def use_backend(backend_name):
from kronos.storage.router import router
# Use only `backend_name` for both reads and writes.
for namespace in settings.namespace_to_streams_configuration:
settings.namespace_to_streams_configuration[namespace] = {
'': {
'backends': {
backend_name: None
},
'read_backend': backend_name
}
}
router.reload()
def use_backend(backend_name):
from kronos.storage.router import router
# Use only `backend_name` for both reads and writes.
for namespace in settings.namespace_to_streams_configuration:
settings.namespace_to_streams_configuration[namespace] = {
'': {
'backends': {
backend_name: None
},
'read_backend': backend_name
}
}
router.reload()
def test_overlapping_shards(self):
""" Tests that changing bucket widths doesn\'t break shit. First inserts
events into two contiguous shards when the bucket width is 2 seconds. Then
it changes bucket width to 4 and inserts the same events again, causing
all events to go into a single bucket. At the end it checks if the first
bucket has 3x the number of events as the seconds bucket. """
stream_name = "TestCassandraBackend_test_overlapping_shards"
stream = self.namespace.get_stream(stream_name, self.width, self.shards)
for i in xrange(60):
self.put(stream_name, [{TIMESTAMP_FIELD: epoch_time_to_kronos_time(i % 4)}])
events = self.get(stream_name, 0, epoch_time_to_kronos_time(4))
self.assertEqual(len(events), 60)
# Change default width to be 4 seconds instead of 2.
settings.storage.cassandra.timewidth_seconds = 4
router.reload()
# All of these events should now go into bucket with start time 0.
for i in xrange(60):
self.put(stream_name, [{TIMESTAMP_FIELD: epoch_time_to_kronos_time(i % 4)}])
events = self.get(stream_name, 0, epoch_time_to_kronos_time(4))
self.assertEqual(len(events), 120)
shard_to_events = defaultdict(int)
for start_time in (0, 2):
# Fetch events directly from each shard.
for shard in xrange(self.shards):
stream_shard = StreamShard(
stream.namespace,
stream_name,
epoch_time_to_kronos_time(start_time),
self.width,
shard,
False,
MAX_LIMIT,
100,
)
events = stream_shard.iterator(
uuid_from_time(start_time, UUIDType.LOWEST), uuid_from_time(start_time + 4)
)
shard_to_events[start_time] += len(list(events))
self.assertEqual(len(shard_to_events), 2)
self.assertEqual(shard_to_events[0], 90) # 30 + 60.
self.assertEqual(shard_to_events[2], 30) # 30 + 0.
self.assertEqual(sum(shard_to_events.itervalues()), 120)
# Revert default width settings.
settings.storage.cassandra.timewidth_seconds = 2
router.reload()
def test_overlapping_shards(self):
''' Tests that changing bucket widths doesn\'t break shit. First inserts
events into two contiguous shards when the bucket width is 2 seconds. Then
it changes bucket width to 4 and inserts the same events again, causing
all events to go into a single bucket. At the end it checks if the first
bucket has 3x the number of events as the seconds bucket. '''
stream_name = 'TestCassandraBackend_test_overlapping_shards'
stream = self.namespace.get_stream(stream_name, self.width, self.shards)
for i in xrange(60):
self.put(stream_name,
[{TIMESTAMP_FIELD: epoch_time_to_kronos_time(i % 4)}])
events = self.get(stream_name, 0, epoch_time_to_kronos_time(4))
self.assertEqual(len(events), 60)
# Change default width to be 4 seconds instead of 2.
settings.storage.cassandra.timewidth_seconds = 4
router.reload()
# All of these events should now go into bucket with start time 0.
for i in xrange(60):
self.put(stream_name, [{TIMESTAMP_FIELD:
epoch_time_to_kronos_time(i % 4)}])
events = self.get(stream_name, 0, epoch_time_to_kronos_time(4))
self.assertEqual(len(events), 120)
shard_to_events = defaultdict(int)
for start_time in (0, 2):
# Fetch events directly from each shard.
for shard in xrange(self.shards):
stream_shard = StreamShard(stream.namespace, stream_name,
epoch_time_to_kronos_time(start_time),
self.width, shard, False,
MAX_LIMIT, 100)
events = stream_shard.iterator(uuid_from_time(start_time,
UUIDType.LOWEST),
uuid_from_time(start_time + 4))
shard_to_events[start_time] += len(list(events))
self.assertEqual(len(shard_to_events), 2)
self.assertEqual(shard_to_events[0], 90) # 30 + 60.
