Question-26: You are employed by a firm that offers a Television Over-the-Top (OOT) application and hosts millions of short films similar to tiny TV. In addition, your employer utilizes Google Kubernetes Engine (GKE) as a platform for all of its workloads. Workloads that are batch, stateful, and stateless are all distributed over a single huge GKE cluster that your firm uses. A single node pool of 200 nodes is used as the configuration for the GKE cluster. The costs associated with maintaining this cluster are too high for your firm, yet you do not want to sacrifice availability. What is it that you ought to do?
A. Build a second GKE cluster, and dedicate it solely to handling batch workloads. Distribute the original two hundred nodes evenly between the two clusters.
B. Establish boundaries for the amount of processing power and memory used by each namespace in the cluster. Set a CPU and memory limit on each and every Pod that you configure.
C. Configure a HorizontalPodAutoscaler for all stateless workloads and for all compatible stateful workloads. Configure the cluster to use node auto scaling.
D. Make the node pool use preemptible virtual machines (VMs).
Correct Answer
Get All 340 Questions and Answer for Google Professional Cloud Architect
: 3 Explanation: The third choice, which does not need significant changes to the existing Kubernetes setup, is the one that should be chosen. Option 1 is not required since you may utilize a variety of alternative node pools each with their own set of settings. Option 2 optimises resource consumption by reducing the amount of CPU and memory that is used by your current node pool. However, this option does not inevitably reduce costs and should still be evaluated. Option 3: This seems to be a pretty solid option. Autoscaling both the workloads and the node pools inside your infrastructure makes the whole system more elastic and provides you with the option to continue relying on the same node pool. Option 4: This may not be the best choice depending on the kind of task you have to do. Stateful workloads are not well-suited for operation on preemptible VMs; however, batch and stateless workloads can typically manage this rather well. Batch and stateless workloads are not well-suited for operation on preemptible VMs. Because there is only one correct response that will be approved, and because the business does not want to jeopardize the availability of the application, the corporation has decided that HPA is the best solution for autoscaling pods. It is necessary for the nodes that make up the GKE cluster to be autoscaled. This is done with cost efficiency in mind. As a result, the choice that should be made is option C, which is to configure a HorizontalPodAutoscaler for all stateless workloads as well as any stateful workloads that are compatible. Make sure that the cluster is set up to utilize auto scaling for the nodes.