def run_test(
self,
subscriptions: Collection[Subscription],
start: timedelta,
end: timedelta,
expected: Collection[ScheduledTask[Subscription]],
sort_key=None,
) -> None:
store = RedisSubscriptionDataStore(
redis_client,
self.dataset,
self.partition_id,
)
for subscription in subscriptions:
store.create(subscription.identifier.uuid, subscription.data)
scheduler = SubscriptionScheduler(
store,
self.partition_id,
timedelta(minutes=1),
DummyMetricsBackend(strict=True),
)
result = list(scheduler.find(self.build_interval(start, end)))
if sort_key:
result.sort(key=sort_key)
assert result == expected
def run_test(
self,
subscriptions: Collection[Subscription],
start: timedelta,
end: timedelta,
expected: Collection[ScheduledSubscriptionTask],
sort_key: Optional[Callable[[ScheduledSubscriptionTask],
Tuple[datetime, uuid.UUID]]] = None,
) -> None:
tick = self.build_tick(start, end)
store = RedisSubscriptionDataStore(
redis_client,
self.entity_key,
self.partition_id,
)
for subscription in subscriptions:
store.create(subscription.identifier.uuid, subscription.data)
scheduler = SubscriptionScheduler(
EntityKey.EVENTS,
store,
self.partition_id,
timedelta(minutes=1),
DummyMetricsBackend(strict=True),
)
result = list(scheduler.find(tick))
if sort_key:
result.sort(key=sort_key)
assert result == expected
def __init__(
self,
entity_key: EntityKey,
mode: SchedulingWatermarkMode,
schedule_ttl: int,
stale_threshold_seconds: Optional[int],
partitions: int,
producer: Producer[KafkaPayload],
scheduled_topic_spec: KafkaTopicSpec,
metrics: MetricsBackend,
) -> None:
self.__mode = mode
self.__stale_threshold_seconds = stale_threshold_seconds
self.__partitions = partitions
self.__producer = producer
self.__scheduled_topic_spec = scheduled_topic_spec
self.__metrics = metrics
self.__buffer_size = settings.SUBSCRIPTIONS_ENTITY_BUFFER_SIZE.get(
entity_key.value, settings.SUBSCRIPTIONS_DEFAULT_BUFFER_SIZE)
self.__schedulers = {
index: SubscriptionScheduler(
entity_key,
RedisSubscriptionDataStore(redis_client, entity_key,
PartitionId(index)),
partition_id=PartitionId(index),
cache_ttl=timedelta(seconds=schedule_ttl),
metrics=self.__metrics,
)
for index in range(self.__partitions)
}
def test_subscription_worker_consistent(
subscription_data: SubscriptionData) -> None:
state.set_config("event_subscription_non_consistent_sample_rate", 1)
broker: Broker[SubscriptionTaskResult] = Broker(MemoryMessageStorage(),
TestingClock())
result_topic = Topic("subscription-results")
broker.create_topic(result_topic, partitions=1)
frequency = timedelta(minutes=1)
evaluations = 1
subscription = Subscription(
SubscriptionIdentifier(PartitionId(0), uuid1()),
subscription_data,
)
store = DummySubscriptionDataStore()
store.create(subscription.identifier.uuid, subscription.data)
metrics = TestingMetricsBackend()
dataset = get_dataset("events")
worker = SubscriptionWorker(
dataset,
ThreadPoolExecutor(),
{
0:
SubscriptionScheduler(store, PartitionId(0), timedelta(),
DummyMetricsBackend(strict=True))
},
broker.get_producer(),
result_topic,
metrics,
)
now = datetime(2000, 1, 1)
tick = Tick(
offsets=Interval(0, 1),
timestamps=Interval(now - (frequency * evaluations), now),
)
worker.process_message(Message(Partition(Topic("events"), 0), 0, tick,
now))
time.sleep(0.1)
assert (len([
m for m in metrics.calls
if isinstance(m, Increment) and m.name == "consistent"
]) == 1)
def subscriptions(
*,
dataset_name: str,
topic: Optional[str],
partitions: Optional[int],
commit_log_topic: Optional[str],
commit_log_groups: Sequence[str],
consumer_group: str,
auto_offset_reset: str,
bootstrap_servers: Sequence[str],
max_batch_size: int,
max_batch_time_ms: int,
max_query_workers: Optional[int],
schedule_ttl: int,
result_topic: Optional[str],
log_level: Optional[str],
delay_seconds: Optional[int],
) -> None:
"""Evaluates subscribed queries for a dataset."""
