Interview prep guide / Apache Spark / Updated September 2026 for Spark 4.2
Apache Spark interview guide: 161 questions, and the follow-up behind each one
Written for data engineer and data architect loops. Each question carries three things: the answer a strong candidate gives, the follow-up the interviewer asks next, and the reply that loses the room. Fourteen sections go from the driver down to incident drills and a coding round.
How to use this without wasting the hours
- Read the question. Say your answer out loud before you open the card. Then compare.
- Open the card. If your answer missed what the follow-up tests, you were not done.
- Mark it drilled only when you can give the strong answer in under 90 seconds without reading.
- Start the session clock. The rail shows whether your pace finishes the guide in the time you have.
What to do after the last card
Reading is the cheap part. Pick the three sections where the follow-up caught you most often and do them again with the clock running. Then take one incident drill and reproduce it locally: a 4 core Spark session with a deliberately skewed key is enough to see the straggler task, the spill, and the fix. That paragraph of "here is what I saw in the UI and here is what I changed" is worth more in a loop than any list of configs.
Two disciplines that separate people who pass Spark rounds from people who know Spark: answer in the order of what happens (driver, plan, stages, shuffle, executor), and quantify. File sizes, partition counts, executor memory, batch duration. Engineers who put numbers on things get believed.