Question 40

Several engineers edit a shared Glue script and unreviewed changes frequently break production jobs. Which development workflow helps?

  1. Git pull request review
  2. Unit tests with fixture data
  3. Immutable deployment artifact
  4. CI pipeline integration tests

Explanation

Correct Answer

A:

Git branches and pull requests support code review before changes merge into a shared baseline; in this scenario, use branches with peer review and required tests before merging, making intended changes and reviewers auditable.

Incorrect Answers

B:

Unit tests validate transformation functions against representative inputs and expected results; the current need is for reviewing ETL changes before production release, not catching mapping regressions before deployment.

C:

Versioned packages and immutable images let a pipeline redeploy the exact artifact tested previously; this can support preventing drift between staging and production ETL code, but would not address the need for reviewing ETL changes before production release.

D:

Integration tests exercise real interfaces between components in an isolated environment; this helps with validating source-to-target contracts before release, not reviewing ETL changes before production release.

Question 41

Records already arrive in Kinesis Data Streams, but the function is never invoked because no stream consumer integration exists. What must be added?

  1. Kinesis event source mapping
  2. Lambda batch failure reporting
  3. Kinesis iterator age metric
  4. Lambda parallelization factor

Explanation

Correct Answer

A:

A Lambda event source mapping reads Kinesis records, checkpoints progress and invokes the function in batches; in this scenario, map the stream to the Lambda function and allow the execution role to read records; Lambda manages the polling loop.

Incorrect Answers

B:

Lambda can report individual failed record IDs for supported stream batch processing; this helps with retrying only failed records in a stream batch, not connecting a Lambda processor to a Kinesis shard.

C:

IteratorAge measures how far behind a stream-reading Lambda consumer is in processing records; although useful for diagnosing lag in a Lambda stream consumer, it leaves the need for connecting a Lambda processor to a Kinesis shard unmet.

D:

The parallelization factor raises concurrent Lambda batches processed per Kinesis shard while preserving partition-key ordering; it would target increasing consumer throughput on busy shards, while the question is about connecting a Lambda processor to a Kinesis shard.

Question 42

A data pipeline connects to RDS using a password that rotates automatically and must never appear in deployment templates. Which store fits?

  1. SSM Parameter Store SecureString
  2. IAM role credentials instead of access keys
  3. KMS encryption context policy
  4. Secrets Manager secret

Explanation

Correct Answer

D:

Secrets Manager stores and retrieves sensitive credentials and supports managed rotation patterns; in this scenario, retrieve the secret at runtime through scoped IAM and coordinate application-side connection refresh after rotation.

Incorrect Answers

A:

Systems Manager Parameter Store can hold encrypted configuration values as SecureString parameters; its benefit is storing a rarely changed encrypted ETL token; that is different from protecting a rotating production database password.

B:

IAM role sessions provide automatically rotated temporary credentials to supported AWS workloads; the current need is for protecting a rotating production database password, not authenticating an AWS SDK call without storing an AWS secret.

C:

KMS key policies can evaluate encryption-context conditions to narrow permitted cryptographic use; this can support scoping decrypt permission to the intended encrypted dataset, but would not address the need for protecting a rotating production database password.

Question 43

Containerized JSON parsers fail with out-of-memory errors on larger files although the ECS service remains healthy. Which adjustment should be tested?

  1. ECS Service Auto Scaling
  2. EKS Horizontal Pod Autoscaler
  3. ECS task CPU and memory sizing
  4. Container image multi-stage build

Explanation

Correct Answer

C:

An ECS task definition reserves CPU and memory for its container workloads; in this scenario, increase task memory to match measured peak working set and confirm supported CPU-memory combinations before redeployment.

Incorrect Answers

A:

ECS Service Auto Scaling changes the desired count of running service tasks based on configured metrics; this helps with adding parallel workers when a queue backlog rises, not giving each transformation container enough compute and RAM.

B:

Kubernetes HPA adjusts replica counts of a target workload using metrics and configured policies; although useful for scaling Kubernetes transformation pods based on utilization, it leaves the need for giving each transformation container enough compute and RAM unmet.

