def attach_all_parts():
detached_parts = clickhouse.query(chi["metadata"]["name"], "SELECT name FROM system.detached_parts WHERE database='default' AND table='test' AND reason=''", pod=detached_pod)
all_parts = ""
for part in detached_parts.splitlines():
all_parts += f"ALTER TABLE default.test ATTACH PART '{part}';"
if all_parts.strip() != "":
clickhouse.query(chi["metadata"]["name"], all_parts, pod=detached_pod)
def create_part_and_detach():
clickhouse.query(chi["metadata"]["name"], "INSERT INTO default.test SELECT now(), number FROM numbers(100)", pod=detached_pod)
part_name = clickhouse.query(
chi["metadata"]["name"],
sql="SELECT name FROM system.parts WHERE database='default' AND table='test' ORDER BY modification_time DESC LIMIT 1",
pod=detached_pod
)
clickhouse.query(chi["metadata"]["name"], f"ALTER TABLE default.test DETACH PART '{part_name}'", pod=detached_pod)
def prepare_table_for_backup(backup_pod, chi, rows=1000):
backup_name = f'test_backup_{time.strftime("%Y-%m-%d_%H%M%S")}'
clickhouse.query(
chi['metadata']['name'],
"CREATE TABLE IF NOT EXISTS default.test_backup ON CLUSTER 'all-sharded' (i UInt64) ENGINE MergeTree() ORDER BY tuple();"
f"INSERT INTO default.test_backup SELECT number FROM numbers({rows})",
pod=backup_pod)
return backup_name
def run_queries_with_priority():
sql = ""
for i in range(50):
sql += f"SET priority={i % 20};SELECT uniq(number) FROM numbers(20000000):"
cmd = f"echo \\\"{sql} SELECT 1\\\" | xargs -i'{{}}' --no-run-if-empty -d ':' -P 20 clickhouse-client --time -m -n -q \\\"{{}}\\\""
kubectl.launch(f"exec {priority_pod} -- bash -c \"{cmd}\"", timeout=120)
clickhouse.query(
chi["metadata"]["name"],
"SELECT event_time, CurrentMetric_QueryPreempted FROM system.metric_log WHERE CurrentMetric_QueryPreempted > 0",
host=priority_svc,
)
def test_longest_running_query(self, prometheus_operator_spec, clickhouse_operator_spec, chi):
long_running_pod, long_running_svc, _, _ = alerts.random_pod_choice_for_callbacks(chi)
# 600s trigger + 2*30s - double prometheus scraping interval
clickhouse.query(chi["metadata"]["name"], "SELECT now(),sleepEachRow(1),number FROM system.numbers LIMIT 660",
host=long_running_svc, timeout=670)
with Then("check ClickHouseLongestRunningQuery firing"):
fired = alerts.wait_alert_state("ClickHouseLongestRunningQuery", "firing", True, labels={"hostname": long_running_svc},
time_range='30s')
assert fired, error("can't get ClickHouseLongestRunningQuery alert in firing state")
with Then("check ClickHouseLongestRunningQuery gone away"):
resolved = alerts.wait_alert_state("ClickHouseLongestRunningQuery", "firing", False, labels={"hostname": long_running_svc})
assert resolved, error("can't check ClickHouseLongestRunningQuery alert is gone away")
def test_backup_size(self, chi, minio_spec):
decrease_pod, _, increase_pod, _ = alerts.random_pod_choice_for_callbacks(
chi)
backup_cases = {
decrease_pod: {
'rows': (10000, 1000),
'decrease': True,
},
increase_pod: {
'rows': (1000, 10000),
'decrease': False,
},
}
for backup_pod in backup_cases:
decrease = backup_cases[backup_pod]['decrease']
for backup_rows in backup_cases[backup_pod]['rows']:
backup_name = prepare_table_for_backup(backup_pod,
chi,
rows=backup_rows)
exec_on_backup_container(
backup_pod,
f'curl -X POST -sL "http://127.0.0.1:7171/backup/create?name={backup_name}"'
)
wait_backup_command_status(backup_pod,
f'create {backup_name}',
expected_status='success')
if decrease:
clickhouse.query(chi['metadata']['name'],
f"TRUNCATE TABLE default.test_backup",
pod=backup_pod)
time.sleep(15)
fired = alerts.wait_alert_state(
"ClickHouseBackupSizeChanged",
"firing",
expected_state=True,
