DATABRICKS-CERTIFIED-DATA-ENGINEER-ASSOCIATE Real Exam Questions and

Question 1 Single choice

A data engineer needs to find all customers who have never placed an order. The customers table contains customer_id, customer_name, and other customer attributes. The orders table contains order_id, customer_id, and other order attributes. Both tables share customer_id as the join key. The result must include only customer columns with no order-related columns in the output.

Which expression exactly achieves this requirement?

A
Inner join, then filter for NULL order_id
B
Left anti-join from customers to orders
C
Left outer join from orders to customers
D
Full outer join, then filter where order_id IS NULL

Show answer and explanation

Correct answer: B

Explanation

A left anti-join returns rows from the left DataFrame only when no matching key exists on the right. Placing customers on the left and orders on the right therefore selects customers whose customer_id has no order match. Anti-join output contains only the left-side customer columns, so no additional projection is needed to remove order attributes.

Question 2 Single choice

A governance team is evaluating whether to use Unity Catalog attribute-based access control (ABAC) policies or manually applied row filters and column masks to protect sensitive data across their catalog.

Why should the team use ABAC policies instead of manually applied row filters and column masks?

A
ABAC policies use governed tags to dynamically match tables and columns, so a single policy defined on a catalog automatically applies to all current and future tables within that catalog without per-table configuration.
B
ABAC policies allow row filters and column masks to be written in Python and Scala directly, whereas manually applied filters and masks can only use SQL user-defined functions.
C
ABAC policies can be applied to views and materialized views in addition to tables, whereas manually applied row filters and column masks can only be applied to tables.
D
ABAC policies support applying multiple distinct row filters to the same table for a single user at query time, whereas manually applied row filters are limited to one filter per table.

Show answer and explanation

Correct answer: A

Explanation

ABAC policies use governed tags as dynamic matching criteria for tables and columns. Defining one policy at the catalog level therefore extends its protection to matching current and future objects in that catalog. This removes the need to attach and maintain equivalent row filters or column masks on every table separately.

Question 3 Single choice

A data engineering team is using Kafka to capture event data and then ingest it into Databricks. The team wants to be able to see these historical events. Medallion architecture is already in place. The team wants to be mindful of costs.

Where should this historical event data be stored?

A
Gold
B
Silver
C
Bronze
D
Raw layer

Show answer and explanation

Correct answer: C

Explanation

The Bronze layer preserves ingested event records in their raw or near-raw form, making it the appropriate place to retain Kafka history before cleansing. Silver is intended for refined data and Gold for consumer-level outputs. Keeping the original events in Bronze also avoids unnecessary transformations for archival access.

Question 4 Single choice

A data engineer wants to create a data entity from a couple of tables. The data entity must be used by other data engineers in other sessions. It also must be saved to a physical location.

Which of the following data entities should the data engineer create?

A
Database
B
Function
C
View
D
Temporary view
E
Table

Show answer and explanation

Correct answer: E

Explanation

A table is persistent across sessions and stores its data at a physical location. Those two properties allow other data engineers to use the entity later without depending on the creator's active session. A temporary view is session-scoped, while a regular view represents a saved query and does not itself satisfy the requirement to store physical data.

Question 5 Single choice

Which SQL code snippet will correctly demonstrate a Data Definition Language (DDL) operation used to create a table?

A
CREATE TABLE employees (
id INT,
name STRING
);
B
DROP TABLE employees;
C
ALTER TABLE employees ADD COLUMN salary DECIMAL(10,2);
D
INSERT INTO employees (id, name) VALUES (1 'Alice');

Show answer and explanation

Correct answer: A

Explanation

CREATE TABLE employees (id INT, name STRING); is a Data Definition Language operation because it creates a table object and defines its column names and types. INSERT is a data-manipulation operation, while DROP and ALTER are also DDL operations but do not demonstrate the requested creation of a new table.

Question 6 Single choice

A data engineer and data analyst are working together on a data pipeline. The data engineer is working on the raw, bronze, and silver layers of the pipeline using Python, and the data analyst is working on the gold layer of the pipeline using SQL. The raw source of the pipeline is a streaming input. They now want to migrate their pipeline to use Delta Live Tables.

