AWS Certified Solutions Architect - Associate SAA-C03 Practice Test 46
Question 450
A DynamoDB Streams-driven Lambda commits a search-index update, then times out before its event batch is acknowledged. Later the same records are retried. How should the search-index update behave?
- Delete the source DynamoDB table after each batch to prevent event retries.
- Treat each retried stream record as a brand-new change requiring another irreversible counter increment.
- Use idempotent index operations and a replay-aware failure strategy for retried stream batches.
- Assume exactly-once search-index commits are guaranteed merely because DynamoDB Streams retains sequence numbers.
Explanation
Correct Answer
C:Stream event-source mappings may retry failed batches; updates should converge correctly even if a record is processed again.
Incorrect Answers
A:Deleting production data is not an event replay control and introduces far more serious data loss.
B:Blind duplicate increments can corrupt derived state when the original effect already committed before timeout.
D:Sequence information supports tracking but does not atomically coordinate the downstream index commit and Lambda batch acknowledgement.
Question 451
A telemetry stream already uses Kinesis Data Streams with correctly selected partition keys. Five separate analytics consumers all call GetRecords and now compete for shared shard read throughput; two must process events with low delay without taking capacity from the others. Which stream-consumer capability addresses that competition?
- Increase the DNS TTL of the stream endpoint so each consumer gets dedicated shard bandwidth.
- Register eligible enhanced fan-out consumers and use SubscribeToShard to obtain dedicated per-consumer read throughput, while evaluating the extra cost.
- Move all consumers to the same polling loop and claim that each gains an independent 2 MB/s shard quota.
- Change only the record retention duration, which assigns a separate read channel to every current consumer.
Explanation
Correct Answer
B:Enhanced fan-out provides dedicated throughput per registered consumer per shard and pushes records instead of requiring GetRecords polling. It reduces consumer competition but introduces consumer-shard and retrieval charges.
Incorrect Answers
A:DNS TTL does not control stream shard read quotas.
C:Shared polling does not create separate throughput allocations for independent consumers.
D:Retention changes how long records can be replayed, not the number of concurrent dedicated read channels.
Question 452
A Kinesis Data Streams application assigns the same partition key to every sensor event. One shard becomes hot and throttles writes although other shards have capacity. What publisher-side change helps distribute traffic?
- Set the stream retention to a longer period; retention automatically redistributes every new partition key across shards.
- Change only record timestamps so the same partition key hashes to different shards every second.
- Continue using one universal partition key and increase unrelated EC2 CPU capacity only.
- Choose partition keys with adequate cardinality and distribution, consistent with the required per-entity ordering scope.
Explanation
Correct Answer
D:Kinesis maps records to shards using partition keys. More balanced keys spread writes while preserving ordering for records that truly belong together.
Incorrect Answers
A:Retention controls how far back readers can replay and does not fix a hot producer partition.
B:Hash routing is based on the partition key, not an arbitrary timestamp payload attribute.
C:Producer compute does not distribute a single hot hash key across Kinesis shards.
Question 453
A Kinesis reader may be down for forty hours and must replay the missed records when it resumes. The stream currently retains data for twenty-four hours. What configuration should the architect review?
- Increase Kinesis stream retention to a supported period that covers the outage and recovery interval.
- Enable an SQS visibility timeout on the stream; Kinesis retained records inherit SQS queue semantics.
- Increase the producer's EC2 Auto Scaling desired count without changing the stream retention setting.
- Use a shorter retention window to speed up reader recovery of records that have already expired.
Explanation
Correct Answer
A:Consumer checkpointing helps resume reads only while original stream records remain retained. An adequate retention window is essential for replay.
Incorrect Answers
B:Kinesis retention/replay is distinct from SQS message invisibility after individual receives.
C:More producer instances do not preserve older records after the configured Kinesis retention period ends.
D:Decreasing retention removes historical data sooner and makes extended outage recovery worse.
Question 454
A Kinesis Data Streams pipeline hits write throughput limits while EC2 producers remain mostly idle. The team assumes adding more EC2 instances will automatically expand stream capacity. What should they assess?
- Add producer EC2 instances without examining Kinesis throttling because source compute automatically increases shard count.
- Review stream capacity mode, shard or on-demand scaling behavior and Kinesis throughput metrics and quotas.
- Increase S3 object storage capacity for an unrelated bucket, which raises Kinesis write capacity in the same Region.
- Turn off all stream monitoring to remove throughput limits observed in CloudWatch graphs.
Explanation
Correct Answer
B:Kinesis stream ingress is bounded by its service capacity configuration and partition distribution, independently of how many producer VMs exist.
Incorrect Answers
A:Higher producer capacity cannot bypass a fixed stream throughput ceiling and may intensify throttling.
C:Object storage capacity and stream record throughput are separate service limits and resources.
D:Suppressing metrics does not change service capacity or the underlying throttled API behavior.
Question 455
A financial processing task must run at 02:00 every weekday. Today a cron process on one EC2 instance triggers it, and jobs are missed when that host is replaced. What managed scheduling choice removes this dependency?
- Use a Lambda function that sleeps continuously for 24 hours, then starts the job.
- Increase the EC2 root volume size; a larger disk guarantees all future cron executions.
- Create an Amazon EventBridge Scheduler schedule targeting the supported job endpoint.
- Keep the single-instance cron and turn off EC2 health replacement to avoid schedule interruption.
