We have seen previously that by using orchestration tools such as Ansible we can reduce the work required to deploy, manage, and maintain a Ceph cluster. We have also seen how these tools can help you discover available hardware resources and deploy Ceph to them.

However, using Ansible to configure bare-metal servers still results in a very static deployment, possibly not best suited for today's more dynamic workloads. Designing Ansible playbooks also needs to take into account several different Linux distributions and also any changes that may occur between different releases; systemd is a great example of this. Furthermore, a lot of development in orchestration tools needs to be customized to handle discovering, deploying, and managing Ceph. This is a common theme that the Ceph developers have thought about; with the use of Linux containers and their associated orchestration platforms, they hope to improve Ceph's deployment experience.

One such approach, which has been selected as the preferred option, is to join forces with a project called Rook. Rook works with the container management platform Kubernetes to automate the deployment, configuration, and consumption of Ceph storage. If you were to draw up a list of requirements and features which a custom Ceph orchestration and management framework would need to implement, you would likely design something which functions in a similar fashion to Kubernetes. So it makes sense to build functionality on top of the well-established Kubernetes project, and Rook does exactly that.

One major benefit of running Ceph in containers is that is allows collocation of services on the same hardware. Traditionally in Ceph clusters it was expected that Ceph monitors would run on dedicated hardware; when utilizing containers this requirement is removed. For smaller clusters, this can amount to a large saving in the cost of running and purchasing servers. If resources permit, other container-based workloads could also be allowed to run across the Ceph hardware, further increasing the Return on Investment for the hardware purchase. The use of Docker containers reserves the required hardware resources so that workloads cannot impact each other.

To better understand how these two technologies work with Ceph, we first need to cover Kubernetes in more detail and actual containers themselves.