CertSafari

    Free Microsoft Certified: Fabric Data Engineer Associate (DP-700) Sample Questions

    35 free sample questions from our bank of 347+, covering every exam domain, with answers and detailed explanations. Updated September 2026.

    Practice all 347 questions in quiz mode

    Domain 1: Implement and manage an analytics solution

    Subdomain 1.4: Orchestrate processes

    1.An operations team wants a nightly pipeline to keep running indefinitely with no planned end. When they open the pipeline's schedule configuration, they find they cannot leave the end date blank. What should they do to achieve an effectively open-ended schedule?

    1. A.Set the schedule's end date far into the future, such as 01/01/2099, since Fabric schedules always require both a start and an end date.
    2. B.Leave the end date field empty and save the schedule, since Fabric interprets a blank end date as an open-ended recurrence.
    3. C.Configure an interval-based schedule instead, since interval-based schedules support truly open-ended, unbounded recurrence windows.
    4. D.Attach an event-based trigger in addition to the schedule, since combining both removes the end-date requirement for the fixed schedule.
    Show answer & explanation

    Correct answer: A — Set the schedule's end date far into the future, such as 01/01/2099, since Fabric schedules always require both a start and an end date.

    • A. Setting the end date far in the future is correct because Fabric requires both a start and end date for a fixed schedule, with no option for a truly open-ended schedule.
    • B. Leaving the end date blank is incorrect because Fabric does not accept an empty end date; a start and end date must both be specified when configuring a schedule.
    • C. Interval-based schedules still require configured, non-overlapping fixed intervals and are not a way to bypass the end-date requirement of a fixed schedule.
    • D. Adding an event-based trigger runs alongside a schedule but does not change or remove the end-date requirement on the fixed schedule itself.

    Subdomain 1.4: Orchestrate processes

    2.An engineer is comparing the legacy and newer preview versions of the Invoke pipeline activity while designing a multi-workspace orchestration pattern. Which of the following are accurate capabilities or configuration options relevant to this comparison? (Select all that apply)(Select 4)

    1. A.The newer preview Invoke pipeline activity can invoke pipelines across Fabric workspaces, including those originating from ADF or Synapse.
    2. B.The legacy Invoke pipeline activity can only monitor the parent pipeline's status, not the invoked child pipeline's status.
    3. C.Both the legacy and preview Invoke pipeline activities support choosing an existing pipeline or creating a new one from the activity.
    4. D.The invoked pipeline can be configured to run and complete before the parent continues, or to run in parallel with later activities.
    5. E.The legacy Invoke pipeline activity automatically upgrades itself to the preview version the first time it successfully runs.
    6. F.Only the legacy Invoke pipeline activity supports waiting for the child pipeline's completion before the parent proceeds.
    Show answer & explanation

    Correct answers: A, B, C, D — The newer preview Invoke pipeline activity can invoke pipelines across Fabric workspaces, including those originating from ADF or Synapse.; The legacy Invoke pipeline activity can only monitor the parent pipeline's status, not the invoked child pipeline's status.; Both the legacy and preview Invoke pipeline activities support choosing an existing pipeline or creating a new one from the activity.; The invoked pipeline can be configured to run and complete before the parent continues, or to run in parallel with later activities.

    • A. This is correct: the newer preview version extends Invoke pipeline to call pipelines across Fabric workspaces and even pipelines originating from ADF or Synapse.
    • B. This is correct: the legacy activity can only monitor the parent pipeline and cannot report on or invoke ADF or Synapse pipelines like the preview version can.
    • C. This is correct: both versions let the author pick an existing pipeline from a dropdown or create a new one directly from the activity configuration.
    • D. This is correct: the activity can be set to wait on completion of the invoked pipeline, or to continue so the invoked pipeline runs in parallel with subsequent activities.
    • E. This is incorrect: there is no automatic upgrade behavior; choosing between legacy and preview is a manual activity selection made by the pipeline author.
    • F. This is incorrect: waiting for completion versus continuing in parallel is a configuration option available on the activity generally, not exclusive to the legacy version.

    Subdomain 1.1: Configure Microsoft Fabric workspace settings

    3.A workspace admin enabled shortcut caching with a 7-day retention period for a lakehouse's external shortcuts. A file behind one shortcut is accessed once on day 1 and then not touched again until day 6. According to how the cache retention period behaves, what happens to that file's cached copy?

    1. A.The file remains cached because each access resets the retention countdown, so accessing it on day 6 restarts the 7-day window from that point
    2. B.The file is evicted from the cache automatically at the end of day 7 regardless of the access on day 6, since retention counts from the original cache time
    3. C.The file is evicted after day 1 plus 7 days because only the first access into the cache counts toward the retention period
    4. D.The file is never evicted automatically; only a manual Reset cache action removes files once caching has been enabled for a shortcut
    Show answer & explanation

    Correct answer: A — The file remains cached because each access resets the retention countdown, so accessing it on day 6 restarts the 7-day window from that point

    • A. The retention period is described as resetting on each access to the file, so accessing the file again on day 6 restarts the countdown from that point rather than letting the original day-1 window expire on day 8.
    • B. This describes a fixed expiry from the original cache time, but the retention period is reset by access rather than being a fixed one-time countdown, so the file would not be evicted at the end of day 7 given the day 6 access.
    • C. This also treats the countdown as fixed from the first access, but the documented behavior is that every access resets the retention period, not just the initial one.
    • D. Reset cache is a manual action that clears all cached files on demand, but the retention period setting also drives automatic eviction after the configured number of days without access, so eviction is not manual-only.

    Subdomain 1.1: Configure Microsoft Fabric workspace settings

    4.A workspace admin is enabling workspace identity so that both Airflow DAGs and other automation can authenticate to Fabric items without stored credentials. Which of the following Fabric item types are listed as supported for authentication via workspace identity in Apache Airflow jobs? (Choose 3)(Select 3)

    1. A.Lakehouses
    2. B.Semantic model refresh
    3. C.Pipelines
    4. D.Power BI report subscriptions
    5. E.Domain settings configuration
    Show answer & explanation

    Correct answers: A, B, C — Lakehouses; Semantic model refresh; Pipelines

    • A. Correct. Lakehouses are explicitly listed as a Fabric service that workspace identity supports authenticating to from Airflow jobs.
    • B. Correct. Semantic model refresh is explicitly listed among the Fabric services that workspace identity can authenticate to and trigger from an Airflow DAG.
    • C. Correct. Pipelines are explicitly listed as a supported Fabric service for workspace identity authentication in Airflow jobs.
    • D. Power BI report subscriptions are an email/notification delivery feature for reports, not one of the Fabric item types listed as supported for workspace identity authentication from Airflow.
    • E. Domain settings configuration is an administrative action performed by domain or Fabric admins in the admin portal; it is not a Fabric item type that an Airflow DAG authenticates to and runs.

