ClickHouse/tests/integration/test_s3_zero_copy_replication/test.py

import datetime
import logging
import time

import pytest
from helpers.cluster import ClickHouseCluster

logging.getLogger().setLevel(logging.INFO)
logging.getLogger().addHandler(logging.StreamHandler())


@pytest.fixture(scope="module")
def cluster():
    try:
        cluster = ClickHouseCluster(__file__)
        cluster.add_instance("node1", main_configs=["configs/config.d/s3.xml"], macros={'replica': '1'},
                             with_minio=True,
                             with_zookeeper=True)
        cluster.add_instance("node2", main_configs=["configs/config.d/s3.xml"], macros={'replica': '2'},
                             with_minio=True,
                             with_zookeeper=True)
        logging.info("Starting cluster...")
        cluster.start()
        logging.info("Cluster started")

        yield cluster
    finally:
        cluster.shutdown()


def get_large_objects_count(cluster, size=100, folder='data'):
    minio = cluster.minio_client
    counter = 0
    for obj in minio.list_objects(cluster.minio_bucket, '{}/'.format(folder)):
        if obj.size >= size:
            counter = counter + 1
    return counter


def wait_for_large_objects_count(cluster, expected, size=100, timeout=30):
    while timeout > 0:
        if get_large_objects_count(cluster, size=size) == expected:
            return
        timeout -= 1
        time.sleep(1)
    assert get_large_objects_count(cluster, size=size) == expected


def wait_for_active_parts(node, num_expected_parts, table_name, timeout=30):
    deadline = time.monotonic() + timeout
    num_parts = 0
    while time.monotonic() < deadline:
        num_parts_str = node.query("select count() from system.parts where table = '{}' and active".format(table_name))
        num_parts = int(num_parts_str.strip())
        if num_parts == num_expected_parts:
            return

        time.sleep(0.2)

    assert num_parts == num_expected_parts


# Result of `get_large_objects_count` can be changed in other tests, so run this case at the beginning
@pytest.mark.order(0)
@pytest.mark.parametrize(
    "policy", ["s3"]
)
def test_s3_zero_copy_replication(cluster, policy):
    node1 = cluster.instances["node1"]
    node2 = cluster.instances["node2"]

    node1.query(
        """
        CREATE TABLE s3_test ON CLUSTER test_cluster (id UInt32, value String)
        ENGINE=ReplicatedMergeTree('/clickhouse/tables/s3_test', '{}')
        ORDER BY id
        SETTINGS storage_policy='{}'
        """
            .format('{replica}', policy)
    )

    node1.query("INSERT INTO s3_test VALUES (0,'data'),(1,'data')")
    node2.query("SYSTEM SYNC REPLICA s3_test")
    assert node1.query("SELECT * FROM s3_test order by id FORMAT Values") == "(0,'data'),(1,'data')"
    assert node2.query("SELECT * FROM s3_test order by id FORMAT Values") == "(0,'data'),(1,'data')"

    # Based on version 21.x - should be only 1 file with size 100+ (checksums.txt), used by both nodes
    assert get_large_objects_count(cluster) == 1

    node2.query("INSERT INTO s3_test VALUES (2,'data'),(3,'data')")
    node1.query("SYSTEM SYNC REPLICA s3_test")

    assert node2.query("SELECT * FROM s3_test order by id FORMAT Values") == "(0,'data'),(1,'data'),(2,'data'),(3,'data')"
    assert node1.query("SELECT * FROM s3_test order by id FORMAT Values") == "(0,'data'),(1,'data'),(2,'data'),(3,'data')"

    # Based on version 21.x - two parts
    wait_for_large_objects_count(cluster, 2)

    node1.query("OPTIMIZE TABLE s3_test FINAL")

    # Based on version 21.x - after merge, two old parts and one merged
    wait_for_large_objects_count(cluster, 3)

    # Based on version 21.x - after cleanup - only one merged part
    wait_for_large_objects_count(cluster, 1, timeout=60)

    node1.query("DROP TABLE IF EXISTS s3_test NO DELAY")
    node2.query("DROP TABLE IF EXISTS s3_test NO DELAY")


def test_s3_zero_copy_on_hybrid_storage(cluster):
    node1 = cluster.instances["node1"]
    node2 = cluster.instances["node2"]

    node1.query(
        """
        CREATE TABLE hybrid_test ON CLUSTER test_cluster (id UInt32, value String)
        ENGINE=ReplicatedMergeTree('/clickhouse/tables/hybrid_test', '{}')
        ORDER BY id
        SETTINGS storage_policy='hybrid'
        """
            .format('{replica}')
    )

    node1.query("INSERT INTO hybrid_test VALUES (0,'data'),(1,'data')")
    node2.query("SYSTEM SYNC REPLICA hybrid_test")

