In this chapter, you will learn how to create real Swarm clusters with thousands of nodes; specifically we'll cover the following topics:
- Tools to deploy large Swarms
- Swarm2k: One of the largest Swarm mode cluster ever built, made of 2,300 nodes
- Swarm3k: The second experiment, a cluster with 4,700 nodes
- How to plan hardware resources
- HA cluster topologies
- Swarm infrastructures management, networking, and security
- Monitoring dashboards
- What you learned from the Swarm2k and Swarm3k experiments
With Swarm Mode, we can easily design a production-grade cluster.
The principles and architecture we're illustrating here are important in general and give a foundation on how to architect production installations, regardless of the tools. However, from a practical point of view, the tools to use are also important.
At the time of writing this book, Docker Machine was not the ideal single tool to use for large swarms setups, so we're demonstrating our production-scale deployments with a tool born alongside with this book that we already introduced in Chapter 1, Welcome to Docker Swarm: belt (https://github.com/chanwit/belt). We'll use it in conjunction with Docker Machine, Docker Networking, and the DigitalOcean's doctl command.
In Chapter 5, Administer a Swarm Cluster you'll learn how it's possible to automate the creation of Swarms; especially, how to quickly join a massive number of workers with scripts and other mechanisms, such as Ansible.
