# Scheduling a changeset

URL: https://kywi-docs-production.up.railway.app/governing/scheduling-a-changeset
Type: docPage

Schedule sets a changeset's publish time, but nothing fires it unless the same background sweep that handles content's Published At is running.

Scheduling a changeset sets a time for it to publish later instead of right now. An editor or above can set it, same as building the changeset in the first place.

## What Schedule actually does

Setting a schedule only records the target time — it doesn’t queue the publish anywhere by itself. A changeset goes live once a recurring background sweep notices its scheduled time has passed and it’s still sitting open, waiting. That’s the same sweep that handles individual pages’ Published At and Expires At dates (see [Publishing, scheduling and expiry](/authoring/publishing-scheduling-and-expiry/)) — there’s no separate scheduler just for changesets.

## Nothing fires it unless the sweep is running

The background sweep isn’t started automatically — whoever set this site up has to turn it on. If a changeset’s scheduled time comes and goes and it’s still sitting open, don’t assume the date was wrong: ask whoever operates the site whether the sweep is actually running before you keep re-checking it.

## A scheduled publish is not recorded in the audit log

A publish fired by the sweep has no user behind it, and it does not go through the route a person presses the button on, so it writes nothing to the audit log: no entry against the changeset, and none against the pages it published. A changeset published by hand records both, naming the person who did it.

That matters most when the changeset carries edits that were waiting in the approvals queue. Publishing clears them from the queue either way, but on the scheduled path nothing is left saying who approved them. If the trail matters more than the timing, publish it yourself.

## A stuck changeset backs off instead of spamming

If a scheduled changeset fails to publish — an item in it can no longer be applied, say — it isn’t retried on every single sweep tick forever. The wait between retries doubles each time it fails again, capped at once an hour, so a changeset that can’t go live stays visibly stuck at open without flooding anything in the meantime.

That one stuck changeset doesn’t hold up the others, either: each scheduled changeset publishes independently, so the ones after it still go out on time even while it sits waiting.
