titan-iac/services/keycloak/oneoffs/ldap-federation-job.yaml

# services/keycloak/oneoffs/ldap-federation-job.yaml
# One-off job for sso/keycloak-ldap-federation-12.
# Purpose: keycloak ldap federation 12 (see container args/env in this file).
# Run by setting spec.suspend to false, reconcile, then set it back to true.
# Safe to delete the finished Job/pod; it should not run continuously.
apiVersion: batch/v1
kind: Job
metadata:
  name: keycloak-ldap-federation-12
  namespace: sso
spec:
  suspend: true
  backoffLimit: 2
  template:
    metadata:
      annotations:
        vault.hashicorp.com/agent-inject: "true"
        vault.hashicorp.com/agent-pre-populate-only: "true"
        vault.hashicorp.com/role: "sso"
        vault.hashicorp.com/agent-inject-secret-keycloak-env.sh: "kv/data/atlas/shared/keycloak-admin"
        vault.hashicorp.com/agent-inject-template-keycloak-env.sh: |
          {{ with secret "kv/data/atlas/shared/keycloak-admin" }}
          export KEYCLOAK_ADMIN="{{ .Data.data.username }}"
          export KEYCLOAK_ADMIN_USER="{{ .Data.data.username }}"
          export KEYCLOAK_ADMIN_PASSWORD="{{ .Data.data.password }}"
          {{ end }}
          {{ with secret "kv/data/atlas/sso/keycloak-db" }}
          export KC_DB_URL_DATABASE="{{ .Data.data.POSTGRES_DATABASE }}"
          export KC_DB_USERNAME="{{ .Data.data.POSTGRES_USER }}"
          export KC_DB_PASSWORD="{{ .Data.data.POSTGRES_PASSWORD }}"
          {{ end }}
          {{ with secret "kv/data/atlas/shared/portal-e2e-client" }}
          export PORTAL_E2E_CLIENT_ID="{{ .Data.data.client_id }}"
          export PORTAL_E2E_CLIENT_SECRET="{{ .Data.data.client_secret }}"
          {{ end }}
          {{ with secret "kv/data/atlas/sso/openldap-admin" }}
          export LDAP_ADMIN_PASSWORD="{{ .Data.data.LDAP_ADMIN_PASSWORD }}"
          export LDAP_CONFIG_PASSWORD="{{ .Data.data.LDAP_CONFIG_PASSWORD }}"
          export LDAP_BIND_PASSWORD="${LDAP_ADMIN_PASSWORD}"
          {{ end }}
          {{ with secret "kv/data/atlas/shared/postmark-relay" }}
          export KEYCLOAK_SMTP_USER="{{ index .Data.data "apikey" }}"
          export KEYCLOAK_SMTP_PASSWORD="{{ index .Data.data "apikey" }}"
          {{ end }}
    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: hardware
                    operator: In
                    values: ["rpi5","rpi4"]
                  - key: node-role.kubernetes.io/worker
                    operator: Exists
      restartPolicy: OnFailure
      serviceAccountName: sso-vault
      containers:
        - name: configure
          image: python:3.11-alpine
          imagePullPolicy: IfNotPresent
          env:
            - name: KEYCLOAK_SERVER
              value: http://keycloak.sso.svc.cluster.local
            - name: KEYCLOAK_REALM
              value: atlas
            - name: LDAP_URL
              value: ldap://openldap.sso.svc.cluster.local:389
            - name: LDAP_BIND_DN
              value: cn=admin,dc=bstein,dc=dev
            - name: LDAP_USERS_DN
              value: ou=users,dc=bstein,dc=dev
            - name: LDAP_GROUPS_DN
              value: ou=groups,dc=bstein,dc=dev
          command: ["/bin/sh", "-c"]
          args:
            - |
              set -euo pipefail
              . /vault/secrets/keycloak-env.sh
              python - <<'PY'
              import json
              import os
              import time
              import urllib.parse
              import urllib.error
              import urllib.request

              base_url = os.environ["KEYCLOAK_SERVER"].rstrip("/")
              realm = os.environ["KEYCLOAK_REALM"]
              admin_user = os.environ["KEYCLOAK_ADMIN_USER"]
              admin_password = os.environ["KEYCLOAK_ADMIN_PASSWORD"]

