Skip to main content
Set retries and retry_interval (milliseconds) on a task. A delivery fails on any non-2xx response or transport error; Schedy waits and tries again until the count is exhausted, then marks the task failed. Each try is recorded as an attempt in the task’s log, with its status code, error, and duration.

​
Retry mode

retry_mode selects how the delay between retries is computed. It defaults to fixed.
  • fixed waits retry_interval between every attempt.
  • exponential waits min(retry_interval * 2^n, cap) with full jitter: each wait is a random point in [0, that ceiling], so a struggling endpoint gets backed off instead of hammered at a constant rate, and retries from many clients spread out instead of synchronising. The cap is fixed at 5 minutes.
Retries happen inline within a single run, so the backoff horizon is bounded by the process lifetime: if the process restarts mid-backoff, the task is re-queued from the start rather than resuming its remaining waits.

​
Retry-After

When a delivery fails with 429 Too Many Requests or 503 Service Unavailable and the response carries a Retry-After header, Schedy honors it. The requested wait becomes a floor under the next retry’s delay: the retry never fires sooner than the server asked, but a strategy that would wait longer anyway still does. Both header forms are accepted (Retry-After: 120 and an HTTP-date), and the wait is capped at the 5 minute backoff ceiling so a misbehaving endpoint cannot park a delivery slot indefinitely. Other statuses ignore the header.

​
Failure callback

Set SCHEDY_ON_FAILURE_URL and a task that exhausts its retries POSTs a single, best-effort notification there before going quiet, so a permanent failure at 3am is not silent:
The callback is fire-and-forget: it is never retried, and a failing callback never triggers a callback about itself. It is failure-only - success does not call back. A task can override the server-wide URL with its own on_failure_url (an absolute http(s) URL):
When set, the callback for that task goes only to on_failure_url; SCHEDY_ON_FAILURE_URL remains the fallback for tasks that don’t set one.