mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-10 01:25:21 +00:00
415 lines
13 KiB
Python
415 lines
13 KiB
Python
import pytest
|
|
|
|
from helpers.cluster import ClickHouseCluster, CLICKHOUSE_CI_MIN_TESTED_VERSION
|
|
|
|
uuids = []
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
def cluster():
|
|
try:
|
|
cluster = ClickHouseCluster(__file__)
|
|
cluster.add_instance(
|
|
"node1",
|
|
main_configs=["configs/storage_conf.xml", "configs/no_async_load.xml"],
|
|
with_nginx=True,
|
|
use_old_analyzer=True,
|
|
)
|
|
cluster.add_instance(
|
|
"node2",
|
|
main_configs=["configs/storage_conf_web.xml", "configs/no_async_load.xml"],
|
|
with_nginx=True,
|
|
stay_alive=True,
|
|
with_zookeeper=True,
|
|
use_old_analyzer=True,
|
|
)
|
|
cluster.add_instance(
|
|
"node3",
|
|
main_configs=["configs/storage_conf_web.xml", "configs/no_async_load.xml"],
|
|
with_nginx=True,
|
|
with_zookeeper=True,
|
|
use_old_analyzer=True,
|
|
)
|
|
|
|
cluster.add_instance(
|
|
"node4",
|
|
main_configs=["configs/storage_conf.xml", "configs/no_async_load.xml"],
|
|
with_nginx=True,
|
|
stay_alive=True,
|
|
with_installed_binary=True,
|
|
image="clickhouse/clickhouse-server",
|
|
tag=CLICKHOUSE_CI_MIN_TESTED_VERSION,
|
|
)
|
|
cluster.add_instance(
|
|
"node5",
|
|
main_configs=["configs/storage_conf.xml", "configs/no_async_load.xml"],
|
|
with_nginx=True,
|
|
use_old_analyzer=True,
|
|
)
|
|
|
|
cluster.start()
|
|
|
|
def create_table_and_upload_data(node, i):
|
|
node.query(
|
|
f"CREATE TABLE data{i} (id Int32) ENGINE = MergeTree() ORDER BY id SETTINGS storage_policy = 'def', min_bytes_for_wide_part=1;"
|
|
)
|
|
|
|
node.query("SYSTEM STOP MERGES")
|
|
|
|
for _ in range(10):
|
|
node.query(
|
|
f"INSERT INTO data{i} SELECT number FROM numbers(500000 * {i+1})"
|
|
)
|
|
node.query(f"SELECT * FROM data{i} ORDER BY id")
|
|
|
|
metadata_path = node.query(
|
|
f"SELECT data_paths FROM system.tables WHERE name='data{i}'"
|
|
)
|
|
metadata_path = metadata_path[
|
|
metadata_path.find("/") : metadata_path.rfind("/") + 1
|
|
]
|
|
print(f"Metadata: {metadata_path}")
|
|
|
|
node.exec_in_container(
|
|
[
|
|
"bash",
|
|
"-c",
|
|
"/usr/bin/clickhouse static-files-disk-uploader --test-mode --url http://nginx:80/test1 --metadata-path {}".format(
|
|
metadata_path
|
|
),
|
|
],
|
|
user="root",
|
|
)
|
|
parts = metadata_path.split("/")
|
|
print(f"UUID: {parts[3]}")
|
|
return parts[3]
|
|
|
|
node1 = cluster.instances["node1"]
|
|
|
|
global uuids
|
|
for i in range(2):
|
|
uuid = create_table_and_upload_data(node1, i)
|
|
uuids.append(uuid)
|
|
|
|
node4 = cluster.instances["node4"]
|
|
|
|
uuid = create_table_and_upload_data(node4, 2)
|
|
uuids.append(uuid)
|
|
|
|
yield cluster
|
|
|
|
finally:
|
|
cluster.shutdown()
|
|
|
|
|
|
@pytest.mark.parametrize("node_name", ["node2"])
|
|
def test_usage(cluster, node_name):
|
|
node1 = cluster.instances["node1"]
