Question 390
A thousand nightly independent jobs all start at midnight and overload the metadata database. Which scheduling option smooths launch peaks?
- Airflow task retry policy
- Step Functions execution history
- Airflow dependency trigger rule
- EventBridge Scheduler flexible time window
- Glue conditional trigger
Explanation
Correct Answer
D:Flexible time windows let a scheduled invocation occur within a configured interval instead of at an exact instant; in this scenario, allow a flexible invocation window for workloads without strict start-time requirements to distribute schedule pressure.
Incorrect Answers
A:Airflow tasks support retry counts and delays within a DAG's execution schedule; it would target retrying a transiently failing DAG task, while the question is about spreading noncritical batch launches to avoid thundering herds.
B:Step Functions maintains execution event history for supported workflows and helps locate failed states; its benefit is tracing the precise failure transition in a workflow; that is different from spreading noncritical batch launches to avoid thundering herds.
C:Airflow trigger rules determine when a downstream task may run relative to upstream task states; the current need is for spreading noncritical batch launches to avoid thundering herds, not ensuring a cleanup task runs after upstream failures.
E:Glue conditional triggers can start downstream jobs when designated upstream jobs or crawlers reach configured states; this helps with running a publish job only after upstream Glue success, not spreading noncritical batch launches to avoid thundering herds.
Question 391
A state machine intentionally retries individual task errors, and an alert must occur only after the execution ultimately fails. What event should drive it?
- EventBridge event-to-target routing
- SQS durable queue
- SNS topic fan-out
- Step Functions workflow status alert
- CloudWatch alarm action
Explanation
Correct Answer
D:An EventBridge rule can match Step Functions execution status changes for monitoring and notifications; in this scenario, filter for the terminal failed execution status and route it to notification, avoiding premature alarms on handled retries.
Incorrect Answers
A:EventBridge rules route matching state-change events to designated consumers; it would target notifying only after a specific job state transition, while the question is about notifying when a whole workflow fails rather than one retrying task.
B:SQS retains messages for asynchronous processing by consumers that may be temporarily unavailable; its benefit is buffering remediation work until a worker can run; that is different from notifying when a whole workflow fails rather than one retrying task.
C:SNS publishes notifications to multiple subscribed endpoints or queues using topic subscriptions; the current need is for notifying when a whole workflow fails rather than one retrying task, not delivering one alert to several independent recipients.
E:CloudWatch alarms evaluate metrics against thresholds and can invoke configured notification actions; this helps with alerting after repeated error metric breaches, not notifying when a whole workflow fails rather than one retrying task.
Question 392
Sensor readings have a repeating 60-second calibration cycle and an audit currently selects every sixtieth reading. What risk should be evaluated?
- Stratified random sampling
- Reservoir sampling
- Random sampling
- Spark key salting
- Systematic sampling caveat
Explanation
Correct Answer
E:Systematic sampling takes observations at regular intervals and may bias results when the stream is periodic; in this scenario, sampling at the cycle interval may overrepresent one phase; randomize selection or choose a stratified design that covers phases.
Incorrect Answers
A:Stratified sampling draws observations from defined subgroups to preserve rare but important categories; this helps with assessing a dataset with very uneven country distributions, not avoiding periodic bias when checking machine readings.
B:Reservoir sampling can maintain a bounded random sample from a stream of unknown total length; although useful for sampling a large unknown-length event stream with fixed memory, it leaves the need for avoiding periodic bias when checking machine readings unmet.
C:Simple random sampling selects records with equal inclusion probability when an unbiased population estimate is needed; it would target building a general quality sample without subgroup constraints, while the question is about avoiding periodic bias when checking machine readings.
D:Key salting distributes records for a heavily skewed join key across multiple synthetic partitions; its benefit is reducing one-key hotspots in a distributed join; that is different from avoiding periodic bias when checking machine readings.
Question 393
An Athena table has accurate partitions but a query without date filters scans the full historical lake. What SQL practice reduces work?
- Partition pruning predicate
- Athena partition projection
- MSCK REPAIR TABLE
- ALTER TABLE ADD PARTITION
- Glue crawler partition update
Explanation
Correct Answer
A:Query filters on partition columns allow compatible engines to avoid scanning unrelated object prefixes; in this scenario, add selective partition-column predicates so the planner can prune partitions; metadata accuracy alone cannot limit a broad query.
Incorrect Answers
B:Partition projection computes partition locations from table properties instead of storing every partition explicitly; this can support avoiding per-day catalog updates for predictable partitions, but would not address the need for reducing reads after partition registration.
C:Athena's MSCK REPAIR TABLE can discover compatible Hive-style S3 partitions and add missing catalog entries; this helps with registering newly created Hive-style partition folders, not reducing reads after partition registration.
