mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-06 15:42:39 +00:00
b75963d370
This PR formats all the `*.py` files found under the `tests/integration` folder. It also reorders the imports and cleans up a bunch of unused imports. The formatting also takes care of other things like wrapping lines and fixing spaces and indents such that the tests look more readable.
86 lines
2.6 KiB
Python
86 lines
2.6 KiB
Python
# pylint: disable=unused-argument
|
|
# pylint: disable=redefined-outer-name
|
|
# pylint: disable=line-too-long
|
|
|
|
import pytest
|
|
|
|
from helpers.cluster import ClickHouseCluster
|
|
|
|
cluster = ClickHouseCluster(__file__)
|
|
|
|
n1 = cluster.add_instance('n1', main_configs=['configs/remote_servers.xml'])
|
|
n2 = cluster.add_instance('n2', main_configs=['configs/remote_servers.xml'])
|
|
|
|
params = pytest.mark.parametrize('cluster,q', [
|
|
('internal_replication', 0),
|
|
('no_internal_replication', 1),
|
|
])
|
|
|
|
|
|
@pytest.fixture(scope='module', autouse=True)
|
|
def start_cluster():
|
|
try:
|
|
cluster.start()
|
|
yield cluster
|
|
finally:
|
|
cluster.shutdown()
|
|
|
|
|
|
def create_tables(cluster):
|
|
n1.query('DROP TABLE IF EXISTS data')
|
|
n2.query('DROP TABLE IF EXISTS data')
|
|
n1.query('DROP TABLE IF EXISTS dist')
|
|
|
|
n1.query('CREATE TABLE data (key Int) Engine=Memory()')
|
|
n2.query('CREATE TABLE data (key Int) Engine=Memory()')
|
|
n1.query("""
|
|
CREATE TABLE dist AS data
|
|
Engine=Distributed(
|
|
{cluster},
|
|
currentDatabase(),
|
|
data,
|
|
rand()
|
|
)
|
|
""".format(cluster=cluster))
|
|
|
|
|
|
def insert_data(cluster, **settings):
|
|
create_tables(cluster)
|
|
n1.query('INSERT INTO dist SELECT * FROM numbers(10)', settings=settings)
|
|
n1.query('SYSTEM FLUSH DISTRIBUTED dist')
|
|
|
|
|
|
@params
|
|
def test_prefer_localhost_replica_1(cluster, q):
|
|
insert_data(cluster)
|
|
assert int(n1.query('SELECT count() FROM data')) == 10
|
|
assert int(n2.query('SELECT count() FROM data')) == 10 * q
|
|
|
|
|
|
@params
|
|
def test_prefer_localhost_replica_1_load_balancing_in_order(cluster, q):
|
|
insert_data(cluster, load_balancing='in_order')
|
|
assert int(n1.query('SELECT count() FROM data')) == 10
|
|
assert int(n2.query('SELECT count() FROM data')) == 10 * q
|
|
|
|
|
|
@params
|
|
def test_prefer_localhost_replica_0_load_balancing_nearest_hostname(cluster, q):
|
|
insert_data(cluster, load_balancing='nearest_hostname', prefer_localhost_replica=0)
|
|
assert int(n1.query('SELECT count() FROM data')) == 10
|
|
assert int(n2.query('SELECT count() FROM data')) == 10 * q
|
|
|
|
|
|
@params
|
|
def test_prefer_localhost_replica_0_load_balancing_in_order(cluster, q):
|
|
insert_data(cluster, load_balancing='in_order', prefer_localhost_replica=0)
|
|
assert int(n1.query('SELECT count() FROM data')) == 10 * q
|
|
assert int(n2.query('SELECT count() FROM data')) == 10
|
|
|
|
|
|
@params
|
|
def test_prefer_localhost_replica_0_load_balancing_in_order_sync(cluster, q):
|
|
insert_data(cluster, load_balancing='in_order', prefer_localhost_replica=0, insert_distributed_sync=1)
|
|
assert int(n1.query('SELECT count() FROM data')) == 10 * q
|
|
assert int(n2.query('SELECT count() FROM data')) == 10
|