Databricks Certified Data Engineer Associate Practice Exam

What the Databricks Certified Data Engineer Associate exam covers

• Databricks Lakehouse Platform180 questions
• ELT with Spark SQL and Python217 questions
• Incremental Data Processing165 questions
• Production Pipelines119 questions
• Data Governance67 questions

Free Databricks Certified Data Engineer Associate sample questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 748.

1. Question 1Data Governance

   What is the primary purpose of Unity Catalog within the Databricks Lakehouse Platform?

   • ATo provide a single, centralized governance layer for data and AI assets across every workspace in an accountCorrect
   • BTo provide a dedicated cluster type reserved for streaming Structured Streaming jobs
   • CTo replace Delta Lake's transaction log with a faster metadata format
   • DTo automatically tune Spark shuffle partitions during query execution
   ✓ Correct answer: A

   Unity Catalog provides one place to manage permissions, auditing, lineage, and discovery for catalogs, schemas, tables, and other assets, and that governance applies consistently to every workspace attached to the same metastore. It is a governance and security concept, not a compute engine or performance-tuning feature.

   Why the other options are wrong

   • BCluster types (all-purpose vs job) are a compute concept unrelated to governance; Unity Catalog does not define cluster types.
   • CUnity Catalog governs Delta tables; it does not replace or alter the underlying Delta transaction log.
   • DShuffle partition tuning is a Spark performance setting, not a Unity Catalog governance function.
2. Question 2Incremental Data Processing

   Besides Delta and cloud file sources, which built-in format is commonly used only for local testing of a streaming pipeline?

   • Aparquet-test
   • Bconsole-source
   • Csql
   • DrateCorrect
   ✓ Correct answer: D

   The rate source produces rows at a configurable rate purely for testing and demos; it has no real-world data behind it. It is useful for experimenting with triggers, output modes, and windowed aggregations without needing a real file or messaging source.

   Why the other options are wrong

   • AThere is no 'parquet-test' source format in Spark.
   • B'console' is a sink format for debugging output, not a source format named 'console-source'.
   • C'sql' is not a Structured Streaming source format.
3. Question 3Incremental Data ProcessingSelect all that apply

   In the context of stateful streaming aggregations, which statements about watermarks are correct? (Choose TWO)

   • AA watermark is required before a streaming query can be started at all, even without aggregation
   • BA watermark defines a threshold for how late event-time data can arrive and still be included in an active windowCorrect
   • CWatermarks guarantee that no late data will ever be dropped
   • DWatermarks help bound the size of state that must be retained for closed windowsCorrect
   • EWatermarks are configured on the query's trigger, not on a DataFrame column
   ✓ Correct answer: B, D

   withWatermark() tells Spark how long past the maximum observed event time it should keep accepting late data for a window before considering that window closed, which lets Spark safely drop the state associated with closed windows and cap overall memory usage.

   Why the other options are wrong

   • ANon-aggregating queries do not require a watermark in order to run.
   • CData arriving later than the watermark threshold can indeed be dropped from consideration; there is no absolute guarantee against ever dropping late data.
   • EWatermarks are defined using withWatermark() on an event-time column of the DataFrame, not as a trigger setting.
4. Question 4Incremental Data Processing

   A source directory stores files in a structure like /data/year=2026/month=07/day=28/file.json, and the team wants Auto Loader to automatically populate year, month, and day as columns derived from the file path rather than parsing them out of the file contents. Which Auto Loader option supports this?

   • AcloudFiles.partitionColumnsCorrect
   • BcloudFiles.inferColumnTypes
   • CcloudFiles.validateOptions
   • DcloudFiles.allowOverwrites
   ✓ Correct answer: A

   When source files are organized in Hive-style partition directories (such as year=2026/month=07/day=28), setting cloudFiles.partitionColumns lets Auto Loader extract those path segments and add them as columns in the resulting DataFrame, instead of requiring that information to be parsed out of each file's contents.

   Why the other options are wrong

   • BcloudFiles.inferColumnTypes controls whether inferred columns get specific data types instead of defaulting to string; it does not extract values from the file path.
   • CcloudFiles.validateOptions is not a real Auto Loader configuration option.
   • DcloudFiles.allowOverwrites controls whether Auto Loader can reprocess a file that was modified in place, unrelated to deriving columns from a path.
5. Question 5Databricks Lakehouse PlatformSelect all that apply

   Which TWO SQL warehouse types require compute instances to be provisioned inside the customer's own cloud account (i.e., are not serverless)? (Choose TWO)

   • AClassicCorrect
   • BProCorrect
   • CServerless
   • DA cluster pool
   • EAn all-purpose cluster
   ✓ Correct answer: A, B

   Unlike the Serverless tier, both the Classic and Pro SQL warehouse types provision their underlying cloud instances inside the customer's own cloud account, which means they take longer to start up than serverless compute.

   Why the other options are wrong

   • CServerless is the tier that avoids provisioning instances in the customer's own account; it runs on Databricks-managed compute.
   • DA cluster pool is a general mechanism for pre-warming instances for clusters, not a named SQL warehouse tier.
   • EAn all-purpose cluster is a notebook/job compute resource, not one of the Databricks SQL warehouse tiers.
6. Question 6Databricks Lakehouse Platform

   Which scenario is a typical, expected example of Delta schema evolution?

