Lab 2C — SQS decoupling

15 timed scenario questions (~24 min, ~96 sec each). Reinforces night-13-sqs-decoupling.md and night-13-lab-sqs-setup scripts. Answer without peeking at night-13-quiz-answers.json first.

1. 1.Global Ski Atlas runs Iceland as a monthly Fargate batch job (Night 12). A separate stats-uploader worker should process completions without polling ecs:DescribeTasks in a loop. What integration service best buffers completion events between the batch job and the uploader?

   A.Amazon SNS — fan-out push to email and Lambda when the batch job completesB.AWS Step Functions Express — sync workflow invoked from the Iceland container on exitC.Amazon SQS — producer sends or EventBridge routes; consumer polls at its own paceD.Amazon API Gateway REST with VPC Link and ECS integration for completion callbacks

2. 2.A Lambda consumer processes SQS messages that take up to 3 minutes. The queue visibility timeout is 30 seconds. What problem will you see in production?

   A.Messages are permanently deleted from the queue after 30 seconds regardless of outcomeB.The same message becomes visible again while Lambda still runs — duplicate processingC.SQS extends visibility automatically for all Lambdas without event source mapping setupD.FIFO ordering breaks because visibility timeout must be at least 60 seconds on standard

3. 3.The Iceland completion queue receives one message per successful monthly run. Order between months does not matter; maximum throughput is desired. Which queue type should you create?

   A.SQS standard queue — at-least-once delivery, best-effort ordering, highest throughputB.SQS FIFO queue without content-based deduplication — strict order for monthly runsC.Amazon MQ for RabbitMQ — managed broker with durable queues and dead-letter exchangesD.DynamoDB stream with Lambda event source mapping — stream records trigger the uploader

4. 4.night-13-lab-sqs-setup.sh sets maxReceiveCount to 3 on the main queue with a redrive policy pointing at saa-study-gsa-iceland-completion-dlq. What does this accomplish?

   A.Silently deletes poison messages from the main queue after three receives without configuring any DLQB.Automatically retries failed EventBridge rule invocations and resends every event to the main queueC.Guarantees exactly-once delivery of every message that reaches the DLQ after three triesD.Moves messages to the DLQ after three failed processing attempts (received but not deleted)

5. 5.An EventBridge rule targets an SQS queue with FailedInvocations in CloudWatch. The rule event pattern is correct. What is the most common fix on the queue side?

   A.Enable SSE-SQS encryption on the queue and attach a customer-managed KMS key policy for EventBridgeB.Attach the AmazonSQSFullAccess managed policy to the default event bus resource policy for the accountC.Queue policy allowing events.amazonaws.com sqs:SendMessage with aws:SourceArn = rule ARND.Rename the queue with a .fifo suffix and enable content-based deduplication for targets

6. 6.A stats-uploader worker polls an empty completion queue thousands of times per hour, increasing cost and API throttling risk. Which ReceiveMessage parameter reduces empty responses?

   A.DelaySeconds set to 900 on SendMessage — delays new message visibility for 15 minutesB.WaitTimeSeconds long polling (up to 20) — SQS holds the request until a message arrivesC.MessageGroupId on a standard queue — groups empty polls into one long-poll batch per groupD.CloudWatch alarm on ApproximateNumberOfMessages — suppresses ReceiveMessage when empty

7. 7.You need a single worker pool to absorb spikes from many producers without push delivery to every worker. SNS already fans out to five Lambdas. What should sit in front of one consolidated worker tier?

   A.SQS standard queue — workers pull messages; the buffer absorbs producer bursts safelyB.A second SNS topic subscribed to the first — chains fan-out so five Lambdas forward to oneC.EventBridge custom event bus with no targets — buffers events until a rule is attached laterD.AWS AppSync GraphQL subscription — WebSocket push from producers to one consolidated pool

8. 8.A malformed JSON message causes the consumer to throw on every attempt. The operator wants to inspect failures without blocking the main queue forever. Which configuration pair is required?

