> ## Documentation Index > > Fetch the complete documentation index at: https://lithi.ai/llms.txt > > Use this file to discover all available pages before exploring further. --- title: Pause, revoke and recover a schedule description: >- Pause a schedule to stop new occurrences, learn what happens to work already running, then read back both the pause and the resume before you trust either. route: /docs/compute/automation/pause-revoke-and-recover page_id: docs-compute-automation-pause-revoke-and-recover page_type: howTo content_layer: guided documentation surface_profile: guided_docs audience: Administrators stopping or restarting an automation under pressure voice: D reader_question: How do I stop an automation now without breaking the work already running? primary_action: Choose an adapter for your data source_locale: en-US source_status: APPROVED source_version: docs-compute-automation-pause-revoke-and-recover-en-us-2026-09-05-v1 last_updated: '2026-09-05' robots: index claim_registry_pinned: true claim_registry_resync: "npm run governed:tsx -- scripts/build-compute-public-claim-registry.ts --write" --- ## Before you start You need an administration credential and the schedule's identifier. A work key cannot pause, resume or revoke a schedule. Decide first which of three things you want: stop new occurrences, stop a job that is already running, or remove the schedule permanently. They are separate acts. ## Pause stops new occurrences only Pause is the fast, reversible act. It stops the schedule from creating further occurrences. It does nothing to a job that is already admitted. An admitted job keeps its own snapshot and its own authority, and it finishes or fails on its own terms. Use pause when you want to stop the bleeding while you investigate. It is the one you can undo. ## What happens to work in flight An occurrence that is already a job is unaffected by the pause. Follow it through its normal states rather than trying to stop it from the schedule. If you need that job stopped as well, cancel the job itself. Cancellation is a job operation, and it moves through `CANCELLING` before it reaches `CANCELLED`. Two things survive a pause and are worth knowing. - A running job still holds its input snapshot. Pausing the schedule does not change the bytes it reads. - A delivery already in progress still writes. Stop it at the destination if you need to, not at the schedule. ## Credentials expire while you are paused A pause can outlive the authority a schedule depends on. That is the most common reason a resume looks broken. Delivery authority is deliberately time-bounded. `AUTHORITY_EXPIRED` means the permission lapsed, not that a result was lost, and the fix is to renew it and retry the delivery. Before you resume, check the source credential reference, the destination authority and the spend ceiling. Renew whichever has lapsed, then resume. See [how to renew a destination's authority](/docs/compute/configure/destinations) if the destination is the one that expired. ## Revoke is forward, never an undo Revoke removes the schedule. It does not reach back and undo occurrences that already ran, results that were delivered or charges that were incurred. Revoke, then read the schedule back, then read its events. Do it in that order, because the response only tells you the request was accepted. If a revoke refuses with a conflict, the schedule moved since your plan. Inspect it, plan again, and revoke the version you actually looked at. ## Resume and read both states back Resume is a separate act with its own read-back. Do not assume the pause reversed cleanly because the call returned. Read back four things after a resume. 1. The schedule state is active again. 2. The next occurrence times match what you expect, in the time zone you set. 3. The catch-up rule is what you want, because it decides whether missed occurrences now run. 4. The spend ceiling and any approval are current, not the ones from before the pause. Catch-up is the item that surprises people. Resuming a schedule with catch-up enabled after a long pause can create a burst of occurrences at once. ## Confirm success Watch the first occurrence after the resume. It should show a fresh quote, a current policy read and one admitted job. Compare its identity against the last occurrence before the pause. They must differ; a resumed schedule creates new occurrences, it does not reissue old ones. Keep the receipt reference from the apply. It is what you quote to support if the schedule behaves differently afterwards. [Choose an adapter for your data](/docs/compute/integrations)