Renew a lease or let work expire

Renew an existing lease only when the operator still needs the work. A lease is a cleanup deadline for controller-launched services, not a guarantee that their task will finish correctly. Start with an existing service identity, permission to prolong it, and the actual CLI on PATH.

Review before extending

1. Inspect the exact service. Compare owner, purpose, working directory and current process identity with the ongoing task.

2. Read its lease and any maximum-runtime limit. A renewal cannot remove a separately configured hard runtime ceiling.

3. Decide whether the work is still needed. Do not renew every visible agent's services just to clear attention indicators.

4. If an extension is authorized, replace SERVICE_ID below with the inspected identity:

   fleetctl show "SERVICE_ID"
   fleetctl renew "SERVICE_ID" --lease 30m
   fleetctl show "SERVICE_ID"

Check the new deadline

Read the updated record rather than assuming the requested duration was accepted unchanged. Renewals are capped by configuration. Record the actual expiry if another agent will depend on it.

If the work is no longer needed, an unrenewed controller-launched service is subject to lease cleanup even when its launching agent has disappeared. Use an explicit authorized stop when you need a deliberate immediate shutdown rather than treating lease expiry as a precise scheduling mechanism.

Paused services retain ports and memory. The supervisor leaves paused work alone; expiry cleanup resumes on unpause. An expired claimed record is reported but not signalled, because a claim grants no control authority.

When renewal fails

A missing, stale or rejected record needs a fresh identity review. Do not replace the service with a direct background shell process or remove its limits to evade a refusal. If the controller is unavailable, diagnose it before assuming the old service stopped.

See lease syntax and safe service controls.