rally_runtime/jobs

Durable background job runner backed by SQLite.

Jobs are rows in the system DB’s jobs table. A poller actor claims ready jobs with UPDATE…RETURNING, runs the handler, and marks them completed or schedules a retry. Failed jobs get quadratic backoff (5s, 20s, 45s, 80s) and are dead-lettered after max_attempts. Running jobs have a claimed_at lease so they can be reclaimed if the process crashes. Use run_once for deterministic testing without the polling actor.

Types

Job

</>

pub type Job {
  Job(id: Int, name: String, payload: BitArray, attempts: Int)
}

Constructors

Job(id: Int, name: String, payload: BitArray, attempts: Int)

JobHandler

</>

pub type JobHandler =
  fn(String, BitArray) -> Result(Nil, String)

Values

enqueue

</>

pub fn enqueue(
  db db: sqlight.Connection,
  name name: String,
  payload payload: BitArray,
  run_at run_at: Int,
) -> Nil

enqueue_in

</>

pub fn enqueue_in(
  db db: sqlight.Connection,
  name name: String,
  payload payload: BitArray,
  delay_seconds delay_seconds: Int,
) -> Nil

run_once

</>

pub fn run_once(
  db db: sqlight.Connection,
  handler handler: fn(String, BitArray) -> Result(Nil, String),
) -> Nil

run_once_at

</>

pub fn run_once_at(
  db db: sqlight.Connection,
  now now: Int,
  handler handler: fn(String, BitArray) -> Result(Nil, String),
) -> Nil

start_runner

</>

pub fn start_runner(
  db db: sqlight.Connection,
  handler handler: fn(String, BitArray) -> Result(Nil, String),
) -> Result(actor.Started(Nil), actor.StartError)