Microsoft Fabric interview questions usually cover six areas: OneLake and shortcuts, Lakehouse vs Warehouse, pipelines and Dataflows Gen2, Spark notebooks, Real-Time Intelligence, and security, capacity and CI/CD. Many of these topics map to the DP-700 Fabric Data Engineer exam.
What interview panels ask about Microsoft Fabric
Microsoft Fabric brings data integration, engineering, warehousing, real-time analytics and Power BI into one SaaS platform built on OneLake. Since Microsoft retired DP-203 and made DP-700 the data engineer certification, Fabric questions have moved from “nice to have” to a standard part of Azure data engineering interviews.
Panels test three things. Can you explain the architecture — OneLake, workspaces, capacities and items? Can you choose the right item — Lakehouse or Warehouse, pipeline or Dataflow Gen2, notebook or SQL? And can you run it in production — security, monitoring, capacity and deployment?
If you have worked with Synapse or Databricks, many concepts carry over: Spark, Delta tables and medallion layers. Focus on what is different in Fabric. For the rest of the interview, see our Azure Data Engineer interview questions, and to learn Fabric hands-on, join our Azure Data Engineer Training in Hyderabad. Official references: Microsoft Fabric documentation on Microsoft Learn and the official DP-700 study guide.
🏞️ Seven sections, Q1 to Q32, each with an answer, a real-time example and an interview tip. Practise Microsoft Fabric interview questions with live labs.
Book a free demoMicrosoft Fabric and OneLake basics
Start here. Every Fabric interview opens with these concepts.
Q1. What is Microsoft Fabric?
🌱 Fresher · In simple words: Fabric is one Microsoft platform where you bring in, store, transform and report on data — no separate tools to connect.
Answer. Microsoft Fabric is an end-to-end, software-as-a-service analytics platform. It brings Data Factory, Data Engineering, Data Warehouse, Data Science, Real-Time Intelligence and Power BI experiences together on one shared storage layer, OneLake, with common security, governance and capacity-based billing.
Real-time example. A company that used ADF, Synapse and Power BI separately moves to Fabric so that ingestion, transformation and reports share one copy of the data.
Interview tip. Use the words “SaaS” and “OneLake” in the first sentence.
Q2. What is OneLake?
🌱 Fresher · In simple words: OneLake is one shared data lake for the whole company — think of it as “OneDrive for data”.
Answer. OneLake is a single, logical data lake for the whole organisation, provided automatically with every Fabric tenant. It is built on ADLS Gen2, stores tabular data in the open Delta Parquet format, and is organised by workspaces and items.
Real-time example. A Lakehouse table written by a Spark notebook is immediately readable by the Warehouse SQL engine and Power BI, without copying it.
Interview tip. Say “one copy of data, many engines” — that is the core idea.
Q3. What are OneLake shortcuts?
🌱 Fresher · In simple words: A shortcut shows data from another place inside Fabric without copying it.
Answer. Shortcuts are references that make data stored elsewhere appear inside a Lakehouse or KQL database without copying it. Targets can be other OneLake locations or external storage such as ADLS Gen2, Amazon S3 and Google Cloud Storage.
Real-time example. An existing ADLS Gen2 data lake is exposed through a shortcut, so Fabric notebooks read it in place during a phased migration.
Interview tip. Contrast shortcuts (no copy) with mirroring (replicated copy).
Q4. What is mirroring in Fabric?
🚀 Experienced · In simple words: Mirroring keeps a live copy of a database inside OneLake automatically.
Answer. Mirroring continuously replicates data from operational databases — for example Azure SQL Database, Azure Cosmos DB or Snowflake — into OneLake in Delta format in near real time, so it can be analysed without building ETL pipelines.
Real-time example. An order system’s Azure SQL Database is mirrored so analysts query near-real-time data without loading the production database.
Interview tip. Mention that it reduces custom CDC pipeline work.
Q5. What are workspaces and capacities?
🌱 Fresher · In simple words: A workspace is like a project folder. A capacity is the compute power — and the bill — behind it.
Answer. A workspace is a container for Fabric items such as lakehouses, warehouses, pipelines, notebooks and reports, with its own access roles. A capacity is the pool of compute (measured in capacity units, sold as F SKUs) that the workspace runs on.
Real-time example. Dev, test and prod workspaces are assigned to different capacities so heavy development work does not slow production reports.
Interview tip. Know that capacity size drives both performance and cost.
