Streaming

Cards

1. Which job is fine as a daily batch rather than a stream? easy Multiple choice
2. A Kafka topic has 12 partitions, and a consumer group has 20 consumers. How many of them receive messages? easy Multiple choice
3. Kafka keeps messages after consumers read them. What does that make possible? easy Flashcard
4. A topic must take 300 MB/s, and one consumer instance can process about 10 MB/s. How many partitions does the topic need so consumers keep up? easy Estimate
5. Counting events in fixed, non-overlapping one-minute buckets (12:00–12:01, 12:01–12:02, …) uses a _____ window. A "last 5 minutes, updated every minute" count uses a _____ window. easy Fill in the blank
6. Why do analytics windows usually use event time rather than processing time? medium Flashcard
7. What ordering does Kafka guarantee, and how do you keep one user's events in order? medium Flashcard
8. A Kafka topic with _____ keeps at least the latest message for each key and removes older ones, so the topic stays a complete snapshot of current values. medium Fill in the blank
9. A Kafka consumer can commit its offset before or after processing a message. What does each give you? medium Flashcard
10. A million view events a second must update per-video counts. Why aggregate in the stream before writing to a database? medium Flashcard
11. A Kafka cluster takes 50 MB/s, keeps data for 7 days, and stores 3 replicas. How much disk does that need? medium Estimate
12. You want to group each user's clicks into visits that end after 30 minutes without a click. Which window fits? medium Multiple choice
13. What is change data capture (CDC), and why use it to keep a search index in sync? hard Flashcard
14. A stream job keeps counts in memory. How does it recover after a crash without double-counting? hard Flashcard
15. In a Kappa architecture (stream processing only, no separate batch layer), how do you recompute results after fixing a bug? hard Multiple choice
16. An event arrives after its window's result was already emitted. What can a stream job do with it? hard Flashcard
17. A consumer joins a Kafka consumer group and partitions are rebalanced. What can that cause? hard Multiple choice
18. A stream processor's running estimate that no more events older than time T will arrive is called a _____. Events older than it that arrive anyway are _____. hard Fill in the blank