FAQ

Schedule did not fire

Check these in order:

  1. Timezone. tz is set, and set to what you meant. A missing timezone is the most common cause, and daylight-saving transitions are the most common moment for it to surface.
  2. enabled = true.
  3. Registration. The named agent exists, or the named workflow is registered for the system principal.
  4. last_err. The schedule's own error state usually names the problem outright.

Check last_err first if the schedule fired before and then stopped. A schedule that has never fired is more likely a timezone or registration problem.

  • Config is read at startup. The server must be restarted after adding or changing a schedule file.
  • Instances must agree. Every server instance should run the same config revision.

A forbid schedule returning a conflict is the overlap policy working.

One case specific to agent-created jobs (scheduled_jobs): the model scheduled a past datetime. Models date-guess from stale training data, so an unanchored "tomorrow" can land in the past. Creation fails rather than silently never firing: the error names the current server time, so the model self-corrects in one retry - and every agent prompt carries today's date, which keeps the first guess anchored.

Next

On this page