Databricks Certified Data Engineer Associate

Last checked October 22, 2026 against the official Databricks exam page.

1. Question 1Data Transformation and Modeling

   A silver Delta table must be refreshed from a daily batch of changed customer rows. Matching customer_id rows should be updated and new ones inserted. Which statement does this?

   1. AINSERT OVERWRITE target SELECT * FROM updates, which replaces every existing customer row in the target
   2. BMERGE INTO target USING updates ON customer_id WHEN MATCHED THEN UPDATE WHEN NOT MATCHED THEN INSERT
   3. CCREATE OR REPLACE TABLE target AS SELECT * FROM updates, which rebuilds the table from the batch alone
   4. DINSERT INTO target SELECT * FROM updates, which appends every batch row to the existing table
   Show answer and explanationHide answer

   Correct answer: B

   MERGE INTO applies updates to matched keys and inserts unmatched ones in a single atomic transaction, which is the upsert the requirement describes. INSERT OVERWRITE drops the existing customers that are not in the batch, and a plain INSERT adds duplicates of rows already present. CREATE OR REPLACE rebuilds the table from the batch alone and loses the rest.

2. Question 2Governance and Security

   A group needs to read one table, sales.orders.transactions, in Unity Catalog, but must not be able to create or change tables. Which grants are correct?

   1. AUSE CATALOG and USE SCHEMA on the parents, plus SELECT on the table
   2. BGRANT MODIFY on table sales.orders.transactions to the group, so the group can read the table and also write to it
   3. CGRANT ALL PRIVILEGES on catalog sales to the group, so that access to all its schemas is covered for future work
   4. DGRANT SELECT on catalog sales to the group, which is inherited by every table and view below it in the catalog
   Show answer and explanationHide answer

   Correct answer: C

   Reading a table in Unity Catalog needs USE CATALOG and USE SCHEMA on its parents plus SELECT on the table, and nothing more. ALL PRIVILEGES on the catalog gives create and modify rights the group should not have. MODIFY allows writes, which the requirement excludes.

3. Question 3Data Ingestion and Loading

   A team must ingest JSON files that land continuously in cloud object storage. They need schema inference, tracking of files already processed, and no rescans of old files. Which feature fits best?

   1. AAuto Loader using the cloudFiles source with a schema location
   2. BA nightly spark.read of the full folder with overwrite mode on the output
   3. CA Python task that lists the folder with dbutils.fs.ls and appends each new file
   4. DA Delta MERGE keyed on file name, built from a manually maintained file list
   Show answer and explanationHide answer

   Correct answer: A

   Auto Loader uses the cloudFiles source to discover new files incrementally, stores the inferred schema in the schema location, and records which files are already processed. A nightly full read rescans everything, and listing folders by hand rebuilds bookkeeping that Auto Loader already does. A MERGE on a manually maintained list is fragile and pushes that bookkeeping onto the engineer.

4. Question 4Working with Lakeflow Jobs

   A Lakeflow Job has three tasks: ingest, transform and publish. Transform must run only after ingest succeeds, and each run should use compute created for that run and removed afterwards. Which configuration is correct?

   1. ASchedule each task as its own notebook at a fixed time with no dependencies
   2. BDefine task dependencies and attach a job cluster to the tasks
   3. CPut all three steps in one notebook cell and rely on a single retry setting
   4. DAttach all tasks to one all-purpose cluster that is started manually before each run
   Show answer and explanationHide answer

   Correct answer: B

   Task dependencies enforce the order, and a job cluster is created for the run and terminated when it finishes, which matches the cost requirement. Fixed-time separate schedules cannot guarantee that transform waits for ingest. An all-purpose cluster is meant for interactive work and is more expensive for scheduled runs.

5. Question 5Troubleshooting, Monitoring, and Optimization

   In the Spark UI, one task in a shuffle stage runs far longer than the rest because a single join key holds most of the rows. Which fix addresses the root cause?

   1. AEnable adaptive query execution skew join handling to split the heavy partition
   2. BIncrease the driver memory only, since the driver coordinates every shuffle in the job
   3. CReduce the shuffle partitions to one so that all rows are processed in one task
   4. DConvert the source Delta table to CSV so that the reader splits the data more evenly
   Show answer and explanationHide answer

   Correct answer: A

   Adaptive query execution's skew join handling splits oversized partitions at runtime, which spreads the heavy key across several tasks. Driver memory does not change how rows are distributed among executors. A single shuffle partition makes the skew worse by putting all rows in one task.