ClickHouse/tests/integration/test_replicated_database/test.py

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1548 lines
63 KiB
Python
Raw Normal View History

import os
import shutil
import time
2020-10-26 15:12:16 +00:00
import re
import pytest
2022-07-27 17:15:00 +00:00
import threading
from helpers.cluster import ClickHouseCluster
2021-02-11 22:23:40 +00:00
from helpers.test_tools import assert_eq_with_retry, assert_logs_contain
from helpers.network import PartitionManager
2021-11-15 14:07:08 +00:00
test_recover_staled_replica_run = 1
cluster = ClickHouseCluster(__file__)
main_node = cluster.add_instance(
"main_node",
main_configs=["configs/config.xml"],
user_configs=["configs/settings.xml"],
with_zookeeper=True,
stay_alive=True,
macros={"shard": 1, "replica": 1},
)
dummy_node = cluster.add_instance(
"dummy_node",
main_configs=["configs/config.xml"],
user_configs=["configs/settings2.xml"],
with_zookeeper=True,
stay_alive=True,
macros={"shard": 1, "replica": 2},
)
competing_node = cluster.add_instance(
"competing_node",
main_configs=["configs/config.xml"],
user_configs=["configs/settings.xml"],
with_zookeeper=True,
stay_alive=True,
macros={"shard": 1, "replica": 3},
)
snapshotting_node = cluster.add_instance(
"snapshotting_node",
main_configs=["configs/config.xml"],
user_configs=["configs/settings.xml"],
with_zookeeper=True,
macros={"shard": 2, "replica": 1},
)
snapshot_recovering_node = cluster.add_instance(
"snapshot_recovering_node",
2024-06-03 18:51:09 +00:00
main_configs=["configs/config.xml"],
user_configs=["configs/settings.xml"],
with_zookeeper=True,
)
all_nodes = [
main_node,
dummy_node,
competing_node,
snapshotting_node,
snapshot_recovering_node,
]
2021-02-02 19:39:04 +00:00
2024-05-13 13:46:25 +00:00
bad_settings_node = cluster.add_instance(
2024-05-14 15:22:36 +00:00
"bad_settings_node",
2024-06-03 18:51:09 +00:00
main_configs=["configs/config2.xml"],
2024-05-13 13:46:25 +00:00
user_configs=["configs/inconsistent_settings.xml"],
with_zookeeper=True,
macros={"shard": 1, "replica": 4},
)
uuid_regex = re.compile("[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}")
2020-10-26 15:12:16 +00:00
def assert_create_query(nodes, table_name, expected):
replace_uuid = lambda x: re.sub(uuid_regex, "uuid", x)
2020-11-24 10:24:39 +00:00
query = "show create table {}".format(table_name)
2020-10-26 15:12:16 +00:00
for node in nodes:
assert_eq_with_retry(node, query, expected, get_result=replace_uuid)
@pytest.fixture(scope="module")
def started_cluster():
try:
cluster.start()
yield cluster
finally:
cluster.shutdown()
def test_flatten_nested(started_cluster):
main_node.query(
"CREATE DATABASE create_replicated_table ENGINE = Replicated('/test/create_replicated_table', 'shard1', 'replica' || '1');"
)
dummy_node.query(
"CREATE DATABASE create_replicated_table ENGINE = Replicated('/test/create_replicated_table', 'shard1', 'replica2');"
)
main_node.query(
"CREATE TABLE create_replicated_table.replicated_table (d Date, k UInt64, i32 Int32) ENGINE=ReplicatedMergeTree ORDER BY k PARTITION BY toYYYYMM(d);"
)
main_node.query(
"CREATE MATERIALIZED VIEW create_replicated_table.mv ENGINE=ReplicatedMergeTree ORDER BY tuple() AS select d, cast([(k, toString(i32))] as Nested(a UInt64, b String)) from create_replicated_table.replicated_table"
)
assert main_node.query(
"show create create_replicated_table.mv"
) == dummy_node.query("show create create_replicated_table.mv")
main_node.query("DROP DATABASE create_replicated_table SYNC")
dummy_node.query("DROP DATABASE create_replicated_table SYNC")
2020-05-24 17:12:24 +00:00
def test_create_replicated_table(started_cluster):
main_node.query(
"CREATE DATABASE create_replicated_table ENGINE = Replicated('/test/create_replicated_table', 'shard1', 'replica' || '1');"
)
dummy_node.query(
"CREATE DATABASE create_replicated_table ENGINE = Replicated('/test/create_replicated_table', 'shard1', 'replica2');"
)
assert (
"Explicit zookeeper_path and replica_name are specified"
in main_node.query_and_get_error(
"CREATE TABLE create_replicated_table.replicated_table (d Date, k UInt64, i32 Int32) "
"ENGINE=ReplicatedMergeTree('/test/tmp', 'r') ORDER BY k PARTITION BY toYYYYMM(d);"
)
)
2021-03-17 14:29:24 +00:00
assert (
"Explicit zookeeper_path and replica_name are specified"
in main_node.query_and_get_error(
"CREATE TABLE create_replicated_table.replicated_table (d Date, k UInt64, i32 Int32) "
2022-06-24 17:10:33 +00:00
"ENGINE=ReplicatedMergeTree('/test/tmp', 'r') ORDER BY k PARTITION BY toYYYYMM(d);"
)
)
2021-03-17 14:29:24 +00:00
2022-06-24 17:10:33 +00:00
assert (
"This syntax for *MergeTree engine is deprecated"
in main_node.query_and_get_error(
"CREATE TABLE create_replicated_table.replicated_table (d Date, k UInt64, i32 Int32) "
2022-06-24 17:10:33 +00:00
"ENGINE=ReplicatedMergeTree('/test/tmp/{shard}', '{replica}', d, k, 8192);"
)
)
2021-02-02 19:39:04 +00:00
main_node.query(
"CREATE TABLE create_replicated_table.replicated_table (d Date, k UInt64, i32 Int32) ENGINE=ReplicatedMergeTree ORDER BY k PARTITION BY toYYYYMM(d);"
)
expected = (
"CREATE TABLE create_replicated_table.replicated_table\\n(\\n `d` Date,\\n `k` UInt64,\\n `i32` Int32\\n)\\n"
2022-04-13 14:51:59 +00:00
"ENGINE = ReplicatedMergeTree(\\'/clickhouse/tables/{uuid}/{shard}\\', \\'{replica}\\')\\n"
"PARTITION BY toYYYYMM(d)\\nORDER BY k\\nSETTINGS index_granularity = 8192"
)
2023-03-02 17:35:56 +00:00
assert_create_query(
[main_node, dummy_node], "create_replicated_table.replicated_table", expected
)
2020-10-26 15:12:16 +00:00
# assert without replacing uuid
2023-03-02 17:35:56 +00:00
assert main_node.query(
"show create create_replicated_table.replicated_table"
2023-03-02 17:35:56 +00:00
) == dummy_node.query("show create create_replicated_table.replicated_table")
main_node.query("DROP DATABASE create_replicated_table SYNC")
dummy_node.query("DROP DATABASE create_replicated_table SYNC")
2020-05-24 17:12:24 +00:00
@pytest.mark.parametrize("engine", ["MergeTree", "ReplicatedMergeTree"])
2020-11-24 10:24:39 +00:00
def test_simple_alter_table(started_cluster, engine):
database = f"test_simple_alter_table_{engine}"
main_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica1');"
)
dummy_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica2');"
)
2021-02-02 19:39:04 +00:00
# test_simple_alter_table
name = f"{database}.alter_test"
main_node.query(
"CREATE TABLE {} "
"(CounterID UInt32, StartDate Date, UserID UInt32, VisitID UInt32, NestedColumn Nested(A UInt8, S String), ToDrop UInt32) "
"ENGINE = {} PARTITION BY StartDate ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID);".format(
name, engine
)
)
2020-11-24 10:24:39 +00:00
main_node.query("ALTER TABLE {} ADD COLUMN Added0 UInt32;".format(name))
main_node.query("ALTER TABLE {} ADD COLUMN Added2 UInt32;".format(name))
main_node.query(
"ALTER TABLE {} ADD COLUMN Added1 UInt32 AFTER Added0;".format(name)
)
main_node.query(
"ALTER TABLE {} ADD COLUMN AddedNested1 Nested(A UInt32, B UInt64) AFTER Added2;".format(
name
)
)
main_node.query(
"ALTER TABLE {} ADD COLUMN AddedNested1.C Array(String) AFTER AddedNested1.B;".format(
name
)
)
main_node.query(
"ALTER TABLE {} ADD COLUMN AddedNested2 Nested(A UInt32, B UInt64) AFTER AddedNested1;".format(
name
)
)
full_engine = (
engine
if not "Replicated" in engine
2022-04-13 14:51:59 +00:00
else engine + "(\\'/clickhouse/tables/{uuid}/{shard}\\', \\'{replica}\\')"
)
expected = (
"CREATE TABLE {}\\n(\\n `CounterID` UInt32,\\n `StartDate` Date,\\n `UserID` UInt32,\\n"
" `VisitID` UInt32,\\n `NestedColumn.A` Array(UInt8),\\n `NestedColumn.S` Array(String),\\n"
" `ToDrop` UInt32,\\n `Added0` UInt32,\\n `Added1` UInt32,\\n `Added2` UInt32,\\n"