   A.Message retention of 60 seconds on the main queue only — poison messages expire quicklyB.SNS topic subscribed to the main queue with email — failure notifications to operatorsC.Visibility timeout set to 0 — messages return to the queue immediately on every receiveD.DLQ plus RedrivePolicy maxReceiveCount on the source queue for poison-message isolation

9. 9.A payment ledger requires strict ordering per customer account and no duplicate debits. Throughput is moderate. Which SQS option fits?

   A.Standard queue with a random MessageDeduplicationId on each SendMessage callB.Standard queue with DelaySeconds 300 — staggers debits five minutes apart per customerC.FIFO queue with MessageGroupId per customer and content-based deduplication enabledD.EventBridge schedule expression rate(1 second) — serializes ledger writes via timed triggers

10. 10.A consumer receives a message, crashes before calling DeleteMessage, and visibility timeout is 60 seconds. What happens to the message?

    A.The message is permanently removed — unacknowledged receives count as fully processedB.After visibility timeout it becomes visible again for another consumer — at-least-onceC.SQS immediately moves the message to a DLQ even when no redrive policy is configuredD.EventBridge detects the consumer crash and republishes a duplicate ECS Task State Change

11. 11.Night 12 EventBridge schedule triggers Iceland RunTask. Night 13 adds an ECS Task State Change rule → SQS. How does that split responsibilities?

    A.Night 12 answers when to run; Night 13 decouples post-success work from the batch containerB.Night 13 replaces the Night 12 schedule rule — only one EventBridge rule may exist per event busC.SQS becomes the scheduler and calls ecs:RunTask directly instead of relying on the EventBridge cronD.Night 13 removes the need for iam:PassRole on the ECS execution and task roles for RunTask

12. 12.An exam question shows producers in us-east-1 and a consumer in eu-west-1 that must share a work queue with minimal latency to the consumer. What is the best answer?

    A.Single SQS queue in us-east-1 — cross-Region ReceiveMessage is free with negligible latencyB.FIFO queue with a global endpoint — AWS routes polls to the nearest Region automaticallyC.EventBridge default bus in us-east-1 — replicates all events automatically to every RegionD.SQS queue near the consumer (or replicate events) — SQS is Regional; cross-Region adds cost

13. 13.After sending ten messages you call GetQueueAttributes for ApproximateNumberOfMessages and see 7, then 10 a few seconds later without new sends. Why?

    A.SQS lost three messages due to the approximate counter rounding down under high send volumeB.The DLQ redrive policy absorbed three messages automatically without any consumer receivesC.The attribute is approximate — eventually consistent; not a real-time exact count for examsD.Visibility timeout expired on three in-flight messages and SQS deleted them from the queue

14. 14.night-13-lab-sqs-setup uses an event pattern on ECS Task State Change with containers.exitCode [0] and taskDefinitionArn prefix for globalskiatlas-backend-k8s-iceland. A task exits 1 after S3 upload fails. What happens?

    A.The failed message is sent directly to the DLQ because exit code 1 matches the redrive policy filterB.No completion message is enqueued — pattern requires exit 0; add a failure path (Night 14 SNS)C.EventBridge retries ecs:RunTask automatically until the container exits successfully with code 0D.SQS receives the ECS event anyway with exitCode 1 because the rule matches all STOPPED tasks

15. 15.Night 14 will add SNS so one Iceland failure alert reaches email and a Lambda simultaneously. Why is SNS a better fit than a single SQS queue for that requirement?

    A.SNS fan-out push delivers one publish to many subscribers without shared-queue pollingB.SQS cannot integrate with Lambda — only SNS supports Lambda as an event notification targetC.SNS guarantees strict FIFO ordering across all email and Lambda subscribers for the topicD.SQS standard queues natively support email protocol — one queue delivers to inbox and Lambda