mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-27 01:51:59 +00:00
139 lines
4.9 KiB
Python
139 lines
4.9 KiB
Python
import pytest
|
|
from helpers.cluster import ClickHouseCluster
|
|
from helpers.test_tools import TSV
|
|
|
|
|
|
cluster = ClickHouseCluster(__file__)
|
|
|
|
node_ll = cluster.add_instance(
|
|
"node_ll",
|
|
main_configs=[
|
|
"configs/remote_servers.xml",
|
|
],
|
|
user_configs=[
|
|
"configs/allow_database_types.xml",
|
|
],
|
|
macros={"replica": "node_ll", "shard": "shard"},
|
|
with_zookeeper=True,
|
|
)
|
|
|
|
node_no_ll = cluster.add_instance(
|
|
"node_no_ll",
|
|
main_configs=[
|
|
"configs/no_lazy_load.xml",
|
|
"configs/remote_servers.xml",
|
|
],
|
|
user_configs=[
|
|
"configs/allow_database_types.xml",
|
|
],
|
|
macros={"replica": "node_no_ll", "shard": "shard"},
|
|
with_zookeeper=True,
|
|
)
|
|
|
|
instances = [node_ll, node_no_ll]
|
|
|
|
|
|
@pytest.fixture(scope="module", autouse=True)
|
|
def started_cluster():
|
|
try:
|
|
cluster.start()
|
|
|
|
# "sleep(1)" is here to make loading of dictionaries a bit slower for this test.
|
|
instances[0].query(
|
|
"CREATE VIEW src ON CLUSTER 'cluster' AS SELECT number AS key, number * number + 1 + sleep(1) AS value FROM numbers(10)"
|
|
)
|
|
|
|
# "dict_get_user" can only call dictGet(), nothing more.
|
|
instances[0].query("CREATE USER dictget_user ON CLUSTER 'cluster'")
|
|
instances[0].query(
|
|
"GRANT dictGet ON atomicdb.dict TO dictget_user ON CLUSTER 'cluster'"
|
|
)
|
|
instances[0].query(
|
|
"GRANT dictGet ON repldb.dict TO dictget_user ON CLUSTER 'cluster'"
|
|
)
|
|
|
|
instances[0].query("CREATE DATABASE atomicdb ON CLUSTER 'cluster'")
|
|
instances[0].query(
|
|
"CREATE DATABASE repldb ON CLUSTER 'cluster' ENGINE=Replicated('/clickhouse/path/','{shard}','{replica}')"
|
|
)
|
|
|
|
yield cluster
|
|
|
|
finally:
|
|
instances[0].query("DROP TABLE IF EXISTS src ON CLUSTER 'cluster'")
|
|
instances[0].query("DROP USER IF EXISTS dictget_user ON CLUSTER 'cluster'")
|
|
instances[0].query("DROP DATABASE IF EXISTS atomicdb ON CLUSTER 'cluster'")
|
|
instances[0].query("DROP DATABASE IF EXISTS repldb ON CLUSTER 'cluster'")
|
|
|
|
cluster.shutdown()
|
|
|
|
|
|
@pytest.fixture(autouse=True)
|
|
def cleanup_after_test():
|
|
try:
|
|
yield
|
|
finally:
|
|
instances[0].query("DROP DICTIONARY IF EXISTS dict ON CLUSTER 'cluster'")
|
|
|
|
|
|
def get_status(instance, dictionary_name):
|
|
return instance.query(
|
|
"SELECT status FROM system.dictionaries WHERE name='" + dictionary_name + "'"
|
|
).rstrip("\n")
|
|
|
|
|
|
@pytest.mark.parametrize(
|
|
"database, instance_to_create_dictionary, instances_to_check",
|
|
[
|
|
("atomicdb", node_ll, [node_ll]),
|
|
("atomicdb", node_no_ll, [node_no_ll]),
|
|
("repldb", node_ll, [node_ll, node_no_ll]),
|
|
("repldb", node_no_ll, [node_ll, node_no_ll]),
|
|
],
|
|
)
|
|
def test_create_or_replace(database, instance_to_create_dictionary, instances_to_check):
|
|
num_steps = 2
|
|
dict_uuids = {}
|
|
for step in range(0, num_steps):
|
|
create_dictionary_query = f"CREATE OR REPLACE DICTIONARY {database}.dict (key Int64, value Int64) PRIMARY KEY key SOURCE(CLICKHOUSE(HOST 'localhost' PORT tcpPort() USER 'default' TABLE 'src' PASSWORD '' DB 'default')) LIFETIME(0) LAYOUT(FLAT())"
|
|
instance_to_create_dictionary.query(create_dictionary_query)
|
|
|
|
for instance in instances_to_check:
|
|
if instance != instance_to_create_dictionary:
|
|
instance.query(f"SYSTEM SYNC DATABASE REPLICA {database}")
|
|
|
|
dict_uuid = instance.query(
|
|
f"SELECT uuid FROM system.tables WHERE database='{database}' AND table='dict'"
|
|
).strip("\n")
|
|
|
|
dict_db, dict_name, dict_status = (
|
|
instance.query(
|
|
f"SELECT database, name, status FROM system.dictionaries WHERE uuid='{dict_uuid}'"
|
|
)
|
|
.strip("\n")
|
|
.split("\t")
|
|
)
|
|
assert dict_db == database
|
|
assert dict_name == "dict"
|
|
|
|
# "uuid" must be the same for all the dictionaries created at the same "step" and different for the dictionaries created at different steps.
|
|
if step in dict_uuids:
|
|
assert dict_uuids[step] == dict_uuid
|
|
dict_uuids[step] = dict_uuid
|
|
assert dict_uuid not in [
|
|
dict_uuids[prev_step] for prev_step in range(0, step)
|
|
]
|
|
|
|
expected_dict_status = (
|
|
["NOT_LOADED"] if instance == node_ll else ["LOADING", "LOADED"]
|
|
)
|
|
assert dict_status in expected_dict_status
|
|
|
|
for instance in instances_to_check:
|
|
select_query = f"SELECT arrayJoin([0, 5, 7, 11]) as key, dictGet({database}.dict, 'value', key)"
|
|
expected_result = TSV([[0, 1], [5, 26], [7, 50], [11, 0]])
|
|
assert instance.query(select_query) == expected_result
|
|
assert instance.query(select_query, user="dictget_user") == expected_result
|
|
|
|
instance_to_create_dictionary.query(f"DROP DICTIONARY IF EXISTS {database}.dict")
|