mirror of https://github.com/zulip/zulip.git
776 lines
32 KiB
Python
776 lines
32 KiB
Python
import base64
|
|
import datetime
|
|
import os
|
|
import signal
|
|
import time
|
|
from collections import defaultdict
|
|
from contextlib import contextmanager, suppress
|
|
from inspect import isabstract
|
|
from typing import Any, Callable, Dict, Iterator, List, Mapping, Optional
|
|
from unittest.mock import MagicMock, patch
|
|
|
|
import orjson
|
|
from django.conf import settings
|
|
from django.db.utils import IntegrityError
|
|
from django.test import override_settings
|
|
|
|
from zerver.lib.email_mirror import RateLimitedRealmMirror
|
|
from zerver.lib.email_mirror_helpers import encode_email_address
|
|
from zerver.lib.queue import MAX_REQUEST_RETRIES
|
|
from zerver.lib.rate_limiter import RateLimiterLockingError
|
|
from zerver.lib.remote_server import PushNotificationBouncerRetryLaterError
|
|
from zerver.lib.send_email import EmailNotDeliveredError, FromAddress
|
|
from zerver.lib.test_classes import ZulipTestCase
|
|
from zerver.lib.test_helpers import mock_queue_publish
|
|
from zerver.models import (
|
|
NotificationTriggers,
|
|
PreregistrationUser,
|
|
ScheduledMessageNotificationEmail,
|
|
UserActivity,
|
|
UserProfile,
|
|
get_client,
|
|
get_realm,
|
|
get_stream,
|
|
)
|
|
from zerver.tornado.event_queue import build_offline_notification
|
|
from zerver.worker import queue_processors
|
|
from zerver.worker.queue_processors import (
|
|
EmailSendingWorker,
|
|
FetchLinksEmbedData,
|
|
LoopQueueProcessingWorker,
|
|
MissedMessageWorker,
|
|
QueueProcessingWorker,
|
|
get_active_worker_queues,
|
|
)
|
|
|
|
Event = Dict[str, Any]
|
|
|
|
|
|
class FakeClient:
|
|
def __init__(self, prefetch: int = 0) -> None:
|
|
self.queues: Dict[str, List[Dict[str, Any]]] = defaultdict(list)
|
|
|
|
def enqueue(self, queue_name: str, data: Dict[str, Any]) -> None:
|
|
self.queues[queue_name].append(data)
|
|
|
|
def start_json_consumer(
|
|
self,
|
|
queue_name: str,
|
|
callback: Callable[[List[Dict[str, Any]]], None],
|
|
batch_size: int = 1,
|
|
timeout: Optional[int] = None,
|
|
) -> None:
|
|
chunk: List[Dict[str, Any]] = []
|
|
queue = self.queues[queue_name]
|
|
while queue:
|
|
chunk.append(queue.pop(0))
|
|
if len(chunk) >= batch_size or not len(queue):
|
|
callback(chunk)
|
|
chunk = []
|
|
|
|
def local_queue_size(self) -> int:
|
|
return sum(len(q) for q in self.queues.values())
|
|
|
|
|
|
@contextmanager
|
|
def simulated_queue_client(client: FakeClient) -> Iterator[None]:
|
|
with patch.object(queue_processors, "SimpleQueueClient", lambda *args, **kwargs: client):
|
|
yield
|
|
|
|
|
|
class WorkerTest(ZulipTestCase):
|
|
def test_UserActivityWorker(self) -> None:
|
|
fake_client = FakeClient()
|
|
|
|
user = self.example_user("hamlet")
|
|
UserActivity.objects.filter(
|
|
user_profile=user.id,
|
|
client=get_client("ios"),
|
|
).delete()
|
|
|
|
data = dict(
|
|
user_profile_id=user.id,
|
|
client_id=get_client("ios").id,
|
|
time=time.time(),
|
|
query="send_message",
|
|
)
|
|
fake_client.enqueue("user_activity", data)
