Question bank
Databricks Certified Data Engineer Associate
43 Databricks Certified Data Engineer Associate questions with the correct answer and a written explanation, organised across 7 exam domains.
- 43 questions
- 7 domains
- Foundation level
certilumQuestion index
Jump straight to any question. Answers and explanations are inline.
Data Ingestion and Loading
8 questions in this domain
- Question 1Single Choice
Which cloud storage URI format is correct for Azure Data Lake Storage Gen2?
- Awasbs://container@storage.blob.core.windows.net/path
- Babfss://container@storage.dfs.core.windows.net/pathCorrect answer
- Cazure://container@storage.dfs.core.windows.net/path
- Dhttps://storageaccount.blob.core.windows.net/path
Why this is the answer
The correct URI for ADLS Gen2 is abfss://, which uses the Azure Blob File System driver.
- Question 2Multiple Select
Which of the following file formats are natively supported by Auto Loader?
- ACSVCorrect answer
- BParquetCorrect answer
- CExcel
- DJSONCorrect answer
Why this is the answer
Auto Loader supports CSV, Parquet, JSON, Avro, ORC, and text files. Excel is not supported natively.
- Question 3Interactive
Based on the provided Auto Loader code, what is the purpose of the 'cloudFiles.schemaLocation' option?
- ASpecifies where ingested data is written
- BStores inferred schema for evolutionCorrect answer
- CSets the checkpoint location for fault tolerance
- DDefines the cloud storage URI
Why this is the answer
The schemaLocation stores the inferred schema persistently, enabling schema evolution across multiple batches.
- Question 4Single Choice
Which of the following is the recommended approach for incremental file ingestion in Databricks?
- ASpark Structured Streaming
- BAuto Loader (cloudFiles)Correct answer
- CCOPY INTO
- Ddbutils.fs.ls()
Why this is the answer
Auto Loader (cloudFiles) is purpose-built for incremental file ingestion, handling schema evolution and exactly-once semantics.
- Question 5Multiple Select
Which of the following are valid output modes for Structured Streaming?
- AappendCorrect answer
- BcompleteCorrect answer
- Coverwrite
- DupdateCorrect answer
Why this is the answer
Structured Streaming supports append, complete, and update output modes. Overwrite is a batch-mode option.
- Question 6Interactive
In the JDBC read code, if the 'id' column is of type VARCHAR, what will happen when the query is executed?
- ARuns successfully with 10 partitions
- BFails because partitionColumn must be numeric, date, or timestampCorrect answer
- CIgnores partitionColumn and uses single partition
- DConverts VARCHAR to integer automatically
Why this is the answer
partitionColumn requires an integer, date, or timestamp column to split ranges; a string column causes an error.
- Question 7Single Choice
What is the default behavior of Delta Lake when writing a DataFrame with a schema that does not match the target table?
- AAutomatically merges schemas
- BFails with an AnalysisExceptionCorrect answer
- COverwrites the table schema
- DReorders columns to match
Why this is the answer
Delta Lake enforces strict schema match by default; mismatches cause a write failure unless mergeSchema is enabled.
- Question 8Multiple Select
Which of the following are benefits of using Auto Loader over standard batch reads?
- AAutomatic file discovery for new filesCorrect answer
- BBuilt-in schema evolutionCorrect answer
- CRequires less cloud storage permissions
- DExactly-once processing with checkpointingCorrect answer
Why this is the answer
Auto Loader automatically discovers new files, supports schema evolution, and provides exactly-once guarantees via checkpointing.
Data Transformation and Modeling
9 questions in this domain
- Question 9Single Choice
What is the primary purpose of the Delta transaction log?
- AStore data files
- BRecord metadata and enable ACID transactionsCorrect answer
- CProvide schema inference
- DManage cluster resources
Why this is the answer
The transaction log records every change, enabling ACID transactions and time travel.
- Question 10Multiple Select
Which statements are true about managed Delta tables in Databricks? (Select two)
- AData is stored in a user-specified location
- BDropping the table deletes the underlying dataCorrect answer
- CThey can be shared across workspaces without additional configuration
- DThey are created using CREATE TABLE without LOCATIONCorrect answer
Why this is the answer
Managed tables store data in the default DBFS location and dropping removes data. They do not specify a LOCATION.
- Question 11Interactive
Given a Delta table with many small files, which command should be run to improve read performance?
- AVACUUM
- BOPTIMIZECorrect answer
- CMERGE
- DDESCRIBE HISTORY
Why this is the answer
OPTIMIZE compacts small files into larger ones, improving scan performance.
