AWS Certified Data Engineer - Associate (DEA-C01) practice questions

Question 1 · Data Ingestion and Transformation

Willowbrook Freight streams truck telemetry into a Kinesis data stream with 8 shards, using the tractor unit's depot code as the partition key. Three of the depots generate roughly 70 percent of all events, and the writing application receives frequent ProvisionedThroughputExceededException errors even though total stream ingest is far below the aggregate 8 MB/s limit. What change should the data engineer make first?

1. A.Change the partition key to a higher-cardinality value, such as the vehicle ID with a random suffix
2. B.Switch the stream to extended data retention so throttled records can be replayed from an earlier point
3. C.Increase the producer's retry count and add exponential backoff so throttled records eventually succeed
4. D.Move the consumer applications to enhanced fan-out so each consumer receives its own read throughput

Question 2 · Data Store Management

Aldersgate Publishing stores 80 TB of archived manuscript scans in S3 Standard. Access is genuinely unpredictable: a title may be untouched for two years and then read daily for a month when a film adaptation is announced. Retrieval must stay immediate, and the team does not want to build access-pattern tooling. Which storage approach fits best?

1. A.A lifecycle rule to S3 Standard-IA at 30 days combined with a custom access-tracking Lambda
2. B.A lifecycle rule transitioning objects to S3 Glacier Deep Archive after 90 days of age
3. C.S3 One Zone-Infrequent Access, accepting single-AZ durability in exchange for a lower rate
4. D.S3 Intelligent-Tiering, which moves objects between access tiers based on observed usage

Question 3 · Data Operations and Support

The on-call team at Larkspur Payments wants to be paged when a Kinesis-based pipeline falls behind, defined as consumers being more than 15 minutes behind the tip of the stream. Which CloudWatch alarm most directly encodes that condition?

1. A.An alarm on the consumer Lambda's Duration metric exceeding its p99 baseline by 50 percent
2. B.An alarm on GetRecords.IteratorAgeMilliseconds exceeding 900000 for two consecutive periods
3. C.An alarm on IncomingRecords dropping below the seven-day average for the same hour
4. D.An alarm on WriteProvisionedThroughputExceeded exceeding zero for two consecutive periods

Question 4 · Data Security and Governance

Wexley Health governs a lake with 300 tables across marketing, clinical, and finance domains. New analysts should receive access by data classification rather than by naming individual tables, and classifications change as tables are reclassified. Which Lake Formation approach scales best?

1. A.Create one Lake Formation database per classification and physically copy tables into the matching database
2. B.Grant named-resource permissions on each table and re-run a script whenever a classification changes
3. C.Define LF-Tags such as classification and domain, attach them to resources, and grant permissions on tag expressions
4. D.Create one IAM role per table with an inline policy listing the table's S3 prefix and catalog ARN

Question 5 · Data Ingestion and Transformation

A nightly AWS Glue Spark job at Corvid Analytics reads a growing S3 prefix of order files and writes them to a curated zone. Each run currently reprocesses every file ever landed, so runtime has grown from 6 minutes to 90 minutes over the year. The team wants each run to process only files that arrived since the previous successful run, without maintaining its own state table.

1. A.Set the job's worker type to G.2X and raise the number of workers to shorten the scan
2. B.Configure the job to run on a shorter schedule so each execution handles a smaller batch
3. C.Add a Glue crawler that runs before the job so the catalog reflects only new partitions
4. D.Enable job bookmarks on the Glue job so it tracks previously processed S3 objects between runs

Question 6 · Data Store Management

In a Redshift warehouse at Pemberly Outfitters, a 4 billion row `fact_sales` table is joined to a 900 row `dim_channel` table in nearly every dashboard query. Query plans show a large DS_BCAST_INNER step and high network cost. Which distribution setup addresses this?

1. A.Set fact_sales to DISTSTYLE ALL and dim_channel to DISTSTYLE KEY on channel_id
2. B.Set both tables to DISTSTYLE EVEN so rows spread uniformly across all slices
3. C.Set both tables to DISTSTYLE KEY on channel_id so matching rows share a slice
4. D.Set dim_channel to DISTSTYLE ALL and keep fact_sales distributed on a high-cardinality key

Question 7 · Data Operations and Support

A critical Glue job at Wrenfield Chemicals occasionally fails when a shared JDBC source is briefly unavailable. The team wants the job retried automatically a couple of times, and if it still fails, a message posted to an SNS topic that pages the on-call engineer. Which configuration achieves this with managed services?

1. A.Set maximum retries to 2 and add an EventBridge rule on Glue FAILED job state change targeting the SNS topic
2. B.Enable continuous logging on the job and create a CloudWatch Logs metric filter on the word Exception
3. C.Schedule a second Glue job 30 minutes later that checks the first job's run status and notifies on failure
4. D.Wrap each JDBC read in a Python try/except loop inside the job script and call SNS Publish in the handler

Question 8 · Data Security and Governance

A Glue job in the analytics account at Fernhollow Group must read objects from an S3 bucket owned by the ingestion account. The bucket is encrypted with SSE-S3 and the analytics account owns the Glue job role. Which configuration grants the required cross-account read?

1. A.An S3 Access Grants instance in the analytics account mapped to the ingestion bucket prefix
2. B.An IAM policy on the Glue role in the analytics account allowing s3:GetObject on the bucket ARN
3. C.A bucket policy in the ingestion account allowing the analytics account root principal, with no role-side policy needed
4. D.A bucket policy allowing the analytics Glue role, plus an IAM policy on that role allowing the reads