" `AddedNested1.A` Array(UInt32),\\n `AddedNested1.B` Array(UInt64),\\n `AddedNested1.C` Array(String),\\n"
" `AddedNested2.A` Array(UInt32),\\n `AddedNested2.B` Array(UInt64)\\n)\\n"
"ENGINE = {}\\nPARTITION BY StartDate\\nORDER BY (CounterID, StartDate, intHash32(UserID), VisitID)\\n"
"SETTINGS index_granularity = 8192".format(name, full_engine)
)
2020-10-26 15:12:16 +00:00
2020-11-24 10:24:39 +00:00
assert_create_query([main_node, dummy_node], name, expected)
2020-05-24 17:12:24 +00:00
2021-02-02 19:39:04 +00:00
# test_create_replica_after_delay
competing_node.query(
f"CREATE DATABASE IF NOT EXISTS {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica3');"
)
2020-05-24 17:12:24 +00:00
2020-11-24 10:24:39 +00:00
main_node.query("ALTER TABLE {} ADD COLUMN Added3 UInt32;".format(name))
main_node.query("ALTER TABLE {} DROP COLUMN AddedNested1;".format(name))
main_node.query("ALTER TABLE {} RENAME COLUMN Added1 TO AddedNested1;".format(name))
full_engine = (
engine
if not "Replicated" in engine
2022-04-13 14:51:59 +00:00
else engine + "(\\'/clickhouse/tables/{uuid}/{shard}\\', \\'{replica}\\')"
)
expected = (
"CREATE TABLE {}\\n(\\n `CounterID` UInt32,\\n `StartDate` Date,\\n `UserID` UInt32,\\n"
" `VisitID` UInt32,\\n `NestedColumn.A` Array(UInt8),\\n `NestedColumn.S` Array(String),\\n"
" `ToDrop` UInt32,\\n `Added0` UInt32,\\n `AddedNested1` UInt32,\\n `Added2` UInt32,\\n"
" `AddedNested2.A` Array(UInt32),\\n `AddedNested2.B` Array(UInt64),\\n `Added3` UInt32\\n)\\n"
"ENGINE = {}\\nPARTITION BY StartDate\\nORDER BY (CounterID, StartDate, intHash32(UserID), VisitID)\\n"
"SETTINGS index_granularity = 8192".format(name, full_engine)
)
2020-05-24 17:12:24 +00:00
2020-11-24 10:24:39 +00:00
assert_create_query([main_node, dummy_node, competing_node], name, expected)
main_node.query(f"DROP DATABASE {database} SYNC")
dummy_node.query(f"DROP DATABASE {database} SYNC")
competing_node.query(f"DROP DATABASE {database} SYNC")
2021-02-02 19:39:04 +00:00
@pytest.mark.parametrize("engine", ["MergeTree", "ReplicatedMergeTree"])
def test_delete_from_table(started_cluster, engine):
database = f"delete_from_table_{engine}"
main_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica1');"
)
dummy_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard2', 'replica1');"
)
name = f"{database}.delete_test"
main_node.query(
"CREATE TABLE {} "
"(id UInt64, value String) "
2023-01-31 14:35:08 +00:00
"ENGINE = {} PARTITION BY id%2 ORDER BY (id);".format(name, engine)
)
main_node.query("INSERT INTO TABLE {} VALUES(1, 'aaaa');".format(name))
main_node.query("INSERT INTO TABLE {} VALUES(2, 'aaaa');".format(name))
dummy_node.query("INSERT INTO TABLE {} VALUES(1, 'bbbb');".format(name))
dummy_node.query("INSERT INTO TABLE {} VALUES(2, 'bbbb');".format(name))
2023-03-01 20:12:00 +00:00
main_node.query("DELETE FROM {} WHERE id=2;".format(name))
2023-01-31 14:35:08 +00:00
expected = "1\taaaa\n1\tbbbb"
2023-01-31 14:35:08 +00:00
table_for_select = name
if not "Replicated" in engine:
table_for_select = f"cluster('{database}', {name})"
for node in [main_node, dummy_node]:
2023-01-31 14:35:08 +00:00
assert_eq_with_retry(
node,
"SELECT * FROM {} ORDER BY id, value;".format(table_for_select),
expected,
)
main_node.query(f"DROP DATABASE {database} SYNC")
dummy_node.query(f"DROP DATABASE {database} SYNC")
def get_table_uuid(database, name):
return main_node.query(
f"SELECT uuid FROM system.tables WHERE database = '{database}' and name = '{name}'"
).strip()
@pytest.fixture(scope="module", name="attachable_part")
def fixture_attachable_part(started_cluster):
main_node.query(f"CREATE DATABASE testdb_attach_atomic ENGINE = Atomic")
main_node.query(
f"CREATE TABLE testdb_attach_atomic.test (CounterID UInt32) ENGINE = MergeTree ORDER BY (CounterID)"
)
main_node.query(f"INSERT INTO testdb_attach_atomic.test VALUES (123)")
main_node.query(
f"ALTER TABLE testdb_attach_atomic.test FREEZE WITH NAME 'test_attach'"
)
table_uuid = get_table_uuid("testdb_attach_atomic", "test")
return os.path.join(
main_node.path,
f"database/shadow/test_attach/store/{table_uuid[:3]}/{table_uuid}/all_1_1_0",
)
@pytest.mark.parametrize("engine", ["MergeTree", "ReplicatedMergeTree"])
def test_alter_attach(started_cluster, attachable_part, engine):
database = f"alter_attach_{engine}"
main_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica1');"
)
dummy_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica2');"
)
2021-11-11 08:12:54 +00:00
main_node.query(
f"CREATE TABLE {database}.alter_attach_test (CounterID UInt32) ENGINE = {engine} ORDER BY (CounterID)"
)
table_uuid = get_table_uuid(database, "alter_attach_test")
# Provide and attach a part to the main node
shutil.copytree(
attachable_part,
os.path.join(
main_node.path,
f"database/store/{table_uuid[:3]}/{table_uuid}/detached/all_1_1_0",
),
)
main_node.query(f"ALTER TABLE {database}.alter_attach_test ATTACH PART 'all_1_1_0'")
# On the main node, data is attached
2023-07-19 08:47:53 +00:00
assert (
main_node.query(f"SELECT CounterID FROM {database}.alter_attach_test")
== "123\n"
)
# On the other node, data is replicated only if using a Replicated table engine
if engine == "ReplicatedMergeTree":
2023-07-19 08:47:53 +00:00
assert (
dummy_node.query(f"SELECT CounterID FROM {database}.alter_attach_test")
== "123\n"
)
else:
2023-07-19 08:47:53 +00:00
assert (
dummy_node.query(f"SELECT CounterID FROM {database}.alter_attach_test")
== ""
)
main_node.query(f"DROP DATABASE {database} SYNC")
dummy_node.query(f"DROP DATABASE {database} SYNC")
@pytest.mark.parametrize("engine", ["MergeTree", "ReplicatedMergeTree"])
def test_alter_drop_part(started_cluster, engine):
database = f"alter_drop_part_{engine}"
main_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica1');"
)
dummy_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica2');"
)
2021-11-11 08:12:54 +00:00
part_name = "all_0_0_0" if engine == "ReplicatedMergeTree" else "all_1_1_0"
main_node.query(
f"CREATE TABLE {database}.alter_drop_part (CounterID UInt32) ENGINE = {engine} ORDER BY (CounterID)"
)
main_node.query(f"INSERT INTO {database}.alter_drop_part VALUES (123)")
if engine == "MergeTree":
dummy_node.query(f"INSERT INTO {database}.alter_drop_part VALUES (456)")
2024-01-10 10:07:13 +00:00
else:
main_node.query(f"SYSTEM SYNC REPLICA {database}.alter_drop_part PULL")
main_node.query(f"ALTER TABLE {database}.alter_drop_part DROP PART '{part_name}'")
assert main_node.query(f"SELECT CounterID FROM {database}.alter_drop_part") == ""
if engine == "ReplicatedMergeTree":
# The DROP operation is still replicated at the table engine level
2023-07-19 08:47:53 +00:00
assert (
dummy_node.query(f"SELECT CounterID FROM {database}.alter_drop_part") == ""
)
else:
2023-03-02 17:35:56 +00:00
assert (
dummy_node.query(f"SELECT CounterID FROM {database}.alter_drop_part")
2023-03-02 17:35:56 +00:00
== "456\n"
)
main_node.query(f"DROP DATABASE {database} SYNC")
dummy_node.query(f"DROP DATABASE {database} SYNC")
@pytest.mark.parametrize("engine", ["MergeTree", "ReplicatedMergeTree"])
def test_alter_detach_part(started_cluster, engine):
database = f"alter_detach_part_{engine}"
main_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica1');"
)
dummy_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica2');"
)
2021-11-11 08:12:54 +00:00
part_name = "all_0_0_0" if engine == "ReplicatedMergeTree" else "all_1_1_0"
main_node.query(
f"CREATE TABLE {database}.alter_detach (CounterID UInt32) ENGINE = {engine} ORDER BY (CounterID)"
)
main_node.query(f"INSERT INTO {database}.alter_detach VALUES (123)")
if engine == "MergeTree":
dummy_node.query(f"INSERT INTO {database}.alter_detach VALUES (456)")
else:
main_node.query(f"SYSTEM SYNC REPLICA {database}.alter_detach PULL")