self.assertEqual(shard_to_events[2], 30) # 30 + 0.
self.assertEqual(sum(shard_to_events.itervalues()), 120)
# Revert default width settings.
settings.storage.cassandra.timewidth_seconds = 2
router.reload()
def test_stream_splitting(self):
""" Inserts events that fall into multiple time buckets and ensures that
each time bucket has the right number of events at the end and all the
events in each bucket fall into its time range. """
stream_name = "TestCassandraBackend_test_stream_splitting"
stream = self.namespace.get_stream(stream_name, self.width, self.shards)
settings.storage.cassandra.timewidth_seconds = 2
router.reload()
# Each bucket interval has width of 2 seconds, so this should distribute
# events in 5 buckets: [0, 2), [2, 4), [4, 6), [6, 8), [8, 10).
for i in xrange(100):
self.put(stream_name, [{TIMESTAMP_FIELD: epoch_time_to_kronos_time(i % 10)}])
events = self.get(stream_name, 0, epoch_time_to_kronos_time(10))
self.assertEqual(len(events), 100)
bucket_to_events = defaultdict(list)
for start_time in (0, 2, 4, 6, 8):
# Fetch events directly from each bucket.
for shard in xrange(self.shards):
stream_shard = StreamShard(
stream.namespace,
stream_name,
epoch_time_to_kronos_time(start_time),
self.width,
shard,
False,
MAX_LIMIT,
100,
)
events = stream_shard.iterator(
uuid_from_time(start_time, UUIDType.LOWEST), uuid_from_time(start_time + self.width_seconds)
)
bucket_to_events[start_time].extend(marshal.loads(event.json) for event in events)
num_events = 0
for start_time, events in bucket_to_events.iteritems():
# Each bucket should have 20 events and they must fall in the bucket's
# time range.
self.assertEqual(len(events), 20)
for event in events:
time = kronos_time_to_epoch_time(event[TIMESTAMP_FIELD])
self.assertTrue(time >= start_time)
self.assertTrue(time < start_time + self.width_seconds)
num_events += len(events)
self.assertEqual(num_events, 100)
def test_stream_splitting(self):
''' Inserts events that fall into multiple time buckets and ensures that
each time bucket has the right number of events at the end and all the
events in each bucket fall into its time range. '''
stream_name = 'TestCassandraBackend_test_stream_splitting'
stream = self.namespace.get_stream(stream_name, self.width, self.shards)
settings.storage.cassandra.timewidth_seconds = 2
router.reload()
# Each bucket interval has width of 2 seconds, so this should distribute
# events in 5 buckets: [0, 2), [2, 4), [4, 6), [6, 8), [8, 10).
for i in xrange(100):
self.put(stream_name, [{TIMESTAMP_FIELD:
epoch_time_to_kronos_time(i % 10)}])
events = self.get(stream_name, 0, epoch_time_to_kronos_time(10))
self.assertEqual(len(events), 100)
bucket_to_events = defaultdict(list)
for start_time in (0, 2, 4, 6, 8):
# Fetch events directly from each bucket.
for shard in xrange(self.shards):
stream_shard = StreamShard(stream.namespace, stream_name,
epoch_time_to_kronos_time(start_time),
self.width, shard, False,
MAX_LIMIT, 100)
events = stream_shard.iterator(
uuid_from_time(start_time, UUIDType.LOWEST),
uuid_from_time(start_time + self.width_seconds))
bucket_to_events[start_time].extend(marshal.loads(event.json)
for event in events)
num_events = 0
for start_time, events in bucket_to_events.iteritems():
# Each bucket should have 20 events and they must fall in the bucket's
# time range.
self.assertEqual(len(events), 20)
for event in events:
time = kronos_time_to_epoch_time(event[TIMESTAMP_FIELD])
self.assertTrue(time >= start_time)
self.assertTrue(time < start_time + self.width_seconds)
num_events += len(events)
self.assertEqual(num_events, 100)
def reload_router(kill_update_thread=False):
router.reload()
backend = router.get_backend('elasticsearch')
if kill_update_thread:
backend.index_manager.kill_rollover_worker()
def reload_router(kill_update_thread=False):
router.reload()
backend = router.get_backend('elasticsearch')
if kill_update_thread:
backend.index_manager.kill_rollover_worker()
