Question 200
A state machine invokes a throttled service that usually recovers within seconds. Which error-handling design is appropriate?
- SQS ChangeMessageVisibility heartbeat
- Idempotent processing checkpoint
- SQS dead-letter queue
- Step Functions retry with backoff
- Multi-AZ dependent services
Explanation
Correct Answer
D:State machine Retry rules can handle selected errors with intervals and exponential backoff; in this scenario, define a selective Retry policy and a failure path so short-lived errors are retried without restarting the entire pipeline.
Incorrect Answers
A:The SQS ChangeMessageVisibility API can dynamically extend the invisibility period of an in-flight message as a variable-length task proceeds; although useful for extending SQS visibility for a task whose running time varies widely, it leaves the need for recovering from transient downstream API failures unmet.
B:Durable checkpoints plus idempotent sink writes limit duplicate effects after task retries; it would target making a restarted pipeline safe to replay, while the question is about recovering from transient downstream API failures.
C:A dead-letter queue collects messages that repeatedly fail processing after configured receive attempts; its benefit is isolating permanently failing messages for investigation; that is different from recovering from transient downstream API failures.
E:Using supported multi-AZ data services can reduce interruption from a single Availability Zone failure; this can support keeping stateful dependencies available during one AZ outage, but would not address the need for recovering from transient downstream API failures.
Question 201
Analysts need a purchased market-data dataset from a commercial provider rather than a direct internal-account table grant. Which service is relevant?
- AWS Data Exchange subscription
- Lake Formation cross-account sharing
- S3 access point scoped sharing
- Redshift producer datashare authorization
Explanation
Correct Answer
A:AWS Data Exchange provides ways for subscribers to discover and access licensed third-party datasets; in this scenario, subscribe to the approved data product and consume it via its supported delivery method and licensing conditions.
Incorrect Answers
B:Lake Formation can govern cross-account catalog resource permissions for supported data-lake sharing workflows; the current need is for obtaining an external provider's licensed dataset under managed terms, not granting another account access to approved governed S3 tables.
C:S3 access points can provide separate access and network controls for applications using a common bucket; this can support sharing one lake bucket with several differently authorized applications, but would not address the need for obtaining an external provider's licensed dataset under managed terms.
D:Redshift producer administrators grant and authorize consumer access to datashares containing specified warehouse objects; this helps with authorizing a new cross-account consumer of an existing Redshift datashare, not obtaining an external provider's licensed dataset under managed terms.
Question 202
A web team already uses GraphQL and must retrieve selected attributes of entities through a managed endpoint. Which service is suitable?
- S3 presigned GET URL
- Lambda query facade
- API Gateway REST API integration
- AppSync GraphQL resolver
Explanation
Correct Answer
D:AppSync supports GraphQL operations with resolvers that access data sources; in this scenario, define a GraphQL schema and resolvers for authorized fields, letting clients request precisely their needed response structure.
Incorrect Answers
A:A presigned S3 URL grants time-limited access to a specific object using the signer's permissions; the current need is for exposing a typed GraphQL interface to clients, not securely delivering a large generated report file.
B:A Lambda handler can validate inputs, query a datastore and serialize responses for API callers; this can support hiding database credentials and SQL behind a service endpoint, but would not address the need for exposing a typed GraphQL interface to clients.
C:API Gateway can expose authenticated HTTP methods that route to a backed data service; this helps with publishing filtered datasets through a governed API, not exposing a typed GraphQL interface to clients.
Question 203
The same aggregation is queried hundreds of times per hour while its source facts update only daily. Which Redshift object can help?
- Redshift materialized view
- Redshift datashare
- Redshift federated query
- Redshift Spectrum external tables
Explanation
Correct Answer
A:A materialized view persists computed query results for repeated retrieval and planned refresh; in this scenario, maintain a materialized result with a refresh schedule matched to freshness needs, reducing repeated aggregation cost.
