Cloud Engineer Case Studies & Practical Application 3 — Questions and Answers
Question 1: A gaming company's leaderboard API on GKE handles 10k RPS normally but fails during tournament events with 100k RPS. The team already uses horizontal pod autoscaling. What additional change addresses the traffic spike?
- Pre-scale the deployment before the event and configure cluster autoscaler with surge nodes (Correct answer)
- Switch from GKE Standard to GKE Autopilot
- Increase pod CPU limits to reduce throttling
- Add a Redis cache in front of the leaderboard database
Correct answer: Pre-scale the deployment before the event and configure cluster autoscaler with surge nodes
HPA reacts to existing load; pre-scaling ensures capacity is ready before the event, and cluster autoscaler surge nodes prevent node provisioning delays from limiting pod scheduling.
Question 2: An enterprise runs Terraform to manage 500 AWS resources. A junior engineer accidentally runs `terraform destroy` on a production workspace. Which preventive control would have stopped this?
- Using Terraform Cloud with Sentinel policies that block destroy plans in the production workspace (Correct answer)
- Storing the Terraform state in an S3 bucket with versioning enabled
- Using Terraform modules for all resources
- Requiring peer review on all Terraform pull requests
Correct answer: Using Terraform Cloud with Sentinel policies that block destroy plans in the production workspace
Sentinel policies in Terraform Cloud can enforce rules like blocking destroy operations in designated workspaces before a plan is applied.
Question 3: A company's Azure AKS cluster costs are 60% higher than budgeted. Analysis shows average node CPU utilization at 15% and memory at 20%. What is the most effective cost reduction strategy?
- Right-size pods with VPA recommendations and enable cluster autoscaler to scale down idle nodes (Correct answer)
- Purchase 3-year Azure Reserved VM Instances for current node count
- Move the AKS cluster to a lower-cost Azure region
- Switch from Azure CNI to kubenet networking
Correct answer: Right-size pods with VPA recommendations and enable cluster autoscaler to scale down idle nodes
Low utilization means pods are over-provisioned; VPA right-sizing packs more pods per node, and cluster autoscaler removes underutilized nodes, directly reducing the node count.
Question 4: A data engineering team ingests 1TB of JSON logs daily into AWS S3, then queries them with Athena. Queries scan the full 1TB each time, costing $5/query. What single change reduces query cost most?
- Convert logs to Parquet format with Snappy compression and partition by date (Correct answer)
- Move logs from S3 Standard to S3 Intelligent-Tiering
- Enable S3 Select on the bucket
- Increase Athena query concurrency
Correct answer: Convert logs to Parquet format with Snappy compression and partition by date
Parquet's columnar format allows Athena to scan only required columns, and date partitioning prunes irrelevant partitions; together they can reduce data scanned by 95%+.
Question 5: A SaaS platform on GCP needs to send transactional emails but Gmail flags them as spam. SPF is configured. What additional DNS/mail configurations should the engineer verify?
- Configure DKIM signing and a DMARC policy for the sending domain (Correct answer)
- Switch to sending emails via Cloud Functions instead of Cloud Run
- Add the mail server IP to Google's Safe Browsing list
- Enable VPC Service Controls on the mail service
Correct answer: Configure DKIM signing and a DMARC policy for the sending domain
DKIM cryptographically signs emails so receiving servers can verify authenticity, and DMARC tells receivers how to handle failures; together they significantly reduce spam classification.
Question 6: A multi-region AWS deployment uses Route 53 latency-based routing. Users in Europe report that they are sometimes routed to us-east-1 instead of eu-west-1. What is the most likely cause?
- The eu-west-1 health check is failing, causing Route 53 to failover to us-east-1 (Correct answer)
- Latency-based routing does not support European regions
- Route 53 TTL is too low, causing frequent resolver changes
- The CloudFront distribution is overriding Route 53 decisions
Correct answer: The eu-west-1 health check is failing, causing Route 53 to failover to us-east-1
Route 53 latency routing respects health checks; if the eu-west-1 endpoint health check fails, Route 53 routes users to the next healthy region regardless of latency.
Question 7: A company migrates on-premises VMs to Azure using Azure Migrate. Post-migration, the application team reports that internal service discovery no longer works. What is the most likely root cause?
- The migrated VMs use hardcoded IP addresses that differ in the Azure VNet (Correct answer)
- Azure Migrate does not support Windows Server workloads
- Azure NSGs block all internal traffic by default
- The VMs were migrated to the wrong subscription
Correct answer: The migrated VMs use hardcoded IP addresses that differ in the Azure VNet
On-premises services often use static IPs for internal communication; after migration to Azure, VMs receive new private IPs, breaking hardcoded references unless DNS or configuration is updated.
Cloud Engineer Case Studies & Practical Application 3
A gaming company's leaderboard API on GKE handles 10k RPS normally but fails during tournament events with 100k RPS.
The team already uses horizontal pod autoscaling.
What additional change addresses the traffic spike?