Free Databricks Certified Associate Developer for Apache Spark Practice Questions

Sample Practice Questions

Try these 10 sample questions from the Databricks Certified Associate Developer for Apache Spark practice bank. Each includes the answer and explanation. Continue with the interactive practice above to review the full question set.

1. 1What enables Spark to process different portions of a dataset in parallel across a cluster?

   1. A. Replicating the complete dataset into every task
   2. B. Partitioning the collection across cluster nodes
   3. C. Executing every transformation in the driver thread
   4. D. Converting every input record into a SQL table

   Answer: B

   Explanation: The guide describes an RDD as "a collection of elements partitioned across the nodes of the cluster that can be operated on in parallel." Partitioning distributes the work; full replication and driver-only execution are not the mechanism.

   Reference: RDD Programming Guide — Spark 4.0.1 Documentation — Overview

2. 2With spark.sql.sources.default unchanged, which format does spark.read.load(path) use by default?

   1. A. JSON
   2. B. CSV
   3. C. Parquet
   4. D. ORC

   Answer: C

   Explanation: The guide specifies "the default data source (parquet unless otherwise configured by spark.sql.sources.default)". An omitted format therefore selects Parquet under this condition.

   Reference: Generic Load/Save Functions — Spark SQL 4.0.1 — Generic Load/Save Functions — default source

3. 3A DataFrame has an amount column. With col imported, which expression returns a DataFrame where amount is doubled under the same column name?

   1. A. df.withColumnRenamed("amount", "amount * 2")
   2. B. df.withColumn("amount", col("amount") * 2)
   3. C. df.drop("amount").withColumn("amount", col("amount") * 2)
   4. D. df.withColumn("double_amount", col("amount") * 2)

   Answer: B

   Explanation: withColumn returns a DataFrame by "adding a column or replacing the existing column that has the same name." Keeping the name amount replaces its expression; double_amount would add a different column.

   Reference: DataFrame — PySpark 4.0.1 API — DataFrame.withColumn

4. 4A filtered DataFrame has 100 partitions. The developer wants 10 partitions using a narrow operation without a shuffle. Which call meets that goal?

   1. A. df.repartition(10)
   2. B. df.coalesce(10)
   3. C. df.repartition(10, 'key')
   4. D. df.repartitionByRange(10, 'key')

   Answer: B

   Explanation: The coalesce API says a reduction "results in a narrow dependency" and "there will not be a shuffle". Repartitioning redistributes data using a shuffle.

   Reference: DataFrame.coalesce — PySpark 4.0.1 API — DataFrame.coalesce — narrow dependency

5. 5In Structured Streaming's programming model, how is a live input data stream represented?

   1. A. As a fixed table that never accepts new rows
   2. B. As a driver-local Python list rebuilt manually on every trigger
   3. C. As an unbounded table that is continuously appended
   4. D. As a single SQL scalar value replacing all previous records

   Answer: C

   Explanation: The guide treats a stream as a "table that is being continuously appended." Queries run incrementally over that unbounded input.

   Reference: Structured Streaming — Getting Started — Spark 4.0.1 — Programming Model

6. 6Which architectural change does Spark Connect provide for client applications?

   1. A. It requires the client application to run in the same process as the Spark driver
   2. B. It separates client applications from the Spark driver using a client-server architecture
   3. C. It runs all DataFrame computations in the client process while the server stores query text
   4. D. It removes the driver and has clients execute directly inside individual worker JVMs

   Answer: B

   Explanation: The overview describes a "decoupled client-server architecture that allows remote connectivity to Spark clusters". Clients are separated from the driver process.

   Reference: Spark Connect Overview — Spark 4.0.1 — Spark Connect Overview — introduction

7. 7How does the data distribution of a pandas-on-Spark DataFrame compare with that of an ordinary pandas DataFrame?

   1. A. Both keep their data on a single machine
   2. B. Pandas-on-Spark is distributed; ordinary pandas is single-machine
   3. C. Pandas-on-Spark is single-machine; ordinary pandas is distributed
   4. D. Both distribute their data across multiple machines

   Answer: B

   Explanation: The guide says "the former is distributed and the latter is in a single machine." Distributed execution is the relevant scalability distinction.

   Reference: From/to pandas and PySpark DataFrames — PySpark 4.0.1 — pandas — distributed versus single-machine data

8. 8Which resilience benefit does Spark provide for RDD-based processing when a cluster node fails?

   1. A. It automatically recovers affected RDD data
   2. B. It guarantees that no task will execute again
   3. C. It requires users to restore every lost partition manually
   4. D. It requires the full application to restart from its first input record

   Answer: A

   Explanation: The guide states, "Finally, RDDs automatically recover from node failures." This provides recovery rather than a guarantee that tasks never need to run again.

   Reference: RDD Programming Guide — Spark 4.0.1 Documentation — Overview

9. 9Which code explicitly reads newline-delimited JSON records from /data/events.json as a DataFrame?

   1. A. spark.read.load("/data/events.json", format="json")
   2. B. spark.read.load("/data/events.json", format="parquet")
   3. C. spark.read.load("/data/events.json", format="orc")
   4. D. spark.read.load("/data/events.json", format="text")

   Answer: A

   Explanation: The guide introduces its example with "To load a JSON file you can use:". The shown load call supplies the JSON format explicitly. Other format choices use different parsers; text would return lines rather than parsed JSON fields.

   Reference: Generic Load/Save Functions — Spark SQL 4.0.1 — Manually Specifying Options — Python JSON example

10. 10Which call renames the existing customer column to customer_id without changing its values?

    1. A. df.withColumnRenamed("customer_id", "customer")
    2. B. df.withColumn("customer_id", lit("customer"))
    3. C. df.drop("customer")
    4. D. df.withColumnRenamed("customer", "customer_id")

    Answer: D

    Explanation: withColumnRenamed(existing, new) "Returns a new DataFrame by renaming an existing column." The original name must precede the replacement name.

    Reference: DataFrame — PySpark 4.0.1 API — DataFrame.withColumnRenamed(existing, new)