- Question 12Single Choice
What happens when you try to write data with a mismatched schema to a Delta table?
- AThe write failsCorrect answer
- BThe schema is automatically evolved
- CThe data is stored in a separate table
- DThe write succeeds with nulls
Why this is the answer
Delta Lake enforces schema on write by default; mismatched columns cause failure.
- Question 13Multiple Select
Which Delta Live Tables expectation actions will drop failing rows without causing pipeline failure? (Select two)
- Aexpect_or_fail
- Bexpect_or_dropCorrect answer
- Cexpect
- Dexpect_all
Why this is the answer
expect_or_drop silently drops failing rows; expect_all applies multiple rules but does not drop on its own.
- Question 14Interactive
You have a Delta table queried with filters on date and region. Which column would be most effective for partitioning to improve query performance?
- Auser_id
- BdateCorrect answer
- Cregion
- Damount
Why this is the answer
Partitioning on low-cardinality columns used in filters, like date, improves query pruning.
- Question 15Single Choice
How long are Delta table transaction log records retained by default for time travel?
- A30 days
- B7 daysCorrect answer
- C1 day
- DForever
Why this is the answer
The default retention duration for Delta transaction log is 7 days.
- Question 16Multiple Select
Which statements are true about PySpark DataFrames? (Select two)
- ATransformations are executed immediately
- BCreating a DataFrame reads all data into memory
- CBuilt-in functions are generally more efficient than Python UDFsCorrect answer
- DDataFrames are immutableCorrect answer
Why this is the answer
PySpark DataFrames are immutable and transformations are lazy. Built-in functions run in optimized JVM code.
- Question 17Interactive
In SCD Type 2, what does the UPDATE clause in this MERGE statement accomplish?
- AIt updates the current row's attribute values in place
- BIt sets is_current to false to expire the current rowCorrect answer
- CIt inserts a new row for the change
- DIt deletes the old row
Why this is the answer
In SCD Type 2, the existing row is marked as expired (is_current = false) before inserting a new version.
Databricks Intelligence Platform
3 questions in this domain
- Question 18Single Choice
What is the primary purpose of the workspace root storage in a Databricks workspace?
- ATo store production data tables and analytics results
- BTo serve as the default location for DBFS, cluster logs, and system tablesCorrect answer
- CTo host Unity Catalog metadata and permission definitions
- DTo provide a globally accessible bucket shared across all workspaces
Why this is the answer
The workspace root storage is an initial cloud storage bucket automatically created during workspace provisioning. It is used for DBFS, cluster logs, and system tables, but it is not intended for production data. Best practice is to use external locations governed by Unity Catalog for production workloads.
- Question 19Multiple Select
Which two of the following authentication methods can be used to set up the Databricks command line interface?
- APersonal Access Token (PAT) generated from the workspaceCorrect answer
- BEnvironment variables DATABRICKS_HOST and DATABRICKS_TOKENCorrect answer
- CUsing the workspace root storage credential
- DUsing the cluster’s IAM role attached to the workspace
Why this is the answer
The Databricks CLI can be authenticated using a personal access token (PAT) or environment variables (DATABRICKS_HOST and DATABRICKS_TOKEN). Workspace root storage credentials and cluster IAM roles are not used for CLI authentication. This is covered in the section on navigating the UI and using the CLI.
- Question 20Interactive
What happens when you execute the following Databricks CLI command?
- AIt creates a new job definition with ID 123
- BIt triggers an immediate run of the existing job with ID 123Correct answer
- CIt lists all runs associated with job ID 123
- DIt submits a one-time job without saving a definition
Why this is the answer
The command 'databricks jobs run-now --job-id 123' triggers an existing job (defined with ID 123) to start a run immediately. It does not create a new job, list runs, or submit a one-time job. The 'run-now' command is used to execute a job on demand.
Governance and Security
7 questions in this domain
- Question 21Single Choice
What is the top-level container in Unity Catalog's three-level namespace?
- AMetastoreCorrect answer
- BCatalog
- CSchema
- DTable
Why this is the answer
A metastore is the top-level container for metadata in Unity Catalog, housing catalogs, schemas, and tables.
- Question 22Multiple Select
Which of the following are valid Unity Catalog privileges?
- ASELECTCorrect answer
- BMODIFYCorrect answer
- CUSAGECorrect answer
- DOWNERSHIP
Why this is the answer
SELECT, MODIFY, and USAGE are standard Unity Catalog privileges; OWNERSHIP is a special authority that cannot be granted.
- Question 23Interactive
Which SQL command correctly grants SELECT on the table 'sales' to the group 'analysts'?
