Question 130

A workload needs a custom Spark distribution configuration and side libraries not available in the managed Glue runtime. Which environment fits?

  1. Lambda per-record enrichment
  2. Redshift SQL transformations
  3. Amazon EMR customized frameworks
  4. AWS Glue Spark ETL

Explanation

Correct Answer

C:

Amazon EMR provides controlled Hadoop and Spark clusters with configurable frameworks and libraries; in this scenario, use EMR to select appropriate cluster software and instance resources, accepting greater configuration and operations responsibility.

Incorrect Answers

A:

Lambda is suited to short event-driven transformations within its runtime and resource limits; although useful for enriching individual small records after an event, it leaves the need for running specialized large-scale distributed transformations unmet.

B:

Redshift SQL operations can transform warehouse-resident data without exporting it to a separate engine; it would target reshaping existing warehouse tables in place, while the question is about running specialized large-scale distributed transformations.

D:

AWS Glue provides managed Spark-based ETL jobs and integration with the Data Catalog; the current need is for running specialized large-scale distributed transformations, not running recurring serverless-style Spark transformations.

Question 131

Analysts want to see null frequencies and value distributions while designing a repeatable cleaning recipe. Which tool fits?

  1. DataBrew visual profiling
  2. QuickSight SPICE dataset
  3. Athena result visualization downstream
  4. Amazon QuickSight dashboard

Explanation

Correct Answer

A:

DataBrew's visual interface can inspect column distributions and transformations before publishing cleaned output; in this scenario, inspect profiles and samples visually, then codify approved transformations in the preparation recipe.

Incorrect Answers

B:

SPICE provides in-memory accelerated storage for supported QuickSight analytic datasets; it would target speeding repeated dashboard interactions on imported results, while the question is about exploring raw field problems interactively.

C:

Athena produces query result datasets that can be connected to separate reporting and visualization tools; its benefit is preparing lake queries for a dashboard dataset; that is different from exploring raw field problems interactively.

D:

QuickSight builds governed analytic dashboards and visualizations from configured data sources; the current need is for exploring raw field problems interactively, not publishing metrics to business users through dashboards.

Question 132

An MWAA DAG starts on time but its S3 extract operator fails with a Python stack trace. Where should engineers look first?

  1. Step Functions failed state details
  2. Glue job run metrics
  3. MWAA worker task log
  4. MWAA scheduler log

Explanation

Correct Answer

C:

Worker task logs record errors raised while an Airflow operator executes; in this scenario, read worker/task logs for that DAG run and correlate any IAM or connection errors with the operator configuration.

Incorrect Answers

A:

Execution details identify state-level errors and task transitions in a Step Functions workflow; the current need is for diagnosing a failed task with running schedulers, not locating which branch and integration caused a state-machine failure.

B:

Glue job run data includes execution duration, failures and relevant service metrics for job diagnosis; this can support determining whether an ETL problem is resource-related, but would not address the need for diagnosing a failed task with running schedulers.

D:

Scheduler logs show why Apache Airflow DAG runs may not be getting scheduled; although useful for diagnosing DAG runs that never enter the queue, it leaves the need for diagnosing a failed task with running schedulers unmet.

Question 133

A code change accidentally converts blank currency fields into zeros instead of errors. Which test type should detect the change?

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

Explanation

Correct Answer

D:

Unit tests validate transformation functions against representative inputs and expected results; in this scenario, add deterministic transformation tests for null, malformed and boundary inputs in the development pipeline.

Incorrect Answers

A:

Integration tests exercise real interfaces between components in an isolated environment; this helps with validating source-to-target contracts before release, not catching mapping regressions before deployment.

B:

Versioned packages and immutable images let a pipeline redeploy the exact artifact tested previously; although useful for preventing drift between staging and production ETL code, it leaves the need for catching mapping regressions before deployment unmet.

C:

Git branches and pull requests support code review before changes merge into a shared baseline; it would target reviewing ETL changes before production release, while the question is about catching mapping regressions before deployment.

Question 134

One malformed record causes repeated Lambda retries of an otherwise valid Kinesis batch. Which response capability narrows retries?

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

Explanation

Correct Answer

D:

Lambda can report individual failed record IDs for supported stream batch processing; in this scenario, implement partial-batch response behavior with the relevant event-source configuration, so acknowledged records do not need reprocessing.

Incorrect Answers

A:

IteratorAge measures how far behind a stream-reading Lambda consumer is in processing records; it would target diagnosing lag in a Lambda stream consumer, while the question is about retrying only failed records in a stream batch.

B:

The parallelization factor raises concurrent Lambda batches processed per Kinesis shard while preserving partition-key ordering; its benefit is increasing consumer throughput on busy shards; that is different from retrying only failed records in a stream batch.

C:

A Lambda event source mapping reads Kinesis records, checkpoints progress and invokes the function in batches; the current need is for retrying only failed records in a stream batch, not connecting a Lambda processor to a Kinesis shard.

Question 135

A small internal ETL job needs an encrypted configuration value but no managed automatic rotation workflow. What service is suitable?

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

Explanation

Correct Answer

D:

Systems Manager Parameter Store can hold encrypted configuration values as SecureString parameters; in this scenario, store the value as SecureString with appropriate KMS and parameter permissions, retrieving it securely at runtime.

Incorrect Answers

A:

KMS key policies can evaluate encryption-context conditions to narrow permitted cryptographic use; this helps with scoping decrypt permission to the intended encrypted dataset, not storing a rarely changed encrypted ETL token.