main_node.query(f"ALTER TABLE {database}.alter_detach DETACH PART '{part_name}'")
detached_parts_query = f"SELECT name FROM system.detached_parts WHERE database='{database}' AND table='alter_detach'"
assert main_node.query(detached_parts_query) == f"{part_name}\n"
if engine == "ReplicatedMergeTree":
# The detach operation is still replicated at the table engine level
assert dummy_node.query(detached_parts_query) == f"{part_name}\n"
else:
assert dummy_node.query(detached_parts_query) == ""
main_node.query(f"DROP DATABASE {database} SYNC")
dummy_node.query(f"DROP DATABASE {database} SYNC")
@pytest.mark.parametrize("engine", ["MergeTree", "ReplicatedMergeTree"])
def test_alter_drop_detached_part(started_cluster, engine):
database = f"alter_drop_detached_part_{engine}"
main_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica1');"
)
dummy_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica2');"
)
2021-11-11 08:12:54 +00:00
part_name = "all_0_0_0" if engine == "ReplicatedMergeTree" else "all_1_1_0"
main_node.query(
f"CREATE TABLE {database}.alter_drop_detached (CounterID UInt32) ENGINE = {engine} ORDER BY (CounterID)"
)
main_node.query(f"INSERT INTO {database}.alter_drop_detached VALUES (123)")
2023-03-02 17:35:56 +00:00
main_node.query(
f"ALTER TABLE {database}.alter_drop_detached DETACH PART '{part_name}'"
2023-03-02 17:35:56 +00:00
)
if engine == "MergeTree":
dummy_node.query(f"INSERT INTO {database}.alter_drop_detached VALUES (456)")
2023-03-02 17:35:56 +00:00
dummy_node.query(
f"ALTER TABLE {database}.alter_drop_detached DETACH PART '{part_name}'"
2023-03-02 17:35:56 +00:00
)
main_node.query(
f"ALTER TABLE {database}.alter_drop_detached DROP DETACHED PART '{part_name}'"
2023-03-02 17:35:56 +00:00
)
detached_parts_query = f"SELECT name FROM system.detached_parts WHERE database='{database}' AND table='alter_drop_detached'"
assert main_node.query(detached_parts_query) == ""
assert dummy_node.query(detached_parts_query) == f"{part_name}\n"
main_node.query(f"DROP DATABASE {database} SYNC")
dummy_node.query(f"DROP DATABASE {database} SYNC")
2021-11-11 08:12:54 +00:00
@pytest.mark.parametrize("engine", ["MergeTree", "ReplicatedMergeTree"])
def test_alter_drop_partition(started_cluster, engine):
database = f"alter_drop_partition_{engine}"
main_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica1');"
)
dummy_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard1', 'replica2');"
)
snapshotting_node.query(
f"CREATE DATABASE {database} ENGINE = Replicated('/test/{database}', 'shard2', 'replica1');"
)
main_node.query(
f"CREATE TABLE {database}.alter_drop (CounterID UInt32) ENGINE = {engine} ORDER BY (CounterID)"
)
main_node.query(f"INSERT INTO {database}.alter_drop VALUES (123)")
if engine == "MergeTree":
dummy_node.query(f"INSERT INTO {database}.alter_drop VALUES (456)")
snapshotting_node.query(f"INSERT INTO {database}.alter_drop VALUES (789)")
main_node.query(
f"ALTER TABLE {database}.alter_drop ON CLUSTER {database} DROP PARTITION ID 'all'",
settings={"replication_alter_partitions_sync": 2},
)
assert (
main_node.query(
f"SELECT CounterID FROM clusterAllReplicas('{database}', {database}.alter_drop)"
)
== ""
)
assert dummy_node.query(f"SELECT CounterID FROM {database}.alter_drop") == ""
main_node.query(f"DROP DATABASE {database}")
dummy_node.query(f"DROP DATABASE {database}")
snapshotting_node.query(f"DROP DATABASE {database}")
def test_alter_fetch(started_cluster):
main_node.query(
"CREATE DATABASE alter_fetch ENGINE = Replicated('/test/alter_fetch', 'shard1', 'replica1');"
)
dummy_node.query(
"CREATE DATABASE alter_fetch ENGINE = Replicated('/test/alter_fetch', 'shard1', 'replica2');"
)
2021-11-11 08:12:54 +00:00
main_node.query(
"CREATE TABLE alter_fetch.fetch_source (CounterID UInt32) ENGINE = ReplicatedMergeTree ORDER BY (CounterID)"
)
main_node.query(
"CREATE TABLE alter_fetch.fetch_target (CounterID UInt32) ENGINE = ReplicatedMergeTree ORDER BY (CounterID)"
)
main_node.query("INSERT INTO alter_fetch.fetch_source VALUES (123)")
table_uuid = get_table_uuid("alter_fetch", "fetch_source")
main_node.query(
f"ALTER TABLE alter_fetch.fetch_target FETCH PART 'all_0_0_0' FROM '/clickhouse/tables/{table_uuid}/{{shard}}' "
)
detached_parts_query = "SELECT name FROM system.detached_parts WHERE database='alter_fetch' AND table='fetch_target'"
assert main_node.query(detached_parts_query) == "all_0_0_0\n"
assert dummy_node.query(detached_parts_query) == ""
main_node.query("DROP DATABASE alter_fetch SYNC")
dummy_node.query("DROP DATABASE alter_fetch SYNC")
def test_alters_from_different_replicas(started_cluster):
main_node.query(
"CREATE DATABASE alters_from_different_replicas ENGINE = Replicated('/test/alters_from_different_replicas', 'shard1', 'replica1');"
)
dummy_node.query(
"CREATE DATABASE alters_from_different_replicas ENGINE = Replicated('/test/alters_from_different_replicas', 'shard1', 'replica2');"
)
2021-11-11 08:12:54 +00:00
2021-02-02 19:39:04 +00:00
# test_alters_from_different_replicas
competing_node.query(
"CREATE DATABASE alters_from_different_replicas ENGINE = Replicated('/test/alters_from_different_replicas', 'shard1', 'replica3');"
)
2021-02-02 19:39:04 +00:00
main_node.query(
"CREATE TABLE alters_from_different_replicas.concurrent_test "
"(CounterID UInt32, StartDate Date, UserID UInt32, VisitID UInt32, NestedColumn Nested(A UInt8, S String), ToDrop UInt32) "
2022-06-23 08:37:52 +00:00
"ENGINE = MergeTree PARTITION BY toYYYYMM(StartDate) ORDER BY (CounterID, StartDate, intHash32(UserID), VisitID);"
)
main_node.query(
"CREATE TABLE alters_from_different_replicas.dist AS alters_from_different_replicas.concurrent_test ENGINE = Distributed(alters_from_different_replicas, alters_from_different_replicas, concurrent_test, CounterID)"
)
2021-02-02 19:39:04 +00:00
2021-02-09 15:14:20 +00:00
dummy_node.stop_clickhouse(kill=True)
settings = {"distributed_ddl_task_timeout": 5}
2024-03-04 19:22:18 +00:00
assert "is not finished on 1 of 3 hosts" in competing_node.query_and_get_error(
"ALTER TABLE alters_from_different_replicas.concurrent_test ADD COLUMN Added0 UInt32;",
settings=settings,
)
settings = {
"distributed_ddl_task_timeout": 5,
"distributed_ddl_output_mode": "null_status_on_timeout",
}
2022-06-29 14:27:21 +00:00
assert "shard1\treplica2\tQUEUED\t" in main_node.query(
"ALTER TABLE alters_from_different_replicas.concurrent_test ADD COLUMN Added2 UInt32;",
settings=settings,
)
settings = {
"distributed_ddl_task_timeout": 5,
"distributed_ddl_output_mode": "never_throw",
}
2022-06-29 14:27:21 +00:00
assert "shard1\treplica2\tQUEUED\t" in competing_node.query(
"ALTER TABLE alters_from_different_replicas.concurrent_test ADD COLUMN Added1 UInt32 AFTER Added0;",
settings=settings,
)
2021-02-09 15:14:20 +00:00
dummy_node.start_clickhouse()
main_node.query(
"ALTER TABLE alters_from_different_replicas.concurrent_test ADD COLUMN AddedNested1 Nested(A UInt32, B UInt64) AFTER Added2;"
)
competing_node.query(
"ALTER TABLE alters_from_different_replicas.concurrent_test ADD COLUMN AddedNested1.C Array(String) AFTER AddedNested1.B;"
)
main_node.query(
"ALTER TABLE alters_from_different_replicas.concurrent_test ADD COLUMN AddedNested2 Nested(A UInt32, B UInt64) AFTER AddedNested1;"
)
2020-06-07 11:26:42 +00:00
expected = (
"CREATE TABLE alters_from_different_replicas.concurrent_test\\n(\\n `CounterID` UInt32,\\n `StartDate` Date,\\n `UserID` UInt32,\\n"
" `VisitID` UInt32,\\n `NestedColumn.A` Array(UInt8),\\n `NestedColumn.S` Array(String),\\n `ToDrop` UInt32,\\n"
" `Added0` UInt32,\\n `Added1` UInt32,\\n `Added2` UInt32,\\n `AddedNested1.A` Array(UInt32),\\n"
