AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 57
Question 560
Global Accelerator sends users to both US and EU backend ALBs. If the US Region fails, all US user traffic may shift to the EU region. The EU ALB's targets are sized for only its normal users. What failure readiness issue exists?
- Assume the accelerator automatically multiplies EU EC2 and relational database capacity without service quotas or scaling configuration.
- Reserve or rapidly provision enough healthy secondary-region application and database capacity to absorb redirected traffic.
- Configure the EU ALB to respond 200 to all health checks while dropping excess business requests.
- Disable all autoscaling in the EU Region because static capacity always grows automatically after DNS failure.
Explanation
Correct Answer
B:Traffic steering can move connections but cannot create backend throughput or database headroom in a Region that is undersized for failover.
Incorrect Answers
A:Global Accelerator routes traffic and does not bypass downstream provisioning or database limits.
C:False health makes routing decisions misleading and does not expand usable application capacity.
D:Without capacity planning, a surviving Region can saturate during a global traffic shift.
Question 561
An accelerator endpoint is healthy at the TCP listener but the application fails payment requests because its regional database has crashed. The team expects Global Accelerator to repair SQL state by itself. What should be investigated?
- Review endpoint health criteria and the resilience of the underlying database, application and regional dependencies.
- Move the accelerator's global IP address to a different static value to repair the original Region's database files.
- Disable all endpoint health reporting; absent health data makes payment requests complete successfully.
- Enable an additional accelerator listener and assume every failed database transaction is committed automatically.
Explanation
Correct Answer
A:A healthy network listener is not proof of a usable business path. Acceleration cannot restore a failed database or rewrite application transaction state.
Incorrect Answers
B:Changing client ingress IPs is unrelated to recovering failed database storage or SQL engines.
C:Hiding failure signals does not fix dependencies and may route more traffic into an unavailable payment service.
D:Network listeners do not execute SQL recovery or supply a missing authoritative database writer.
Question 562
An Application Load Balancer target group has several registered application instances across two enabled Availability Zones. During a shared dependency failure, every target fails its application health check. Operators expect the ALB to block all forwarding until one target recovers. Which behavior must the incident response plan account for?
- The ALB always returns a load-balancer error without contacting any target whenever all registered targets fail health checks.
- When all registered targets in the group are unhealthy in the enabled Availability Zones, the ALB can fail open and route requests to those unhealthy registered targets.
- The ALB automatically launches replacement EC2 instances in a new AWS Region even if no Auto Scaling or disaster recovery configuration exists.
- Every health-check failure instructs Route 53 to permanently delete all DNS records for the application across all Regions.
Explanation
Correct Answer
B:AWS documents that when a target group contains only unhealthy registered targets, the ALB routes requests to those targets despite their status. Recovery planning must restore dependency health rather than assume health checks form an absolute stop.
Incorrect Answers
A:ALB generally selects healthy registered targets when they exist; however, all-target-unhealthy conditions have a documented fail-open exception. An unconditional no-forwarding assumption is unsafe.
C:An ALB does not independently create a cross-Region recovery fleet. Capacity replacement and regional disaster recovery require separately configured infrastructure and controls.
D:ALB target health checks do not automatically delete all public DNS records. DNS-based health routing is a separate service behavior, and blanket permanent deletion is not the ALB fail-open mechanism.
Question 563
An ALB health check marks instances unhealthy during brief ten-millisecond CPU hiccups. Another proposal probes every hour and waits all day before recovery. How should health thresholds and intervals be selected?
- Set all thresholds to zero because that guarantees instantaneous perfect diagnosis with no false alarms.
- Disable health checking entirely because no checks provide the fastest and safest failover for customers.
- Tune supported interval, timeout and healthy/unhealthy thresholds against measured startup, transient-error and recovery behavior.
- Set one-hour check intervals for a service with a seconds-level recovery objective and expect immediate detection.
Explanation
Correct Answer
C:Overly aggressive checks create false removals; excessively slow checks delay failure detection. The settings should match service behavior and SLA.
Incorrect Answers
A:Health checks require supported nonzero timing and cannot mathematically eliminate both false positives and delay.