|
|
|
|
# Now process the event a second time and confirm count goes
|
|
# up. Ideally, we'd use an event with a slightly newer
|
|
# time, but it's not really important.
|
|
fake_client.enqueue("user_activity", data)
|
|
with simulated_queue_client(fake_client):
|
|
worker = queue_processors.UserActivityWorker()
|
|
worker.setup()
|
|
worker.start()
|
|
activity_records = UserActivity.objects.filter(
|
|
user_profile=user.id,
|
|
client=get_client("ios"),
|
|
)
|
|
self.assert_length(activity_records, 1)
|
|
self.assertEqual(activity_records[0].count, 2)
|
|
|
|
def test_missed_message_worker(self) -> None:
|
|
cordelia = self.example_user("cordelia")
|
|
hamlet = self.example_user("hamlet")
|
|
othello = self.example_user("othello")
|
|
|
|
hamlet1_msg_id = self.send_personal_message(
|
|
from_user=cordelia,
|
|
to_user=hamlet,
|
|
content="hi hamlet",
|
|
)
|
|
|
|
hamlet2_msg_id = self.send_personal_message(
|
|
from_user=cordelia,
|
|
to_user=hamlet,
|
|
content="goodbye hamlet",
|
|
)
|
|
|
|
hamlet3_msg_id = self.send_personal_message(
|
|
from_user=cordelia,
|
|
to_user=hamlet,
|
|
content="hello again hamlet",
|
|
)
|
|
|
|
othello_msg_id = self.send_personal_message(
|
|
from_user=cordelia,
|
|
to_user=othello,
|
|
content="where art thou, othello?",
|
|
)
|
|
|
|
hamlet_event1 = dict(
|
|
user_profile_id=hamlet.id,
|
|
message_id=hamlet1_msg_id,
|
|
trigger=NotificationTriggers.PRIVATE_MESSAGE,
|
|
)
|
|
hamlet_event2 = dict(
|
|
user_profile_id=hamlet.id,
|
|
message_id=hamlet2_msg_id,
|
|
trigger=NotificationTriggers.PRIVATE_MESSAGE,
|
|
mentioned_user_group_id=4,
|
|
)
|
|
othello_event = dict(
|
|
user_profile_id=othello.id,
|
|
message_id=othello_msg_id,
|
|
trigger=NotificationTriggers.PRIVATE_MESSAGE,
|
|
)
|
|
|
|
events = [hamlet_event1, hamlet_event2, othello_event]
|
|
|
|
fake_client = FakeClient()
|
|
for event in events:
|
|
fake_client.enqueue("missedmessage_emails", event)
|
|
|
|
mmw = MissedMessageWorker()
|
|
batch_duration = datetime.timedelta(
|
|
seconds=hamlet.email_notifications_batching_period_seconds
|
|
)
|
|
assert (
|
|
hamlet.email_notifications_batching_period_seconds
|
|
== othello.email_notifications_batching_period_seconds
|
|
)
|
|
|
|
class MockTimer:
|
|
is_running = False
|
|
|
|
def is_alive(self) -> bool:
|
|
return self.is_running
|
|
|
|
def start(self) -> None:
|
|
self.is_running = True
|
|
|
|
timer = MockTimer()
|
|
timer_mock = patch(
|
|
"zerver.worker.queue_processors.Timer",
|
|
return_value=timer,
|
|
)
|
|
|
|
send_mock = patch(
|
|
"zerver.lib.email_notifications.do_send_missedmessage_events_reply_in_zulip",
|
|
)
|
|
|
|
bonus_event_hamlet = dict(
|
|
user_profile_id=hamlet.id,
|
|
message_id=hamlet3_msg_id,
|
|
trigger=NotificationTriggers.PRIVATE_MESSAGE,
|
|
)
|
|
|
|
def check_row(
|
|
row: ScheduledMessageNotificationEmail,
|
|
scheduled_timestamp: datetime.datetime,
|
|
mentioned_user_group_id: Optional[int],
|
|
) -> None:
|
|
self.assertEqual(row.trigger, NotificationTriggers.PRIVATE_MESSAGE)
|
|
self.assertEqual(row.scheduled_timestamp, scheduled_timestamp)
|
|
self.assertEqual(row.mentioned_user_group_id, mentioned_user_group_id)
|
|
|
|
with send_mock as sm, timer_mock as tm:
|
|
with simulated_queue_client(fake_client):
|
|
self.assertFalse(timer.is_alive())
|
|
|
|
time_zero = datetime.datetime(2021, 1, 1, tzinfo=datetime.timezone.utc)
|
|
expected_scheduled_timestamp = time_zero + batch_duration
|
|
with patch("zerver.worker.queue_processors.timezone_now", return_value=time_zero):
|
|
mmw.setup()
|
|
mmw.start()
|
|
|
|
# The events should be saved in the database
|
|
hamlet_row1 = ScheduledMessageNotificationEmail.objects.get(
|
|
user_profile_id=hamlet.id, message_id=hamlet1_msg_id
|
|
)
|
|
check_row(hamlet_row1, expected_scheduled_timestamp, None)
|
|
|
|
hamlet_row2 = ScheduledMessageNotificationEmail.objects.get(
|
|
user_profile_id=hamlet.id, message_id=hamlet2_msg_id
|
|
)
|
|
check_row(hamlet_row2, expected_scheduled_timestamp, 4)
|
|
|
|
othello_row1 = ScheduledMessageNotificationEmail.objects.get(
|
|
user_profile_id=othello.id, message_id=othello_msg_id
|
|
)
|
|
check_row(othello_row1, expected_scheduled_timestamp, None)
|
|
|
|
# Additionally, the timer should have be started
|
|
self.assertTrue(timer.is_alive())