" `AddedNested1.B` Array(UInt64),\\n `AddedNested1.C` Array(String),\\n `AddedNested2.A` Array(UInt32),\\n"
" `AddedNested2.B` Array(UInt64)\\n)\\n"
2022-06-23 08:37:52 +00:00
"ENGINE = MergeTree\\nPARTITION BY toYYYYMM(StartDate)\\nORDER BY (CounterID, StartDate, intHash32(UserID), VisitID)\\nSETTINGS index_granularity = 8192"
)
2020-06-07 11:26:42 +00:00
2023-03-02 17:35:56 +00:00
assert_create_query(
[main_node, competing_node],
"alters_from_different_replicas.concurrent_test",
expected,
)
2020-06-20 15:38:20 +00:00
2021-02-02 19:39:04 +00:00
# test_create_replica_after_delay
main_node.query("DROP TABLE alters_from_different_replicas.concurrent_test SYNC")
main_node.query(
"CREATE TABLE alters_from_different_replicas.concurrent_test "
"(CounterID UInt32, StartDate Date, UserID UInt32, VisitID UInt32, NestedColumn Nested(A UInt8, S String), ToDrop UInt32) "
"ENGINE = ReplicatedMergeTree ORDER BY CounterID;"
)
2020-10-26 15:12:16 +00:00
expected = (
"CREATE TABLE alters_from_different_replicas.concurrent_test\\n(\\n `CounterID` UInt32,\\n `StartDate` Date,\\n `UserID` UInt32,\\n"
" `VisitID` UInt32,\\n `NestedColumn.A` Array(UInt8),\\n `NestedColumn.S` Array(String),\\n `ToDrop` UInt32\\n)\\n"
2022-04-13 14:51:59 +00:00
"ENGINE = ReplicatedMergeTree(\\'/clickhouse/tables/{uuid}/{shard}\\', \\'{replica}\\')\\nORDER BY CounterID\\nSETTINGS index_granularity = 8192"
)
2020-10-26 15:12:16 +00:00
2023-03-02 17:35:56 +00:00
assert_create_query(
[main_node, competing_node],
"alters_from_different_replicas.concurrent_test",
expected,
)
2020-06-20 15:38:20 +00:00
main_node.query(
"INSERT INTO alters_from_different_replicas.dist (CounterID, StartDate, UserID) SELECT number, addDays(toDate('2020-02-02'), number), intHash32(number) FROM numbers(10)"
)
2021-02-02 19:39:04 +00:00
# test_replica_restart
main_node.restart_clickhouse()
2020-10-26 15:12:16 +00:00
expected = (
"CREATE TABLE alters_from_different_replicas.concurrent_test\\n(\\n `CounterID` UInt32,\\n `StartDate` Date,\\n `UserID` UInt32,\\n"
" `VisitID` UInt32,\\n `NestedColumn.A` Array(UInt8),\\n `NestedColumn.S` Array(String),\\n `ToDrop` UInt32\\n)\\n"
2022-04-13 14:51:59 +00:00
"ENGINE = ReplicatedMergeTree(\\'/clickhouse/tables/{uuid}/{shard}\\', \\'{replica}\\')\\nORDER BY CounterID\\nSETTINGS index_granularity = 8192"
)
2021-02-02 19:39:04 +00:00
# test_snapshot_and_snapshot_recover
snapshotting_node.query(
"CREATE DATABASE alters_from_different_replicas ENGINE = Replicated('/test/alters_from_different_replicas', 'shard2', 'replica1');"
)
snapshot_recovering_node.query(
"CREATE DATABASE alters_from_different_replicas ENGINE = Replicated('/test/alters_from_different_replicas', 'shard2', 'replica2');"
)
2023-03-02 17:35:56 +00:00
assert_create_query(
all_nodes, "alters_from_different_replicas.concurrent_test", expected
)
2020-11-27 14:04:03 +00:00
main_node.query("SYSTEM FLUSH DISTRIBUTED alters_from_different_replicas.dist")
main_node.query(
"ALTER TABLE alters_from_different_replicas.concurrent_test UPDATE StartDate = addYears(StartDate, 1) WHERE 1"
)
2023-03-02 17:35:56 +00:00
res = main_node.query(
"ALTER TABLE alters_from_different_replicas.concurrent_test DELETE WHERE UserID % 2"
)
assert (
2022-06-29 14:27:21 +00:00
"shard1\treplica1\tOK" in res
and "shard1\treplica2\tOK" in res
and "shard1\treplica3\tOK" in res
)
2022-06-29 14:27:21 +00:00
assert "shard2\treplica1\tOK" in res and "shard2\treplica2\tOK" in res
2021-02-08 19:36:17 +00:00
expected = (
"1\t1\tmain_node\n"
"1\t2\tdummy_node\n"
"1\t3\tcompeting_node\n"
"2\t1\tsnapshotting_node\n"
"2\t2\tsnapshot_recovering_node\n"
)
assert (
main_node.query(
"SELECT shard_num, replica_num, host_name FROM system.clusters WHERE cluster='alters_from_different_replicas'"
)
== expected
)
2020-11-29 11:45:32 +00:00
2021-02-02 19:39:04 +00:00
# test_drop_and_create_replica
main_node.query("DROP DATABASE alters_from_different_replicas SYNC")
main_node.query(
"CREATE DATABASE alters_from_different_replicas ENGINE = Replicated('/test/alters_from_different_replicas', 'shard1', 'replica1');"
)
2020-10-26 15:12:16 +00:00
expected = (
"CREATE TABLE alters_from_different_replicas.concurrent_test\\n(\\n `CounterID` UInt32,\\n `StartDate` Date,\\n `UserID` UInt32,\\n"
" `VisitID` UInt32,\\n `NestedColumn.A` Array(UInt8),\\n `NestedColumn.S` Array(String),\\n `ToDrop` UInt32\\n)\\n"
2022-04-13 14:51:59 +00:00
"ENGINE = ReplicatedMergeTree(\\'/clickhouse/tables/{uuid}/{shard}\\', \\'{replica}\\')\\nORDER BY CounterID\\nSETTINGS index_granularity = 8192"
)
2020-10-26 15:12:16 +00:00
2023-03-02 17:35:56 +00:00
assert_create_query(
[main_node, competing_node],
"alters_from_different_replicas.concurrent_test",
expected,
)
assert_create_query(
all_nodes, "alters_from_different_replicas.concurrent_test", expected
)
2021-02-02 19:39:04 +00:00
for node in all_nodes:
node.query("SYSTEM SYNC REPLICA alters_from_different_replicas.concurrent_test")
2021-02-02 19:39:04 +00:00
expected = (
"0\t2021-02-02\t4249604106\n"
"1\t2021-02-03\t1343103100\n"
"4\t2021-02-06\t3902320246\n"
"7\t2021-02-09\t3844986530\n"
"9\t2021-02-11\t1241149650\n"
)
2020-10-26 15:12:16 +00:00
assert_eq_with_retry(
dummy_node,
"SELECT CounterID, StartDate, UserID FROM alters_from_different_replicas.dist ORDER BY CounterID",
expected,
)
main_node.query("DROP DATABASE alters_from_different_replicas SYNC")
dummy_node.query("DROP DATABASE alters_from_different_replicas SYNC")
competing_node.query("DROP DATABASE alters_from_different_replicas SYNC")
snapshotting_node.query("DROP DATABASE alters_from_different_replicas SYNC")
snapshot_recovering_node.query("DROP DATABASE alters_from_different_replicas SYNC")
2020-06-20 15:38:20 +00:00
2022-07-29 16:33:16 +00:00
def create_some_tables(db):
settings = {
"distributed_ddl_task_timeout": 0,
"allow_experimental_object_type": 1,
"allow_suspicious_codecs": 1,
}
main_node.query(f"CREATE TABLE {db}.t1 (n int) ENGINE=Memory", settings=settings)
dummy_node.query(
f"CREATE TABLE {db}.t2 (s String) ENGINE=Memory", settings=settings
)
main_node.query(
f"CREATE TABLE {db}.mt1 (n int) ENGINE=MergeTree order by n",
settings=settings,
)
dummy_node.query(
f"CREATE TABLE {db}.mt2 (n int) ENGINE=MergeTree order by n",
settings=settings,
)
main_node.query(
f"CREATE TABLE {db}.rmt1 (n int) ENGINE=ReplicatedMergeTree order by n",
settings=settings,
)
dummy_node.query(
f"CREATE TABLE {db}.rmt2 (n int CODEC(ZSTD, ZSTD, ZSTD(12), LZ4HC(12))) ENGINE=ReplicatedMergeTree order by n",
settings=settings,
)
main_node.query(
f"CREATE TABLE {db}.rmt3 (n int, json Object('json') materialized '') ENGINE=ReplicatedMergeTree order by n",
settings=settings,
)
dummy_node.query(
f"CREATE TABLE {db}.rmt5 (n int) ENGINE=ReplicatedMergeTree order by n",
settings=settings,
)
main_node.query(
f"CREATE MATERIALIZED VIEW {db}.mv1 (n int) ENGINE=ReplicatedMergeTree order by n AS SELECT n FROM recover.rmt1",
settings=settings,
)
dummy_node.query(
f"CREATE MATERIALIZED VIEW {db}.mv2 (n int) ENGINE=ReplicatedMergeTree order by n AS SELECT n FROM recover.rmt2",
settings=settings,
)
main_node.query(
f"CREATE DICTIONARY {db}.d1 (n int DEFAULT 0, m int DEFAULT 1) PRIMARY KEY n "
"SOURCE(CLICKHOUSE(HOST 'localhost' PORT 9000 USER 'default' TABLE 'rmt1' PASSWORD '' DB 'recover')) "
"LIFETIME(MIN 1 MAX 10) LAYOUT(FLAT())"
)
dummy_node.query(
f"CREATE DICTIONARY {db}.d2 (n int DEFAULT 0, m int DEFAULT 1) PRIMARY KEY n "
"SOURCE(CLICKHOUSE(HOST 'localhost' PORT 9000 USER 'default' TABLE 'rmt2' PASSWORD '' DB 'recover')) "
"LIFETIME(MIN 1 MAX 10) LAYOUT(FLAT())"
)
2022-07-29 16:33:16 +00:00
# These tables are used to check that DatabaseReplicated correctly renames all the tables in case when it restores from the lost state
def create_table_for_exchanges(db):