B:Without checks, a load balancer cannot reliably distinguish application failure from a valid target.
D:Slow periodic sampling conflicts directly with a short fault detection and recovery requirement.
Question 564
A public ALB serves a critical customer API, but its target group contains only one EC2 instance in one AZ. Managers call the application highly available because the ALB is managed. What is the missing architectural redundancy?
- Create a second listener pointing to the same single target and treat that as failure-domain redundancy.
- Turn off ALB health checks so the only target remains registered while its host is unavailable.
- Increase the one target's EBS size and assume the ALB creates replacement app servers in another AZ.
- Deploy multiple healthy target instances across at least two eligible AZs with suitable Auto Scaling.
Explanation
Correct Answer
D:A resilient load balancer cannot keep serving a business workload if its only backend target is unavailable during a zonal event.
Incorrect Answers
A:Two listener rules still depend on the exact same failed backend application instance.
B:Keeping a broken target registered does not restore responses or reduce the single-instance fault domain.
C:Block-storage sizing cannot add healthy backend compute across independent zones.
Question 565
After a deployment, an ALB reports spikes of HTTP 502/503 from one application service while target health remains green. The incident commander assumes every 5xx proves a network security-group block. What diagnostic distinction matters?
- Correlate ALB HTTP status and target-response metrics with network connectivity and application logs; 5xx need not mean a network reachability failure.
- Remove all healthy targets, because having zero targets guarantees no future client-facing 5xx errors.
- Treat all 5xx responses as proof that the target security group has denied every incoming TCP packet.
- Ignore target application logs and adjust only S3 lifecycle rules as the primary network incident fix.
Explanation
Correct Answer
A:An application can be reachable yet fail internally or return gateway errors. HTTP telemetry and connectivity checks answer different failure questions.
Incorrect Answers
B:With no serving targets the ALB cannot satisfy users and will continue returning errors.
C:A reachable target may generate HTTP errors for application, dependency or protocol reasons without a packet-level deny.
D:Object lifecycle settings do not explain HTTP responses from a live service target.
Question 566
A fleet of restartable compute workers runs as Spot Instances in an EC2 Auto Scaling group across several capacity pools. Worker shutdown needs a brief drain period. Waiting for a two-minute interruption notice often does not leave enough time to start replacements and safely transfer jobs. Which supported mechanism can attempt replacement before an at-risk instance is interrupted?
- Use an Auto Scaling health-check grace period to predict Spot interruptions and launch a new instance while the existing node is still healthy.
- Switch the group to a single lowest-price Spot pool so Auto Scaling always finds interruption-free replacement capacity.
- Disable EC2 rebalance recommendation events and rely exclusively on the regular post-termination health-check replacement path.
- Enable EC2 Auto Scaling Capacity Rebalancing so rebalance recommendations can trigger proactive replacement attempts, together with diverse capacity pools and interruption-tolerant drain logic.
Explanation
Correct Answer
D:Capacity Rebalancing can launch a replacement when Spot sends a rebalance recommendation, allowing earlier workload drain. It remains best-effort: recommendation timing, market availability and graceful shutdown duration still matter.
Incorrect Answers
A:A health-check grace period defers some health evaluations during startup; it does not consume Spot rebalance recommendations or initiate proactive at-risk capacity replacement.
B:A single lowest-price Spot pool can concentrate interruption risk. New Spot capacity is never guaranteed interruption-free and should use suitable diversified allocation choices.
C:Disabling recommendations forfeits the early warning. Post-interruption health replacement may restore desired count but cannot provide the requested proactive lead time.
Question 567
Following an AZ recovery, replacement EC2 instances enter service but need about seven minutes to warm application caches and pass ALB health checks. The Auto Scaling group uses a 60-second health-check grace period, so new instances are replaced repeatedly before they can become healthy. What setting should be corrected?
- Disable the group’s health checks permanently so even crashed nodes are retained.
- Set an appropriate Auto Scaling health-check grace period that covers expected initialization, while maintaining functioning load-balancer and instance health checks.
- Create an unrelated S3 inventory report and assume it changes instance health status.