    Subdomain 1.2: Implement lifecycle management in Fabric

    5.Which statements about deploying content through a Fabric deployment pipeline are accurate? (Select all that apply)(Select 3)

    1. A.Deploying through the pipeline UI updates the target stage's workspace content but does not update an associated app; the app must be updated separately through the deployment pipelines API
    2. B.Items that exist in the source stage but are unpaired with the target stage are created as new duplicate copies rather than overwriting anything
    3. C.A pipeline can have anywhere from two to ten stages, with three stages provided as the default starting point
    4. D.Deploying to a stage automatically deletes any item in that stage that does not also exist in the source stage
    5. E.Renaming a paired item in one stage breaks its pairing with the corresponding item in the adjacent stage
    Show answer & explanation

    Correct answers: A, B, C — Deploying through the pipeline UI updates the target stage's workspace content but does not update an associated app; the app must be updated separately through the deployment pipelines API; Items that exist in the source stage but are unpaired with the target stage are created as new duplicate copies rather than overwriting anything; A pipeline can have anywhere from two to ten stages, with three stages provided as the default starting point

    • A. Correct. Deployment through the UI copies workspace content only; updating an associated app's content or settings requires a separate manual step through the deployment pipelines API.
    • B. Correct. When source-stage content has no existing pairing in the target stage, deployment creates a new duplicate copy there instead of overwriting an existing item.
    • C. Correct. Pipelines support two to ten stages, and three stages are the default starting point when a pipeline is first created.
    • D. Incorrect. Items that exist only in the target stage and not in the source stage remain untouched by the deployment; they are not automatically deleted.
    • E. Incorrect. Pairing is preserved across renames, so a paired item can be given a different display name without losing its pairing with the adjacent stage's item.

    Subdomain 1.2: Implement lifecycle management in Fabric

    6.A report named `Sales Overview` in the Development stage of a pipeline was already deployed once and is paired with its Test-stage counterpart. Afterward, the Development-stage report is renamed to `Regional Sales Overview`. What happens the next time the team deploys from Development to Test?

    1. A.The paired item in Test is still matched to it internally and gets overwritten, even though the display names no longer match.
    2. B.Fabric treats the renamed item as a new report, so the deployment creates a duplicate copy alongside the original Sales Overview item already in Test.
    3. C.The deployment fails with a display-name conflict error and blocks publishing until someone renames the Test report to match Development.
    4. D.The pairing between the two items is broken by the rename, so it must be manually re-established using the deployment pipeline's compare view.
    Show answer & explanation

    Correct answer: A — The paired item in Test is still matched to it internally and gets overwritten, even though the display names no longer match.

    • A. Correct. Paired items remain paired even after a rename, so the two items can have different display names and the deployment still overwrites the existing paired copy instead of duplicating it.
    • B. Incorrect. A duplicate copy is only created when items were never paired in the first place, not because an already-paired item was renamed.
    • C. Incorrect. Fabric does not require display names to match for paired items, so a rename does not trigger a naming conflict error during deployment.
    • D. Incorrect. Pairing persists automatically through renames and does not need to be manually reset in the compare view.

    Subdomain 1.1: Configure Microsoft Fabric workspace settings

    7.A data engineering team's notebooks frequently spend three minutes waiting for a Spark session to start because an administrator switched the workspace default pool from the automatically created starter pool to a custom pool sized for a batch workload. The team wants the fast five-to-ten-second session start back for everyday interactive notebook work, without giving up the ability to size nodes for the batch workload later. What should the workspace admin do?

    1. A.Set the workspace default pool back to the Starter Pool, and keep the custom Spark pool available so notebooks can select it when larger node sizing is needed.
    2. B.Reduce the custom pool's maximum node count to one and set it as the workspace default so single-node sessions start in five to ten seconds, matching the starter pool's cold-start time.
    3. C.Disable the Customize compute configurations for items toggle so every notebook automatically inherits the capacity SKU's baseline Spark compute settings across the workspace.
    4. D.Enable dynamic allocation of executors on the custom pool so Spark scales the executor count down to zero between interactive notebook runs and back up on the next query.
    Show answer & explanation

    Correct answer: A — Set the workspace default pool back to the Starter Pool, and keep the custom Spark pool available so notebooks can select it when larger node sizing is needed.

    • A. The starter pool is a prehydrated live pool sized for the capacity SKU, which is exactly why it gives the fast 5-10 second start; switching the workspace default back to it restores that experience while the custom pool remains available for items that need specific node sizing.
    • B. Shrinking the custom pool to a single node still requires provisioning a fresh cluster on each session start, so it does not reproduce the prehydrated starter pool experience and would undersize the batch workload the pool was created for.
    • C. That toggle only controls whether individual items can override session-level driver and executor sizing; it does not change which pool (starter or custom) is used as the workspace default, so it would not restore the fast start.
    • D. Dynamic allocation adjusts the number of executors within a running session based on data volume, but it does not change cluster provisioning time, so sessions on the custom pool would still take longer to start than the starter pool.

    Subdomain 1.3: Configure security and governance

    8.A finance warehouse table stores customer phone numbers. The requirement is that a masked query should still show the area code but hide the remaining digits behind a fixed placeholder, for example turning 555.123.1234 into 555.1XXXXXXX. Which masking function meets this requirement?

    1. A.The default masking function, which fully replaces the entire phone number column with a fixed value regardless of its original content.
    2. B.The email masking function, which formats the phone number as a partial address ending in a constant '.com' suffix, ignoring its digits.
    3. C.The custom string masking function, which exposes a configurable prefix and suffix while replacing the middle with a fixed padding string.
    4. D.The random masking function, which substitutes the phone number with a random numeric value drawn from configured minimum and maximum range.
    Show answer & explanation

    Correct answer: C — The custom string masking function, which exposes a configurable prefix and suffix while replacing the middle with a fixed padding string.

    • A. Default masking replaces the whole value uniformly, such as turning it into repeated X characters or a fixed string, so it cannot preserve the area code portion as required.
    • B. Email masking always formats the result as a single-letter-plus-domain email pattern, which doesn't apply to a phone number and cannot preserve a partial numeric prefix.
    • C. Custom string masking, defined with a prefix and suffix length plus a padding string, is exactly the function used to keep the area code visible while masking the remaining digits.
    • D. Random masking substitutes numeric columns with a randomly generated value in a range, which hides the entire number rather than preserving a visible area code.