setup_logging(log_level)
setup_sentry()
dataset = get_dataset(dataset_name)
storage = dataset.get_default_entity().get_writable_storage()
assert (
storage is not None
), f"Dataset {dataset_name} does not have a writable storage by default."
loader = enforce_table_writer(dataset).get_stream_loader()
commit_log_topic_spec = loader.get_commit_log_topic_spec()
assert commit_log_topic_spec is not None
result_topic_spec = loader.get_subscription_result_topic_spec()
assert result_topic_spec is not None
metrics = MetricsWrapper(
environment.metrics,
"subscriptions",
tags={
"group": consumer_group,
"dataset": dataset_name
},
)
consumer = TickConsumer(
SynchronizedConsumer(
KafkaConsumer(
build_kafka_consumer_configuration(
loader.get_default_topic_spec().topic,
consumer_group,
auto_offset_reset=auto_offset_reset,
bootstrap_servers=bootstrap_servers,
), ),
KafkaConsumer(
build_kafka_consumer_configuration(
commit_log_topic_spec.topic,
f"subscriptions-commit-log-{uuid.uuid1().hex}",
auto_offset_reset="earliest",
bootstrap_servers=bootstrap_servers,
), ),
(Topic(commit_log_topic) if commit_log_topic is not None else
Topic(commit_log_topic_spec.topic_name)),
set(commit_log_groups),
),
time_shift=(timedelta(seconds=delay_seconds *
-1) if delay_seconds is not None else None),
)
producer = ProducerEncodingWrapper(
KafkaProducer(
build_kafka_producer_configuration(
loader.get_default_topic_spec().topic,
bootstrap_servers=bootstrap_servers,
override_params={
"partitioner": "consistent",
"message.max.bytes": 50000000, # 50MB, default is 1MB
},
)),
SubscriptionTaskResultEncoder(),
)
executor = ThreadPoolExecutor(max_workers=max_query_workers)
logger.debug("Starting %r with %s workers...", executor,
getattr(executor, "_max_workers", 0))
metrics.gauge("executor.workers", getattr(executor, "_max_workers", 0))
with closing(consumer), executor, closing(producer):
from arroyo import configure_metrics
configure_metrics(StreamMetricsAdapter(metrics))
batching_consumer = StreamProcessor(
consumer,
(Topic(topic) if topic is not None else Topic(
loader.get_default_topic_spec().topic_name)),
BatchProcessingStrategyFactory(
SubscriptionWorker(
dataset,
executor,
{
index: SubscriptionScheduler(
RedisSubscriptionDataStore(redis_client, dataset,
PartitionId(index)),
PartitionId(index),
cache_ttl=timedelta(seconds=schedule_ttl),
metrics=metrics,
)
for index in
range(partitions if partitions is not None else loader.
get_default_topic_spec().partitions_number)
},
producer,
Topic(result_topic) if result_topic is not None else Topic(
result_topic_spec.topic_name),
metrics,
),
max_batch_size,
max_batch_time_ms,
),
)
def handler(signum: int, frame: Optional[Any]) -> None:
batching_consumer.signal_shutdown()
signal.signal(signal.SIGINT, handler)
signal.signal(signal.SIGTERM, handler)
batching_consumer.run()
def subscriptions(
*,
dataset_name: str,
topic: Optional[str],
partitions: Optional[int],
commit_log_topic: Optional[str],
commit_log_groups: Sequence[str],
consumer_group: str,
auto_offset_reset: str,
bootstrap_servers: Sequence[str],
max_batch_size: int,
max_batch_time_ms: int,
schedule_ttl: int,
result_topic: Optional[str],
log_level: Optional[str],
) -> None:
"""Evaluates subscribed queries for a dataset."""