- Set the grace period to zero seconds to guarantee each new instance receives more warm-up time.
Explanation
Correct Answer
B:The grace period prevents premature ASG replacement when an instance has entered service but is still initializing and not yet passing ELB checks. It should reflect observed startup time rather than hiding sustained faults.
Incorrect Answers
A:Disabling meaningful health evaluation prevents the ASG from replacing genuinely unhealthy instances.
C:Inventory metadata from S3 cannot alter load-balancer health behavior for new instances.
D:Zero seconds provides no health-check grace and can make the premature replacement loop worse.
Question 568
An EC2 Auto Scaling group serves sporadic online auctions. Each newly launched application instance needs twelve minutes to download model files and initialize large in-memory indexes, so unexpected demand spikes exhaust the SLA before fresh instances can serve. The health-check grace period is already long enough to prevent premature replacement. Operations wants preinitialized capacity without paying for a second continuously running fleet. Which design most directly shortens scale-out readiness?
- Shorten the health-check grace period to zero so every newly launched instance is counted as healthy without waiting for application startup.
- Increase the load balancer deregistration delay so new instances start serving before initialization finishes.
- Configure an EC2 Auto Scaling warm pool of initialized stopped or hibernated instances, and use launch lifecycle hooks to finish bootstrap before a warm instance enters service.
- Increase the Auto Scaling group minimum to the maximum possible traffic capacity and keep all instances running at full price permanently.
Explanation
Correct Answer
C:An EC2 Auto Scaling warm pool can retain preinitialized instances stopped or hibernated for faster reuse. Lifecycle hooks guard long user-data/bootstrap work before instances are stopped and before they are admitted into service.
Incorrect Answers
A:A grace period changes when Auto Scaling evaluates instance health; it cannot eliminate twelve minutes of bootstrap work. Shortening it may reintroduce unhealthy-instance churn.
B:Deregistration delay controls draining existing targets during removal, not the readiness time of a newly launched application instance.
D:A constantly running maximum-size fleet could hide cold starts but contradicts the requirement to avoid paying for large unused EC2 capacity and does not use the available warm-pool optimization.
Question 569
A two-AZ Auto Scaling group successfully replaces servers after failure, but a user loses an in-progress shopping cart held only in the failed process memory and a long upload is interrupted. What assumption should be corrected?
- Assume replacement instances inherit all previous RAM and unfinished TCP connections when a failing host is terminated.
- Auto Scaling replacement does not preserve instance-local state or every in-flight request; use shared persistence and graceful retry/drain patterns.
- Increase the Auto Scaling maximum size and assume this copies uncommitted carts from failed nodes.
- Disable persistent database storage because it delays instance replacement and compute recovery.
Explanation
Correct Answer
B:New healthy compute can restore capacity, but stateful operations need independent data persistence and interruption-safe client or worker behavior.
Incorrect Answers
A:New EC2 instances start new processes and cannot transparently recover arbitrary failed-host memory and sockets.
C:Desired and maximum counts control capacity, not application transaction-state replication.
D:Removing durable state would worsen business data loss and does not protect in-progress operations.
Question 570
A nonprofit needs disaster recovery for an application with a four-hour RTO and limited monthly budget. It can afford a small always-running database and minimal core services in a secondary Region, scaling the full web tier only after a disaster. Which DR pattern is appropriate?
- Use a pilot-light design with core replicated services and tested capacity scale-up during activation.
- Keep no secondary data or infrastructure configuration, and assume a blank account can meet four-hour recovery automatically.
- Use only DNS failover to a Region containing no database, application images or compute capacity plan.
- Run an equal full-scale active-active stack in both Regions, despite the stated need to minimize recurring cost.
Explanation
Correct Answer
A:Pilot light keeps a minimal secondary foundation active, trading longer recovery activation for lower steady-state compute cost than warm full stacks.
Incorrect Answers
B:Without core data and tested deployment automation, restoring a functional service may exceed the permitted recovery window.
C:DNS routing cannot supply the absent core dependencies needed to serve customers after failover.
D:Full active-active capacity is generally costlier and provides stronger availability than the budget-constrained scenario requires.