settings = {"distributed_ddl_task_timeout": 0}
for table in ["a1", "a2", "a3", "a4", "a5", "a6"]:
main_node.query(
f"CREATE TABLE {db}.{table} (s String) ENGINE=ReplicatedMergeTree order by s",
settings=settings,
)
2022-07-29 16:33:16 +00:00
def test_recover_staled_replica(started_cluster):
main_node.query(
"CREATE DATABASE recover ENGINE = Replicated('/clickhouse/databases/recover', 'shard1', 'replica1');"
)
started_cluster.get_kazoo_client("zoo1").set(
"/clickhouse/databases/recover/logs_to_keep", b"10"
)
dummy_node.query(
"CREATE DATABASE recover ENGINE = Replicated('/clickhouse/databases/recover', 'shard1', 'replica2');"
)
2022-07-29 16:33:16 +00:00
settings = {"distributed_ddl_task_timeout": 0}
create_some_tables("recover")
create_table_for_exchanges("recover")
2022-07-29 16:33:16 +00:00
for table in ["t1", "t2", "mt1", "mt2", "rmt1", "rmt2", "rmt3", "rmt5"]:
main_node.query(f"INSERT INTO recover.{table} VALUES (42)")
for table in ["t1", "t2", "mt1", "mt2"]:
dummy_node.query(f"INSERT INTO recover.{table} VALUES (42)")
for i, table in enumerate(["a1", "a2", "a3", "a4", "a5", "a6"]):
main_node.query(f"INSERT INTO recover.{table} VALUES ('{str(i + 1) * 10}')")
for table in ["rmt1", "rmt2", "rmt3", "rmt5"]:
main_node.query(f"SYSTEM SYNC REPLICA recover.{table}")
for table in ["a1", "a2", "a3", "a4", "a5", "a6"]:
main_node.query(f"SYSTEM SYNC REPLICA recover.{table}")
2021-02-11 22:23:40 +00:00
with PartitionManager() as pm:
pm.drop_instance_zk_connections(dummy_node)
dummy_node.query_and_get_error("RENAME TABLE recover.t1 TO recover.m1")
main_node.query_with_retry(
"RENAME TABLE recover.t1 TO recover.m1", settings=settings
)
main_node.query_with_retry(
"ALTER TABLE recover.mt1 ADD COLUMN m int", settings=settings
)
main_node.query_with_retry(
"ALTER TABLE recover.rmt1 ADD COLUMN m int", settings=settings
)
main_node.query_with_retry(
"RENAME TABLE recover.rmt3 TO recover.rmt4", settings=settings
)
2021-03-16 10:00:49 +00:00
main_node.query_with_retry("DROP TABLE recover.rmt5", settings=settings)
main_node.query_with_retry("DROP DICTIONARY recover.d2", settings=settings)
main_node.query_with_retry(
"CREATE DICTIONARY recover.d2 (n int DEFAULT 0, m int DEFAULT 1) PRIMARY KEY n "
"SOURCE(CLICKHOUSE(HOST 'localhost' PORT 9000 USER 'default' TABLE 'rmt1' PASSWORD '' DB 'recover')) "
"LIFETIME(MIN 1 MAX 10) LAYOUT(FLAT());",
settings=settings,
)
inner_table = (
".inner_id."
+ dummy_node.query_with_retry(
"SELECT uuid FROM system.tables WHERE database='recover' AND name='mv1'"
).strip()
)
main_node.query_with_retry(
f"ALTER TABLE recover.`{inner_table}` MODIFY COLUMN n int DEFAULT 42",
settings=settings,
)
main_node.query_with_retry(
2023-11-30 18:42:12 +00:00
"ALTER TABLE recover.mv1 MODIFY QUERY SELECT m as n FROM recover.rmt1",
settings=settings,
)
main_node.query_with_retry(
"RENAME TABLE recover.mv2 TO recover.mv3",
settings=settings,
)
main_node.query_with_retry(
"CREATE TABLE recover.tmp AS recover.m1", settings=settings
)
2021-03-16 10:00:49 +00:00
main_node.query_with_retry("DROP TABLE recover.tmp", settings=settings)
main_node.query_with_retry(
"CREATE TABLE recover.tmp AS recover.m1", settings=settings
)
2021-03-16 10:00:49 +00:00
main_node.query_with_retry("DROP TABLE recover.tmp", settings=settings)
main_node.query_with_retry(
"CREATE TABLE recover.tmp AS recover.m1", settings=settings
)
main_node.query("EXCHANGE TABLES recover.a1 AND recover.a2", settings=settings)
main_node.query("EXCHANGE TABLES recover.a3 AND recover.a4", settings=settings)
main_node.query("EXCHANGE TABLES recover.a5 AND recover.a4", settings=settings)
main_node.query("EXCHANGE TABLES recover.a6 AND recover.a3", settings=settings)
main_node.query("RENAME TABLE recover.a6 TO recover.a7", settings=settings)
main_node.query("RENAME TABLE recover.a1 TO recover.a8", settings=settings)
assert (
main_node.query(
"SELECT name FROM system.tables WHERE database='recover' AND name NOT LIKE '.inner_id.%' ORDER BY name"
)
== "a2\na3\na4\na5\na7\na8\nd1\nd2\nm1\nmt1\nmt2\nmv1\nmv3\nrmt1\nrmt2\nrmt4\nt2\ntmp\n"
)
query = (
"SELECT name, uuid, create_table_query FROM system.tables WHERE database='recover' AND name NOT LIKE '.inner_id.%' "
"ORDER BY name SETTINGS show_table_uuid_in_table_create_query_if_not_nil=1"
)
2021-02-11 22:23:40 +00:00
expected = main_node.query(query)
assert_eq_with_retry(dummy_node, query, expected)
assert (
main_node.query(
"SELECT count() FROM system.tables WHERE database='recover' AND name LIKE '.inner_id.%'"
)
== "2\n"
)
assert (
dummy_node.query(
"SELECT count() FROM system.tables WHERE database='recover' AND name LIKE '.inner_id.%'"
)
== "2\n"
)
2021-02-11 22:23:40 +00:00
# Check that Database Replicated renamed all the tables correctly
for i, table in enumerate(["a2", "a8", "a5", "a7", "a4", "a3"]):
assert (
dummy_node.query(f"SELECT * FROM recover.{table}") == f"{str(i + 1) * 10}\n"
)
for table in [
"m1",
"t2",
"mt1",
"mt2",
"rmt1",
"rmt2",
"rmt4",
"d1",
"d2",
"mv1",
"mv3",
]:
assert main_node.query(f"SELECT (*,).1 FROM recover.{table}") == "42\n"
for table in ["t2", "rmt1", "rmt2", "rmt4", "d1", "d2", "mt2", "mv1", "mv3"]:
assert (
dummy_node.query(f"SELECT '{table}', (*,).1 FROM recover.{table}")
== f"{table}\t42\n"
)
for table in ["m1", "mt1"]:
assert dummy_node.query(f"SELECT count() FROM recover.{table}") == "0\n"
2021-11-15 14:07:08 +00:00
global test_recover_staled_replica_run
assert (
dummy_node.query(
"SELECT count() FROM system.tables WHERE database='recover_broken_tables'"
)
== f"{test_recover_staled_replica_run}\n"
)
assert (
2022-04-14 13:31:49 +00:00
dummy_node.query(
"SELECT count() FROM system.tables WHERE database='recover_broken_replicated_tables'"
)
== f"{test_recover_staled_replica_run}\n"
)
2021-11-15 14:07:08 +00:00
test_recover_staled_replica_run += 1
print(dummy_node.query("SHOW DATABASES"))
print(dummy_node.query("SHOW TABLES FROM recover_broken_tables"))
print(dummy_node.query("SHOW TABLES FROM recover_broken_replicated_tables"))
table = dummy_node.query(
"SHOW TABLES FROM recover_broken_tables LIKE 'mt1_41_%' LIMIT 1"
).strip()
assert (
dummy_node.query(f"SELECT (*,).1 FROM recover_broken_tables.{table}") == "42\n"
)
table = dummy_node.query(
"SHOW TABLES FROM recover_broken_replicated_tables LIKE 'rmt5_41_%' LIMIT 1"
).strip()
assert (
dummy_node.query(f"SELECT (*,).1 FROM recover_broken_replicated_tables.{table}")
== "42\n"
)
2021-02-11 22:23:40 +00:00
2021-03-08 17:26:38 +00:00
expected = "Cleaned 6 outdated objects: dropped 1 dictionaries and 3 tables, moved 2 tables"
2021-02-11 22:23:40 +00:00
assert_logs_contain(dummy_node, expected)
dummy_node.query("DROP TABLE recover.tmp")
assert_eq_with_retry(
main_node,
"SELECT count() FROM system.tables WHERE database='recover' AND name='tmp'",
"0\n",
)
2021-11-11 08:12:54 +00:00
main_node.query("DROP DATABASE recover SYNC")
dummy_node.query("DROP DATABASE recover SYNC")
2021-02-11 22:23:40 +00:00
def test_recover_staled_replica_many_mvs(started_cluster):
main_node.query("DROP DATABASE IF EXISTS recover_mvs")
dummy_node.query("DROP DATABASE IF EXISTS recover_mvs")
main_node.query_with_retry(
"CREATE DATABASE IF NOT EXISTS recover_mvs ENGINE = Replicated('/clickhouse/databases/recover_mvs', 'shard1', 'replica1');"
)
started_cluster.get_kazoo_client("zoo1").set(
"/clickhouse/databases/recover_mvs/logs_to_keep", b"10"
)
dummy_node.query_with_retry(
"CREATE DATABASE IF NOT EXISTS recover_mvs ENGINE = Replicated('/clickhouse/databases/recover_mvs', 'shard1', 'replica2');"
)
settings = {"distributed_ddl_task_timeout": 0}
with PartitionManager() as pm:
pm.drop_instance_zk_connections(dummy_node)
dummy_node.query_and_get_error("RENAME TABLE recover_mvs.t1 TO recover_mvs.m1")
