Quiz · Hard
Scenario questions on high availability, disaster recovery, pricing, IAM, compliance and migration strategies.
Which database setting provides automatic failover to a standby in another AZ?
Read replica
Multi-AZ deployment
Snapshot export
Parameter group
Multi-AZ keeps a synchronous standby for HA. Read replicas are asynchronous copies for scaling reads.
A company must keep a web app running if an entire availability zone fails. Which design is best?
Many large instances in one AZ
One large instance with a backup script
Instances in at least two AZs behind a load balancer
Instances in one AZ with an Elastic IP
Spreading across AZs with a load balancer removes the AZ as a single point of failure; scaling up in one AZ does not.
A batch image-rendering job can be restarted at any time and is not time-critical. Which EC2 purchasing option minimizes cost?
Spot Instances
On-Demand
Dedicated Hosts
Reserved Instances with a 3-year term
Spot offers deep discounts for interruption-tolerant jobs. Reserved or dedicated options cost more for a short flexible job.
A steady production database will run for at least three years. What is the most economical pricing model?
On-Demand
Spot
Pay per request
Reserved capacity / Savings Plan
Committing to 1 or 3 years gives a large discount for predictable baseline workloads.
Which strategy moves an application to the cloud with no code changes (lift and shift)?
Refactor
Repurchase
Rehost
Retire
Rehosting moves as is. Refactoring re-architects and repurchasing replaces it with a SaaS product.
Which disaster recovery strategy keeps a scaled-down but complete copy of the system running?
Backup and restore
Warm standby
Pilot light
Cold site
Warm standby runs a smaller full environment that can be scaled up quickly. A pilot light has only the core components running.
A service has 99.9% availability. About how much downtime per year does that allow?
8.76 hours
52.6 minutes
3.65 days
5.26 minutes
0.001 x 525,600 minutes = 525.6 minutes, which is about 8.76 hours. 99.99% is 52.6 minutes.
Two dependent services each have 99.9% availability and are called in series. What is the approximate combined availability?
99.9%
99.99%
100%
99.8%
Availabilities multiply: 0.999 x 0.999 = 0.998. Serial dependencies reduce availability.
An S3 bucket containing customer files was exposed publicly. What is the most likely root cause?
The provider's data center was hacked
An EC2 instance ran out of memory
A misconfigured bucket policy or ACL allowing public access
DNS failed
Most storage leaks come from customer-side misconfiguration. Block public access settings prevent this class of error.
Which AWS service records API calls to show who deleted a resource?
CloudWatch
CloudTrail
Route 53
Trusted Advisor
CloudTrail logs API activity. CloudWatch collects performance metrics and logs.
An application on EC2 needs to read from S3. What is the most secure way to give it credentials?
Attach an IAM role to the instance
Hard-code access keys in the code
Email keys to the developer
Use the root account keys
Roles provide temporary credentials automatically; hard-coded keys leak easily.
Which Well-Architected pillar focuses on running workloads at the lowest price point that meets requirements?
Reliability
Performance Efficiency
Operational Excellence
Cost Optimization
Cost Optimization is about avoiding unnecessary spend; Reliability is about recovering from failures.
A company wants to avoid being tied to a single provider. Which combination helps most?
Provider-specific proprietary databases
A single-region deployment
Containers with Kubernetes and Terraform
Hard-coded provider SDK calls everywhere
Open, portable tooling reduces switching cost. Proprietary services increase lock-in.
What should you check first when a cloud bill unexpectedly doubles?
Reboot all servers
Cost explorer by service and tag to find the growing resource
Disable MFA
Increase instance sizes
Break down spend by service, account and tag to find the source (such as data egress, idle volumes or runaway autoscaling).
A function needs to run for 40 minutes processing a large video. Why is Lambda a poor fit?
The maximum execution time is 15 minutes
Lambda cannot access S3
Lambda has no CPU
Lambda only runs Java
For long jobs use containers, batch services or EC2. Lambda is intended for short event-driven tasks.
Which pairing of Kubernetes terms is correct?
Pod: physical server; Node: container
Node: image registry; Pod: load balancer
Deployment: cluster storage; Pod: network
Pod: group of containers; Node: machine that runs pods
Pods run on nodes. A Deployment manages replicas of pods.
Which control would best prevent a leaked access key from being useful to an attacker?
A longer instance name
More CPU on the instance
Least-privilege policies plus MFA and key rotation
A different AMI
Limiting what a key can do and rotating it reduces damage. Over-broad admin keys are the worst case.
A healthcare company wants a cloud region in its own country because law requires patient data to stay there. Which concept is this?
Elasticity
Data residency / sovereignty
Rightsizing
Vendor lock-in
Regulations may require data to remain in specific jurisdictions.
Which is a benefit of using a CDN for a global website?
Lower latency by serving cached content from edge locations
Removes the need for DNS
Eliminates the origin server
Doubles database capacity
CDNs cache content near users and shield the origin from load, but the origin still exists for uncached content.