DBX-DEP Databricks Data Engineer Professional
- Vendor
- Databricks
- Level
- Professional
- Questions
- 60
- Time
- 120 min
- Domains
- 9
- In the bank
- 360
Practice 360 questionsTake a timed mock examDay-by-day study plan
Guides for this exam
- Declarative Pipelines vs Lakeflow Jobs vs Auto Loader vs COPY INTO: Which One a Databricks Question Wants · 한국어
- 8 Mistakes That Repeat Across Every Databricks Exam, and How to Spot Each One in the Question · 한국어
- DBX-DEP Study Guide: The Professional Exam Rewards the Smallest Action That Fixes the Problem · 한국어
The exam at a glance
| Level | Professional |
|---|---|
| Questions | 60 |
| Time | 120 minutes (about 2 minutes per question) |
| Passing mark | Not published by the vendor |
Domains and weights
| Domain | Weight | Practice questions here |
|---|---|---|
| D1 · Developing Code for Data Processing using Python and SQL | 23% | 83 |
| D2 · Data Ingestion & Acquisition | 12% | 43 |
| D3 · Data Manipulation | 12% | 43 |
| D4 · Monitoring and Alerting | 10% | 36 |
| D5 · Cost & Performance Optimization | 15% | 54 |
| D6 · Ensuring Data Security and Compliance | 8% | 29 |
| D7 · Data Governance | 5% | 18 |
| D8 · Debugging and Deploying | 10% | 36 |
| D9 · Data Modeling | 5% | 18 |
D1 · Developing Code for Data Processing using Python and SQL 23%
Code questions show a pipeline, job or bundle that misbehaves or must change, often with a snippet of YAML, Python or SQL, and ask for the fix. How Declarative Pipelines, Lakeflow Jobs and bundles resolve paths, dependencies and targets usually decides. Strong options change the one setting that explains the symptom; weaker ones rewrite the pipeline or switch compute without cause.
What to watch for
- In a bundle, paths and includes resolve relative to the file that declares them; validate errors usually point at the wrong relative path.
- Declarative Pipelines manage dependencies and checkpoints for you; know when a streaming table, a materialized view or a plain view is the right object.
- Serverless compute removes cluster tuning but not code problems: read the error before changing the compute.
Tasks in the official exam guide (12)
- 1.1 Implement scalable Python project structures for Declarative Automation Bundles
- 1.2 Troubleshoot dependency conflicts and installation failures for external libraries
- 1.3 Develop UDFs with Pandas, Python and SQL, including Unity Catalog functions
- 1.4 Create production-ready streaming pipelines with Lakeflow Pipelines and Auto Loader
- 1.5 Create and automate ETL workloads with Lakeflow Jobs via UI, API and CLI
- 1.6 Choose between a streaming table and a materialized view
- 1.7 Implement CDC pipelines with AUTO CDC APIs (SCD Type 1 and 2)
- 1.8 Choose between Structured Streaming and Spark Declarative Pipelines
- 1.9 Create Lakeflow Jobs with control-flow operators (If/Else, For Each)
- 1.10 Choose compute and configuration for environments and dependencies
- 1.11 Develop unit and integration tests for pipeline code
- 1.12 Apply Structured Streaming stateful-processing semantics
D2 · Data Ingestion & Acquisition 12%
Ingestion questions describe a source such as Kafka, a database or files, and a change: a new topic, missed data, a schema change. How streaming checkpoints and offsets behave usually decides. Strong options work with the checkpoint (or start a new one on purpose); weaker ones change options that a running checkpoint ignores.
What to watch for
- Starting options such as startingOffsets apply only to a new checkpoint; replaying lost data needs a new checkpoint or a separate backfill.
- Lakeflow Connect and Lakehouse Federation solve different problems: copying data in versus querying it where it lives.
- Schema changes in the source need an explicit evolution or rescue setting, not a silent drop.
Tasks in the official exam guide (5)
- 2.1 Ingest diverse formats from message buses and cloud storage
- 2.2 Configure incremental CDC pipelines with Delta or Iceberg targets
- 2.3 Configure CDC ingestion from relational databases with Lakeflow Connect
- 2.4 Implement OpenSharing and Clean Rooms for privacy-preserving collaboration
- 2.5 Configure Lakehouse Federation with governance
D3 · Data Manipulation 12%
Manipulation questions give a table shape and a target result and ask for the SQL or pipeline step that produces it. Nulls, ordering and semi-structured data usually decide. Strong options use the function that fits the shape, such as a window function or VARIANT access; weaker ones add joins or collect data to the driver.
What to watch for
- Carrying the last known value forward is a window function with IGNORE NULLS, not a self-join.
- max_by, array functions and ordered aggregates give the latest values per key in one GROUP BY.
- Pipeline expectations can warn, drop or fail on bad rows: pick the action the requirement names.
Tasks in the official exam guide (4)
D4 · Monitoring and Alerting 10%
Monitoring questions describe a failed run, a cost question or an audit need and ask where to look or what to set up. Which record keeps the information, and for how long, usually decides. Strong options read the right system table, event log or API; weaker ones rely on UI views that keep data only for a limited time.
What to watch for
- Each iteration of a For each task is its own run: find the failed iteration before rerunning everything.
- System tables such as system.billing.usage answer cost questions across workspaces; the Jobs UI does not keep run output forever.
- Databricks SQL alerts fit thresholds on query results; pipeline event logs fit data quality and update history.