Incorrect Answers
B:Redshift data sharing lets authorized consumers query shared warehouse data without making full copies; this helps with exposing governed datasets to a separate Redshift consumer, not caching a recurring expensive analytic result inside the warehouse.
C:Redshift federated query reads supported operational relational databases from analytic SQL; although useful for joining current operational rows from a supported database without bulk replication, it leaves the need for caching a recurring expensive analytic result inside the warehouse unmet.
D:Redshift Spectrum queries external data in S3 via catalog-backed external table definitions; it would target joining S3 lake datasets without first loading them into warehouse tables, while the question is about caching a recurring expensive analytic result inside the warehouse.
Question 204
Security staff must run historical queries across collected control-plane events without assembling raw S3 log files manually. Which feature helps?
- CloudTrail trail to S3
- CloudTrail S3 data events
- CloudTrail Lake event data store
- CloudTrail management events
Explanation
Correct Answer
C:CloudTrail Lake stores queryable event data for SQL-based investigation over configured event sources; in this scenario, query an appropriately configured CloudTrail Lake event data store, respecting selected event categories and retention periods.
Incorrect Answers
A:An organization trail can deliver selected CloudTrail records to a centralized S3 destination for retention and auditing; this can support consolidating API audit records from multiple accounts, but would not address the need for searching years of selected API activity with SQL.
B:CloudTrail data events can record object-level operations for selected S3 resources when configured; this helps with auditing individual object reads in a sensitive lake, not searching years of selected API activity with SQL.
D:Management events record control-plane API operations such as creating or updating AWS resources; it would target auditing who modified a Kinesis stream configuration, while the question is about searching years of selected API activity with SQL.
Question 205
A data scientist needs interactive encoding and cleaning steps before exporting a prepared dataset. Which tool is suitable?
- Notebook exploratory validation
- DataBrew missing-value transform
- SageMaker Data Wrangler transformations
- Athena conditional validation SQL
Explanation
Correct Answer
C:Data Wrangler offers visual data preparation transformations in supported SageMaker data workflows; in this scenario, use the visual transformation interface and export repeatable preparation logic rather than manually editing every raw file.
Incorrect Answers
A:Notebooks enable iterative code, plots and checks for exploratory analysis of sampled datasets; the current need is for preparing features or exploratory datasets through a visual workflow, not investigating anomalies with repeatable interactive code.
B:DataBrew recipes support data cleaning operations for missing or inconsistent tabular field values; this can support standardizing blanks without custom ETL code, but would not address the need for preparing features or exploratory datasets through a visual workflow.
D:Athena SQL can detect rule violations using predicates and aggregations over cataloged data; although useful for quantifying invalid rows before publishing an S3 dataset, it leaves the need for preparing features or exploratory datasets through a visual workflow unmet.
Question 206
All IoT devices send a single partition key and one shard saturates despite available stream capacity. What producer change is needed?
- SQS standard queue fan-in
- Separate registered stream consumers
- Kinesis partition key design
- Kinesis enhanced fan-out
- Kafka consumer groups
Explanation
Correct Answer
C:A Kinesis partition key determines shard assignment and per-key ordering for records; in this scenario, use a higher-cardinality key that distributes records while retaining the ordering guarantees needed within each logical entity.
Incorrect Answers
A:A standard SQS queue buffers messages from multiple producers for a pool of competing worker consumers; this can support decoupling many asynchronous producers from workers, but would not address the need for spreading heavy producer traffic across shards.
B:Applications can register distinct Kinesis stream consumers to obtain their own enhanced fan-out paths; this helps with creating separately managed read identities for independent applications, not spreading heavy producer traffic across shards.
D:Enhanced fan-out gives registered consumers dedicated throughput for a Kinesis Data Streams shard; it would target isolating high-throughput independent stream readers, while the question is about spreading heavy producer traffic across shards.
E:Kafka consumer groups divide topic partitions among members within a group while different groups read independently; its benefit is scaling workers for one Kafka subscription; that is different from spreading heavy producer traffic across shards.
Question 207
A reporting query needs one row per store per day with total revenue, not all transaction rows. Which SQL clause fits?
