feat(k8s): cluster naming (#243)

* chore(netbird): move to kebab naming

Render NetBird object names (groups, setup keys, policies, networks,
network-resources) as lowercase-kebab instead of UPPER_SNAKE, e.g.
YUCCA_PROD_HTZ_FSN1_MGMT → yucca-prod-htz-fsn1-mgmt. The 1Password setup-key
item titles stay UPPER_SNAKE (decoupled) so CI/ansible/talos op:// consumers
keep resolving.

Pin the futo-org/netbird provider to 1.0.2, which fixes the group
resources TF→API decode so a resource-tag group (htz-fsn1 `resources`) can be
renamed in place — no name pin needed.

* update locks

* chore(naming): naming names

CLAUDE.md

@@ -187,3 +187,19 @@ highest tag to staging → production is gated behind a reviewed `promote-produc
   trailing commas, width 120.
 - Generated files are eslint-ignored: `**/fetch-client.ts`, `packages/web/src/locales`, `dist`,
   `build`, `.svelte-kit`.
+
+### Naming
+
+- **Cluster names are themed by workload.** Kubernetes clusters → **Star Wars** (`luke` =
+  staging, `father` = the soon-to-be prod). Ceph clusters → **Dune** (`sietch`, `spice`, …).
+  Choose the next themed name when standing up a cluster; it's the Talos/Ceph cluster name +
+  the `<clustername>` hostname segment.
+- **Node hostnames** follow `<product>-<provider>-<region>-<clustername>-<role>-<nodename>` —
+  e.g. a staging Talos node is `yucca-int-aus-luke-k8s-<word>`:
+  - `product` = `yucca`; `role` = the workload segment (`k8s` for Talos nodes, `ceph` for Ceph).
+  - `provider` / `region` = the **3-letter** `provider_code` / `region_code` from the region's
+    `region.hcl` (austin = `int`/`aus`, htz-fsn1 = `htz`/`fsn`).
+  - `<clustername>` = the themed cluster name (above).
+  - `<nodename>` = auto-picked from the shared name inventory
+    (`tf/shared/modules/node-names/wordlist.txt`) — deterministic per cluster, unique within it;
+    pass an explicit node `name` to override.

tf/deployment/dev/global/dns/region_inputs.tf

@@ -51,6 +51,12 @@ variable "provider_code" {
   default     = null
 }
 
+variable "region_code" {
+  description = "3-letter region segment of node hostnames (null for global)."
+  type        = string
+  default     = null
+}
+
 variable "domain" {
   description = "Region FQDN suffix (null for global)."
   type        = string

tf/deployment/prod/htz-fsn1/region.hcl

@@ -3,6 +3,7 @@ locals {
   role          = "primary"
   site_id       = 40
   datacenter    = "fsn1"
-  provider_code = "htz"
+  provider_code = "htz" # 3-letter provider segment of node hostnames
+  region_code   = "fsn" # 3-letter region segment of node hostnames
   domain        = "prod.fsn1.htz.futo.cloud"
 }

tf/deployment/staging/austin/ceph/region_inputs.tf

@@ -51,6 +51,12 @@ variable "provider_code" {
   default     = null
 }
 
+variable "region_code" {
+  description = "3-letter region segment of node hostnames (null for global)."
+  type        = string
+  default     = null
+}
+
 variable "domain" {
   description = "Region FQDN suffix (null for global)."
   type        = string

tf/deployment/staging/austin/region.hcl

@@ -7,6 +7,7 @@ locals {
   role          = "primary"
   site_id       = null # austin is not a fabric-managed site (no switch fabric)
   datacenter    = "austin"
-  provider_code = "int"
+  provider_code = "int" # 3-letter provider segment of node hostnames
+  region_code   = "aus" # 3-letter region segment of node hostnames
   domain        = "staging.austin.int.futo.cloud"
 }

tf/deployment/staging/austin/talos/clusters.auto.tfvars

@@ -7,7 +7,9 @@
 # reachable across the install reboot.
 
 clusters = {
-  yucca-staging = {
+  # Star Wars-themed cluster name (staging = luke). Node hostnames derive as
+  # yucca-int-aus-luke-k8s-<name> (name auto-picked from the shared inventory).
+  luke = {
     talos_version      = "1.13.4" # latest stable (v1.13.4); default k8s = 1.36.1
     kubernetes_version = "v1.36.1"
 
