mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-12-14 18:32:29 +00:00
103eb17107
After this patch SYSTEM FLUSH LOGS will force system.*_log table creatoins (only for enabled tables of course). This will avoid such hacks like: set log_queries=1; select 1; system flush logs; This is the second version of the patch, since first has a deadlock [1] (reported by @alesapin). This version does not use separate lock and do not issue CREATE TABLE from the query execution thread, instead it activate flushing thread to do so, hence this should be the same as if the table will be created during flush. [1]: https://gist.github.com/alesapin/d915d1deaa27d49aa31223daded02be2 Fixes: #11563 Cc: @alesapin Cc: @nikitamikhaylov
59 lines
1.6 KiB
Python
59 lines
1.6 KiB
Python
# pylint: disable=line-too-long
|
|
# pylint: disable=unused-argument
|
|
# pylint: disable=redefined-outer-name
|
|
|
|
import pytest
|
|
from helpers.cluster import ClickHouseCluster
|
|
|
|
cluster = ClickHouseCluster(__file__)
|
|
node = cluster.add_instance('node_default')
|
|
|
|
system_logs = [
|
|
# disabled by default
|
|
('system.part_log', 0),
|
|
('system.text_log', 0),
|
|
|
|
# enabled by default
|
|
('system.query_log', 1),
|
|
('system.query_thread_log', 1),
|
|
('system.trace_log', 1),
|
|
('system.metric_log', 1),
|
|
]
|
|
|
|
# Default timeout for flush is 60
|
|
# decrease timeout for the test to show possible issues.
|
|
timeout = pytest.mark.timeout(30)
|
|
|
|
@pytest.fixture(scope='module', autouse=True)
|
|
def start_cluster():
|
|
try:
|
|
cluster.start()
|
|
yield cluster
|
|
finally:
|
|
cluster.shutdown()
|
|
|
|
@pytest.fixture(scope='function')
|
|
def flush_logs():
|
|
node.query('SYSTEM FLUSH LOGS')
|
|
|
|
@timeout
|
|
@pytest.mark.parametrize('table,exists', system_logs)
|
|
def test_system_logs(flush_logs, table, exists):
|
|
q = 'SELECT * FROM {}'.format(table)
|
|
if exists:
|
|
node.query(q)
|
|
else:
|
|
assert "Table {} doesn't exist".format(table) in node.query_and_get_error(q)
|
|
|
|
# Logic is tricky, let's check that there is no hang in case of message queue
|
|
# is not empty (this is another code path in the code).
|
|
@timeout
|
|
def test_system_logs_non_empty_queue():
|
|
node.query('SELECT 1', settings={
|
|
# right now defaults are the same,
|
|
# this set explicitly to avoid depends from defaults.
|
|
'log_queries': 1,
|
|
'log_queries_min_type': 'QUERY_START',
|
|
})
|
|
node.query('SYSTEM FLUSH LOGS')
|