Data Operations and Support
Free AWS Certified Data Engineer - Associate practice — 6 questions on Data Operations and Support, with explanations. No sign-up. Full 12-question mixed test →
A Step Functions state machine orchestrates a Glue ETL job that loads records into a Redshift table. When the Glue job step fails with a transient service error, the built-in Retry policy re-invokes the job, which occasionally causes duplicate rows to appear in the target table because the previous partial run had already written some data. Which change BEST ensures idempotency while keeping the retry benefit?
A staging table plus a keyed MERGE/UPSERT (or DELETE-then-INSERT) pattern guarantees that re-running the same job after a retry produces the same end state, eliminating duplicates regardless of how many times the step executes.
A Lambda function consuming from a Kinesis Data Stream has its timeout set to the default 3 seconds. During peak traffic, CloudWatch Logs show frequent "Task timed out after 3.00 seconds" errors, and the stream's GetRecords.IteratorAgeMilliseconds metric climbs into the hours. Actual average processing time per batch under normal load is under 1 second, but batches occasionally take up to 45 seconds during traffic spikes. Which change BEST resolves this issue?
The root cause is a Lambda timeout configured lower than actual worst-case processing time. Raising the timeout to match real workload behavior (and adjusting batch size if needed) stops the timeout errors, which in turn lets IteratorAge recover as records are successfully processed instead of retried.
You are writing an AWS Glue Data Quality ruleset (DQDL) to enforce two requirements on the "order_total" column: every non-null value must be greater than 0, and no more than 2% of rows may contain a null value in that column. Which ruleset correctly expresses BOTH requirements?
ColumnValues "order_total" > 0 checks the value constraint on non-null entries, and Completeness "order_total" > 0.98 expresses the requirement that at least 98% of rows have a non-null value (i.e., no more than 2% null) — both are valid DQDL rule types used correctly.
A Redshift cluster uses manual WLM with a single query queue configured for concurrency level 5. During normal operation queries run without delay, but during month-end reporting the queue frequently backs up and users report long wait times. Which change BEST resolves the spike-driven queuing while keeping ongoing cost under control?
Concurrency Scaling automatically adds transient burst clusters to absorb short-lived spikes in concurrent read queries, and Automatic WLM dynamically manages memory and concurrency — together this resolves month-end queuing without paying for extra capacity the rest of the month.
An Athena workgroup runs a daily analytical query against a 5 TB dataset stored as uncompressed CSV in S3, scanning the entire dataset each time at a cost of roughly $25 per query. Analysts consistently filter results by "region" and "order_date". Which action would MOST reduce ongoing query costs with minimal continued maintenance?
Converting to columnar, compressed Parquet and partitioning by the columns analysts actually filter on lets Athena prune partitions and read far fewer bytes per query, producing a large, durable reduction in cost-per-query for this recurring workload.
A long-running EMR cluster performs nightly Spark ETL jobs lasting about 4 hours. Task nodes represent roughly 70% of the cluster's hourly cost and currently run On-Demand. The Spark jobs use checkpointing and can tolerate node interruption without data loss. Which configuration change BEST reduces cost while maintaining acceptable reliability?
Task nodes hold no HDFS data and are the largest cost driver, making them ideal candidates for Spot. A diversified instance fleet across multiple instance types and AZs reduces the risk of simultaneous interruption, keeping core/master On-Demand preserves cluster stability, and Managed Scaling automatically right-sizes capacity — this is the AWS-recommended pattern for cost-optimized, interruption-tolerant transient workloads.
Ready for the real thing?
The full course: two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed explanations.
$109.99 $34.99 with code FREETEST33 — valid through September 14.
Get my $34.99 deal →