I'm managing a number of client websites, mainly WordPress and Laravel, on separate Hetzner CX32 instances—4 vCPU, 8 GB RAM, and 80 GB disk. Currently, I'm running about 25 of these, and it's become quite chaotic trying to manage them all individually. To simplify my setup, I'm considering launching a lightweight k3s cluster with Rook Ceph for shared storage across nodes. I'm thinking about starting with 5 Hetzner CX42 instances—each having 8 vCPU, 16 GB RAM, and 160 GB disk—and plan to scale as more clients come in. My questions are: 1. Is this hardware sufficient for running k3s with Rook Ceph reliably for production work? 2. What's the recommended minimum to prevent future issues? 3. Are there any specific challenges I should anticipate while using Ceph on Hetzner, like networking or disk performance? I want to avoid overbuilding but also don't want to be overwhelmed by under-provisioning. Any insights from those with similar setups would be greatly appreciated!
1 Answer
Honestly, I recommend against going the Ceph route. Instead, you could use the Hetzner CSI driver to utilize their block storage directly. It's a lot faster and much more reliable.
I need shared storage to avoid the hassle of volumes getting attached and detached every time a pod moves. Having shared storage makes scaling much smoother for all my sites.