Examworthyexamworthy.comDatabricks Certified Data Engineer Associate cheat sheet
Databricks
Free to share. Examworthy is not affiliated with or endorsed by Databricks; Data-Engineer-Associate and related marks belong to their respective owners.
At a glance
Format: Multiple choice, online proctored or at a test center
Domain weight map
Heaviest first - spend your time hereHow this exam thinks
This exam rewards having built a pipeline, not having read about one. Most questions resolve to a single mechanism that a practitioner already relies on.
Spot the trap
Tempting wrong answers, and why they failTempting but wrong
Under ANSI behaviour, does CAST to a decimal type silently substitute NULL when it cannot parse a value?
Why it fails
No. That NULL-on-failure behaviour belonged to the legacy non-ANSI mode. With ANSI behaviour enabled, a CAST that cannot represent its input in the target type raises a runtime error and aborts the query, so TRY_CAST is required if you need a null result instead of a failure.
Data Transformation and Modeling
Tempting but wrong
If COPY INTO is rerun against a folder that still contains previously loaded files, does it duplicate those rows in the target table?
Why it fails
No. That behavior describes a plain read-and-append of a folder, not COPY INTO. COPY INTO maintains its own per-target record of already-ingested files and skips them, so files retained in the folder are not reloaded.
Data Ingestion and Loading
Tempting but wrong
If a task is skipped because its dependency sat on the untaken branch of a condition task, does that skip count as a failed dependency and fail the downstream task and the run?
Why it fails
No. A skipped task is not a failure state. The default Run if condition does block a dependent task whose upstream was skipped, but the dependent task is itself marked skipped, not failed, and skipped tasks do not cause the overall run to be reported as failed.
Working with Lakeflow Jobs
Tempting but wrong
Does a SELECT grant on a table only take effect for the table's owner, requiring a separate metastore-level grant for anyone else to read it?
Why it fails
No. A SELECT grant applies directly to the named principal, whether a user, group, or service principal, regardless of who owns the table. Unity Catalog has no requirement to duplicate a grant at the metastore level for ordinary reads; a table-level SELECT is sufficient once the traversal privileges on the catalog and schema are also in place.
Governance and Security
Tempting but wrong
Can a pull request be opened, reviewed and merged entirely from inside a Databricks Git folder's dialog?
Why it fails
No. Even though the dialog can reference pull requests, it has no review surface of its own. Opening, approving and merging a pull request always happens in the linked Git provider, not in the workspace.
Implementing CI/CD
Tempting but wrong
Is it true that setting one timeout raises a notification and a run is only cancelled once a second, higher timeout is also configured on the same job?
Why it fails
No. A job carries a single timeout value, and reaching it cancels the run outright rather than merely notifying. There is no two-tier timeout mechanism. Alerting without cancellation requires the separate warning threshold setting, not a lower timeout paired with a higher one.
Troubleshooting, Monitoring, and Optimization
Tempting but wrong
If delta.logRetentionDuration is set to thirty days, does a time travel query to a twenty-day-old version always succeed?
Why it fails
No. Log retention only guarantees the log entry describing that version is kept, not the data files it references. If VACUUM has already deleted those files under a shorter file retention setting, the read fails even though the log still names the version.
Databricks Intelligence Platform
Tempting but wrong
If you pass both a subset list and a how argument to DataFrameNaFunctions.drop, does the how argument override the subset and force the check across every column in the DataFrame?
Why it fails
No. subset and how are independent settings: subset picks which columns are inspected for nulls, and how sets the threshold applied within that selection. Supplying how never widens or cancels the subset restriction.
Data Transformation and Modeling
Key terms
Exam-day rules
- Read the requirement in the last line first, then judge each option against it. Several options will be true statements about the platform and only one answers what was asked.
- Ask which option you have actually watched happen. This exam is built on the gap between what sounds reasonable and what the engine does.
- On any access question, walk the whole path. A SELECT grant with a missing USE CATALOG or USE SCHEMA above it is the exam's favourite trap.
- When a scenario names file volume or arrival pattern, it is choosing between Auto Loader and COPY INTO. Let the growth of the directory decide.
- For a row-count question, work out the answer on paper before reading the options. The options are designed to be individually plausible.
Revision schedule
- Week 1Get current on the platform and its 2026 names
- Week 1Build one real pipeline end to end
- Weeks 2-3Drill ingestion and transformation, the two heaviest domains
- Week 3Work through orchestration and governance
- Week 4Cover CI/CD and troubleshooting