mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-15 20:24:07 +00:00
135 lines
4.3 KiB
Python
135 lines
4.3 KiB
Python
import time
|
|
|
|
import pytest
|
|
|
|
from helpers.cluster import ClickHouseCluster
|
|
|
|
cluster = ClickHouseCluster(__file__)
|
|
node = cluster.add_instance(
|
|
"node",
|
|
main_configs=["configs/asynchronous_metrics_update_period_s.xml"],
|
|
)
|
|
|
|
|
|
@pytest.fixture(scope="module")
|
|
def started_cluster():
|
|
try:
|
|
cluster.start()
|
|
yield cluster
|
|
|
|
finally:
|
|
cluster.shutdown()
|
|
|
|
|
|
def greater(a, b):
|
|
return b > a
|
|
|
|
|
|
def lesser(a, b):
|
|
return b < a
|
|
|
|
|
|
def query_until_condition(a, b, condition, retries=20, timeout=60, delay=0.5):
|
|
"""
|
|
|
|
:param a: could be an input lambda that returns an int or just an int
|
|
:param b: could be an input lambda that returns an int or just an int
|
|
:param condition: lambda that returns a boolean after comparing a and b
|
|
:param retries: number of times to retry until the condition is met
|
|
:param timeout: time in seconds after which stop retrying
|
|
:param delay: time in seconds between each retry
|
|
:return: values of a and b (value post evaluation if lambda)
|
|
"""
|
|
retries_done = 0
|
|
start_time = time.time()
|
|
while True:
|
|
res1 = a() if callable(a) else a
|
|
res2 = b() if callable(b) else b
|
|
if condition(res1, res2):
|
|
return res1, res2
|
|
retries_done += 1
|
|
if retries_done >= retries or (time.time() - start_time) > timeout:
|
|
return res1, res2
|
|
time.sleep(delay)
|
|
|
|
|
|
def test_total_pk_bytes_in_memory_fields(started_cluster):
|
|
query_create = """CREATE TABLE test_pk_bytes
|
|
(
|
|
a UInt64,
|
|
b UInt64
|
|
)
|
|
Engine=MergeTree()
|
|
ORDER BY a SETTINGS index_granularity=1"""
|
|
node.query(query_create)
|
|
|
|
query_pk_bytes = "SELECT value FROM system.asynchronous_metrics WHERE metric = 'TotalPrimaryKeyBytesInMemory';"
|
|
query_pk_bytes_allocated = """SELECT value FROM system.asynchronous_metrics
|
|
WHERE metric = 'TotalPrimaryKeyBytesInMemoryAllocated';"""
|
|
|
|
# query for metrics before inserting anything into the table
|
|
pk_bytes_before = int(node.query(query_pk_bytes).strip())
|
|
pk_bytes_allocated_before = int(node.query(query_pk_bytes_allocated).strip())
|
|
|
|
# insert data into the table and select
|
|
node.query(
|
|
"""INSERT INTO test_pk_bytes SELECT number + 20, number * 20 from numbers(1000000)"""
|
|
)
|
|
|
|
node.query("""SELECT * FROM test_pk_bytes where a > 1000000""")
|
|
|
|
# functions to query primary key bytes used and allocated in memory
|
|
def res_pk_bytes():
|
|
return int(node.query(query_pk_bytes).strip())
|
|
|
|
def res_pk_bytes_allocated():
|
|
return int(node.query(query_pk_bytes_allocated).strip())
|
|
|
|
# query again after data insertion (make a reasonable amount of retries)
|
|
# metrics should be greater after inserting data
|
|
pk_bytes_before, pk_bytes_after = query_until_condition(
|
|
pk_bytes_before, res_pk_bytes, condition=greater
|
|
)
|
|
assert pk_bytes_after > pk_bytes_before
|
|
|
|
pk_bytes_allocated_before, pk_bytes_allocated_after = query_until_condition(
|
|
pk_bytes_allocated_before, res_pk_bytes_allocated, condition=greater
|
|
)
|
|
assert pk_bytes_allocated_after > pk_bytes_allocated_before
|
|
|
|
# insert some more data
|
|
node.query(
|
|
"""INSERT INTO test_pk_bytes SELECT number + 100, number * 200 from numbers(1000000)"""
|
|
)
|
|
node.query("""SELECT * FROM test_pk_bytes""")
|
|
|
|
# query again and compare the metrics.
|
|
# metrics should be greater after inserting more data
|
|
pk_bytes_after, pk_bytes_after_2 = query_until_condition(
|
|
pk_bytes_after, res_pk_bytes, condition=greater
|
|
)
|
|
assert pk_bytes_after_2 > pk_bytes_after
|
|
|
|
pk_bytes_allocated_after, pk_bytes_allocated_after_2 = query_until_condition(
|
|
pk_bytes_allocated_after, res_pk_bytes_allocated, condition=greater
|
|
)
|
|
assert pk_bytes_allocated_after_2 > pk_bytes_allocated_after
|
|
|
|
# drop all the data
|
|
node.query("TRUNCATE table test_pk_bytes;")
|
|
|
|
# query again and compare the metrics.
|
|
# metrics should be lesser after dropping some data
|
|
before_drop, after_drop = query_until_condition(
|
|
pk_bytes_after_2, res_pk_bytes, condition=lesser
|
|
)
|
|
assert before_drop > after_drop
|
|
|
|
before_drop, after_drop = query_until_condition(
|
|
pk_bytes_allocated_after_2, res_pk_bytes_allocated, condition=lesser
|
|
)
|
|
assert before_drop > after_drop
|
|
|
|
# finally drop the table
|
|
node.query("DROP table test_pk_bytes;")
|