Graceful shutdown

await q.shutdown(10_000) // wait up to 10s for in-flight jobs

shutdown(graceMs):

1. Stops polling for new jobs immediately.
2. Fires signal.abort() on every in-flight job's AbortSignal.
3. Waits up to graceMs for handlers to return.
4. Closes the database pool.

Wire it to SIGTERM

const q = await Eddyq.connect(process.env.DATABASE_URL!)
// ... q.work(...), await q.start() ...

for (const sig of ['SIGINT', 'SIGTERM']) {
  process.once(sig, async () => {
    await q.shutdown(10_000)
    process.exit(0)
  })
}

Inside a handler

Always pass signal through to anything that supports it — fetch, AbortController, your DB driver's cancellation API.

q.work('long.task', async ({ signal }) => {
  const res = await fetch(url, { signal })
  // signal will throw on shutdown, your handler unwinds, eddyq retries the job
})

What happens if a job exceeds the grace period?

The connection closes mid-execution. The job remains running until its lease expires (default 5 minutes), then becomes available for retry.