assert result_topic is not None
setup_logging(log_level)
setup_sentry()
dataset = get_dataset(dataset_name)
if not bootstrap_servers:
bootstrap_servers = settings.DEFAULT_DATASET_BROKERS.get(
dataset_name, settings.DEFAULT_BROKERS
)
loader = enforce_table_writer(dataset).get_stream_loader()
consumer = TickConsumer(
SynchronizedConsumer(
KafkaConsumer(
build_kafka_consumer_configuration(
bootstrap_servers,
consumer_group,
auto_offset_reset=auto_offset_reset,
),
PassthroughCodec(),
),
KafkaConsumer(
build_kafka_consumer_configuration(
bootstrap_servers,
f"subscriptions-commit-log-{uuid.uuid1().hex}",
auto_offset_reset="earliest",
),
CommitCodec(),
),
(
Topic(commit_log_topic)
if commit_log_topic is not None
else Topic(loader.get_commit_log_topic_spec().topic_name)
),
set(commit_log_groups),
)
)
producer = KafkaProducer(
{
"bootstrap.servers": ",".join(bootstrap_servers),
"partitioner": "consistent",
"message.max.bytes": 50000000, # 50MB, default is 1MB
},
SubscriptionResultCodec(),
)
with closing(consumer), closing(producer):
batching_consumer = BatchingConsumer(
consumer,
(
Topic(topic)
if topic is not None
else Topic(loader.get_default_topic_spec().topic_name)
),
SubscriptionWorker(
SubscriptionExecutor(
dataset,
ThreadPoolExecutor(
max_workers=settings.SUBSCRIPTIONS_MAX_CONCURRENT_QUERIES
),
),
{
index: SubscriptionScheduler(
RedisSubscriptionDataStore(
redis_client, dataset, PartitionId(index)
),
PartitionId(index),
cache_ttl=timedelta(seconds=schedule_ttl),
)
for index in range(
partitions
if partitions is not None
else loader.get_default_topic_spec().partitions_number
)
},
producer,
Topic(result_topic),
),
max_batch_size,
max_batch_time_ms,
create_metrics(
"snuba.subscriptions",
tags={"group": consumer_group, "dataset": dataset_name},
),
)
def handler(signum, frame) -> None:
batching_consumer.signal_shutdown()
signal.signal(signal.SIGINT, handler)
signal.signal(signal.SIGTERM, handler)
batching_consumer.run()
def __init__(
self,
dataset: Dataset,
entity_names: Sequence[str],
partitions: int,
max_concurrent_queries: int,
total_concurrent_queries: int,
producer: Producer[KafkaPayload],
metrics: MetricsBackend,
stale_threshold_seconds: Optional[int],
result_topic: str,
schedule_ttl: int,
scheduling_mode: Optional[SchedulingWatermarkMode] = None,
) -> None:
# TODO: self.__partitions might not be the same for each entity
self.__partitions = partitions
self.__entity_names = entity_names
self.__metrics = metrics
entity_keys = [EntityKey(entity_name) for entity_name in self.__entity_names]
self.__schedulers = [
{
index: SubscriptionScheduler(
entity_key,
RedisSubscriptionDataStore(
redis_client, entity_key, PartitionId(index)
),
partition_id=PartitionId(index),
cache_ttl=timedelta(seconds=schedule_ttl),
metrics=self.__metrics,
)
for index in range(self.__partitions)
}
for entity_key in entity_keys
]
# Just apply the max buffer size if they are configured differently
# for each entity that is being run together
self.__buffer_size = max(
[
settings.SUBSCRIPTIONS_ENTITY_BUFFER_SIZE.get(
entity_key.value, settings.SUBSCRIPTIONS_DEFAULT_BUFFER_SIZE
)
for entity_key in entity_keys
]
)
self.__executor_factory = SubscriptionExecutorProcessingFactory(
max_concurrent_queries,
total_concurrent_queries,
dataset,
entity_names,
producer,
metrics,
stale_threshold_seconds,
result_topic,
)
if scheduling_mode is not None:
self.__mode = scheduling_mode
else:
modes = {
self._get_entity_watermark_mode(entity_key)
for entity_key in entity_keys
}
mode = modes.pop()
assert len(modes) == 0, "Entities provided do not share the same mode"
self.__mode = mode
def test_subscription_worker(broker: Broker[SubscriptionTaskResult], ) -> None:
result_topic = Topic("subscription-results")
broker.create_topic(result_topic, partitions=1)
frequency = timedelta(minutes=1)
evaluations = 3
subscription = Subscription(
SubscriptionIdentifier(PartitionId(0), uuid1()),
SubscriptionData(
project_id=1,
conditions=[],
aggregations=[["count()", "", "count"]],
time_window=timedelta(minutes=60),
resolution=frequency,
),
)
store = DummySubscriptionDataStore()
store.create(subscription.identifier.uuid, subscription.data)
metrics = DummyMetricsBackend(strict=True)
dataset = get_dataset("events")
worker = SubscriptionWorker(
dataset,
ThreadPoolExecutor(),
{
0: SubscriptionScheduler(store, PartitionId(0), timedelta(),
metrics)
},
broker.get_producer(),
result_topic,
metrics,
)
now = datetime(2000, 1, 1)
tick = Tick(
offsets=Interval(0, 1),
timestamps=Interval(now - (frequency * evaluations), now),
)
result_futures = worker.process_message(
Message(Partition(Topic("events"), 0), 0, tick, now))
