Merge pull request 49: 建立 DNS source of truth

2026-09-13 15:42:12 +00:00
parent bdacc03e74 e2be2ca045
commit d256682d11
14 changed files with 324 additions and 35 deletions
@@ -6,11 +6,13 @@ on:
    branches: [main]
    paths:
      - 'infrastructure/**/ansible/**'
+      - 'infrastructure/dns/**'
      - '.ansible-lint'
      - '.gitea/workflows/ansible.yml'
  pull_request:
    paths:
      - 'infrastructure/**/ansible/**'
+      - 'infrastructure/dns/**'
      - '.ansible-lint'
      - '.gitea/workflows/ansible.yml'

@@ -48,6 +50,7 @@ jobs:
      - name: ansible-lint
        run: |
          export PATH="$HOME/.local/bin:$PATH"
+          export ANSIBLE_COLLECTIONS_PATH="$PWD/infrastructure/samba-ad/ansible/collections:/root/.ansible/collections"
          rc=0
          for p in infrastructure/openbao infrastructure/samba-ad infrastructure/proxmox; do
            echo "::group::$p"
@@ -55,3 +58,22 @@ jobs:
            echo "::endgroup::"
          done
          exit $rc
+
+  collection-test:
+    runs-on: self-hosted
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Install ansible-core
+        run: |
+          python3 -m pip install --user --break-system-packages \
+            --index-url https://pypi.org/simple --quiet uv==0.11.7
+          export PATH="$HOME/.local/bin:$PATH"
+          uv tool install ansible-core --quiet
+
+      - name: Run ansible-test
+        working-directory: infrastructure/samba-ad/ansible/collections/ansible_collections/ddupan/homelab
+        run: |
+          export PATH="$HOME/.local/bin:$PATH"
+          ansible-test sanity --venv --requirements --python 3.12 --color no
+          ansible-test units --venv --requirements --python 3.12 --color no
@@ -0,0 +1,29 @@
+# DNS 声明与权威边界
+
+`records.yml` 是 homelab DNS 的唯一声明清单，但不是 DNS 服务本身。不同视图仍由最适合
+它们的后端提供：
+
+| 视图 | 权威或递归服务 | 配置方式 |
+|---|---|---|
+| 公网 `ddupan.top` | Cloudflare | Terraform；尚待完整导入已有记录 |
+| AD `ad.ddupan.top` | Samba internal DNS | `samba_dns_record` Ansible module |
+| LAN split horizon | Blocky | 尚待从 inventory 渲染或校验 |
+| Kubernetes Pod split horizon | CoreDNS | 尚待从 inventory 渲染或校验 |
+
+## 安全边界
+
+- Samba module 只管理 `homelab_dns.samba.records` 明确列出的 RRset，不遍历或清理 zone。
+- `_ldap`、`_kerberos`、域控制器 locator 等由 Samba 自动维护的记录不进入 inventory。
+- 一个受管 RRset 默认使用 `exact: true`：同名同类型的额外值会被删除，但其他名称和类型
+  不受影响。
+- DHCP 切换不属于本阶段。Blocky 仍未成为 LAN 客户端的正式 resolver。
+
+## 分阶段接管
+
+1. 用 Samba module 接管现有静态 A RRset，首次 check mode 应为零变更。
+2. 将 Cloudflare 已有 tunnel DNS 记录导入 Terraform state。
+3. 让 Blocky 与 CoreDNS 从 `split_horizon.records` 生成配置或执行 CI 一致性检查。
+4. 验证公网、LAN、Pod、AD 四个视图后，再单独修改 DHCP。
+
+当前 inventory 已明确暴露一个既有差异：`obj.ddupan.top` 在 Blocky 中存在，但 CoreDNS
+尚无对应覆盖。本阶段不会偷偷修复它；后续在两个 resolver 同时接管时统一修复。
@@ -0,0 +1,44 @@
+---
+# Homelab DNS desired state. This file is the canonical inventory; individual
+# backends consume only the views they own.
+homelab_dns:
+  samba:
+    # Samba remains authoritative for the AD zone. Only these explicitly listed
+    # RRsets are reconciled; Samba-generated AD/Kerberos records are untouched.
+    records:
+      - { zone: ad.ddupan.top, name: bao, type: A, values: [192.168.10.8] }
+      - { zone: ad.ddupan.top, name: pve1, type: A, values: [192.168.10.4] }
+      - { zone: ad.ddupan.top, name: pve2, type: A, values: [192.168.10.7] }
+      - { zone: ad.ddupan.top, name: pve3, type: A, values: [192.168.10.9] }
+      - { zone: ad.ddupan.top, name: retrolab, type: A, values: [10.60.0.10] }
+      - { zone: ad.ddupan.top, name: netbox, type: A, values: [192.168.10.127] }
+      - { zone: ad.ddupan.top, name: s3, type: A, values: [192.168.10.127] }
+
+  split_horizon:
+    # LAN and pod resolvers should eventually render the same set from here.
+    # Adoption of Blocky/CoreDNS is deliberately a separate change.
+    records:
+      - { name: git.ddupan.top, type: A, values: [192.168.10.127] }
+      - { name: auth.ddupan.top, type: A, values: [192.168.10.127] }
+      - { name: obj.ddupan.top, type: A, values: [192.168.10.127] }
+
+  public:
+    # Names expected at Cloudflare. Terraform adoption is a separate change;
+    # complete RRsets here make the current ownership gap explicit.
+    records:
+      - name: auth.ddupan.top
+        type: CNAME
+        values: [ff392451-b0b1-45bb-964e-6d9372c3a9e3.cfargotunnel.com]
+        proxied: true
+      - name: git.ddupan.top
+        type: CNAME
+        values: [ff392451-b0b1-45bb-964e-6d9372c3a9e3.cfargotunnel.com]
+        proxied: true
+      - name: obj.ddupan.top
+        type: CNAME
+        values: [ff392451-b0b1-45bb-964e-6d9372c3a9e3.cfargotunnel.com]
+        proxied: true
+      - name: e5renew.ddupan.top
+        type: CNAME
+        values: [ff392451-b0b1-45bb-964e-6d9372c3a9e3.cfargotunnel.com]
+        proxied: true
@@ -54,6 +54,7 @@ second run is a no-op there and only reconciles config drift.
 |---|---|
 | `roles/dc_vm/` | creates the DC VM locally via libvirt + cloud-init |
 | `roles/samba_ad_dc/` | provisions the DC (§1) — idempotent |
+| `ansible/collections/ansible_collections/ddupan/homelab/` | repository-local collection containing the Samba DNS module and its `ansible-test` tests |
 | `roles/samba_ad_dc/tasks/verify.yml` | smoke tests (`--tags verify`) |
 | `roles/samba_ad_dc/tasks/legacy.yml` | opt-in retro-client protocols (§ Retro clients) |
 | `roles/windows_vm/` | unattended-installs the Windows Server 2025 admin box (§2.0) |
@@ -62,6 +63,9 @@ second run is a no-op there and only reconciles config drift.
 | `group_vars/all/vault.yml` | admin passwords (gitignored; encrypt with ansible-vault) |
 | `inventory/hosts.yml` | DC + Windows hosts |

