Create a drift schedule#

  1. Open Delivery Settings and the drift-schedule controls.
  2. Name the schedule and choose the project.
  3. Enter the cron expression and explicit time zone.
  4. Choose observe, plan or reconcile mode. Start with observe when validating a new schedule.
  5. Set jitter, concurrency and enablement, then save.
  6. Check the next evaluation and last-result information on the project’s Drift view.
A drift schedule combines a project, cron expression, time zone and remediation mode. New schedules remain inactive until enabled.
A drift schedule combines a project, cron expression, time zone and remediation mode. New schedules remain inactive until enabled. View full size ↗

Choose the mode deliberately#

ModeIntent
observeObserve divergence without assuming a corrective deployment is authorized.
planProduce proposed-change evidence for review.
reconcileRequest corrective delivery through the normal safety and policy gates.

Reconcile can lead to actual cluster changes. It still depends on the project, target, approval, policy and change-window rules. A drift badge with an old observation time should be refreshed before acting.

Configure a change window#

  1. Open change-window management in Delivery Settings.
  2. Choose the environment and, optionally, a single project.
  3. Select maintenance or freeze, then enter the time zone, days, start/end times and a reason.
  4. Save and verify the resulting window against the intended local time.
  5. Check a blocked run’s reason before changing a window; a freeze may be working exactly as intended.
Use a change window to define the affected environment/project, time zone and permitted or frozen interval.
Use a change window to define the affected environment/project, time zone and permitted or frozen interval. View full size ↗

Pause or change automation#

Disable the specific schedule when you need to stop future evaluations. Review project polling/automation separately; disabling a drift schedule is not the same as disabling source polling. Existing runs need their own state-aware cancellation/recovery decisions.