|
|
|
|
# If another event is received, test that it gets saved with the same
|
|
# `expected_scheduled_timestamp` as the earlier events.
|
|
fake_client.enqueue("missedmessage_emails", bonus_event_hamlet)
|
|
self.assertTrue(timer.is_alive())
|
|
few_moments_later = time_zero + datetime.timedelta(seconds=3)
|
|
with patch(
|
|
"zerver.worker.queue_processors.timezone_now", return_value=few_moments_later
|
|
):
|
|
# Double-calling start is our way to get it to run again
|
|
mmw.start()
|
|
hamlet_row3 = ScheduledMessageNotificationEmail.objects.get(
|
|
user_profile_id=hamlet.id, message_id=hamlet3_msg_id
|
|
)
|
|
check_row(hamlet_row3, expected_scheduled_timestamp, None)
|
|
|
|
# Now let us test `maybe_send_batched_emails`
|
|
# If called too early, it shouldn't process the emails.
|
|
one_minute_premature = expected_scheduled_timestamp - datetime.timedelta(seconds=60)
|
|
with patch(
|
|
"zerver.worker.queue_processors.timezone_now", return_value=one_minute_premature
|
|
):
|
|
mmw.maybe_send_batched_emails()
|
|
self.assertEqual(ScheduledMessageNotificationEmail.objects.count(), 4)
|
|
|
|
# If called after `expected_scheduled_timestamp`, it should process all emails.
|
|
one_minute_overdue = expected_scheduled_timestamp + datetime.timedelta(seconds=60)
|
|
with self.assertLogs(level="INFO") as info_logs, patch(
|
|
"zerver.worker.queue_processors.timezone_now", return_value=one_minute_overdue
|
|
):
|
|
mmw.maybe_send_batched_emails()
|
|
self.assertEqual(ScheduledMessageNotificationEmail.objects.count(), 0)
|
|
|
|
self.assert_length(info_logs.output, 2)
|
|
self.assertIn(
|
|
f"INFO:root:Batch-processing 3 missedmessage_emails events for user {hamlet.id}",
|
|
info_logs.output,
|
|
)
|
|
self.assertIn(
|
|
f"INFO:root:Batch-processing 1 missedmessage_emails events for user {othello.id}",
|
|
info_logs.output,
|
|
)
|
|
|
|
# All batches got processed. Verify that the timer isn't running.
|
|
self.assertEqual(mmw.timer_event, None)
