Apply Recommendations
Applying a recommendation changes the target Kubernetes workload. EcoScale is designed to make that action deliberate: review the recommendation, confirm the workload, apply the change, and validate the outcome.
What Apply Does
Section titled “What Apply Does”When a recommendation is applied, EcoScale patches the target workload with selected compute values. The action can be scoped to specific containers and resource settings, which lets operators avoid all-or-nothing updates.
This is an operational change, not a reporting action. Treat it with the same discipline as a deployment, especially for production workloads.
Review Before Applying
Section titled “Review Before Applying”| Review area | What to confirm |
|---|---|
| Workload identity | Confirm the cluster, namespace, WorkloadScaler, target workload kind, and target workload name. |
| Container scope | Confirm the container names that will receive updated resources. |
| Resource delta | Compare current and recommended CPU and memory requests and limits. |
| Cost impact | Confirm that the projected savings or cost increase is understood. |
| Reliability risk | Check workload criticality, recent incidents, and sensitivity to memory pressure or CPU throttling. |
If the change affects a critical service, align with the service owner and choose a safe change window. If the recommendation includes a large memory reduction, validate carefully before applying.
Validate After Applying
Section titled “Validate After Applying”The apply workflow is not complete until the workload remains healthy after the change. Watch pod restarts, scheduling behavior, memory pressure, CPU throttling symptoms, latency, and error rates.
Expected savings should also be verified. EcoScale helps estimate cost impact, but operational validation confirms whether the change achieved the intended outcome without degrading performance.
Rollout Strategy
Section titled “Rollout Strategy”Start with low-risk workloads that have clear waste and stable behavior. Apply recommendations manually, monitor the result, and use those successes to define which workloads are ready for automation.
This approach creates a durable optimization program: manual action is used to build trust, and automation is reserved for recommendations that repeatedly prove safe.