Explanation
Correct Answer
C:Scheduler invokes supported AWS targets at defined times without relying on a single long-running cron host, while supporting retries and timing configuration.
Incorrect Answers
A:A Lambda execution cannot serve as an unlimited always-on timer and is ill-suited to long waits.
B:Disk capacity alone cannot protect scheduled invocation against instance failure or restart.
D:Suppressing host replacement introduces availability risk and does not make one operating-system scheduler resilient.
Question 456
An EventBridge Scheduler rule targets a Step Functions state machine in another account. The schedule is created, but invocations fail with AccessDenied because the scheduler has no right to start executions. What is missing?
- Change the state machine's display name and expect the role policy to be generated automatically.
- Embed the root account's long-term access key into the schedule description field.
- Increase Scheduler flexible time window to twelve hours to remove IAM authorization errors.
- Configure a suitable IAM execution role assumed by Scheduler with permission for the target action and any cross-account resource requirements.
Explanation
Correct Answer
D:The Scheduler service uses an invocation role to call permitted targets. A schedule definition does not implicitly grant rights to start state-machine executions.
Incorrect Answers
A:Renaming a target does not grant the scheduler principal events or Step Functions API permissions.
B:Static administrator credentials in schedule metadata are insecure and do not implement the expected invocation-role trust model.
C:A schedule window affects delivery timing, not permission to invoke the protected target.
Question 457
A hospital schedules a data synchronization task at 09:00 local time each day, including daylight-saving transitions. A colleague specifies 09:00 UTC with a flexible twelve-hour execution window. Which Scheduler settings better honor the requirement?
- Hardcode UTC at 09:00 because all local civil times remain identical to UTC across daylight saving.
- Set the flexible window to the maximum solely because that always executes at the exact requested minute.
- Set the intended time-zone-aware schedule and an appropriate execution window for the deadline sensitivity.
- Use an EC2 Auto Scaling cooldown as the replacement for EventBridge calendar and time-zone semantics.
Explanation
Correct Answer
C:Scheduler can express schedules in supported local time zones. A flexible window must be chosen to respect the actual business deadline rather than arbitrarily delaying execution.
Incorrect Answers
A:Time zone offsets may change seasonally, so fixed UTC and local 09:00 are not generally equivalent.
B:A flexible window permits service execution over an interval rather than guaranteeing an exact instantaneous run.
D:Auto Scaling cooldown governs capacity changes and does not define a daily local-time job schedule.
Question 458
EventBridge Scheduler invokes a payment reconciliation job during a transient service outage. The job must not vanish without investigation after its allowed retries end. Which failure policy should be included?
- Configure a supported retry policy and dead-letter queue for failed schedule target deliveries.
- Make every failed target return HTTP 200 regardless of whether reconciliation executed correctly.
- Increase the schedule description length so AWS automatically retains failed job payloads indefinitely.
- Set the retry count to zero and disable all failure destinations because failed schedules are always replayed automatically later.
Explanation
Correct Answer
A:Scheduler retry controls transient invocation attempts, while an SQS DLQ can retain invocation failure information for operational recovery.
Incorrect Answers
B:False success signals conceal business failure and defeat the objective of reliable scheduled completion.
C:Metadata size does not create durable failure recovery or manage configured invocation retries.
D:Without retries or durable failure routing, a missed invocation may require manual intervention and may be lost to the workflow.
Question 459
A genomics processing pipeline already runs container jobs in AWS Batch. A downstream variant-calling job must not begin before the upstream alignment job succeeds; merely submitting both to the same job queue causes occasional out-of-order execution. What should the submission include?
- Set both jobs to the same queue priority and assume priority alone expresses a success-dependent order.
- Increase the downstream job’s vCPU request and assume its scheduler automatically waits for a different job’s success.
- Use a CloudFront cache policy to sequence AWS Batch container jobs based on completed artifacts.
- Submit the downstream AWS Batch job with a dependency on the upstream alignment job ID so it becomes runnable only after the prerequisite succeeds.
Explanation
Correct Answer
D:AWS Batch job dependencies hold a dependent job in PENDING until referenced jobs complete successfully; failure of a dependency prevents the dependent run. This provides scheduling semantics without polling and guessing when work finished.
Incorrect Answers
A:Priority affects runnable-job selection but does not encode a strict dependency on the outcome of another job.
B:Compute resource requests affect scheduling fit rather than establishing required-success relationships.
C:A CDN cache policy does not manage batch execution ordering.
Question 460
An AWS Batch platform must run one class of GPU inference jobs and another class of low-memory CPU transforms. The resources have very different cost and priority profiles. Which Batch constructs should separate scheduling?
- Put all jobs in one process on one untyped EC2 instance because Batch ignores CPU, GPU and memory differences.
- Define suitable job definitions and queues associated with compute environments matched to the resource classes.
- Use only S3 object tags to allocate GPUs to Batch jobs, without job resource definitions.
- Require all GPU tasks to run through API Gateway synchronous integrations until they finish.
Explanation
Correct Answer
B:Job definitions express workload resource needs, while queues and compute environments let the scheduler map workloads to appropriate capacity.
Incorrect Answers
A:Batch can schedule by job resource requirements, and ignoring them risks failed placement and wasted capacity.
C:S3 metadata does not directly request container accelerators or govern Batch placement.
D:Long compute jobs are better scheduled separately than coupled to finite HTTP request timeouts.
