mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-14 19:45:11 +00:00
72 lines
3.1 KiB
Python
72 lines
3.1 KiB
Python
|
#!/usr/bin/env python3
|
||
|
|
||
|
import pytest
|
||
|
import time
|
||
|
from helpers.cluster import ClickHouseCluster
|
||
|
from helpers.network import PartitionManager
|
||
|
import random
|
||
|
import string
|
||
|
|
||
|
cluster = ClickHouseCluster(__file__)
|
||
|
node1 = cluster.add_instance('node1', with_zookeeper=True)
|
||
|
node2 = cluster.add_instance('node2', with_zookeeper=True)
|
||
|
|
||
|
DEFAULT_MAX_THREADS_FOR_FETCH = 3
|
||
|
|
||
|
@pytest.fixture(scope="module")
|
||
|
def started_cluster():
|
||
|
try:
|
||
|
cluster.start()
|
||
|
|
||
|
yield cluster
|
||
|
|
||
|
finally:
|
||
|
cluster.shutdown()
|
||
|
|
||
|
|
||
|
def get_random_string(length):
|
||
|
return ''.join(random.choice(string.ascii_uppercase + string.digits) for _ in range(length))
|
||
|
|
||
|
|
||
|
def test_limited_fetches(started_cluster):
|
||
|
"""
|
||
|
Test checks that that we utilize all available threads for fetches
|
||
|
"""
|
||
|
node1.query("CREATE TABLE t (key UInt64, data String) ENGINE = ReplicatedMergeTree('/clickhouse/test/t', '1') ORDER BY tuple() PARTITION BY key")
|
||
|
node2.query("CREATE TABLE t (key UInt64, data String) ENGINE = ReplicatedMergeTree('/clickhouse/test/t', '2') ORDER BY tuple() PARTITION BY key")
|
||
|
|
||
|
with PartitionManager() as pm:
|
||
|
node2.query("SYSTEM STOP FETCHES t")
|
||
|
node1.query("INSERT INTO t SELECT 1, '{}' FROM numbers(5000)".format(get_random_string(104857)))
|
||
|
node1.query("INSERT INTO t SELECT 2, '{}' FROM numbers(5000)".format(get_random_string(104857)))
|
||
|
node1.query("INSERT INTO t SELECT 3, '{}' FROM numbers(5000)".format(get_random_string(104857)))
|
||
|
node1.query("INSERT INTO t SELECT 4, '{}' FROM numbers(5000)".format(get_random_string(104857)))
|
||
|
node1.query("INSERT INTO t SELECT 5, '{}' FROM numbers(5000)".format(get_random_string(104857)))
|
||
|
node1.query("INSERT INTO t SELECT 6, '{}' FROM numbers(5000)".format(get_random_string(104857)))
|
||
|
pm.add_network_delay(node1, 80)
|
||
|
node2.query("SYSTEM START FETCHES t")
|
||
|
fetches_result = []
|
||
|
background_fetches_metric = []
|
||
|
fetched_parts = set([])
|
||
|
for _ in range(1000):
|
||
|
result = node2.query("SELECT result_part_name FROM system.replicated_fetches").strip().split()
|
||
|
background_fetches_metric.append(int(node2.query("select value from system.metrics where metric = 'BackgroundFetchesPoolTask'").strip()))
|
||
|
if not result:
|
||
|
if len(fetched_parts) == 6:
|
||
|
break
|
||
|
time.sleep(0.1)
|
||
|
else:
|
||
|
for part in result:
|
||
|
fetched_parts.add(part)
|
||
|
fetches_result.append(result)
|
||
|
print(fetches_result[-1])
|
||
|
print(background_fetches_metric[-1])
|
||
|
time.sleep(0.1)
|
||
|
|
||
|
for concurrently_fetching_parts in fetches_result:
|
||
|
if len(concurrently_fetching_parts) > DEFAULT_MAX_THREADS_FOR_FETCH:
|
||
|
assert False, "Found more than {} concurrently fetching parts: {}".format(DEFAULT_MAX_THREADS_FOR_FETCH, ', '.join(concurrently_fetching_parts))
|
||
|
|
||
|
assert max([len(parts) for parts in fetches_result]) == 3, "Strange, but we don't utilize max concurrent threads for fetches"
|
||
|
assert(max(background_fetches_metric)) == 3, "Just checking metric consistent with table"
|