Struct LeaderElection 

pub struct LeaderElection { /* private fields */ }

Leader election using PostgreSQL advisory locks.

Advisory locks provide a simple, reliable way to elect a leader without external coordination services. Key properties:

1. Mutual exclusion: Only one session can hold a given lock ID at a time.
2. Automatic release: If the connection dies, PostgreSQL releases the lock.
3. Non-blocking try: pg_try_advisory_lock returns immediately with success/failure.

Each LeaderRole maps to a unique lock ID, allowing multiple independent leader elections (e.g., separate leaders for cron scheduler and workflow timers).

The is_leader flag uses SeqCst ordering because:

• Multiple threads read this flag to decide whether to execute leader-only code
• We need visibility guarantees across threads immediately after acquiring/releasing
• The performance cost is negligible (leadership changes are rare)

Implementations

impl LeaderElection

pub fn new( pool: PgPool, node_id: NodeId, role: LeaderRole, config: LeaderConfig, ) -> Self

Create a new leader election instance.

pub fn is_leader(&self) -> bool

Check if this node is the leader.

pub fn shutdown_receiver(&self) -> Receiver<bool>

Get a shutdown receiver.

pub fn stop(&self)

Stop the leader election.

pub async fn try_become_leader(&self) -> Result<bool>

Try to acquire leadership.

pub async fn refresh_lease(&self) -> Result<()>

Refresh the leadership lease.

pub async fn release_leadership(&self) -> Result<()>

Release leadership.

pub async fn check_leader_health(&self) -> Result<bool>

Check if the current leader is healthy.

pub async fn get_leader(&self) -> Result<Option<LeaderInfo>>

Get current leader info.

pub async fn run(&self)

Run the leader election loop.