lock_openFree sampleMaintain a Data Analytics Solutionhard
A "Warehouse" holds an Employees table whose Salary and BankAccount columns must be hidden from analysts, while those analysts still need to query every row and all other columns of the table. The masking must apply at the engine, not just in report visuals. Which access control should the engineer configure on the "Warehouse"?
- ACreate a row-level security policy that filters out the rows containing salary and bank account values for the analyst security predicate.
- BHide the Salary and BankAccount columns in the downstream "semantic model" so the analysts no longer see them when they build report visuals.
- CApply a sensitivity label to the Employees table that marks Salary and BankAccount as confidential and blocks the analysts from reading them.
- DGrant the analysts SELECT on the Employees table but DENY column-level permission on the Salary and BankAccount columns so those two columns are not returned.check_circle Correct
Use column-level security to deny specific sensitive columns at the engine while keeping the rest of a warehouse table queryable. Fabric Warehouse column-level security denies SELECT on named columns so the engine omits them from results regardless of the client, which hides Salary and BankAccount without removing rows or relying on report-layer hiding.
Why A is wrong: Row-level security removes whole rows by predicate, but the requirement keeps every row visible and only hides two columns, so a row filter is the wrong granularity here.
Why B is wrong: Hiding columns in the model is cosmetic and report-scoped, so analysts querying the "Warehouse" directly still read the values, failing the engine-level masking requirement.
Why C is wrong: A sensitivity label classifies and can govern export, but it does not strip specific columns from a query result, so it cannot enforce the column hiding described.
Why D is correct: Column-level security in a Fabric "Warehouse" denies access to named columns while leaving the rest of the table queryable, which precisely meets hiding two sensitive columns at the engine.
lock_openFree sampleMaintain a Data Analytics Solutionhard
An external auditing partner needs read access to a single Parquet folder of curated invoices in "OneLake", but must be blocked from every other folder in the same "Lakehouse" and must not be granted any workspace role. The team wants the narrowest possible grant scoped to that one folder. Which approach should the engineer take?
- AAdd the partner to the "Lakehouse" workspace as a Viewer so they inherit read access, then rely on the partner only opening the invoices folder.
- BDefine an OneLake data access role scoped to the invoices folder path and assign the partner to that role so access is limited to that folder only.check_circle Correct
- CBuild row-level security over the invoices table so the partner's predicate returns only invoice rows and no rows from other lakehouse tables.
- DApply a confidential sensitivity label to the other folders so the partner is prevented from opening anything except the invoices folder in the lakehouse.
Use OneLake data access roles to grant folder-level file read on a specific path without assigning a broad workspace role. OneLake data access roles enforce file-level security by binding a role to specific folder paths in a lakehouse, so a partner can read one invoices folder and be denied the rest without any workspace membership.
Why A is wrong: A Viewer workspace role grants read across the whole item and other workspace content, which is far broader than the single folder the requirement permits.
Why B is correct: OneLake data access roles grant folder-level read on a specific path without any workspace role, giving the partner exactly the one invoices folder and nothing else.
Why C is wrong: Row-level security governs query results through the SQL endpoint or model, not raw file access in "OneLake", so it cannot scope file-system read to one folder.
Why D is wrong: Sensitivity labels classify content and influence export controls, but they do not grant or deny file read on folders, so they cannot enforce this folder-scoped access.
lock_openFree sampleMaintain a Data Analytics Solutionmedium
A governance lead must ensure that confidential figures inside a "Power BI" "semantic model" stay protected even after a report consumer exports the data to an Excel workbook saved outside "Microsoft Fabric". The protection should travel with the file and enforce access at the document level. Which capability should the lead apply to the "semantic model"?
- AApply a sensitivity label backed by "Microsoft Purview" Information Protection, so the encryption and usage rights persist on any file exported from the item.check_circle Correct
- BPromote the "semantic model" so consumers see it flagged as a trusted, ready-to-use item when they browse the workspace and OneLake catalogue.
- CCertify the "semantic model" through the tenant endorsement process so a reviewer attests that its figures are authoritative and approved for reuse.
