Data Engineer Associate · 6% of the exam

Databricks Intelligence Platform: free practice questions

5 sample questions from our 66-question bank for this domain — answers and explanations included. These are the same scenario-based style as the real Databricks exam.

1. A data engineer is setting up a Databricks SQL warehouse for a business intelligence team. The team's workload consists of unpredictable bursts of concurrent queries from 20 analysts throughout the business day, with very low activity at night. Which Databricks SQL warehouse configuration would BEST balance performance during peak hours and cost efficiency during off-peak hours?

  • A. A Classic SQL warehouse with auto-stop disabled to ensure analysts never experience cold-start delays.
  • B. A serverless SQL warehouse with auto-stop enabled, allowing Databricks to automatically scale and terminate compute resources based on query demand.✓ Correct
  • C. A Pro SQL warehouse with a fixed cluster size of 4X-Large to handle all concurrent queries at maximum throughput at all times.
  • D. A Classic SQL warehouse sized at X-Small, as smaller warehouse sizes always result in lower costs regardless of query concurrency.
Explanation

Option B is correct: a serverless SQL warehouse automatically scales compute up during peak query bursts and down (including full termination via auto-stop) during idle periods, providing both high performance under load and minimal cost when inactive. This is ideal for unpredictable, bursty workloads. Option A is wrong: disabling auto-stop means the warehouse runs (and incurs cost) 24/7, including overnight when there is low activity — this is not cost-efficient. Option C is wrong: a fixed 4X-Large warehouse is over-provisioned and extremely expensive to run continuously; it does not scale down during off-peak hours. Option D is wrong: a fixed X-Small warehouse will become a bottleneck with 20 concurrent analysts, causing query queuing and poor performance during peak hours regardless of its lower cost.

2. A data engineer runs the following command: CONVERT TO DELTA parquet.`abfss://container@storage.dfs.core.windows.net/raw/events`; What is the result of this operation, and what is an important consideration the engineer should be aware of?

  • A. The command creates a new copy of the data in Delta format at a different location; the original Parquet files are preserved and unmodified.
  • B. The command converts the existing Parquet table in-place to Delta Lake format by creating a _delta_log transaction log on top of the existing files, without rewriting the data files; however, partitioned tables may require CONVERT TO DELTA with the PARTITIONED BY clause.✓ Correct
  • C. The command rewrites all existing Parquet files into a new optimized Delta format, which may require significant additional cloud storage during the conversion process.
  • D. The command permanently deletes the original Parquet files and replaces them with Delta-formatted Snappy-compressed ORC files.
Explanation

CONVERT TO DELTA performs an in-place conversion: it creates a _delta_log directory on top of the existing Parquet files without rewriting them, making the conversion fast and storage-efficient. For partitioned Parquet tables, the engineer must include the PARTITIONED BY clause to correctly register partition schema. Option A is wrong; no copy is made — the conversion is in-place. Option C is incorrect; data files are not rewritten during conversion. Option D is completely wrong; CONVERT TO DELTA does not delete files or convert to ORC format.

3. A data engineer is explaining the Databricks platform architecture to a new hire. She states that user-created data, such as Delta Lake table files, always resides in the customer's own cloud storage account. Which architectural concept does this statement describe?

  • A. Control plane isolation
  • B. Data plane residency✓ Correct
  • C. Cluster autoscaling boundary
  • D. Metastore federation
Explanation

Option B is correct: in Databricks architecture, the data plane is the customer's own cloud account where compute resources (clusters) run and where actual data files are stored in the customer's cloud object storage. The data never leaves the customer's account. Option A is wrong: the control plane refers to the Databricks-managed infrastructure (web application, job scheduler, cluster manager, etc.) that lives in Databricks' own cloud account, not where data resides. Option C is wrong: autoscaling boundaries govern how cluster node counts scale up or down and are unrelated to where data is stored. Option D is wrong: metastore federation is a Unity Catalog concept related to cross-workspace metadata sharing, not data residency.

4. A data engineer is troubleshooting an issue where a Delta Lake table's DESCRIBE HISTORY command shows 2,000 versions, but the team only requires 30 days of time travel history. Running VACUUM has not reduced the number of entries in DESCRIBE HISTORY. Why does VACUUM not remove transaction log history entries?

  • A. VACUUM only works on external tables; managed table history requires the PURGE HISTORY command.
  • B. VACUUM removes unreferenced data files (Parquet files) from the storage location but does NOT remove entries from the transaction log or reduce the output of DESCRIBE HISTORY. Transaction log cleanup is handled separately.✓ Correct
  • C. VACUUM removes both data files and transaction log entries older than the retention threshold, but DESCRIBE HISTORY uses a cached view that requires a cluster restart to reflect the changes.
  • D. VACUUM requires the DELTA.logRetentionDuration property to be set to 0 days before it will clean transaction log entries.
Explanation

Correct: VACUUM deletes unreferenced Parquet data files from the table's storage directory that are older than the retention threshold (default 7 days). It does NOT touch the transaction log (_delta_log). Transaction log entry cleanup is controlled by the table property delta.logRetentionDuration (default 30 days) and is handled automatically by Delta Lake during subsequent write operations—not by VACUUM. Wrong: (A) VACUUM works on both managed and external Delta tables; there is no PURGE HISTORY command. (C) VACUUM does not remove transaction log entries at all, regardless of retention settings; there is no cached view requiring a restart. (D) While delta.logRetentionDuration controls log cleanup, this is separate from VACUUM; setting it to 0 would not cause VACUUM to clean the log—the log cleanup is triggered by write operations, not VACUUM.

5. A data engineer is reviewing Delta Lake internals. She wants to understand what information is stored in the Delta Lake transaction log (also known as the _delta_log directory). Which of the following BEST describes the contents of the transaction log?

  • A. The transaction log stores the actual data rows as compressed Parquet records, organized by transaction order.
  • B. The transaction log contains JSON and Parquet checkpoint files that record every operation (add/remove file, schema change, metadata update) applied to the table, enabling reconstruction of any table version.✓ Correct
  • C. The transaction log is a binary index file that maps row-level primary keys to the physical Parquet file locations for fast point lookups.
  • D. The transaction log stores only schema definitions and table statistics; the history of data operations is maintained separately in the Hive metastore.
Explanation

Correct: The _delta_log directory contains a sequence of JSON commit files (one per transaction) and periodic Parquet checkpoint files. Each JSON entry records the operation performed—such as file additions, file removals, schema changes, and metadata updates—allowing Delta Lake to reconstruct the exact state of the table at any version. Wrong: (A) Actual data rows are stored in Parquet files in the table directory, not in the transaction log. (C) The transaction log is not a row-level index; it tracks file-level operations. Row-level statistics may be embedded in log entries for data skipping, but the log is not a primary-key index. (D) The transaction log is self-contained in the table directory; it does not rely on the Hive metastore for operation history, which is a key advantage of the Delta format.

61 more questions in this domain

Practice the full bank with instant grading, flashcards, and a timed mock exam.

Start practicing free