These dp 600 sample questions are scenario-based items covering security and governance, data preparation, and semantic modeling, the three domains Microsoft publishes for Exam DP-600. Each question below shows the options, the correct answer, and an explanation of why that answer is right and why the closest wrong option is not, so you can check your reasoning, not just your final choice.
These dp 600 sample questions are scenario-based, covering all three DP-600 domains.
Question 1 (Prepare)
A hospital's telemetry data is stored in an Azure SQL database that is inside a virtual network and has no public endpoint. You need to connect Fabric to this database for a near-real-time dashboard. The dashboard requires the lowest possible latency, and you cannot install or maintain gateway software. What should you create?
- A. A VNet data gateway connected to the database's virtual network subnet.
- B. An on-premises data gateway installed on an Azure VM in the database's subnet.
- C. An on-premises data gateway installed on a Windows Server in the hospital.
- D. A Dataflow Gen2 that connects to Azure SQL using the database's public endpoint.
Answer: A
Explanation: The VNet data gateway option is correct because the Azure SQL database is in a virtual network with no public endpoint, and a managed VNet data gateway creates a private, low-latency path from Fabric to that subnet without requiring you to install software. The strongest distractor is the on-premises data gateway installed on an Azure VM in the subnet: it could technically reach the database, but it still requires installing, patching, and maintaining a gateway on a VM, which the scenario prohibits. A dataflow using the public endpoint cannot work because no public endpoint exists. The general rule is to use an on-premises data gateway for on-premises data sources and a VNet data gateway for Azure virtual-network data sources when you need a Microsoft-managed private connection.
Question 2 (Maintain)
Contoso Airlines' Flight Operations workspace contains the Flight Pay report and several sensitive operational dashboards. Union schedulers need to view and build new visuals from the Flight Pay report only. The report and its semantic model cannot be moved or copied because refresh and lineage are managed in this workspace. Schedulers must not be able to see any other workspace items. How should you grant access?
- A. Publish an app that contains only the Flight Pay report, and add the schedulers to the app audience for that report.
- B. Add the schedulers to the Flight Operations workspace as Viewers; this lets them open Flight Pay and prevents them from editing the other dashboards.
- C. Move the Flight Pay report to a new workspace that contains only that report, and add the schedulers as Viewers in the new workspace.
- D. Share the Flight Pay report directly with the schedulers and share its semantic model with Build permission so they can build new visuals.
Answer: D
Explanation: This is correct because the requirement is an item-level boundary: only one report and its semantic model are granted, while the report remains in the original workspace. Sharing the report and semantic model directly is the only option that combines Read on the report and Build on the semantic model without giving a workspace role. The strongest distractor is the workspace Viewer role: a Viewer can open the Flight Pay report and cannot edit it, but workspace roles are aggregate grants across every item in the workspace, so the schedulers would also see the sensitive dashboards, violating the only-this-report constraint. The rule is: use workspace roles when users need blanket access to a whole workspace, and use item-level Share/Manage permissions when access must be capped at one item or semantic model.
Question 3 (Model)
A public-sector agency stores incident records as Delta tables in a Fabric lakehouse. The agency estimates that importing those tables into an in-memory semantic model would require 150 GB. Reports must support row-level detail and return aggregations in under two seconds, and the solution must not create a second copy of the data or require scheduled refresh. Which storage mode should you choose?
- A. DirectQuery
- B. Direct Lake
- C. Composite model
- D. Import
Answer: B
Explanation: Direct Lake is correct because the source is Delta tables in OneLake, an Import copy would exceed the 100 GB limit for imported semantic models, and the requirement forbids a second copy and scheduled refresh. Direct Lake lets the analytics engine read the Delta tables directly from OneLake, so the model is not bound by the Import-mode size quota and does not require a refreshable copy. The strongest distractor is DirectQuery, which also avoids an imported copy but pushes queries to the source and cannot deliver the same high-performance, columnstore-style analysis over 150 GB of data that the scenario requires. The general rule is: use Direct Lake for large Delta data in OneLake that needs fast analysis, use DirectQuery when the source must be queried live or is outside OneLake, and use Import only when the model is small enough and a refreshable copy is acceptable.
Question 4 (Maintain)
An industrial IoT team has a Fabric workspace that powers a live operations dashboard. After a hotfix is merged into the main Git branch, the dashboard must show the corrected model within five minutes. You are configuring version control for this workspace. The solution must preserve full Git history and no one may run a workspace sync by hand. What should you configure?
