DP-750 mock exam: 25 free practice questions

Q1 / 25

Which language is used to query Azure Databricks events after Log Analytics ingests them?

1. AKusto Query Language
2. BSQL warehouse cluster policy syntax
3. CSpark configuration syntax
4. DDatabricks Asset Bundle YAML

Q2 / 25

A source can export CSV or Parquet, sends 500 GB daily, and downstream jobs read a few columns with filters. Which source-file configuration is best?

1. ALand XML files and flatten them during every query
2. BLand JSON documents grouped by transaction identifier
3. CLand Parquet files in date-organized folders
4. DLand CSV files in one flat directory

Q3 / 25

A batch cluster shows sustained memory pressure during peaks, lighter demand between peaks, and strict completion deadlines. Which adjustment best addresses both observations?

1. ALower the maximum workers and ignore memory utilization
2. BShorten auto-termination while commands are still running
3. CUse higher-memory workers and configure autoscaling bounds
4. DUse smaller workers and keep a fixed worker count

Q4 / 25

An unpartitioned Delta table has heavy skew, concurrent writes, and filters that recently shifted to customer and date. Existing data must adopt the new layout. Which design best meets all constraints?

1. ASet customer and date as clustering keys, then run `OPTIMIZE FULL`
2. BPartition by customer and date, then add the same liquid-clustering keys
3. CZ-order customer and date, then enable liquid clustering on that table
4. DSet customer and date as keys, then run only `VACUUM`

Q5 / 25

A security analyst must identify out-of-hours table changes, the initiating principal, and the request origin across workspaces. Which design supplies all three?

1. AQuery DatabricksNotebook by duration and completed task count
2. BQuery DatabricksJobs by schedule, Response, and RequestParams
3. CQuery DatabricksClusters by resize count and cluster status
4. DQuery DatabricksTables by time, Identity, and SourceIPAddress

Q6 / 25

Why does SQL PIVOT require an aggregate function?

1. ASeveral source rows can map to one pivoted output cell
2. BEvery source category must first become a separate input table
3. CEach output column must be converted into a filtering predicate
4. DNull source values must be replaced before columns are created

Q7 / 25

A team is still exploring intermediate results, must loop over an external API, and needs direct control of Spark caching. Which authoring approach should it choose?

1. AUse a materialized view with managed incremental updates
2. BUse a declarative pipeline with automatic dependency analysis
3. CUse a notebook with procedural code
4. DUse a streaming table with append-only processing

Q8 / 25

A managed table's filters and data distribution change frequently. The team wants minimal manual tuning and wants key changes only when expected skipping savings exceed reorganization cost. Which configuration fits?

1. ARemove clustering keys and rely on `VACUUM` to reorganize rows
2. BFix four manual keys and disable workload-history analysis
3. CSchedule Z-ordering and partition the table on the same columns
4. DEnable predictive optimization and configure `CLUSTER BY AUTO`

Q9 / 25

Which access mode provides shared classic compute for multiple concurrent users?

1. ADedicated access mode
2. BServerless performance mode
3. CClassic compute in Standard access mode
4. DSingle-node cluster mode

Q10 / 25

How can a mask use the value of another column in the same table?

1. AStore the other value in the table's descriptive comment
2. BAdd a second row filter to the same protected table
3. CPass the additional column through the USING COLUMNS clause
4. DConvert the mask into a catalog-level SELECT privilege

Q11 / 25

Which declarative object should be selected for an aggregation that must remain synchronized with upstream updates?

1. AA notebook task parameter
2. BAn append-only streaming table
3. CA materialized view
4. DDepends on

Q12 / 25

Which format is the default for tables in Azure Databricks?

1. AApache Iceberg
2. BJSON
3. CCSV
4. DDelta Lake

Q13 / 25

Why is an LTS runtime a suitable choice for operational job compute?