Lakehouse vs Warehouse interview questions
The most common Fabric design question is which storage item to choose.
Q6. What is a Fabric Lakehouse?
🌱 Fresher · In simple words: A Lakehouse holds both files and tables, and you work on it mainly with Spark notebooks.
Answer. A Lakehouse stores both files and Delta tables in OneLake. You load and transform data mainly with Spark notebooks, pipelines and Dataflows Gen2, and every Lakehouse automatically gets a read-only SQL analytics endpoint for T-SQL queries.
Real-time example. Raw JSON files land in the Files area and Spark notebooks turn them into Delta tables in the Tables area.
Interview tip. Mention the SQL analytics endpoint is read-only — a common follow-up.
Q7. What is a Fabric Warehouse?
🌱 Fresher · In simple words: A Warehouse is a SQL-style store where you can insert, update and delete data with T-SQL.
Answer. A Warehouse is a relational data warehouse that supports full T-SQL — DDL and DML, including INSERT, UPDATE, DELETE and multi-table transactions — while storing data as Delta tables in OneLake.
Real-time example. A finance team that works mainly in SQL builds its star schema and stored procedures in a Warehouse.
Interview tip. Say the deciding factor is mainly skills and workload: Spark-first vs T-SQL-first.
Q8. When do you choose Lakehouse vs Warehouse?
🌱 Fresher · In simple words: Spark or Python team, or messy files → Lakehouse. SQL team needing full updates → Warehouse. Many projects use both.
Answer. Choose a Lakehouse for semi-structured or unstructured data, Spark and Python transformation, and data science work. Choose a Warehouse when the team is T-SQL focused and needs full DML and multi-table transactions. Many solutions use both, for example Lakehouses for Bronze and Silver and a Warehouse for Gold.
Real-time example. Engineers clean data in Lakehouses with PySpark, and BI developers build the Gold star schema in a Warehouse with T-SQL.
Interview tip. Give a “both” answer with a reason — it shows design maturity.
Q9. What is Direct Lake mode?
🚀 Experienced · In simple words: Direct Lake lets Power BI read Delta tables straight from OneLake, so reports stay fresh without long refreshes.
Answer. Direct Lake is a Power BI semantic model storage mode that reads Delta tables directly from OneLake. It avoids scheduled import refreshes while giving performance close to import mode, and falls back to DirectQuery (querying the source live, which is slower) in some situations.
Real-time example. A sales report reads Gold Delta tables through Direct Lake, so new data appears soon after the pipeline finishes without a long dataset refresh.
Interview tip. Mention fallback behaviour to show depth.
Q10. What is V-Order?
🚀 Experienced · In simple words: V-Order arranges Parquet files while writing them, so reports can read them faster.
Answer. V-Order is a write-time optimisation for Parquet files that applies special sorting, encoding and compression so Fabric engines such as Power BI and SQL can read the data faster. The files remain standard, open Parquet.
Real-time example. Gold tables are written with V-Order because they are read many times by reports.
Interview tip. Say it trades slightly slower writes for faster reads.
Q11. How do you maintain Lakehouse Delta tables?
🚀 Experienced · In simple words: Regularly join small files and remove old files so your tables stay fast.
Answer. Run table maintenance — OPTIMIZE to compact small files (optionally with V-Order and Z-Order) and VACUUM to remove old unreferenced files — from the Lakehouse UI, notebooks or scheduled pipelines.
Real-time example. A streaming table that writes many small files is optimised every night.
Interview tip. Link small files to slow queries — it is the reason maintenance exists.
Data Factory in Fabric: pipelines, Dataflows Gen2 and notebooks
These questions test how you ingest and transform data inside Fabric.
Q12. Pipeline vs Dataflow Gen2 — what is the difference?
🌱 Fresher · In simple words: Pipelines move data and control the order of steps. Dataflows Gen2 clean data with clicks instead of code.
Answer. A data pipeline orchestrates activities such as Copy, Notebook, Stored Procedure, ForEach and If Condition, and is best for moving data at scale and controlling flow. Dataflow Gen2 is a low-code Power Query transformation tool that writes results to destinations such as a Lakehouse or Warehouse.
Real-time example. A pipeline copies raw files, runs a notebook and then refreshes a Dataflow Gen2 that business users maintain.
Interview tip. Say: pipelines orchestrate, dataflows transform with low code, notebooks transform with code.
