DEX-401 practice questions

Q: Is this DEX-401 practice test free?

Yes. All 15 DEX-401 practice questions on this page are free, with instant scoring and a written explanation for each answer. You do not need an account and there is no time limit.

Q: How much does the real DEX-401 exam cost?

The official price is $200. Cyber VK sells the same official DEX-401 voucher for $179.99 — a saving of $20.01. The voucher is identical to one bought direct and is emailed within minutes of payment.

Q: Are these questions the same as the real DEX-401 exam?

No. These are original practice questions written to match the published exam objectives and difficulty. They are not real exam questions — sharing or reproducing live exam content breaches every certification body's candidate agreement and can void your certification.

Q: How should I use this practice test?

Use it as a readiness check rather than a study tool. Work through the questions without notes, read the explanation for every answer including the ones you got right, and treat any domain where you miss more than a third as the area to revise before booking your exam.

Question 1

Which Delta Lake command should a data engineer run to compact many small files in a table into fewer, larger files?

Accepted Answer

OPTIMIZE performs bin-packing compaction, rewriting small files into larger ones to improve read performance; VACUUM only deletes unreferenced files and does not compact data.

Question 2

A Delta table has a default retention threshold. A data engineer runs VACUUM on it with no RETAIN clause. What is the effect on time travel?

Accepted Answer

VACUUM defaults to a 7-day retention window and removes only unreferenced data files older than that, breaking time travel to versions that depended on those files while leaving newer versions intact.

Question 3

Two concurrent jobs write to the same Delta table: Job A deletes rows where region = 'EU' while Job B appends new rows for region = 'US'. Both use the default isolation level. What is the most likely outcome?

Accepted Answer

Delta uses optimistic concurrency with file-level conflict detection, so an append that adds new files does not conflict with a delete that rewrites unrelated files; conflicts arise only when the transactions read or modify the same files.

Question 4

In Structured Streaming, which output mode writes only the rows that were added to the result table since the last trigger?

Accepted Answer

Append mode emits only newly added rows that will not change again, which is why it requires a watermark for aggregations; Update emits changed rows and Complete rewrites the entire result table.

Question 5

A streaming job aggregates events by a 10-minute tumbling window and uses withWatermark("event_time", "15 minutes"). What happens to an event that arrives 20 minutes after its window's watermark has passed?

Accepted Answer

The watermark defines how long state for a window is retained; once the watermark passes, the window's state is evicted and later events for it are dropped rather than updating the aggregate.

Question 6

A production stream reads from a Delta table and writes to another Delta table using a checkpoint. The engineer changes the query from a simple projection to include a stateful aggregation and restarts the job with the same checkpoint location. What is the expected result?

Accepted Answer

Checkpoints store offsets plus state store metadata tied to the query's plan, so introducing stateful operations to an existing checkpoint is an unsupported query change and the restart fails; a new checkpoint location is required.

Question 7

In the medallion architecture, which layer typically holds raw ingested data in its original form with minimal transformation?

Accepted Answer

The Bronze layer captures raw source data as-is for replay and auditability, while Silver applies cleansing and conformance and Gold serves business-level aggregates.

Question 8

A dimension table must preserve full history of attribute changes so analysts can query customer state as of any past date. Which modeling approach should be implemented in the Silver layer?

Accepted Answer

Type 2 slowly changing dimensions preserve history by adding a new row per change with validity ranges; time travel is bounded by VACUUM retention and cannot serve as a durable history mechanism.

Question 9

A 4 TB Delta table is queried almost exclusively by filters on a high-cardinality customer_id column plus occasional filters on event_date. Partitioning by customer_id has produced hundreds of thousands of tiny files. What is the best remediation?

Accepted Answer

High-cardinality columns are poor partition keys because they create small-file explosion; partitioning on a low-cardinality date column and clustering on customer_id gives effective data skipping without fragmenting the table.

Question 10

Which Unity Catalog object sits between the metastore and schemas in the three-level namespace?

Accepted Answer

Unity Catalog uses catalog.schema.table, so the catalog is the top-level container inside a metastore that holds schemas.

Question 11

A team must expose a sales table so that analysts see only rows for their own region, without maintaining separate copies of the table. What is the appropriate Unity Catalog feature?

Accepted Answer

Dynamic views apply row-level filtering at query time based on the requesting principal's identity or group membership, so one table can serve all regions without duplication.

Question 12

An engineer creates an external table in Unity Catalog pointing at a cloud storage path, then drops the table. Later the same path is registered as a managed table. Which statement correctly describes the data lifecycle?

Accepted Answer

DROP on an external table removes only the metadata and leaves files intact, whereas managed tables live in the metastore-, catalog-, or schema-level managed storage location rather than an arbitrary user-specified path.

Question 13

In a Databricks multi-task job, how does an engineer ensure task B runs only after task A completes successfully?

Accepted Answer

Multi-task jobs use an explicit DAG where declaring task A as a dependency makes task B wait for A's success; time-based offsets do not guarantee ordering or success.

Question 14

A nightly job intermittently fails due to a transient cloud storage error. The team wants automatic recovery and notification only when recovery ultimately fails. What configuration best achieves this?

Accepted Answer

A bounded retry policy handles transient errors automatically, and failure-only notifications fire after retries are exhausted, so the team is alerted only for genuine failures.

Question 15

A batch job reading a Delta source must guarantee that reruns after a partial failure do not duplicate rows in the target table. The source provides a stable business key. Which approach best delivers idempotency?

Accepted Answer

MERGE keyed on a stable business key makes the write idempotent because previously inserted keys match and are skipped on rerun; append duplicates and full overwrite discards unrelated history.

DEX-401 — DEX-401 Databricks Certified Data Engineer Professional Exam Voucher | Buy Online Practice Test

Free Practice Test

DEX-401 practice test — common questions