D:Adding a partition explicitly registers specified partition values and their locations in the catalog; although useful for registering a small known list of partitions precisely, it leaves the need for reducing reads after partition registration unmet.
E:A Glue crawler can discover supported new partitions and maintain catalog metadata during a run; it would target keeping partition metadata synchronized after scheduled ingestion, while the question is about reducing reads after partition registration.
Question 394
A workflow passes huge prior API responses to every following task and starts exceeding payload constraints. What change helps?
- Step Functions execution input/output paths
- Step Functions Choice state
- Step Functions Map state
- Step Functions Catch transition
- Lambda timeout setting
Explanation
Correct Answer
A:InputPath, ResultPath and OutputPath can shape data passed between workflow states; in this scenario, select and reduce state payload fields at each transition, and store large datasets externally with object references.
Incorrect Answers
B:A Choice state routes execution based on a condition in the state input; this can support selecting a branch after checking data type, but would not address the need for retaining only relevant fields between processing tasks.
C:A Map state processes collection items using controlled parallel iterations in a workflow; this helps with applying the same transform to many independent input files, not retaining only relevant fields between processing tasks.
D:Catch handlers move execution to a recovery state when a configured task error occurs; although useful for sending failed processing to a quarantine path, it leaves the need for retaining only relevant fields between processing tasks unmet.
E:A Lambda timeout caps the maximum duration of a single function invocation; it would target preventing a stalled serverless task from running indefinitely, while the question is about retaining only relevant fields between processing tasks.
Question 395
A Glue application writes 'QUARANTINE_THRESHOLD' in logs, and operations need a metric alarm each time it appears. What feature turns logs into metrics?
- CloudWatch subscription filter
- OpenSearch indexed log search
- CloudWatch Logs Insights
- CloudWatch metric filter
- Athena over structured S3 logs
Explanation
Correct Answer
D:A metric filter creates CloudWatch metric datapoints from matching CloudWatch Logs events; in this scenario, create a metric filter for the stable pattern and alarm on emitted counts; avoid relying only on generic job failure metrics.
Incorrect Answers
A:Log subscription filters send matching events from CloudWatch Logs to supported downstream consumers; this helps with streaming selected application errors into a central processor, not alerting on a distinctive application error pattern.
B:OpenSearch can index log documents for search, filtering and operational dashboards; although useful for searching high-cardinality text patterns across indexed application logs, it leaves the need for alerting on a distinctive application error pattern unmet.
C:Logs Insights searches and aggregates records in CloudWatch log groups with query syntax; it would target interactively grouping recent Lambda exception events from CloudWatch log groups, while the question is about alerting on a distinctive application error pattern.
E:Athena can query centrally stored S3 audit logs after their schema is represented in a catalog; the current need is for alerting on a distinctive application error pattern, not running SQL over accumulated S3 log archives.
Question 396
Policy requires data to be encrypted before any object bytes are sent over the network to S3. Which pattern is appropriate?
- Client-side encryption before upload
- TLS for JDBC connections
- Cross-account S3 bucket policy
- Cross-account KMS key policy
- External role IAM KMS permission
Explanation
Correct Answer
A:Client-side encryption transforms plaintext into ciphertext before it leaves the originating application; in this scenario, encrypt in the producer using controlled keys and upload ciphertext, managing decryption securely for authorized consumers.
Incorrect Answers
B:TLS secures supported database client-server communications against interception in transit; the current need is for ensuring raw plaintext never leaves the source environment, not encrypting traffic from Glue to a relational database endpoint.
C:An S3 resource policy can authorize an external principal to read particular object keys; this can support allowing a partner role to reach shared encrypted object paths, but would not address the need for ensuring raw plaintext never leaves the source environment.
D:A customer-managed KMS key policy can authorize a principal or external account for intended cryptographic operations; this helps with granting a partner account permitted access to encrypted lake objects, not ensuring raw plaintext never leaves the source environment.
E:A consumer identity policy can permit use of an externally owned customer-managed KMS key when its key policy also allows it; although useful for enabling a cross-account reader role to call KMS Decrypt, it leaves the need for ensuring raw plaintext never leaves the source environment unmet.
Question 397
A transformation bug affected last week's input, but the fixed job skips those files due to recorded state. Which operation enables replay?
- EMR step logs in S3
- EMR managed scaling policies
- Glue job bookmark reset
- Glue job worker increase
- Glue job data partition tuning
Explanation
Correct Answer
C:Resetting a Glue bookmark permits an ETL job to reprocess previously tracked source data intentionally; in this scenario, reset or rewind bookmark state as appropriate and make target writes idempotent to avoid unintended duplicates.
Incorrect Answers
A:EMR cluster logging can preserve step and application output in an S3 log destination; although useful for investigating a failed historical EMR processing step, it leaves the need for replaying a corrected historical data batch unmet.