              ldap_url = os.environ["LDAP_URL"]
              ldap_bind_dn = os.environ["LDAP_BIND_DN"]
              ldap_bind_password = os.environ["LDAP_BIND_PASSWORD"]
              ldap_users_dn = os.environ["LDAP_USERS_DN"]
              ldap_groups_dn = os.environ["LDAP_GROUPS_DN"]

              def http_json(method: str, url: str, token: str, payload=None):
                  data = None
                  headers = {"Authorization": f"Bearer {token}"}
                  if payload is not None:
                      data = json.dumps(payload).encode()
                      headers["Content-Type"] = "application/json"
                  req = urllib.request.Request(url, data=data, headers=headers, method=method)
                  with urllib.request.urlopen(req, timeout=30) as resp:
                      body = resp.read()
                      if not body:
                          return resp.status, None, dict(resp.headers)
                      return resp.status, json.loads(body.decode()), dict(resp.headers)

              def get_token():
                  token_data = urllib.parse.urlencode(
                      {
                          "grant_type": "password",
                          "client_id": "admin-cli",
                          "username": admin_user,
                          "password": admin_password,
                      }
                  ).encode()
                  token_req = urllib.request.Request(
                      f"{base_url}/realms/master/protocol/openid-connect/token",
                      data=token_data,
                      headers={"Content-Type": "application/x-www-form-urlencoded"},
                      method="POST",
                  )
                  with urllib.request.urlopen(token_req, timeout=30) as resp:
                      token_body = json.loads(resp.read().decode())
                  return token_body["access_token"]

              def wait_for_keycloak():
                  for _ in range(60):
                      try:
                          token = get_token()
                          if token:
                              return token
                      except Exception:
                          time.sleep(2)
                  raise SystemExit("Keycloak not ready")

              token = wait_for_keycloak()

              # Keycloak component "parentId" must be the realm UUID, not the realm name.
              status, realm_rep, _ = http_json(
                  "GET",
                  f"{base_url}/admin/realms/{realm}",
                  token,
              )
              if status != 200 or not realm_rep or not realm_rep.get("id"):
                  raise SystemExit(f"Unable to resolve realm id for {realm} (status={status})")
              realm_id = realm_rep["id"]

              # Some historical LDAP federation components were created with parentId=<realm name>.
              # That makes realm resolution null in Keycloak internals and breaks authentication.
              status, all_components, _ = http_json(
                  "GET",
                  f"{base_url}/admin/realms/{realm}/components",
                  token,
              )
              if status != 200:
                  raise SystemExit(f"Unexpected components response: {status}")
              all_components = all_components or []

              for c in all_components:
                  if c.get("providerId") != "ldap":
                      continue
                  if c.get("providerType") != "org.keycloak.storage.UserStorageProvider":
                      continue
                  if c.get("parentId") == realm_id:
                      continue
                  cid = c.get("id")
                  if not cid:
                      continue
                  print(f"Fixing LDAP federation parentId for {cid} (was {c.get('parentId')})")
                  status, comp, _ = http_json(
                      "GET",
                      f"{base_url}/admin/realms/{realm}/components/{cid}",
                      token,
                  )
                  if status != 200 or not comp:
                      raise SystemExit(f"Unable to fetch component {cid} (status={status})")
                  comp["parentId"] = realm_id
                  status, _, _ = http_json(
                      "PUT",
                      f"{base_url}/admin/realms/{realm}/components/{cid}",
                      token,
                      comp,
                  )
                  if status not in (200, 204):
                      raise SystemExit(f"Unexpected parentId repair status for {cid}: {status}")

              # Find existing LDAP user federation provider (if any)
              status, components, _ = http_json(
                  "GET",
                  f"{base_url}/admin/realms/{realm}/components?type=org.keycloak.storage.UserStorageProvider",
                  token,
              )
              if status != 200:
                  raise SystemExit(f"Unexpected components response: {status}")
              components = components or []

              ldap_components = [c for c in components if c.get("providerId") == "ldap" and c.get("id")]

