Free Microsoft DP-750 Practice Questions

Correct answer: A. Yes

A foreign catalog in Azure Databricks Unity Catalog is specifically designed to expose external database schemas and tables (such as those in Microsoft SQL Server) within the Unity Catalog framework. When you create a foreign catalog connected to DB1, the schemas and tables become queryable in Databricks, appear alongside other Unity Catalog objects in Catalog Explorer, and the data remains in the source SQL Server database without being copied into Databricks-managed storage. This solution directly addresses all three requirements.

Why the other options are wrong

• B. Foreign catalogs are the correct mechanism for exposing external database schemas and tables in Unity Catalog without data duplication.

Correct answer: C. a job cluster that uses autoscaling

Job clusters with autoscaling are purpose-built for running Spark Declarative Pipelines. They automatically scale to handle variable workloads, minimizing costs by terminating after job completion and only running when needed. Unlike all-purpose clusters that remain running and incur continuous charges, job clusters are created for specific jobs and destroyed afterward, providing both cost efficiency and operational simplicity for scheduled pipeline execution.

Why the other options are wrong

• A. SQL warehouses are designed for SQL queries on Delta tables, not for Spark transformation pipelines.
• B. All-purpose clusters with autoscaling remain running between jobs, incurring unnecessary costs and requiring manual management.
• D. Single-node clusters cannot handle the scaling requirements of production pipelines and lack distributed processing capabilities.

Correct answer: B. CREATE SCHEMA catalog1.schema2 -MANAGED LOCATION ‘abfss://[email protected]/data’;

The correct syntax for creating a schema in Unity Catalog with a managed location uses the MANAGED LOCATION clause. The statement CREATE SCHEMA catalog1.schema2 MANAGED LOCATION 'abfss://[email protected]/data' properly specifies that schema2 belongs to catalog1 and designates the Azure Data Lake Storage Gen2 path as the managed location for tables created in this schema.

Why the other options are wrong

• A. The LOCATION clause without MANAGED keyword is deprecated syntax and does not properly designate the managed location in Unity Catalog.
• C. CREATE CATALOG creates a new catalog, not a schema within an existing catalog.
• D. DBPROPERTIES syntax does not support location configuration; the MANAGED LOCATION clause is the correct approach.

Correct answer: C. Upload the libraries to Workspace1 and install the libraries on Cluster1.

Uploading libraries to Workspace1 storage and then installing them on Cluster1 provides centralized management and integrates with Unity Catalog access control. This approach allows administrators to manage library versions, control access through Unity Catalog permissions, and maintain audit trails of library usage across the workspace, ensuring consistent governance and compliance.

Why the other options are wrong

• A. Using pip3 installs libraries directly from PyPI without Unity Catalog governance or centralized version control.
• B. Custom dependency scripts bypass Unity Catalog's access control and governance mechanisms.
• D. Manually installing and restarting the cluster is operationally inefficient and does not leverage Unity Catalog's access control features.

Correct answer: B, E. Enable autoscaling for Cluster1; Configure Cluster1 to terminate after 30 minutes of inactivity.

Enabling autoscaling on Cluster1 (option B) allows it to scale up during increased workloads and scale down during periods of low utilization. Configuring autotermination after 30 minutes of inactivity (option E) ensures the cluster terminates when not in use, eliminating costs during idle periods. Together, these settings provide automatic scaling to meet demand while minimizing costs by shutting down unnecessary resources.

Why the other options are wrong

• A. Disabling Photon acceleration reduces performance optimization and does not address scaling or cost requirements.
• C. Compute policies manage user permissions but do not affect cluster scaling behavior or cost.
• D. Specifying a fixed number of workers prevents autoscaling, directly contradicting the scaling requirement.

Correct answer: A. Configure a checkpoint location for the streaming query.

Configuring a checkpoint location for the streaming query enables exactly-once semantics and idempotent processing. The checkpoint stores the current progress of the streaming job, allowing it to resume from the last processed offset after a cluster restart. Without a checkpoint, the streaming job cannot track its progress and must reprocess from the beginning of the data.

Why the other options are wrong

• B. Increasing trigger interval affects processing frequency but does not prevent reprocessing after restart.
• C. Change data feed captures data modifications but does not prevent reprocessing or track streaming progress.
• D. Watermarks manage late-arriving data in windowed operations but do not prevent reprocessing after cluster restarts.

Correct answer: D. Disable autoscaling and reduce the number of worker nodes.

Since the workload is predictable, does not spike, and CPU utilization is consistently low at 20 percent, the cluster is overprovisioned. Disabling autoscaling and reducing the number of worker nodes ensures the cluster uses only the compute capacity needed for this batch workload. The current node type already meets requirements, so this adjustment directly reduces costs without impacting job duration.