Q13. How do you implement incremental loading in a Fabric pipeline?
🌱 Fresher · In simple words: Load only the rows that changed since the last run, using a saved “last loaded” time.
Answer. Store a watermark (for example the last modified timestamp) in a control table, use a Lookup activity to read it, copy only rows newer than the watermark, MERGE them into the target, and update the watermark after success. Copy job also offers incremental copy options for supported sources.
Real-time example. A nightly pipeline loads only orders changed since the previous run instead of the full table.
Interview tip. Mention that the watermark is updated only after a successful load.
Q14. How do you use Spark notebooks in Fabric?
🌱 Fresher · In simple words: Notebooks are where you write PySpark or SQL code, and pipelines can run them with inputs such as a date.
Answer. Fabric notebooks run Spark with PySpark, Spark SQL, Scala or SparkR, attach to a default Lakehouse, and can be scheduled or called from pipelines with parameters. Environments manage libraries and Spark settings.
Real-time example. A parameterised notebook receives load_date from a pipeline and processes one day of data, which makes backfills easy.
Interview tip. Mention environments for library management — it shows production experience.
Q15. How do you run Spark efficiently in Fabric?
🚀 Experienced · In simple words: Use the right Spark pool and share sessions, so jobs start quicker and use less capacity.
Answer. Use starter pools for quick session start, custom pools when you need specific node sizes, high-concurrency sessions to share a session across notebooks, and the usual Spark practices: filter early, avoid unnecessary shuffles, and optimise table layout.
Real-time example. High-concurrency mode lets several notebooks in a pipeline share one Spark session and reduces start-up time.
Interview tip. Connect Spark choices to capacity consumption.
Q16. How do you build a medallion architecture in Fabric?
🌱 Fresher · In simple words: Keep raw, cleaned and report-ready data in separate Lakehouses — Bronze, Silver and Gold.
Answer. Create separate Lakehouses (often in separate workspaces) for Bronze, Silver and Gold. Ingest raw data into Bronze, clean and conform it into Silver with notebooks or dataflows, and model Gold for reporting in a Lakehouse or Warehouse, orchestrated by pipelines.
Real-time example. Gold tables are consumed by Power BI through Direct Lake.
Interview tip. Explain why separate workspaces help with security and deployment.
Real-Time Intelligence interview questions
Real-time questions are increasingly common, and they are part of the DP-700 skills measured.
Q17. What is Real-Time Intelligence in Fabric?
🌱 Fresher · In simple words: Real-Time Intelligence is Fabric’s toolkit for live, streaming data.
Answer. Real-Time Intelligence is the Fabric workload for streaming and event data. It includes Eventstream for ingesting and routing events, Eventhouse with KQL databases for storage and fast queries, KQL querysets, Real-Time Dashboards, and Activator for triggering actions on data conditions.
Real-time example. IoT sensor events flow through Eventstream into an Eventhouse and a Real-Time Dashboard shows machine temperatures.
Interview tip. Name each component with its job in one line.
Q18. What is an Eventstream?
🌱 Fresher · In simple words: Eventstream collects live events and sends them where you need them.
Answer. An Eventstream captures events from sources such as Azure Event Hubs, IoT Hub, change data capture feeds or custom apps, optionally transforms them with no-code operations, and routes them to destinations such as an Eventhouse, a Lakehouse or Activator.
Real-time example. Website click events from Event Hubs are filtered and sent both to an Eventhouse for live analysis and to a Lakehouse for history.
Interview tip. Mention that one stream can feed several destinations.
Q19. What is an Eventhouse and KQL?
🚀 Experienced · In simple words: An Eventhouse stores very large amounts of event data, and KQL is the quick query language you use to analyse it.
Answer. An Eventhouse is a container for one or more KQL databases optimised for large volumes of time-series and event data. KQL (Kusto Query Language) is the read-optimised query language used to filter, aggregate and analyse that data quickly.
Real-time example. A query using summarize count() by bin(timestamp, 5m) shows events per five minutes over the last hour.
Interview tip. Write one small KQL query in the interview — it stands out.
Q20. What is Activator?
🌱 Fresher · In simple words: Activator watches your data and sends an alert, or starts an action, when a rule is met.
Answer. Activator (formerly Data Activator) monitors data in Fabric and automatically takes actions — such as sending an email or Teams message or starting a pipeline — when conditions you define are met.
Real-time example. An alert fires when a freezer temperature stays above a threshold for ten minutes.
