ClickHouse/tests/integration/test_adaptive_granularity/test.py
bharatnc b75963d370 Format and optimize imports in integration test files
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.
2020-09-15 23:10:54 -07:00

409 lines
19 KiB
Python

import time
import pytest
from helpers.client import QueryRuntimeException, QueryTimeoutExceedException
from helpers.cluster import ClickHouseCluster
from helpers.test_tools import assert_eq_with_retry
cluster = ClickHouseCluster(__file__)
node1 = cluster.add_instance('node1', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True)
node2 = cluster.add_instance('node2', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True)
node3 = cluster.add_instance('node3', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True, image='yandex/clickhouse-server', tag='19.6.3.18',
with_installed_binary=True)
node4 = cluster.add_instance('node4', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True)
node5 = cluster.add_instance('node5', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True, image='yandex/clickhouse-server', tag='19.1.15',
with_installed_binary=True)
node6 = cluster.add_instance('node6', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True)
node7 = cluster.add_instance('node7', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True, image='yandex/clickhouse-server', tag='19.6.3.18', stay_alive=True,
with_installed_binary=True)
node8 = cluster.add_instance('node8', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True, image='yandex/clickhouse-server', tag='19.1.15', stay_alive=True,
with_installed_binary=True)
node9 = cluster.add_instance('node9', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml',
'configs/merge_tree_settings.xml'], with_zookeeper=True,
image='yandex/clickhouse-server', tag='19.1.15', stay_alive=True,
with_installed_binary=True)
node10 = cluster.add_instance('node10', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml',
'configs/merge_tree_settings.xml'], with_zookeeper=True,
image='yandex/clickhouse-server', tag='19.6.3.18', stay_alive=True,
with_installed_binary=True)
node11 = cluster.add_instance('node11', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True, image='yandex/clickhouse-server', tag='19.1.15', stay_alive=True,
with_installed_binary=True)
node12 = cluster.add_instance('node12', main_configs=['configs/remote_servers.xml', 'configs/log_conf.xml'],
with_zookeeper=True, image='yandex/clickhouse-server', tag='19.1.15', stay_alive=True,
with_installed_binary=True)
def prepare_single_pair_with_setting(first_node, second_node, group):
for node in (first_node, second_node):
node.query("CREATE DATABASE IF NOT EXISTS test")
# Two tables with adaptive granularity
first_node.query(
'''
CREATE TABLE table_by_default(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_by_default', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 10485760
'''.format(g=group))
second_node.query(
'''
CREATE TABLE table_by_default(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_by_default', '2')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 10485760
'''.format(g=group))
# Two tables with fixed granularity
first_node.query(
'''
CREATE TABLE table_with_fixed_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_with_fixed_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 0
'''.format(g=group))
second_node.query(
'''
CREATE TABLE table_with_fixed_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_with_fixed_granularity', '2')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 0
'''.format(g=group))
# Two tables with different granularity
with pytest.raises(QueryRuntimeException):
first_node.query(
'''
CREATE TABLE table_with_different_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_with_different_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 10485760
'''.format(g=group))
second_node.query(
'''
CREATE TABLE table_with_different_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_with_different_granularity', '2')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 0
'''.format(g=group))
# Two tables with different granularity, but enabled mixed parts
first_node.query(
'''
CREATE TABLE table_with_mixed_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_with_mixed_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 10485760, enable_mixed_granularity_parts=1
'''.format(g=group))
second_node.query(
'''
CREATE TABLE table_with_mixed_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_with_mixed_granularity', '2')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 0, enable_mixed_granularity_parts=1
'''.format(g=group))
def prepare_single_pair_without_setting(first_node, second_node, group):
for node in (first_node, second_node):
node.query("CREATE DATABASE IF NOT EXISTS test")
# Two tables with fixed granularity
first_node.query(
'''
CREATE TABLE table_with_fixed_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_with_fixed_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