assert result_futures is not None and len(result_futures) == evaluations
# Publish the results.
worker.flush_batch([result_futures])
# Check to make sure the results were published.
# NOTE: This does not cover the ``SubscriptionTaskResultCodec``!
consumer = broker.get_consumer("group")
consumer.subscribe([result_topic])
for i in range(evaluations):
timestamp = now - frequency * (evaluations - i)
message = consumer.poll()
assert message is not None
assert message.partition.topic == result_topic
task, future = result_futures[i]
future_result = request, result = future.result()
assert message.payload.task.timestamp == timestamp
assert message.payload == SubscriptionTaskResult(task, future_result)
# NOTE: The time series extension is folded back into the request
# body, ideally this would reference the timeseries options in
# isolation.
assert (request.body.items() > {
"from_date":
(timestamp - subscription.data.time_window).isoformat(),
"to_date":
timestamp.isoformat(),
}.items())
assert result == {
"meta": [{
"name": "count",
"type": "UInt64"
}],
"data": [{
"count": 0
}],
}
def test_subscription_worker(subscription_data: SubscriptionData) -> None:
broker: Broker[SubscriptionTaskResult] = Broker(MemoryMessageStorage(),
TestingClock())
result_topic = Topic("subscription-results")
broker.create_topic(result_topic, partitions=1)
frequency = timedelta(minutes=1)
evaluations = 3
subscription = Subscription(
SubscriptionIdentifier(PartitionId(0), uuid1()),
subscription_data,
)
store = DummySubscriptionDataStore()
store.create(subscription.identifier.uuid, subscription.data)
metrics = DummyMetricsBackend(strict=True)
dataset = get_dataset("events")
worker = SubscriptionWorker(
dataset,
ThreadPoolExecutor(),
{
0: SubscriptionScheduler(store, PartitionId(0), timedelta(),
metrics)
},
broker.get_producer(),
result_topic,
metrics,
)
now = datetime(2000, 1, 1)
tick = Tick(
offsets=Interval(0, 1),
timestamps=Interval(now - (frequency * evaluations), now),
)
result_futures = worker.process_message(
Message(Partition(Topic("events"), 0), 0, tick, now))
assert result_futures is not None and len(result_futures) == evaluations
# Publish the results.
worker.flush_batch([result_futures])
# Check to make sure the results were published.
# NOTE: This does not cover the ``SubscriptionTaskResultCodec``!
consumer = broker.get_consumer("group")
consumer.subscribe([result_topic])
for i in range(evaluations):
timestamp = now - frequency * (evaluations - i)
message = consumer.poll()
assert message is not None
assert message.partition.topic == result_topic
task, future = result_futures[i]
future_result = request, result = future.result()
assert message.payload.task.timestamp == timestamp
assert message.payload == SubscriptionTaskResult(task, future_result)
# NOTE: The time series extension is folded back into the request
# body, ideally this would reference the timeseries options in
# isolation.
from_pattern = FunctionCall(
String(ConditionFunctions.GTE),
(
Column(None, String("timestamp")),
Literal(Datetime(timestamp - subscription.data.time_window)),
),
)
to_pattern = FunctionCall(
String(ConditionFunctions.LT),
(Column(None, String("timestamp")), Literal(Datetime(timestamp))),
)
condition = request.query.get_condition()
assert condition is not None
conditions = get_first_level_and_conditions(condition)
assert any([from_pattern.match(e) for e in conditions])
assert any([to_pattern.match(e) for e in conditions])
assert result == {
"meta": [{
"name": "count",
"type": "UInt64"
}],
"data": [{
"count": 0
}],
}