- A. Connect the workspace to Git and use a scheduled Power Automate flow to call Update from Git at the top of every hour.
- B. Connect the workspace to Git and have the team run Update from Git in the Source Control panel after each hotfix is merged into main.
- C. Connect the workspace to Git and use a CI/CD pipeline to call the Fabric REST API to run Update from Git after each main merge.
- D. Connect the workspace to Git and use a scheduled automation that runs Update from Git 30 minutes after every main branch merge.
Answer: C
Explanation: This is correct because the CI/CD pipeline automatically performs the explicit Update from Git operation immediately after the merge, so the workspace receives the hotfix without a person starting the sync and within the five-minute window. The manual Update from Git in the Source Control panel option is the strongest distractor: it is the normal way to sync a Fabric workspace, but the scenario eliminates it because the sync must be started by a person. The scheduled options also preserve Git history, but their 30-minute or hourly cadence violates the five-minute requirement. The general rule is that Fabric Git integration is not an event-driven deployment; a connected workspace does not live-track a branch. If an SLA forces fast propagation, automate the Update from Git operation from a CI/CD pipeline.
Question 5 (Prepare)
A telecom operator stores call detail records in a Fabric Lakehouse table. New records arrive continuously, and the CallDurationSeconds column is STRING. Analysts need to query it as INTEGER through the SQL analytics endpoint for ad-hoc reports. The data must remain the single source of truth; the solution may not create a separate table or file and must minimize ongoing Fabric capacity usage. Which approach should you use?
- A. Run a CREATE TABLE AS SELECT in a Fabric Warehouse to cast CallDurationSeconds to Integer in a new table.
- B. Create a Dataflow (Gen2) that changes CallDurationSeconds to Integer in Power Query and loads the result into a new Lakehouse delta table.
- C. Create a Spark notebook that reads the Lakehouse table, casts CallDurationSeconds to Integer, and writes a new delta table in the Lakehouse.
- D. Create a SQL view in the Lakehouse's SQL analytics endpoint that uses TRY_CAST on CallDurationSeconds to expose it as INTEGER.
Answer: D
Explanation: The SQL analytics endpoint view option is correct because it projects CallDurationSeconds as INTEGER without copying the underlying data. TRY_CAST is evaluated when the view is queried, so there is no separate table or file and no refresh compute as new records arrive. The strongest distractor is the Dataflow (Gen2) option: changing a column data type in Power Query is a normal way to perform this transformation, but that approach materializes a second Lakehouse table and would need to be refreshed continuously, violating the no-copy and minimum-capacity constraints. The Warehouse CTAS and Spark table options fail the same way by creating stored copies. General rule: when a conversion is only needed for query-time consumption and the data must remain a single stored copy, use a projected view; materialize only when a physical table or precomputed performance is required.
Question 6 (Maintain)
A regulated financial services firm has a production Fabric workspace named 'Regulatory Reporting' that contains the reports, semantic models, and artifacts the new internal audit team must review. The workspace is part of a deployment pipeline, and reporting must continue without interruption. You cannot recreate the workspace or republish any item. The audit team must automatically see current and newly added workspace content. How should you grant access?
- A. Create a separate workspace, copy the reports, and grant the team Viewer access.
- B. Add the team to the existing production workspace using the Viewer workspace role.
- C. Publish a Fabric app from the workspace and grant the team access to that app.
- D. Share every report and semantic model in the workspace individually with the team.
Answer: B
Explanation: Adding the team with the Viewer role is correct because a workspace role grants the same read-only access to everything in the workspace, including items added later, and does not require recreating or republishing anything. The strongest distractor is the separate-workspace option: it looks like a cleaner security boundary for a regulated firm, but copying production reports would violate the constraint that the workspace cannot be recreated or republished, and a copy would not automatically inherit future items from the original workspace. The general rule is to use a workspace role when users need consistent access to all present and future workspace content; use item-level sharing or apps only when they need a curated subset and you are prepared to update the sharing or app membership as content changes.
Question 7 (Prepare)
At a healthcare provider, you must aggregate 120 daily clinical feeds and land each aggregate in its own Delta table in a Lakehouse. A single Dataflow Gen2 item would minimize cost and maintenance, but it fails because Dataflow Gen2 allows at most 50 output destinations per item. The nightly batch must finish by 06:00. What should you do?
