agentleFS
Aidas-devAidas-dev/k8s-agent-skills/skills/talos-terraform/SKILL.md

talos-terraform

Skill4 starsChanged 5 months agoMITView on GitHub

Talos Linux Terraform provider (siderolabs/talos) — machine secrets, machine configuration, config apply, bootstrap, and the sidero-blessed production pattern (live secrets + patch-based config, upgrades owned elsewhere).

The file

Files
SKILL.md

talos-terraform — Terraform Provider

For CLI operations see talosctl. For machine config YAML see talosconfig.

Provider: siderolabs/talos Latest stable: v0.11.0 (2026-04-27) — SDK targets Talos v1.13 config line Next (alpha): v0.12.0-alpha.x (2026-06-25) — adds talos_machine/talos_cluster, based on Talos v1.14. NOT stable.

Provider Setup

terraform {
  required_providers {
    talos = {
      source  = "siderolabs/talos"
      version = "~> 0.11.0"   # stable
    }
  }
}

Sidero-Blessed Production Pattern

Production-proven pattern for managing an EXISTING Talos cluster with Terraform. Core idea: generate config from live cluster secrets + patch files, apply with the stable talos_machine_configuration_apply resource, and let a separate system own OS/K8s upgrades.

Why NOT talos_machine (alpha)

talos_machine exists only in v0.12.0-alpha. It has a Create/Update bug with image = null — Optional+Computed fields plan as Unknown, and the provider attempts an upgrade with an empty image ref. If OS upgrades are owned elsewhere (e.g. tuppr TalosUpgrade CRs), avoid talos_machine entirely.

Architecture

locals (node map + live secrets + talosconfig)
  → data talos_machine_configuration (per node: live secrets + patch file)
    → resource talos_machine_configuration_apply (apply config to node)
      → apply_mode = staged_if_needing_reboot
      → lifecycle prevent_destroy

The Pattern

# 1. Cluster secrets from the LIVE cluster — NEVER regenerate.
#    Regenerating rotates all certs. Extracted once, gitignored.
locals {
  cluster_secrets = jsondecode(file("${path.module}/local_cluster_secrets.json"))
}

# 2. Client config from the cluster's talosconfig (os:admin).
#    certs.os = machine CA (for config gen) — NOT a valid apid client.
#    apid RBAC requires subject O=os:admin.
locals {
  talos_cfg = yamldecode(file("${path.module}/talosconfig.yaml"))
  talos_client_config = {
    ca_certificate     = local.talos_cfg.contexts[local.talos_cfg.context].ca
    client_certificate = local.talos_cfg.contexts[local.talos_cfg.context].crt
    client_key         = local.talos_cfg.contexts[local.talos_cfg.context].key
  }
}

# 3. Generate machine config per node: live secrets + version contract +
#    the applied config as patch (byte-exact reproduction of node state).
data "talos_machine_configuration" "node" {
  for_each = local.nodes

  cluster_name       = var.cluster_name
  cluster_endpoint   = each.value.endpoint
  machine_type       = each.value.role          # controlplane | worker
  machine_secrets    = local.cluster_secrets
  kubernetes_version = each.value.kubernetes_version
  talos_version      = each.value.talos_version  # RECOMMENDED: pin it

  config_patches = [file("${path.module}/patches/${each.key}-apply.yaml")]
}

# 4. Apply config — config-only, no OS version sync.
resource "talos_machine_configuration_apply" "node" {
  for_each = local.nodes

  node                        = each.value.ip
  client_configuration        = local.talos_client_config
  machine_configuration_input = data.talos_machine_configuration.node[each.key].machine_configuration
  apply_mode                  = "staged_if_needing_reboot"

  lifecycle {
    prevent_destroy = true   # live nodes: reconcile config only, never recreate
  }
}

Key Decisions (from production)

