QuestionQ93

VerifiedOptimizing performance and cost

You are conducting a semi-annual capacity-planning exercise for your flagship service. You anticipate service-user growth of 10% month over month during the next six months. Your fully containerized service runs on Google Cloud Platform (GCP) in a Google Kubernetes Engine (GKE) Standard regional cluster spanning three zones, with cluster autoscaler enabled. You currently use approximately 30% of your total deployed CPU capacity and require resilience to the loss of a zone. You want users to experience minimal negative impact from this growth or from a zone failure, while avoiding unnecessary costs. How should you prepare for the predicted growth?

Explanation

GKE Cluster Autoscaler scales node pools only within their configured minimum and maximum limits, and it adds capacity when Pods cannot be scheduled. Horizontal Pod Autoscaler adjusts workload replica counts in response to CPU, memory, or other supported metrics, allowing application capacity to grow with demand. Confirming the node-pool maximum and validating the expected load ensures that enough capacity remains available when demand rises and when one of the three zones fails, without pre-provisioning unnecessary nodes.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!