Skip to content

Design Docs

This document aims to note-down some quick notes on the general design goals of ClusterTool.

  • We would like to use GO, because it would make it a lot easier to support multiple platforms
  • We also would like to get rid of the requirement to fork a repo
  • We would prefer not to require users to have a complete boatload of folder structure (from said fork) configured for it to even play-nice
  • We would like to cut down on the list of required dependencies and integrate everything into the go-tool as far as we can.

Cluster Design


Storage will be rendered by default by adding OpenEBS LocalPV. This means no clustered-storage is added.

Reasoning behind this:

  • Clustered storage is highly personal
  • Clustered storage has increased complexity


  • Doing R&D to see if VolSync Restic Copy instead of snapshots work well enough


  • We will pack MetalLB as the default LoadBalancer, as it “just works” pretty well
  • We will pack Cilium as CNI, as it’s basically the best CNI out there
  • We will NOT pack ingress by default, as that can quite easily be added after the fact


  • We will add kubeapps, exposed to IP:port by default, to give users an easy copy-paste interface to launch helm-charts
  • We will ship kubernetes-dashboard exposed to IP:Port (different ip than kubeapps) by default as well, to give an easy status and management interface
  • FluxCD bootstrapping will be offered as an option and will automatically also manage the above added manifests


  • A SOPS encryption key will be created by default
  • Regardless of FluxCD install, we will always load the encryption key into the cluster
  • We will offer quick tooling to encrypt and decrypt sensitive data


There are basically multiple parts to it:

Part A. Generating TalosOS configs -> Talhelper and custom code (mixed)

  1. Verify talconfig (only if talos cluster)
  2. Create talsecret (only if talos cluster)
  3. Run complete config generation
  4. Ensure talenv/secrets are also uploaded to the cluster
  5. Ensure default helm values.yaml Files for included charts, are reflected under /manifests

Part B. Bootstrap a Cluster

  1. Applying configs to cluster nodes (if it’s a Talos Cluster) -> Talhelper
  2. Bootstrap Talos if-needed and if it’s a Talos Cluster -> Talhelper
  3. Load our default base charts on bootstrap -> Custom code

Part C.

  1. Update Talos nodes if Talos cluster -> Talhelper with some extra magic
  2. Update k8s on Talos if Talos cluster -> Talos CLI
  3. Update our base charts if needed -> Custom code

Part D. Bootstrapping fluxcd (optional) -> Mostly just a fluxcd command

Other tools:

  • SOPS encryption
  • SOPS decryption
  • Prep working directory

General Code flow for Talos stuff:

  • Select Talos function
  • Have “all” function to select all
  • Render node list function for TUI/GUI from config
  • Or select a list of nodes (now by name, comma sep, future GUI/TUI)
  • Have a healthcheck function that takes node to check
  • Prior to run, run health checks
  • When parsing over, run a healthcheck after each node, warning to confirm if it doesn’t work out
  • Bootstrap only does the first node, after which user can apply

Interesting other features

  • We should craft a function that goes over the IPs listed and fills the storage and networking hardware found on the nodes as comments to the config, so users can easily configure.