DBX-DEA Databricks Data Engineer Associate
- Vendor
- Databricks
- Level
- Associate
- Questions
- 45
- Time
- 90 min
- Domains
- 7
- In the bank
- 285
Practice 285 questionsTake a timed mock examDay-by-day study plan
Guides for this exam
- DBX-DEA Study Guide: Ask "What Happens on the Second Run?" and Half the Exam Answers Itself · 한국어
- 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
The exam at a glance
| Level | Associate |
|---|---|
| Questions | 45 |
| Time | 90 minutes (about 2 minutes per question) |
| Passing mark | Not published by the vendor |
Domains and weights
| Domain | Weight | Practice questions here |
|---|---|---|
| D1 · Databricks Intelligence Platform | 6% | 17 |
| D2 · Data Ingestion and Loading | 21% | 59 |
| D3 · Data Transformation and Modeling | 22% | 63 |
| D4 · Working with Lakeflow Jobs | 16% | 46 |
| D5 · Implementing CI/CD | 10% | 28 |
| D6 · Troubleshooting, Monitoring, and Optimization | 10% | 29 |
| D7 · Governance and Security | 15% | 43 |
D1 · Databricks Intelligence Platform 6%
Platform questions check how a Databricks workspace is put together and what it costs to run. Scenarios cover idle clusters, scheduled ETL on shared compute, failed writes and default table formats. The deciding facts are architectural: which plane runs which service, when a Delta write becomes visible, and which compute type fits interactive versus scheduled work.
What to watch for
- For cost questions, check whether the work is interactive or scheduled; scheduled runs belong on compute that exists only for the run, and idle interactive clusters need auto termination.
- Reason about failed writes through the transaction log: data files are not part of a table until the commit, which also explains how time travel and VACUUM interact.
- Keep the two planes straight: Databricks runs the web application and scheduler, while where Spark processes data depends on classic or serverless compute.
Tasks in the official exam guide (2)
D2 · Data Ingestion and Loading 21%
Ingestion items describe a source (files in cloud storage, a SaaS tool, a database) and an expectation about scale, cost, language or deployment. You choose among COPY INTO, Auto Loader modes, streaming tables and Lakeflow Connect connectors. Most wrong options would load the data, but not in the way Databricks recommends for that volume, team or support need.
What to watch for
- As file counts grow, watch the cost of discovering files: directory listing and file notification with file events behave very differently at high volume.
- Running many COPY INTO commands side by side is not a speed trick; one statement over many files usually serves better, and concurrency is meant for separate producers.
- Note who supports a connector before relying on it, and treat ingestion pipelines as code deployed per environment through bundles and the Databricks CLI.
Tasks in the official exam guide (7)
- 2.1 Apply ingestion patterns (batch, streaming, incremental) and import from files and Lakeflow Connect
- 2.2 Load files incrementally with COPY INTO
- 2.3 Use Auto Loader with schema enforcement and evolution
- 2.4 Configure Lakeflow Connect to ingest enterprise sources
- 2.5 Land data with JDBC/ODBC or REST clients orchestrated by Lakeflow Jobs
- 2.6 Prioritize between ingestion methods by requirements
- 2.7 Ingest semi-structured and unstructured data
D3 · Data Transformation and Modeling 22%
Transformation questions are hands-on: a config setting that seems ignored, a join with a duplicate column, bad rows that must be kept, a dashboard that recomputes a heavy query. You choose the function, setting or table object that gives the required result. Small differences between similar options (describe and summary, a view and a materialized view) decide most items.
What to watch for
- For shuffle tuning in DataFrame or SQL code, check which configuration actually controls shuffle partitions; settings for RDD operations do not reach those queries.
- When invalid rows must be reviewed and not just counted, route every row by one condition and its opposite, so each row lands somewhere.
- Choose gold objects by how they are read and refreshed: frequent reads of a heavy aggregate, an always-current projection and append-only ingestion each suit a different object.