Why the other options are wrong

• A. Photon acceleration improves query performance but does not reduce the number of nodes or lower costs when CPU is underutilized.
• B. Increasing to a larger node type increases both cost and resource utilization, contrary to the cost reduction goal.
• C. Autotermination timeout affects when an idle cluster shuts down but does not address the overprovisioning during active workloads.

Correct answer: A. serverless compute

Serverless compute is purpose-built for Spark Declarative Pipelines and provides automatic scaling with minimal operational overhead. Unlike job clusters that require manual management and configuration, serverless compute automatically handles resource allocation, scaling, and termination. This approach minimizes both administrative effort and operational costs while delivering the autoscaling capability required for variable workloads.

Why the other options are wrong

• B. Single-node clusters cannot scale and lack distributed processing for production pipelines.
• C. All-purpose clusters with autoscaling require manual configuration and management, increasing operational effort.
• D. SQL warehouses are designed for SQL analytics, not Spark transformation pipelines.
• E. Job clusters require explicit configuration and management, whereas serverless compute handles this automatically.

Correct answer: D. Declarative Automation Bundles in Git

Declarative Automation Bundles (DAB) in Git provide a comprehensive ALM solution that meets all stated requirements. DAB supports source control for notebooks, workflows, and jobs through Git integration, enables parameterization via configuration files for different environments (dev, test, prod), and ensures consistent and repeatable deployments through infrastructure-as-code principles. This is the modern recommended approach by Databricks for production ALM workflows.

Why the other options are wrong

• A. Export/import of workspaces is manual, not repeatable, and lacks source control integration for proper version management.
• B. While Terraform can manage Databricks resources, it is not the primary tool for managing notebooks and workflows; DAB is purpose-built for this.
• C. Git folders with manual promotion is inconsistent and error-prone, lacking the automation and parametrization required for production deployments.

Correct answer: B. A new Declarative Automation Bundles project structure is created.

The `databricks bundle init` command initializes a new Declarative Automation Bundles project structure by creating the necessary directory layout, configuration files, and template files required to define Databricks resources. This is the first step in setting up a bundle project before any validation or deployment occurs.

Why the other options are wrong

• A. Bundle validation is performed by `databricks bundle validate`, not `bundle init`.
• C. Deployment occurs with `databricks bundle deploy`, not `bundle init`.
• D. Running jobs or pipelines is done with `databricks bundle run`, not `bundle init`.

Correct answer: D. Grant the analysts the SELECT permission for the Transactions table and apply column masks to email_address and credit_card_number.

Column masks applied to sensitive columns (email_address and credit_card_number) allow business analysts to query the entire Transactions table while automatically redacting sensitive data according to the specified rules (showing only domain portion of email and last four digits of credit card). This approach meets all requirements: analysts see non-sensitive columns fully, sensitive columns are masked per specification, queries execute without errors, and the principle of least privilege is maintained by limiting what analysts see without revoking table access.

Why the other options are wrong

• A. Column-level encryption prevents querying altogether and would cause errors rather than displaying masked values.
• B. Granting SELECT only on non-sensitive columns prevents analysts from querying the full table as required.
• C. Row-level filters restrict which rows are visible, not which columns or which parts of column values are masked.

Correct answer: C, D. delta.logRetentionDuration; delta.deletedFileRetentionDuration

Retaining deleted data for 30 days means old data files and the transaction log that references them must both be kept for 30 days. delta.deletedFileRetentionDuration keeps removed data files for 30 days before VACUUM can delete them, and delta.logRetentionDuration keeps the log history for the same period so that time travel to that data works. On Unity Catalog managed tables, predictive optimization runs VACUUM automatically, so data older than 30 days is removed permanently with minimal effort.

Why the other options are wrong

• A. delta.timeUntilArchived relates to archival storage support and does not control how long deleted data is retained.
• B. delta.enableDeletionVectors changes how deletes are recorded to avoid rewriting files, but it does not set a retention period.
• E. delta.autoOptimize.autoCompact compacts small files and has no effect on how long deleted data is kept.

Correct answer: D. Use Delta Sharing.

Delta Sharing enables sharing curated data with external organizations using their own compute platforms while maintaining central governance through Unity Catalog. It allows recipients to query shared data without needing Databricks compute, requires no administrative overhead for managing external user accounts in Databricks, and provides secure, fine-grained access control. Delta Sharing is the purpose-built solution for cross-organization data sharing with minimal administrative effort.