    Subdomain 1.2: Implement lifecycle management in Fabric

    9.A finance team wants a deployment pipeline with four stages instead of three: Development, Test, UAT, and Production. Is this configuration possible in Fabric?

    1. A.Yes, a pipeline can have between two and ten stages, and stages can be added, deleted, or renamed from the default three.
    2. B.No, Fabric deployment pipelines are fixed at exactly three stages named Development, Test, and Production.
    3. C.Yes, but only if the fourth stage beyond three is created through the deployment pipelines REST API rather than the portal.
    4. D.No, a fourth stage is only possible by chaining two separate three-stage pipelines together end to end.
    Show answer & explanation

    Correct answer: A — Yes, a pipeline can have between two and ten stages, and stages can be added, deleted, or renamed from the default three.

    • A. Correct. A pipeline supports between two and ten stages; three stages are provided as a default starting point, but stages can be added, removed, or renamed to fit a team's release process.
    • B. Incorrect. Three stages are only the default; the stage count and names are configurable within the supported two-to-ten range.
    • C. Incorrect. Additional stages can be added directly through the portal when creating or editing the pipeline; the API isn't a requirement for going beyond three stages.
    • D. Incorrect. Chaining separate pipelines is unnecessary since a single pipeline natively supports up to ten stages.

    Subdomain 1.4: Orchestrate processes

    10.A workspace needs one workflow that copies raw files from an on-premises source, waits for that copy to finish, branches into a different path if the copy activity fails, then calls a notebook and a Dataflow Gen2 in sequence. Which orchestration item should coordinate this end-to-end control flow?

    1. A.A pipeline, because its activities, dependencies, and conditional paths are designed to sequence and branch across heterogeneous steps.
    2. B.A Dataflow Gen2, because its query dependency graph can sequence the on-premises copy, branch on failure, and invoke downstream notebook queries.
    3. C.A notebook, because a single Spark session can call the on-premises copy job, branch on error, and chain the Dataflow Gen2 transformation through library imports.
    4. D.An Eventhouse update policy, because it can chain the incoming copy, branch on failure, and apply a downstream notebook transformation automatically.
    Show answer & explanation

    Correct answer: A — A pipeline, because its activities, dependencies, and conditional paths are designed to sequence and branch across heterogeneous steps.

    • A. A pipeline is correct because it is the item built for orchestration: activity dependencies, success/failure branching, and invoking notebooks or Dataflow Gen2 as steps in a controlled sequence.
    • B. Dataflow Gen2 focuses on data transformation queries, not on branching control flow or invoking other items conditionally, so it cannot coordinate this workflow.
    • C. A notebook executes Spark code and is not designed to orchestrate conditional branching across unrelated items like copy activities and Dataflow Gen2.
    • D. An Eventhouse update policy transforms incoming rows within Real-Time Intelligence and has no concept of branching pipeline-style control flow.

    Subdomain 1.3: Configure security and governance

    11.A data governance team applies a sensitivity label with an associated Microsoft Purview protection policy to a lakehouse. A user later exports the lakehouse's data to a .csv file and shares that file outside the organization. Which statement correctly describes the label's effect in this scenario?

    1. A.The label's access control does not extend to the .csv export, so the exported file carries no built-in protection from the Fabric sensitivity label.
    2. B.The label automatically re-applies its protection policy inside the exported .csv file, restricting who can open it outside the organization.
    3. C.Fabric blocks the export operation entirely, since any item carrying a sensitivity label is permanently barred from export to CSV or any other format.
    4. D.The label converts the .csv export into an encrypted Power BI Desktop file so external recipients must authenticate before opening it.
    Show answer & explanation

    Correct answer: A — The label's access control does not extend to the .csv export, so the exported file carries no built-in protection from the Fabric sensitivity label.

    • A. Fabric access control from sensitivity labels applies within the tenant and to a limited set of supported export paths; exporting to a .csv file isn't one of the supported paths, so the exported file leaves without the label's protection.
    • B. Protection policies don't automatically travel into unsupported export formats like .csv; only certain export paths such as Excel or PDF carry the label and its access control forward.
    • C. Fabric doesn't universally block exports of labeled items; it instead issues a warning in some cases while still allowing the export to proceed for unsupported protection paths.
    • D. Exporting to .csv doesn't convert the file into a Power BI Desktop file format, and Fabric doesn't automatically wrap arbitrary exports in a new file type with authentication.

    Subdomain 1.3: Configure security and governance

    12.A sales lakehouse table stores order records for every region in one company. Regional managers should each see only the rows for their own region when they query the table through the SQL analytics endpoint, and this restriction must hold regardless of which reporting tool they use. Which control should be configured on the table?

    1. A.Configure row-level security within a OneLake security role so that a predicate filters visible rows to each manager's own region for every supported query engine.
    2. B.Configure dynamic data masking on the region column so unauthorized managers see a masked placeholder value instead of the true region name across all returned rows.
    3. C.Create a separate workspace per region and grant each regional manager the Contributor role only in the workspace holding that region's copy of the table.
    4. D.Apply a sensitivity label to the table that restricts access to users belonging to the same Microsoft Entra region-based security group as the data owner.
    Show answer & explanation

    Correct answer: A — Configure row-level security within a OneLake security role so that a predicate filters visible rows to each manager's own region for every supported query engine.

    • A. Row-level security defined inside a OneLake security role filters which rows a member can see based on a predicate, and this filtering is enforced consistently across the engines that support it, including the SQL analytics endpoint.
    • B. Dynamic data masking hides column values, not rows, so a masked region column would still return every order row to every manager instead of limiting rows to their own region.
    • C. Splitting the table across regional workspaces would require duplicating and maintaining the data in multiple places instead of applying row-level filtering to one shared table.
    • D. Sensitivity labels classify and can encrypt items as a whole, but they don't filter individual rows of a table based on a manager's region.

    Subdomain 1.1: Configure Microsoft Fabric workspace settings

    13.A workspace runs mixed workloads: lightweight PySpark transformations that finish in minutes, and an occasional large batch job that needs more executors. The admin wants the batch job to acquire additional nodes automatically while it runs and release them when it finishes, without manually resizing the pool before and after each run. Which custom Spark pool setting should the admin enable?

    1. A.Autoscale, so the pool acquires nodes up to the configured maximum and releases them when done.
    2. B.High concurrency mode, so multiple notebooks share one running Spark session and its allocated nodes.
    3. C.Set maximum job lifetime, so Fabric terminates the job once it exceeds the duration limit set.
    4. D.Automatic logging for machine learning experiments, so resource usage is captured for review.
    Show answer & explanation

    Correct answer: A — Autoscale, so the pool acquires nodes up to the configured maximum and releases them when done.