    assert node1.query("SELECT * FROM hybrid_test ORDER BY id FORMAT Values") == "(0,'data'),(1,'data')"
    assert node2.query("SELECT * FROM hybrid_test ORDER BY id FORMAT Values") == "(0,'data'),(1,'data')"

    assert node1.query("SELECT partition_id,disk_name FROM system.parts WHERE table='hybrid_test' FORMAT Values") == "('all','default')"
    assert node2.query("SELECT partition_id,disk_name FROM system.parts WHERE table='hybrid_test' FORMAT Values") == "('all','default')"

    node1.query("ALTER TABLE hybrid_test MOVE PARTITION ID 'all' TO DISK 's31'")

    assert node1.query("SELECT partition_id,disk_name FROM system.parts WHERE table='hybrid_test' FORMAT Values") == "('all','s31')"
    assert node2.query("SELECT partition_id,disk_name FROM system.parts WHERE table='hybrid_test' FORMAT Values") == "('all','default')"

    # Total objects in S3
    s3_objects = get_large_objects_count(cluster, size=0)

    node2.query("ALTER TABLE hybrid_test MOVE PARTITION ID 'all' TO DISK 's31'")

    assert node1.query("SELECT partition_id,disk_name FROM system.parts WHERE table='hybrid_test' FORMAT Values") == "('all','s31')"
    assert node2.query("SELECT partition_id,disk_name FROM system.parts WHERE table='hybrid_test' FORMAT Values") == "('all','s31')"

    # Check that after moving partition on node2 no new obects on s3
    wait_for_large_objects_count(cluster, s3_objects, size=0)

    assert node1.query("SELECT * FROM hybrid_test ORDER BY id FORMAT Values") == "(0,'data'),(1,'data')"
    assert node2.query("SELECT * FROM hybrid_test ORDER BY id FORMAT Values") == "(0,'data'),(1,'data')"

    node1.query("DROP TABLE IF EXISTS hybrid_test NO DELAY")
    node2.query("DROP TABLE IF EXISTS hybrid_test NO DELAY")


def insert_data_time(node, table, number_of_mb, time, start=0):
    values = ','.join(f"({x},{time})" for x in range(start, int((1024 * 1024 * number_of_mb) / 8) + start + 1))
    node.query(f"INSERT INTO {table} VALUES {values}")


def insert_large_data(node, table):
    tm = time.mktime((datetime.date.today() - datetime.timedelta(days=7)).timetuple())
    insert_data_time(node, table, 1, tm, 0)
    tm = time.mktime((datetime.date.today() - datetime.timedelta(days=3)).timetuple())
    insert_data_time(node, table, 1, tm, 1024*1024)
    tm = time.mktime(datetime.date.today().timetuple())
    insert_data_time(node, table, 10, tm, 1024*1024*2)


@pytest.mark.parametrize(
    ("storage_policy", "large_data", "iterations"),
    [
        ("tiered", False, 10),
        ("tiered_copy", False, 10),
        ("tiered", True, 3),
        ("tiered_copy", True, 3),
    ]
)
def test_s3_zero_copy_with_ttl_move(cluster, storage_policy, large_data, iterations):
    node1 = cluster.instances["node1"]
    node2 = cluster.instances["node2"]

    node1.query("DROP TABLE IF EXISTS ttl_move_test NO DELAY")
    node2.query("DROP TABLE IF EXISTS ttl_move_test NO DELAY")

    for i in range(iterations):
        node1.query(
            """
            CREATE TABLE ttl_move_test ON CLUSTER test_cluster (d UInt64, d1 DateTime)
            ENGINE=ReplicatedMergeTree('/clickhouse/tables/ttl_move_test', '{}')
            ORDER BY d
            TTL d1 + INTERVAL 2 DAY TO VOLUME 'external'
            SETTINGS storage_policy='{}'
            """
                .format('{replica}', storage_policy)
        )

        if large_data:
            insert_large_data(node1, 'ttl_move_test')
        else:
            node1.query("INSERT INTO ttl_move_test VALUES (10, now() - INTERVAL 3 DAY)")
            node1.query("INSERT INTO ttl_move_test VALUES (11, now() - INTERVAL 1 DAY)")

        node1.query("OPTIMIZE TABLE ttl_move_test FINAL")
        node2.query("SYSTEM SYNC REPLICA ttl_move_test")

        if large_data:
            assert node1.query("SELECT count() FROM ttl_move_test FORMAT Values") == "(1572867)"
            assert node2.query("SELECT count() FROM ttl_move_test FORMAT Values") == "(1572867)"
        else:
            assert node1.query("SELECT count() FROM ttl_move_test FORMAT Values") == "(2)"
            assert node2.query("SELECT count() FROM ttl_move_test FORMAT Values") == "(2)"
            assert node1.query("SELECT d FROM ttl_move_test ORDER BY d FORMAT Values") == "(10),(11)"
            assert node2.query("SELECT d FROM ttl_move_test ORDER BY d FORMAT Values") == "(10),(11)"

        node1.query("DROP TABLE IF EXISTS ttl_move_test NO DELAY")
        node2.query("DROP TABLE IF EXISTS ttl_move_test NO DELAY")


