Problem:
I’m looking to bring wider Kubernetes hosts into my foreman estate, I’m playing around in a test env before taking this forward, but I’d appreciate thoughts and input to try to focus the approach to evaluate if this could be a production pattern.
I’m currently running Kubernetes hosts with a traditional foreman workflow provision host - traditional OS, puppet configures the node, it joins the cluster done.
I’d like to try a new pattern of provisioning a Fedora CoreOS host, doing a super basic / small puppet run (configure ssh keys for example) join the Kubernetes cluster and register (just for visibility).
I’ve not had a positive experience with CoreOS provisioning on Foreman and not really done much with it and Foreman for a while, but things have moved on, now seems the time to try to utilise Fedora CoreOS as an ephemeral platform for K8s and I’ve love to try to keep host provisioning and management simplified and centralised in foreman. I’m looking to provision virtual and physical hosts with FCOS (never done physical hosts on FCSOS before).
Documentation seems quite fragmented and often outdated, so I’d love some advice / pointers on
- solid / current docs of foreman FCOS provisioning
- opinion and experience on a puppet run inside FOCS to configure
- physical hardware provisioning FCOS with foreman
- Lifecycle FCOS management with Foreman (not Katello) - workflow and tooling
Any real world experience positive and negative of this pattern or similar patterns I’d love to hear from you.