This page is for developers preparing for the Databricks Certified Associate Developer for Apache Spark exam. It lists the exam format, fee, time limit and retake rules from Databricks, maps each official content area to a free practice set, and walks through sample questions with explanations. The questions are written by us and are not real exam items.
Facts verified Nov 8, 2026 against Databricks. Dates that have passed are removed automatically.
Facts verified Nov 8, 2026 against Databricks. Dates that have passed are removed automatically.
| Item | Detail | Official source |
|---|---|---|
| Number of questions | 45 scored questions (up to 55 total) | Databricks |
| Time limit | 90 minutes | Databricks |
| Passing score | Not published | Databricks |
| Exam fee | USD 200 | Databricks |
| Retake policy | 14-day wait period between attempts | Databricks |
| Validity | 2 years | Databricks |
| Renewal | Recertify every 2 years by passing current live exam | Databricks |
| Delivery | Online proctored or test center | Databricks |
| Eligibility | No prerequisites; 6+ months experience recommended | Databricks |
| Score release | Unofficial immediately; badge after proctor review | Databricks |
Facts verified Nov 8, 2026 against Databricks. Dates that have passed are removed automatically.
Interactive Apache Spark resources on PracticeTestGeeks, free to start:
According to Databricks, βTotal number of questions you may see: Up to 55 questions (45 scored plus up to 10 unscored items)β[1] Practice with the free Apache Spark practice test on PracticeTestGeeks (9 questions with answer explanations).
The Apache Spark Developer Associate exam is run by Databricks. It tests whether you can build and reason about Spark applications using the DataFrame API, Spark SQL, Structured Streaming and related tools.
Format: you may see up to 55 questions, of which 45 are scored. The rest are unscored items that you cannot tell apart from the others. The time limit is 90 minutes. You can sit the exam online with a proctor or at a test center. The fee is USD 200, plus applicable taxes as required by local law.
Eligibility: there are no prerequisites. Databricks recommends 6+ months of hands-on experience with the tasks in the exam guide.
Scoring: Databricks does not publish the passing score, because it is set through statistical analysis and can change as questions are updated. Unofficial results are shown right after the exam. For a passing result, Databricks reviews the proctor recording before it issues your digital badge.
Retakes and renewal: you can retake the exam as many times as you like, but you must wait 14 days between attempts. The certification is valid for 2 years. To recertify you must pass the current live version of the exam.
Check the official Databricks exam guide and certification FAQ before you register, since fees, policies and the outline can change.
The content areas below are copied from the official Apache Spark content outline published by Databricks, with the share of the exam each area carries. The right-hand column shows which of our practice sets covers it.
| Official content area | Weight | Practice on this site |
|---|---|---|
| Apache Spark Architecture and Components[2] | 20% | Mixed practice tests above Know what the driver, executors and cluster manager each do, and how a job is split into stages and tasks. |
| Developing Apache Sparkβ’ DataFrame/DataSet API Applications[2] | 30% | Apache Spark Spark SQL and DataFrames Practise selecting, filtering, joining, grouping and writing DataFrames, and know which operations are transformations and which are actions. |
| Structured Streaming[2] | 10% | Apache Spark Spark Streaming Understand triggers, output modes, checkpoints and how a streaming query differs from a batch query. |
| Troubleshooting and Tuning Apache Spark DataFrame API Applications[2] | 10% | Apache Spark Spark Performance Tuning Learn to recognise shuffles, skew and caching trade-offs, and which configuration setting addresses each symptom. |
| Using Pandas API on Apache Spark[2] | 5% | Mixed practice tests above Compare pandas-style code with native DataFrame code, and know when data is distributed and when it is collected to the driver. |
| Using Spark Connect to deploy applications[2] | 5% | Mixed practice tests above Understand the client-server split in Spark Connect and how it changes where your code runs. |
| Using Spark SQL[2] | 20% | Apache Spark Spark SQL and DataFrames Write and read SQL for joins, aggregates, window functions and views, and relate each query to its DataFrame equivalent. |
These are practice questions written for this site, not questions from the real exam. Each explanation shows why the right answer is right and why the others are wrong.