    • A. Autoscale is the setting that lets a custom pool acquire new nodes within the specified maximum while a job runs and retire them after execution completes, which is exactly the elastic behavior described.
    • B. High concurrency lets several notebooks share one Spark session for efficiency, but it does not add or remove nodes based on a single job's changing resource demand.
    • C. Maximum job lifetime is a guardrail that cancels a job once it runs longer than a configured duration; it stops runaway jobs but does not scale compute up or down during execution.
    • D. Autologging captures parameters, metrics, and output items for machine learning training runs; it is an observability feature and has no effect on node acquisition or release.

    Subdomain 1.3: Configure security and governance

    14.A team builds a Power BI report on top of a Fabric lakehouse. The lakehouse has a sensitivity label applied, and separately a Power BI dataset elsewhere in the tenant also carries its own label. Which downstream inheritance behavior is accurate by default?

    1. A.The lakehouse's label propagates downstream to its Power BI report, but a Power BI item's label doesn't propagate into Fabric items.
    2. B.Both the lakehouse's label and the Power BI dataset's label propagate freely in both directions between Fabric items and Power BI items.
    3. C.Neither label propagates downstream to any dependent item unless an administrator manually reapplies the label on every new dependent item.
    4. D.Only the Power BI dataset's label propagates downstream, while a Fabric item's label never propagates to any other item, Fabric or Power BI.
    Show answer & explanation

    Correct answer: A — The lakehouse's label propagates downstream to its Power BI report, but a Power BI item's label doesn't propagate into Fabric items.

    • A. Downstream inheritance supports propagation from a Fabric item to another Fabric item and from a Fabric item to a Power BI item, but propagation from a Power BI item back into a Fabric item is not supported.
    • B. Inheritance is directional rather than bidirectional; a Power BI item's label doesn't flow back into Fabric items even though a Fabric item's label can flow into Power BI items.
    • C. Downstream inheritance is enabled by default and applies automatically to dependent items without requiring an administrator to manually reapply the label each time.
    • D. Fabric item labels do propagate downstream to dependent Fabric and Power BI items; it's specifically the reverse direction, from Power BI into Fabric, that isn't supported.

    Domain 2: Ingest and transform data

    Subdomain 2.2: Ingest and transform batch data

    15.A team configures Fabric mirroring against an on-premises SAP source. Because SAP stores data in a proprietary format that OneLake engines cannot read directly, how does mirroring make the SAP data available in OneLake?

    1. A.Mirroring continuously replicates the source data into OneLake as analytics-ready Delta tables, since the proprietary format cannot be read in place
    2. B.Mirroring creates a shortcut that reads the SAP storage files directly, since shortcuts can parse any proprietary source format without any conversion step at all
    3. C.Mirroring pauses replication and instead schedules a nightly Copy activity to move the SAP tables into the lakehouse on a fixed cadence
    4. D.Mirroring exposes the SAP tables only as external T-SQL objects through PolyBase, leaving the underlying rows unreplicated in OneLake
    Show answer & explanation

    Correct answer: A — Mirroring continuously replicates the source data into OneLake as analytics-ready Delta tables, since the proprietary format cannot be read in place

    • A. When the source format is proprietary, mirroring uses the replication mechanism, copying catalog metadata and continuously writing the data into OneLake as Delta tables so it becomes analytics-ready.
    • B. Shortcut-based metadata mirroring is reserved for sources that already store data in an open format OneLake can read directly, such as Databricks or Snowflake; SAP's proprietary format rules this path out.
    • C. Mirroring is a continuous, turnkey replication capability, not a scheduled Copy activity; describing it as a nightly copy job misstates how the feature operates.
    • D. Mirroring is not implemented through PolyBase external tables, and it does replicate rows into OneLake rather than leaving them unreplicated behind a query layer.

    Subdomain 2.2: Ingest and transform batch data

    16.A gold-layer lakehouse table must combine product data from a manufacturing system and a marketing system, joined by a shared SKU, and roll it up into a single wide table so downstream Power BI reports avoid multi-table joins. Which action is being performed here?

    1. A.Denormalizing the data, since related attributes from multiple normalized sources are combined into one wide table for simpler downstream queries
    2. B.Mirroring the data, since two operational systems are being continuously replicated into a single shared catalog without any transformation applied
    3. C.Creating a shortcut, since the SKU field is being used to reference the marketing system's storage location without copying any rows
    4. D.Applying change data capture, since the goal is to detect which SKU rows changed between the manufacturing and marketing systems
    Show answer & explanation

    Correct answer: A — Denormalizing the data, since related attributes from multiple normalized sources are combined into one wide table for simpler downstream queries

    • A. Combining related attributes from separate normalized sources on a shared key into one wide table is the definition of denormalization, done here to simplify downstream reporting joins.
    • B. Mirroring surfaces an external database or catalog as-is in OneLake; it does not describe joining and flattening two sources' columns into a new combined table.
    • C. A shortcut only references data at its source location; it does not perform a join across two systems or produce a new combined, wide table.
    • D. Change data capture is a technique for identifying changed rows in a source over time; it is not the mechanism used to combine columns from two systems into one table.

    Subdomain 2.2: Ingest and transform batch data

    17.A workspace's items, including pipelines, notebooks, and Dataflow Gen2 definitions, are connected to an Azure DevOps Git repository for version control. After a developer commits a change and it is deployed from the Dev stage to the Test stage of a deployment pipeline, what happens to the data stored in the workspace's lakehouses?

    1. A.The item definitions are promoted to Test, but the lakehouse data itself is not copied; only the notebook, pipeline, and dataflow definitions move
    2. B.Every row of lakehouse data is copied from the Dev stage to the Test stage automatically as part of that same deployment pipeline stage promotion step
    3. C.The deployment pipeline deletes all Dev-stage data once Test receives the promoted definitions, to avoid storing duplicate copies
    4. D.Git tracks and version-controls the lakehouse data files themselves, so the Test stage receives a Git-committed snapshot of the data
    Show answer & explanation

    Correct answer: A — The item definitions are promoted to Test, but the lakehouse data itself is not copied; only the notebook, pipeline, and dataflow definitions move

    • A. Fabric's Git integration and deployment pipelines track item definitions such as notebooks, pipelines, and dataflows; data stays in OneLake and is not copied between stages when definitions are promoted.
    • B. Deployment pipeline stage promotion does not copy lakehouse data; it only carries over item definitions, so describing an automatic full data copy misstates the behavior.
    • C. Promoting definitions to Test has no effect on Dev-stage data; deployment pipelines do not delete source data as part of promoting item definitions.
    • D. Git tracks item definitions, not the underlying data files in OneLake; lakehouse data is not committed to or restored from the Git repository.

