raft

README

CLOUD COMPUTING - UE20CS351

Project: Implementing Raft Logic in Go

In this project, you will:

1. Learn the basics of GoLang.
2. Understand the basic logic behind Raft
3. Implement part of the logic behind raft, in GoLang, for leader election and log replication.

What you're given:

You are provided with a GoLang project structure, which, when complete, will allow you to successfully demonstrate leader election and log replication via Raft. However, parts of the code are deliberately missing; Your job is to fill it in, and make sure that expected behaviour is observed in scenarios such as network partitioning.

.
├── go.mod
├── NodeLogs
│   ├── 0
│   ├── 1
│   ├── 2
│   ├── 3
│   └── 4
├── raft_cluster.go
├── raft_election_logic.go
├── raft_leader_logic.go
├── raft_node.go
├── raft_rpc_handlers.go
├── raft_test.go
├── README.md
├── server_setup.go
└── verbose
    ├── 1.log
    └── 2.log

The files server_setup.go, raft_node.go and raft_cluster.go require no modification. raft_node.go, however, contains vital information about the persistent state of a raft node itself, and is worth going through to better understand the flow of the code.

How you should start:

Start off by installing GoLang. Here is a link to the official guide.

Once you've installed Go, it's a good idea to familiarise yourself with Raft. The Raft Paper itself, in conjunction with the interactive visualisation at https://raft.github.io/, is a major help there.

You must also familiarise yourself with the basics of GoLang. Here is a great guide to help you get started. Of course, you'll probably learn most by actually playing around with code.

Familiarise yourself with Raft Leader Election and Log Replication, i.e up until but not including Section 7 of the paper; you do not need to familiarise yourself with Log Compaction.

The scenarios we deal with here do not include complete node failure, although it is trivial to account for such a case by asking recovered nodes to replay their logs. Instead, we deal with partitioned nodes; i.e, a 'disconnected node' is a node that is still functioning, but is cut off from the rest of the cluster. Think along the lines of 'its internet failed.' KNOWING THIS IS IMPORTANT FOR YOUR EVALUATION.

What's expected of you:

You are expected to implement 4 specific functionalities:

1. Implement the becomeFollower function entirely.
2. Implement the logic for a follower node to handle a received RequestVote function from a candidate.
3. Implement the logic for a candidate to handle a reply to the above RequestVote it sent out to its peers, be it successfully or unsuccessfully.
4. Implement the logic for the leader to commit its log successfully in the event of majority confirmation, or fail in case majority confirmation is not received.

These four functionalities' locations become apparent once you go through the codebase, and read through the Raft paper.

What you can use to test your code out:

go test -v -race -run Test1 > verbose/1.log

go test -v -race -run Test2 > verbose/2.log

The above command executes a couple of scenarios for your code. Read through the logs generated in the verbose directory for Test1 , and make sure to have all 5 files open in NodeLogs and watch them update in real time for Test2 (like this:)

You could also instead tail -f the files with tmux, like you did for Experiment 4.

For evaluation, you will need to walk through the output you got from the Tests 1 and 2 and explain why it's expected behaviour. You will also need to then execute test-cases given to you later and do the same. (Test 3 given to you will fail by default; it's your job to explain why and how to fix it.) (refer to the last paragraph of 'how you should start')

**Done by : **

DISHA SUNIL NIKAM : PES1UG20CS133

DIVYAM VIJAY : PES1UG20CS136

GAURAV DNYANESH MAHAJAN : PES1UG20CS150

SANIKA M RANGAYYAN : PES1UG20CS901

Documentation

Index

• Constants
• type AppendEntriesArgs
• type AppendEntriesReply
• type Cluster
  • func NewCluster(t *testing.T, n int) *Cluster
  • func (this *Cluster) DisconnectPeer(id int)
  • func (this *Cluster) ReconnectPeer(id int)
  • func (this *Cluster) Shutdown()
  • func (this *Cluster) SubmitClientCommand(serverId int, cmd interface{}) bool
• type LogEntry
• type RaftNode
  • func NewRaftNode(id int, peersIds []int, server *Server, ready <-chan interface{}) *RaftNode
  • func (this *RaftNode) GetNodeState() (id int, term int, isLeader bool)
  • func (this *RaftNode) HandleAppendEntries(args AppendEntriesArgs, reply *AppendEntriesReply) error
  • func (this *RaftNode) HandleRequestVote(args RequestVoteArgs, reply *RequestVoteReply) error
  • func (this *RaftNode) KillNode()
  • func (this *RaftNode) ReceiveClientCommand(command interface{}) bool
• type RequestVoteArgs
• type RequestVoteReply
• type Server
  • func NewServer(serverId int, peersIds []int, ready <-chan interface{}, minRPCLatency int) *Server
  • func (this *Server) AppendEntries(args AppendEntriesArgs, reply *AppendEntriesReply) error
  • func (this *Server) ConnectToPeer(peerId int, addr net.Addr) error
  • func (this *Server) DisconnectAll()
  • func (this *Server) DisconnectPeer(peerId int) error
  • func (this *Server) GetCurrentAddress() net.Addr
  • func (this *Server) RequestVote(args RequestVoteArgs, reply *RequestVoteReply) error
  • func (this *Server) SendRPCCallTo(id int, serviceMethod string, args interface{}, reply interface{}) error
  • func (this *Server) Serve()
  • func (this *Server) Shutdown()