   • AAn upstream source starts sending a new column, and with mergeSchema enabled it is automatically added to the target tableCorrect
   • BA column's data type is silently changed from integer to an incompatible struct type without any configuration
   • CThe entire table is dropped and recreated every time any column changes
   • DRows with mismatched types are deleted automatically instead of being written
   ✓ Correct answer: A

   The common, supported schema-evolution scenario is a new column appearing in the source and being added to the target table's schema when mergeSchema (or autoMerge) is enabled. This is additive and backward compatible, unlike arbitrary incompatible type changes.

   Why the other options are wrong

   • BIncompatible type changes like integer-to-struct are not silently allowed by evolution.
   • CSchema evolution does not require dropping and recreating the table.
   • DMismatched rows are not silently deleted; enforcement blocks the write instead.
7. Question 7Production Pipelines

   Which Python DLT decorator applies a dictionary of multiple named expectations at once and fails the update if any rule is violated?

   • A@dlt.expect_all_or_failCorrect
   • B@dlt.expect_all
   • C@dlt.expect_or_fail
   • D@dlt.expect_dict_or_fail
   ✓ Correct answer: A

   The 'all' family of decorators (expect_all, expect_all_or_drop, expect_all_or_fail) let a single decorator apply several named rules at once, matching the corresponding single-rule decorator's enforcement level for every rule in the dictionary.

   Why the other options are wrong

   • B@dlt.expect_all applies multiple expectations but only in warn mode, without failing the update.
   • C@dlt.expect_or_fail applies to a single expectation, not a dictionary of multiple rules.
   • D@dlt.expect_dict_or_fail is not a real decorator in the DLT Python API.
8. Question 8Production Pipelines

   Where can an engineer see a chronological list of every past execution of a specific Databricks Job, including start time, duration, and outcome?

   • AThe job's run history view in the Jobs UICorrect
   • BThe Delta table's transaction log (_delta_log) for the target table
   • CThe workspace's Unity Catalog metastore configuration page
   • DThe cluster's Spark UI environment tab
   ✓ Correct answer: A

   This run history lets engineers review when each run started, how long it took, whether it succeeded or failed, and drill into individual task logs and outcomes, which is central to operating and troubleshooting a production job over time.

   Why the other options are wrong

   • BThe Delta transaction log tracks table-level file and schema changes, not job execution history.
   • CThe metastore configuration page manages catalog/schema governance settings, not job run history.
   • DThe Spark UI environment tab shows cluster configuration details for a session, not a job's historical run list.
9. Question 9ELT with Spark SQL and Python

   A table sales.orders has columns (order_id, customer_id, order_total) in that order. A data engineer wants to insert a new row while explicitly specifying only order_id and order_total, leaving customer_id to receive its default or NULL value. Which syntax supports this?

   • AINSERT INTO sales.orders (order_id, order_total) VALUES ('O1', 99.50);Correct
   • BINSERT INTO sales.orders VALUES ('O1', 99.50);
   • CINSERT INTO sales.orders SKIP (customer_id) VALUES ('O1', 99.50);
   • DINSERT INTO sales.orders EXCEPT (customer_id) VALUES ('O1', 99.50);
   ✓ Correct answer: A

   Providing an explicit column list in parentheses after the table name lets the VALUES clause supply only those named columns, leaving the rest at default or NULL.

   Why the other options are wrong

   • BWithout an explicit column list, VALUES positions must match all declared columns in order, which would misalign values against this table's three columns.
   • CSKIP is not valid INSERT INTO syntax for omitting columns.
   • DEXCEPT is a set operator for combining query results; it is not valid syntax within an INSERT INTO column list.
10. Question 10ELT with Spark SQL and Python

    Which snippet correctly takes the result of a SQL query and continues processing it using the PySpark DataFrame API?

    • Aresult_df = spark.sql("SELECT region, amount FROM sales"); result_df.filter(col("amount") > 100)Correct
    • Bresult_df = spark.sql("SELECT region, amount FROM sales"); result_df.execute()
    • Cspark.sql("SELECT region, amount FROM sales").toSQL()
    • Dresult_df = spark.sql("SELECT region, amount FROM sales").decode()
    ✓ Correct answer: A

    Because spark.sql() returns a DataFrame, the result can be assigned to a variable and immediately used with any DataFrame API method, such as filter(), just like a DataFrame built through .read or transformations.

    Why the other options are wrong

    • Bexecute() is not a valid DataFrame method.
    • CtoSQL() is not a valid DataFrame method for this purpose.
    • Ddecode() is not a valid DataFrame method; spark.sql() already returns a structured DataFrame, not encoded text.

Who this Databricks Certified Data Engineer Associate practice exam is for

This practice set is for anyone preparing for the Databricks Certified Data Engineer Associate exam at the intermediate level - from first-time candidates building a foundation to experienced Data practitioners doing a final review before test day. If you learn best by working through realistic questions and reading why each answer is right or wrong, it is built for you.

How to use this Databricks Certified Data Engineer Associate practice exam

1. Start with the free sample questions above to gauge your current baseline.
2. Read the full explanation on every question, including why each wrong option is wrong.
3. Track your weak domains and focus your study where you are losing the most marks.
4. Once you are scoring consistently well, take a timed, full-length mock exam.
5. Use your readiness score to decide when you are ready to book the real Databricks Certified Data Engineer Associate exam.

Related Data resources

• Databricks Certified Data Engineer Associate study guideKey concepts
• Data practice examsAll Data
• Certification pathWhere this fits
• Certification exam guides & tipsBlog
• Plans & pricingFree & paid
• SnowPro Core (COF-C02) practice examRelated

Databricks Certified Data Engineer Associate practice exam FAQ