ClickHouse/tests/integration/test_parallel_replicas_distributed_skip_shards/test.py
2023-12-13 14:42:06 +01:00

162 lines
5.0 KiB
Python

import pytest
from helpers.cluster import ClickHouseCluster
from helpers.client import QueryRuntimeException
cluster = ClickHouseCluster(__file__)
# create only 2 nodes out of 3 nodes in cluster with 1 shard
# and out of 6 nodes in first shard in cluster with 2 shards
node1 = cluster.add_instance(
"n1", main_configs=["configs/remote_servers.xml"], with_zookeeper=True
)
node2 = cluster.add_instance(
"n2", main_configs=["configs/remote_servers.xml"], with_zookeeper=True
)
@pytest.fixture(scope="module", autouse=True)
def start_cluster():
try:
cluster.start()
yield cluster
finally:
cluster.shutdown()
def create_tables(cluster, table_name):
# create replicated tables
node1.query(f"DROP TABLE IF EXISTS {table_name} SYNC")
node2.query(f"DROP TABLE IF EXISTS {table_name} SYNC")
node1.query(
f"CREATE TABLE IF NOT EXISTS {table_name} (key Int64, value String) Engine=ReplicatedMergeTree('/test_parallel_replicas/shard1/{table_name}', 'r1') ORDER BY (key)"
)
node2.query(
f"CREATE TABLE IF NOT EXISTS {table_name} (key Int64, value String) Engine=ReplicatedMergeTree('/test_parallel_replicas/shard1/{table_name}', 'r2') ORDER BY (key)"
)
# create distributed table
node1.query(f"DROP TABLE IF EXISTS {table_name}_d SYNC")
node1.query(
f"""
CREATE TABLE {table_name}_d AS {table_name}
Engine=Distributed(
{cluster},
currentDatabase(),
{table_name},
key
)
"""
)
# populate data
node1.query(f"INSERT INTO {table_name} SELECT number, number FROM numbers(1000)")
node2.query(f"INSERT INTO {table_name} SELECT -number, -number FROM numbers(1000)")
node1.query(f"INSERT INTO {table_name} SELECT number, number FROM numbers(3)")
@pytest.mark.parametrize(
"prefer_localhost_replica",
[
pytest.param(0),
pytest.param(1),
],
)
def test_skip_unavailable_shards(start_cluster, prefer_localhost_replica):
cluster = "test_multiple_shards_multiple_replicas"
table_name = "test_table"
create_tables(cluster, table_name)
expected_result = f"2003\t-999\t999\t3\n"
# w/o parallel replicas
assert (
node1.query(
f"SELECT count(), min(key), max(key), sum(key) FROM {table_name}_d settings skip_unavailable_shards=1"
)
== expected_result
)
# parallel replicas
assert (
node1.query(
f"SELECT count(), min(key), max(key), sum(key) FROM {table_name}_d",
settings={
"allow_experimental_parallel_reading_from_replicas": 2,
"max_parallel_replicas": 3,
"prefer_localhost_replica": prefer_localhost_replica,
"skip_unavailable_shards": 1,
"connections_with_failover_max_tries": 0, # just don't wait for unavailable replicas
},
)
== expected_result
)
@pytest.mark.parametrize(
"prefer_localhost_replica",
[
pytest.param(0),
pytest.param(1),
],
)
def test_error_on_unavailable_shards(start_cluster, prefer_localhost_replica):
cluster = "test_multiple_shards_multiple_replicas"
table_name = "test_table"
create_tables(cluster, table_name)
# w/o parallel replicas
with pytest.raises(QueryRuntimeException):
node1.query(
f"SELECT count(), min(key), max(key), sum(key) FROM {table_name}_d settings skip_unavailable_shards=0"
)
# parallel replicas
with pytest.raises(QueryRuntimeException):
node1.query(
f"SELECT count(), min(key), max(key), sum(key) FROM {table_name}_d",
settings={
"allow_experimental_parallel_reading_from_replicas": 2,
"max_parallel_replicas": 3,
"prefer_localhost_replica": prefer_localhost_replica,
"skip_unavailable_shards": 0,
},
)
@pytest.mark.parametrize(
"skip_unavailable_shards",
[
pytest.param(0),
pytest.param(1),
],
)
def test_no_unavailable_shards(start_cluster, skip_unavailable_shards):
cluster = "test_single_shard_multiple_replicas"
table_name = "test_table"
create_tables(cluster, table_name)
expected_result = f"2003\t-999\t999\t3\n"
# w/o parallel replicas
assert (
node1.query(
f"SELECT count(), min(key), max(key), sum(key) FROM {table_name}_d settings skip_unavailable_shards={skip_unavailable_shards}"
)
== expected_result
)
# parallel replicas
assert (
node1.query(
f"SELECT count(), min(key), max(key), sum(key) FROM {table_name}_d",
settings={
"allow_experimental_parallel_reading_from_replicas": 2,
"max_parallel_replicas": 3,
"prefer_localhost_replica": 0,
"skip_unavailable_shards": skip_unavailable_shards,
},
)
== expected_result
)