Examkin

Soru

Zorluk: ZorPlanning Google Kubernetes Engine (GKE) Cluster Architectures

A digital health platform is designing a Google Kubernetes Engine (GKE) architecture to run two workloads with contrasting operational requirements:

• Workload 1: A legacy telemetry engine that requires custom Linux kernel (sysctl) parameters for network socket optimization.
• Workload 2: A batch data analysis service that is stateless, fault-tolerant, and requires rapid scaling at minimal cost.

The engineering team wants to minimize infrastructure management where possible while fully supporting both workloads. Which cluster architecture and node pool strategy should the cloud engineer select?

  1. Deploy a GKE Standard cluster; configure a custom node pool with tuned sysctl parameters for Workload 1, and a secondary node pool using Spot VMs with Cluster Autoscaler for Workload 2.Cevap
  2. B
    Deploy a GKE Autopilot cluster; provision Workload 1 on standard Autopilot pods and Workload 2 on Spot pods to eliminate node management overhead.
  3. C
    Deploy a GKE Standard cluster with a single Spot VM node pool for both workloads, utilizing Horizontal Pod Autoscaler (HPA) to scale node capacity as demand increases.
  4. D
    Deploy a GKE Autopilot cluster using Spot instances for Workload 1, and configure a GKE Standard node pool within the same cluster for Workload 2.

Cevap

Deploying a GKE Standard cluster with a customized node pool for kernel parameter tuning and a Spot VM node pool with Cluster Autoscaler for batch workloads satisfies all requirements.
GKE Standard must be selected because Workload 1 requires low-level kernel parameter tuning (sysctl), which is not permitted in GKE Autopilot mode. Provisioning a dedicated node pool for Workload 1 accommodates these kernel settings, while a separate Spot VM node pool with Cluster Autoscaler enabled optimizes costs and handles dynamic scaling for Workload 2's stateless, fault-tolerant batch process.

Adım Adım Çözüm

1
Evaluate the kernel customization requirements of Workload 1.
GKE Autopilot restricts node-level access and kernel modifications (such as sysctl settings). Therefore, a GKE Standard cluster is mandatory.
Autopilot automates node management by locking down node configuration, whereas Standard permits custom node pool settings.
2
Determine node pool configuration for Workload 1.
Configure a dedicated GKE Standard node pool with the necessary sysctl parameters applied during node provisioning.
Isolating Workload 1 to its own node pool ensures node-level settings do not impact other workloads.
3
Evaluate cost and scaling requirements for Workload 2.
Create a secondary node pool using Spot VMs paired with GKE Cluster Autoscaler.
Spot VMs deliver significantly discounted compute for fault-tolerant, stateless batch workloads, and Cluster Autoscaler handles dynamic node capacity changes.

Anahtar Kavram

Selecting between GKE Autopilot and GKE Standard based on node-level customization needs (sysctl, custom OS/node settings) vs. operational management overhead, alongside node pool specialization (Spot VMs and autoscaling).
Bu soruyu puanla

Konular

Planning Google Kubernetes Engine (GKE) Cluster Architectures

Bu konudan daha fazla soru