DE-PRO practice questions 191–200

1. Question 191 · 202 in bank

   A data engineering team is configuring access controls in Databricks Unity Catalog. They grant the SELECT privilege on the sales catalog to the analyst_group, expecting that members of this group will automatically have SELECT access to all current and future schemas, tables, and views within the sales catalog. What describes the privilege inheritance behavior in Unity Catalog?

   • APrivileges in Unity Catalog do not cascade; SELECT most be explicitly granted on each schema and table, even if granted at the catalog level.
   • BGranting SELECT on a catalog automatically applies SELECT to all current and future schemas, tables, and views within that catalog.
   • CGranting SELECT at the catalog level applies to existing schemas and tables but not to those created in the future.
   • DPrivileges granted at the schema level override any catalog-level privileges and prevent access unless explicitly revoked.

   Show answer

   Answer: B

   Next question

2. Question 192 · 202 in bank

   An organization processes customer data from web and mobile applications. Data includes names, emails phone numbers, and location history. Data arrives both as Batch files from an SFTP drop (daily) and Streaming JSON events from Kafka (real-time). To comply with internal data privacy policies, the following requirements must be met: • Personally identifiable information (PII) like email, phone_number, and ip_address must be masked or anonymized before storage • Both batch and streaming pipelines must apply consistent PII handling • Masking logic must be auditable and reproducible • The masked data must still be usable for downstream analytics How should the data engineer design a compliant data pipeline on Databricks that supports both batch and streaming modes, applies data masking to PII, and maintains traceability of transformations for audits?

   • AIngest both batch and streaming data using Lakeflow Spark Declarative Pipelines, and apply masking via Unity Catalog column masks at read time to avoid modifying the data during ingestion.
   • BLoad batch data with notebooks and ingest streaming data with SQL Warehouses; use Unity Catalog column masks on Silver tables to redact fields after storage.
   • CUse Lakeflow Spark Declarative Pipelines for batch and streaming ingestion define a PII masking function, and apply it during Bronze ingestion before writing to Delta Lake.
   • DAllow PII to be stored unmasked in Bronze for lineage tracking, then apply masking logic in Gold tables used for reporting.

   Show answer

   Answer: C

   Next question

3. Question 193 · 202 in bank

   A data architect is designing a Databricks solution to efficiently process data for different business requirements. In which scenario should a data engineer use a materialized view compared to a streaming table?

   • APrecomputing complex aggregations and joins from multiple large tables to accelerate BI dashboard performance.
   • BProcessing high-volume, continuous clickstream data from a website to monitor user behavior in real-time.
   • CIngesting data from Apache Kafka topics with sub second processing requirements for immediate alerting,
   • DImplementing a CDC (Change Data Capture) pipeline that needs to detect and respond to database changes within seconds.

   Show answer

   Answer: A

   Next question

4. Question 194 · 202 in bank

   A data engineering team has a time-consuming data ingestion job. It has three data sources, and each data ingestion notebook takes about one hour to load new data. The job had been working fine, loading data every midnight. However, one day they received an alert message stating the job had failed. By investigating the cause, they figured out that the failed notebook code was updated, and a new required configuration was introduced to parameterize a hardcoded value without notice. Now, the team has to fix the issue as quickly as possible to load the latest data from the failing data source. Which action should the team take in this scenario?

   • AUpdate the task by adding the missing task parameter, and manually run the job.
   • BShare the analysis with the failing notebook owner so that they can fix it quickly.
   • CRepair the run with the new parameter.
   • DRepair the run with the new parameter, and update the task by adding the missing task parameter.

   Show answer

   Answer: D

   Next question

5. Question 195 · 202 in bank

   A data engineer is masking the provided data column containing the email address. The goal is to have an output of the same length for all rows, while keeping different outputs for different values. Which SQL function should be used to achieve this?

   • Ahash(email)
   • Bmask(email, ‘?’)
   • Csha1(‘email’)
   • Dsha2(email,0)

   Show answer

   Answer: D

   Next question

6. Question 196 · 202 in bank

   A data engineer is implementing Unity Catalog governance for a multi-team environment. Data scientists need interactive clusters for basic data exploration tasks, while automated ETL jobs require dedicated processing. How should the data engineer configure cluster isolation policies to enforce least privilege and ensure Unity Catalog compliance?