sleep_time=settings.prometheus_scrape_interval,
labels={"pod_name": backup_pod},
time_range='60s')
assert fired, error(
f"can't get ClickHouseBackupSizeChanged alert in firing state, decrease={decrease}"
)
with Then("check ClickHouseBackupSizeChanged gone away"):
resolved = alerts.wait_alert_state("ClickHouseBackupSizeChanged",
"firing",
expected_state=False,
labels={"pod_name": backup_pod})
assert resolved, error(
f"can't get ClickHouseBackupSizeChanged alert is gone away, decrease={decrease}"
)
def reboot_clickhouse_and_distributed_exection():
# we need 70 delayed files for catch
insert_sql = 'INSERT INTO default.test_distr(event_time, test) SELECT now(), number FROM system.numbers LIMIT 10000'
select_sql = 'SELECT count() FROM default.test_distr'
with Then("reboot clickhouse-server pod"):
kubectl.launch(
f"exec -n {kubectl.namespace} {restarted_pod} -c clickhouse-pod -- kill 1",
ok_to_fail=True,
)
with Then("Insert to distributed table"):
clickhouse.query(chi["metadata"]["name"], insert_sql, host=delayed_pod, ns=kubectl.namespace)
with Then("Select from distributed table"):
clickhouse.query_with_error(chi["metadata"]["name"], select_sql, host=delayed_pod,
ns=kubectl.namespace)
def insert_replicated_data(chi, create_tables, insert_tables):
with When(f'create if not exists replicated tables {create_tables}'):
for table in create_tables:
clickhouse.create_table_on_cluster(
chi,
'all-sharded',
f'default.{table}',
f'(id UInt64) ENGINE=ReplicatedMergeTree(\'/clickhouse/tables/default.{table}/{{shard}}\',\'{{replica}}\') ORDER BY (id)',
if_not_exists=True,
)
with When(f'insert tables data {insert_tables}'):
for table in insert_tables:
clickhouse.query(
chi['metadata']['name'],
f'INSERT INTO default.{table} SELECT rand()+number FROM numbers(1000)',
pod="chi-test-cluster-for-zk-default-0-1-0")
def insert_many_parts_to_clickhouse():
stop_merges = "SYSTEM STOP MERGES default.test;"
min_block = "SET max_block_size=1; SET max_insert_block_size=1; SET min_insert_block_size_rows=1;"
with When(f"Insert to MergeTree table {parts_limits} parts"):
r = parts_limits
sql = stop_merges + min_block + f"INSERT INTO default.test(event_time, test) SELECT now(),number FROM system.numbers LIMIT {r};"
clickhouse.query(chi_name, sql, host=selected_svc, ns=kubectl.namespace)
sql = min_block + "INSERT INTO default.test(event_time, test) SELECT now(), number FROM system.numbers LIMIT 1;"
clickhouse.query_with_error(chi_name, sql, host=selected_svc, ns=kubectl.namespace)
with Then(f"wait prometheus_scrape_interval={prometheus_scrape_interval}*2 sec"):
time.sleep(prometheus_scrape_interval * 2)
with Then("after 21.8 InsertedRows include system.* rows"):
for i in range(35):
sql = min_block + "INSERT INTO default.test(event_time, test) SELECT now(), number FROM system.numbers LIMIT 1;"
clickhouse.query_with_error(chi_name, sql, host=selected_svc, ns=kubectl.namespace)
def test_replicas_max_absolute_delay(self, prometheus_operator_spec, clickhouse_operator_spec, chi):
stop_replica_pod, stop_replica_svc, insert_pod, insert_svc = alerts.random_pod_choice_for_callbacks(chi)
clickhouse.create_table_on_cluster(
chi,
'all-replicated', 'default.test_repl',
'(event_time DateTime, test UInt64) ' +
'ENGINE ReplicatedMergeTree(\'/clickhouse/tables/{installation}-{shard}/test_repl\', \'{replica}\') ORDER BY tuple()'
)
prometheus_scrape_interval = 15
def restart_clickhouse_and_insert_to_replicated_table():
with When(f"stop replica fetches on {stop_replica_svc}"):
sql = "SYSTEM STOP FETCHES default.test_repl"
kubectl.launch(
f"exec -n {kubectl.namespace} {stop_replica_pod} -c clickhouse-pod -- clickhouse-client -q \"{sql}\"",
