Chapter 8 Coordinate Your Services with Consensus

Distributed Services with Go — by Travis Jeffery (56 / 84)

👈 What You Learned | TOC | What Is Raft and How Does It Work? 👉

Distributed services are like commercial kitchens. Imagine a small restaurant opens up with one stove and one cook. Patrons discover the restaurant and tell their friends, and business is booming. But the kitchen struggles with the mob of customers and sometimes the stove breaks, forcing the restaurant to close for the night and lose business. So the restaurant hires two more cooks and buys two more stoves. The cooks keep up with orders now but they make mistakes: they mix up appetizers and entrees; they mix up tables; they make double of one order while forgetting to make another. They lack coordination. So the kitchen hires a chef to oversee and coordinate the kitchen. When an order comes in, the chef divides the order and assigns the appetizers, entrees, and deserts to the cooks who prepare the food timely and correctly. The patrons love the fast, quality service, and the kitchen becomes world-renowned.

In this chapter, we look at the chef of distributed services: consensus. Consensus algorithms are tools used to get distributed services to agree on shared state even in the face of failures. In ​Request Discovered Services and Replicate Logs​, we naively implemented replication in our service, and the servers replicate each other in a cycle, making infinite copies of the same data. We need to put the servers in…