Databricks Certified Data Engineer Professional

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

1. Question 1Developing Code for Data Processing using Python and SQL

   A pipeline must upsert daily change records into a Delta table keyed on customer_id. Duplicate keys can appear within one batch. Which statement produces a correct result?

   1. AUse MERGE with a deduplicated source, so each customer_id appears at most once before the match is evaluated
   2. BUse INSERT INTO with the batch as-is, then rely on a later OPTIMIZE to remove the duplicate rows automatically
   3. CUse MERGE with duplicate keys in the source, since Delta picks the latest row for each key without any rule
   4. DUse COPY INTO each night with the full history, which overwrites the table and removes any duplicates from the batch
   Show answer and explanationHide answer

   Correct answer: A

   MERGE fails when multiple source rows match one target row, so the source must be deduplicated to one row per key first, often by ordering on a timestamp. INSERT leaves duplicates, duplicate keys in MERGE cause errors rather than an automatic choice, and OPTIMIZE does not remove logical duplicates.

2. Question 2Data Ingestion & Acquisition

   A source drops thousands of JSON files into a cloud storage folder each hour. The pipeline must ingest each file exactly once, scale without listing the whole folder, and tolerate new fields in the JSON. Which approach fits?

   1. AA streaming job that reads the folder with a fixed schema and drops any file that contains a field it does not know
   2. BA nightly batch that reads every file in the folder and appends all of them to the target table each time
   3. CA manual COPY INTO run by an engineer each hour, with the file list pasted into the notebook as a parameter
   4. DAuto Loader with cloudFiles, a checkpoint location, and schema evolution enabled for new fields
   Show answer and explanationHide answer

   Correct answer: D

   Auto Loader tracks processed files through its checkpoint, scales with directory notifications or incremental listing, and can evolve the schema for new fields. Nightly full appends duplicate data, manual COPY INTO depends on people, and a fixed schema that drops files loses data.

3. Question 3Cost & Performance Optimisation

   A nightly job runs on a large all-purpose cluster that sits idle for most of the day, and costs have risen. The workload is scheduled and predictable. Which change most directly reduces cost?

   1. AKeep the all-purpose cluster running at its maximum size so that jobs start instantly every time they are scheduled
   2. BMove the scheduled work to a job cluster or serverless compute that starts for the run and stops when it finishes
   3. CRun the job on a single-node cluster that the team starts manually each night, without any scheduling or policy
   4. DAdd more workers to the all-purpose cluster so that the job finishes sooner and the cluster is turned off earlier
   Show answer and explanationHide answer

   Correct answer: B

   Job or serverless compute starts for the scheduled run and stops after it, which removes idle cost. Keeping a large cluster always on maximizes idle spend, manual single-node runs add operational work, and more workers raise cost without removing idle time.

4. Question 4Data Transformation, Cleansing, and Quality

   A declarative pipeline must drop rows with a null customer_id and record how many rows were dropped in the pipeline's quality metrics. Which feature fits?

   1. AAn expectation with a drop action on customer_id IS NOT NULL, which records the dropped rows in the pipeline event log
   2. BA Python try and except block around each row that silently skips the bad rows without logging anything at all
   3. CA separate job that counts the nulls each night and emails the total to the data owner, with no change to the pipeline
   4. DA WHERE clause in the final view with no record of the removed rows, since the metric is not needed for the team
   Show answer and explanationHide answer

   Correct answer: A

   An expectation with a drop action removes the failing rows and records quality metrics in the pipeline event log. A WHERE clause removes rows without metrics, silent exception handling hides the loss, and a nightly email job is detached from the pipeline.

5. Question 5Ensuring Data Security and Compliance

   A table holds personal data, and only a compliance group may see the email column. Other analysts can query the table. Which Unity Catalog feature enforces this?

   1. AA comment on the email column asking analysts not to read it, with no enforcement on the table itself
   2. BA notebook setting that hides the output of the email column in the notebooks analysts run on their own laptops
   3. CA separate copy of the table with the email column removed, shared with all analysts and kept up to date by hand
   4. DA column mask function that returns masked values to users outside the compliance group
   Show answer and explanationHide answer

   Correct answer: D

   A column mask enforces access at query time based on the user's group, so unauthorized users receive masked values. Comments do not enforce anything, a manual copy drifts and duplicates data, and notebook settings do not protect data in queries.