Which of the following changes will need to be made to the pipeline when migrating to Delta Live Tables?

A
None of these changes will need to be made
B
The pipeline will need to stop using the medallion-based multi-hop architecture
C
The pipeline will need to be written entirely in SQL
D
The pipeline will need to use a batch source in place of a streaming source
E
The pipeline will need to be written entirely in Python

Show answer and explanation

Correct answer: A

Explanation

Delta Live Tables can support the existing medallion-based multi-hop design, accept a streaming source, and define pipeline datasets with both Python and SQL. Therefore, the engineer can continue implementing raw through Silver logic in Python while the analyst defines Gold logic in SQL; none of the listed architectural or language changes is required.

Question 7 Single choice

A data engineer is cleaning a bronze table that receives the same customer records from multiple source systems. Duplicate rows have the same customer_id and email but different ingestion_timestamp values. The silver table should contain only one record per unique combination of customer_id and email.

Which PySpark operation correctly deduplicates based on the business keys?

A
df.dropDuplicates(['customer_id', 'email'])
B
df.groupBy('customer_id', 'email').agg(max('ingestion_timestamp').alias ('latest_ts'))
C
df.select('customer_id', 'email').distinct()
D
df.distinct()

Show answer and explanation

Correct answer: A

Explanation

df.dropDuplicates(['customer_id', 'email']) defines the duplicate key as the two business columns rather than the entire row. Differences in ingestion_timestamp therefore do not create separate Silver records, and one complete record is retained for each unique customer_id and email pair. A full distinct() would still treat the timestamps as meaningful differences.

Question 8 Single choice

A data engineer is standardizing repository layouts for multiple teams adopting Databricks Asset Bundles. The engineer wants to ensure every project has a single authoritative configuration file at the repository root that defines the bundle name, targets, workspace settings, permissions, and resource mappings (for jobs and pipelines).

What strategy should the data engineer use to meet the goal?

A
Place multiple databricks.yml files under each subfolder (for example, jobs/pipelines/, workspace/) and merge them at deploy time using the include mapping.
B
Place exactly one databricks.yml at the repository root; it is the main configuration file and may reference additional configuration files via the include mapping.
C
Place a databricks.yml in a databricks/hidden folder at the repository root; only hidden locations are valid for bundle configs.
D
Place a databricks.yaml at the repository root and optional databricks.yml in subfolders; the CLI prefers .yaml over.yml when both exist.

Show answer and explanation

Correct answer: B

Explanation

The repository root should contain exactly one authoritative databricks.yml as the bundle's main configuration file. It establishes the bundle identity and can define targets, workspace settings, permissions, and resource mappings. When configuration is split for maintainability, the root file may reference additional files through include without creating competing main files.

Question 9 Single choice

Which of the following commands will return the location of database customer360?

A
DESCRIBE LOCATION customer360;
B
DROP DATABASE customer360;
C
DESCRIBE DATABASE customer360;
D
ALTER DATABASE customer360 SET DBPROPERTIES ('location' = '/user'};
E
USE DATABASE customer360;

Show answer and explanation

Correct answer: C

Explanation

DESCRIBE DATABASE customer360; returns descriptive metadata for that database, including its storage location. The statement inspects the database without changing it. USE only changes the active database, DROP removes it, and ALTER changes properties rather than serving as the command for retrieving its current metadata.

Question 10 Single choice

A data engineer converts an external table to a managed table. A structured streaming job that reads from this table continues running during the conversion. After the conversion completes, the streaming job stops processing new records.

How should the data engineer resolve this?

A
Grant the streaming job additional permissions on the new managed storage location
B
Restart the streaming job, so it picks up the new managed table location
C
Run REFRESH TABLE on the converted table to update the streaming checkpoint
D
Delete the streaming checkpoint directory and let the job reprocess from the beginning

Show answer and explanation

Correct answer: B

Explanation

Converting the table changes its managed storage location while the running stream still has execution state tied to the location it opened previously. Restarting the streaming job causes it to resolve and use the converted managed table location. The existing checkpoint need not be deleted, so the job can resume without intentionally reprocessing the entire source history.