Tasks in the official exam guide (5)
- 4.1 Use system tables for cost analysis, auditing and workload monitoring
- 4.2 Use REST APIs, CLI and SDK to monitor jobs and pipelines
- 4.3 Use pipeline event logs to monitor health and data quality
- 4.4 Use Lakehouse alerts to monitor business and operational signals
- 4.5 Use the Lakeflow Jobs UI and Jobs API to monitor runs
D5 · Cost & Performance Optimization 15%
Optimization questions describe slow queries or rising cost on Delta tables and ask which layout or setting helps. How the table is queried and maintained usually decides. Strong options let liquid clustering, predictive optimization or caching do the work they are built for; weaker ones keep manual OPTIMIZE and ZORDER jobs that fight them.
What to watch for
- Liquid clustering replaces partitioning and ZORDER for most new tables; choose keys from the columns queries filter on.
- Predictive optimization runs maintenance on serverless compute billed to the account, not to your cluster.
- The disk cache speeds repeated reads on the same compute; it does not help a single cold scan.
Tasks in the official exam guide (6)
- 5.1 Explain how managed tables, Predictive Optimization and Liquid Clustering reduce overhead
- 5.2 Choose the Delta optimization technique for a table access pattern
- 5.3 Understand how caching improves repeated reads
- 5.4 Apply Change Data Feed for incremental downstream processing
- 5.5 Use the query profile to identify performance bottlenecks
- 5.6 Compare Liquid Clustering with partitioning and ZORDER
D6 · Ensuring Data Security and Compliance 8%
Security questions describe a group that needs to do one thing with a job, pipeline or table and must not do more. Which permission level grants exactly that usually decides. Strong options pick the narrowest ACL or Unity Catalog privilege, or a row filter or column mask; weaker ones hand out ownership or admin rights.
What to watch for
- Job and pipeline permission levels (CAN VIEW, CAN MANAGE RUN, CAN MANAGE) separate seeing, running and changing.
- Row filters and column masks protect data in place; attribute-based policies apply them by tag across many tables.
- Run jobs as a service principal so access does not depend on one person's account.
Tasks in the official exam guide (5)
- 6.1 Apply least-privilege ACLs to Unity Catalog and workspace objects
- 6.2 Apply ABAC policies with governed tags for row filters and column masks
- 6.3 Apply anonymization and pseudonymization methods
- 6.4 Implement a compliant pipeline with PII detection and masking
- 6.5 Implement a data purging strategy for retention and deletion policies
D7 · Data Governance 5%
Governance questions describe stewards, catalogs and schemas and ask what a grant or privilege really allows. How Unity Catalog privileges inherit and who may grant them usually decides. Strong options grant at the level that covers future objects too; weaker ones grant object by object or assume MANAGE includes reading data.
What to watch for
- Privileges granted on a catalog or schema apply to objects created later; USE CATALOG and USE SCHEMA are still needed to reach them.
- MANAGE lets a principal grant and manage an object, not read its data.
- Comments, tags and the information schema make data discoverable; they do not grant access.
Tasks in the official exam guide (2)
D8 · Debugging and Deploying 10%
Deployment questions describe Git folders, bundles and CI workflows, and ask how to ship changes safely or debug a run. Which tool owns the deployment usually decides. Strong options deploy with bundles to a target under a service principal and use the Spark UI or system tables to debug; weaker ones edit production by hand.
What to watch for
- Git folders are for development; a production deployment from CI should use bundles, not pull into a shared folder.
- Bundle targets keep dev and prod settings apart in one project.
- For slow stages, read the Spark UI (skew, spill, shuffle) before changing cluster size.
Tasks in the official exam guide (4)
D9 · Data Modeling 5%
Modeling questions describe how tables grow and are read, and ask for the table design or maintenance that keeps them healthy. Small files, history and query patterns usually decide. Strong options match managed tables, clustering, materialized views or metric views to the use; weaker ones copy data into extra tables to work around a design.
What to watch for
- Frequent small appends create many small files: managed tables with predictive optimization, or scheduled OPTIMIZE for external tables, fix it.
- Table size on storage includes history; VACUUM and retention settings decide how much stays.
- Metric views and materialized views give consistent business measures without each team rebuilding them.
Tasks in the official exam guide (2)
A study order that works
The 25-day study plan for DBX-DEP takes the domains in this order. Each day has a short reading, key terms and a quiz from the bank.
- Days 1–6: D1 · Developing Code for Data Processing using Python and SQL
- Days 7–9: D2 · Data Ingestion & Acquisition
- Days 10–12: D3 · Data Manipulation
- Days 13–15: D4 · Monitoring and Alerting
- Days 16–18: D5 · Cost & Performance Optimization
- Days 19–20: D6 · Ensuring Data Security and Compliance
- Days 21–23: D8 · Debugging and Deploying
- Day 24: D9 · Data Modeling
- Day 25: Final Review: Decision Cheat-Sheet Across the Nine Domains
Recent changes to these questions
- : 4 answer choice sets changed, 4 answer keys corrected, 5 explanations rewritten
- : 60 questions added
- : 2 answer choice sets changed, 2 answer keys corrected, 2 explanations rewritten
- : Exam added to the bank (300 questions added)
About these questions
Every question is original, written from the public exam guide and checked against the vendor's own documentation. None comes from a real exam. Each one has an explanation of why the answer is right and why the other options are wrong. How the questions are written.