B:

IAM role sessions provide automatically rotated temporary credentials to supported AWS workloads; although useful for authenticating an AWS SDK call without storing an AWS secret, it leaves the need for storing a rarely changed encrypted ETL token unmet.

C:

Secrets Manager stores and retrieves sensitive credentials and supports managed rotation patterns; it would target protecting a rotating production database password, while the question is about storing a rarely changed encrypted ETL token.

Question 136

A containerized worker processes SQS messages reliably but queue age rises during daytime demand. What can add processing capacity automatically?

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

Explanation

Correct Answer

D:

ECS Service Auto Scaling changes the desired count of running service tasks based on configured metrics; in this scenario, scale the service task count from queue-backed demand or suitable utilization, while considering downstream quota limits.

Incorrect Answers

A:

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 adding parallel workers when a queue backlog rises.

B:

Kubernetes HPA adjusts replica counts of a target workload using metrics and configured policies; the current need is for adding parallel workers when a queue backlog rises, not scaling Kubernetes transformation pods based on utilization.

C:

An ECS task definition reserves CPU and memory for its container workloads; this can support giving each transformation container enough compute and RAM, but would not address the need for adding parallel workers when a queue backlog rises.

Question 137

A pipeline sends medical narratives for summarization but must exclude identifying fields from inputs. What should happen upstream?

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

Explanation

Correct Answer

D:

Sensitive fields should be removed or minimized before they are sent to a model where policy requires it; in this scenario, apply deterministic detection and masking appropriate to the data policy before generating prompts, then audit what is transmitted.

Incorrect Answers

A:

Model responses require parsing and validation against an expected structured contract before ingestion; although useful for preventing unstructured model responses from breaking ETL, it leaves the need for keeping customer identifiers out of model prompts unmet.

B:

A review queue can route uncertain automated extractions to humans instead of blindly publishing them; it would target avoiding unchecked facts in ambiguous document extraction, while the question is about keeping customer identifiers out of model prompts.

C:

Amazon Bedrock can invoke foundation models to classify or summarize unstructured text in a pipeline; its benefit is deriving categories from free-form support notes; that is different from keeping customer identifiers out of model prompts.

Question 138

An API always fetches orders by tenant and recent date but the proposed DynamoDB design requires full scans and client-side joins. What redesign helps?

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

Explanation

Correct Answer

B:

DynamoDB tables are designed around known partition/sort key access patterns rather than normalized joins; in this scenario, co-locate related items or add indexes aligned with key queries and monitor partition hot spots.

Incorrect Answers

A:

A star schema links facts to dimensions for analytic joins and common reporting aggregations; it would target modeling sales facts for BI measures by customer and product, while the question is about optimizing predictable NoSQL key queries.

C:

Surrogate keys can stabilize warehouse joins when operational business identifiers change; the current need is for optimizing predictable NoSQL key queries, not preserving historical dimension references through source ID changes.

D:

Lake Formation grants operate on registered data-lake catalog resources and can enforce fine-grained policies; this can support controlling which analysts can read governed S3 tables, but would not address the need for optimizing predictable NoSQL key queries.

Question 139

A curated dataset needs a new optional attribute while old records remain queryable by downstream consumers. Which table capability helps?

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

Explanation

Correct Answer

A:

Iceberg supports schema evolution using field identities rather than relying only on physical column order; in this scenario, apply supported Iceberg schema evolution and validate engine compatibility before changing downstream contracts.

Incorrect Answers

B:

Amazon S3 Tables provides purpose-built table buckets for Apache Iceberg with automatic file compaction and snapshot maintenance; the current need is for adding columns without rewriting every historical file, not hosting Iceberg tables with managed storage-level maintenance in Amazon S3.

C:

Iceberg can evolve partition specifications without rewriting historical data immediately into one fixed layout; this can support changing partition strategy as data volumes grow, but would not address the need for adding columns without rewriting every historical file.

D:

Iceberg tracks table snapshots and metadata so readers can use consistent committed table states; this helps with keeping readers consistent during an object-based table update, not adding columns without rewriting every historical file.

Question 140

Analysts need to detect repeated email values after normalization, and policy forbids recovering the original addresses from outputs. What approach might fit?

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

Explanation

Correct Answer

B:

A cryptographic hash maps inputs to fixed outputs without keeping an ordinary reversible mapping; in this scenario, apply a suitable one-way, keyed or salted hashing design as required; consider low-entropy input guessing risks before relying on hashes alone.

Incorrect Answers

A:

Masking policies can return altered column representations to database users based on defined privileges; its benefit is obscuring sensitive fields at query time for nonprivileged roles; that is different from comparing known identifiers without storing plaintext values.

C:

Macie can examine supported S3 data for sensitive information and report discovery findings; this can support detecting unexpected sensitive content in S3 objects, but would not address the need for comparing known identifiers without storing plaintext values.

D:

Aggregation can report measurements across groups while excluding direct person-level rows from published outputs; this helps with publishing population statistics without individual record detail, not comparing known identifiers without storing plaintext values.

E:

Tokenization replaces sensitive identifiers with controlled surrogate values and protects the mapping separately; although useful for linking customer events without exposing raw identifiers, it leaves the need for comparing known identifiers without storing plaintext values unmet.

AWS Certified Data Engineer - Associate DEA-C01 Practice Tests