- A. Use three Dataflow Gen2 items, distribute the 120 destinations, and run them in parallel.
- B. Increase the Fabric capacity SKU so one Dataflow item can publish all 120 destinations.
- C. Create one Dataflow Gen2 item per feed and run the items in parallel.
- D. Use one Dataflow item with a single Delta destination and expose aggregates as SQL views.
Answer: A
Explanation: The correct answer is to use multiple Dataflow Gen2 items because the 50-output-destination limit is a per-item ceiling: a single dataflow item cannot exceed it no matter how much capacity is available. The strongest distractor is the capacity-increase option, which treats the failure as an aggregate capacity problem; raising the SKU adds CU throughput but does not raise the number of destinations allowed in one dataflow item. The approach that writes to one Delta table and exposes views avoids the destination cap but does not produce the required 120 Delta tables. A separate item per feed is within the limit but conflicts with the stated cost and maintenance goal. The general rule is to separate limits that are global capacity limits from limits that apply per item, per partition, or per connection; for a per-item limit, scale by adding more items, not more capacity.
Question 8 (Model)
Your agency is creating a shared semantic model in Fabric so all Power BI reports use one version of its on-premises SQL Server data. The Database Administrator has confirmed that the warehouse has no public endpoint and is not in an Azure virtual network. You need to make the model refreshable in the Fabric service. What should you use to connect to the warehouse?
- A. An on-premises data gateway installed on a domain-joined server that can reach the warehouse.
- B. A Power BI template (.pbit) file with the semantic model definition and report layout.
- C. A virtual network data gateway attached to a subnet in the warehouse's Azure virtual network.
- D. A Power BI data source (.pbids) file that stores the warehouse connection for report authors.
Answer: A
Explanation: This is correct because the warehouse is private and on-premises, with no public endpoint and no presence in an Azure VNet, so the Fabric service needs an on-premises data gateway installed on a machine that can reach the warehouse. The strongest distractor is the virtual network data gateway: it is also a private-connection pattern, but it only works for data sources inside an Azure virtual network, so it is eliminated by the stated constraint. The .pbids file and .pbit file options are reusable assets, but they support Desktop authoring and would not make one central semantic model refreshable in the Fabric service. General rule: in Fabric and Power BI, cloud services reach on-premises private data sources through an on-premises data gateway, while VNet data gateways are only for data sources deployed in Azure VNets.
Question 9 (Prepare)
A SaaS startup stores tenant usage data in an on-premises SQL Server. A compliance contract requires that the usage data remains in that on-premises SQL Server, yet executives want near-real-time Power BI reports that combine it with data already in Fabric. The query is a simple filter on a wide fact table. What should you create?
- A. A view in the on-premises SQL Server that filters the fact table, and connect to it from Power BI through an on-premises data gateway using DirectQuery.
- B. A view in the on-premises SQL Server that filters the fact table, and use an on-premises data gateway in a Dataflow Gen2 to load it into a Lakehouse table.
- C. A stored procedure in the on-premises SQL Server whose result set is copied by a Fabric pipeline into a Fabric Warehouse, then connect to it using Direct Lake.
- D. An inline table-valued function in the on-premises SQL Server, and call it from a Fabric pipeline through a VNet data gateway into a Lakehouse table.
Answer: A
Explanation: Correct is the DirectQuery view option. The compliance rule says the data must stay on-premises, and the business needs near-real-time access, so the data must not be materialised in OneLake; DirectQuery sends queries to the view while leaving the data in the source. The strongest distractor is the Dataflow Gen2 option: it correctly uses a view and a gateway but then loads a copy into Lakehouse, which violates the stated compliance constraint. The stored-procedure and VNet gateway options are also wrong—the procedure copies data into a warehouse, and a VNet data gateway serves Azure VNet sources, not on-premises SQL Server. The general rule is: when a source must remain in place and the latency target requires live reads, use a DirectQuery-compatible object such as a view with an on-premises data gateway; reserve stored procedures and Dataflows for cases where copying is allowed.
Next steps
See our DP-600 practice test, study plan, or exam format.
Exam facts on this page come from Microsoft’s own pages (official exam page). Prices, versions and policies change — confirm there before you book.