for identifier in ["1", "2", "3", "4"]:
main_node.query(
f"CREATE TABLE recover_mvs.rmt{identifier} (n int) ENGINE=ReplicatedMergeTree ORDER BY n",
settings=settings,
)
print("Created tables")
for identifier in ["1", "2", "3", "4"]:
main_node.query(
f"CREATE TABLE recover_mvs.mv_inner{identifier} (n int) ENGINE=ReplicatedMergeTree ORDER BY n",
settings=settings,
)
for identifier in ["1", "2", "3", "4"]:
main_node.query_with_retry(
f"""CREATE MATERIALIZED VIEW recover_mvs.mv{identifier}
TO recover_mvs.mv_inner{identifier}
AS SELECT * FROM recover_mvs.rmt{identifier}""",
settings=settings,
)
print("Created MVs")
for identifier in ["1", "2", "3", "4"]:
main_node.query_with_retry(
f"""CREATE VIEW recover_mvs.view_from_mv{identifier}
AS SELECT * FROM recover_mvs.mv{identifier}""",
settings=settings,
)
print("Created Views on top of MVs")
for identifier in ["1", "2", "3", "4"]:
main_node.query_with_retry(
f"""CREATE MATERIALIZED VIEW recover_mvs.cascade_mv{identifier}
ENGINE=MergeTree() ORDER BY tuple()
POPULATE AS SELECT * FROM recover_mvs.mv_inner{identifier};""",
settings=settings,
)
print("Created cascade MVs")
for identifier in ["1", "2", "3", "4"]:
main_node.query_with_retry(
f"""CREATE VIEW recover_mvs.view_from_cascade_mv{identifier}
AS SELECT * FROM recover_mvs.cascade_mv{identifier}""",
settings=settings,
)
print("Created Views on top of cascade MVs")
for identifier in ["1", "2", "3", "4"]:
main_node.query_with_retry(
f"""CREATE MATERIALIZED VIEW recover_mvs.double_cascade_mv{identifier}
ENGINE=MergeTree() ORDER BY tuple()
POPULATE AS SELECT * FROM recover_mvs.`.inner_id.{get_table_uuid("recover_mvs", f"cascade_mv{identifier}")}`""",
settings=settings,
)
print("Created double cascade MVs")
for identifier in ["1", "2", "3", "4"]:
main_node.query_with_retry(
f"""CREATE VIEW recover_mvs.view_from_double_cascade_mv{identifier}
AS SELECT * FROM recover_mvs.double_cascade_mv{identifier}""",
settings=settings,
)
print("Created Views on top of double cascade MVs")
# This weird table name is actually makes sence because it starts with letter `a` and may break some internal sorting
main_node.query_with_retry(
"""
CREATE VIEW recover_mvs.anime
AS
SELECT n
FROM
(
SELECT *
FROM
(
SELECT *
FROM
(
SELECT *
FROM recover_mvs.mv_inner1 AS q1
INNER JOIN recover_mvs.mv_inner2 AS q2 ON q1.n = q2.n
) AS new_table_1
INNER JOIN recover_mvs.mv_inner3 AS q3 ON new_table_1.n = q3.n
) AS new_table_2
INNER JOIN recover_mvs.mv_inner4 AS q4 ON new_table_2.n = q4.n
)
""",
settings=settings,
)
print("Created final boss")
for identifier in ["1", "2", "3", "4"]:
main_node.query_with_retry(
f"""CREATE DICTIONARY recover_mvs.`11111d{identifier}` (n UInt64)
PRIMARY KEY n
SOURCE(CLICKHOUSE(HOST 'localhost' PORT tcpPort() TABLE 'double_cascade_mv{identifier}' DB 'recover_mvs'))
LAYOUT(FLAT()) LIFETIME(1)""",
settings=settings,
)
print("Created dictionaries")
for identifier in ["1", "2", "3", "4"]:
main_node.query_with_retry(
f"""CREATE VIEW recover_mvs.`00000vd{identifier}`
AS SELECT * FROM recover_mvs.`11111d{identifier}`""",
settings=settings,
)
print("Created Views on top of dictionaries")
dummy_node.query("SYSTEM SYNC DATABASE REPLICA recover_mvs")
query = "SELECT name FROM system.tables WHERE database='recover_mvs' ORDER BY name"
assert main_node.query(query) == dummy_node.query(query)
main_node.query("DROP DATABASE IF EXISTS recover_mvs")
dummy_node.query("DROP DATABASE IF EXISTS recover_mvs")
2021-02-15 00:04:46 +00:00
def test_startup_without_zk(started_cluster):
with PartitionManager() as pm:
pm.drop_instance_zk_connections(main_node)
err = main_node.query_and_get_error(
"CREATE DATABASE startup ENGINE = Replicated('/clickhouse/databases/startup', 'shard1', 'replica1');"
)
Fix test_replicated_database::test_startup_without_zk flakiness CI founds [1] E assert 'ZooKeeper' in "Received exception from server (version 24.1.1):\nCode: 999. DB::Exception: Received from 172.16.1.6:9000. Coordination::Exception. Coordination::Exception: Coordination error: Connection loss, path /clickhouse/databases/startup. Stack trace:\n\n0. ? @ 0x000000000c734f1b in /usr/bin/clickhouse\n1. ? @ 0x000000000fbef3f4 in /usr/bin/clickhouse\n2. ? @ 0x000000000fbeebc8 in /usr/bin/clickhouse\n3. ? @ 0x0000000012ded0f9 in /usr/bin/clickhouse\n4. ? @ 0x0000000012deaa71 in /usr/bin/clickhouse\n5. ? @ 0x00000000107e51e8 in /usr/bin/clickhouse\n6. ? @ 0x00000000107eb87a in /usr/bin/clickhouse\n7. ? @ 0x0000000011213ec3 in /usr/bin/clickhouse\n8. ? @ 0x00000000111e9acf in /usr/bin/clickhouse\n9. ? @ 0x000000001120e85e in /usr/bin/clickhouse\n10. ? @ 0x000000001174f0a2 in /usr/bin/clickhouse\n11. ? @ 0x0000000011748c9a in /usr/bin/clickhouse\n12. ? @ 0x00000000126fb5a9 in /usr/bin/clickhouse\n13. ? @ 0x00000000127130f9 in /usr/bin/clickhouse\n14. ? @ 0x000000001517bf12 in /usr/bin/clickhouse\n15. ? @ 0x000000001517cd11 in /usr/bin/clickhouse\n16. ? @ 0x00000000152740c7 in /usr/bin/clickhouse\n17. ? @ 0x00000000152726bc in /usr/bin/clickhouse\n18. ? @ 0x00007f7f36485ac3\n19. ? @ 0x00007f7f36517660\n. (KEEPER_EXCEPTION)\n(query: CREATE DATABASE startup ENGINE = Replicated('/clickhouse/databases/startup', 'shard1', 'replica1');)\n" [1]: https://s3.amazonaws.com/clickhouse-test-reports/58610/441a3fbb9b1c088f528841727f013b829f63fc8a/integration_tests__release__[3_4]/integration_run_test_replicated_database_test_py_0.log Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
2024-01-10 11:43:30 +00:00
assert "ZooKeeper" in err or "Coordination::Exception" in err
main_node.query(
"CREATE DATABASE startup ENGINE = Replicated('/clickhouse/databases/startup', 'shard1', 'replica1');"
)
2022-08-12 12:16:50 +00:00
main_node.query(
"CREATE TABLE startup.rmt (n int) ENGINE=ReplicatedMergeTree order by n"
)
2021-02-15 00:04:46 +00:00
main_node.query("INSERT INTO startup.rmt VALUES (42)")
with PartitionManager() as pm:
pm.drop_instance_zk_connections(main_node)
2022-08-20 11:52:10 +00:00
main_node.restart_clickhouse(stop_start_wait_sec=60)
2021-02-15 00:04:46 +00:00
assert main_node.query("SELECT (*,).1 FROM startup.rmt") == "42\n"
# we need to wait until the table is not readonly
2022-08-19 09:39:59 +00:00
main_node.query_with_retry("INSERT INTO startup.rmt VALUES(42)")
2022-08-19 09:39:59 +00:00
main_node.query_with_retry("CREATE TABLE startup.m (n int) ENGINE=Memory")
2021-02-15 00:04:46 +00:00
main_node.query("EXCHANGE TABLES startup.rmt AND startup.m")
assert main_node.query("SELECT (*,).1 FROM startup.m") == "42\n"
2021-11-11 13:36:10 +00:00
main_node.query("DROP DATABASE startup SYNC")
2021-08-16 18:30:53 +00:00
2021-08-16 18:30:53 +00:00
def test_server_uuid(started_cluster):
uuid1 = main_node.query("select serverUUID()")
uuid2 = dummy_node.query("select serverUUID()")
assert uuid1 != uuid2
main_node.restart_clickhouse()
uuid1_after_restart = main_node.query("select serverUUID()")
assert uuid1 == uuid1_after_restart
def test_sync_replica(started_cluster):
main_node.query(
"CREATE DATABASE test_sync_database ENGINE = Replicated('/test/sync_replica', 'shard1', 'replica1');"
)
dummy_node.query(
"CREATE DATABASE test_sync_database ENGINE = Replicated('/test/sync_replica', 'shard1', 'replica2');"
)
number_of_tables = 1000
settings = {"distributed_ddl_task_timeout": 0}
with PartitionManager() as pm:
pm.drop_instance_zk_connections(dummy_node)
for i in range(number_of_tables):
main_node.query(
"CREATE TABLE test_sync_database.table_{} (n int) ENGINE=MergeTree order by n".format(
i
),
settings=settings,
)
2022-05-06 16:37:20 +00:00
# wait for host to reconnect
dummy_node.query_with_retry("SELECT * FROM system.zookeeper WHERE path='/'")
dummy_node.query("SYSTEM SYNC DATABASE REPLICA test_sync_database")
2024-03-08 16:43:10 +00:00
