Question

Difficulty: MediumManaging Google Kubernetes Engine Resources

An operations team runs a batch processing pipeline on a Google Kubernetes Engine (GKE) Standard cluster. During workload spikes, newly created Pods remain stuck in a Pending state because current nodes lack unallocated CPU capacity. Additionally, the team needs to reduce infrastructure spending specifically for fault-tolerant batch jobs without risking critical core services. Which TWO configurations should the team implement to address these operational requirements? (Select TWO.)

  1. Enable Cluster Autoscaler on the node pool to automatically provision additional Compute Engine instances when Pods cannot be scheduled due to insufficient resource requests.Answer
  2. B
    Configure a Horizontal Pod Autoscaler (HPA) targeting node CPU utilization to automatically add node instances when existing nodes reach capacity.
  3. Create a separate node pool utilizing Spot VMs for the batch workloads, and apply matching taints and tolerations to isolate fault-tolerant jobs onto these lower-cost nodes.Answer
  4. D
    Migrate the cluster to GKE Autopilot to configure custom host OS kernel parameters while eliminating node pool management overhead.

Answer

The correct operational choices are enabling Cluster Autoscaler on the node pool to provision nodes for unschedulable Pods, and deploying a dedicated Spot VM node pool with taints and tolerations to lower costs for fault-tolerant workloads.
Enabling Cluster Autoscaler directly addresses unschedulable Pending Pods by automatically expanding the node pool when CPU requests exceed available capacity. Additionally, creating a Spot VM node pool with appropriate taints and tolerations isolates fault-tolerant batch workloads onto lower-cost preemptible infrastructure, achieving the cost-reduction goal without affecting core workloads.

Step-by-Step Solution

1
Address the Pod scheduling bottleneck caused by node CPU allocation limits.
Enabling Cluster Autoscaler allows GKE to detect Pods in a Pending state due to resource deficits and automatically scale up node pool capacity.
Horizontal Pod Autoscaler (HPA) increases Pod counts, which would worsen node resource starvation rather than solving node capacity constraints.
2
Optimize infrastructure costs for non-critical, fault-tolerant batch workloads.
Provisioning a secondary node pool using Spot VMs reduces compute costs significantly. Applying taints to the Spot node pool and tolerations to the batch Pods prevents critical workloads from landing on preemptible hardware.
Spot VMs are ideal for workloads that can tolerate unexpected interruptions, providing significant discounts compared to standard VM instances.

Key Concept

GKE Cluster Autoscaler vs HPA scaling boundaries and Spot VM node pool isolation
Estimated Time:1m 30s
Rate this question