ok_to_fail=True, timeout=600,
)
sql = "INSERT INTO default.test_repl SELECT now(), number FROM numbers(100000)"
kubectl.launch(
f"exec -n {kubectl.namespace} {insert_pod} -c clickhouse-pod -- clickhouse-client -q \"{sql}\"",
)
with Then("check ClickHouseReplicasMaxAbsoluteDelay firing"):
fired = alerts.wait_alert_state(
"ClickHouseReplicasMaxAbsoluteDelay", "firing", True, labels={"hostname": stop_replica_svc},
time_range='60s', sleep_time=prometheus_scrape_interval * 2,
callback=restart_clickhouse_and_insert_to_replicated_table
)
assert fired, error("can't get ClickHouseReadonlyReplica alert in firing state")
clickhouse.query(
chi["metadata"]["name"],
"SYSTEM START FETCHES; SYSTEM RESTART REPLICAS; SYSTEM SYNC REPLICA default.test_repl",
host=stop_replica_svc, timeout=240
)
with Then("check ClickHouseReplicasMaxAbsoluteDelay gone away"):
resolved = alerts.wait_alert_state("ClickHouseReplicasMaxAbsoluteDelay", "firing", False, labels={"hostname": stop_replica_svc})
assert resolved, error("can't check ClickHouseReplicasMaxAbsoluteDelay alert is gone away")
clickhouse.drop_table_on_cluster(chi, 'all-replicated', 'default.test_repl')
def test_distributed_files_to_insert(self, prometheus_operator_spec, clickhouse_operator_spec, chi):
delayed_pod, delayed_svc, restarted_pod, restarted_svc = alerts.random_pod_choice_for_callbacks(chi)
clickhouse.create_distributed_table_on_cluster(chi)
insert_sql = 'INSERT INTO default.test_distr(event_time, test) SELECT now(), number FROM system.numbers LIMIT 1000'
clickhouse.query(
chi["metadata"]["name"], 'SYSTEM STOP DISTRIBUTED SENDS default.test_distr',
pod=delayed_pod, ns=kubectl.namespace
)
files_to_insert_from_metrics = 0
files_to_insert_from_disk = 0
tries = 0
# we need more than 50 delayed files for catch
while files_to_insert_from_disk <= 55 and files_to_insert_from_metrics <= 55 and tries < 500:
kubectl.launch(
f"exec -n {kubectl.namespace} {restarted_pod} -c clickhouse-pod -- kill 1",
ok_to_fail=True,
)
clickhouse.query(chi["metadata"]["name"], insert_sql, pod=delayed_pod, host=delayed_pod, ns=kubectl.namespace)
files_to_insert_from_metrics = clickhouse.query(
chi["metadata"]["name"], "SELECT value FROM system.metrics WHERE metric='DistributedFilesToInsert'",
pod=delayed_pod, ns=kubectl.namespace
)
files_to_insert_from_metrics = int(files_to_insert_from_metrics)
files_to_insert_from_disk = int(kubectl.launch(
f"exec -n {kubectl.namespace} {delayed_pod} -c clickhouse-pod -- bash -c 'ls -la /var/lib/clickhouse/data/default/test_distr/*/*.bin 2>/dev/null | wc -l'",
ok_to_fail=False,
))
with When("reboot clickhouse-server pod"):
fired = alerts.wait_alert_state(
"ClickHouseDistributedFilesToInsertHigh", "firing", True,
labels={"hostname": delayed_svc, "chi": chi["metadata"]["name"]}
)
assert fired, error("can't get ClickHouseDistributedFilesToInsertHigh alert in firing state")
kubectl.wait_pod_status(restarted_pod, "Running", ns=kubectl.namespace)
clickhouse.query(
chi["metadata"]["name"], 'SYSTEM START DISTRIBUTED SENDS default.test_distr',
pod=delayed_pod, ns=kubectl.namespace
)
with Then("check ClickHouseDistributedFilesToInsertHigh gone away"):
resolved = alerts.wait_alert_state("ClickHouseDistributedFilesToInsertHigh", "firing", False, labels={"hostname": delayed_svc})
assert resolved, error("can't check ClickHouseDistributedFilesToInsertHigh alert is gone away")
clickhouse.drop_distributed_table_on_cluster(chi)
def check_zk_root_znode(chi, pod_count, retry_count=5):
for pod_num in range(pod_count):
out = ""
expected_out = ""
for i in range(retry_count):
if keeper_type == "zookeeper-operator":
expected_out = "[clickhouse, zookeeper, zookeeper-operator]"
