Amazon AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 34

Question 330

A nightly workflow must apply the same transformation to 150 independent S3 files, wait for results, and control maximum parallel executions. Which Step Functions feature fits?

Hard-code 150 sequential Lambda calls without failure branch information.
Send 150 independent Lambda invocations directly and infer final completion only from the count of API invocation acceptances without tracking each result.
Run an S3 bucket replication rule and assume it executes the custom transform code.
Use a Map state with supported concurrency and failure controls for the independent items.

Explanation

Correct Answer

A Map state fans out similar work with bounded concurrency and tracks each item’s results, instead of wiring 150 bespoke branches.

Incorrect Answers

This unnecessarily serializes independent tasks and increases orchestration maintenance.

Invocation acceptance is not a coordinated run receipt for per-file outcomes and concurrency. A Step Functions Map state supplies bounded fan-out and workflow item results.

Replication copies supported object versions, not arbitrary application transforms.

Question 331

When an encrypted receipt is saved into S3, a scanner should run asynchronously, but the upload API should return without waiting for image analysis. Which integration is suitable?

Make the browser invoke scanner and object upload synchronously and report failure if scanner is slow.
Use an S3 storage-class transition rule as the only scanner trigger.
Configure a supported S3 ObjectCreated notification to SQS/EventBridge/Lambda, with relevant filters and idempotent handling.
Use an S3 bucket policy to grant PutObject to the scanner and assume the policy itself asynchronously invokes the scanner on each upload.

Explanation

Correct Answer

Object notifications allow storage success and independent downstream processing; event delivery can be duplicated and filters matter.

Incorrect Answers

The service is then coupled to scanner availability.

Lifecycle transition does not invoke arbitrary analysis functions.

S3 policies authorize storage API operations; they do not emit processing work. S3 events integrated with SQS/EventBridge/Lambda decouple upload from scanning.

Question 332

The company runs thousands of independent nightly Monte Carlo simulations in containers. They may wait in a queue and require elastic compute rather than interactive web serving. Which AWS service should orchestrate jobs?

Use AWS Batch job queues and compute environments with suitable resource requirements and retry policy.
Use an ALB listener rule as the only job scheduler.
Reserve one EC2 web application instance for all simulations and never queue work.
Use an EC2 Auto Scaling group of always-running HTTP servers and keep an open request for every nightly simulation instead of job queuing.

Explanation

Correct Answer

Batch schedules queued container jobs onto managed capacity and allows work to scale independently of user requests.

Incorrect Answers

ALB routes requests to live targets, not batch container jobs.

One synchronous host cannot efficiently handle large independent overnight jobs.

ASG capacity alone cannot coordinate independent queued container jobs, retries and resource requests. AWS Batch provides a job scheduling boundary for the batch workload.

Question 333

A producer needs to submit a 5 MB PDF for asynchronous processing. Its event metadata is small, but SQS message payload size is limited. Which handoff avoids attempting to inline a large document?

Store the PDF in S3 and place a durable object reference plus integrity/version metadata on the SQS message.
Base64 encode the 5 MB PDF directly into a standard SQS message and assume encoding removes size limits.
Divide one PDF into arbitrary unrelated SQS messages with no manifest or reconstruction contract.
Increase the SQS queue visibility timeout and assume that timing change permits a single oversized 5-MB message payload.

Explanation

Correct Answer

Object payload stays in S3; the queue carries a bounded small pointer that workers resolve under authorization.

Incorrect Answers

Encoding increases payload size rather than bypassing queue limits.

Without an explicit assembly protocol workers cannot recover the original object reliably.

Visibility controls how long received messages are hidden while workers process them; it cannot raise the maximum message size. S3 payload plus an SQS reference avoids the limit.

Question 334

A billing consumer may receive the same paid-order event twice because the queue uses at-least-once delivery. Which transactional design prevents duplicate charges?

Rely on the queue visibility timeout as proof the event can never recur.
Use a business-operation idempotency key and check/commit payment state atomically at the authoritative ledger boundary.
Create a fresh payment ID for every retry of one order.
Treat SQS message deletion after handler receipt as proof the payment provider never committed the same business transaction earlier.

Explanation

Correct Answer

An event delivery can repeat; deduping the business transaction, not merely the message receipt, protects customer funds.

Incorrect Answers

At-least-once semantics allow duplicates despite appropriate visibility.

That can turn retries into separately chargeable operations.

Queue acknowledgement and payment settlement are separate operations. A durable idempotency key checked atomically at the payment ledger prevents duplicate charges across retries.

Question 335

An image processor depends on a third-party service that is temporarily unavailable. New uploads must still be accepted and existing jobs retried without the upload producer becoming unavailable. Which design is preferable?

Retry the external service indefinitely from the user’s open upload request.
Silently discard jobs when the external provider returns 503.
Queue work durably and have the independent consumer use bounded backoff, idempotency and DLQ recovery.
Restart every uploader whenever the downstream worker receives a timeout.

Explanation

Correct Answer

The ingestion path stays independent, while consumer retry policy addresses remote outages without losing accepted work.