|
|
|
|
# Hacky test coming up! We want to test the try-except block in the consumer which handles
|
|
# IntegrityErrors raised when the message was deleted before it processed the notification
|
|
# event.
|
|
# However, Postgres defers checking ForeignKey constraints to when the current transaction
|
|
# commits. This poses some difficulties in testing because of Django running tests inside a
|
|
# transaction which never commits. See https://code.djangoproject.com/ticket/22431 for more
|
|
# details, but the summary is that IntegrityErrors due to database constraints are raised at
|
|
# the end of the test, not inside the `try` block. So, we have the code inside the `try` block
|
|
# raise `IntegrityError` by mocking.
|
|
def raise_error(**kwargs: Any) -> None:
|
|
raise IntegrityError
|
|
|
|
fake_client.enqueue("missedmessage_emails", hamlet_event1)
|
|
|
|
with patch(
|
|
"zerver.models.ScheduledMessageNotificationEmail.objects.create",
|
|
side_effect=raise_error,
|
|
), self.assertLogs(level="DEBUG") as debug_logs:
|
|
mmw.start()
|
|
self.assertIn(
|
|
"DEBUG:root:ScheduledMessageNotificationEmail row could not be created. The message may have been deleted. Skipping event.",
|
|
debug_logs.output,
|
|
)
|
|
|
|
# Check that the frequency of calling maybe_send_batched_emails is correct (5 seconds)
|
|
self.assertEqual(tm.call_args[0][0], 5)
|
|
|
|
# Verify the payloads now
|
|
args = [c[0] for c in sm.call_args_list]
|
|
arg_dict = {
|
|
arg[0].id: dict(
|
|
missed_messages=arg[1],
|
|
count=arg[2],
|
|
)
|
|
for arg in args
|
|
}
|
|
|
|
hamlet_info = arg_dict[hamlet.id]
|
|
self.assertEqual(hamlet_info["count"], 3)
|
|
self.assertEqual(
|
|
{m["message"].content for m in hamlet_info["missed_messages"]},
|
|
{"hi hamlet", "goodbye hamlet", "hello again hamlet"},
|
|
)
|
|
|
|
othello_info = arg_dict[othello.id]
|
|
self.assertEqual(othello_info["count"], 1)
|
|
self.assertEqual(
|
|
{m["message"].content for m in othello_info["missed_messages"]},
|
|
{"where art thou, othello?"},
|
|
)
|
|
|
|
with send_mock as sm, timer_mock as tm:
|
|
with simulated_queue_client(fake_client):
|
|
time_zero = datetime.datetime(2021, 1, 1, tzinfo=datetime.timezone.utc)
|
|
# Verify that we make forward progress if one of the messages throws an exception
|
|
fake_client.enqueue("missedmessage_emails", hamlet_event1)
|
|
fake_client.enqueue("missedmessage_emails", hamlet_event2)
|
|
fake_client.enqueue("missedmessage_emails", othello_event)
|
|
with patch("zerver.worker.queue_processors.timezone_now", return_value=time_zero):
|
|
mmw.setup()
|
|
mmw.start()
|
|
|
|
def fail_some(user: UserProfile, *args: Any) -> None:
|
|
if user.id == hamlet.id:
|
|
raise RuntimeError
|
|
|
|
sm.side_effect = fail_some
|
|
one_minute_overdue = expected_scheduled_timestamp + datetime.timedelta(seconds=60)
|
|
with patch(
|
|
"zerver.worker.queue_processors.timezone_now", return_value=one_minute_overdue
|
|
), self.assertLogs(level="ERROR") as error_logs:
|
|
mmw.maybe_send_batched_emails()
|
|
self.assertIn(
|
|
"ERROR:root:Failed to process 2 missedmessage_emails for user 10",
|
|
error_logs.output[0],
|
|
)
|
|
self.assertEqual(ScheduledMessageNotificationEmail.objects.count(), 0)
|
|
|
|
def test_push_notifications_worker(self) -> None:
|
|
"""
|
|
The push notifications system has its own comprehensive test suite,
|
|
so we can limit ourselves to simple unit testing the queue processor,
|
|
without going deeper into the system - by mocking the handle_push_notification
|
|
functions to immediately produce the effect we want, to test its handling by the queue
|
|
processor.
