Scaling your services is easy when your services are stateless. However, in real life applications have to deal with state - how do you scale stateful software systems and their data? This is where it may become scary - while modern storage solutions (from Etcd to Cassandra) promise to solve this problem for us, operating these solutions may be tricky. Even if you don’t have to reimplement underlying consensus algorithms (like Paxos and Raft), you still need to understand how they work to operate, debug, and scale your systems. This talk explains the mechanics of different consensus algorithms in everyday terms. This talk is not an academic paper: instead of providing formal proofs and scientific language, it focuses on practical aspects of consensus algorithms’ usage. It explains actionable insights in (relatively) simple words without pitching a concrete approach or technology as a magic cure.