B:EMR managed scaling adjusts available node capacity under configured bounds for supported clusters; it would target matching variable Spark compute demand without constant oversizing, while the question is about replaying a corrected historical data batch.
D:More appropriate Glue worker resources can help when Spark tasks are genuinely CPU or memory constrained; the current need is for replaying a corrected historical data batch, not reducing a resource-limited Glue job runtime.
E:Spark partition sizing determines whether distributed ETL tasks can run evenly and make good use of available executor slots; this can support raising parallel task utilization when a Glue read produces too few partitions, but would not address the need for replaying a corrected historical data batch.
Question 398
Country codes contain whitespace and inconsistent case, causing avoidable failed membership checks. What preparation step helps?
- DataBrew profile statistics
- Glue Data Quality evaluation results
- Glue Data Quality ruleset
- DataBrew validation recipe
- Quality quarantine S3 prefix
Explanation
Correct Answer
D:A DataBrew recipe records transformations that can normalize inconsistent values and produce an approved output; in this scenario, apply a reviewed trimming and normalization recipe before measuring whether values meet the approved code set.
Incorrect Answers
A:DataBrew profile jobs summarize null values, distributions and other properties of tabular input; this can support measuring baseline data characteristics before quality rule design, but would not address the need for standardizing known fields before final quality evaluation.
B:Quality evaluation results record which configured checks pass and fail for a tested dataset; this helps with tracking rule failures across successive daily runs, not standardizing known fields before final quality evaluation.
C:Glue Data Quality supports rulesets for evaluating data quality against supported datasets and ETL flows; although useful for applying repeatable quality checks in a Glue pipeline, it leaves the need for standardizing known fields before final quality evaluation unmet.
E:A separate S3 prefix can preserve rejected records for controlled remediation without polluting curated queries; its benefit is keeping invalid records available for debugging but out of production analytics; that is different from standardizing known fields before final quality evaluation.
Question 399
The same automation role serves multiple teams, and requests should be evaluated against the team identity supplied in the role session. Which feature fits?
- Session tag authorization
- Cross-account bucket policy
- S3 access point policy
- IAM ABAC with principal tags
- Lake Formation LF-Tag based access
- IAM RBAC role assignment
Explanation
Correct Answer
A:Session tags can carry approved attributes in assumed-role sessions for ABAC decisions; in this scenario, pass controlled session tags through permitted role assumption and use matching condition keys in policies, limiting who may set those tags.
Incorrect Answers
B:An S3 bucket resource policy can allow a designated external account or role to access selected objects; the current need is for propagating per-session team attributes to authorization rules, not authorizing direct read access from another AWS account.
C:S3 access points offer distinct access controls and network policies for shared S3 datasets; this can support exposing one bucket to applications with different data-access boundaries, but would not address the need for propagating per-session team attributes to authorization rules.
D:Attribute-based access control compares principal and resource attributes through supported IAM condition keys; this helps with mapping ETL roles to same-team tagged resources, not propagating per-session team attributes to authorization rules.
E:LF-Tag authorization grants governed table access based on metadata classifications and expressions; although useful for authorizing lake assets by governed data classification, it leaves the need for propagating per-session team attributes to authorization rules unmet.
F:Role-based access control groups permissions according to job functions or workload responsibilities; it would target giving production operators a stable predefined permission set, while the question is about propagating per-session team attributes to authorization rules.
Question 400
Five ETL teams need the same reviewed S3 access statements and independent manual copies have drifted. What IAM policy form helps?
- IAM managed versus inline policy
- IAM identity policy attachment
- SageMaker Unified Studio domain
- SageMaker domain unit hierarchy
- SageMaker Unified Studio project membership
- Service-linked IAM role
Explanation
Correct Answer
A:Customer-managed IAM policies can be versioned and reused across roles, unlike role-specific inline policies; in this scenario, use a shared customer-managed policy with controlled review and versioning while attaching it only to intended roles.
Incorrect Answers
B:An identity-based IAM policy grants allowed API actions to a user, group or role subject to other policy checks; this helps with authorizing a data-processing role for a narrow AWS API action, not reusing controlled permissions across multiple data pipeline roles.
C:A Unified Studio domain defines an overarching environment for governed data and analytics workspaces; although useful for establishing a common organizational boundary for governed projects, it leaves the need for reusing controlled permissions across multiple data pipeline roles unmet.
D:Domain units group related teams and help structure delegated governance within a domain; it would target separating business-unit stewardship responsibilities, while the question is about reusing controlled permissions across multiple data pipeline roles.
E:Project membership controls collaboration and access within the project's governed workspace; its benefit is limiting an analyst's membership to one approved project; that is different from reusing controlled permissions across multiple data pipeline roles.
F:Service-linked roles are IAM roles predefined for supported AWS services to act on an account's behalf; the current need is for reusing controlled permissions across multiple data pipeline roles, not allowing an AWS managed service to perform its own required operations.