|
|
"""
|
|
fake_client = FakeClient()
|
|
|
|
def fake_publish(
|
|
queue_name: str, event: Dict[str, Any], processor: Callable[[Any], None]
|
|
) -> None:
|
|
fake_client.enqueue(queue_name, event)
|
|
|
|
def generate_new_message_notification() -> Dict[str, Any]:
|
|
return build_offline_notification(1, 1)
|
|
|
|
def generate_remove_notification() -> Dict[str, Any]:
|
|
return {
|
|
"type": "remove",
|
|
"user_profile_id": 1,
|
|
"message_ids": [1],
|
|
}
|
|
|
|
with simulated_queue_client(fake_client):
|
|
worker = queue_processors.PushNotificationsWorker()
|
|
worker.setup()
|
|
with patch(
|
|
"zerver.worker.queue_processors.handle_push_notification"
|
|
) as mock_handle_new, patch(
|
|
"zerver.worker.queue_processors.handle_remove_push_notification"
|
|
) as mock_handle_remove, patch(
|
|
"zerver.worker.queue_processors.initialize_push_notifications"
|
|
):
|
|
event_new = generate_new_message_notification()
|
|
event_remove = generate_remove_notification()
|
|
fake_client.enqueue("missedmessage_mobile_notifications", event_new)
|
|
fake_client.enqueue("missedmessage_mobile_notifications", event_remove)
|
|
|
|
worker.start()
|
|
mock_handle_new.assert_called_once_with(event_new["user_profile_id"], event_new)
|
|
mock_handle_remove.assert_called_once_with(
|
|
event_remove["user_profile_id"], event_remove["message_ids"]
|
|
)
|
|
|
|
with patch(
|
|
"zerver.worker.queue_processors.handle_push_notification",
|
|
side_effect=PushNotificationBouncerRetryLaterError("test"),
|
|
) as mock_handle_new, patch(
|
|
"zerver.worker.queue_processors.handle_remove_push_notification",
|
|
side_effect=PushNotificationBouncerRetryLaterError("test"),
|
|
) as mock_handle_remove, patch(
|
|
"zerver.worker.queue_processors.initialize_push_notifications"
|
|
):
|
|
event_new = generate_new_message_notification()
|
|
event_remove = generate_remove_notification()
|
|
fake_client.enqueue("missedmessage_mobile_notifications", event_new)
|
|
fake_client.enqueue("missedmessage_mobile_notifications", event_remove)
|
|
|
|
with mock_queue_publish(
|
|
"zerver.lib.queue.queue_json_publish", side_effect=fake_publish
|
|
), self.assertLogs("zerver.worker.queue_processors", "WARNING") as warn_logs:
|
|
worker.start()
|
|
self.assertEqual(mock_handle_new.call_count, 1 + MAX_REQUEST_RETRIES)
|
|
self.assertEqual(mock_handle_remove.call_count, 1 + MAX_REQUEST_RETRIES)
|
|
self.assertEqual(
|
|
warn_logs.output,
|
|
[
|
|
"WARNING:zerver.worker.queue_processors:Maximum retries exceeded for trigger:1 event:push_notification",
|
|
]
|
|
* 2,
|
|
)
|
|
|
|
@patch("zerver.worker.queue_processors.mirror_email")
|
|
def test_mirror_worker(self, mock_mirror_email: MagicMock) -> None:
|
|
fake_client = FakeClient()
|
|
stream = get_stream("Denmark", get_realm("zulip"))
|
|
stream_to_address = encode_email_address(stream)
|
|
data = [
|
|
dict(
|
|
msg_base64=base64.b64encode(b"\xf3test").decode(),
|
|
time=time.time(),
|
|
rcpt_to=stream_to_address,
|
|
),
|
|
] * 3
|
|
for element in data:
|
|
fake_client.enqueue("email_mirror", element)
|
|
|
|
with simulated_queue_client(fake_client):
|
|
worker = queue_processors.MirrorWorker()
|
|
worker.setup()
|
|
worker.start()
|
|
|
|
self.assertEqual(mock_mirror_email.call_count, 3)
|
|
|
|
@patch("zerver.worker.queue_processors.mirror_email")
|
|
@override_settings(RATE_LIMITING_MIRROR_REALM_RULES=[(10, 2)])
|
|