- AGRANT SELECT ON TABLE sales TO analysts;Correct answer
- BGRANT SELECT ON sales TO analysts;
- CALLOW SELECT ON sales TO analysts;
- DGRANT READ ON TABLE sales TO analysts;
Why this is the answer
The correct syntax uses GRANT SELECT ON TABLE followed by the table name and TO with the principal.
- Question 24Single Choice
Which privilege is needed to see a schema's existence without accessing its data?
- ASELECT
- BUSAGECorrect answer
- CMODIFY
- DCREATE
Why this is the answer
USAGE on a schema permits listing its objects but not reading data; SELECT is needed for data access.
- Question 25Multiple Select
Which statements about Unity Catalog tags are true?
- ATags can be applied at catalog, schema, table, and column levels.Correct answer
- BTags propagate automatically from a catalog to all child objects.
- CTags can be referenced by attribute-based access control (ABAC) policies.Correct answer
- DTable tags are set using the ALTER TABLE command with SET TAGS.Correct answer
Why this is the answer
Tags are scoped per object (no inheritance), usable in ABAC, and managed via ALTER TABLE ... SET TAGS.
- Question 26Interactive
Which SQL syntax correctly sets a tag on a table?
- AALTER TABLE sales SET TAGS ('sensitivity' = 'high');Correct answer
- BALTER TABLE sales SET TAG ('sensitivity' = 'high');
- CALTER TABLE sales ADD TAGS ('sensitivity' = 'high');
- DALTER TABLE sales UPDATE TAGS ('sensitivity' = 'high');
Why this is the answer
The correct syntax uses SET TAGS with a comma-separated list of key=value pairs inside parentheses.
- Question 27Single Choice
Who can access the workspace audit logs?
- AAll workspace users
- BWorkspace Admin onlyCorrect answer
- CMetastore Admin only
- DBoth Workspace Admin and Metastore Admin
Why this is the answer
Audit logs are accessible only to Workspace Admins for security; other roles require explicit privileges.
Implementing CI/CD
5 questions in this domain
- Question 28Single Choice
What is the primary purpose of integrating Git with Databricks?
- ATo enable collaboration, reproducibility, and automated deploymentsCorrect answer
- BTo store notebooks directly in DBFS for easy sharing
- CTo schedule jobs without any manual intervention
- DTo import data from external sources into the workspace
Why this is the answer
Git integration enables collaboration, reproducibility, and automation, which are essential for CI/CD pipelines.
- Question 29Multiple Select
Which two actions are required to update a remote Git repository from a Databricks Repo?
- ACommit the changes locallyCorrect answer
- BPush the committed changesCorrect answer
- CPull the latest changes from remote
- DCreate a pull request on the remote provider
Why this is the answer
To update the remote repository, you must first commit locally, then push. Pulling only fetches updates; PRs are for merging branches.
- Question 30Interactive
What does the command 'databricks bundle deploy -t prod' achieve?
- AIt deploys the bundle artifacts and configures resources in the production workspaceCorrect answer
- BIt validates the bundle YAML syntax without connecting to a workspace
- CIt runs the job defined in the bundle immediately
- DIt destroys all resources that were previously deployed by the bundle
Why this is the answer
The 'deploy' command uploads artifacts and creates or updates Databricks resources in the specified target environment.
- Question 31Single Choice
Which authentication method is recommended for automating Databricks deployments in CI/CD?
- AService principal tokenCorrect answer
- BPersonal user password
- CWorkspace admin credentials
- DSSH key pair for the cluster
Why this is the answer
Service principal tokens are not tied to a user, making them secure and ideal for automated pipelines.
- Question 32Multiple Select
Which two features are true about Databricks Asset Bundles?
- AThey use a declarative YAML file called databricks.ymlCorrect answer
- BThey support environment-specific variables through targetsCorrect answer
- CThey execute notebooks during the deployment process
- DThey require manual resource creation in the workspace after each commit
Why this is the answer
DABs are declarative IaC with a YAML manifest and support parameterized targets. They do not execute code at deploy time.
Troubleshooting, Monitoring, and Optimization
4 questions in this domain
- Question 33Single Choice
In the context of analyzing Spark performance, what kind of information does the Spark UI provide real-time visibility into?
- AOnly driver logs and environmental configurations
- BJob, stage, and task execution detailsCorrect answer
- COnly storage memory and cache metrics
- DOnly SQL query plans and DAG visualization
Why this is the answer
The Spark UI provides real-time visibility into jobs, stages, tasks, storage, SQL execution, and environment configuration.