    Subdomain 2.2: Ingest and transform batch data

    18.OneLake shortcuts physically copy the referenced files into OneLake storage at the moment the shortcut is created, so the source system's storage account can be deleted immediately afterward without any data loss.

    1. A.True
    2. B.False
    Show answer & explanation

    Correct answer: B — False

    • A. This statement is false: a shortcut is a reference to data that stays at its source, not a physical copy, so deleting the source storage would remove the data the shortcut points to.
    • B. The statement is false. Shortcuts add a reference to selected data in the OneLake namespace while the data itself remains at the source; no copy is created at shortcut-creation time, so removing the source breaks the shortcut.

    Subdomain 2.3: Ingest and transform streaming data

    19.An Eventhouse references rarely queried archival data through a OneLake shortcut, and the team's priority is minimizing cost rather than query speed. What should they do with the shortcut?

    1. A.Keep the shortcut as a standard, non-accelerated reference, since occasional queries do not justify the OneLake Premium cache cost.
    2. B.Enable query acceleration on the shortcut so every query benefits from Eventhouse-level indexing regardless of how rarely it runs.
    3. C.Mirror the archival source into the Eventhouse so the data is fully replicated and always available without any external calls.
    4. D.Ingest the archival data into a native Eventhouse table on a nightly schedule to guarantee the fastest possible query times.
    Show answer & explanation

    Correct answer: A — Keep the shortcut as a standard, non-accelerated reference, since occasional queries do not justify the OneLake Premium cache cost.

    • A. Correct. Leaving the shortcut as a standard, non-accelerated reference avoids the OneLake Premium cache cost, which is appropriate when queries against the archival data are infrequent.
    • B. Incorrect. Query acceleration adds caching and indexing cost that is wasted on data that is rarely queried, working against the stated cost priority.
    • C. Incorrect. Mirroring replicates the entire source continuously and adds storage and compute cost that is unnecessary for rarely accessed archival data.
    • D. Incorrect. A nightly ingestion job adds ongoing pipeline maintenance and storage duplication that is not justified for infrequently queried data.

    Subdomain 2.3: Ingest and transform streaming data

    20.A clickstream analysis must group each user's events into a session that closes after 30 minutes of inactivity, regardless of how long the session actually runs. Which windowing approach should be used?

    1. A.Apply a session window with a thirty-minute timeout so a session closes after that period of inactivity.
    2. B.Apply a tumbling window with a thirty-minute size so every session boundary aligns to a fixed clock interval.
    3. C.Apply a hopping window with a thirty-minute size and a five-minute hop to detect inactivity between clicks.
    4. D.Apply a sliding window keyed on event count instead of elapsed time to approximate session boundaries.
    Show answer & explanation

    Correct answer: A — Apply a session window with a thirty-minute timeout so a session closes after that period of inactivity.

    • A. Correct. A session window groups events by periods of activity and closes the session once a 30-minute gap of inactivity occurs, matching the scenario exactly.
    • B. Incorrect. A tumbling window closes on a fixed clock interval regardless of activity, which would split or merge sessions incorrectly.
    • C. Incorrect. A hopping window advances on a fixed overlapping schedule and does not close based on an inactivity gap.
    • D. Incorrect. A sliding window reacts to changes in window contents, not to elapsed inactivity time, so it cannot define a 30-minute session boundary.

    Subdomain 2.3: Ingest and transform streaming data

    21.A Spark Structured Streaming job uses the following code: ```python query = (df.writeStream .format("delta") .outputMode("append") .trigger(processingTime="1 minute") .start("Tables/bronze_events")) ``` After a manual restart, the job reprocesses data it had already written instead of resuming cleanly. Which setting is missing from this code?

    1. A.The `.option("checkpointLocation", ...)` setting, which Spark needs to persist offsets and resume exactly where it left off.
    2. B.The `.outputMode("append")` setting, which should instead be set to `"complete"` for this delta sink to work correctly.
    3. C.The `.trigger(processingTime="1 minute")` setting, which should instead be replaced with a `once` trigger for this scenario.
    4. D.The `.format("delta")` setting, which should instead be set to `"parquet"` since delta does not support streaming writes.
    Show answer & explanation

    Correct answer: A — The `.option("checkpointLocation", ...)` setting, which Spark needs to persist offsets and resume exactly where it left off.

    • A. Correct. Without a `checkpointLocation`, Spark has no persisted record of offsets or progress, so a restart has no way to know what was already written and reprocesses data.
    • B. Incorrect. `append` mode is appropriate for writing new streaming rows to a Delta table and is not the cause of the reprocessing behavior described.
    • C. Incorrect. The processing-time trigger controls how often micro-batches run, not whether the job can resume state after a restart.
    • D. Incorrect. Delta format fully supports streaming writes and is the correct sink format here; it is not the source of the reprocessing issue.

    Subdomain 2.3: Ingest and transform streaming data

    22.A data engineering team wants to compute a rolling anomaly score over an event stream using a custom Python machine learning library and stateful logic across events, then persist the results to a lakehouse table. Which engine should they choose?

    1. A.Use Spark Structured Streaming in a Fabric notebook with a stateful transformation, checkpointing results to the lakehouse table.
    2. B.Use an Eventstream with the no-code event processor's aggregate operator to compute the rolling anomaly score for each event.
    3. C.Use a KQL update policy on the raw ingestion table to transform every incoming row into a fully computed anomaly score column value.
    4. D.Use a Dataflow Gen2 with a Power Query custom function scheduled every five minutes to recompute the anomaly score.
    Show answer & explanation

    Correct answer: A — Use Spark Structured Streaming in a Fabric notebook with a stateful transformation, checkpointing results to the lakehouse table.

    • A. Correct. Spark Structured Streaming in a notebook supports custom Python libraries and stateful transformations, with a checkpoint location ensuring the job resumes correctly against the lakehouse sink.
    • B. Incorrect. The no-code aggregate operator only supports built-in functions like sum, average, minimum, and maximum, not custom Python ML logic.
    • C. Incorrect. A KQL update policy runs a KQL transformation query, not arbitrary Python machine learning code, and targets Eventhouse tables rather than a lakehouse.
    • D. Incorrect. A five-minute scheduled Dataflow Gen2 is batch-oriented and does not support stateful, per-event streaming computation with a custom ML library.

