Option C

C is correct A. Configure the horizontal pod autoscaler to use the average response time from the Liveness and Readiness Probes. --> using health check as a trigger of scaling is weird. if the response time of the health check is delayed, it may be caused by resources issues such as CPU, memories, and so on. so you should use such values as SLIs. B. Configure the vertical pod autoscaler in GKE and enable the cluster autoscaler to scale the cluster as pods expand. --> it doesn't referred to pod autoscaling. D. Expose the NGINX stats endpoint and configure the horizontal pod autoscaler to use the request metrics exposed by the NGINX deployment. --> if you use request metrics as SLIs, you should use custom metrics as SLIs. it is a little bit redundant.

option C

C. Install the Stackdriver custom metrics adapter and configure a horizontal pod autoscaler to use the number of requests provided by the GCLB. To scale the deployment of the application's frontend using an appropriate Service Level Indicator (SLI), we need to monitor the traffic coming to the application. One way to do this is to install the Stackdriver custom metrics adapter, which provides visibility into GCLB metrics such as request counts, bytes sent and received, and active connections. We can then configure a horizontal pod autoscaler (HPA) to scale the number of pods based on the request count coming through the GCLB, which will help to ensure that our application is always available to handle the incoming traffic.

C is the answer.

I will go with C

C is the correct answer as per Google documentation

C is correct

Option B is incorrect because there will be no benefit in vertically scaling the front end if the requests are very high. Because in that case the network is the bottleneck and not the instance resources. In my opinion the Answer is C.

c looks more feasible to me. but also B is in small favour. Horizontal pod autoscaler will scale based on the custom metrics such as requests per second(i.e. no. of requests). but vertical autoscaling is also useful feature for the frontend application. as it requires more resource needs if the traffic is high. Since vertical autoscaling will first delete the pod and adjust the cpu and memory to recreate the pod, It can cause downtime for that duration and not recommended. Hence, More inclined towards answer C. Still, reply my answer with your explanation please https://cloud.google.com/kubernetes-engine/docs/concepts/horizontalpodautoscaler#overview

B is correct front-end web applications Scale based on the number of incoming request. so need vertical scaling Back-end Batch Processing (Scale Horizontally) reference https://docs.rightscale.com/faq/What_is_auto-scaling.html

C https://cloud.google.com/kubernetes-engine/docs/tutorials/autoscaling-metrics

According to Google https://cloud.google.com/kubernetes-engine/docs/tutorials/autoscaling-metrics#cpu_1 Horizontal Pod Autoscalers can scale based on CPU utilization natively, so the Custom Metrics Adapter is not needed, therefor C doesn't fit .

C - https://cloud.google.com/kubernetes-engine/docs/tutorials/autoscaling-metrics You want to scale horizontally

Option B