I'm planning to implement Ceph in a Kubernetes environment after learning the basics on bare metal. Generally, Rook is the go-to method for deploying Ceph on Kubernetes, but I've come across several comments suggesting it might not be the best approach. I'm curious to hear from anyone who has used Ceph without Rook. Are these warnings legitimate, or is it possible to run Ceph on Kubernetes without it?
2 Answers
If you're managing Ceph within Kubernetes, I strongly recommend using Rook. It's designed for that environment and makes things a lot simpler. However, you can run Ceph outside of K8s using other methods like cephadm or Ansible if you prefer.
I have Ceph running on a Proxmox cluster, and for Kubernetes, I connected it using the ceph-csi-operator. Everything's been smooth for us, very straightforward to set up by following the instructions on their GitHub repo.
Just to clarify, do you have Ceph on Proxmox, and your Kubernetes cluster uses that Ceph? Or is Ceph set up on both systems?