|
|
node4 = cluster.instances["node4"]
|
|
node2 = cluster.instances[node_name]
|
|
global uuids
|
|
assert len(uuids) == 3
|
|
for i in range(3):
|
|
node2.query(
|
|
"""
|
|
CREATE TABLE test{} UUID '{}'
|
|
(id Int32) ENGINE = MergeTree() ORDER BY id
|
|
SETTINGS storage_policy = 'web';
|
|
""".format(
|
|
i, uuids[i]
|
|
)
|
|
)
|
|
|
|
result = node2.query("SELECT * FROM test{} settings max_threads=20".format(i))
|
|
|
|
result = node2.query("SELECT count() FROM test{}".format(i))
|
|
assert int(result) == 5000000 * (i + 1)
|
|
|
|
result = node2.query(
|
|
"SELECT id FROM test{} WHERE id % 56 = 3 ORDER BY id".format(i)
|
|
)
|
|
node = node1
|
|
if i == 2:
|
|
node = node4
|
|
|
|
assert result == node.query(
|
|
"SELECT id FROM data{} WHERE id % 56 = 3 ORDER BY id".format(i)
|
|
)
|
|
|
|
result = node2.query(
|
|
"SELECT id FROM test{} WHERE id > 789999 AND id < 999999 ORDER BY id".format(
|
|
i
|
|
)
|
|
)
|
|
assert result == node.query(
|
|
"SELECT id FROM data{} WHERE id > 789999 AND id < 999999 ORDER BY id".format(
|
|
i
|
|
)
|
|
)
|
|
|
|
# to check right handling of paths in disk web
|
|
node2.query("SELECT count() FROM system.remote_data_paths")
|
|
|
|
node2.query("DROP TABLE test{} SYNC".format(i))
|
|
print(f"Ok {i}")
|
|
|
|
|
|
def test_incorrect_usage(cluster):
|
|
node2 = cluster.instances["node3"]
|
|
global uuids
|
|
node2.query(
|
|
"""
|
|
CREATE TABLE test0 UUID '{}'
|
|
(id Int32) ENGINE = MergeTree() ORDER BY id
|
|
SETTINGS storage_policy = 'web';
|
|
""".format(
|
|
uuids[0]
|
|
)
|
|
)
|
|
|
|
result = node2.query("SELECT count() FROM test0")
|
|
assert int(result) == 5000000
|
|
|
|
result = node2.query_and_get_error("ALTER TABLE test0 ADD COLUMN col1 Int32 first")
|
|
assert "Table is read-only" in result
|
|
|
|
result = node2.query_and_get_error("TRUNCATE TABLE test0")
|
|
assert "Table is read-only" in result
|
|
|
|
result = node2.query_and_get_error("OPTIMIZE TABLE test0 FINAL")
|
|
assert "Table is in readonly mode due to static storage" in result
|
|
|
|
node2.query("DROP TABLE test0 SYNC")
|
|
|
|
|
|
@pytest.mark.parametrize("node_name", ["node2"])
|
|
def test_cache(cluster, node_name):
|
|
node1 = cluster.instances["node1"]
|
|
node4 = cluster.instances["node4"]
|
|
node2 = cluster.instances[node_name]
|
|
global uuids
|
|
assert len(uuids) == 3
|
|
for i in range(3):
|
|
node2.query(
|
|
"""
|
|
CREATE TABLE test{} UUID '{}'
|
|
(id Int32) ENGINE = MergeTree() ORDER BY id
|
|
SETTINGS storage_policy = 'cached_web';
|
|
""".format(
|
|
i, uuids[i]
|
|
)
|
|
)
|
|
|
|
result = node2.query(
|
|
"""
|
|
SYSTEM DROP FILESYSTEM CACHE;
|
|
SELECT count() FROM system.filesystem_cache;
|
|
"""
|
|
)
|
|
assert int(result) == 0
|
|
|
|
result = node2.query("SELECT * FROM test{} settings max_threads=20".format(i))
|
|
|
|
result = node2.query(
|
|
"""
|
|
SELECT count() FROM system.filesystem_cache;