assert "2\n" == main_node.query(
"SELECT sum(is_active) FROM system.clusters WHERE cluster='test_sync_database'"
)
assert dummy_node.query(
"SELECT count() FROM system.tables where database='test_sync_database'"
).strip() == str(number_of_tables)
assert main_node.query(
"SELECT count() FROM system.tables where database='test_sync_database'"
).strip() == str(number_of_tables)
2022-05-06 16:37:20 +00:00
engine_settings = {"default_table_engine": "ReplicatedMergeTree"}
dummy_node.query(
"CREATE TABLE test_sync_database.table (n int, primary key n) partition by n",
settings=engine_settings,
)
main_node.query("INSERT INTO test_sync_database.table SELECT * FROM numbers(10)")
dummy_node.query("TRUNCATE TABLE test_sync_database.table", settings=settings)
dummy_node.query(
"ALTER TABLE test_sync_database.table ADD COLUMN m int", settings=settings
)
main_node.query(
"SYSTEM SYNC DATABASE REPLICA ON CLUSTER test_sync_database test_sync_database"
)
lp1 = main_node.query(
"select value from system.zookeeper where path='/test/sync_replica/replicas/shard1|replica1' and name='log_ptr'"
2022-05-06 16:37:20 +00:00
)
lp2 = main_node.query(
"select value from system.zookeeper where path='/test/sync_replica/replicas/shard1|replica2' and name='log_ptr'"
2022-05-06 16:37:20 +00:00
)
max_lp = main_node.query(
"select value from system.zookeeper where path='/test/sync_replica/' and name='max_log_ptr'"
2022-05-06 16:37:20 +00:00
)
assert lp1 == max_lp
assert lp2 == max_lp
2022-07-27 17:15:00 +00:00
main_node.query("DROP DATABASE test_sync_database SYNC")
dummy_node.query("DROP DATABASE test_sync_database SYNC")
2022-07-27 17:15:00 +00:00
def test_force_synchronous_settings(started_cluster):
main_node.query(
"CREATE DATABASE test_force_synchronous_settings ENGINE = Replicated('/clickhouse/databases/test2', 'shard1', 'replica1');"
)
dummy_node.query(
"CREATE DATABASE test_force_synchronous_settings ENGINE = Replicated('/clickhouse/databases/test2', 'shard1', 'replica2');"
)
snapshotting_node.query(
"CREATE DATABASE test_force_synchronous_settings ENGINE = Replicated('/clickhouse/databases/test2', 'shard2', 'replica1');"
)
main_node.query(
"CREATE TABLE test_force_synchronous_settings.t (n int) ENGINE=ReplicatedMergeTree('/test/same/path/{shard}', '{replica}') ORDER BY tuple()"
)
main_node.query(
"INSERT INTO test_force_synchronous_settings.t SELECT * FROM numbers(10)"
)
snapshotting_node.query(
"INSERT INTO test_force_synchronous_settings.t SELECT * FROM numbers(10)"
)
snapshotting_node.query(
"SYSTEM SYNC DATABASE REPLICA test_force_synchronous_settings"
)
dummy_node.query("SYSTEM SYNC DATABASE REPLICA test_force_synchronous_settings")
snapshotting_node.query("SYSTEM STOP MERGES test_force_synchronous_settings.t")
def start_merges_func():
time.sleep(5)
snapshotting_node.query("SYSTEM START MERGES test_force_synchronous_settings.t")
start_merges_thread = threading.Thread(target=start_merges_func)
start_merges_thread.start()
settings = {
"mutations_sync": 2,
"database_replicated_enforce_synchronous_settings": 1,
}
main_node.query(
"ALTER TABLE test_force_synchronous_settings.t UPDATE n = n * 10 WHERE 1",
settings=settings,
)
assert "10\t450\n" == snapshotting_node.query(
"SELECT count(), sum(n) FROM test_force_synchronous_settings.t"
)
start_merges_thread.join()
def select_func():
dummy_node.query(
"SELECT sleepEachRow(1) FROM test_force_synchronous_settings.t SETTINGS function_sleep_max_microseconds_per_block = 0"
2022-07-27 17:15:00 +00:00
)
select_thread = threading.Thread(target=select_func)
select_thread.start()
settings = {"database_replicated_enforce_synchronous_settings": 1}
snapshotting_node.query(
"DROP TABLE test_force_synchronous_settings.t SYNC", settings=settings
)
main_node.query(
"CREATE TABLE test_force_synchronous_settings.t (n String) ENGINE=ReplicatedMergeTree('/test/same/path/{shard}', '{replica}') ORDER BY tuple()"
)
select_thread.join()
2022-07-29 16:33:16 +00:00
def test_recover_digest_mismatch(started_cluster):
main_node.query("DROP DATABASE IF EXISTS recover_digest_mismatch")
dummy_node.query("DROP DATABASE IF EXISTS recover_digest_mismatch")
2022-07-29 16:33:16 +00:00
main_node.query(
"CREATE DATABASE recover_digest_mismatch ENGINE = Replicated('/clickhouse/databases/recover_digest_mismatch', 'shard1', 'replica1');"
)
dummy_node.query(
"CREATE DATABASE recover_digest_mismatch ENGINE = Replicated('/clickhouse/databases/recover_digest_mismatch', 'shard1', 'replica2');"
)
create_some_tables("recover_digest_mismatch")
main_node.query("SYSTEM SYNC DATABASE REPLICA recover_digest_mismatch")
dummy_node.query("SYSTEM SYNC DATABASE REPLICA recover_digest_mismatch")
2022-07-29 16:33:16 +00:00
ways_to_corrupt_metadata = [
"mv /var/lib/clickhouse/metadata/recover_digest_mismatch/t1.sql /var/lib/clickhouse/metadata/recover_digest_mismatch/m1.sql",
"sed --follow-symlinks -i 's/Int32/String/' /var/lib/clickhouse/metadata/recover_digest_mismatch/mv1.sql",
"rm -f /var/lib/clickhouse/metadata/recover_digest_mismatch/d1.sql",
2023-08-14 17:06:32 +00:00
"rm -rf /var/lib/clickhouse/metadata/recover_digest_mismatch/", # Will trigger "Directory already exists"
"rm -rf /var/lib/clickhouse/store",
2022-07-29 16:33:16 +00:00
]
for command in ways_to_corrupt_metadata:
print(f"Corrupting data using `{command}`")
2022-07-29 16:33:16 +00:00
need_remove_is_active_node = "rm -rf" in command
dummy_node.stop_clickhouse(kill=not need_remove_is_active_node)
dummy_node.exec_in_container(["bash", "-c", command])
query = (
"SELECT name, uuid, create_table_query FROM system.tables WHERE database='recover_digest_mismatch' AND name NOT LIKE '.inner_id.%' "
"ORDER BY name SETTINGS show_table_uuid_in_table_create_query_if_not_nil=1"
)
expected = main_node.query(query)
if need_remove_is_active_node:
2022-07-29 16:33:16 +00:00
# NOTE Otherwise it fails to recreate ReplicatedMergeTree table due to "Replica already exists"
main_node.query(
"SYSTEM DROP REPLICA '2' FROM DATABASE recover_digest_mismatch"
)
# There is a race condition between deleting active node and creating it on server startup
# So we start a server only after we deleted all table replicas from the Keeper
dummy_node.start_clickhouse()
2022-07-29 16:33:16 +00:00
assert_eq_with_retry(dummy_node, query, expected)
main_node.query("DROP DATABASE IF EXISTS recover_digest_mismatch")
dummy_node.query("DROP DATABASE IF EXISTS recover_digest_mismatch")
print("Everything Okay")
2023-06-22 17:26:42 +00:00
def test_replicated_table_structure_alter(started_cluster):
main_node.query("DROP DATABASE IF EXISTS table_structure")
dummy_node.query("DROP DATABASE IF EXISTS table_structure")
main_node.query(
"CREATE DATABASE table_structure ENGINE = Replicated('/clickhouse/databases/table_structure', 'shard1', 'replica1');"
)
dummy_node.query(
"CREATE DATABASE table_structure ENGINE = Replicated('/clickhouse/databases/table_structure', 'shard1', 'replica2');"
)
competing_node.query(
"CREATE DATABASE table_structure ENGINE = Replicated('/clickhouse/databases/table_structure', 'shard1', 'replica3');"
)
competing_node.query("CREATE TABLE table_structure.mem (n int) ENGINE=Memory")
dummy_node.query("DETACH DATABASE table_structure")
settings = {"distributed_ddl_task_timeout": 0}
2023-06-22 17:26:42 +00:00
main_node.query(
"CREATE TABLE table_structure.rmt (n int, v UInt64) ENGINE=ReplicatedReplacingMergeTree(v) ORDER BY n",
settings=settings,
)
competing_node.query("SYSTEM SYNC DATABASE REPLICA table_structure")
competing_node.query("DETACH DATABASE table_structure")
2023-06-22 17:26:42 +00:00
main_node.query(
"ALTER TABLE table_structure.rmt ADD COLUMN m int", settings=settings
)
main_node.query(
