DP-700: Microsoft Fabric Data Engineer Associate Practice Exam

What the DP-700 exam covers

• Implement and manage an analytics solution263 questions
• Ingest and transform data263 questions
• Monitor and optimize an analytics solution232 questions

Free DP-700 sample questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 758.

1. Question 1Implement and manage an analytics solution

   You need to grant a data engineer the ability to create, edit, and delete all items in a Fabric workspace, including managing other users' access, but you must NOT make them the owner of the workspace. Which workspace role should you assign?

   • AViewer
   • BContributor
   • CMemberCorrect
   • DAdmin
   • EGuest
   ✓ Correct answer: C

   Member can view, create, edit, and delete items and can add other users (as Members, Contributors, or Viewers). This meets the requirement of managing access without conferring full Admin/owner-level control such as deleting the workspace itself.

   Why the other options are wrong

   • AViewer can only read/consume content and cannot create or edit items.
   • BContributor can create and edit items but cannot add or manage other users' access.
   • DAdmin has full control including deleting the workspace and would effectively act as an owner-level role, which the requirement excludes.
   • EGuest is not a Fabric workspace role; workspace roles are Admin, Member, Contributor, and Viewer.
2. Question 2Implement and manage an analytics solutionSelect all that apply

   You created a SECURITY POLICY with a FILTER PREDICATE on a Warehouse fact table. A member of the Workspace Admin role reports they can still see all rows. Which TWO statements explain this correctly?

   • AMembers mapped to elevated workspace roles that grant administrator-level access can bypass RLS filter predicates on the WarehouseCorrect
   • BRLS filter predicates never apply to any user who has SELECT permission
   • CThe policy may have been created with STATE = OFF, so no predicates are enforced
   • DFilter predicates are ignored for the db_owner / administrator context, so highly privileged principals see all rowsCorrect
   • ERLS predicates only apply to service principals, never to interactive users
   ✓ Correct answer: A, D

   RLS in the Warehouse restricts standard principals, but members acting with administrator/db_owner-level rights (such as elevated workspace roles) are not filtered and see all rows. That is by design for high-privilege contexts. To restrict everyone you must test with a low-privilege principal.

   Why the other options are wrong

   • BOrdinary users with only SELECT are exactly who RLS filters; the statement is false.
   • CSTATE = OFF would disable enforcement, but the scenario says the admin specifically sees all rows while others presumably do not, so the admin bypass (A and D) is the correct explanation; C describes a different failure and is not selected here.
   • ERLS applies to interactive users too, not just service principals.
3. Question 3Implement and manage an analytics solution

   For a strong information-protection posture in Fabric, a security architect proposes: (1) enable sensitivity labels with mandatory labeling, (2) configure labels with encryption for the most sensitive tiers, and (3) enable label-based export protection. Which single overarching Microsoft service defines and centrally manages the labels and their protection settings used by all three?

   • AMicrosoft Purview Information ProtectionCorrect
   • BAzure Key Vault
   • CMicrosoft Defender for Cloud Apps
   • DAzure Monitor
   ✓ Correct answer: A

   Sensitivity labels, their protection (encryption) settings, and the policies that publish and enforce them (including mandatory labeling) are all authored and managed in Microsoft Purview Information Protection, which Fabric consumes for its information-protection features.

   Why the other options are wrong

   • BAzure Key Vault stores secrets and keys; it does not define sensitivity labels or their policies.
   • CDefender for Cloud Apps focuses on cloud app security and can act on labels but does not define the label taxonomy.
   • DAzure Monitor handles telemetry and alerting, not sensitivity label management.
4. Question 4Ingest and transform data

   You want to validate a large COPY INTO command's file access and formatting without actually inserting rows into the target. What is the correct way to test this in Fabric Warehouse?

   • ARun COPY INTO against a throwaway staging table (or a small file subset) and inspect results, since COPY INTO has no dry-run/validate-only modeCorrect
   • BAdd VALIDATION_MODE = 'RETURN_ERRORS' to perform a dry run
   • CAdd DRYRUN = TRUE to the WITH clause
   • DPrefix the statement with EXPLAIN to load nothing and report format errors
   ✓ Correct answer: A

   Fabric Warehouse COPY INTO does not provide a validate-only/dry-run mode. The practical approach is to load into a temporary or throwaway staging table (optionally with a small file subset and a low MAXERRORS) to confirm access and formatting before the real load.

   Why the other options are wrong

   • BVALIDATION_MODE = 'RETURN_ERRORS' is Snowflake syntax, not Fabric COPY INTO.
   • CDRYRUN is not a COPY INTO option.
   • DEXPLAIN does not turn COPY INTO into a no-op validation and is not used this way.
5. Question 5Ingest and transform dataSelect all that apply

   Which statements about Fabric Data Factory Copy activity STAGING are correct? (Choose 2.)

   • AStaging can enable performant bulk loads (e.g., PolyBase/COPY-style) into supported analytical sinksCorrect
   • BStaging is useful when the source is behind an on-premises data gateway and the sink cannot be loaded directlyCorrect
   • CStaging always reduces cost because it removes the need for any compute
   • DStaging is mandatory for every Lakehouse sink
   ✓ Correct answer: A, B

   Interim staging lands data in a staging store so the service can then perform an optimized bulk load into the sink, and it is the recommended approach when a gateway source cannot feed the sink directly.

