Struct LeaderElection

pub struct LeaderElection {
    node_id: Arc<str>,
    state: Arc<RwLock<SharedState>>,
    last_heartbeat: Arc<RwLock<Instant>>,
}

Manages leader election in the OmniOrchestrator cluster.

The LeaderElection module is responsible for determining which node in the cluster should act as the leader. It implements a simple deterministic leader election algorithm based on node IDs to ensure that exactly one node assumes the leader role.

Leader election is a critical component in distributed systems that ensures:

• Coordination responsibilities are clearly assigned
• A single point of truth exists for cluster-wide decisions
• System stability is maintained through consistent leadership

The election process runs periodically to accommodate cluster changes such as nodes joining or leaving the system.

Fields

node_id: Arc<str>

Unique identifier for the current node

state: Arc<RwLock<SharedState>>

Shared state that tracks leadership status and cluster information

last_heartbeat: Arc<RwLock<Instant>>

Timestamp of the last heartbeat received This can be used for more sophisticated leader election algorithms that take into account node responsiveness

Implementations

impl LeaderElection

pub fn new(node_id: Arc<str>, state: Arc<RwLock<SharedState>>) -> Self

Creates a new LeaderElection instance.

Initializes the leader election module with the current node’s identity and a reference to the shared state. The last_heartbeat is initialized to the current time.

Arguments

• node_id - Unique identifier for the current node
• state - Shared state for tracking leadership status

Returns

A new LeaderElection instance ready to begin the election process

pub async fn start(&self)

Starts the leader election process.

This method begins a continuous cycle of leader elections at a fixed interval. Once started, it will periodically execute the election_cycle method to determine the current leader based on the existing cluster composition.

The election happens every 5 seconds, which provides a balance between responsiveness to cluster changes and system overhead.

Note

This method runs indefinitely in a loop and should typically be spawned in its own task or thread.

async fn election_cycle(&self)

Performs a single leader election cycle.

This method implements the core leader election algorithm, which follows these steps:

1. Retrieve all nodes in the cluster
2. Sort the nodes by ID for deterministic selection
3. Select the first node in the sorted list as the leader
4. Update the shared state with the election results

The algorithm is intentionally simple and deterministic, ensuring that all nodes will independently arrive at the same conclusion about who the leader is, without requiring additional communication.

Special Cases

• If the cluster contains only one node, that node becomes the leader.
• If the cluster contains no nodes (which shouldn’t happen as the current node should always be included), the current node becomes the leader by default.

Side Effects

• Updates the shared state to reflect the new leader
• Logs information about the election process and results