    Subdomain 2.1: Design and implement loading patterns

    23.A `Customer` dimension tracks a customer's `LoyaltyTier` attribute. The business wants historical sales reports to always reflect the loyalty tier the customer held at the time of each sale, even after the customer's tier changes later. Which change-handling strategy satisfies this requirement?

    1. A.Type 2: expire the current dimension row version and insert a new versioned row whenever the tier changes.
    2. B.Type 1: simply overwrite the existing dimension row's tier value whenever the source tier changes specifically.
    3. C.Type 3: add a `PreviousLoyaltyTier` column that stores only the one immediately prior value specifically in this case.
    4. D.Leave the attribute completely unmanaged so the source system's latest value always overwrites the dimension row directly.
    Show answer & explanation

    Correct answer: A — Type 2: expire the current dimension row version and insert a new versioned row whenever the tier changes.

    • A. Slowly changing dimension type 2 preserves history by expiring the prior row version and inserting a new row for each change, and fact rows loaded before the change keep pointing at the surrogate key for the old version. This is exactly what lets historical reports show the loyalty tier that applied at the time of each sale.
    • B. Type 1 overwrites the attribute in place with no versioning, so every historical fact row that joins to this dimension member would immediately show the customer's current tier instead of the tier at the time of sale. This does not meet the requirement to preserve point-in-time history.
    • C. Type 3 only retains the single most recent prior value in an extra column, so it cannot reconstruct the tier at an arbitrary point further back in history. It falls short of full historical accuracy for every past sale.
    • D. Letting the source value flow through unmanaged behaves like an uncontrolled type 1 overwrite, with no mechanism to preserve any historical version of the attribute. Past sales would report the customer's current tier rather than their tier at the time of the sale.

    Subdomain 2.1: Design and implement loading patterns

    24.A reviewer is checking a newly implemented dimension load against the `Customer` dimension's SCD type 2 requirement. Which of the following correctly describe how the change should be processed? (Choose 2.)(Select 2)

    1. A.Expire the current row version by setting an end-date column and a current-flag column to false now.
    2. B.Insert a new row for the changed member, with its start date set to the prior version's end date.
    3. C.Update the existing row's attributes in place without inserting any additional rows at all specifically.
    4. D.Delete the previous version's row entirely so only the newest version remains in the table.
    5. E.Add a new column to the table for every possible historical value the attribute could ever take.
    Show answer & explanation

    Correct answers: A, B — Expire the current row version by setting an end-date column and a current-flag column to false now.; Insert a new row for the changed member, with its start date set to the prior version's end date.

    • A. Correct. Processing an SCD type 2 change starts by expiring the current version: the end-date validity column is set to the ETL processing date and the current-flag column is set to false, marking that version as no longer active.
    • B. Correct. After expiring the old version, a new row is inserted for the changed member with its start date aligned to the prior version's end date and its current-flag set to true, which is what creates the new active version while preserving the old one for history.
    • C. Incorrect. Updating the existing row's attributes in place with no new row inserted is the SCD type 1 pattern, which overwrites history rather than preserving it, and does not satisfy an SCD type 2 requirement.
    • D. Incorrect. SCD type 2 processing keeps the previous version's row rather than deleting it; deleting it would destroy exactly the historical record the versioning pattern is meant to preserve.
    • E. Incorrect. Adding a dedicated column per possible historical value is neither the type 2 nor any standard SCD pattern; type 2 preserves history through additional rows, not additional columns.

    Subdomain 2.1: Design and implement loading patterns

    25.Statement: An Eventstream permanently stores every event it processes, so historical events can always be queried directly from the Eventstream item itself, even without routing the data to any destination.

    1. A.True
    2. B.False
    Show answer & explanation

    Correct answer: B — False

    • A. True is incorrect. Eventstreams stream events through memory and forward them to configured destinations; they do not retain events for later querying on the Eventstream item itself.
    • B. False. To make streaming events queryable over time, they must be routed to a destination such as an Eventhouse; the Eventstream itself is a pass-through processing layer, not a storage layer.

    Domain 3: Monitor and optimize an analytics solution

    Subdomain 3.2: Identify and resolve errors

    26.Which of the following statements about monitoring an eventhouse's system overview page in Microsoft Fabric are accurate? (Select all that apply.)(Select 3)

    1. A.The Top 10 ingested databases tile reports ingested row counts and ingestion errors, though it currently reports only partial ingestion errors rather than a fully complete count.
    2. B.The Eventhouse schema changes section only tracks database creation and deletion events, and it does not surface any table-level or materialized-view-level changes.
    3. C.The Activity in minutes by application chart always shows every application that queried the eventhouse individually, with no grouping or aggregation of lower-volume applications.
    4. D.Several system overview tiles, including Ingestion and Top 10 queried databases, expose an Open query option revealing the exact underlying KQL query used to compute that tile.
    5. E.The eventhouse storage breakdown separates the original uncompressed size from the compressed size, and reports a Premium storage figure representing the portion cached in the high-performance tier.
    Show answer & explanation

    Correct answers: A, D, E — The Top 10 ingested databases tile reports ingested row counts and ingestion errors, though it currently reports only partial ingestion errors rather than a fully complete count.; Several system overview tiles, including Ingestion and Top 10 queried databases, expose an Open query option revealing the exact underlying KQL query used to compute that tile.; The eventhouse storage breakdown separates the original uncompressed size from the compressed size, and reports a Premium storage figure representing the portion cached in the high-performance tier.

    • A. This matches documented behavior: the Top 10 ingested databases tile highlights ingested rows and ingestion errors, and the documentation explicitly notes that currently only partial ingestion errors are reported.
    • B. This understates documented behavior: the Eventhouse schema changes section also tracks creating, altering, or deleting tables, external tables, materialized views, and functions, not only database creation and deletion.
    • C. This overstates documented behavior: the Activity in minutes by application chart shows the top five applications by activity volume and groups the remaining applications under an Other category rather than listing every application individually.
    • D. This matches documented behavior: tiles such as Ingestion, Top 10 queried databases, Top 10 ingested databases, and Activity in minutes by top users each expose an Open query option in their ellipsis menu.
    • E. This matches documented behavior: the eventhouse storage tile separates original size from compressed size and reports the amount of Premium, high-performance-tier storage utilized.

    Subdomain 3.2: Identify and resolve errors

    27.For each of the following statements, select Yes if the statement is true. Otherwise, select No. Statement: A stored procedure in a Fabric Warehouse can safely wrap a cross-database update to two warehouses in the same workspace inside a single BEGIN DISTRIBUTED TRANSACTION block, because distributed transactions are fully supported for coordinating commits across multiple Fabric warehouses.