@@ -105,9 +107,9 @@ clusters = {
     }
 
     nodes = [
-      { name = "staging-cp1", role = "control-plane", address = "10.10.10.47" },
-      { name = "staging-cp2", role = "control-plane", address = "10.10.10.242" },
-      { name = "staging-cp3", role = "control-plane", address = "10.10.10.117" },
+      { role = "control-plane", address = "10.10.10.47" },
+      { role = "control-plane", address = "10.10.10.242" },
+      { role = "control-plane", address = "10.10.10.117" },
     ]
   }
 }

tf/deployment/staging/austin/talos/main.tf

@@ -3,6 +3,8 @@ module "cluster" {
   source   = "../../../../shared/modules/talos-baremetal"
 
   cluster_name       = each.key
+  provider_code      = var.provider_code
+  region_code        = var.region_code
   talos_version      = each.value.talos_version
   kubernetes_version = each.value.kubernetes_version
   talos_schematic_id = each.value.talos_schematic_id

tf/deployment/staging/austin/talos/region_inputs.tf

@@ -51,6 +51,12 @@ variable "provider_code" {
   default     = null
 }
 
+variable "region_code" {
+  description = "3-letter region segment of node hostnames (null for global)."
+  type        = string
+  default     = null
+}
+
 variable "domain" {
   description = "Region FQDN suffix (null for global)."
   type        = string

tf/deployment/staging/austin/talos/variables.tf

@@ -154,7 +154,7 @@ variable "netbird_operator_api_token" {
 }
 
 variable "clusters" {
-  description = "Map of bare-metal Talos cluster specs keyed by short cluster name. Declarative: add/modify an entry in clusters.auto.tfvars + tf:apply."
+  description = "Map of bare-metal Talos cluster specs keyed by themed cluster name (Star Wars; e.g. 'luke'). Declarative: add/modify an entry in clusters.auto.tfvars + tf:apply."
   type = map(object({
     talos_version      = string
     kubernetes_version = optional(string)

tf/deployment/staging/austin/talos/versions.tf

@@ -5,6 +5,11 @@ terraform {
       source  = "siderolabs/talos"
       version = "~> 0.11"
     }
+    # Hostname picks for the talos nodes (node-names module → random_shuffle).
+    random = {
+      source  = "hashicorp/random"
+      version = "~> 3.6"
+    }
     helm = {
       source  = "hashicorp/helm"
       version = "~> 3.1"

tf/deployment/staging/global/dns/region_inputs.tf

@@ -51,6 +51,12 @@ variable "provider_code" {
   default     = null
 }
 
+variable "region_code" {
+  description = "3-letter region segment of node hostnames (null for global)."
+  type        = string
+  default     = null
+}
+
 variable "domain" {
   description = "Region FQDN suffix (null for global)."
   type        = string

tf/deployment/terragrunt.hcl

@@ -54,6 +54,7 @@ locals {
   site_id       = lookup(local.region_meta, "site_id", null)
   datacenter    = lookup(local.region_meta, "datacenter", null)
   provider_code = lookup(local.region_meta, "provider_code", null)
+  region_code   = lookup(local.region_meta, "region_code", null)
   domain        = lookup(local.region_meta, "domain", null)
 }
 
@@ -100,5 +101,6 @@ inputs = {
   site_id       = local.site_id
   datacenter    = local.datacenter
   provider_code = local.provider_code
+  region_code   = local.region_code
   domain        = local.domain
 }

