Google Professional Cloud DevOps Engineer Question 152
Select 3Google Cloud PlatformYour team has observed intermittent downtime in a production application hosted on Google Kubernetes Engine (GKE). Logs show that pods are being evicted due to resource contention, and your monitoring system indicates that the cluster nodes frequently run out of memory. As a Google Professional Cloud DevOps Engineer, what actions should you take to resolve this issue and prevent further downtime?
- A
Configure resource requests and limits for the application pods to ensure fair resource allocation.
- B
Upgrade the GKE cluster nodes to a machine type with higher memory capacity.
- C
Enable Kubernetes Horizontal Pod Autoscaler (HPA) to automatically scale pods based on CPU and memory usage.
- D
Increase the number of nodes in the GKE cluster by manually scaling the node pool.
- E
Migrate the application to a serverless environment such as Cloud Run to avoid managing infrastructure manually.
Show answer and explanation
Correct answers: A, B, D
Explanation
The observed downtime is due to resource contention and memory exhaustion in the GKE cluster. To resolve this, you should take actions that directly address resource allocation and availability. Configuring resource requests and limits ensures fair resource distribution among pods. Upgrading node machine types and increasing the number of nodes in the cluster provide additional resources to handle workload demands. While other options like HPA and serverless migration might provide benefits in different contexts, they are not the most direct solutions to the specific issue described.
- A. Correct.
Correct. Configuring resource requests and limits ensures that pods have guaranteed access to the resources they need while preventing resource contention caused by competing workloads.
- B. Correct.
Correct. Upgrading to nodes with higher memory capacity can help prevent resource exhaustion, particularly if the workloads have high memory demands.
- C. Incorrect.
Incorrect. While enabling Horizontal Pod Autoscaler (HPA) can optimize scaling for CPU and memory usage in certain scenarios, it does not directly address the issue of resource contention on individual nodes.
- D. Correct.
Correct. Increasing the number of nodes in the cluster by scaling the node pool provides additional memory and CPU resources, reducing the likelihood of resource contention.
- E. Incorrect.
Incorrect. Migrating to a serverless environment like Cloud Run may resolve some infrastructure management challenges but does not directly address the root cause of resource contention within the GKE cluster.