@pytest.mark.parametrize(
    ("large_data", "iterations"),
    [
        (False, 10),
        (True, 3),
    ]
)
def test_s3_zero_copy_with_ttl_delete(cluster, large_data, iterations):
    node1 = cluster.instances["node1"]
    node2 = cluster.instances["node2"]

    node1.query("DROP TABLE IF EXISTS ttl_delete_test NO DELAY")
    node2.query("DROP TABLE IF EXISTS ttl_delete_test NO DELAY")

    for i in range(iterations):
        node1.query(
            """
            CREATE TABLE ttl_delete_test ON CLUSTER test_cluster (d UInt64, d1 DateTime)
            ENGINE=ReplicatedMergeTree('/clickhouse/tables/ttl_delete_test', '{}')
            ORDER BY d
            TTL d1 + INTERVAL 2 DAY
            SETTINGS storage_policy='tiered'
            """
                .format('{replica}')
        )

        if large_data:
            insert_large_data(node1, 'ttl_delete_test')
        else:
            node1.query("INSERT INTO ttl_delete_test VALUES (10, now() - INTERVAL 3 DAY)")
            node1.query("INSERT INTO ttl_delete_test VALUES (11, now() - INTERVAL 1 DAY)")

        node1.query("OPTIMIZE TABLE ttl_delete_test FINAL")
        node2.query("SYSTEM SYNC REPLICA ttl_delete_test")

        if large_data:
            assert node1.query("SELECT count() FROM ttl_delete_test FORMAT Values") == "(1310721)"
            assert node2.query("SELECT count() FROM ttl_delete_test FORMAT Values") == "(1310721)"
        else:
            assert node1.query("SELECT count() FROM ttl_delete_test FORMAT Values") == "(1)"
            assert node2.query("SELECT count() FROM ttl_delete_test FORMAT Values") == "(1)"
            assert node1.query("SELECT d FROM ttl_delete_test ORDER BY d FORMAT Values") == "(11)"
            assert node2.query("SELECT d FROM ttl_delete_test ORDER BY d FORMAT Values") == "(11)"

        node1.query("DROP TABLE IF EXISTS ttl_delete_test NO DELAY")
        node2.query("DROP TABLE IF EXISTS ttl_delete_test NO DELAY")


def test_s3_zero_copy_concurrent_merge(cluster):
    node1 = cluster.instances["node1"]
    node2 = cluster.instances["node2"]

    node1.query("DROP TABLE IF EXISTS concurrent_merge NO DELAY")
    node2.query("DROP TABLE IF EXISTS concurrent_merge NO DELAY")

    for node in (node1, node2):
        node.query(
        """
        CREATE TABLE concurrent_merge (id UInt64)
        ENGINE=ReplicatedMergeTree('/clickhouse/tables/concurrent_merge', '{replica}')
        ORDER BY id
        SETTINGS index_granularity=2, storage_policy='s3', remote_fs_execute_merges_on_single_replica_time_threshold=1
        """
    )

    node1.query("system stop merges")
    node2.query("system stop merges")

    # This will generate two parts with 20 granules each
    node1.query("insert into concurrent_merge select number from numbers(40)")
    node1.query("insert into concurrent_merge select number + 1 from numbers(40)")

    wait_for_active_parts(node2, 2, 'concurrent_merge')

    # Merge will materialize default column, it should sleep every granule and take 20 * 2 * 0.1 = 4 sec.
    node1.query("alter table concurrent_merge add column x UInt32 default sleep(0.1)")

    node1.query("system start merges")
    node2.query("system start merges")

    # Now, the merge should start.
    # Because of remote_fs_execute_merges_on_single_replica_time_threshold=1,
    # only one replica will start merge instantly.
    # The other replica should wait for 1 sec and also start it.
    # That should probably cause a data race at s3 storage.
    # For now, it does not happen (every blob has a random name, and we just have a duplicating data)
    node1.query("optimize table concurrent_merge final")

    wait_for_active_parts(node1, 1, 'concurrent_merge')
    wait_for_active_parts(node2, 1, 'concurrent_merge')

    for node in (node1, node2):
        assert node.query('select sum(id) from concurrent_merge').strip() == '1600'