Why the other options are wrong

• A. Creating a SQL warehouse requires the external organization to use Databricks compute and involves managing external user access, increasing administrative overhead.
• B. Granting workspace access to external users requires managing external identities and compute provisioning, not suitable for organizations using their own platforms.
• C. Moving data to SFTP loses centralized governance capabilities and is not integrated with Unity Catalog.
• E. Lakeflow Connect is for building data pipelines within Databricks, not for sharing data with external organizations.

Correct answer: B. No

Lakeflow Connect is a data integration tool designed for moving and replicating data between systems, not for exposing remote schemas directly as Unity Catalog objects. It would copy or move data rather than creating a live foreign connection. To expose DB1 schemas and tables directly in Unity Catalog without copying data, a foreign catalog connection is required.

Correct answer: D. continuous

A continuous trigger is the appropriate choice for processing new records as soon as they become available. Continuous triggers ensure that Lakeflow Jobs run perpetually and process incoming data with minimal latency, providing near real-time ingestion. This is ideal for streaming data scenarios where records must be processed immediately upon arrival rather than waiting for scheduled intervals or manual invocation.

Why the other options are wrong

• A. Manual triggers require user intervention and cannot process records automatically as they arrive.
• B. File arrival triggers are designed for batch file ingestion and may introduce delays compared to continuous processing.
• C. Scheduled triggers process data on fixed time intervals and cannot react to new records arriving between scheduled runs.

Correct answer: C. Grant User1 the CAN MANAGE permission for Cluster1.

The CAN MANAGE permission for Cluster1 grants User1 the ability to install packages and manage cluster libraries, which is the specific capability needed to install Maven packages on all cluster nodes. This permission follows the principle of least privilege by granting only the minimal access required for the task without providing broader administrative capabilities or resource-group-level permissions.

Why the other options are wrong

• A. Assigning the Contributor role at the resource group level provides far too much access beyond installing packages and violates least privilege.
• B. CAN RESTART permission only allows restarting the cluster and does not include library or package management capabilities.
• D. Assigning the workspace admins role provides excessive permissions beyond what is needed for installing packages on a single cluster.

Correct answer: D. Create a new catalog.

Creating a new catalog for each department meets all requirements. Catalogs provide logical isolation of data assets, independent access control management, and centralized governance through Unity Catalog. Each department's catalog can have separate permissions and schemas while maintaining a unified governance model across the workspace. This approach is more efficient than creating separate workspaces or storage accounts and provides better organization than using dynamic views.

Why the other options are wrong

• A. Creating separate storage accounts adds infrastructure complexity and doesn't leverage Unity Catalog's centralized governance capabilities.
• B. Dynamic views provide row-level filtering but do not logically isolate data assets or provide independent access management.
• C. Creating separate workspaces defeats the purpose of centralized governance and creates unnecessary infrastructure overhead.

Correct answer: A. Values appear in both tables more than once.

The INTERSECT operator returns only the values that appear in both result sets. The left side finds values in Column1 of Table1 that occur more than once, and the right side finds values in Column2 of Table2 that occur more than once. The intersection of these two queries returns only the values that appear in both tables and occur more than once in each respective table, meaning these values appear in both tables more than once.

Why the other options are wrong

• B. INTERSECT returns only common values, not values that appear in either table; UNION would return values from either table.
• C. INTERSECT requires values to be in both result sets; a value appearing only in Table2 would not satisfy this condition.
• D. The result includes values that appear in both tables more than once, not just Table1 alone.

Correct answer: C. an external location that uses a storage credential

An external location configured with a storage credential is required for the COPY INTO statement to access data in Azure Data Lake Storage Gen2. External locations define the path to cloud storage and include authentication credentials that allow Databricks to read from or write to the specified container. This is a fundamental Unity Catalog component needed for any operation that accesses data outside Databricks-managed storage.

Why the other options are wrong

• A. An external table named Customer would not directly enable the COPY INTO command to access the source file; the external location and credential are what enable the access.
• B. Lakeflow ingestion gateway is not the appropriate mechanism for COPY INTO operations.
• D. A volume is a storage container within Unity Catalog but does not provide the necessary authentication mechanism for accessing external cloud storage paths.

Correct answer: C. Auto Loader

Auto Loader is the Azure Databricks solution designed for ingesting new data from cloud storage into managed Delta tables with near-real-time latency. It automatically detects and processes new files as they arrive, requires minimal custom code, and handles the complexity of incremental ingestion and schema evolution. It is purpose-built for scenarios where new rows are appended frequently to cloud storage.

