1.py

Note : After starting your implementation, few seconds will be given for your system to stabilize and elect a leader before evaluation starts.