+Static DNS records are declared once in `../../dns/records.yml`. The role manages
+only those RRsets and deliberately leaves Samba-generated AD locator records alone.
+
 ---

 ## 0. Decisions — fill these in before touching anything
@@ -1,6 +1,7 @@
 [defaults]
 inventory = inventory/hosts.yml
 roles_path = roles
+collections_paths = collections
 host_key_checking = False
 callback_result_format = yaml
 nocows = True
@@ -0,0 +1 @@
+tests/output/
@@ -0,0 +1,4 @@
+# ddupan.homelab
+
+Repository-local Ansible collection for homelab infrastructure modules. It is consumed
+directly through the `collections_paths` setting and is not published.
@@ -0,0 +1,16 @@
+---
+namespace: ddupan
+name: homelab
+version: 0.1.0
+readme: README.md
+authors:
+  - panxiao81
+description: Ansible plugins used to manage the ddupan.top homelab.
+license:
+  - GPL-3.0-or-later
+tags:
+  - infrastructure
+  - dns
+repository: https://git.ddupan.top/panxiao81/homelab-infra
+build_ignore:
+  - .git
@@ -0,0 +1,159 @@
+#!/usr/bin/python
+
+# Copyright: (c) 2026, homelab-infra contributors
+# GNU General Public License v3.0+ (see COPYING or https://www.gnu.org/licenses/gpl-3.0.txt)
+
+DOCUMENTATION = r"""
+---
+module: samba_dns_record
+short_description: Reconcile one Samba internal DNS RRset
+author:
+  - Pan Xiao (@panxiao81)
+description:
+  - Reconciles only the named RRset and never deletes an undeclared zone or name.
+  - Runs C(samba-tool dns) on a Samba AD domain controller using its machine account.
+options:
+  server:
+    description: DNS server accepted by C(samba-tool dns).
+    type: str
+    required: true
+  zone:
+    description: Authoritative DNS zone.
+    type: str
+    required: true
+  name:
+    description: Record owner relative to the zone.
+    type: str
+    required: true
+  type:
+    description: DNS record type.
+    type: str
+    choices: [A, AAAA, CNAME, PTR, TXT]
+    required: true
+  values:
+    description: Complete desired value set for this owner and type.
+    type: list
+    elements: str
+    default: []
+  state:
+    description: Whether the desired RRset is present or absent.
+    type: str
+    choices: [present, absent]
+    default: present
+  exact:
+    description: Remove live values not present in C(values).
+    type: bool
+    default: true
+"""
+
+EXAMPLES = r"""
+- name: Reconcile an A RRset
+  samba_dns_record:
+    server: dc1.ad.ddupan.top
+    zone: ad.ddupan.top
+    name: pve1
+    type: A
+    values: [192.168.10.4]
+    exact: true
+"""
+
+RETURN = r"""
+before:
+  description: Values observed before reconciliation.
+  type: list
+  returned: always
+after:
+  description: Values expected after reconciliation.
+  type: list
+  returned: always
+"""
+
+import re
+
+from ansible.module_utils.basic import AnsibleModule
+
+
+ABSENT_ERRORS = (
+    "WERR_DNS_ERROR_NAME_DOES_NOT_EXIST",
+    "WERR_DNS_ERROR_RECORD_DOES_NOT_EXIST",
+    "WERR_DNS_ERROR_NXDOMAIN",
+)
+
+
+def normalize_value(record_type, value):
+    value = value.strip()
+    if record_type in ("CNAME", "PTR"):
+        return value.rstrip(".").lower()
+    if record_type == "TXT" and len(value) >= 2 and value[0] == value[-1] == '"':
+        return value[1:-1]
+    return value.lower() if record_type == "AAAA" else value
+
+
+def parse_query(stdout, record_type):
+    values = []
+    prefix = re.compile(r"^\s*%s:\s*(.*?)\s*(?:\([^)]*\))?\s*$" % re.escape(record_type))
+    for line in stdout.splitlines():
