Validate homelab inventory with YAML parser

jeannie

@@ -73,11 +73,26 @@ mapping = {
     "pimox.default_worker_count": "LAB_PIMOX_WORKER_COUNT",
 }
 
+def flatten(document, prefix=""):
+    values = {}
+    if isinstance(document, dict):
+        for key, value in document.items():
+            path = f"{prefix}.{key}" if prefix else str(key)
+            if isinstance(value, dict):
+                values.update(flatten(value, path))
+            elif isinstance(value, list):
+                values[path] = ",".join(str(item) for item in value)
+            elif value is not None:
+                values[path] = str(value)
+    return values
+
+
+def parse_simple_inventory(path):
     stack = []
     values = {}
     pattern = re.compile(r"^(\s*)([A-Za-z_][A-Za-z0-9_]*):(?:\s*(.*?))?\s*$")
 
-with open(inventory_file, encoding="utf-8") as handle:
+    with open(path, encoding="utf-8") as handle:
         for raw_line in handle:
             if not raw_line.strip() or raw_line.lstrip().startswith("#") or raw_line.lstrip().startswith("- "):
                 continue
@@ -89,14 +104,23 @@ with open(inventory_file, encoding="utf-8") as handle:
             value = (match.group(3) or "").strip()
             while stack and stack[-1][0] >= indent:
                 stack.pop()
-        path = ".".join([item[1] for item in stack] + [key])
+            current_path = ".".join([item[1] for item in stack] + [key])
             if value == "":
                 stack.append((indent, key))
                 continue
             if " #" in value:
                 value = value.split(" #", 1)[0].strip()
-        value = value.strip("\"'")
+            values[current_path] = value.strip("\"'")
-        values[path] = value
+    return values
+
+
+try:
+    import yaml
+except ImportError:
+    values = parse_simple_inventory(inventory_file)
+else:
+    with open(inventory_file, encoding="utf-8") as handle:
+        values = flatten(yaml.safe_load(handle) or {})
 
 for path, env_name in mapping.items():
     value = values.get(path)
@@ -1302,6 +1326,10 @@ homelab_preflight() {
     echo "${phase^} preflight checks passed."
 }
 
+inventory_check() {
+    "${REPO_ROOT}/scripts/validate-homelab-inventory" "${HOMELAB_INVENTORY_FILE:-${REPO_ROOT}/homelab.yml}"
+}
+
 run_pimox_pipeline() {
     local mode="${LAB_PIMOX_PIPELINE:-true}"
     local pimox_host="${LAB_PIMOX_HOST:-${TF_VAR_pimox_host:-192.168.100.80}}"
@@ -5199,6 +5227,9 @@ case "${1:-}" in
     preflight)
         homelab_preflight
         ;;
+    inventory-check)
+        inventory_check
+        ;;
     fix-debian-docker-root)
         fix_debian_docker_root
         ;;
@@ -5224,7 +5255,7 @@ case "${1:-}" in
         nuke
         ;;
     *)
-        echo "Usage: $0 {up|plan [all|provisioning|cluster|platform|apps|edge]|rebuild-cluster|stop-cluster|start-cluster|status|apps|website-translation-model|website-ollama-listen|ollama-setup|deploy-gitea|rpi-services|bootstrap-gitea-repo|backup-gitea|drill-gitea-restore|install-gitea-runner|move-prometheus-stack-workers|doctor-versions|doctor-edge|doctor-gitea|doctor-rpi|doctor-cluster|preflight|fix-debian-docker-root|secrets-init|secrets-check|tailnet-policy-check|ai-index|ai-check|openwrt|nuke}"
+        echo "Usage: $0 {up|plan [all|provisioning|cluster|platform|apps|edge]|rebuild-cluster|stop-cluster|start-cluster|status|apps|website-translation-model|website-ollama-listen|ollama-setup|deploy-gitea|rpi-services|bootstrap-gitea-repo|backup-gitea|drill-gitea-restore|install-gitea-runner|move-prometheus-stack-workers|doctor-versions|doctor-edge|doctor-gitea|doctor-rpi|doctor-cluster|preflight|inventory-check|fix-debian-docker-root|secrets-init|secrets-check|tailnet-policy-check|ai-index|ai-check|openwrt|nuke}"
         exit 1
         ;;
 esac