'''.format(g=group))
second_node.query(
'''
CREATE TABLE table_with_fixed_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/{g}/table_with_fixed_granularity', '2')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes = 0
'''.format(g=group))
@pytest.fixture(scope="module")
def start_static_cluster():
try:
cluster.start()
prepare_single_pair_with_setting(node1, node2, "shard1")
prepare_single_pair_with_setting(node3, node4, "shard2")
prepare_single_pair_without_setting(node5, node6, "shard3")
yield cluster
finally:
cluster.shutdown()
@pytest.mark.parametrize(
('first_node', 'second_node', 'table'),
[
(node1, node2, 'table_by_default'),
(node1, node2, 'table_with_fixed_granularity'),
(node3, node4, 'table_by_default'),
(node3, node4, 'table_with_fixed_granularity'),
(node5, node6, 'table_with_fixed_granularity'),
]
)
def test_different_versions_cluster(start_static_cluster, first_node, second_node, table):
counter = 1
for n1, n2 in ((first_node, second_node), (second_node, first_node)):
n1.query("INSERT INTO {tbl} VALUES (toDate('2018-10-01'), {c1}, 333), (toDate('2018-10-02'), {c2}, 444)".format(
tbl=table, c1=counter * 2, c2=counter * 2 + 1))
n2.query("SYSTEM SYNC REPLICA {tbl}".format(tbl=table))
assert_eq_with_retry(n2, "SELECT count() from {tbl}".format(tbl=table), str(counter * 2))
n1.query("DETACH TABLE {tbl}".format(tbl=table))
n2.query("DETACH TABLE {tbl}".format(tbl=table))
n1.query("ATTACH TABLE {tbl}".format(tbl=table))
n2.query("ATTACH TABLE {tbl}".format(tbl=table))
assert_eq_with_retry(n1, "SELECT count() from {tbl}".format(tbl=table), str(counter * 2))
assert_eq_with_retry(n2, "SELECT count() from {tbl}".format(tbl=table), str(counter * 2))
n1.query("OPTIMIZE TABLE {tbl} FINAL".format(tbl=table))
n2.query("SYSTEM SYNC REPLICA {tbl}".format(tbl=table))
assert_eq_with_retry(n1, "SELECT count() from {tbl}".format(tbl=table), str(counter * 2))
assert_eq_with_retry(n2, "SELECT count() from {tbl}".format(tbl=table), str(counter * 2))
counter += 1
@pytest.fixture(scope="module")
def start_dynamic_cluster():
try:
cluster.start()
node7.query(
'''
CREATE TABLE table_with_default_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/7/table_with_default_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
''')
node7.query(
'''
CREATE TABLE table_with_adaptive_default_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/7/table_with_adaptive_default_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
SETTINGS index_granularity_bytes=10485760
''')
node8.query(
'''
CREATE TABLE table_with_default_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/8/table_with_default_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
''')
node9.query(
'''
CREATE TABLE table_with_default_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/9/table_with_default_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
''')
node10.query(
'''
CREATE TABLE table_with_default_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/10/table_with_default_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
''')
node11.query(
'''
CREATE TABLE table_with_default_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/shard11/table_with_default_granularity', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
''')
node12.query(
'''
CREATE TABLE table_with_default_granularity(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/shard11/table_with_default_granularity', '2')
PARTITION BY toYYYYMM(date)
ORDER BY id
''')
yield cluster
finally:
cluster.shutdown()
@pytest.mark.parametrize(
('n', 'tables'),
[
(node7, ['table_with_default_granularity', 'table_with_adaptive_default_granularity']),
(node8, ['table_with_default_granularity']),
]
)
def test_version_single_node_update(start_dynamic_cluster, n, tables):
for table in tables:
n.query(
"INSERT INTO {tbl} VALUES (toDate('2018-10-01'), 1, 333), (toDate('2018-10-02'), 2, 444)".format(tbl=table))
n.restart_with_latest_version()
for table in tables:
assert n.query("SELECT count() from {tbl}".format(tbl=table)) == '2\n'
n.query(
"INSERT INTO {tbl} VALUES (toDate('2018-10-01'), 3, 333), (toDate('2018-10-02'), 4, 444)".format(tbl=table))
assert n.query("SELECT count() from {tbl}".format(tbl=table)) == '4\n'
@pytest.mark.parametrize(
('node',),
[
(node9,),
(node10,)
]
)
def test_mixed_granularity_single_node(start_dynamic_cluster, node):
node.query(
"INSERT INTO table_with_default_granularity VALUES (toDate('2018-10-01'), 1, 333), (toDate('2018-10-02'), 2, 444)")
node.query(
"INSERT INTO table_with_default_granularity VALUES (toDate('2018-09-01'), 1, 333), (toDate('2018-09-02'), 2, 444)")
def callback(n):
n.replace_config("/etc/clickhouse-server/merge_tree_settings.xml",
"<yandex><merge_tree><enable_mixed_granularity_parts>1</enable_mixed_granularity_parts></merge_tree></yandex>")
n.replace_config("/etc/clickhouse-server/config.d/merge_tree_settings.xml",
"<yandex><merge_tree><enable_mixed_granularity_parts>1</enable_mixed_granularity_parts></merge_tree></yandex>")