DecisionWhy
Secrets from live cluster (extracted, gitignored)Regenerating = cert rotation = cluster outage
os:admin client config from talosconfigcerts.os (machine CA) is NOT accepted by apid RBAC
talos_machine_configuration_apply (stable)talos_machine is alpha + broken with image = null
apply_mode = staged_if_needing_rebootConfig-only change applies immediately; reboot-requiring change stages for next reboot
No image setOS version NOT managed by TF — a dedicated upgrade system owns upgrades (see tuppr skill: TalosUpgrade/KubernetesUpgrade CRs)
config_patches = the applied config as patchByte-exact reproduction of what's on the node
prevent_destroy = trueLive nodes never destroyed/recreated by TF

Core Resources (v0.11.0 stable)

ResourcePurpose
talos_machine_secretsGenerate cluster secrets (bootstrap only — never on existing cluster!)
talos_machine_configuration (data source)Generate machine config for a node type
talos_machine_configuration_applyApply config to a node (stable, production path)
talos_machine_bootstrapBootstrap etcd (first CP only)
talos_cluster_kubeconfigFetch kubeconfig
talos_image_factory_schematicImage factory schematic (secureboot)

Data sources: talos_client_configuration, talos_cluster_health, talos_machine_disks, Image Factory (versions/extensions/overlays/urls).

Ephemeral variants (new in 0.11): talos_machine_configuration, talos_machine_secrets, talos_client_configuration, talos_cluster_kubeconfig — keep secrets OUT of state. Requires Terraform 1.10+. Use _wo (write-only) attributes (Terraform 1.11+) with them.

Bootstrap Workflow (new cluster)

# 1. Generate secrets ONCE (fresh cluster only)
resource "talos_machine_secrets" "this" {}

# 2. Generate config per node type
data "talos_machine_configuration" "controlplane" {
  cluster_name     = "example-cluster"
  cluster_endpoint = "https://10.5.0.2:6443"
  machine_type     = "controlplane"
  machine_secrets  = talos_machine_secrets.this.machine_secrets
  talos_version    = "v1.13.8"          # pin it — new provider SDK enables new config features
  config_patches = [file("patches/controlplane.yaml")]
}

# 3. Apply config to first CP (insecure maintenance mode)
resource "talos_machine_configuration_apply" "first_cp" {
  node                        = "10.5.0.2"
  client_configuration        = talos_machine_secrets.this.client_configuration
  machine_configuration_input = data.talos_machine_configuration.controlplane.machine_configuration
}

# 4. Bootstrap etcd
resource "talos_machine_bootstrap" "this" {
  depends_on           = [talos_machine_configuration_apply.first_cp]
  node                 = "10.5.0.2"
  client_configuration = talos_machine_secrets.this.client_configuration
}

Why pin talos_version?

Without it, a new provider version with a new Talos SDK major enables new machineconfig features by default → unexpected behavior. Example values: v1.12, v1.13.8, 1.13.

Common Mistakes

  • Regenerating secrets on an existing cluster — talos_machine_secrets is bootstrap-only. On a live cluster, extract secrets from the cluster and store gitignored. Regenerating rotates all certs → outage.
  • Using certs.os as apid client — The machine CA cert is for config generation, not apid auth. apid RBAC requires O=os:admin. Use the cluster's talosconfig client cert.
  • talos_machine with image = null — Alpha resource, Create/Update bug: Optional+Computed plans Unknown, provider tries empty-image upgrade. Use talos_machine_configuration_apply and let a dedicated system own upgrades.
  • No talos_version pin — New provider + new SDK major silently enables new config features. Always set it.
  • No prevent_destroy — TF can destroy/recreate live nodes. Always set lifecycle { prevent_destroy = true } for production nodes.
  • apply_mode default — auto reboots immediately when needed. staged_if_needing_reboot avoids surprise reboots on config-only changes.
  • kubernetes_version mismatch — In talos_machine_configuration, kubernetes_version only sets component images at bootstrap; it does NOT trigger upgrades. Sync it with whatever owns upgrades.
  • Client-side vs write-only confusion — With ephemeral resources, use _wo attributes everywhere to keep secrets out of state (requires Terraform 1.11+).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.