Interview tip. Mention the rename; some job descriptions still use the old name.
Security, governance and monitoring
Production questions separate experienced candidates from beginners.
Q21. What workspace roles exist in Fabric?
🌱 Fresher · In simple words: Four roles, from most to least power: Admin, Member, Contributor and Viewer.
Answer. Admin, Member, Contributor and Viewer. Admins manage the workspace and its access, Members can also share content, Contributors can create and edit items, and Viewers can only view and read content.
Real-time example. Data engineers are Contributors in the dev workspace but only Viewers in prod, where deployments run through a pipeline.
Interview tip. Follow with least privilege: give item-level access when a workspace role is too broad.
Q22. How do you secure data at row and column level?
🚀 Experienced · In simple words: Show each user only the rows and columns they are allowed to see.
Answer. In a Warehouse or SQL analytics endpoint, use T-SQL row-level security, column-level security and dynamic data masking. OneLake security roles can restrict access to folders and tables in a Lakehouse, and Power BI semantic models can apply row-level security for report users.
Real-time example. Regional managers see only their region’s rows, and the phone column is masked for analysts.
Interview tip. Say which layer you secure and why — engine level vs report level.
Q23. How do sensitivity labels work in Fabric?
🚀 Experienced · In simple words: Sensitivity labels tag data as Confidential (or similar) so it stays protected wherever it goes.
Answer. Sensitivity labels from Microsoft Purview Information Protection can be applied to Fabric items to classify data such as Confidential, and labels can flow to downstream items and exported files.
Real-time example. A Gold Lakehouse with salary data is labelled Confidential, and reports built on it inherit the label.
Interview tip. Mention Purview — governance questions usually expect it.
Q24. How do you monitor Fabric workloads?
🌱 Fresher · In simple words: Use the Monitoring hub to check job runs and the Capacity Metrics app to check usage.
Answer. Use the Monitoring hub to track pipeline, notebook and dataflow runs, the Spark UI and run details for Spark jobs, and the Fabric Capacity Metrics app to see capacity usage, throttling and which items consume the most.
Real-time example. The Capacity Metrics app showed one unoptimised notebook consuming most of the capacity during business hours.
Interview tip. Always mention the Capacity Metrics app — it is Fabric-specific.
Q25. What happens when a capacity is overloaded?
🚀 Experienced · In simple words: If you use more capacity than you bought for too long, Fabric slows down or delays jobs.
Answer. Fabric smooths short bursts of usage over time, but sustained overuse leads to throttling: interactive operations are delayed and later background jobs can be rejected until usage recovers. You respond by optimising workloads, rescheduling jobs or scaling the capacity.
Real-time example. Moving heavy refreshes to off-peak hours removed daytime throttling.
Interview tip. Use the words smoothing (spreading short spikes over time) and throttling (slowing jobs down).
CI/CD and deployment in Fabric
Deployment questions show whether you have worked on a real team project.
Q26. How does Git integration work in Fabric?
🚀 Experienced · In simple words: Connect the workspace to Git so every change is saved, reviewed and versioned.
Answer. A workspace can be connected to a branch in Azure DevOps or GitHub. Supported items are serialised to the repository, so you can commit changes, use branches and pull requests, and update the workspace from Git.
Real-time example. Each developer works in a feature-branch workspace, then raises a pull request into the main branch.
Interview tip. Mention pull requests and code review, not only “we use Git”.
Q27. What are deployment pipelines?
🚀 Experienced · In simple words: Deployment pipelines move tested work from Dev to Test to Production.
Answer. Deployment pipelines promote content between stages such as Development, Test and Production workspaces, compare differences between stages, and use deployment rules to change settings such as data source connections per environment.
Real-time example. A release moves tested pipelines and reports from Test to Production in one deployment.
Interview tip. Explain how Git and deployment pipelines work together.
Q28. How do you manage environment-specific settings?
🚀 Experienced · In simple words: Never hard-code environment names — pass them as settings, so the same code works in every environment.
Answer. Use deployment rules, parameters in pipelines and notebooks, variable libraries where available, and separate connections per environment, so the same code runs in dev, test and prod without manual edits.
Real-time example. The notebook reads a parameter for the target Lakehouse instead of a hard-coded name.
Interview tip. Hard-coded environment values are a red flag; say how you avoid them.
Practise with a trainer Fabric labs, DP-700 preparation and mock interviews in our Azure Data Engineer Training in Hyderabad.