Why the other options are wrong

• A. Scheduled batch jobs introduce latency and require custom scheduling logic; they do not meet near-real-time requirements as effectively as Auto Loader.
• B. Azure Data Factory is a separate service that adds operational complexity and overhead compared to the native Auto Loader functionality in Databricks.
• D. External tables reference files without copying data; they do not meet the requirement to store data in managed Delta tables.

Correct answer: C. Add a NOT NULL constraint to transaction_id and a CHECK constraint to amount.

In Azure Databricks with Unity Catalog, table-level data quality enforcement is implemented using SQL constraints. NOT NULL constraints prevent null values in specified columns, and CHECK constraints enforce conditions on column values. Adding a NOT NULL constraint to transaction_id and a CHECK constraint (amount > 0) to the amount column will automatically reject invalid records at write time, meeting all three requirements without requiring application-level validation logic.

Why the other options are wrong

• A. SELECT with WHERE conditions only validates during reads, not at write time, so invalid data would still be stored in the table.
• B. Creating a view that filters data does not prevent invalid records from being written to the underlying table; the view only hides them from queries.
• D. Row-level security controls who can access data, not what data can be written; it does not enforce data quality constraints.

Correct answer: C. INSERT INTO

The INSERT INTO command appends new records to an existing table while preserving all historical data. Since the daily CSV file contains only new records with no updates to existing rows, INSERT INTO is the appropriate command that minimizes processing effort by performing a simple append operation without requiring any update or merge logic.

Why the other options are wrong

• A. UPDATE is used to modify existing rows; it is not appropriate for loading new records and would require logic to match and update rows, adding unnecessary complexity.
• B. INSERT OVERWRITE replaces all existing data in the table, which contradicts the requirement to preserve existing historical data.

Correct answer: C. USE CATALOG, USE SCHEMA, and SELECT

To query data in catalogs and their schemas, the service principal requires USE CATALOG permission to access each catalog, USE SCHEMA permission to access each schema within those catalogs, and SELECT permission on the tables to retrieve data. These three permissions are the minimum required hierarchy to enable query access while following the principle of least privilege by not granting unnecessary administrative or write permissions.

Why the other options are wrong

• A. USE SCHEMA and SELECT alone omit the USE CATALOG permission needed to access catalogs themselves.
• B. USE CATALOG and SELECT omit the USE SCHEMA permission required to access individual schemas within the catalog.
• D. USE CATALOG and USE SCHEMA alone omit the SELECT permission required to actually query table data.

Correct answer: B. No

A Databricks access connector is an Azure resource that manages authentication and authorization to cloud services, but it does not itself expose external database schemas and tables as Unity Catalog objects. To expose an external SQL Server database as queryable Unity Catalog objects, a foreign catalog or external connection through Unity Catalog must be configured, not just an access connector.

Correct answer: C. INSERT INTO sales.schema1.table2 SELECT * FROM sales.schema1.table1;

The INSERT INTO command appends new rows to an existing table while preserving all existing data. This is the correct choice because the requirement explicitly states to preserve existing data in table2 and only add rows from table1. INSERT INTO is also the most efficient option as it only performs an append operation without scanning or rewriting existing data.

Why the other options are wrong

• A. INSERT OVERWRITE replaces all existing data in table2 with data from table1, which violates the requirement to preserve existing data.
• B. CREATE TABLE would fail because table2 already exists; it does not preserve existing data in any case.
• D. CREATE OR REPLACE TABLE drops and recreates the table, destroying all existing historical data in table2.

Correct answer: B. No

Creating a native catalog in Unity Catalog does not meet the requirements. A native catalog is for organizing Unity Catalog-managed objects within Databricks storage, not for connecting to external SQL Server databases. To expose SQL Server schemas and tables alongside Unity Catalog objects without copying data, you need to create a foreign catalog that connects to the external database using a connection to DB1. This allows querying the remote database objects as if they were Unity Catalog objects while keeping data in the source system.

Why the other options are wrong

• A. A native catalog is used for organizing Databricks-managed objects, not for exposing external database schemas and tables.

Correct answer: B. an Azure Storage account

Azure Storage accounts are the most cost-effective solution for storing audit logs that are infrequently accessed and only reviewed during security investigations. Storage accounts offer the lowest ingestion and retention costs compared to other options, support 90-day retention policies through lifecycle management, and provide durable, scalable storage for large volumes of audit data that don't require real-time querying or complex analysis.

Why the other options are wrong

• A. Azure Event Hubs is designed for high-throughput streaming and event processing, making it unnecessarily expensive for write-once audit logs.
• C. Log Analytics is optimized for regular querying and monitoring, which increases costs for logs only reviewed during investigations.
• D. Azure Monitor metrics is not designed for storing raw audit logs and would require additional components for log storage.