D:

Multi-stage Docker builds separate build tools from the runtime image to reduce final image contents; its benefit is keeping transformation images lean and faster to distribute; that is different from giving each transformation container enough compute and RAM.

Question 44

A data lake contains millions of support descriptions that require topic labels for analytics, not model training. What approach could add enrichment?

  1. PII redaction before LLM call
  2. Human review for low-confidence extraction
  3. Schema validation of LLM output
  4. Bedrock text classification request

Explanation

Correct Answer

D:

Amazon Bedrock can invoke foundation models to classify or summarize unstructured text in a pipeline; in this scenario, invoke an appropriate Bedrock model with bounded prompts and validate structured output before writing labels to a curated table.

Incorrect Answers

A:

Sensitive fields should be removed or minimized before they are sent to a model where policy requires it; the current need is for deriving categories from free-form support notes, not keeping customer identifiers out of model prompts.

B:

A review queue can route uncertain automated extractions to humans instead of blindly publishing them; this can support avoiding unchecked facts in ambiguous document extraction, but would not address the need for deriving categories from free-form support notes.

C:

Model responses require parsing and validation against an expected structured contract before ingestion; this helps with preventing unstructured model responses from breaking ETL, not deriving categories from free-form support notes.

Question 45

A warehouse must aggregate revenue across customers, products and dates while retaining shared descriptive attributes. What modeling pattern fits?

  1. Surrogate warehouse dimension key
  2. Redshift star schema
  3. DynamoDB access-pattern modeling
  4. Lake Formation governed tables

Explanation

Correct Answer

B:

A star schema links facts to dimensions for analytic joins and common reporting aggregations; in this scenario, place measures and foreign keys in a fact table and reuse dimension tables for reporting filters and groupings.

Incorrect Answers

A:

Surrogate keys can stabilize warehouse joins when operational business identifiers change; this can support preserving historical dimension references through source ID changes, but would not address the need for modeling sales facts for BI measures by customer and product.

C:

DynamoDB tables are designed around known partition/sort key access patterns rather than normalized joins; although useful for optimizing predictable NoSQL key queries, it leaves the need for modeling sales facts for BI measures by customer and product unmet.

D:

Lake Formation grants operate on registered data-lake catalog resources and can enforce fine-grained policies; it would target controlling which analysts can read governed S3 tables, while the question is about modeling sales facts for BI measures by customer and product.

Question 46

Batch writers replace partitions while Athena users query an S3-backed table and need consistent results. Which table format feature helps?

  1. Amazon S3 Tables table bucket
  2. Iceberg partition evolution
  3. Iceberg schema evolution
  4. Apache Iceberg snapshots

Explanation

Correct Answer

D:

Iceberg tracks table snapshots and metadata so readers can use consistent committed table states; in this scenario, use Iceberg transactional snapshots with a compatible catalog and engine so readers see committed table versions.

Incorrect Answers

A:

Amazon S3 Tables provides purpose-built table buckets for Apache Iceberg with automatic file compaction and snapshot maintenance; this helps with hosting Iceberg tables with managed storage-level maintenance in Amazon S3, not keeping readers consistent during an object-based table update.

B:

Iceberg can evolve partition specifications without rewriting historical data immediately into one fixed layout; although useful for changing partition strategy as data volumes grow, it leaves the need for keeping readers consistent during an object-based table update unmet.

C:

Iceberg supports schema evolution using field identities rather than relying only on physical column order; it would target adding columns without rewriting every historical file, while the question is about keeping readers consistent during an object-based table update.

Question 47

Data scientists must join transactions across days without seeing raw customer account numbers, and an approved re-identification service must retain controlled mapping access. Which transformation technique fits?

  1. Redshift dynamic data masking
  2. One-way hashing of identifiers
  3. Amazon Macie sensitive data discovery
  4. Deterministic PII tokenization
  5. Data anonymization aggregation

Explanation

Correct Answer

D:

Tokenization replaces sensitive identifiers with controlled surrogate values and protects the mapping separately; in this scenario, tokenize the stable identifier and keep re-identification privileges restricted so analytical joins can use a consistent surrogate.

Incorrect Answers

A:

Masking policies can return altered column representations to database users based on defined privileges; this helps with obscuring sensitive fields at query time for nonprivileged roles, not linking customer events without exposing raw identifiers.

