Files
shopdb-flask/tests/test_plugins/test_endpoints_enroll_limits.py
cproudlock 066a19052b
Some checks failed
CI / backend (push) Has been cancelled
CI / naming (push) Has been cancelled
CI / gates (push) Has been cancelled
CI / frontend (push) Has been cancelled
CI / lean-build (push) Has been cancelled
Act on the wj architectural review: races, limits, pool and contract gaps
Findings F2, F4, F5, F6, F7, F8, F9, F10, F11 and F13 from
wj-architectural-review.md. F1/F1a (the non-sargable LOWER pattern) and F3
(the warranty sync holding a pooled connection) are being handled separately
and are untouched here.

- A collector check-in that raced another for the same PC lost everything:
  the unique index on the application fingerprint raised, nothing caught it,
  and the whole check-in rolled back, discarding state, events and hardware,
  answering the agent with form copy meant for a person. The insert now runs
  in a savepoint, reconciles what the other transaction already wrote, one row
  per savepoint so a second collision skips a row rather than the batch, and
  the check-in commits with a warning. This is the shape the delete side of
  the same function already had.
- Enrollment is deliberately unauthenticated, so it now carries a per-address
  rate limit and a cap on pending requests, and the cap is a setting an admin
  can see and change. A flood can no longer bury the approval queue or grow
  the table without bound.
- The fleet alert poll no longer formats debug strings on every request.
- BusinessUnit is on the contract surface (0.34.0), so the endpoints plugin
  stops reaching into a core table with raw SQL.
- A test walks every asset type descriptor and proves the endpoint it names
  really refuses a caller without the declared permission.
- The plugin settings routes now need a permission; they were telling anyone
  which modules a site runs and where its components live.
- The connection pool is sized explicitly with a short timeout, and the
  arithmetic against the server's connection ceiling is written next to the
  worker count instead of being left to defaults nobody chose.
- Fleet alerts carry both the correct lowercase keys and the old camelCase
  pair. Trays older than 1.7.0 match keys case-sensitively and would have gone
  quiet on every PC not yet upgraded; the old pair goes when they are.
- The dev server proxy points at localhost again, with environment variables
  for the remote-box workflow, and one developer's address is out of the
  server certificate page.
2026-09-16 08:57:04 -04:00

