Kafka Streams

7 h
requires KFK-07

From your world. dbt joins between a fact and a dimension, but continuous

Explain this to a teammate who knows dbt well. Cover how a Kafka Streams KStream-KTable join enriches orders with customers. Also cover why it is not just "a join in SQL."

Rubric — a good answer covers:

KStream = append-only facts (orders). KTable = latest value per key, a changelog (customers, a conformed dimension).
The stream drives the join. Each order looks up the current customer. A change in the table alone emits nothing.
Co-partitioning (same key, same partition count) is required. The other option is a GlobalKTable.
State lives in a state store that a compacted changelog topic backs. A crashed instance recovers by replaying the changelog.
Event-time windows (not processing time) keep aggregations correct when records arrive in a different order.
TopologyTestDriver runs the topology in-process (no broker, no RocksDB), deterministically in event time. This is how you unit-test the logic.
Analogy: it is the fact-to-dimension lookup in dbt, but continuous and incremental. It always uses the latest dimension value at the time the event arrives.