def test_mirror_worker_rate_limiting(self, mock_mirror_email: MagicMock) -> None:
|
|
fake_client = FakeClient()
|
|
realm = get_realm("zulip")
|
|
RateLimitedRealmMirror(realm).clear_history()
|
|
stream = get_stream("Denmark", realm)
|
|
stream_to_address = encode_email_address(stream)
|
|
data = [
|
|
dict(
|
|
msg_base64=base64.b64encode(b"\xf3test").decode(),
|
|
time=time.time(),
|
|
rcpt_to=stream_to_address,
|
|
),
|
|
] * 5
|
|
for element in data:
|
|
fake_client.enqueue("email_mirror", element)
|
|
|
|
with simulated_queue_client(fake_client), self.assertLogs(
|
|
"zerver.worker.queue_processors", level="WARNING"
|
|
) as warn_logs:
|
|
start_time = time.time()
|
|
with patch("time.time", return_value=start_time):
|
|
worker = queue_processors.MirrorWorker()
|
|
worker.setup()
|
|
worker.start()
|
|
# Of the first 5 messages, only 2 should be processed
|
|
# (the rest being rate-limited):
|
|
self.assertEqual(mock_mirror_email.call_count, 2)
|
|
|
|
# If a new message is sent into the stream mirror, it will get rejected:
|
|
fake_client.enqueue("email_mirror", data[0])
|
|
worker.start()
|
|
self.assertEqual(mock_mirror_email.call_count, 2)
|
|
|
|
# However, message notification emails don't get rate limited:
|
|
with self.settings(EMAIL_GATEWAY_PATTERN="%s@example.com"):
|
|
address = "mm" + ("x" * 32) + "@example.com"
|
|
event = dict(
|
|
msg_base64=base64.b64encode(b"\xf3test").decode(),
|
|
time=time.time(),
|
|
rcpt_to=address,
|
|
)
|
|
fake_client.enqueue("email_mirror", event)
|
|
worker.start()
|
|
self.assertEqual(mock_mirror_email.call_count, 3)
|
|
|
|
# After some time passes, emails get accepted again:
|
|
with patch("time.time", return_value=(start_time + 11.0)):
|
|
fake_client.enqueue("email_mirror", data[0])
|
|
worker.start()
|
|
self.assertEqual(mock_mirror_email.call_count, 4)
|
|
|
|
# If RateLimiterLockingError is thrown, we rate-limit the new message:
|
|
with patch(
|
|
"zerver.lib.rate_limiter.RedisRateLimiterBackend.incr_ratelimit",
|
|
side_effect=RateLimiterLockingError,
|
|
):
|
|
with self.assertLogs("zerver.lib.rate_limiter", "WARNING") as mock_warn:
|
|
fake_client.enqueue("email_mirror", data[0])
|
|
worker.start()
|
|
self.assertEqual(mock_mirror_email.call_count, 4)
|
|
self.assertEqual(
|
|
mock_warn.output,
|
|
[
|
|
"WARNING:zerver.lib.rate_limiter:Deadlock trying to incr_ratelimit for RateLimitedRealmMirror:zulip"
|
|
],
|
|
)
|
|
self.assertEqual(
|
|
warn_logs.output,
|
|
[
|
|
"WARNING:zerver.worker.queue_processors:MirrorWorker: Rejecting an email from: None to realm: zulip - rate limited."
|
|
]
|
|
* 5,
|
|
)
|
|
|
|
def test_email_sending_worker_retries(self) -> None:
|
|
"""Tests the retry_send_email_failures decorator to make sure it
|
|
retries sending the email 3 times and then gives up."""
|
|
fake_client = FakeClient()
|
|
|
|
data = {
|
|
"template_prefix": "zerver/emails/confirm_new_email",
|
|
"to_emails": [self.example_email("hamlet")],
|
|
"from_name": "Zulip Account Security",
|
|
"from_address": FromAddress.NOREPLY,
|
|
"context": {},
|
|
}
|
|
fake_client.enqueue("email_senders", data)
|
|
|
|
def fake_publish(
|
|
queue_name: str, event: Dict[str, Any], processor: Optional[Callable[[Any], None]]
|
|
) -> None:
|
|
fake_client.enqueue(queue_name, event)
|
|
|
|
with simulated_queue_client(fake_client):
|
|
worker = queue_processors.EmailSendingWorker()
|
|
worker.setup()
|
|
with patch(
|
|
