Policies

Policies govern how Wakeplane handles concurrent execution, missed runs, timeouts, and failures. They are defined per schedule and apply to every occurrence.

Overlap Policy

forbid (default)

The new run is not claimed until the active count drops below max_concurrency. The pending run waits in the queue.

allow

New runs start regardless of how many are already active, up to max_concurrency.

queue_latest

All pending runs except the most recent one are skipped. Only the latest pending run is dispatched when capacity opens.

replace

Active runs receive a cancellation signal. All pending runs except the most recent are skipped. The latest pending run is dispatched once the active run exits.

replace is cooperative and best-effort. If the active executor does not stop promptly, behavior degrades toward queue_latest.

Misfire Policy

run_once_if_late (default)

Run exactly one overdue occurrence, even if multiple were missed.

skip

Skip all overdue occurrences.

catch_up

Materialize a run for every missed occurrence.

Timeout

policy.timeout_seconds sets a deadline for the executor. Default: 300 seconds.

Max Concurrency

policy.max_concurrency sets the maximum number of simultaneously active runs for a schedule. Default: 1.

Retry

retry:
  max_attempts: 5
  strategy: exponential
  initial_delay_seconds: 30
  max_delay_seconds: 900

Retries use exponential backoff and exhausted runs become dead letters. Cancellation is not retried.

Policy Interaction Example

A schedule with overlap: forbid, misfire: run_once_if_late, and retry.max_attempts: 3 will materialize one overdue run after an outage, retry failures with backoff, and dead-letter the occurrence after the retry budget is exhausted.