    1. A.Yes
    2. B.No
    Show answer & explanation

    Correct answer: B — No

    • A. Marking this true would contradict the documented limitation that distributed transactions, including BEGIN DISTRIBUTED TRANSACTION, are explicitly not supported in Fabric Data Warehouse.
    • B. This is correct: Fabric Data Warehouse's documented limitations state that distributed transactions are not supported, so wrapping a cross-warehouse update in one is not a safe or valid pattern.

    Subdomain 3.3: Optimize performance

    28.An application performs thousands of individual single-row `INSERT` statements against a warehouse table throughout the day, and query latency against that table has degraded over time. What is the recommended fix?

    1. A.Pre-stage the rows and batch them into fewer, larger `INSERT` statements (or use `COPY INTO` for bulk loads) instead of issuing many small transactions.
    2. B.Increase the warehouse's retention period so more historical Parquet file versions remain available for time-travel queries against the table.
    3. C.Switch every `INSERT` to run inside its own explicit `BEGIN TRANSACTION`/`COMMIT` block to guarantee snapshot isolation for each row.
    4. D.Add a clustered index on the table's primary key column so single-row inserts are routed directly to the correct file without scanning.
    Show answer & explanation

    Correct answer: A — Pre-stage the rows and batch them into fewer, larger `INSERT` statements (or use `COPY INTO` for bulk loads) instead of issuing many small transactions.

    • A. Each small INSERT/UPDATE/DELETE generates a new Parquet file, so batching writes into larger transactions (or using `COPY INTO`) reduces file fragmentation and the resulting query latency.
    • B. Increasing retention only affects how long old file versions are kept for time travel; it does not reduce fragmentation from ongoing small-row inserts and can increase storage use.
    • C. Fabric Data Warehouse already wraps every statement in snapshot-isolated transactions; wrapping each single-row insert explicitly does not change the file fragmentation caused by their small size.
    • D. Primary key and unique key constraints are not enforced in Fabric Data Warehouse, and traditional clustered indexes are not the mechanism that controls file layout here.

    Subdomain 3.3: Optimize performance

    29.Which of the following statements correctly describe how Eventhouse data tiers and query acceleration behave? (Select all that apply.)(Select 3)

    1. A.Data ingested into an Eventhouse is written into both a premium SSD-backed hot cache tier and standard storage.
    2. B.The cache period on a caching policy can be tuned per table, with a configurable range rather than a single fixed value for every table.
    3. C.A query acceleration policy on a OneLake shortcut caches data as it lands, aiming for performance close to natively ingested Eventhouse data.
    4. D.Once data ages out of the hot cache tier, it becomes permanently unqueryable and must be re-ingested to run any further queries against it.
    5. E.Query acceleration is only available for native Eventhouse tables and cannot be applied to OneLake shortcuts pointing at external data.
    Show answer & explanation

    Correct answers: A, B, C — Data ingested into an Eventhouse is written into both a premium SSD-backed hot cache tier and standard storage.; The cache period on a caching policy can be tuned per table, with a configurable range rather than a single fixed value for every table.; A query acceleration policy on a OneLake shortcut caches data as it lands, aiming for performance close to natively ingested Eventhouse data.

    • A. Eventhouse ingestion writes data into OneLake cache storage on premium SSD as well as standard storage, giving it both a fast hot tier and durable standard storage.
    • B. The cache period for a table's caching policy is configurable per table, with a wide allowed range rather than one fixed value applied globally.
    • C. Query acceleration caches shortcut data as it lands in OneLake specifically to approach the query performance of data ingested directly into the Eventhouse.
    • D. Data that ages out of the SSD hot cache remains queryable from standard storage, just with slower query performance, rather than becoming permanently inaccessible.
    • E. Query acceleration is specifically designed for OneLake shortcuts, letting shortcut-referenced data approach native ingestion performance; it is not restricted to native Eventhouse tables only.

    Subdomain 3.3: Optimize performance

    30.A Spark job writes a gold-layer table that is consumed almost exclusively by a Power BI Direct Lake semantic model, with minimal downstream Spark or SQL analytics endpoint traffic. What should you do to best serve this consumer?

    1. A.Enable V-Order on the Spark-written table (or the `readHeavyForPBI` profile) so its write-time sorting and compression reduce transcoding overhead for Direct Lake.
    2. B.Leave V-Order disabled and rely only on adaptive target file size, since V-Order provides no measurable benefit for any Fabric consumer beyond the SQL analytics endpoint.
    3. C.Partition the table by report slicer columns so Direct Lake can prune entire partitions instead of relying on any write-time file optimization.
    4. D.Convert the table to a mirrored database so Fabric manages its Delta layout and compaction automatically, replacing Spark-side V-Order tuning entirely.
    Show answer & explanation

    Correct answer: A — Enable V-Order on the Spark-written table (or the `readHeavyForPBI` profile) so its write-time sorting and compression reduce transcoding overhead for Direct Lake.

    • A. V-Order's write-time sort and compression reduces the transcoding work Direct Lake performs loading Parquet data into its in-memory format, which matters most for a mostly Direct Lake-consumed table.
    • B. V-Order specifically improves Direct Lake and other read-oriented consumption; guidance recommends against enabling it solely for Spark or SQL analytics endpoint performance, not that it has no benefit anywhere.
    • C. Partitioning on report-specific slicer columns is not the recommended lever for Direct Lake performance and can fragment the table without addressing the transcoding overhead V-Order targets.
    • D. Converting to a mirrored database changes the entire ingestion pattern and is not a substitute for enabling V-Order on an existing Spark-written gold table serving Direct Lake.

    Subdomain 3.2: Identify and resolve errors

    31.A team creates a same-tenant OneLake-to-OneLake shortcut and wants each downstream user to see only the rows they are individually permitted to see at the source lakehouse, without the team having to manage a separate set of permissions on the shortcut itself. Which authentication model should they choose, and why?

    1. A.Passthrough authentication, because it forwards the querying user's own identity to the target location, so the source system's existing access controls apply directly without needing to be replicated.
    2. B.Delegated authentication, because it uses one fixed connection identity for the shortcut, and that identity automatically inherits the union of every downstream user's individual permissions granted at the source lakehouse.
    3. C.Passthrough authentication, because it grants every downstream user who can open the shortcut full read access to all tables in the target lakehouse, regardless of their individual row-level or column-level permissions at the source.
    4. D.Delegated authentication, because only delegated shortcuts support OneLake workspace security roles for the target lakehouse, and passthrough shortcuts cannot have any row-level or column-level filtering applied to the underlying tables.
    Show answer & explanation

    Correct answer: A — Passthrough authentication, because it forwards the querying user's own identity to the target location, so the source system's existing access controls apply directly without needing to be replicated.