B:

A cryptographic hash maps inputs to fixed outputs without keeping an ordinary reversible mapping; although useful for comparing known identifiers without storing plaintext values, it leaves the need for linking customer events without exposing raw identifiers unmet.

C:

Macie can examine supported S3 data for sensitive information and report discovery findings; it would target detecting unexpected sensitive content in S3 objects, while the question is about linking customer events without exposing raw identifiers.

E:

Aggregation can report measurements across groups while excluding direct person-level rows from published outputs; the current need is for linking customer events without exposing raw identifiers, not publishing population statistics without individual record detail.

Question 48

A Glue job reaches the database host but fails during driver initialization for a proprietary SQL engine. What should be checked?

  1. Glue VPC routing and DNS
  2. JDBC driver compatibility
  3. Secrets-based JDBC authentication
  4. ODBC connector configuration

Explanation

Correct Answer

B:

JDBC requires a driver compatible with the source database and its supported protocol; in this scenario, supply a supported JDBC driver and connection properties that match the source protocol and Glue runtime.

Incorrect Answers

A:

Glue network connections need paths and DNS resolution to reach databases inside a VPC; this helps with resolving a JDBC timeout to an internal data source, not making a Glue job connect to a vendor database.

C:

ETL connections can retrieve database credentials from a secret rather than embedding plaintext; it would target separating sensitive credentials from ETL source code, while the question is about making a Glue job connect to a vendor database.

D:

ODBC connection settings rely on a driver and data-source configuration compatible with the application runtime; its benefit is integrating a tool that expects an ODBC data source; that is different from making a Glue job connect to a vendor database.

Question 49

A Kinesis consumer can be offline for three days and must resume at its last safely processed record without skipping events. Which TWO design controls cover both requirements? Choose TWO.

  1. Kinesis retention adjustment
  2. Managed Service for Apache Flink keyed windows
  3. Stateless transformation
  4. Durable consumer checkpoints
  5. Idempotent output writes

Explanation

Correct Answers

A:

Stream retention controls how long published records remain available for replay in Kinesis Data Streams; in this scenario, increase the retention period beyond the recovery objective; a consumer cannot replay records that have already aged out.

D:

Checkpoints track a stream application's committed read position separately from the stream itself; for the additional requirement of resuming from the last successfully processed record, commit the last safe stream offset or sequence position alongside durable processing semantics, then resume from that boundary.

Incorrect Answers

B:

Managed Service for Apache Flink can maintain keyed window aggregates with checkpoint-based state recovery across streaming records; this helps with running recoverable per-customer rolling-window totals in managed Flink, not extending the recovery window for an offline consumer.

C:

Stateless processors derive each output solely from their current input and need no cross-event accumulated state; although useful for applying independent per-record normalization rules, it leaves the need for extending the recovery window for an offline consumer unmet.

E:

Idempotent writes make replaying the same logical input safe without duplicating downstream effects; its benefit is preventing duplicate target rows after event redelivery; that is different from extending the recovery window for an offline consumer.

Question 50

An analyst can connect to Redshift but receives permission denied when querying a finance table. Which authorization layer should change?

  1. Redshift dynamic data masking
  2. Redshift database role
  3. Redshift GRANT on schema objects
  4. Redshift row-level security policy

Explanation

Correct Answer

C:

Redshift SQL grants determine which database principals can access schemas, tables and other objects; in this scenario, grant the necessary schema usage and table SELECT privileges to the database role rather than expanding AWS control-plane permissions.

Incorrect Answers

A:

Redshift dynamic data masking can change the presentation of sensitive columns according to database policy; this helps with hiding customer email values from ordinary warehouse readers, not giving a reporting role SELECT on one curated table.

B:

A Redshift database role groups database privileges that can be granted to multiple database users; although useful for administering common warehouse table permissions centrally, it leaves the need for giving a reporting role SELECT on one curated table unmet.

D:

Redshift row-level security can restrict table rows returned for principals under applied policies; its benefit is limiting warehouse query results to a user's assigned tenant; that is different from giving a reporting role SELECT on one curated table.

AWS Certified Data Engineer - Associate DEA-C01 Practice Tests