Request Discovered Services and Replicate Logs

Distributed Services with Go — by Travis Jeffery (53 / 84)

👈 Discover Services with Serf | TOC | Test Discovery and the Service End-to-End 👉

Let’s build on our service discovery to add replication in our service so that we store multiple copies of the log data when we have multiple servers in a cluster. Replication makes our service more resilient to failures. For example, if a node’s disk fails and we can’t recover its data, replication can save our butts because it ensures that there’s a copy saved on another disk.

In the next chapter, we’ll coordinate the servers so our replication will have a defined leader-follower relationship, but for now we simply want the servers to replicate each other when they discover each other and not worry about whether they should, like the scientists from Jurassic Park. Our goal for the rest of this chapter is to build something simple that makes use of our service’s discovery and sets us up for our coordinated replication in the next chapter.

Discovery alone isn’t useful — so what if a bunch of computers discover each other and they just sit there doing nothing? Discovery is important because the discovery events trigger other processes in our service like replication and consensus. When servers discover other servers, we want to trigger the servers to replicate. We need a component in our service that handles when a server joins (or leaves) the cluster and begins (or ends)…