1. AIt removes the need to test production upgrades
2. BIt emphasizes extended compatibility and stability
3. CIt automatically converts every cluster to serverless
4. DIt enables Photon on GPU clusters

Q14 / 25

Two recipients need different regional rows from one table, the provider wants one maintained data source, and access must be reviewable later. Which implementation best meets the requirements?

1. ACreate a full physical copy per recipient and disable audit logging for the shares
2. BShare a dynamic view filtered by recipient properties and monitor sharing audit events
3. CShare the base table to both recipients and describe regional limits in a comment
4. DGrant both recipients the entire schema so future tables inherit access

Q15 / 25

Which recovery behavior does a continuous job use after repeated failures?

1. AManual repair of every failed task before another run
2. BImmediate cancellation of the continuous trigger configuration
3. CUnlimited task retries at one fixed interval
4. DWhole-job retries with exponential backoff

Q16 / 25

What does Delta schema enforcement do when incoming data has an unexpected column?

1. ADrops the column and records a warning metric
2. BRejects the mismatched write by default
3. CAdds the column to every target by default
4. DConverts the column into a primary key

Q17 / 25

Two teams share a cluster but require incompatible versions of the same Python package. Each version is needed in only one team's notebook session, and the shared cluster configuration must remain unchanged. Which approach best fits?

1. AUse one unpinned compute-scoped PyPI package
2. BPut both wheels in a volume and install both cluster-wide
3. CUse notebook-scoped installations for the separate sessions
4. DInstall both versions as compute-scoped libraries

Q18 / 25

Which Unity Catalog principal represents an individual person?

1. AA service principal created for an automated application
2. BA securable table that receives object-level privileges
3. CA user principal for a person, managed through Microsoft Entra ID
4. DA group principal whose members inherit assigned privileges

Q19 / 25

A workflow has two independent ingestion tasks, one transform that needs both outputs, and cleanup that must run even after a failure. Which dependency design satisfies the ordering and failure requirements?

1. AChain the ingestions, run transform after the first, and set cleanup to All succeeded
2. BRun all four tasks without dependencies and use a shared job parameter for order
3. CRun ingestions in parallel, make transform depend on both, and set cleanup to All done
4. DRun transform with both ingestions, then set cleanup to At least one failed

Q20 / 25

Which pipeline component stores source credentials and endpoint information?

1. AA full refresh scheduling window
2. BA destination table's SCD type
3. CA destination column exclusion list
4. DA reusable Lakeflow Connect connection

Q21 / 25

A notebook workload uses RDD APIs. Which compute selection supports that requirement?

1. AClassic compute with Dedicated access mode
2. BA serverless SQL warehouse with Photon
3. CClassic compute with Standard access mode
4. DServerless notebook compute with automatic scaling

Q22 / 25

A regulated table needs 30 days of time travel, VACUUM runs frequently, and unreferenced files must remain usable throughout that period. Which setup is appropriate?

1. ASet both log and deleted-file retention to 30 days before vacuuming
2. BEnable column lineage and leave both retention properties unchanged
3. CKeep deleted-file retention at 7 days and raise only log retention to 30 days
4. DRun VACUUM RETAIN 0 HOURS and rely on the transaction log for old data

Q23 / 25

Which run result follows when an operator stops an active Lakeflow Job run?

1. AThe active run is skipped
2. BThe active run is canceled
3. CThe active run succeeds with failures
4. DThe active run is repaired

Q24 / 25

What does the availableNow trigger do?

1. ARewrites the complete target after each event
2. BRuns continuously without a termination point
3. CProcesses the full current backlog and then stops
4. DProcesses one deprecated micro-batch and stops

Q25 / 25

A Runtime 13.3 LTS cluster uses Standard access mode, but setup validation cannot reach the database and authentication also fails. What should the engineer verify first?

1. ACreate the foreign catalog before repairing reachability, then rely on table privileges for authentication
2. BDowngrade to Runtime 12.2 LTS, switch access mode, and retry without changing the network rules
3. CReplace the stored credentials with a table `LOCATION` and remove the firewall exception
4. DFirewall and port rules, stored credentials, and accepted Databricks source addresses