scripts/render-docs

@@ -99,6 +99,8 @@ main() {
         return 1
     fi
 
+    "${REPO_ROOT}/scripts/validate-homelab-inventory" "${INVENTORY_FILE}"
+
     main_script="$(read_inventory_value metadata.main_script)"
     if [[ -z "${main_script}" ]]; then
         printf 'homelab.yml metadata.main_script must not be empty\n' >&2

scripts/validate-homelab-inventory

@@ -0,0 +1,228 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+REPO_ROOT="$(cd "${SCRIPT_DIR}/.." && pwd)"
+INVENTORY_FILE="${1:-${REPO_ROOT}/homelab.yml}"
+
+if [[ ! -s "${INVENTORY_FILE}" ]]; then
+    printf 'Missing homelab inventory: %s\n' "${INVENTORY_FILE}" >&2
+    exit 1
+fi
+
+if command -v python3 >/dev/null 2>&1 &&
+    python3 - "${INVENTORY_FILE}" <<'PY'
+import ipaddress
+import sys
+
+try:
+    import yaml
+except ImportError:
+    raise SystemExit(42)
+
+inventory_file = sys.argv[1]
+
+with open(inventory_file, encoding="utf-8") as handle:
+    document = yaml.safe_load(handle)
+
+failures = []
+
+
+def get(path):
+    value = document
+    for part in path.split("."):
+        if not isinstance(value, dict) or part not in value:
+            return None
+        value = value[part]
+    return value
+
+
+def require(path, expected_type=str):
+    value = get(path)
+    if value is None or value == "":
+        failures.append(f"{path} is required")
+        return None
+    if expected_type is not None and not isinstance(value, expected_type):
+        failures.append(f"{path} must be {expected_type.__name__}, got {type(value).__name__}")
+        return None
+    return value
+
+
+if not isinstance(document, dict):
+    failures.append("inventory root must be a YAML mapping")
+else:
+    for path in (
+        "metadata.main_script",
+        "domain.base",
+        "domain.public_url",
+        "domain.gitea_url",
+        "network.lan_cidr",
+        "network.lan_ip_prefix",
+        "network.metallb.traefik_ip",
+        "hosts.debian.lan_ip",
+        "hosts.debian.tailscale_ip",
+        "hosts.debian.docker_root",
+        "hosts.rpi4.lan_ip",
+        "hosts.rpi4.tailscale_ip",
+        "hosts.opi5_pimox.lan_ip",
+        "hosts.opi5_pimox.worker_storage",
+        "hosts.oci_edge.public_ip",
+        "services.gitea.root_url",
+        "services.local_registry.endpoint",
+        "services.traefik.load_balancer_ip",
+    ):
+        require(path)
+    require("pimox.default_worker_count", None)
+
+    subdomains = require("domain.subdomains", list)
+    if isinstance(subdomains, list):
+        for index, value in enumerate(subdomains):
+            if not isinstance(value, str) or not value:
+                failures.append(f"domain.subdomains[{index}] must be a non-empty string")
+
+    for path in (
+        "network.lan_cidr",
+        "network.metallb.traefik_ip",
+        "hosts.debian.lan_ip",
+        "hosts.debian.tailscale_ip",
+        "hosts.rpi4.lan_ip",
+        "hosts.rpi4.tailscale_ip",
+        "hosts.opi5_pimox.lan_ip",
+        "hosts.oci_edge.public_ip",
+        "services.traefik.load_balancer_ip",
+    ):
+        value = get(path)
+        if not value:
+            continue
+        try:
+            if path.endswith("lan_cidr"):
+                ipaddress.ip_network(str(value), strict=False)
+            else:
+                ipaddress.ip_address(str(value))
+        except ValueError:
