Python for DevOps

Chapter 12. Container Orchestration: Kubernetes

If you are experimenting with Docker, or if running a set of Docker containers on a single machine is all you need, then Docker and Docker Compose would be sufficient for your needs. However, as soon as you move from the number 1 (single machine) to the number 2 (multiple machines), you need to start worrying about orchestrating the containers across the network. For production scenarios, this is a given. You need at least two machines to achieve fault tolerance/high availability.

In our age of cloud computing, the recommended way of scaling an infrastructure is “out” (also referred to as “horizontal scalability”), by adding more instances to your overall system, as opposed to the older way of scaling “up” (or “vertical scalability”), by adding more CPUs and memory to a single instance. A Docker orchestration platform uses these many instances or nodes as sources of raw resources (CPU, memory, network) that it then allocates to individual containers running within the platform. This ties into what we mentioned in Chapter 11 in regards to the advantages of using containers over classic virtual machines (VMs): the raw resources at your disposal will be better utilized because containers can get these resources allocated to them on a much more granular basis than VMs, and you will get more bang for your infrastructure buck.

There has also been a shift from provisioning servers for specific purposes and running specific software packages ...