   Why the other options are wrong

   • CStaging adds an intermediate copy and storage/compute cost; it does not remove compute or always reduce cost.
   • DStaging is optional, not mandatory; many Lakehouse loads run without it.
6. Question 6Monitor and optimize an analytics solution

   The Spark UI shows a stage with very low task parallelism: only 4 tasks are running though the pool has 32 available cores. The input is a single 3 GB gzip-compressed CSV file. Why is parallelism limited, and what is the fix?

   • AGzip is not splittable, so the file is read by a single task; convert/repartition to a splittable format (e.g. Parquet/Delta) or split the source so more tasks can read in parallelCorrect
   • BThe pool is misconfigured and must be recreated
   • CAQE is limiting parallelism; disable AQE
   • DResult-set caching is holding the cores
   ✓ Correct answer: A

   A single gzip-compressed file cannot be split across tasks because gzip is not a splittable codec, so only one task reads it regardless of available cores, leaving parallelism starved. Converting the data to a splittable columnar format like Parquet/Delta (or repartitioning after the initial read, or splitting the source into multiple files) lets many tasks read concurrently and use the full pool.

   Why the other options are wrong

   • BThe pool is fine; the limitation is the non-splittable input format, not pool configuration.
   • CAQE does not force single-task reads of a gzip file; the codec's non-splittability does.
   • DResult-set caching is a Warehouse feature and does not consume Spark cores.
7. Question 7Implement and manage an analytics solution

   In OneLake security (data access roles) for a lakehouse, a user is a member of a role that grants read on /Files/sales but is NOT granted any access to /Files/hr. When they browse the lakehouse, what should they experience for the hr folder?

   • AThey can read hr because workspace membership always overrides OneLake roles
   • BThey cannot read the hr folder data because they lack a role granting access to itCorrect
   • CThey can read hr but only for the first 1,000 rows
   • DOneLake automatically grants read to all folders once any role is assigned
   ✓ Correct answer: B

   With OneLake security enabled, users can access only the folders/paths their assigned data access roles grant. Lacking a role for /Files/hr, the user is denied access to that folder's data.

   Why the other options are wrong

   • AWhen OneLake security is enforced, it restricts data access even for members; it is not simply overridden by workspace role for granted paths.
   • CThere is no partial 1,000-row allowance from OneLake security.
   • DAssigning one role does not implicitly grant access to all other folders.
8. Question 8Implement and manage an analytics solution

   Business requirement: an operational database in Fabric must serve high-concurrency single-row lookups and writes from an app, while its data also becomes queryable for analytics in OneLake automatically. Which Fabric item best fits the operational side?

   • AA warehouse
   • BA SQL database in FabricCorrect
   • CA KQL database
   • DA Dataflow Gen2
   ✓ Correct answer: B

   SQL database in Fabric is built for transactional, high-concurrency operational workloads (row lookups and writes) and automatically replicates its data into OneLake as Delta so it is instantly queryable for analytics, matching both requirements.

   Why the other options are wrong

   • AThe warehouse is an analytical MPP store, not tuned for high-concurrency single-row OLTP writes.
   • CA KQL database targets append-heavy telemetry analytics, not transactional app writes.
   • DA Dataflow Gen2 is an ETL transform tool, not an operational database.
9. Question 9Ingest and transform data

   You must copy data from a source into a Fabric Warehouse where the Copy activity's sink can leverage COPY INTO under the hood for best load performance. Which sink type/setting enables this?

   • ASet the sink to the Warehouse and allow the Copy activity to use the COPY (bulk) load methodCorrect
   • BSet the sink to a CSV file in the Lakehouse Files area
   • CSet the sink to a Power BI semantic model
   • DSet the sink to a KQL database table
   ✓ Correct answer: A

   When the Copy activity sink is a Fabric Warehouse, it can use the COPY-based bulk load method (optionally with staging) to maximize load throughput into the Warehouse table.

   Why the other options are wrong

   • BA CSV file sink does not load into a Warehouse table via COPY INTO.
   • CA semantic model is a Power BI dataset, not a Warehouse bulk-load sink.
   • DA KQL database is not a Warehouse and does not use Warehouse COPY INTO.
10. Question 10Monitor and optimize an analytics solution

    In the Capacity Metrics app, you observe long red bars indicating the capacity was in an overloaded state, followed by 'carryforward' being applied to subsequent periods. What does carryforward represent?

    • AConsumption that exceeded the limit and is added to future timepoints until repaidCorrect
    • BUnused CUs that roll over as bonus capacity
    • CThe number of concurrent Spark sessions allowed
    • DA discount applied when a reservation is purchased
    ✓ Correct answer: A

    When smoothing spreads over-limit usage, the excess consumption is carried forward and added to later timepoints. If carryforward accumulates past thresholds, throttling stages engage. It represents debt, not bonus capacity.

    Why the other options are wrong

    • BUnused CUs are not banked as bonus capacity; carryforward is over-usage, not surplus.
    • CConcurrency limits are separate from carryforward accounting.
    • DReservation discounts are billing concepts, not CU carryforward.

DP-700 practice exam FAQ