- Question 34Multiple Select
Which two statements correctly describe benefits of Z-order clustering on Delta tables?
- AIt reduces the amount of data scanned via data skippingCorrect answer
- BIt completely replaces the need for table partitioning
- CIt colocates correlated data values within the same filesCorrect answer
- DIt automatically runs VACUUM to remove old files after optimization
Why this is the answer
Z-order clustering improves data skipping and colocates similar values in files. Partitioning remains beneficial and VACUUM must be run separately.
- Question 35Interactive
Examine the code below. What is the purpose of adding the 'salt' column in this data skew mitigation technique?
- ATo increase the total number of shuffle partitions
- BTo artificially distribute skewed keys across more partitionsCorrect answer
- CTo reduce the number of shuffle partitions
- DTo enable the broadcast join optimization
Why this is the answer
Adding a random salt to skewed join keys allows data to be distributed more evenly across partitions, mitigating skew.
- Question 36Single Choice
When a job fails with a 'CLUSTER_UNAVAILABLE' error, which log source should you first examine to diagnose the issue?
- ADriver logs
- BWorker logs (executor stderr)
- CCluster event logsCorrect answer
- DSpark UI stages tab
Why this is the answer
Cluster event logs capture infrastructure errors such as cluster unavailability or resource exhaustion, making them the first source for such failures.
Working with Lakeflow Jobs
7 questions in this domain
- Question 37Single Choice
What is the recommended cluster type for running production jobs in Databricks to ensure isolation and cost efficiency?
- AAll-purpose cluster
- BJob clusterCorrect answer
- CHigh-concurrency cluster
- DInteractive cluster
Why this is the answer
Job clusters are ephemeral, isolated, and auto-terminated, making them cost-effective and secure for production jobs. All-purpose clusters are shared and incur idle costs.
- Question 38Multiple Select
Which two statements correctly describe job triggers in Databricks?
- AScheduled jobs use cron expressions and always run regardless of previous outcomes.Correct answer
- BEvent-driven triggers can be configured for file arrival or Delta table updates.Correct answer
- CConcurrent runs are enabled by default for scheduled jobs.
- DManual triggers require a valid cron expression.
Why this is the answer
Cron triggers are stateless (correct). Event-driven triggers support file arrival/Delta table (correct). Concurrent runs are disabled by default. Manual triggers do not use cron.
- Question 39Interactive
The following JSON represents a Databricks job definition. What is missing to ensure the job retries twice on failure?
- AAdd 'max_retries': 2 to the task definition.Correct answer
- BAdd 'retry_policy': {'max_retries': 2} to the job level.
- CAdd 'timeout_seconds': 3600 to limit execution time.
- DAdd 'schedule': {'quartz_cron_expression': '0 0 */2 * * ?'} to run every 2 hours.
Why this is the answer
The field 'max_retries' specifies the number of retry attempts. The default is 0; setting it to 2 enables two retries.
- Question 40Single Choice
What is the purpose of a job cluster's 'autotermination_minutes' setting?
- AIt sets the maximum run time of the job.
- BIt defines the idle time before the cluster is automatically terminated.Correct answer
- CIt determines the interval between retries.
- DIt configures the time zone for the job schedule.
Why this is the answer
Autotermination_minutes specifies how long a job cluster stays alive after the last activity before being shut down to save costs.
- Question 41Multiple Select
Which two options are valid destinations for job notifications in Databricks?
- AEmailCorrect answer
- BSMTP Relay
- CWebhookCorrect answer
- DDatabricks CLI
Why this is the answer
Email and webhook are directly configurable in job notifications. SMTP Relay is not a configurable destination; Databricks provides built-in email relay.
- Question 42Interactive
In the following sample notebook command, what is missing to read a job parameter named 'source_path'?
- ADefine the widget using dbutils.widgets.text("source_path", "") before retrieving the parameter.Correct answer
- BThe code is correct; no changes are needed.
- CReplace dbutils.job.taskValues with dbutils.widgets.get.
- DSet 'source_path' as a Spark configuration property.
Why this is the answer
The widget must be defined using 'dbutils.widgets.text' before calling 'get', otherwise the parameter is not recognized. This is the standard way to parameterize notebooks.
- Question 43Single Choice
What is the default number of retries for a task in a Databricks job if not explicitly configured?
- A0Correct answer
- B3
- C10
- D5
Why this is the answer
The default retry count is 0, meaning no automatic retries are performed unless explicitly configured.
Frequently asked questions
Everything about the Databricks Certified Data Engineer Associate question bank and how to use it.