Azure Databricks interview questionsScenario-based Microsoft Fabric interview questions
Panels judge your reasoning and trade-offs more than a single correct answer.
Q29. A company wants to move from Azure Synapse to Fabric. How would you approach it?
🚀 Experienced · In simple words: Move step by step, reuse existing data with shortcuts, and compare results before switching over.
Answer. Inventory existing pipelines, Spark notebooks, SQL pools and reports; use shortcuts to read existing ADLS Gen2 data without copying; migrate pipelines and notebooks in phases; rebuild dedicated SQL pool models in a Warehouse; validate results side by side; then switch reports to Direct Lake.
Real-time example. Running old and new pipelines in parallel for two weeks proves the numbers match before cut-over.
Interview tip. Phased migration and validation are what panels want to hear.
Q30. Design a near-real-time sales dashboard in Fabric.
🚀 Experienced · In simple words: Send live events to an Eventhouse for the dashboard and to a Lakehouse for history, with alerts on top.
Answer. Ingest order events with Eventstream, store them in an Eventhouse for real-time queries, show them on a Real-Time Dashboard, also land them in a Lakehouse for history, and use Activator to alert when sales drop below a threshold.
Real-time example. Store managers see orders per minute and receive a Teams alert when a store stops sending sales.
Interview tip. Separate the hot path (Eventhouse) from the history path (Lakehouse).
Q31. Reports are slow on a Gold Lakehouse. What do you check?
🚀 Experienced · In simple words: Check the files, then the Power BI model, then the capacity — in that order.
Answer. Check table maintenance and small files, whether V-Order is applied, whether the semantic model uses Direct Lake or is falling back to DirectQuery, the model design (star schema, relationships), and capacity throttling in the Capacity Metrics app.
Real-time example. Compacting small files and fixing a many-to-many relationship made the report fast again.
Interview tip. Cover data layout, model design and capacity in that order.
Q32. How would you prepare for the DP-700 exam alongside interviews?
🌱 Fresher · In simple words: Follow Microsoft’s DP-700 study guide, practise in a Fabric trial and build one complete project.
Answer. Follow the official Microsoft study guide for DP-700, practise in a Fabric trial or capacity, focus on the three skill areas — implementing and managing an analytics solution, ingesting and transforming data, and monitoring and optimising — and build one end-to-end project you can explain.
Real-time example. One project covering pipelines, Lakehouse, Warehouse, Eventstream and security answers most interview and exam topics.
Interview tip. Read our DP-700 certification guide for the exam details.
If you want guided practice on Fabric with real projects, see our Azure Data Engineer Training in Hyderabad.
Key takeaways: Microsoft Fabric Interview Questions
• Microsoft Fabric interview questions focus on OneLake, Lakehouse vs Warehouse, pipelines, Real-Time Intelligence, security and CI/CD.
• Explain shortcuts vs mirroring and Lakehouse vs Warehouse clearly — they are asked in almost every round.
• Know the Fabric-specific operations tools: Monitoring hub, Capacity Metrics app, Git integration and deployment pipelines.
• For real-time questions, name Eventstream, Eventhouse, KQL and Activator with one use each.
• Many Fabric interview topics overlap with DP-700, so preparing for one helps with the other.
Frequently asked questions
Are Microsoft Fabric interview questions asked for Azure Data Engineer roles?
Yes. Since DP-700 replaced DP-203 as the data engineer certification, many Azure data engineering roles now include Fabric questions alongside ADF, Databricks and SQL.
Do I need Power BI knowledge for Fabric interviews?
You need the basics: semantic models, Direct Lake and how reports read Gold tables. Deep report design is usually expected from BI developers rather than data engineers.
Is KQL asked in Fabric interviews?
Increasingly, yes, especially for roles involving streaming data. Learn basic KQL operators such as where, summarize, project and bin.
Which is more important for interviews, Lakehouse or Warehouse?
Both. Lakehouse questions are more common for Spark-focused roles, Warehouse questions for SQL-focused roles, and design questions often expect you to use both.
How is DP-700 related to these questions?
DP-700 is the Fabric Data Engineer Associate exam. Its skills measured cover many of the same topics, so these questions also help your exam preparation.
Based on what we teach in our batches. Last updated 31 October 2026. We review this Microsoft Fabric Interview Questions guide regularly. Spotted something outdated? Tell us.