Rebalancing
Redistribute workloads using a reviewed set of migration operations.
Rebalancing proposes migrations that improve capacity distribution while preserving required placement constraints. It operates within a region and does not add capacity or change droplet sizes.
Create a proposal
Open Fleet → Region → Rebalance and select the node pools and objective. Objectives include reducing reservation imbalance, consolidating eligible workloads, and restoring placement after maintenance.
The proposal records source and destination nodes, affected droplets, estimated transfer volume, required headroom, and excluded workloads. It uses configured reservations for admission and observed metrics for supported ranking inputs. A momentary CPU sample is not treated as guaranteed future demand.
Constraints
All proposed destinations must satisfy scheduling, storage access, CPU compatibility, and live migration requirements. Cordoned or unhealthy nodes cannot receive workloads.
Required affinity and spread policies remain active. Local-only volumes and exclusive devices can exclude droplets from live rebalancing. The proposal lists those exclusions instead of silently stopping the workloads.
Review and execute
Set a migration concurrency limit, bandwidth policy, and failure threshold. Review the full resource list, then submit the proposal. Execution revalidates capacity before each move. A stale proposal can lose destinations to unrelated placements and must be refreshed.
Each migration has its own operation ID. Successful moves remain applied when another move fails. The parent result distinguishes completed, skipped, failed, and canceled children.
Automatic schedules
An automatic policy requires explicit scope, a recurring window, movement limits, and an allowed objective. Use a canary pool and bounded movement before enabling it across production capacity.
Policies should avoid frequent movement caused by short-lived load changes. The preview reports the resources selected for the current evaluation. A plan starts only when its preflight and policy conditions are satisfied.
Capacity effects
Migration temporarily requires capacity at both source and destination. A nearly full fleet can be unable to rebalance even when the desired final layout would fit. Add temporary headroom or stop eligible workloads through a separate reviewed procedure.
Consolidation can reduce the number of occupied nodes. It does not retire, power off, or remove billing for those nodes automatically. Follow Node maintenance when taking consolidated hardware out of service.
Cancel or recover
Canceling prevents new migrations and requests cancellation from eligible active children. It does not restore the original layout. Inspect the resulting node reservations before submitting another proposal.