Resilience and high availability account for 26% of the exam. These three questions hit the most commonly tested patterns: database failover, compute scaling, and DNS-level routing.
Question 1
An application requires automatic database failover within 2 minutes across Availability Zones with no manual intervention. Which RDS deployment option meets this requirement?
- A) RDS Single-AZ with automated backups
- B) RDS Multi-AZ deployment
- C) RDS Read Replica in a different AZ
- D) RDS with manual snapshot restore
Answer: B: RDS Multi-AZ deployment
RDS Multi-AZ maintains a synchronous standby replica in a different Availability Zone. If the primary fails, AWS automatically promotes the standby and updates the DNS endpoint with no application changes required, no manual steps. Failover typically completes in 60–120 seconds.
Why the distractors are wrong:
- Single-AZ with automated backups: Backups enable point-in-time recovery but not automatic failover. Restoring from backup takes minutes to hours.
- Read Replica: Uses asynchronous replication. Read Replicas must be manually promoted to primary and involve potential data loss (replication lag).
- Manual snapshot restore: Slowest option. Creates an entirely new RDS instance.
Key distinction to memorise:
| Feature | Purpose | Replication | Automatic Failover? |
|---|---|---|---|
| Multi-AZ | High Availability | Synchronous | ✅ Yes |
| Read Replica | Read Scaling / DR | Asynchronous | ❌ No (manual promotion) |
| Automated Backups | Recovery | Point-in-time | ❌ No |
Question 2
An e-commerce site experiences unpredictable traffic spikes during flash sales. Which AWS feature automatically adjusts EC2 capacity to maintain a target CPU utilisation of 60%?
- A) Elastic Load Balancing
- B) EC2 Auto Scaling with Target Tracking policy
- C) AWS Lambda
- D) Amazon CloudFront
Answer: B: EC2 Auto Scaling with Target Tracking policy
Target Tracking automatically calculates and applies the required scaling actions to keep a chosen metric at your target value. Set it to 60% CPU and Auto Scaling adds instances when demand rises and removes instances when demand drops, with no manual scaling rules needed.
Why the distractors are wrong:
- Elastic Load Balancing: Distributes traffic across existing instances. It does not add or remove instances.
- AWS Lambda: Serverless compute that scales independently. Does not manage EC2 fleets.
- Amazon CloudFront: A CDN that caches and delivers content. Does not affect EC2 capacity.
The three Auto Scaling policy types:
| Policy | How It Works | Best For |
|---|---|---|
| Target Tracking | Keeps a metric at a target value | Most use cases (simplest) |
| Step Scaling | Scales in steps based on alarm thresholds | When you need fine-grained control |
| Scheduled | Scales at specific times | Predictable load patterns |
</details>On the exam: "automatically adjusts capacity based on a metric" = Target Tracking. "Scales at a specific time" = Scheduled scaling.
Question 3
A company wants Route 53 to automatically direct users to a secondary website in a different region whenever health checks detect the primary site is unavailable. Which routing policy should they configure?
- A) Latency-based routing
- B) Weighted routing
- C) Failover routing
- D) Geolocation routing
Answer: C: Failover routing
Route 53 Failover routing has a Primary and Secondary record. Route 53 continuously runs health checks against the primary. If the primary fails, DNS automatically returns the secondary record, routing users to the backup site with no manual intervention.
Why the distractors are wrong:
- Latency-based — Sends users to the lowest-latency region. Doesn't respect health: if the low-latency region is down, it may still route there.
- Weighted routing — Splits traffic by percentage (e.g. 90/10 for A/B testing). No health-based switching.
- Geolocation routing — Routes based on the user's country or continent. No automatic failover.
All Route 53 routing policies — quick summary:
| Policy | Use Case |
|---|---|
| Simple | Single resource, no health checks |
| Failover | Active/passive HA |
| Latency | Route to lowest-latency region |
| Weighted | A/B testing, gradual migration |
| Geolocation | Different content per country/continent |
| Geoproximity | Route based on geographic bias (Traffic Flow) |
| Multivalue Answer | Return multiple healthy records |
Key Takeaways
- RDS Multi-AZ = synchronous replication + automatic failover. Read Replicas = asynchronous + must be manually promoted
- "Automatically scale to keep CPU at X%" = EC2 Auto Scaling Target Tracking policy
- "Automatically route away from a failed site" = Route 53 Failover routing