tf/shared/modules/ceph-cluster/main.tf

@@ -15,41 +15,33 @@ terraform {
   }
 }
 
-locals {
-  # Wordlist for auto-assigning host names. Names are unique within a cluster
-  # (not globally) — sietch-ceph-laurel and a hypothetical future
-  # a second cluster's *-ceph-laurel could coexist, disambiguated by FQDN.
-  wordlist = compact(split("\n", file("${path.module}/wordlist.txt")))
-
-  # Operator-declared names are reserved — strip them from the pool so
-  # auto-picked names can't collide within this cluster.
-  explicit_names  = [for h in var.hosts : h.name if h.name != null]
-  available_words = tolist(setsubtract(toset(local.wordlist), toset(local.explicit_names)))
-}
-
-# Full shuffle of available words, seeded by cluster name. Position-indexed:
-# host[i] with name == null uses result[i]. Adding hosts at the tail is safe;
-# existing positions keep their names across applies. Bumping name_seed forces
-# a re-roll of the whole cluster (not recommended once hosts exist).
-resource "random_shuffle" "names" {
-  input        = local.available_words
-  result_count = length(local.available_words)
-  keepers = {
+# Auto host names come from the shared node-names inventory (the wordlist used to
+# live here; it now backs both talos + ceph). cluster_name seeds the shuffle so each
+# cluster gets its own permutation; explicit host names are excluded from the pool.
+module "names" {
+  source       = "../node-names"
   cluster_name = var.cluster_name
   name_seed    = var.name_seed
+  names        = [for h in var.hosts : h.name]
 }
+
+# The wordlist + shuffle moved into node-names — preserve the existing shuffle state
+# so host names don't re-randomize on this refactor.
+moved {
+  from = random_shuffle.names
+  to   = module.names.random_shuffle.names
 }
 
 locals {
   # Resolve each host's final name: explicit or auto-picked from shuffle.
   hosts_computed = [
     for i, h in var.hosts : {
-      name           = h.name != null ? h.name : random_shuffle.names.result[i]
+      name           = module.names.resolved[i]
       bond_ip        = h.bond_ip
       bootstrap      = coalesce(h.bootstrap, false)
       roles          = h.roles
-      hostname_short = "${var.cluster_name}-${var.role_in_hostname}-${h.name != null ? h.name : random_shuffle.names.result[i]}"
-      fqdn           = "${var.cluster_name}-${var.role_in_hostname}-${h.name != null ? h.name : random_shuffle.names.result[i]}.${var.domain}"
+      hostname_short = "${var.cluster_name}-${var.role_in_hostname}-${module.names.resolved[i]}"
+      fqdn           = "${var.cluster_name}-${var.role_in_hostname}-${module.names.resolved[i]}.${var.domain}"
     }
   ]
 

tf/shared/modules/node-names/main.tf

@@ -0,0 +1,51 @@
+# node-names — the shared name inventory for auto-naming nodes/hosts. Resolves a
+# per-slot list of names: an explicit override where given, else a deterministic
+# pick from the wordlist. Names are unique within a cluster (the shuffle is seeded
+# by cluster_name; explicit names are excluded from the pool so they can't collide).
+# Position-indexed: slot[i] with no explicit name gets result[i], so appending
+# slots at the tail is stable (existing slots keep their names across applies).
+terraform {
+  required_version = ">= 1.6"
+  required_providers {
+    random = {
+      source  = "hashicorp/random"
+      version = "~> 3.6"
+    }
+  }
+}
+
+variable "cluster_name" {
+  type        = string
+  description = "Seeds the shuffle so each cluster gets its own permutation."
+}
+
+variable "name_seed" {
+  type        = string
+  default     = "v1"
+  description = "Bump to re-roll a cluster's auto-names. Do NOT change once nodes are deployed (renames every auto-named node)."
+}
+
+variable "names" {
+  type        = list(string)
+  description = "Per-slot name: an explicit string to reserve, or null to auto-pick. Resolved positionally in `resolved`."
+}
+
+locals {
+  wordlist        = compact(split("\n", file("${path.module}/wordlist.txt")))
+  explicit_names  = [for n in var.names : n if n != null]
+  available_words = tolist(setsubtract(toset(local.wordlist), toset(local.explicit_names)))
+}
+
+resource "random_shuffle" "names" {
+  input        = local.available_words
+  result_count = length(local.available_words)
+  keepers = {
+    cluster_name = var.cluster_name
+    name_seed    = var.name_seed
+  }
+}
+
+output "resolved" {
+  description = "The resolved name per input slot (explicit where given, else auto-picked), in order."
+  value       = [for i, n in var.names : n != null ? n : random_shuffle.names.result[i]]
+}

tf/shared/modules/talos-baremetal/main.tf

@@ -11,17 +11,35 @@
 # kernel `ip=` cmdline. Nodes are dialed directly at their maintenance IP,
 # which is also pinned as the post-install static IP on the bond.
 
