Struct Executor 

pub struct Executor { /* private fields */ }

Executor configuration for a durable worker.

Implementations

impl Executor

pub fn new(db: DatabaseConnection, executor_id: String) -> Self

pub fn db(&self) -> &DatabaseConnection

pub fn executor_id(&self) -> &str

pub async fn heartbeat(&self) -> Result<(), DurableError>

Write (or upsert) a heartbeat row for this executor.

pub fn start_heartbeat(&self, config: &HeartbeatConfig) -> JoinHandle<()>

Spawn a background tokio task that calls heartbeat() every config.heartbeat_interval.

Returns a JoinHandle so the caller can abort it on graceful shutdown.

pub async fn recover_stale_tasks( &self, staleness_threshold: Duration, ) -> Result<Vec<RecoveredTask>, DurableError>

Claim stale tasks from dead workers in a single atomic operation. Uses FOR UPDATE SKIP LOCKED so multiple executors can run this concurrently without blocking each other — each executor claims a disjoint set of tasks.

Catches three cases:

1. Tasks whose executor_id has a stale heartbeat (> staleness_threshold)
2. Tasks whose executor_id has no heartbeat row at all
3. Orphaned tasks with NULL executor_id that have been RUNNING too long

Claimed tasks go directly to RUNNING (no intermediate PENDING state) with started_at reset and executor_id set to this executor.

pub async fn recover(&self) -> Result<Vec<RecoveredTask>, DurableError>

Claim timed-out tasks in a single atomic operation. Uses FOR UPDATE SKIP LOCKED for safe concurrent claiming.

A task is considered timed-out if:

• status = 'RUNNING' and started_at IS NOT NULL
• timeout_ms has elapsed, or deadline_epoch_ms has passed

Tasks without timeout_ms set are never considered timed-out.