   • AConfigure all dusters with NO_ISOLATION_SHARED access modes since Unity Catalog works with any cluster configuration
   • BAllow all users to create any duster type and rely on manual configuration to enable Unity Catalog access modes.
   • CCreate compute policies with STANDARD access mode for interactive 'workloads and dedicated access mode for automated jobs.
   • DUse only DEDICATED access mode for both interactive workloads and automated jobs to maximize security isolation.

   Show answer

   Answer: C

   Next question

7. Question 197 · 202 in bank

   A data engineer, while designing a Pandas UDF to process financial time series data with complex calculations that require maintaining state across rows within each stock symbol group must ensure the function is efficient and scalable. Which approach will solve the problem with minimum overhead while preserving data integrity?

   • AUse a SCALAR_ITER Pandas UDF with iterator-based processing, implementing state management through persistent storage (Delta tables) that gets updated after each batch to maintain continuity across iterator chunks.
   • BUse a GROUPED_AGG UPF that processes each stock symbol group independently, maintaining state through intermediate aggregation results that get passed between successive UDF calls via broadcast variables.
   • CUse a SCALAR Pandas UDF that processes the entire dataset at once, implementing custom partitioning logic within the UDF to group by stock symbol and maintain state using global variables shared across all executor processes.
   • DUse ApplyInPandas method on spark dataframe that receives all rows for each stock symbol as a pandas DataFrame, allowing processing within each group while maintaining state variables local to each group’s processing function.

   Show answer

   Answer: D

   Next question

8. Question 198 · 202 in bank

   A data engineering team is in the process of migrating off its legacy Hadcop platform. As part of the process, they are evaluating the different storage formats to see which offers the best query performance. Their legacy platform utilizes ORC and RCFile formats. They converted a subset of their data to Delta Lake and ran a side-by-side benchmark, and found that their queries performed significantly better. Upon investigation, they discovered that the query plans were different, and the queries that read from the Delta Lake tables leveraged a Shuffle Hash Join, whereas the legacy formats used Sort Merge Joins. The queries that read from the Delta Lake tables also read less data. Which reason could be attributed to the difference in query performance?

   • AThe queries against the ORC tables leveraged the dynamic data skipping optimization but not the dynamic file pruning optimization.
   • BThe queries against the Delta Lake tables were able to leverage the dynamic file pruning optimization.
   • CDelta Lake enables data skipping and file pruning using a vectorized Parquet reader.
   • DShuffle Hash Joins are always more efficient than Sort Merge Joins.

   Show answer

   Answer: B

   Next question

9. Question 199 · 202 in bank

   A data engineer is using Lakeflow Declarative Pipeline to propagate row deletions from a source bronze table (user_bronze) to a target silver table (user_silver). The developer wants deletions in user_bronze to automatically delete corresponding rows in user_silver during pipeline execution. Which configuration ensures deletions in the bronze table are propagated to the silver table?

   • AUse apply_changes without CDF and filter rows where _soft_deleted is true.
   • BEnable CDF on user_silver, read its transaction log, and use MERGE to sync deletions.
   • CEnable CDF on user_bronze, read its CDF stream, and use apply_changes with apply_as_deletes for user_silver.
   • DConfigure VACUUM on user_bronze to delete files, then rebuild user_silver from scratch.

   Show answer

   Answer: C

   Next question

10. Question 200 · 202 in bank

    A data engineer is designing a Lakeflow Spark Declarative Pipeline to process streaming order data. The pipeline uses Auto Loader to ingest data and must enforce data quality by ensuring customer_id and amount are greater than zero. Invalid records should be dropped. Which Lakeflow Spark Declarative Pipelines configurations implement this requirement using Python?

    • A@dlt.table @dlt.expect_or_drop("valid_customer", "customer_id IS NOT NULL") @dlt.expect_or_drop("valid_amount", "amount > 0") def silver_orders(): return dlt.read_stream("bronze_orders")
    • B@dlt.table def silver_orders(): return dlt.read_stream("bronze_orders").expect_or_drop("valid_customer", "customer_id IS NOT NULL").expect_or_drop("valid_amount", "amount > 0")
    • C@dlt.table @dlt.expect("valid_customer", "customer_id IS NOT NULL") @dlt.expect("valid_amount", "amount > 0") def silver_orders(): return dlt.read_stream("bronze_orders")
    • D@dlt.table def silver_orders(): return dlt.read_stream("bronze_orders").expect("valid_customer", "customer_id IS NOT NULL").expect("validamount", "amount > 0")

    Show answer

    Answer: A

    Next question

Keyboard:J Kmove between questionsA–Dselect an answerRshow or hide the answer← →change page

DE-PRO practice, question 191. 202 questions available.