              # Select a canonical LDAP federation provider deterministically.
              # Duplicate LDAP providers can cause Keycloak admin/user queries to fail if any one of them is misconfigured.
              candidates = []
              for c in ldap_components:
                  if c.get("name") not in ("openldap", "ldap"):
                      continue
                  cfg = c.get("config") or {}
                  if (cfg.get("connectionUrl") or [None])[0] == ldap_url:
                      candidates.append(c)
              if not candidates:
                  candidates = [c for c in ldap_components if c.get("name") in ("openldap", "ldap")]
              candidates.sort(key=lambda x: x.get("id", ""))
              ldap_component = candidates[0] if candidates else None
              ldap_component_id = ldap_component["id"] if ldap_component else None

              desired = {
                  "name": "openldap",
                  "providerId": "ldap",
                  "providerType": "org.keycloak.storage.UserStorageProvider",
                  "parentId": realm_id,
                  "config": {
                      "enabled": ["true"],
                      "priority": ["0"],
                      "importEnabled": ["true"],
                      "editMode": ["WRITABLE"],
                      "syncRegistrations": ["true"],
                      "vendor": ["other"],
                      "connectionUrl": [ldap_url],
                      "bindDn": [ldap_bind_dn],
                      "bindCredential": [ldap_bind_password],
                      "authType": ["simple"],
                      "usersDn": [ldap_users_dn],
                      "searchScope": ["1"],
                      "pagination": ["true"],
                      "usernameLDAPAttribute": ["uid"],
                      "rdnLDAPAttribute": ["uid"],
                      "uuidLDAPAttribute": ["entryUUID"],
                      "userObjectClasses": ["inetOrgPerson, organizationalPerson, person, top"],
                      "trustEmail": ["true"],
                      "useTruststoreSpi": ["never"],
                      "connectionPooling": ["true"],
                      "cachePolicy": ["DEFAULT"],
                      "useKerberosForPasswordAuthentication": ["false"],
                      "allowKerberosAuthentication": ["false"],
                  },
              }

              if ldap_component:
                  desired["id"] = ldap_component["id"]
                  print(f"Updating LDAP federation provider: {desired['id']}")
                  status, _, _ = http_json(
                      "PUT",
                      f"{base_url}/admin/realms/{realm}/components/{desired['id']}",
                      token,
                      desired,
                  )
                  if status not in (200, 204):
                      raise SystemExit(f"Unexpected update status: {status}")
              else:
                  print("Creating LDAP federation provider")
                  status, _, headers = http_json(
                      "POST",
                      f"{base_url}/admin/realms/{realm}/components",
                      token,
                      desired,
                  )
                  if status not in (201, 204):
                      raise SystemExit(f"Unexpected create status: {status}")
                  location = headers.get("Location", "")
                  if location:
                      ldap_component_id = location.rstrip("/").split("/")[-1]

              # Ensure a basic LDAP group mapper exists (optional but harmless).
              if not ldap_component_id:
                  print("WARNING: unable to determine LDAP component id; skipping group mapper")
                  raise SystemExit(0)

              status, components, _ = http_json(
                  "GET",
                  f"{base_url}/admin/realms/{realm}/components?type=org.keycloak.storage.ldap.mappers.LDAPStorageMapper",
                  token,
              )
              components = components or []
              group_mapper = None
              for c in components:
                  if c.get("name") == "openldap-groups" and c.get("parentId") == ldap_component_id:
                      group_mapper = c
                      break

              mapper_payload = {
                  "name": "openldap-groups",
                  "providerId": "group-ldap-mapper",
                  "providerType": "org.keycloak.storage.ldap.mappers.LDAPStorageMapper",
                  "parentId": ldap_component_id,
                  "config": {
                      "groups.dn": [ldap_groups_dn],
                      "group.name.ldap.attribute": ["cn"],
                      "group.object.classes": ["groupOfNames"],
                      "membership.ldap.attribute": ["member"],
                      "membership.attribute.type": ["DN"],
                      "mode": ["LDAP_ONLY"],
                      "user.roles.retrieve.strategy": ["LOAD_GROUPS_BY_MEMBER_ATTRIBUTE"],
                      "preserve.group.inheritance": ["true"],
                  },
              }