- Athena SQL view
- Redshift SQL window function
- SQL PIVOT transformation
- SQL GROUP BY aggregate
Explanation
Correct Answer
D:GROUP BY collapses input rows into groups and computes aggregate measures per distinct key; in this scenario, group by store and date and aggregate the amount measure with any required null and duplication handling.
Incorrect Answers
A:An Athena view stores a query definition for reuse over underlying tables; although useful for exposing a stable SQL projection over lake data, it leaves the need for producing daily totals per store unmet.
B:Window functions calculate values across related rows while preserving row-level output; it would target computing a rolling average for each device, while the question is about producing daily totals per store.
C:Pivoting turns distinct category values into separate columns for cross-tab reports in supported SQL engines; its benefit is presenting monthly category totals as columns; that is different from producing daily totals per store.
Question 208
A Lambda handler spends most of its duration decompressing payloads and runs with very low configured memory. What resource setting should be benchmarked?
- Lambda asynchronous invocation destinations
- Reserved concurrency
- Provisioned concurrency
- Lambda memory/CPU setting
Explanation
Correct Answer
D:Lambda allocates more CPU capacity as the configured memory size increases; in this scenario, test higher memory settings for the associated CPU increase and compare duration and cost, not just memory use.
Incorrect Answers
A:Lambda asynchronous destinations can send execution success or failure records to supported targets; it would target recording failed asynchronous enrichments for repair, while the question is about speeding CPU-intensive parsing in a Lambda transform.
B:Reserved concurrency allocates a concurrency limit and reservation to a specific Lambda function; its benefit is protecting a downstream database from too many simultaneous function calls; that is different from speeding CPU-intensive parsing in a Lambda transform.
C:Provisioned concurrency pre-initializes a configured number of Lambda execution environments to reduce cold-start delay; the current need is for speeding CPU-intensive parsing in a Lambda transform, not stabilizing low-latency synchronous API responses.
Question 209
A new dt=2026-11-06 folder exists in S3, but a catalog-backed query cannot see the partition. What metadata must be updated?
- Glue partition metadata
- Athena named database
- Glue Data Catalog table
- Data Catalog table version history
Explanation
Correct Answer
A:Catalog partition definitions map partition values to locations so compatible engines can prune object sets; in this scenario, add or discover the new partition metadata so queries map its date value to the correct object location.
Incorrect Answers
B:An Athena database is a logical namespace for tables in the underlying catalog; the current need is for registering new date partitions before reading them, not organizing named datasets for a SQL workgroup.
C:A Data Catalog table describes a dataset's location, columns and serialization for compatible query engines; this can support sharing an S3 table definition across Athena and Glue, but would not address the need for registering new date partitions before reading them.
D:Data Catalog versions can preserve prior table metadata definitions for certain schema-management workflows; this helps with inspecting a prior table schema after a crawler update, not registering new date partitions before reading them.
Question 210
An EventBridge rule intermittently fails to invoke a restricted endpoint and losing these failed events is unacceptable. What safeguard helps?
- SNS subscription filtering
- CloudWatch alarm evaluation periods
- CloudWatch alarm composite
- EventBridge failed-event DLQ
Explanation
Correct Answer
D:A rule target can use an SQS dead-letter queue to retain undelivered events after retry exhaustion; in this scenario, configure a target dead-letter queue and investigate delivery failures with suitable permissions and alerts.
Incorrect Answers
A:SNS subscription filter policies select which topic messages are delivered to specific subscribers; its benefit is routing alerts by data pipeline ownership; that is different from capturing notifications that fail delivery to the target.
B:Alarm evaluation periods govern how many datapoints must breach a threshold before the alarm changes state; the current need is for capturing notifications that fail delivery to the target, not suppressing alerts for harmless single-sample spikes.
C:Composite alarms combine the states of other alarms using logical expressions; this can support paging only when error rate and backlog are both unhealthy, but would not address the need for capturing notifications that fail delivery to the target.