|
|
"""
|
|
)
|
|
assert int(result) > 0
|
|
|
|
result = node2.query("SELECT count() FROM test{}".format(i))
|
|
assert int(result) == 5000000 * (i + 1)
|
|
|
|
result = node2.query(
|
|
"SELECT id FROM test{} WHERE id % 56 = 3 ORDER BY id".format(i)
|
|
)
|
|
|
|
node = node1
|
|
if i == 2:
|
|
node = node4
|
|
|
|
assert result == node.query(
|
|
"SELECT id FROM data{} WHERE id % 56 = 3 ORDER BY id".format(i)
|
|
)
|
|
|
|
result = node2.query(
|
|
"SELECT id FROM test{} WHERE id > 789999 AND id < 999999 ORDER BY id".format(
|
|
i
|
|
)
|
|
)
|
|
assert result == node.query(
|
|
"SELECT id FROM data{} WHERE id > 789999 AND id < 999999 ORDER BY id".format(
|
|
i
|
|
)
|
|
)
|
|
|
|
node2.query("DROP TABLE test{} SYNC".format(i))
|
|
print(f"Ok {i}")
|
|
|
|
|
|
def test_unavailable_server(cluster):
|
|
"""
|
|
Regression test for the case when clickhouse-server simply ignore when
|
|
server is unavailable on start and later will simply return 0 rows for
|
|
SELECT from table on web disk.
|
|
"""
|
|
node2 = cluster.instances["node2"]
|
|
global uuids
|
|
node2.query(
|
|
"""
|
|
CREATE TABLE test0 UUID '{}'
|
|
(id Int32) ENGINE = MergeTree() ORDER BY id
|
|
SETTINGS storage_policy = 'web';
|
|
""".format(
|
|
uuids[0]
|
|
)
|
|
)
|
|
node2.stop_clickhouse()
|
|
try:
|
|
# NOTE: you cannot use separate disk instead, since MergeTree engine will
|
|
# try to lookup parts on all disks (to look unexpected disks with parts)
|
|
# and fail because of unavailable server.
|
|
node2.exec_in_container(
|
|
[
|
|
"bash",
|
|
"-c",
|
|
"sed -i 's#http://nginx:80/test1/#http://nginx:8080/test1/#' /etc/clickhouse-server/config.d/storage_conf_web.xml",
|
|
]
|
|
)
|
|
with pytest.raises(Exception):
|
|
# HTTP retries with backup can take awhile
|
|
node2.start_clickhouse(start_wait_sec=120, retry_start=False)
|
|
assert node2.contains_in_log(
|
|
"Caught exception while loading metadata.*Connection refused"
|
|
)
|
|
assert node2.contains_in_log(
|
|
"Failed to make request to 'http://nginx:8080/test1/.*'. Error: 'Connection refused'. Failed at try 10/10."
|
|
)
|
|
finally:
|
|
node2.exec_in_container(
|
|
[
|
|
"bash",
|
|
"-c",
|
|
"sed -i 's#http://nginx:8080/test1/#http://nginx:80/test1/#' /etc/clickhouse-server/config.d/storage_conf_web.xml",
|
|
]
|
|
)
|
|
node2.start_clickhouse()
|
|
node2.query("DROP TABLE test0 SYNC")
|
|
|
|
|
|
def test_replicated_database(cluster):
|
|
node1 = cluster.instances["node3"]
|
|
node1.query(
|
|
"CREATE DATABASE rdb ENGINE=Replicated('/test/rdb', 's1', 'r1')",
|
|
)
|
|
|
|
global uuids
|
|
node1.query(
|
|
"""
|
|
CREATE TABLE rdb.table0 UUID '{}'
|
|
(id Int32) ENGINE = MergeTree() ORDER BY id
|
|
SETTINGS storage_policy = 'web';
|
|
""".format(
|
|
uuids[0]
|
|
)
|
|
)
|
|
|
|
node2 = cluster.instances["node2"]
|
|
node2.query(
|
|
"CREATE DATABASE rdb ENGINE=Replicated('/test/rdb', 's1', 'r2')",