Incorrect Answers

That couples acceptance to an unavailable consumer and creates unbounded latency.

Data loss violates durable acceptance.

Producer restart does not recover queued independent processing.

Question 336

A large upload worker spends most time waiting on slow external storage, so CPU-based scaling remains below threshold. Which trigger best addresses latency and queued-job SLO?

Scale only on average CPU, leaving the known I/O delay invisible.
Increase the worker count only after CPU target tracking exceeds 80% while leaving the known backlog and oldest-message age unmonitored.
Scale consumers using backlog per active worker or age-of-oldest-message relative to target processing time.
Increase task logging verbosity as the only capacity change.

Explanation

Correct Answer

Queue demand is directly measured even when per-worker CPU remains low.

Incorrect Answers

CPU is a poor signal for this bottleneck.

I/O-bound queue consumers can fall behind while CPU remains low. Scaling needs backlog-per-worker or message age signals to address the actual capacity demand.

Logging may diagnose but does not process queued jobs.

Question 337

An ECS service moves among VPC subnets and tasks have ephemeral addresses. Clients must use a stable logical name without adding a separate self-managed mesh. Which approach fits the existing ECS operations?

Use ECS Service Connect namespace and client aliases for discovery across supported ECS services.
Embed each new container IP in all downstream AMIs whenever a task is replaced.
Use Cloud Map registrations but configure ECS clients with retired literal task IP addresses rather than using the intended Service Connect namespace/client aliases.
Use a different root IAM password as the discovery mechanism for every task.

Explanation

Correct Answer

Logical endpoints remain stable as service tasks and addresses change, with ECS-managed connectivity.

Incorrect Answers

Frequent redeployments preserve brittle coupling.

External service discovery helps only when callers use its logical names. A task-IP literal bypasses Service Connect and breaks when ECS replaces tasks.

Identity credentials are not endpoint routing.

Question 338

A suite includes ECS, EC2, and custom applications that all require registry-based service discovery. Teams do not want each client hardcoding addresses. Which general AWS discovery abstraction fits across supported platforms?

Require ECS Service Connect without assessing non-ECS workloads that need discovery.
Use AWS Cloud Map service namespaces and registered healthy instances that clients resolve.
Create a static S3 object list and require every client to parse it at boot forever.
Create separate static Route 53 A records for every ephemeral ECS/EC2 task and leave them unmaintained after instance replacement.

Explanation

Correct Answer

Cloud Map provides discovery metadata beyond one ECS-specific client mesh integration.

Incorrect Answers

Service Connect is focused on supported ECS services and does not automatically cover arbitrary applications.

A static artifact becomes stale as instances change.

Static records are prone to staleness. Cloud Map service discovery supplies updated healthy registrations suitable for heterogeneous application clients.

Question 339

An ALB-backed stateless tier is planned for horizontal scaling. Some developers propose sticky sessions to preserve cart state solely in the process. Which long-term scaling approach is safer?

Persist cart/session state in a secured shared store or supported stateless token so routing affinity is optional.
Depend on indefinite ALB stickiness and retain the cart only in volatile process RAM.
Disable health checks to avoid switching users off a failed target.
Increase Route 53 TTL until browsers always use one EC2 address.

Explanation

Correct Answer

Independence from one host prevents cart loss during health-based failover and lets instances scale freely.

Incorrect Answers

Load balancer affinity cannot recover data if that host dies.

Failed targets need removal; suppressing health checks sacrifices availability.

DNS caching does not preserve application session data.

Question 340

After changing a synchronous image API to an SQS-backed pipeline, which observation best confirms that HTTP requests can succeed while workers are temporarily stopped?

The uploader returns success only after the image thumbnail finishes rendering.
A test upload is accepted and placed on the target queue, then processes successfully after workers resume without resubmitting the HTTP request.
The S3 bucket has the intended name but no message or output appears.
Worker CPU falls when workers are stopped and the upload endpoint also fails immediately.

Explanation

Correct Answer

The experiment directly demonstrates producer/consumer decoupling and durable pending work.

Incorrect Answers

That remains synchronous dependence rather than proof of queue buffering.

Resource naming alone does not show accepted work survives a consumer outage.

That indicates the failure remains coupled to request acceptance.

Pass Your Amazon AWS Certified Solutions Architect - Associate SAA-C03 Exam Easy!

Question 330

Explanation

Correct Answer

Incorrect Answers

Question 331

Explanation

Correct Answer

Incorrect Answers

Question 332

Explanation

Correct Answer

Incorrect Answers

Question 333

Explanation

Correct Answer

Incorrect Answers

Question 334

Explanation

Correct Answer

Incorrect Answers

Question 335

Explanation

Correct Answer

Incorrect Answers

Question 336

Explanation

Correct Answer

Incorrect Answers

Question 337

Explanation

Correct Answer

Incorrect Answers

Question 338

Explanation

Correct Answer

Incorrect Answers

Question 339

Explanation

Correct Answer

Incorrect Answers

Question 340

Explanation

Correct Answer

Incorrect Answers

AWS Certified Solutions Architect - Associate SAA-C03 Practice Tests