    • A. This matches documented behavior: in the passthrough model, the shortcut passes the user's own identity to the target system, so any user can only see data they already have access to there, with no need to redefine access controls.
    • B. Delegated authentication uses a single fixed connection identity for all downstream users rather than resolving and combining each individual user's own permissions, so it does not produce per-user filtering automatically the way passthrough does.
    • C. Passthrough authentication is documented as enforcing the target system's existing access controls per user, not granting blanket full access regardless of each user's actual permissions at the source.
    • D. Same-tenant OneLake shortcuts can use either passthrough or delegated authentication, and passthrough shortcuts are not documented as incompatible with row-level or column-level filtering enforced at the source system itself.

    Subdomain 3.3: Optimize performance

    32.A Spark notebook reads a large DataFrame once and then reuses it across five separate downstream transformations and actions, each of which currently re-reads and re-computes the source data from scratch. What should you do to reduce this redundant work?

    1. A.Cache or persist the DataFrame after the initial read so subsequent transformations reuse the materialized result instead of recomputing it from the source each time.
    2. B.Increase the number of executor cores allocated to the session, since more cores automatically eliminate the need to recompute a DataFrame across every downstream action.
    3. C.Rewrite each downstream transformation as a separate notebook so they run in parallel Spark sessions instead of sharing one session's execution plan.
    4. D.Disable adaptive query execution for the session, since AQE forces Spark to re-evaluate the full DataFrame lineage on every action by default.
    Show answer & explanation

    Correct answer: A — Cache or persist the DataFrame after the initial read so subsequent transformations reuse the materialized result instead of recomputing it from the source each time.

    • A. Caching or persisting a DataFrame after its first computation stores the result in memory or on disk, so later actions reuse it instead of recomputing the full lineage each time.
    • B. Adding executor cores increases parallel processing capacity but does not stop Spark's lazy evaluation from recomputing an uncached DataFrame's lineage on every subsequent action.
    • C. Splitting the work into separate notebooks and sessions adds overhead and loses any in-memory reuse rather than solving the redundant recomputation problem.
    • D. Adaptive query execution optimizes shuffle and join strategies at runtime; it does not control whether an uncached DataFrame's lineage is recomputed across separate actions.

    Subdomain 3.1: Monitor Fabric items

    33.A data engineer wants to be alerted the instant a specific event type appears anywhere in a live event stream, without writing a KQL query or building a dashboard first, using the fastest built-in path available for a stream already surfaced in Real-Time hub. What is the correct entry point for setting that alert?

    1. A.Open the stream's detail page in Real-Time hub, then select the Set alert action there
    2. B.Open the Monitoring Hub Activities page and select Configure notifications for the stream
    3. C.Open the source Eventstream item's Git integration pane and add an alert branch policy
    4. D.Open the workspace's Domain settings page and enable the built-in stream alerting toggle
    Show answer & explanation

    Correct answer: A — Open the stream's detail page in Real-Time hub, then select the Set alert action there

    • A. Real-Time hub's stream detail page has a Set alert action that is the documented, no-code entry point for creating a Data Activator alert on that stream.
    • B. Configure notifications on the Activities page manages scheduled-item failure emails, not per-event alerting on a live streaming data feed.
    • C. Git integration manages version control for workspace items and has no relationship to setting a data-driven alert on a stream's events.
    • D. Domain settings organize workspaces for governance purposes and do not contain a stream alerting toggle.

    Subdomain 3.1: Monitor Fabric items

    34.A workspace lead needs to quickly find every job that failed in the finance workspace within the last day, out of dozens of jobs running across several workspaces the lead has access to. Rather than scrolling through the whole Activities table, which combination of Monitoring Hub filters narrows the list to exactly that set?

    1. A.Status = Failed, Location = the finance workspace, Start time = Last 24 hours
    2. B.Item type = Pipeline only, Submitted by = the lead's own account, Start time = Last 7 days
    3. C.Status = In progress, Location = all workspaces, Start time = Last 24 hours
    4. D.Keyword search for the word 'finance' with no Status or Start time filter applied
    Show answer & explanation

    Correct answer: A — Status = Failed, Location = the finance workspace, Start time = Last 24 hours

    • A. Setting Status to Failed, Location to the finance workspace, and Start time to the last 24 hours is the exact combination documented for finding recent failures in one workspace.
    • B. Restricting to pipelines submitted by the lead's own account would miss failures from other item types or other users, and misses the actual failure filter entirely.
    • C. In progress status shows currently running jobs, not jobs that already failed, so this filter set answers a different question than the one asked.
    • D. A keyword search on a workspace name only matches loaded rows containing that text and does not filter by status or time at all, so it would miss most relevant failures.

    Subdomain 3.1: Monitor Fabric items

    35.A platform engineer is choosing which Data Activator reflex action to configure for a new alert and wants to know the realistic menu of choices. Which of the following are documented action types available once a Data Activator trigger condition fires? (Choose 3.)(Select 3)

    1. A.Send an email directly to the person who needs to know about the trigger
    2. B.Post a message into a Teams channel so the wider team can see it
    3. C.Run a Fabric item such as a pipeline, a notebook, or a Spark job
    4. D.Automatically roll back the last Git commit made to the workspace item
    5. E.Directly rewrite the sensitivity label applied to the triggering item
    Show answer & explanation

    Correct answers: A, B, C — Send an email directly to the person who needs to know about the trigger; Post a message into a Teams channel so the wider team can see it; Run a Fabric item such as a pipeline, a notebook, or a Spark job

    • A. Sending an email is one of the documented actions a triggered reflex can take to notify someone directly.
    • B. Posting a Teams message is one of the documented actions a triggered reflex can take to notify a channel.
    • C. Running a Fabric item lets the reflex directly execute a pipeline, notebook, or Spark job in response to the trigger.
    • D. There is no documented reflex action that performs a Git rollback; source control operations are not part of Data Activator's action set.
    • E. There is no documented reflex action that edits sensitivity labels; label assignment is a Purview governance function, not a Data Activator action.

    Want the full experience?

    These are just samples. Practice the full Microsoft Certified: Fabric Data Engineer Associate (DP-700) question bank in quiz mode — free, no signup, with domain practice and exam simulation.

    Practice all 347 questions in quiz mode