node.restart_with_latest_version(callback_onstop=callback)
node.query("SYSTEM RELOAD CONFIG")
assert_eq_with_retry(node,
"SELECT value FROM system.merge_tree_settings WHERE name='enable_mixed_granularity_parts'",
'1')
assert node.query("SELECT count() from table_with_default_granularity") == '4\n'
node.query(
"INSERT INTO table_with_default_granularity VALUES (toDate('2018-10-01'), 3, 333), (toDate('2018-10-02'), 4, 444)")
assert node.query("SELECT count() from table_with_default_granularity") == '6\n'
node.query("OPTIMIZE TABLE table_with_default_granularity PARTITION 201810 FINAL")
assert node.query("SELECT count() from table_with_default_granularity") == '6\n'
path_to_merged_part = node.query(
"SELECT path FROM system.parts WHERE table = 'table_with_default_granularity' AND active=1 ORDER BY partition DESC LIMIT 1").strip()
node.exec_in_container(["bash", "-c", "find {p} -name '*.mrk2' | grep '.*'".format(
p=path_to_merged_part)]) # check that we have adaptive files
path_to_old_part = node.query(
"SELECT path FROM system.parts WHERE table = 'table_with_default_granularity' AND active=1 ORDER BY partition ASC LIMIT 1").strip()
node.exec_in_container(["bash", "-c", "find {p} -name '*.mrk' | grep '.*'".format(
p=path_to_old_part)]) # check that we have non adaptive files
node.query("ALTER TABLE table_with_default_granularity UPDATE dummy = dummy + 1 WHERE 1")
# still works
assert node.query("SELECT count() from table_with_default_granularity") == '6\n'
node.query("ALTER TABLE table_with_default_granularity MODIFY COLUMN dummy String")
node.query("ALTER TABLE table_with_default_granularity ADD COLUMN dummy2 Float64")
# still works
assert node.query("SELECT count() from table_with_default_granularity") == '6\n'
@pytest.mark.skip(reason="flaky")
def test_version_update_two_nodes(start_dynamic_cluster):
node11.query(
"INSERT INTO table_with_default_granularity VALUES (toDate('2018-10-01'), 1, 333), (toDate('2018-10-02'), 2, 444)")
node12.query("SYSTEM SYNC REPLICA table_with_default_granularity", timeout=20)
assert node12.query("SELECT COUNT() FROM table_with_default_granularity") == '2\n'
def callback(n):
n.replace_config("/etc/clickhouse-server/merge_tree_settings.xml",
"<yandex><merge_tree><enable_mixed_granularity_parts>0</enable_mixed_granularity_parts></merge_tree></yandex>")
n.replace_config("/etc/clickhouse-server/config.d/merge_tree_settings.xml",
"<yandex><merge_tree><enable_mixed_granularity_parts>0</enable_mixed_granularity_parts></merge_tree></yandex>")
node12.restart_with_latest_version(callback_onstop=callback)
node12.query(
"INSERT INTO table_with_default_granularity VALUES (toDate('2018-10-01'), 3, 333), (toDate('2018-10-02'), 4, 444)")
node11.query("SYSTEM SYNC REPLICA table_with_default_granularity", timeout=20)
assert node11.query("SELECT COUNT() FROM table_with_default_granularity") == '4\n'
node12.query(
'''
CREATE TABLE table_with_default_granularity_new(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/shard11/table_with_default_granularity_new', '2')
PARTITION BY toYYYYMM(date)
ORDER BY id
''')
node11.query(
'''
CREATE TABLE table_with_default_granularity_new(date Date, id UInt32, dummy UInt32)
ENGINE = ReplicatedMergeTree('/clickhouse/tables/test/shard11/table_with_default_granularity_new', '1')
PARTITION BY toYYYYMM(date)
ORDER BY id
''')
node12.query(
"INSERT INTO table_with_default_granularity_new VALUES (toDate('2018-10-01'), 1, 333), (toDate('2018-10-02'), 2, 444)")
with pytest.raises(QueryTimeoutExceedException):
node11.query("SYSTEM SYNC REPLICA table_with_default_granularity_new", timeout=20)
node12.query(
"INSERT INTO table_with_default_granularity_new VALUES (toDate('2018-10-01'), 3, 333), (toDate('2018-10-02'), 4, 444)")
node11.restart_with_latest_version(callback_onstop=callback) # just to be sure
for i in range(3):
try:
node11.query("SYSTEM SYNC REPLICA table_with_default_granularity_new", timeout=120)
node12.query("SYSTEM SYNC REPLICA table_with_default_granularity_new", timeout=120)
break
except Exception as ex:
print("Exception during replica sync", ex)
node11.query("SYSTEM RESTART REPLICA table_with_default_granularity_new")
node12.query("SYSTEM RESTART REPLICA table_with_default_granularity_new")
time.sleep(2 * i)
assert node11.query("SELECT COUNT() FROM table_with_default_granularity_new") == "4\n"
assert node12.query("SELECT COUNT() FROM table_with_default_granularity_new") == "4\n"
node11.query(
"INSERT INTO table_with_default_granularity VALUES (toDate('2018-10-01'), 5, 333), (toDate('2018-10-02'), 6, 444)")
for i in range(3):
try:
node12.query("SYSTEM SYNC REPLICA table_with_default_granularity", timeout=120)
break
except Exception as ex:
print("Exception during replica sync", ex)
node11.query("SYSTEM RESTART REPLICA table_with_default_granularity")
node12.query("SYSTEM RESTART REPLICA table_with_default_granularity")
time.sleep(2 * i)
assert node12.query("SELECT COUNT() FROM table_with_default_granularity") == '6\n'