              if group_mapper:
                  mapper_payload["id"] = group_mapper["id"]
                  mapper_payload["parentId"] = group_mapper.get("parentId", mapper_payload["parentId"])
                  print(f"Updating LDAP group mapper: {mapper_payload['id']}")
                  status, _, _ = http_json(
                      "PUT",
                      f"{base_url}/admin/realms/{realm}/components/{mapper_payload['id']}",
                      token,
                      mapper_payload,
                  )
                  if status not in (200, 204):
                      raise SystemExit(f"Unexpected group mapper update status: {status}")
              else:
                  print("Creating LDAP group mapper")
                  status, _, _ = http_json(
                      "POST",
                      f"{base_url}/admin/realms/{realm}/components",
                      token,
                      mapper_payload,
                  )
                  if status not in (201, 204):
                      raise SystemExit(f"Unexpected group mapper create status: {status}")

              def ensure_user_attr_mapper(name: str, ldap_attr: str, user_attr: str):
                  mapper = None
                  for c in components:
                      if c.get("name") == name and c.get("parentId") == ldap_component_id:
                          mapper = c
                          break

                  payload = {
                      "name": name,
                      "providerId": "user-attribute-ldap-mapper",
                      "providerType": "org.keycloak.storage.ldap.mappers.LDAPStorageMapper",
                      "parentId": ldap_component_id,
                      "config": {
                          "ldap.attribute": [ldap_attr],
                          "user.model.attribute": [user_attr],
                          "read.only": ["false"],
                          "always.read.value.from.ldap": ["false"],
                          "is.mandatory.in.ldap": ["false"],
                      },
                  }

                  if mapper:
                      payload["id"] = mapper["id"]
                      payload["parentId"] = mapper.get("parentId", payload["parentId"])
                      print(f"Updating LDAP user mapper: {payload['id']} ({name})")
                      status, _, _ = http_json(
                          "PUT",
                          f"{base_url}/admin/realms/{realm}/components/{payload['id']}",
                          token,
                          payload,
                      )
                      if status not in (200, 204):
                          raise SystemExit(f"Unexpected user mapper update status for {name}: {status}")
                  else:
                      print(f"Creating LDAP user mapper: {name}")
                      status, _, _ = http_json(
                          "POST",
                          f"{base_url}/admin/realms/{realm}/components",
                          token,
                          payload,
                      )
                      if status not in (201, 204):
                          raise SystemExit(f"Unexpected user mapper create status for {name}: {status}")

              ensure_user_attr_mapper("openldap-email", "mail", "email")
              ensure_user_attr_mapper("openldap-first-name", "givenName", "firstName")
              ensure_user_attr_mapper("openldap-last-name", "sn", "lastName")

              # Cleanup duplicate LDAP federation providers and their child components (mappers, etc).
              # Keep only the canonical provider we updated/created above.
              try:
                  status, fresh_components, _ = http_json(
                      "GET",
                      f"{base_url}/admin/realms/{realm}/components",
                      token,
                  )
                  if status != 200:
                      raise Exception(f"unexpected components status {status}")
                  fresh_components = fresh_components or []

                  dup_provider_ids = []
                  for c in fresh_components:
                      if c.get("providerId") != "ldap":
                          continue
                      if c.get("providerType") != "org.keycloak.storage.UserStorageProvider":
                          continue
                      cid = c.get("id")
                      if not cid or cid == ldap_component_id:
                          continue
                      dup_provider_ids.append(cid)

                  if dup_provider_ids:
                      for pid in dup_provider_ids:
                          # Delete child components first.
                          for child in fresh_components:
                              if child.get("parentId") != pid:
                                  continue
                              child_id = child.get("id")
                              if not child_id:
                                  continue
                              try:
                                  http_json(
                                      "DELETE",
                                      f"{base_url}/admin/realms/{realm}/components/{child_id}",
                                      token,
                                  )
                              except urllib.error.HTTPError as e:
                                  print(f"WARNING: failed to delete LDAP child component {child_id} (status={e.code})")
                          try:
                              http_json(
                                  "DELETE",
                                  f"{base_url}/admin/realms/{realm}/components/{pid}",
                                  token,
                              )
                          except urllib.error.HTTPError as e:
                              print(f"WARNING: failed to delete duplicate LDAP provider {pid} (status={e.code})")
                      print(f"Cleaned up {len(dup_provider_ids)} duplicate LDAP federation providers")
              except Exception as e:
                  print(f"WARNING: LDAP cleanup failed (continuing): {e}")
              PY
          volumeMounts: