fix: preserve data granularity and recover tasks safely
This commit is contained in:
@@ -79,3 +79,64 @@ def test_task_pause_resume_reuses_persisted_work_without_duplicate_entries(tmp_p
|
||||
assert jobs.get(job["id"])["completed_units"] == 2
|
||||
jobs.close()
|
||||
|
||||
|
||||
def test_service_shutdown_pauses_instead_of_cancelling_persisted_job(tmp_path):
|
||||
db, settings = Database(tmp_path / 'studio.db'), SettingsStore(tmp_path)
|
||||
entered = threading.Event()
|
||||
def runner(jid, params, manager):
|
||||
entered.set()
|
||||
while True:
|
||||
manager.checkpoint(jid)
|
||||
time.sleep(.01)
|
||||
manager = JobManager(db, settings, runner=runner)
|
||||
job = manager.start({'mode': 'catalog', 'year': 2027})
|
||||
assert entered.wait(2)
|
||||
manager.close()
|
||||
assert Database(tmp_path / 'studio.db').job_get(job['id'])['status'] == 'paused'
|
||||
|
||||
|
||||
def test_business_rate_limit_uses_bounded_backoff_without_skipping_a_page():
|
||||
from studio.network import RateLimited
|
||||
class Limited(CatalogTransport):
|
||||
def __init__(self):
|
||||
self.calls, self.waits = 0, []
|
||||
def request(self, *args, **kwargs):
|
||||
self.calls += 1
|
||||
return Response({'flag': False, 'msg': '访问太频繁,请稍后重试'})
|
||||
def sleep(self, seconds):
|
||||
self.waits.append(seconds)
|
||||
transport = Limited()
|
||||
with pytest.raises(RateLimited):
|
||||
list(CatalogClient(transport).schools('11'))
|
||||
assert transport.calls == 3
|
||||
assert transport.waits == [5, 10]
|
||||
|
||||
|
||||
def test_recovery_waits_for_failed_worker_to_exit_before_launching_another(tmp_path):
|
||||
db, settings = Database(tmp_path / 'studio.db'), SettingsStore(tmp_path)
|
||||
entered, release = threading.Event(), threading.Event()
|
||||
original_log = db.log
|
||||
def guarded_log(jid, level, message):
|
||||
if level == 'error':
|
||||
entered.set()
|
||||
release.wait(2)
|
||||
original_log(jid, level, message)
|
||||
db.log = guarded_log
|
||||
attempts = []
|
||||
def runner(jid, params, manager):
|
||||
attempts.append(jid)
|
||||
if len(attempts) == 1:
|
||||
raise ValueError('failure fixture')
|
||||
manager.complete_unit(jid, 'recovered')
|
||||
manager = JobManager(db, settings, runner=runner)
|
||||
job = manager.start({'mode': 'catalog', 'year': 2027})
|
||||
assert entered.wait(2)
|
||||
threading.Timer(.05, release.set).start()
|
||||
manager.resume(job['id'])
|
||||
deadline = time.monotonic() + 2
|
||||
while manager.get(job['id'])['status'] != 'completed' and time.monotonic() < deadline:
|
||||
time.sleep(.02)
|
||||
assert manager.get(job['id'])['status'] == 'completed'
|
||||
assert len(attempts) == 2
|
||||
manager.close()
|
||||
|
||||
|
||||
Reference in New Issue
Block a user