Databricks Interview Vault | 500+ Q/A

Spark & Databricks Interview Lab — 350+ Real Q&A with Working Code

The complete interview preparation system for Data Engineers — 200 production scenarios + 146 hands-on coding questions, all in one searchable file.

Why this exists

Interviews for Spark and Databricks roles have changed. Nobody asks "What is an RDD?" anymore. They ask:

• "One task runs 45 minutes while 199 finish in seconds — what's happening?"
• "Your MERGE keeps throwing ConcurrentAppendException. Walk me through it."
• "The bill doubled last month. Where do you look first?"

These are scenario rounds — and they're where most candidates freeze, because tutorials teach syntax, not judgment. This lab is built entirely around that gap.

What you get — section by section

📌 Part I — 200 Real-World Scenario Questions

100 on Spark Core & Performance: data skew and salting, broadcast join traps, shuffle spill and the memory math behind it, OOM exit code 137, caching mistakes, AQE behavior, small-files problems, partition sizing heuristics, plan bloat from withColumn loops, GC pressure, speculative execution, driver-side bottlenecks — through to the classic closer, "walk me through what happens when you submit a job."

100 on Delta, Streaming & Platform: MERGE patterns and multi-clause ordering, optimistic concurrency and conflict engineering, time travel and VACUUM policy, schema evolution, Z-order vs liquid clustering, exactly-once streaming anatomy, watermarks and output modes, Kafka configs, Auto Loader modes, DLT, Unity Catalog governance (row filters, lineage, service principals), Workflows, serverless vs classic, spot strategy, cost attribution, Delta Sharing — ending with a full greenfield lakehouse design capstone.

Every single answer follows one structure: Do → Don't → Why → Use case. That's deliberate — it's the answer shape interviewers score highest, and after 200 repetitions it becomes how you naturally talk.

🐍 Part II — 50 Python Questions (Mid → Advanced)

Decorators with arguments, generators and memory traps, the mutable-default-argument bug, LRU caching, GIL implications, context managers, sliding windows, rate limiters — the exact patterns that show up in DE coding rounds. Every answer is runnable code with a Copy button.

🗃️ Part III — 49 SQL / Spark SQL Questions (Mid → Advanced)

Window functions in depth, gaps & islands, retention cohorts, sessionization, SCD Type 2 in pure SQL, dedup keeping latest, GROUPING SETS, NULL logic traps, Delta MERGE syntax, reading query plans before you ship.

⚡ Part IV — 47 PySpark Questions (Mid → Advanced)

pandas UDFs, mapInPandas with pre-loaded ML models, foreachBatch with idempotent MERGE, Auto Loader, stream-stream joins, watermark dedup, DeltaTable API for SCD2, JDBC partitioned reads, bucketed joins, unit-testing PySpark — production-grade code, not toy examples.

How it works

• One HTML file. Opens in any browser — laptop, tablet, phone. No login, no app, no internet needed after download.
• Searchable sidebar + difficulty filters (Mid / Advanced) so you can drill exactly what your interview needs.
• Copy button on every code answer — paste straight into a Databricks notebook and run it.
• Yours forever. No subscription, no expiry.

Who this is for

• Data Engineers with 2–10 years of experience targeting product companies, GCCs, or Databricks-heavy roles
• Engineers moving from on-prem/Informatica/SQL backgrounds into Spark who need the "production judgment" layer fast
• Anyone with an interview in 1–2 weeks who needs structured, high-density prep instead of 40 hours of videos

Not for you if: you've never touched Spark at all — this assumes basic familiarity and builds interview-level depth on top.

How to prepare with it

1. Week 1: Read 20 scenarios a day. Say answers out loud in the Do/Don't/Why/Use case structure before reading them.
2. Week 2: Work through the coding sections in a notebook — run every snippet, break it, fix it.
3. Day before: Use the search box to re-drill your weak topics (search "skew", "MERGE", "watermark"...).

Built by an engineer who's sat on both sides of the interview table. These aren't scraped question lists — every scenario is written from real production failures and real interview rounds, answered the way strong candidates actually answer them.