|
|
)
|
|
node2.query("SYSTEM SYNC DATABASE REPLICA rdb")
|
|
|
|
assert node1.query("SELECT count() FROM rdb.table0") == "5000000\n"
|
|
assert node2.query("SELECT count() FROM rdb.table0") == "5000000\n"
|
|
|
|
node1.query("DROP DATABASE rdb SYNC")
|
|
node2.query("DROP DATABASE rdb SYNC")
|
|
|
|
|
|
def test_page_cache(cluster):
|
|
node = cluster.instances["node2"]
|
|
global uuids
|
|
assert len(uuids) == 3
|
|
for i in range(3):
|
|
node.query(
|
|
"""
|
|
CREATE TABLE test{} UUID '{}'
|
|
(id Int32) ENGINE = MergeTree() ORDER BY id
|
|
SETTINGS storage_policy = 'web';
|
|
""".format(
|
|
i, uuids[i]
|
|
)
|
|
)
|
|
|
|
result1 = node.query(
|
|
f"SELECT sum(cityHash64(*)) FROM test{i} SETTINGS use_page_cache_for_disks_without_file_cache=1 -- test cold cache"
|
|
)
|
|
result2 = node.query(
|
|
f"SELECT sum(cityHash64(*)) FROM test{i} SETTINGS use_page_cache_for_disks_without_file_cache=1 -- test warm cache"
|
|
)
|
|
result3 = node.query(
|
|
f"SELECT sum(cityHash64(*)) FROM test{i} SETTINGS use_page_cache_for_disks_without_file_cache=0 -- test no cache"
|
|
)
|
|
|
|
assert result1 == result3
|
|
assert result2 == result3
|
|
|
|
node.query("SYSTEM FLUSH LOGS")
|
|
|
|
def get_profile_events(query_name):
|
|
text = node.query(
|
|
f"SELECT ProfileEvents.Names, ProfileEvents.Values FROM system.query_log ARRAY JOIN ProfileEvents WHERE query LIKE '% -- {query_name}' AND type = 'QueryFinish'"
|
|
)
|
|
res = {}
|
|
for line in text.split("\n"):
|
|
if line == "":
|
|
continue
|
|
name, value = line.split("\t")
|
|
res[name] = int(value)
|
|
return res
|
|
|
|
ev1 = get_profile_events("test cold cache")
|
|
assert ev1.get("PageCacheChunkMisses", 0) > 0
|
|
assert (
|
|
ev1.get("DiskConnectionsCreated", 0) + ev1.get("DiskConnectionsReused", 0)
|
|
> 0
|
|
)
|
|
|
|
ev2 = get_profile_events("test warm cache")
|
|
assert ev2.get("PageCacheChunkDataHits", 0) > 0
|
|
assert ev2.get("PageCacheChunkMisses", 0) == 0
|
|
assert (
|
|
ev2.get("DiskConnectionsCreated", 0) + ev2.get("DiskConnectionsReused", 0)
|
|
== 0
|
|
)
|
|
|
|
ev3 = get_profile_events("test no cache")
|
|
assert ev3.get("PageCacheChunkDataHits", 0) == 0
|
|
assert ev3.get("PageCacheChunkMisses", 0) == 0
|
|
assert (
|
|
ev3.get("DiskConnectionsCreated", 0) + ev3.get("DiskConnectionsReused", 0)
|
|
> 0
|
|
)
|
|
|
|
node.query("DROP TABLE test{} SYNC".format(i))
|
|
print(f"Ok {i}")
|
|
|
|
|
|
def test_config_reload(cluster):
|
|
node1 = cluster.instances["node5"]
|
|
table_name = "config_reload"
|
|
|
|
global uuids
|
|
node1.query(
|
|
f"""
|
|
DROP TABLE IF EXISTS {table_name};
|
|
CREATE TABLE {table_name} UUID '{uuids[0]}'
|
|
(id Int32) ENGINE = MergeTree() ORDER BY id
|
|
SETTINGS disk = disk(type=web, endpoint='http://nginx:80/test1/');
|
|
"""
|
|
)
|
|
|
|
node1.query("SYSTEM RELOAD CONFIG")
|
|
node1.query(f"DROP TABLE {table_name} SYNC")
|