Programming a distributed database is simple. Write a database, glue some RPC on it, and test a few queries. In reality, the implementation of distributed databases are difficult, inelegant and are barely recognizable compared to their single-node counterparts. In this talk, we'll see how real world distributed database actually work, communicate and are tested. We'll trade RPC for messaging, unit-tests for QuickCheck, and micro-benchmarks for multi-week stress tests.