If you've prepped for a NoSQL interview by memorizing "CAP theorem stands for Consistency, Availability, Partition Tolerance," you've done maybe 10% of the work. The actual interview question is almost never the definition. It's the classification problem: given MongoDB, DynamoDB, or Cassandra, tell me whether it leans CP or AP and why, then explain what a user actually experiences when they read from a replica that hasn't caught up yet.
That pattern repeats across every database. MongoDB questions test whether you know when to embed versus reference data, and whether you can reason about shard key cardinality instead of just naming the feature. Redis questions are rapid-fire "what structure would you use for X", and the wrong answer between allkeys-lru and volatile-lru can look like you've never actually run Redis in production. DynamoDB questions assume you understand there are no joins, so single table design and Query versus Scan aren't trivia, they're the difference between a fast production system and a scan that silently reads your entire table.
I put together 42 questions across these three databases, covering aggregation pipelines, replica sets, distributed locks, cache stampedes, hot partitions, and RCU/WCU math, the stuff that actually comes up once you're past the "what is NoSQL" opener.
Full questions and answers here: https://devencyclopedia.com/blog/nosql-interview-questions
Top comments (0)