Ancestors

Written by Rachel on 2025-02-02 at 04:22

Reading a post on cluster-api, but not sure if it'll convince me to rebuild the Talos cluster using it

https://a-cup-of.coffee/blog/talos-capi-proxmox/

(@mmeier@social.mei-home.net you might find this interesting)

=> More informations about this toot | More toots from rachel@transitory.social

Written by Michael on 2025-02-02 at 08:22

@rachel That was definitely an interesting read!

On the one hand, I really like the fact that they could just define the machines and the entire cluster in a Yaml file. Definitely fewer lines than my Ansible playbooks. But I'm still a bit wary about the fact that you can't log into Talos machines and that you need an entire other k8s cluster to use the Cluster API.

But even with that, it's still tempting. But I should probably first get my Raspberry Pis under management of some kind.

=> More informations about this toot | More toots from mmeier@social.mei-home.net

Written by Rachel on 2025-02-02 at 12:25

@mmeier@social.mei-home.net yup you'd need something additional because you wouldn't be using the Proxmox infra provider

I might give it a shot and see if I can recreate the new Talos cluster in parallel

=> More informations about this toot | More toots from rachel@transitory.social

Written by Rachel on 2025-02-02 at 13:04

@mmeier@social.mei-home.net ok ok if this works I have an amusing idea

I already have that other Talos cluster right? I'll use that as the cluster-api controller for the CAPI cluster, I can just delete the extra worker nodes later

I can also either run vault there for the CAPI cluster, or just use it as secret store via the Kubernetes provider in ExternalSecrets operator

I could also maybe run ArgoCD on that control cluster?

The main trick is gonna be the specialty nodes, like the rook nodes and GPU nodes because they need devices passed in

=> More informations about this toot | More toots from rachel@transitory.social

Written by Michael on 2025-02-02 at 15:55

@rachel That does sound like a good idea, combining all the "foundational" tools into one cluster. Would you do over multiple nodes, or reduce it to a single one?

=> More informations about this toot | More toots from mmeier@social.mei-home.net

Toot

Written by Michael on 2025-02-02 at 15:56

@rachel I've also seen that there are Infrastructure controllers for Tinkerbell and MaaS, both of which I want to look into at some point to make my baremetal host deployments a more complicated. 😁

=> More informations about this toot | More toots from mmeier@social.mei-home.net

Descendants