DynamoDB resilience — PITR + Streams + stub stream Lambda

15 timed scenario questions (~24 min, ~96 sec each). Reinforces night-18-dynamodb-resilience.md and night-18-lab-dynamodb-setup scripts. Answer without peeking at night-18-quiz-answers.json first.

1. 1.An operator accidentally deletes items from saa-study-gsa-wiki-views. PITR is enabled. What is the correct recovery action?

   A.Run restore-table-to-point-in-time to a new table name within the 35-day windowB.Enable the stream TRIM_HORIZON starting position to replay deleted rows in placeC.Call UpdateItem on each missing pageId with ConditionExpression attribute_not_existsD.Disable PITR, run on-demand backup, and import the snapshot over the same table name

2. 2.night-18-lab enables StreamViewType NEW_AND_OLD_IMAGES on saa-study-gsa-wiki-views. The stub Lambda logs old and new viewCount on MODIFY. Which view type would hide the previous count?

   A.OLD_IMAGE only — emits the item before the change but not the new image after the updateB.KEYS_ONLY — emits partition key fields only with no attribute payload in the recordC.NEW_IMAGE only — emits the item after the change but not the previous image before itD.NEW_AND_OLD_IMAGES — includes both old and new images on every stream event type by default

3. 3.You create a DynamoDB stream event source mapping with StartingPosition=LATEST. You then run -BumpView on wiki/iceland. An earlier bump from before the mapping never invoked Lambda. Why?

   A.PITR must be disabled while DynamoDB stream event source mappings remain active on the tableB.On-demand tables cannot emit MODIFY stream events until provisioned capacity mode is enabledC.LATEST delivers only records after the mapping is created — prior stream data is skippedD.Lambda polls an SQS queue behind the stream and the completion queue was empty at mapping time

4. 4.GSA prod wiki-api increments page views with UpdateItem. night-18-lab uses ADD viewCount :inc. Why is ADD preferred over read-then-SET for hot counters?

   A.ADD requires a TransactWriteItems call and therefore guarantees cross-table consistency alwaysB.ADD only works on sort-key attributes, which prevents hot partition mistakes on pageId keysC.ADD triggers Kinesis Data Streams automatically for analytics export of every counter incrementD.ADD atomically increments without a separate GetItem — avoids lost updates under concurrency

5. 5.A stream processor Lambda fails mid-batch. DynamoDB stream event source mappings retry. What design property keeps audit logging safe?

   A.Writing the same log line twice is fine — the handler prints idempotent summaries onlyB.Set batch size to 1 and disable retries on the DynamoDB stream event source mapping entirelyC.Use FIFO SQS between the stream and Lambda to force exactly-once delivery of every recordD.Switch StreamViewType to KEYS_ONLY so retries carry no duplicate attribute payload at all

6. 6.Compare tonight's stub stream Lambda to Night 17 saa-study-gsa-stats-uploader. What is the trigger difference?

   A.Both use EventBridge schedules that fire every minute against their target functionsB.Night 18 reacts to DynamoDB row changes; Night 17 polls SQS completion messagesC.Night 18 uses Aurora logical replication slots; Night 17 uses DynamoDB Streams onlyD.Both use API Gateway HTTP API synchronous invoke paths for every incoming request

7. 7.Global Ski Atlas needs wiki read/write in us-east-1 and eu-west-1 with local low latency. Tables are DynamoDB. Which feature fits?

   A.Aurora Global Database for WikiPages items stored as JSON rows in PostgreSQLB.RDS read replica in eu-west-1 attached directly to the wiki DynamoDB table endpointC.DynamoDB global tables with multi-Region active-active replication for local accessD.S3 Cross-Region Replication on the DynamoDB table export bucket only for DR reads

8. 8.A Lambda in Night 9 private subnets calls DynamoDB in the same Region. NAT Gateway charges are high on GetItem traffic. What reduces DynamoDB egress cost?

   A.Enable DynamoDB Accelerator (DAX) in the public subnet to cache GetItem responsesB.Move the table to Aurora Serverless v2 in the same VPC to avoid DynamoDB API callsC.Attach an internet gateway directly to the private subnet route table for DynamoDBD.Add a DynamoDB gateway VPC endpoint and route DynamoDB prefix-list traffic locally

9. 9.night-18-lab uses PAY_PER_REQUEST on saa-study-gsa-wiki-views. Wiki traffic spikes unpredictably on powder days. Why match on-demand billing here?

   A.On-demand scales read/write capacity per request without provisioning RCU/WCU upfrontB.On-demand removes the need for partition keys on hot counter items stored in the wiki tableC.On-demand automatically enables DynamoDB global tables in a second Region for every new tableD.On-demand requires PITR to stay disabled so continuous backup cost stays predictable monthly

10. 10.You need centralized nightly backups for DynamoDB wiki tables, Aurora resort_stats, and EBS volumes under one retention policy. Which service?

    A.DynamoDB PITR alone — it backs up all AWS data stores in the account automaticallyB.AWS Backup with a plan that selects resources by tag or ARN (Night 19 topic preview)C.CloudTrail with data events enabled on every DynamoDB table stream shard for restoreD.RDS automated backup window extended to 35 days and applied to all AWS services

11. 11.Two Regions write the same item key in a global table within the same second. What conflict resolution applies?

    A.DynamoDB rejects the second write with ConditionalCheckFailedException in both Regions alwaysB.The write from the Region with the lexicographically lower table ARN wins automatically for that itemC.Last writer wins per item — eventual consistency across Regions for conflicting updatesD.Strongly consistent reads block all conflicting writes until a global distributed lock expires

12. 12.resort_stats monthly_runs needs SQL JOIN reports. wiki/iceland viewCount needs millions of atomic increments. Where should each live per GSA split?

    A.Both in Aurora — use JSONB columns for high-volume wiki view counter incrementsB.Both in DynamoDB — model JOINs as multiple Query calls for each reporting requestC.S3 Parquet for both — Athena replaces DynamoDB and Aurora for counters and JOINsD.Aurora for relational resort_stats; DynamoDB for hot wiki view counter increments

13. 13.The stream processor IAM role needs dynamodb:GetRecords on the stream ARN. What else must the role allow for Lambda to emit logs?

    A.logs:CreateLogGroup, logs:CreateLogStream, and logs:PutLogEvents on CloudWatch LogsB.ec2:CreateNetworkInterface for DynamoDB Streams ENI attachment inside the private VPC subnetsC.dynamodb:UpdateItem on the source wiki table for every processed stream record batch itemD.sqs:SendMessage to mirror each stream batch into the Night 13 Iceland completion queue path

14. 14.Maximum continuous backup window for DynamoDB PITR on a single-Region table is how long?

    A.7 days — same as the default RDS automated backup retention period for DB instancesB.35 days — exam standard maximum retention for DynamoDB PITR continuous backupsC.14 days — matches the maximum SQS standard queue message retention period on queuesD.365 days — available only when AWS Backup premium tier is enabled on the AWS account

15. 15.night-18-lab teardown deletes saa-study-gsa-wiki-views. Prod WikiPages is untouched. A teammate asks whether teardown affects Night 17 Aurora. What is true?

    A.Aurora writer endpoint stops whenever any DynamoDB table in the same account is deletedB.Night 17 stats-uploader loses its VPC ENI when the wiki-views study table is removedC.Nothing — Night 18 study table and stream Lambda are independent of Night 16/17 AuroraD.SQS completion queue is deleted automatically together with the DynamoDB study table teardown