Question: What is the default number of partitions when creating an RDD from a collection using sc.parallelize()?
Answer: C. Depends on the SparkContext default parallelism
Explanation:
The correct answer is that the partition count depends on the SparkContext default parallelism. When you call sc.parallelize() without a partition count, Spark uses its default parallelism setting, which usually follows the cores available to the application. A fixed value of 1, 2 or 10 is wrong because Spark does not hard-code a partition count for this call; the number changes with the environment and configuration.
Rule to remember: when you do not pass a partition count, Spark chooses one from the cluster or session settings. If the number matters, set it explicitly and check it with getNumPartitions().
Question: What is the purpose of the spark.sql.shuffle.partitions configuration?
Answer: B. Controls the number of partitions used for shuffle operations like joins and aggregations
Explanation:
The correct answer is that spark.sql.shuffle.partitions controls how many partitions are used after a shuffle, such as in joins and aggregations. The default is 200. Reading files is a different matter, because input partitioning is driven by file and split sizes, so the first option is wrong. Maximum partition size in bytes is a separate setting. The number of concurrent shuffle readers is also not what this setting controls.
Rule to remember: this setting shapes the data after a wide transformation, not the data you read in. Too few partitions give large tasks, and too many give tiny ones.
Question: What is the role of the Spark Driver in a Spark application?
Answer: B. Hosts the SparkContext, maintains the DAG, and coordinates task scheduling across executors
Explanation:
The correct answer is that the Driver hosts the SparkContext, maintains the DAG and coordinates task scheduling across executors. Executing the task computations on partitions is the executors' job, so the first option is wrong. Allocating containers and managing cluster resources belongs to the cluster manager. Storing intermediate shuffle data happens on the executors' side, not in the Driver.
Rule to remember: the Driver plans and coordinates, executors do the work, and the cluster manager hands out resources. Many architecture questions come down to naming which of these three owns a task.
Start with the official Databricks exam guide and note which content areas you can already explain without looking anything up. Then take the practice set that matches your weakest area, such as architecture, SQL and DataFrames, streaming or performance tuning.
After each set, read every explanation, including those for questions you got right. For each miss, write a short note naming the rule you forgot. Then reproduce the idea in a notebook or local Spark session: run the code, look at the query plan, and change one setting at a time to see what happens.
The official outline includes Pandas API on Spark and Spark Connect, which our sets do not cover directly. Cover those with the Databricks and Apache Spark documentation and small hands-on exercises. The Core and RDDs set helps you understand partitions and lazy evaluation, which support many DataFrame questions. The GraphX and MLlib sets are extra practice and are not listed in the official outline.
In the final days, retake sets where you missed questions and go through your notes. Do a timed run to practise pacing against the 90-minute limit, then rest before the exam.
The fee is USD 200, plus applicable taxes as required by local law. Check the official Databricks certification page before you register, since fees can change.
You may see up to 55 questions, of which 45 are scored. The rest are unscored items. The time limit is 90 minutes. The exam can be taken online with a proctor or at a test center.
Databricks does not publish it. It is set through statistical analysis and can change as exams are updated with new questions. Unofficial results appear right after the exam, and a digital badge follows once Databricks reviews the proctor recording for a passing result.
Yes. You can retake it as many times as you want, but you must wait 14 days between attempts. You register again for each attempt, so check the current fee and terms on the Databricks site.
No. Databricks lists no prerequisites, but it recommends 6+ months of hands-on experience with the tasks in the exam guide. Related training is also encouraged by Databricks.
It is valid for 2 years from the date it is awarded. To recertify, you must pass the current live version of the exam, which may differ from the version you originally took.