Resource utilization measures how much processor capacity a workload consumes. In Kubernetes, observe Pod CPU usage under startup, normal, and high load to choose practical resource requests and avoid waste. section 5.2
Use CPU measurements to guide deployment and scaling decisions:
- Run kubectl top pods and observe usage after startup data becomes available; startup memory provides a lower bound for burstable workloads. section 5.2
- Begin with an intentionally high resource request, apply realistic load, and reduce the request toward observed high-water usage. section 5.2
- Use CPU utilization with a Horizontal Pod Autoscaler to add or remove replicas when average usage crosses a target percentage of requested CPU. section 6.2
Scaling a CPU-bound workload
A Deployment requests 250m CPU per Pod, while an HPA maintains between 1 and 10 replicas at a 20% average CPU-utilization target. Repeated requests to a CPU-intensive endpoint raise utilization and cause Kubernetes to create additional Pods.
