Files
shopdb-flask/shopdb/core/api/plugins.py
cproudlock 9b7da9f2d2 Make a plugin toggle take effect without a manual restart
Enabling a plugin could not make its routes live. Flask refuses
register_blueprint once the app has served a request, and each gunicorn worker
holds its own app object, so registering in the worker that served the toggle
would leave the others answering 404 for the same URL.

Worse, the window before the restart was a window in which workers disagreed
about reality. Four defects, all of which answered 200 while going wrong:

- Enabling a not-yet-installed plugin ALWAYS 500'd on a running app:
  install_plugin ran the migrations, updated the registry, and only then called
  register_blueprint and hit Flask's assertion. The toggle half-applied and
  reported failure.
- Enabling one plugin could hang the entire site. @jwt_required does SELECT
  users, so the request's transaction sits open holding a shared metadata lock;
  the plugin's migration runs on its own Alembic connection and needs an
  EXCLUSIVE lock on users for a foreign key. Each waits on the other, and MySQL
  grants metadata locks in request order, so every later login queued behind the
  pending exclusive. Observed twice as a total outage from one plugin toggle.
- Each worker rewrote the WHOLE registry file from a snapshot taken at boot, so
  concurrent enables erased each other: seven enables persisted three, with
  HTTP 200 for all seven.
- Dependency checks read that same stale snapshot, refusing a plugin whose
  dependency another worker had just enabled.

Routes still require the process to restart; what changes is who initiates it.
Each worker fingerprints instance/plugins.json at boot and exits gracefully when
it changes, and the supervisor replaces it. That is the primitive gunicorn's own
--reload uses. Recycles are staggered per worker so the pool rolls rather than
dropping at once, and the exit is scoped to gunicorn: on Windows sys.exit() in a
daemon thread ends the thread rather than the process, so IIS keeps the manual
restart until that path is proven on the Windows VM.

The registry now merges under an exclusive lock and replaces atomically, saves
only the entry being changed, and re-reads before any decision is made from it.

ADR-020 records the decision; ADR-013 is amended where it said the UI must tell
the operator to restart. The restart requirement and the honesty requirement
both stand - only the manual step is removed, and only where it is proven.

(cherry picked from commit 9d6342b3ab)
2026-09-20 09:07:54 -04:00

133 lines
5.5 KiB
Python

"""Plugin introspection + enable/disable API."""
from flask import Blueprint, current_app, request
from flask_jwt_extended import jwt_required
from shopdb.plugins.base import renderable_settings_routes
from shopdb.utils.responses import success_response, error_response, ErrorCodes
from shopdb.utils.authz import require_role
from shopdb.utils.authz import require_permission
plugins_bp = Blueprint('plugins', __name__)
@plugins_bp.route('', methods=['GET'])
@require_permission('settings.view')
def list_plugins():
"""List all discovered plugins (enabled or not) with their metadata and
the framework contract version."""
from shopdb import __contract_version__
pm = current_app.extensions.get('plugin_manager')
plugins = pm.discover_available() if pm else []
plugins.sort(key=lambda p: p['name'])
return success_response({
'contract_version': __contract_version__,
'count': len(plugins),
'plugins': plugins,
})
@plugins_bp.route('/enabled', methods=['GET'])
@jwt_required(optional=True)
def list_enabled_plugins():
"""Return just the names of enabled plugins as a flat array.
Cheap registry read (no DB). Exposed to anonymous callers on purpose:
the navigation endpoint already leaks the same enabled/disabled signal,
and the frontend needs it (including unauthenticated kiosk routes like
/tv) to gate plugin-owned routes. No metadata beyond the names.
"""
pm = current_app.extensions.get('plugin_manager')
names = pm.registry.get_enabled_plugins() if pm else []
return success_response(sorted(names))
@plugins_bp.route('/settings-routes', methods=['GET'])
@require_permission('settings.view')
def get_plugin_settings_routes():
"""Return settings routes contributed by all enabled plugins.
Plugins can register routes to add tabs/pages to settings hubs like
Model Catalog, Asset Config, etc. Frontend router merges these dynamically.
Gated on settings.view, the same permission list_plugins carries. It was
optional-auth and listed as a public read, on the stated reasoning that
unauthenticated pages (kiosks, TV displays) register plugin tabs before
login. Nothing does: the only caller in the frontend is
views/settings/ModelCatalogHub.vue, which sits behind requiresAuth and
requiresAdmin, and no kiosk route touches it. Open, it handed an anonymous
caller the enabled plugin list, the settings hub layout and the internal Vue
component paths - free reconnaissance about which modules a site runs. The
login page never calls this, so gating it costs nothing pre-login.
A route naming a hub that has no code to draw plugin tabs is DROPPED here,
loudly. Returning it would put an entry in a response that nothing reads:
the plugin author sees a correct declaration and an API that accepts it, and
the tab never appears anywhere. Dropping it costs the same missing tab and
leaves a line in the log that says which plugin, which hub, and what the
supported set is. See SETTINGS_HUBS_WITH_PLUGIN_TABS in shopdb/plugins/base.
Returns: { routes: [{hub, path, name, tab, component}, ...] }
"""
pm = current_app.extensions.get('plugin_manager')
if not pm:
return success_response({'routes': []})
routes = []
for plugin_name, plugin in pm.get_all_plugins().items():
# Only include routes from enabled plugins
if not pm.registry.is_enabled(plugin_name):
continue
if not hasattr(plugin, 'get_settings_routes'):
continue
# One broken plugin must not take every settings hub down with it.
try:
plugin_routes = plugin.get_settings_routes()
except Exception:
current_app.logger.exception(
'get_settings_routes failed for plugin %s; skipping it',
plugin_name)
continue
routes.extend(renderable_settings_routes(plugin_name, plugin_routes,
current_app.logger))
return success_response({'routes': routes})
@plugins_bp.route('/<name>', methods=['PUT'])
@jwt_required()
@require_role('admin')
def set_plugin_enabled(name: str):
"""Enable or disable a plugin. Takes effect on the next app restart for
route/navigation changes."""
data = request.get_json() or {}
if 'enabled' not in data:
return error_response(ErrorCodes.VALIDATION_ERROR, 'enabled is required')
pm = current_app.extensions.get('plugin_manager')
if not pm:
return error_response(ErrorCodes.INTERNAL_ERROR,
'Plugin manager unavailable', http_code=500)
want = bool(data['enabled'])
ok = pm.enable_plugin(name) if want else pm.disable_plugin(name)
if not ok:
# enable/disable refused (unknown plugin, or a dependency conflict)
return error_response(
ErrorCodes.CONFLICT,
f"Could not {'enable' if want else 'disable'} '{name}' "
f"(check it exists and dependencies allow it)",
http_code=409
)
return success_response(
{'name': name, 'enabled': want},
# No longer "restart to apply": each worker notices the change to
# plugins.json and recycles itself, so the routes follow on their own
# within seconds. Callers poll /setup/health to know when that has
# finished across the pool.
message=f"Plugin {'enabled' if want else 'disabled'}. "
f"It comes online as the workers pick up the change."
)