Make the k3s API firewall allowlist declarative

The live host restricted 6443/tcp to specific operator addresses, added by hand,
while this role still declared the port open to Anywhere with no source
restriction. The declared config was weaker than reality: running the base role
would have REMOVED the restriction and exposed the Kubernetes API to the
internet. Security was tightened on the host and never fed back into the source
of truth.

Found 2026-08-11 while diagnosing lost cluster access, which turned out to be an
ISP lease rotation (89.244.90.246 -> .236) against a hand-maintained allowlist.

Changes:
- defaults: k3s_api_allowed_sources (empty = 6443 closed to all external
  sources, the safe failure; SSH unaffected so the host stays recoverable) and
  k3s_api_revoked_sources, so rotated addresses are pruned rather than left as
  standing grants to whoever the ISP reassigns them to
- tasks: grant approved sources, then remove any blanket rule, then revoke
  retired ones. Order matters - grants are added before the blanket rule is
  deleted so convergence never opens a window with no API access
- group_vars/all.yaml: the current operator address, plus the two stale grants
  (.246 rotated, 85.132.220.102 historic) marked for revocation
- docs/verification.md: state that 6443 is source-restricted rather than
  listing it as a plainly allowed port

Not yet converged against the live host - the role change is committed but
running it is a production action needing operator approval.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

ansible/inventory/group_vars/all.yaml

@@ -24,3 +24,29 @@ custodian_agent_pubkey: "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIC/V9fe5MGKdhTBz9Kw
 # Private key lives at ~/.ssh/id_ops on the workstation. Never commit the private key.
 ops_bridge_user: tegwick
 ops_bridge_pubkey: "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIKQmXbAVlEa8dzGx8Hk2S7AITpz6sMWdCN0MeMOzL82u ops-bridge@custodian"
+
+# ---------------------------------------------------------------------------
+# k3s API firewall allowlist (6443/tcp)
+#
+# The k3s API is NOT world-reachable. Only these sources may connect.
+# Added 2026-08-11 after discovering the live host was source-restricted by hand
+# while this repo still declared 6443 open to Anywhere — meaning a convergence
+# run would have exposed the Kubernetes API to the internet.
+#
+# These are operator egress addresses, not secrets. They are dynamic: when your
+# ISP rotates the lease, MOVE the old entry to k3s_api_revoked_sources rather
+# than deleting the line, so convergence prunes the stale grant instead of
+# leaving it standing for whoever the address gets reassigned to.
+#
+# Durable alternative worth considering: reach the API over the ops-bridge SSH
+# tunnel instead (see railiance-infra/docs/deploy-stack.md, the
+# k3s-api-coulombcore pattern) and allow no public sources at all.
+k3s_api_allowed_sources:
+  - address: "89.244.90.236"
+    comment: "k3s-api-operator-current"
+
+k3s_api_revoked_sources:
+  - address: "89.244.90.246"
+    comment: "rotated ISP lease, superseded 2026-08-11"
+  - address: "85.132.220.102"
+    comment: "historic operator address, retired"

ansible/roles/base/defaults/main.yml

@@ -0,0 +1,30 @@
+---
+# Base role defaults.
+
+# Source addresses permitted to reach the k3s API (6443/tcp).
+#
+# WHY THIS EXISTS
+# ---------------
+# The k3s API was originally opened to Anywhere by this role, and the source
+# restriction was later added by hand on the live host. That left the declared
+# config WEAKER than reality: re-running this role would have removed the
+# restriction and exposed the Kubernetes API to the internet. Found 2026-08-11.
+#
+# The allowlist is therefore declared here and converged, not hand-edited.
+#
+# Deliberately empty by default. A host that sets no sources gets NO public
+# access to 6443 — which is the safe failure. SSH (22) is unaffected, so a host
+# converged with an empty list is always recoverable.
+#
+# Set the real values in inventory/group_vars/all.yaml. Each entry:
+#   - address: "203.0.113.10"
+#     comment: "k3s-api-operator-workstation"
+k3s_api_allowed_sources: []
+
+# Source addresses whose k3s API access must be REMOVED on convergence.
+#
+# Operator addresses rotate (dynamic ISP leases). Without this, every rotation
+# leaves a standing grant to an address the ISP has since reassigned to someone
+# else. Move an address here when it stops being yours; convergence then prunes
+# it rather than leaving it to accumulate.
+k3s_api_revoked_sources: []

ansible/roles/base/tasks/main.yml

@@ -62,11 +62,48 @@
     rule: allow
     name: OpenSSH
 
-- name: Allow k3s API in UFW
+# k3s API access is source-restricted. See roles/base/defaults/main.yml for why
+# this is declared rather than hand-applied. Order matters below: grants are
+# added BEFORE the blanket rule is removed, so convergence never opens a window
+# in which the operator cannot reach the API.
+
+- name: Allow k3s API from approved operator sources only
   ansible.builtin.ufw:
     rule: allow
     port: '6443'
     proto: tcp
+    from_ip: "{{ item.address }}"
+    comment: "{{ item.comment | default('k3s-api-operator') }}"
+  loop: "{{ k3s_api_allowed_sources }}"
+  loop_control:
+    label: "{{ item.address }}"
+
+- name: Remove blanket k3s API rule if present (must not be world-reachable)
+  ansible.builtin.ufw:
+    rule: allow
+    port: '6443'
+    proto: tcp
+    delete: true
+
+- name: Revoke k3s API access for retired operator sources
+  ansible.builtin.ufw:
+    rule: allow
+    port: '6443'
+    proto: tcp
+    from_ip: "{{ item.address }}"
+    delete: true
+  loop: "{{ k3s_api_revoked_sources }}"
+  loop_control:
+    label: "{{ item.address }}"
+
+- name: Warn when no operator source is allowed to reach the k3s API
+  ansible.builtin.debug:
+    msg: >-
+      k3s_api_allowed_sources is empty, so 6443/tcp is closed to all external
+      sources on this host. This is the safe default, not an error. SSH is
+      unaffected and the host remains recoverable. Set the allowlist in
+      inventory/group_vars/all.yaml to restore API access.
+  when: k3s_api_allowed_sources | length == 0
 
 - name: Allow Flannel VXLAN in UFW
   ansible.builtin.ufw: