Databricks Data Engineer Associate practice exam: 10 free questions

An all-purpose cluster used by an analytics team during office hours keeps running every night and weekend with nothing attached, and the bill shows it. Which setting most directly fixes this?

• ASet auto termination after a period of inactivity
  Correct. Auto termination shuts the cluster down after the configured idle time, so it stops costing money when nobody uses it.
• BEnable autoscaling with a minimum of one worker
  Autoscaling adjusts workers to load, but the driver and minimum workers keep running when idle.
• CSwitch the cluster to Photon acceleration
  Photon speeds up queries. It does not stop an idle cluster.
• DChange the access mode from standard to dedicated
  Access mode is about isolation and sharing. It has no effect on idle time.

A team adds a schedule of every 30 minutes to a Lakeflow Connect ingestion pipeline. How is the schedule run?

• AThe source system pushes data every 30 minutes on its own
  The connector pulls data; the source does not push on the pipeline's schedule.
• BLakeflow Connect creates a Lakeflow Jobs job with the pipeline as a task
  Correct. Each schedule added to the ingestion pipeline creates a job in which the pipeline is a task, and other tasks can be added.
• CA cron job must be created on an all-purpose cluster to call the pipeline
  No cluster cron is needed. Scheduling is handled through Lakeflow Jobs.
• DThe schedule is stored in the destination table properties
  Schedules are not stored in table properties.

A team is evaluating Lakeflow Connect managed connectors for a CRM source. Which TWO statements describe managed connectors? (Choose TWO.)

• AThey require hand-written Structured Streaming code for each table
  Managed connectors are configured through UI, API or bundles, not hand-coded streams.
• BThey read incrementally after the first full load when the source allows it
  Correct. After an initial load, they use change tracking or cursors to read only changed data when possible.
• CThey can read only from files in cloud object storage
  Managed connectors cover SaaS applications and databases, among others.
• DThey copy all data to the DBFS root before writing it to tables
  They write to Unity Catalog tables; the DBFS root is legacy storage.
• EThey produce ingestion pipelines governed by Unity Catalog
  Correct. Managed connectors write to Unity Catalog tables through governed ingestion pipelines.

On classic compute, which Spark property sets the heap memory available to each executor process?

• Aspark.driver.memory
  This sets the memory of the driver process, not of the executors.
• Bspark.sql.shuffle.partitions
  This sets how many partitions a DataFrame shuffle produces, not memory.
• Cspark.sql.autoBroadcastJoinThreshold
  This sets the size limit for automatic broadcast joins, not executor memory.
• Dspark.executor.memory
  Correct. This property sets the JVM heap of each executor. On Databricks it is derived from the worker node type, so a larger node is the usual way to raise it.

A sales DataFrame df has region and amount columns. Which statement returns, for each region, the average amount and the number of orders?

• Adf.agg(avg("amount").alias("avg_amt"), count("*").alias("orders")).groupBy("region")
  agg without groupBy collapses the data into one row with no region column, so the trailing groupBy cannot give per-region results.
• Bdf.groupBy("region").agg(mean("amount").alias("avg_amt"))
  This returns the average per region but not the number of orders.
• Cdf.select(avg("amount").alias("avg_amt"), count("*").alias("orders"))
  This aggregates the whole DataFrame into one row instead of one row per region.
• Ddf.groupBy("region").agg(avg("amount").alias("avg_amt"), count("*").alias("orders"))
  Correct. groupBy defines the groups and agg computes several aggregates per group, here the mean and the row count.

Which Lakeflow Jobs trigger type is data-driven rather than time-based?

• AScheduled, with a simple interval of every hour
  A simple interval is still a clock-based schedule.
• BTable update
  Correct. It starts a run when monitored tables are updated, so data drives the start.
• CScheduled, with a Quartz cron expression
  Cron is a time-based schedule.
• DA manual Run now click
  A manual run is started by a person, not by data.

Which Databricks CLI command checks a bundle's configuration for target prod for errors without deploying anything?

• Adatabricks bundle deploy -t prod
  deploy uploads files and creates or updates resources in the workspace.
• Bdatabricks bundle run -t prod
  run starts a deployed job or pipeline; it needs a resource key and changes nothing about configuration.
• Cdatabricks bundle init
  init creates a new bundle project from a template.
• Ddatabricks bundle validate -t prod
  Correct. validate loads and checks the configuration, resolving variables for the target, without changing the workspace.

In a job run, a task shows the status Upstream failed. What does this mean?

• AThe task did not run because a task it depends on failed
  Correct. The status marks a task that was blocked by a failed dependency.
• BThe task ran and failed while reading an upstream table
  A task that runs and fails shows Failed, with its own error.
• CThe task was not chosen by an If/else branch
  A task on the branch that was not taken is excluded, not marked as an upstream failure.
• DThe task is waiting for its cluster to start
  A task waiting for compute is pending, not failed.

Which PySpark operation introduces a shuffle, shown as an Exchange in the physical plan?

• Adf.filter(F.col("amount") > 0)
  A filter works on each partition independently and moves no data.
• Bdf.withColumn("tax", F.col("amount") * 0.07)
  Adding a column is a narrow transformation within each partition.
• Cdf.groupBy("store_id").agg(F.sum("amount"))
  Correct. Grouping must bring all rows with the same key together, which requires an exchange between executors.
• Ddf.select("store_id", "amount")
  Selecting columns is a narrow transformation that needs no data movement.

In Unity Catalog ABAC, what tells a column mask policy which columns to protect?

• AThe column's data type, matched automatically by Unity Catalog
  Data types are not the matching mechanism for ABAC policies.
• BThe table owner's group membership
  Ownership does not select which columns a policy protects.
• CGoverned tags on the columns, matched by the policy's conditions
  Correct. Policies use conditions on governed tags to find which tables and columns they apply to.
• DThe order of columns in the table schema
  Column position has no role in ABAC.