Feature: Support Joining a New Node as Learner (auto-discovering)

[JoshuaChi]

Summary:

Implement the ability for a new node to join the Raft cluster via a terminal command, with automatic initialization as a learner and subsequent snapshot synchronization from the leader.

Details:

1. A new node can join the cluster by executing a CLI command.
2. Upon joining, the node is automatically assigned the role of a Learner.
3. The leader will detect the new learner and proactively send a snapshot to synchronize its state.
4. Once the snapshot is successfully installed, the learner will perform a readiness check (e.g., comparing log and snapshot metadata with the leader).
5. If the learner determines it is fully up-to-date and consistent, it will request or trigger a promotion to Follower status.

Acceptance Criteria:

• CLI supports adding a new node to the cluster.
• New nodes join as Learner by default.
• Leader sends a snapshot automatically after detecting the new learner.
• Learner installs snapshot and confirms installation success.
• Learner confirms its readiness and can transition to Follower upon agreement with Leader.

[JoshuaChi]

Steps:

1. New node joins cluster with joining status
2. Leader propose membership config change
3. Leader receive majority confirmation for this config change
4. Leader commit this config change
5. New node(Learner) requests initial snapshot from Leader
6. Learner receives Snapshot stream
7. Learner been turned into pending-active status when the gap between last index of local state machine and leader commit index is under acceptance criteria
8. pending-active nodes will be turned into active follower when quorum of the cluster becomes legal(e.g. 3 nodes -> 5 nodes)

[JoshuaChi]

New node joins cluster

Principle:

• Adding new nodes without a Leader is not allowed

---

• Introduce ClusterConfig , ConfigState, JoinRequest and JoinResponse

enum ConfigState {
  // Stable state - cluster uses a single configuration
  STABLE = 0; 
  
  // Configuration Transition State - The cluster is applying configuration changes
  CONFIG_TRANSITION = 1;

  // Election State - The cluster is in the process of leader election
  ELECTION_IN_PROGRESS = 2;
}

• Modify 'ClusterMembershipChangeRequest' by adding change_type
• Modify ClusterMembership

message ClusterMembership {
    uint64 version = 1;
    repeated NodeMeta nodes = 2;
}

• Modify NodeMeta by combining ip and port into one field address

message NodeMeta {
    uint32 id = 1;
    string address = 2; //"ip:port"
    int32 role = 3;
    NodeStatus status = 4; // Add new status fields (such as active/draining)
}

• Refactor Raft Log Entry Payload Definition

Payload {
    Noop,
    Normal,
    Membership,
}

• Linearizable reads only require leadership to be confirmed via heartbeat.
• Manage Peer Channels for Newly Added Learner Nodes