+        match = prefix.match(line)
+        if match:
+            values.append(normalize_value(record_type, match.group(1)))
+    return sorted(set(values))
+
+
+def run(module, args, check_rc=True):
+    command = [module.get_bin_path("samba-tool", required=True), "dns"] + args + ["-P"]
+    return module.run_command(command, check_rc=check_rc)
+
+
+def main():
+    module = AnsibleModule(
+        argument_spec=dict(
+            server=dict(type="str", required=True),
+            zone=dict(type="str", required=True),
+            name=dict(type="str", required=True),
+            type=dict(type="str", required=True, choices=["A", "AAAA", "CNAME", "PTR", "TXT"]),
+            values=dict(type="list", elements="str", default=[]),
+            state=dict(type="str", choices=["present", "absent"], default="present"),
+            exact=dict(type="bool", default=True),
+        ),
+        supports_check_mode=True,
+    )
+
+    p = module.params
+    desired = sorted(set(normalize_value(p["type"], value) for value in p["values"]))
+    if p["state"] == "present" and not desired:
+        module.fail_json(msg="values must not be empty when state=present")
+
+    rc, stdout, stderr = run(
+        module,
+        ["query", p["server"], p["zone"], p["name"], p["type"]],
+        check_rc=False,
+    )
+    if rc == 0:
+        current = parse_query(stdout, p["type"])
+    elif any(marker in stdout + stderr for marker in ABSENT_ERRORS):
+        current = []
+    else:
+        module.fail_json(msg="samba-tool dns query failed", rc=rc, stdout=stdout, stderr=stderr)
+
+    target = desired if p["state"] == "present" else []
+    additions = sorted(set(target) - set(current))
+    removals = sorted(set(current) - set(target)) if p["exact"] or p["state"] == "absent" else []
+    changed = bool(additions or removals)
+
+    if changed and not module.check_mode:
+        for value in removals:
+            run(module, ["delete", p["server"], p["zone"], p["name"], p["type"], value])
+        for value in additions:
+            run(module, ["add", p["server"], p["zone"], p["name"], p["type"], value])
+
+    after = sorted((set(current) - set(removals)) | set(additions))
+    module.exit_json(
+        changed=changed,
+        before=current,
+        after=after,
+        diff={"before": {p["type"]: current}, "after": {p["type"]: after}},
+    )
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,23 @@
+import unittest
+
+from ansible_collections.ddupan.homelab.plugins.modules import samba_dns_record
+
+
+class ParseQueryTests(unittest.TestCase):
+    def test_parses_a_records_and_ignores_other_types(self):
+        output = """  Name=host, Records=3, Children=0
+    A: 192.168.10.4 (flags=f0, serial=1, ttl=900)
+    A: 192.168.10.7 (flags=f0, serial=2, ttl=900)
+    TXT: ignored (flags=f0, serial=3, ttl=900)
+"""
+        self.assertEqual(
+            samba_dns_record.parse_query(output, "A"),
+            ["192.168.10.4", "192.168.10.7"],
+        )
+
+    def test_normalizes_dns_targets(self):
+        output = "    CNAME: Gateway.AD.DDUPAN.TOP. (flags=f0, serial=1, ttl=900)\n"
+        self.assertEqual(
+            samba_dns_record.parse_query(output, "CNAME"),
+            ["gateway.ad.ddupan.top"],
+        )
@@ -10,29 +10,8 @@ samba_ad_dc_ip: "192.168.10.5"
 samba_ad_dns_forwarder: "192.168.10.1"
 samba_ad_reverse_zone: "10.168.192.in-addr.arpa"   # reverse of 192.168.10.0/24