+            failures.append(f"{path} is not a valid IP address/CIDR: {value}")
+
+    worker_count = get("pimox.default_worker_count")
+    if worker_count is not None:
+        try:
+            if int(worker_count) < 0:
+                failures.append("pimox.default_worker_count must be non-negative")
+        except (TypeError, ValueError):
+            failures.append("pimox.default_worker_count must be an integer")
+
+if failures:
+    for failure in failures:
+        print(f"inventory error: {failure}", file=sys.stderr)
+    raise SystemExit(1)
+
+print(f"inventory ok: {inventory_file}")
+PY
+then
+    exit 0
+else
+    status=$?
+    if [[ "${status}" != "42" ]]; then
+        exit "${status}"
+    fi
+fi
+
+if command -v ruby >/dev/null 2>&1; then
+    ruby - "${INVENTORY_FILE}" <<'RUBY'
+require "ipaddr"
+require "yaml"
+
+inventory_file = ARGV.fetch(0)
+document = YAML.load_file(inventory_file)
+failures = []
+
+def get(document, path)
+  path.split(".").reduce(document) do |value, part|
+    return nil unless value.is_a?(Hash) && value.key?(part)
+    value[part]
+  end
+end
+
+required_paths = %w[
+  metadata.main_script
+  domain.base
+  domain.public_url
+  domain.gitea_url
+  network.lan_cidr
+  network.lan_ip_prefix
+  network.metallb.traefik_ip
+  hosts.debian.lan_ip
+  hosts.debian.tailscale_ip
+  hosts.debian.docker_root
+  hosts.rpi4.lan_ip
+  hosts.rpi4.tailscale_ip
+  hosts.opi5_pimox.lan_ip
+  hosts.opi5_pimox.worker_storage
+  hosts.oci_edge.public_ip
+  services.gitea.root_url
+  services.local_registry.endpoint
+  services.traefik.load_balancer_ip
+  pimox.default_worker_count
+]
+
+unless document.is_a?(Hash)
+  failures << "inventory root must be a YAML mapping"
+else
+  required_paths.each do |path|
+    value = get(document, path)
+    failures << "#{path} is required" if value.nil? || value == ""
+  end
+
+  subdomains = get(document, "domain.subdomains")
+  if !subdomains.is_a?(Array)
+    failures << "domain.subdomains must be Array"
+  else
+    subdomains.each_with_index do |value, index|
+      failures << "domain.subdomains[#{index}] must be a non-empty string" unless value.is_a?(String) && !value.empty?
+    end
+  end
+
+  %w[
+    network.metallb.traefik_ip
+    hosts.debian.lan_ip
+    hosts.debian.tailscale_ip
+    hosts.rpi4.lan_ip
+    hosts.rpi4.tailscale_ip
+    hosts.opi5_pimox.lan_ip
+    hosts.oci_edge.public_ip
+    services.traefik.load_balancer_ip
+  ].each do |path|
+    value = get(document, path)
+    next if value.nil? || value == ""
+    begin
+      IPAddr.new(value.to_s)
+    rescue ArgumentError
+      failures << "#{path} is not a valid IP address: #{value}"
+    end
+  end
+
+  cidr = get(document, "network.lan_cidr")
+  begin
+    IPAddr.new(cidr.to_s) if cidr
+  rescue ArgumentError
+    failures << "network.lan_cidr is not a valid CIDR: #{cidr}"
+  end
+
+  worker_count = get(document, "pimox.default_worker_count")
+  failures << "pimox.default_worker_count must be non-negative" if worker_count && worker_count.to_i < 0
+end
+
+if failures.any?
+  failures.each { |failure| warn "inventory error: #{failure}" }
+  exit 1
+end
+
+puts "inventory ok: #{inventory_file}"
+RUBY
+    exit 0
+fi
+
+cat >&2 <<'EOF'
+No real YAML parser is available.
+Install python3-yaml or ruby, then rerun scripts/validate-homelab-inventory.
+EOF
+exit 1