"zerver.lib.send_email.build_email", side_effect=EmailNotDeliveredError
|
|
), mock_queue_publish(
|
|
"zerver.lib.queue.queue_json_publish", side_effect=fake_publish
|
|
), self.assertLogs(
|
|
level="ERROR"
|
|
) as m:
|
|
worker.start()
|
|
self.assertIn("failed due to exception EmailNotDeliveredError", m.output[0])
|
|
|
|
self.assertEqual(data["failed_tries"], 1 + MAX_REQUEST_RETRIES)
|
|
|
|
def test_invites_worker(self) -> None:
|
|
fake_client = FakeClient()
|
|
inviter = self.example_user("iago")
|
|
prereg_alice = PreregistrationUser.objects.create(
|
|
email=self.nonreg_email("alice"), referred_by=inviter, realm=inviter.realm
|
|
)
|
|
PreregistrationUser.objects.create(
|
|
email=self.nonreg_email("bob"), referred_by=inviter, realm=inviter.realm
|
|
)
|
|
invite_expires_in_minutes = 4 * 24 * 60
|
|
data: List[Dict[str, Any]] = [
|
|
dict(
|
|
prereg_id=prereg_alice.id,
|
|
referrer_id=inviter.id,
|
|
invite_expires_in_minutes=invite_expires_in_minutes,
|
|
),
|
|
dict(
|
|
prereg_id=prereg_alice.id,
|
|
referrer_id=inviter.id,
|
|
email_language="en",
|
|
invite_expires_in_minutes=invite_expires_in_minutes,
|
|
),
|
|
# Nonexistent prereg_id, as if the invitation was deleted
|
|
dict(
|
|
prereg_id=-1,
|
|
referrer_id=inviter.id,
|
|
invite_expires_in_minutes=invite_expires_in_minutes,
|
|
),
|
|
]
|
|
for element in data:
|
|
fake_client.enqueue("invites", element)
|
|
|
|
with simulated_queue_client(fake_client):
|
|
worker = queue_processors.ConfirmationEmailWorker()
|
|
worker.setup()
|
|
with patch("zerver.actions.user_settings.send_email"), patch(
|
|
"zerver.worker.queue_processors.send_future_email"
|
|
) as send_mock:
|
|
worker.start()
|
|
self.assertEqual(send_mock.call_count, 2)
|
|
|
|
def test_error_handling(self) -> None:
|
|
processed = []
|
|
|
|
@queue_processors.assign_queue("unreliable_worker", is_test_queue=True)
|
|
class UnreliableWorker(queue_processors.QueueProcessingWorker):
|
|
def consume(self, data: Mapping[str, Any]) -> None:
|
|
if data["type"] == "unexpected behaviour":
|
|
raise Exception("Worker task not performing as expected!")
|
|
processed.append(data["type"])
|
|
|
|
fake_client = FakeClient()
|
|
for msg in ["good", "fine", "unexpected behaviour", "back to normal"]:
|
|
fake_client.enqueue("unreliable_worker", {"type": msg})
|
|
|
|
fn = os.path.join(settings.QUEUE_ERROR_DIR, "unreliable_worker.errors")
|
|
with suppress(FileNotFoundError):
|
|
os.remove(fn)
|
|
|
|
with simulated_queue_client(fake_client):
|
|
worker = UnreliableWorker()
|
|
worker.setup()
|
|
with self.assertLogs(level="ERROR") as m:
|
|
worker.start()
|
|
self.assertEqual(
|
|
m.records[0].message, "Problem handling data on queue unreliable_worker"
|
|
)
|
|
self.assertIn(m.records[0].stack_info, m.output[0])
|
|
|
|
self.assertEqual(processed, ["good", "fine", "back to normal"])
|
|
with open(fn) as f:
|
|
line = f.readline().strip()
|
|
events = orjson.loads(line.split("\t")[1])
|
|
self.assert_length(events, 1)
|
|
event = events[0]
|
|
self.assertEqual(event["type"], "unexpected behaviour")
|
|
|
|
processed = []
|
|
|
|
@queue_processors.assign_queue("unreliable_loopworker", is_test_queue=True)
|
|
class UnreliableLoopWorker(queue_processors.LoopQueueProcessingWorker):
|
|
def consume_batch(self, events: List[Dict[str, Any]]) -> None:
|
|
for event in events:
|
|
if event["type"] == "unexpected behaviour":
|
|
raise Exception("Worker task not performing as expected!")