Tasks in the official exam guide (7)
- 3.1 Clean bronze data into silver tables
- 3.2 Combine DataFrames (joins and unions)
- 3.3 Manipulate columns, rows and table structures
- 3.4 Deduplicate and aggregate data
- 3.5 Understand basic Spark tuning parameters
- 3.6 Build gold layer objects (materialized views, views, streaming tables, tables)
- 3.7 Apply data quality checks and validation rules
D4 · Working with Lakeflow Jobs 16%
Job questions describe when data arrives and how tasks depend on each other, then ask for the trigger, schedule or behavior that follows. Typical cases include batches of files, an external API with nothing to watch, repeated failures in a continuous job and branching task graphs. Read the timing details closely; a trigger that fires too early or too often is the usual wrong pick.
What to watch for
- Data-driven triggers need a table or location in Unity Catalog to observe. If the source is an outside system, a time-based schedule is the natural fit.
- To start once per batch of files, think of debouncing: a timer that resets with each new file, rather than a cap on how often runs may start.
- Trace a task graph one edge at a time and check each task's Run if condition; siblings do not affect each other, but downstream tasks inherit upstream failures.
Tasks in the official exam guide (4)
D5 · Implementing CI/CD 10%
CI/CD questions center on Declarative Automation Bundles, the Databricks CLI and Git folders. A pipeline or team setup is described, and you predict the outcome (which variable value wins, why extra copies appear) or choose the safe option or workflow. Precedence rules and deployment identity explain most surprises, while shared folders and direct commits to protected branches are the usual traps.
What to watch for
- Learn the order in which a bundle resolves variable values, from command-line options through environment variables and target settings down to defaults.
- When a deployment produces unexpected copies of jobs, look at the bundle's root path and at the identity that deployed it.
- Give each developer a personal Git folder and branch, and merge through pull requests on the Git provider instead of sharing one folder.
Tasks in the official exam guide (4)
D6 · Troubleshooting, Monitoring, and Optimization 10%
Troubleshooting scenarios give a symptom (a cluster that will not start, a stage far slower than the rest, a run that never raises an alert) and ask for the cause or the right source of evidence. Good answers read the specific log, metric or system table that explains it. Options that add workers or keep manual maintenance usually miss the cause.
What to watch for
- In the Spark UI, compare the maximum task duration with the upper percentiles; a large gap points to uneven data, which more workers will not fix.
- Know how run states map to notifications, since a run with a failed non-leaf task can still count as successful.
- For long-term trends, query the lakeflow system tables rather than the Jobs UI, and check whether predictive optimization already does a maintenance task you schedule.
Tasks in the official exam guide (5)
- 6.1 Identify job performance trends in run history
- 6.2 Monitor pipeline health in the Lakeflow Jobs UI
- 6.3 Identify bottlenecks (skew, shuffle, spill) in the Spark UI
- 6.4 Understand Liquid Clustering and predictive optimization
- 6.5 Diagnose cluster startup failures, library conflicts and out-of-memory issues
D7 · Governance and Security 15%
Governance questions concern Unity Catalog: which table type to create, which privileges a principal needs, and how masks, row filters and tags behave. Stems are often precise, with a GRANT statement or a function call in each option. The right one gives exactly the access required and relies on account-level identity, while broad grants and workspace-local checks are the common traps.
What to watch for
- List the privileges a task really needs, such as reading and appending, and avoid grants that also hand over ownership-like control.
- Pick managed tables by default, and external tables when the format is not Delta or other systems read the files directly.
- For consistent ABAC policies, control the tag vocabulary itself, and use account-level group checks in masks and filters rather than workspace-local ones.
Tasks in the official exam guide (4)
A study order that works
The 20-day study plan for DBX-DEA takes the domains in this order. Each day has a short reading, key terms and a quiz from the bank.
- Day 1: D1 · Databricks Intelligence Platform
- Days 2–5: D2 · Data Ingestion and Loading
- Days 6–9: D3 · Data Transformation and Modeling
- Days 10–11: D4 · Working with Lakeflow Jobs
- Days 12–13: D5 · Implementing CI/CD
- Days 14–16: D6 · Troubleshooting, Monitoring, and Optimization
- Days 17–19: D7 · Governance and Security
- Day 20: Final Review: Exam Format, Weights and Commonly Confused Pairs
Recent changes to these questions
- : 1 answer choice set changed, 1 answer key corrected, 1 explanation rewritten
- : 60 questions added
- : 2 answer choice sets changed, 2 answer keys corrected, 2 explanations rewritten
- : 1 explanation rewritten
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.