221 lines
8.2 KiB
Python

"""The two bounds on POST /api/endpoints/enroll.
The route takes no auth on purpose - a PC installed without the shared key must
still be able to reach an admin - and it writes a row for every new
(hostname, serialnumber) pair. That is an unauthenticated caller writing to the
database in a loop, and it is also a denial of the enrollment workflow: junk
pending rows bury the real requests an admin has to approve.
Two bounds, and they answer different questions. The per-IP fixed-window limiter
bounds how FAST; the settings-backed cap bounds how MANY rows exist. Neither may
touch a PC already in the workflow, so the cap refuses only a hostname and
serial never seen before and an existing pending or approved row still answers
exactly as it did.
"""
import pytest
from shopdb.core.models import Asset, AssetType, Setting
from plugins.endpoints.api.enrollment import PENDING_CAP_SETTING
from plugins.endpoints.models import EndpointEnrollRequest
ENROLL_URL = '/api/endpoints/enroll'
NONCE = 'n' * 40
@pytest.fixture
def limits(app):
"""Both bounds on, at their shipped defaults, restored afterwards.
The limiter counts in the cache extension, which is process-wide and
outlives a test. Any other test in the same worker that posts to /enroll
spends budget in the same window and IP bucket, so a count-exactly test has
to start from a clean counter - and leave one behind, since it deliberately
fills a bucket. Which tests share a worker depends on the xdist
distribution, so without this the failure is intermittent and looks
unrelated to whatever changed.
"""
from shopdb.api import cache
previous = {key: app.config.get(key) for key in (
'ENDPOINTS_ENROLL_RATELIMIT_ENABLED',
'ENDPOINTS_ENROLL_RATELIMIT_MAX',
'ENDPOINTS_ENROLL_RATELIMIT_WINDOW_SECONDS')}
app.config['ENDPOINTS_ENROLL_RATELIMIT_ENABLED'] = True
with app.app_context():
cache.clear()
yield app
with app.app_context():
cache.clear()
for key, value in previous.items():
if value is None:
app.config.pop(key, None)
else:
app.config[key] = value
def _computer(db, hostname, serialnumber):
from plugins.computers.models import Computer
atype = AssetType.query.filter_by(assettype='computer').first()
if atype is None:
atype = AssetType(assettype='computer')
db.session.add(atype)
db.session.flush()
asset = Asset(assetnumber=hostname, assettypeid=atype.assettypeid,
serialnumber=serialnumber, isactive=True)
db.session.add(asset)
db.session.flush()
db.session.add(Computer(assetid=asset.assetid, hostname=hostname))
db.session.commit()
return asset.assetid
def _body(hostname, serialnumber=None, nonce=NONCE):
return {'hostname': hostname,
'serialnumber': serialnumber or (hostname + '-SER'),
'enrollnonce': nonce, 'agentversion': '1.20.0'}
def _cap(db, value):
Setting.set(PENDING_CAP_SETTING, value, valuetype='integer',
category='plugin.endpoints')
db.session.commit()
def _pending(db, hostname, serialnumber=None, nonce=NONCE,
status=EndpointEnrollRequest.STATUS_PENDING):
from datetime import datetime, timezone
now = datetime.now(timezone.utc).replace(tzinfo=None)
row = EndpointEnrollRequest(
hostname=hostname, serialnumber=serialnumber or (hostname + '-SER'),
enrollnonce=nonce, status=status, requestedat=now)
db.session.add(row)
db.session.commit()
return row
def _code(response):
return response.get_json()['data']['error']['code']
# -- rate limit --------------------------------------------------------------
def test_the_limit_triggers_and_answers_the_standard_shape(client, db, limits):
"""Over budget is a 429 RATE_LIMITED, the same shape as every other
limiter in the product."""
limits.config['ENDPOINTS_ENROLL_RATELIMIT_MAX'] = 3
# An hour, so the fixed window cannot roll over mid-test and reset the
# count. The test is about the budget, not about the clock.
limits.config['ENDPOINTS_ENROLL_RATELIMIT_WINDOW_SECONDS'] = 3600
codes = []
for index in range(5):
response = client.post(ENROLL_URL, json=_body('BAY-%02d' % index))
codes.append(response.status_code)
assert codes[:3] == [202, 202, 202], codes
assert codes[3:] == [429, 429], codes
refused = client.post(ENROLL_URL, json=_body('BAY-99'))
assert refused.status_code == 429
assert _code(refused) == 'RATE_LIMITED'
def test_the_limit_is_off_when_the_config_says_so(client, db, limits):
limits.config['ENDPOINTS_ENROLL_RATELIMIT_ENABLED'] = False
limits.config['ENDPOINTS_ENROLL_RATELIMIT_MAX'] = 1
for index in range(4):
response = client.post(ENROLL_URL, json=_body('OFF-%02d' % index))
assert response.status_code == 202, response.get_data(as_text=True)
# -- pending cap -------------------------------------------------------------
def test_the_cap_refuses_a_new_row(client, db, limits):
"""At the cap, a hostname and serial nobody has seen is told to try later,
and no row is written for it."""
limits.config['ENDPOINTS_ENROLL_RATELIMIT_ENABLED'] = False
_cap(db, 2)
_pending(db, 'OLD-01')
_pending(db, 'OLD-02')
response = client.post(ENROLL_URL, json=_body('NEW-01'))
assert response.status_code == 503, response.get_data(as_text=True)
assert _code(response) == 'ENROLL_PENDING_FULL'
assert EndpointEnrollRequest.query.filter_by(hostname='NEW-01').count() == 0
assert EndpointEnrollRequest.query.count() == 2
def test_the_cap_still_serves_an_existing_pending_row(client, db, limits):
"""A PC already in the queue keeps asking and keeps being answered - the cap
must not lock the fleet out of the workflow it is protecting."""
limits.config['ENDPOINTS_ENROLL_RATELIMIT_ENABLED'] = False
_cap(db, 1)
row = _pending(db, 'OLD-01')
response = client.post(ENROLL_URL, json=_body('OLD-01'))
assert response.status_code == 202, response.get_data(as_text=True)
assert response.get_json()['data']['status'] == 'pending'
db.session.refresh(row)
assert row.lastseenat is not None
def test_the_cap_still_issues_to_an_approved_row(client, db, limits):
"""An approved request hands out its credential at the cap, because that
row is the admin's decision and refusing it would strand the PC."""
limits.config['ENDPOINTS_ENROLL_RATELIMIT_ENABLED'] = False
assetid = _computer(db, 'BAY-07', 'BAY-07-SER')
_cap(db, 1)
_pending(db, 'FILLER-01')
approved = _pending(db, 'BAY-07',
status=EndpointEnrollRequest.STATUS_APPROVED)
approved.matchedassetid = assetid
db.session.commit()
response = client.post(ENROLL_URL, json=_body('BAY-07'))
assert response.status_code == 201, response.get_data(as_text=True)
assert response.get_json()['data']['credential'].startswith('sdbe_')
def test_the_cap_setting_is_honoured(client, db, limits):
"""Raise the setting and the same request that was refused goes through;
0 removes the cap entirely."""
limits.config['ENDPOINTS_ENROLL_RATELIMIT_ENABLED'] = False
_cap(db, 1)
_pending(db, 'OLD-01')
refused = client.post(ENROLL_URL, json=_body('NEW-01'))
assert refused.status_code == 503
_cap(db, 5)
allowed = client.post(ENROLL_URL, json=_body('NEW-01'))
assert allowed.status_code == 202, allowed.get_data(as_text=True)
_cap(db, 0)
uncapped = client.post(ENROLL_URL, json=_body('NEW-02'))
assert uncapped.status_code == 202, uncapped.get_data(as_text=True)
def test_the_cap_is_editable_through_the_enrollment_settings_route(
client, db, auth_headers):
"""A setting nobody can change is a constant with extra steps."""
response = client.put('/api/endpoints/enrollment/settings',
json={'maxpendingenrollments': 42},
headers=auth_headers)
assert response.status_code == 200, response.get_data(as_text=True)
assert response.get_json()['data']['maxpendingenrollments'] == 42
assert int(Setting.get(PENDING_CAP_SETTING)) == 42
bad = client.put('/api/endpoints/enrollment/settings',
json={'maxpendingenrollments': -1}, headers=auth_headers)
assert bad.status_code == 400