Constants

const LogHeartbeatMessages = false

const LogVoteRequestMessages = true

Types

type AppendEntriesArgs

type AppendEntriesArgs struct {
	Term     int
	LeaderId int

	PrevLogIndex int
	PrevLogTerm  int
	Entries      []LogEntry
	LeaderCommit int

	Latency int
}

type AppendEntriesReply

type AppendEntriesReply struct {
	Term    int
	Success bool
}

type Cluster

type Cluster struct {
	// contains filtered or unexported fields
}

func NewCluster

func NewCluster(t *testing.T, n int) *Cluster

func (*Cluster) DisconnectPeer

func (this *Cluster) DisconnectPeer(id int)

DisconnectPeer disconnects a server from all other servers in the nodes.

func (*Cluster) ReconnectPeer

func (this *Cluster) ReconnectPeer(id int)

ReconnectPeer connects a server to all other servers in the nodes.

func (*Cluster) Shutdown

func (this *Cluster) Shutdown()

func (*Cluster) SubmitClientCommand

func (this *Cluster) SubmitClientCommand(serverId int, cmd interface{}) bool

SubmitClientCommand submits the command to serverId.

type LogEntry

type LogEntry struct {
	Command interface{}
	Term    int
}

type RaftNode

type RaftNode struct {
	LOG_ENTRIES bool
	// contains filtered or unexported fields
}

Main Raft Data Structure

func NewRaftNode

func NewRaftNode(id int, peersIds []int, server *Server, ready <-chan interface{}) *RaftNode

Constructor for RaftNodes

func (*RaftNode) GetNodeState

func (this *RaftNode) GetNodeState() (id int, term int, isLeader bool)

GetNodeState reports the state of this RN.

func (*RaftNode) HandleAppendEntries

func (this *RaftNode) HandleAppendEntries(args AppendEntriesArgs, reply *AppendEntriesReply) error

func (*RaftNode) HandleRequestVote

func (this *RaftNode) HandleRequestVote(args RequestVoteArgs, reply *RequestVoteReply) error

RequestVote RPC. This is the function that is executed by the node that RECEIVES the RequestVote.

func (*RaftNode) KillNode

func (this *RaftNode) KillNode()

Kills a RaftNode and sets its state to Dead

func (*RaftNode) ReceiveClientCommand

func (this *RaftNode) ReceiveClientCommand(command interface{}) bool

Either handle Command or tell to divert it to Leader

type RequestVoteArgs

type RequestVoteArgs struct {
	Term         int
	CandidateId  int
	LastLogIndex int
	LastLogTerm  int

	Latency int
}

Handles an incoming RPC RequestVote request

type RequestVoteReply

type RequestVoteReply struct {
	Term        int
	VoteGranted bool
}

type Server

type Server struct {
	RPCServer *rpc.Server
	// contains filtered or unexported fields
}

Server

func NewServer

func NewServer(serverId int, peersIds []int, ready <-chan interface{}, minRPCLatency int) *Server

func (*Server) AppendEntries

func (this *Server) AppendEntries(args AppendEntriesArgs, reply *AppendEntriesReply) error

func (*Server) ConnectToPeer

func (this *Server) ConnectToPeer(peerId int, addr net.Addr) error

func (*Server) DisconnectAll

func (this *Server) DisconnectAll()

func (*Server) DisconnectPeer

func (this *Server) DisconnectPeer(peerId int) error

func (*Server) GetCurrentAddress

func (this *Server) GetCurrentAddress() net.Addr

func (*Server) RequestVote

func (this *Server) RequestVote(args RequestVoteArgs, reply *RequestVoteReply) error

func (*Server) SendRPCCallTo

func (this *Server) SendRPCCallTo(id int, serviceMethod string, args interface{}, reply interface{}) error

func (*Server) Serve

func (this *Server) Serve()

func (*Server) Shutdown

func (this *Server) Shutdown()