|
|
processed.append(event["type"])
|
|
|
|
for msg in ["good", "fine", "unexpected behaviour", "back to normal"]:
|
|
fake_client.enqueue("unreliable_loopworker", {"type": msg})
|
|
|
|
fn = os.path.join(settings.QUEUE_ERROR_DIR, "unreliable_loopworker.errors")
|
|
with suppress(FileNotFoundError):
|
|
os.remove(fn)
|
|
|
|
with simulated_queue_client(fake_client):
|
|
loopworker = UnreliableLoopWorker()
|
|
loopworker.setup()
|
|
with self.assertLogs(level="ERROR") as m:
|
|
loopworker.start()
|
|
self.assertEqual(
|
|
m.records[0].message, "Problem handling data on queue unreliable_loopworker"
|
|
)
|
|
self.assertIn(m.records[0].stack_info, m.output[0])
|
|
|
|
self.assertEqual(processed, ["good", "fine"])
|
|
with open(fn) as f:
|
|
line = f.readline().strip()
|
|
events = orjson.loads(line.split("\t")[1])
|
|
self.assert_length(events, 4)
|
|
|
|
self.assertEqual(
|
|
[event["type"] for event in events],
|
|
["good", "fine", "unexpected behaviour", "back to normal"],
|
|
)
|
|
|
|
def test_timeouts(self) -> None:
|
|
processed = []
|
|
|
|
@queue_processors.assign_queue("timeout_worker", is_test_queue=True)
|
|
class TimeoutWorker(queue_processors.QueueProcessingWorker):
|
|
MAX_CONSUME_SECONDS = 1
|
|
|
|
def consume(self, data: Mapping[str, Any]) -> None:
|
|
if data["type"] == "timeout":
|
|
time.sleep(5)
|
|
processed.append(data["type"])
|
|
|
|
fake_client = FakeClient()
|
|
for msg in ["good", "fine", "timeout", "back to normal"]:
|
|
fake_client.enqueue("timeout_worker", {"type": msg})
|
|
|
|
fn = os.path.join(settings.QUEUE_ERROR_DIR, "timeout_worker.errors")
|
|
with suppress(FileNotFoundError):
|
|
os.remove(fn)
|
|
|
|
with simulated_queue_client(fake_client):
|
|
worker = TimeoutWorker()
|
|
worker.setup()
|
|
worker.ENABLE_TIMEOUTS = True
|
|
with self.assertLogs(level="ERROR") as m:
|
|
worker.start()
|
|
self.assertEqual(
|
|
m.records[0].message,
|
|
"Timed out in timeout_worker after 1 seconds processing 1 events",
|
|
)
|
|
self.assertIn(m.records[0].stack_info, m.output[0])
|
|
|
|
self.assertEqual(processed, ["good", "fine", "back to normal"])
|
|
with open(fn) as f:
|
|
line = f.readline().strip()
|
|
events = orjson.loads(line.split("\t")[1])
|
|
self.assert_length(events, 1)
|
|
event = events[0]
|
|
self.assertEqual(event["type"], "timeout")
|
|
|
|
def test_embed_links_timeout(self) -> None:
|
|
@queue_processors.assign_queue("timeout_worker", is_test_queue=True)
|
|
class TimeoutWorker(FetchLinksEmbedData):
|
|
MAX_CONSUME_SECONDS = 1
|
|
|
|
def consume(self, data: Mapping[str, Any]) -> None:
|
|
# Send SIGALRM to ourselves to simulate a timeout.
|
|
pid = os.getpid()
|
|
os.kill(pid, signal.SIGALRM)
|
|
|
|
fake_client = FakeClient()
|
|
fake_client.enqueue(
|
|
"timeout_worker",
|
|
{
|
|
"type": "timeout",
|
|
"message_id": 15,
|
|
"urls": ["first", "second"],
|
|
},
|
|
)
|
|
|
|
with simulated_queue_client(fake_client):
|
|
worker = TimeoutWorker()
|
|
worker.setup()
|
|
worker.ENABLE_TIMEOUTS = True
|
|
with self.assertLogs(level="WARNING") as m:
|
|
worker.start()
|
|
self.assertEqual(
|
|
m.records[0].message,
|
|
"Timed out in timeout_worker after 1 seconds while fetching URLs for message 15: ['first', 'second']",
|
|
)
|
|
|
|
def test_worker_noname(self) -> None:
|
|
class TestWorker(queue_processors.QueueProcessingWorker):
|
|
def __init__(self) -> None:
|
|
super().__init__()
|
|
|
|
def consume(self, data: Mapping[str, Any]) -> None:
|
|
pass # nocoverage # this is intentionally not called
|
|
|
|
with self.assertRaises(queue_processors.WorkerDeclarationError):
|
|
TestWorker()
|
|
|
|
def test_get_active_worker_queues(self) -> None:
|
|
test_queue_names = set(get_active_worker_queues(only_test_queues=True))
|
|
worker_queue_names = {
|
|
queue_class.queue_name
|
|
for base in [QueueProcessingWorker, EmailSendingWorker, LoopQueueProcessingWorker]
|
|
for queue_class in base.__subclasses__()
|
|
if not isabstract(queue_class)
|
|
}
|
|
|
|
# Verify that the set of active worker queues equals the set
|
|
# of subclasses without is_test_queue set.
|
|
self.assertEqual(set(get_active_worker_queues()), worker_queue_names - test_queue_names)
|