- DEnable row-level security roles on the "semantic model" so each consumer only queries the rows their role is permitted to read.
Use a Purview-backed sensitivity label, not endorsement, to apply persistent document-level encryption that follows data exported out of Fabric. Sensitivity labels integrate with Microsoft Purview Information Protection to embed encryption and usage rights in the file itself, so the protection persists in exported Office documents regardless of where the file is later stored or opened.
Why A is correct: A sensitivity label with protection enforces encryption and rights that follow the content into exported Office files, which is exactly the document-level protection the requirement demands.
Why B is wrong: Promotion only signals trust and discoverability inside the tenant and adds no encryption, so an exported workbook would carry no protection once it leaves the service.
Why C is wrong: Certification is an endorsement that vouches for quality, but it applies no encryption and does nothing to a file once it has been exported out of the service.
Why D is wrong: Row-level security filters rows during live queries inside the model but applies no protection to an exported workbook, so confidential figures would sit unencrypted in the file.
lock_openFree sampleMaintain a Data Analytics Solutionmedium
A team wants to move tested "Power BI" reports and semantic models from a test workspace to a production workspace in a controlled, repeatable way, comparing what differs between the two before each release and promoting with one action. They are not trying to track every file change in source control. Which "Microsoft Fabric" capability best fits promoting content across these stages?
- AConnect each workspace to a Git repository through Git integration and merge branches so that production receives the tested content through pull requests.
- BUse a deployment pipeline with development, test, and production stages to compare the test and production content and promote it in a single action.check_circle Correct
- CExport every report as a .pbix file from test and manually re-upload each one into the production workspace whenever a release is approved.
- DGrant production viewers direct edit access to the test workspace so they consume the tested reports in place without any promotion step at all.
Choose deployment pipelines, not Git integration, when the goal is staged comparison and one-action promotion of content across development, test, and production. Deployment pipelines bind development, test, and production workspaces to stages and surface a diff before promoting items in a single step, which addresses staged release rather than the file-level versioning that Git integration provides.
Why A is wrong: Git integration version-controls workspace items and suits change tracking, but the team explicitly does not want source-control workflows for moving content between stages, so it overshoots the need.
Why B is correct: Deployment pipelines stage workspaces, show what differs between stages, and promote selected content in one action, which is precisely the controlled stage-to-stage release the team wants.
Why C is wrong: Manual export and re-upload is error-prone, gives no stage comparison, and is not the repeatable one-action promotion the requirement describes.
Why D is wrong: Sharing the test workspace blurs the boundary between environments and skips promotion entirely, so production never receives its own governed copy of the content.
lock_openFree sampleMaintain a Data Analytics Solutionmedium
A paginated dataflow refresh is driven by a parameter called RowLimit that is set to a small value in development for fast iteration. When the content reaches production through a deployment pipeline, the production stage must always run with RowLimit set to its full value, regardless of what development holds. How should the engineer enforce the production value on every deployment?
- AChange RowLimit to its full value in the development stage so the larger value simply carries forward unchanged into the production stage.
- BAdd a data source rule on the production stage that remaps the source connection, assuming it also resets the RowLimit parameter to production scale.
- CCreate a parameter rule on the production stage that sets RowLimit to its full production value so every deployment applies that value automatically.check_circle Correct
- DOpen the production item after each deployment and type the full RowLimit value by hand before triggering the scheduled refresh.
Apply a parameter rule on the target stage to force a named parameter to its stage-specific value automatically on each deployment. Parameter rules let a pipeline override the value of a declared parameter for a specific target stage, so the production value is reapplied on every deployment without depending on what earlier stages contain.
Why A is wrong: Raising the development value abandons the fast-iteration setup and still relies on the value carrying forward, which a later development change could silently undo.
Why B is wrong: Data source rules rebind connections, not parameter values, so a data source rule cannot set RowLimit and would leave the parameter at its development value.
Why C is correct: A parameter rule on the target stage overrides a named parameter's value during deployment, so RowLimit is forced to the full production value on every promotion regardless of the development setting.
Why D is wrong: Editing the parameter by hand after every deployment is the manual repetition that parameter rules are designed to eliminate, and it is easy to forget.