Replica capacity is the number of requests per second one workload replica can serve. Measure it under a known load, then use it to estimate how many replicas current traffic requires. section 6.2.1
Use replica capacity to:
- Estimate replicas with replica_count = RPS ÷ replica_capacity. section 6.2.1
- Plan manually by measuring tested capacity, projecting traffic, and increasing replicas accordingly. section 6.2.1
- Drive automatic scaling with a Kubernetes Horizontal Pod Autoscaler using an external RPS metric. section 6.2.1
Scaling an HTTP workload
Expose the workload through a NodePort Service and Ingress, generate traffic with ab, and observe the HPA add replicas as request load rises.