+# Per-node names from the shared inventory (operator override per node, else auto).
+module "names" {
+  source       = "../node-names"
+  cluster_name = var.cluster_name
+  name_seed    = var.name_seed
+  names        = [for n in var.nodes : n.name]
+}
+
 locals {
   cluster_endpoint = coalesce(var.cluster_endpoint, "https://${var.cluster_vip}:6443")
 
   # Prefix length from the subnet CIDR (e.g. "24" from "10.10.10.0/24").
   netmask = split("/", var.subnet_cidr)[1]
 
-  # Stable per-name maps (TF map iteration is alphabetical by key, so plan
+  # Each node's hostname per the fleet convention:
+  #   <product>-<provider>-<region>-<clustername>-<role>-<nodename>
+  #   e.g. yucca-int-aus-luke-k8s-<word>. <nodename> = nodes[i].name or auto-picked.
+  nodes_named = [
+    for i, n in var.nodes : merge(n, {
+      role     = coalesce(n.role, "control-plane")
+      hostname = "${var.product}-${var.provider_code}-${var.region_code}-${var.cluster_name}-${var.role_in_hostname}-${module.names.resolved[i]}"
+    })
+  ]
+
+  # Stable per-hostname maps (TF map iteration is alphabetical by key, so plan
   # ordering is deterministic across runs).
-  node_map        = { for n in var.nodes : n.name => n }
-  cp_node_map     = { for k, n in local.node_map : k => n if coalesce(n.role, "control-plane") == "control-plane" }
-  worker_node_map = { for k, n in local.node_map : k => n if coalesce(n.role, "control-plane") == "worker" }
+  node_map        = { for n in local.nodes_named : n.hostname => n }
+  cp_node_map     = { for k, n in local.node_map : k => n if n.role == "control-plane" }
+  worker_node_map = { for k, n in local.node_map : k => n if n.role == "worker" }
 
   cp_addresses = [for k, n in local.cp_node_map : n.address]
 
@@ -158,7 +176,7 @@ locals {
         apiVersion = "v1alpha1"
         kind       = "HostnameConfig"
         auto       = "off"
-        hostname   = n.name
+        hostname   = n.hostname
       }),
     ]
   }

tf/shared/modules/talos-baremetal/variables.tf

@@ -1,8 +1,39 @@
 variable "cluster_name" {
-  description = "Short cluster identifier (e.g., yucca-staging). Drives the Talos cluster name + node hostname prefix is supplied per-node in `nodes`."
+  description = "Themed cluster name (Star Wars for k8s, e.g. 'luke'/'father'). Used as the Talos cluster name + the <clustername> segment of each node hostname."
   type        = string
 }
 
+# Node hostname convention: <product>-<provider>-<region>-<clustername>-<role>-<nodename>
+#   e.g. yucca-int-aus-luke-k8s-<random>. provider/region are the 3-letter codes
+#   from the region's region.hcl; <nodename> auto-derives from the shared inventory.
+variable "product" {
+  description = "Product segment of node hostnames."
+  type        = string
+  default     = "yucca"
+}
+
+variable "provider_code" {
+  description = "3-letter provider segment of node hostnames (region.hcl provider_code, e.g. 'int' / 'htz')."
+  type        = string
+}
+
+variable "region_code" {
+  description = "3-letter region segment of node hostnames (region.hcl region_code, e.g. 'aus' / 'fsn')."
+  type        = string
+}
+
+variable "role_in_hostname" {
+  description = "Workload-role segment of node hostnames ('k8s' for Talos clusters). Distinct from the per-node control-plane/worker role."
+  type        = string
+  default     = "k8s"
+}
+
+variable "name_seed" {
+  description = "Seed for the per-node random name pick (shared node-names inventory). Do NOT change once nodes are deployed (renames every auto-named node)."
+  type        = string
+  default     = "v1"
+}
+
 variable "talos_version" {
   description = "Talos Linux version to install (e.g., '1.13.3'). Sets machine config schema AND the install image tag (so the installed system is exactly this version, regardless of the maintenance-mode boot image)."
   type        = string
@@ -79,7 +110,7 @@ variable "nodes" {
     and Terraform stays reachable across the install reboot).
   EOT
   type = list(object({
-    name    = string
+    name    = optional(string) # explicit hostname override; null = auto-pick from the shared inventory
     role    = optional(string, "control-plane")
     address = string
   }))

tf/shared/modules/talos-baremetal/versions.tf

@@ -5,5 +5,10 @@ terraform {
       source  = "siderolabs/talos"
       version = "~> 0.11"
     }
+    # Used by the node-names child module (random_shuffle) for hostname picks.
+    random = {
+      source  = "hashicorp/random"
+      version = "~> 3.6"
+    }
   }
 }