"ALTER TABLE table_structure.rmt COMMENT COLUMN v 'version'", settings=settings
)
main_node.query("INSERT INTO table_structure.rmt VALUES (1, 2, 3)")
command = "rm -f /var/lib/clickhouse/metadata/table_structure/mem.sql"
competing_node.exec_in_container(["bash", "-c", command])
competing_node.restart_clickhouse(kill=True)
dummy_node.query("ATTACH DATABASE table_structure")
dummy_node.query("SYSTEM SYNC DATABASE REPLICA table_structure")
dummy_node.query("SYSTEM SYNC REPLICA table_structure.rmt")
assert "1\t2\t3\n" == dummy_node.query("SELECT * FROM table_structure.rmt")
competing_node.query("SYSTEM SYNC DATABASE REPLICA table_structure")
competing_node.query("SYSTEM SYNC REPLICA table_structure.rmt")
2023-06-22 17:26:42 +00:00
# time.sleep(600)
assert "mem" in competing_node.query("SHOW TABLES FROM table_structure")
assert "1\t2\t3\n" == competing_node.query("SELECT * FROM table_structure.rmt")
main_node.query("ALTER TABLE table_structure.rmt ADD COLUMN k int")
main_node.query("INSERT INTO table_structure.rmt VALUES (1, 2, 3, 4)")
dummy_node.query("SYSTEM SYNC DATABASE REPLICA table_structure")
dummy_node.query("SYSTEM SYNC REPLICA table_structure.rmt")
2023-06-22 17:26:42 +00:00
assert "1\t2\t3\t0\n1\t2\t3\t4\n" == dummy_node.query(
"SELECT * FROM table_structure.rmt ORDER BY k"
)
def test_modify_comment(started_cluster):
main_node.query(
"CREATE DATABASE modify_comment_db ENGINE = Replicated('/test/modify_comment', 'shard1', 'replica' || '1');"
)
dummy_node.query(
"CREATE DATABASE modify_comment_db ENGINE = Replicated('/test/modify_comment', 'shard1', 'replica' || '2');"
)
main_node.query(
"CREATE TABLE modify_comment_db.modify_comment_table (d Date, k UInt64, i32 Int32) ENGINE=ReplicatedMergeTree ORDER BY k PARTITION BY toYYYYMM(d);"
)
def restart_verify_not_readonly():
main_node.restart_clickhouse()
assert (
main_node.query(
"SELECT is_readonly FROM system.replicas WHERE table = 'modify_comment_table'"
)
== "0\n"
)
dummy_node.restart_clickhouse()
assert (
dummy_node.query(
"SELECT is_readonly FROM system.replicas WHERE table = 'modify_comment_table'"
)
== "0\n"
)
main_node.query(
"ALTER TABLE modify_comment_db.modify_comment_table COMMENT COLUMN d 'Some comment'"
)
restart_verify_not_readonly()
main_node.query(
"ALTER TABLE modify_comment_db.modify_comment_table MODIFY COMMENT 'Some error comment'"
)
restart_verify_not_readonly()
main_node.query("DROP DATABASE modify_comment_db SYNC")
dummy_node.query("DROP DATABASE modify_comment_db SYNC")
def test_table_metadata_corruption(started_cluster):
main_node.query("DROP DATABASE IF EXISTS table_metadata_corruption")
dummy_node.query("DROP DATABASE IF EXISTS table_metadata_corruption")
main_node.query(
"CREATE DATABASE table_metadata_corruption ENGINE = Replicated('/clickhouse/databases/table_metadata_corruption', 'shard1', 'replica1');"
)
dummy_node.query(
"CREATE DATABASE table_metadata_corruption ENGINE = Replicated('/clickhouse/databases/table_metadata_corruption', 'shard1', 'replica2');"
)
create_some_tables("table_metadata_corruption")
main_node.query("SYSTEM SYNC DATABASE REPLICA table_metadata_corruption")
dummy_node.query("SYSTEM SYNC DATABASE REPLICA table_metadata_corruption")
# Server should handle this by throwing an exception during table loading, which should lead to server shutdown
corrupt = "sed --follow-symlinks -i 's/ReplicatedMergeTree/CorruptedMergeTree/' /var/lib/clickhouse/metadata/table_metadata_corruption/rmt1.sql"
print(f"Corrupting metadata using `{corrupt}`")
dummy_node.stop_clickhouse(kill=True)
dummy_node.exec_in_container(["bash", "-c", corrupt])
query = (
"SELECT name, uuid, create_table_query FROM system.tables WHERE database='table_metadata_corruption' AND name NOT LIKE '.inner_id.%' "
"ORDER BY name SETTINGS show_table_uuid_in_table_create_query_if_not_nil=1"
)
expected = main_node.query(query)
# We expect clickhouse server to shutdown without LOGICAL_ERRORs or deadlocks
dummy_node.start_clickhouse(expected_to_fail=True)
assert not dummy_node.contains_in_log("LOGICAL_ERROR")
fix_corrupt = "sed --follow-symlinks -i 's/CorruptedMergeTree/ReplicatedMergeTree/' /var/lib/clickhouse/metadata/table_metadata_corruption/rmt1.sql"
print(f"Fix corrupted metadata using `{fix_corrupt}`")
dummy_node.exec_in_container(["bash", "-c", fix_corrupt])
dummy_node.start_clickhouse()
assert_eq_with_retry(dummy_node, query, expected)
main_node.query("DROP DATABASE IF EXISTS table_metadata_corruption")
dummy_node.query("DROP DATABASE IF EXISTS table_metadata_corruption")
def test_auto_recovery(started_cluster):
dummy_node.query("DROP DATABASE IF EXISTS auto_recovery")
2024-05-13 13:46:25 +00:00
bad_settings_node.query("DROP DATABASE IF EXISTS auto_recovery")
dummy_node.query(
"CREATE DATABASE auto_recovery ENGINE = Replicated('/clickhouse/databases/auto_recovery', 'shard1', 'replica1');"
)
2024-05-13 13:46:25 +00:00
bad_settings_node.query(
"CREATE DATABASE auto_recovery ENGINE = Replicated('/clickhouse/databases/auto_recovery', 'shard1', 'replica2') SETTINGS max_retries_before_automatic_recovery=3;"
)
dummy_node.query(
"CREATE TABLE auto_recovery.t1 (n int) ENGINE=ReplicatedMergeTree ORDER BY n"
)
dummy_node.query("INSERT INTO auto_recovery.t1 SELECT 42")
# dummy_node has <throw_on_unsupported_query_inside_transaction>0</throw_on_unsupported_query_inside_transaction> (default is 1),
# so it will consider that the setting is changed, and will write it to the DDL entry
2024-05-13 13:46:25 +00:00
# bad_settings_node has implicit_transaction=1, so it will fail and recover from snapshot
dummy_node.query(
"CREATE TABLE auto_recovery.t2 (n int) ENGINE=ReplicatedMergeTree ORDER BY tuple()",
settings={
"throw_on_unsupported_query_inside_transaction": 1,
"distributed_ddl_task_timeout": 0,
},
)
dummy_node.query("INSERT INTO auto_recovery.t2 SELECT 137")
dummy_node.query(
"EXCHANGE TABLES auto_recovery.t1 AND auto_recovery.t2",
settings={"distributed_ddl_task_timeout": 0},
)
2024-05-13 13:46:25 +00:00
bad_settings_node.query(
"SYSTEM SYNC DATABASE REPLICA auto_recovery", settings={"receive_timeout": 60}
)
2024-05-13 13:46:25 +00:00
assert bad_settings_node.contains_in_log(
"Unexpected error (3 times in a row), will try to restart main thread"
)
2024-05-13 13:46:25 +00:00
assert bad_settings_node.contains_in_log("Cannot begin an implicit transaction")
bad_settings_node.query("SYSTEM SYNC REPLICA auto_recovery.t1")
bad_settings_node.query("SYSTEM SYNC REPLICA auto_recovery.t2")
assert "42\n" == dummy_node.query("SELECT * FROM auto_recovery.t2")
assert "137\n" == dummy_node.query("SELECT * FROM auto_recovery.t1")
2024-05-13 13:46:25 +00:00
assert "42\n" == bad_settings_node.query("SELECT * FROM auto_recovery.t2")
assert "137\n" == bad_settings_node.query("SELECT * FROM auto_recovery.t1")
2024-05-28 16:42:10 +00:00
def test_all_groups_cluster(started_cluster):
dummy_node.query("DROP DATABASE IF EXISTS db_cluster")
bad_settings_node.query("DROP DATABASE IF EXISTS db_cluster")
dummy_node.query(
"CREATE DATABASE db_cluster ENGINE = Replicated('/clickhouse/databases/all_groups_cluster', 'shard1', 'replica1');"
)
bad_settings_node.query(
"CREATE DATABASE db_cluster ENGINE = Replicated('/clickhouse/databases/all_groups_cluster', 'shard1', 'replica2');"
)
2024-06-04 18:05:58 +00:00
assert "dummy_node\n" == dummy_node.query(
2024-05-28 16:42:10 +00:00
"select host_name from system.clusters where name='db_cluster' order by host_name"
)
assert "bad_settings_node\n" == bad_settings_node.query(
"select host_name from system.clusters where name='db_cluster' order by host_name"
)
assert "bad_settings_node\ndummy_node\n" == bad_settings_node.query(
"select host_name from system.clusters where name='all_groups.db_cluster' order by host_name"
)