-# Extra A records for non-domain hosts published in the AD DNS zone.
-samba_ad_extra_a_records:
-  - { name: "bao", ip: "192.168.10.8" }            # OpenBao (../openbao), not domain-joined
-  # Proxmox cluster nodes (../proxmox). Not domain-joined; they authenticate
-  # USERS against this DC rather than being members themselves.
-  - { name: "pve1", ip: "192.168.10.4" }
-  - { name: "pve2", ip: "192.168.10.7" }
-  - { name: "pve3", ip: "192.168.10.9" }
-  # Lab VMs on the SDN VNets (routed via the VyOS router, see ../../proxmox).
-  # These are NOT on 192.168.10.0/24, so they have no PTR in the existing
-  # reverse zone — forward resolution only unless a 0.60.10.in-addr.arpa zone
-  # is added later.
-  - { name: "retrolab", ip: "10.60.0.10" }
-  # k3s services exposed on the LAN through the Envoy gateway (../../../platform/envoy-gateway;
-  # Contour was retired 2026-07-25). They all point at the k3s node, which is where
-  # Envoy's LoadBalancer lands; the gateway routes by Host header and serves the
-  # *.ad.ddupan.top wildcard cert.
-  # Adding another such service = one more line here + an HTTPRoute, nothing else.
-  - { name: "netbox", ip: "192.168.10.127" }        # NetBox (../../../apps/netbox)
-  # SeaweedFS S3. Exists so Terraform state does NOT ride the Cloudflare tunnel:
-  # obj.ddupan.top works, but it hairpins through the WAN, and on 2026-07-28 that
-  # path was blackholed for hours by a dead VPN tunnel. State must stay on the LAN.
-  - { name: "s3", ip: "192.168.10.127" }            # SeaweedFS S3 (../../../apps/seaweedfs)
+# Static records now live in ../../dns/records.yml and are reconciled as complete
+# RRsets by the local samba_dns_record module.

 # Support legacy clients (Win9x/NT4/2000/XP)? INSECURE — see README "Retro clients".
 samba_ad_legacy_clients: false
@@ -6,6 +6,8 @@
  hosts: samba_dc
  become: true
  gather_facts: true
+  vars_files:
+    - ../../dns/records.yml
  roles:
    - role: samba_ad_dc
  post_tasks:
@@ -0,0 +1,14 @@
+---
+- name: Reconcile explicitly managed AD DNS RRsets
+  ddupan.homelab.samba_dns_record:
+    server: "{{ samba_ad_dc_ip }}"
+    zone: "{{ item['zone'] }}"
+    name: "{{ item['name'] }}"
+    type: "{{ item['type'] }}"
+    values: "{{ item['values'] }}"
+    state: present
+    exact: true
+  loop: "{{ homelab_dns.samba.records }}"
+  loop_control:
+    label: "{{ item['name'] }}.{{ item['zone'] }} {{ item['type'] }}"
+  tags: [dns]
@@ -198,20 +198,11 @@
    - "'already exists' not in (kms_srv.stderr | default('')) + (kms_srv.stdout | default(''))"
  no_log: true

-# --- Extra A records for non-domain hosts (e.g. OpenBao) -----------------------
+# --- Directory objects and explicitly managed DNS records ----------------------
 - name: Service accounts and RBAC groups
  ansible.builtin.import_tasks: directory_objects.yml
  tags: [directory, accounts]

- name: Register extra A records in the AD DNS zone
-  ansible.builtin.command:
-    cmd: >-
-      samba-tool dns add {{ samba_ad_dc_ip }} {{ samba_ad_realm | lower }}
-      {{ item.name }} A {{ item.ip }} -P
-  loop: "{{ samba_ad_extra_a_records }}"
-  register: extra_a
-  changed_when: "'Record added successfully' in (extra_a.stdout | default(''))"
-  failed_when:
-    - extra_a.rc != 0
-    - "'already exists' not in (extra_a.stderr | default('')) + (extra_a.stdout | default(''))"
+- name: Reconcile static AD DNS records
+  ansible.builtin.import_tasks: dns_records.yml
  tags: [dns]