keeper_cmd = './bin/zkCli.sh ls /'
pod_prefix = "zookeeper"
elif keeper_type == "zookeeper":
expected_out = "[clickhouse, zookeeper]"
keeper_cmd = './bin/zkCli.sh ls /'
pod_prefix = "zookeeper"
else:
expected_out = "clickhouse"
keeper_cmd = "if [[ ! $(command -v zookeepercli) ]]; then "
keeper_cmd += "wget -q -O /tmp/zookeepercli.deb https://github.com/outbrain/zookeepercli/releases/download/v1.0.12/zookeepercli_1.0.12_amd64.deb; "
keeper_cmd += "dpkg -i /tmp/zookeepercli.deb; "
keeper_cmd += "fi; "
keeper_cmd += "zookeepercli -servers 127.0.0.1:2181 -c ls /"
pod_prefix = "clickhouse-keeper"
out = kubectl.launch(
f"exec {pod_prefix}-{pod_num} -- bash -ce '{keeper_cmd}'",
ns=settings.test_namespace,
ok_to_fail=True)
if expected_out in out:
break
else:
with Then(
f"{keeper_type} ROOT NODE not ready, wait {(i + 1) * 3} sec"
):
time.sleep((i + 1) * 3)
assert expected_out in out, f"Unexpected {keeper_type} `ls /` output"
out = clickhouse.query(
chi["metadata"]["name"],
"SELECT count() FROM system.zookeeper WHERE path='/'")
expected_out = {
"zookeeper": "2",
"zookeeper-operator": "3",
"clickhouse-keeper": "1",
}
assert expected_out[keeper_type] == out.strip(
" \t\r\n"
), f"Unexpected `SELECT count() FROM system.zookeeper WHERE path='/'` output {out}"
def wait_clickhouse_cluster_ready(chi):
with Given("All expected pods present in system.clusters"):
all_pods_ready = False
while all_pods_ready is False:
all_pods_ready = True
for pod in chi['status']['pods']:
cluster_response = clickhouse.query(
chi["metadata"]["name"],
"SYSTEM RELOAD CONFIG; SELECT host_name FROM system.clusters WHERE cluster='all-sharded'",
pod=pod
)
for host in chi['status']['fqdns']:
svc_short_name = host.replace(f'.{settings.test_namespace}.svc.cluster.local', '')
if svc_short_name not in cluster_response:
with Then("Not ready, sleep 5 seconds"):
all_pods_ready = False
time.sleep(5)
def wait_clickhouse_no_readonly_replicas(chi, retries=10):
expected_replicas = chi["spec"]["configuration"]["clusters"][0]["layout"][
"replicasCount"]
expected_replicas = "[" + ",".join(["0"] * expected_replicas) + "]"
for i in range(retries):
readonly_replicas = clickhouse.query(
chi['metadata']['name'],
"SELECT groupArray(value) FROM cluster('all-sharded',system.metrics) WHERE metric='ReadonlyReplica'"
)
if readonly_replicas == expected_replicas:
break
else:
with Then(
f"Clickhouse have readonly_replicas={readonly_replicas}, expected={expected_replicas}, Wait for {i*3} seconds"
):
time.sleep(i * 3)
if i == retries - 1:
Fail(
f"ClickHouse ZK failed, readonly_replicas={readonly_replicas}, expected={expected_replicas}"
)
def check_zk_root_znode(chi, pod_count, zk_retry=5):
for pod_num in range(pod_count):
out = ""
for i in range(zk_retry):
out = kubectl.launch(
f"exec zookeeper-{pod_num} -- bash -ce './bin/zkCli.sh ls /'",
ns=settings.test_namespace,
ok_to_fail=True)
if "[clickhouse, zookeeper]" in out:
break
else:
with Then(
f"Zookeeper ROOT NODE not ready, wait { (i+1)*3} sec"
):
time.sleep((i + 1) * 3)
assert "[clickhouse, zookeeper]" in out, "Unexpected `zkCli.sh ls /` output"
out = clickhouse.query(
chi["metadata"]["name"],
"SELECT count() FROM system.zookeeper WHERE path='/'")
assert "2" == out.strip(
" \t\r\n"
), f"Unexpected `SELECT count() FROM system.zookeeper WHERE path='/'` output {out}"
def test_ch_002(self):
kubectl.create_and_check(
"manifests/chi/test-ch-002-row-level.yaml", {
"apply_templates": {"manifests/chit/tpl-clickhouse-21.8.yaml"},
"do_not_delete": 1,
})
chi = "test-ch-002-row-level"
create_table = """create table test (d Date default today(), team LowCardinality(String), user String) Engine = MergeTree() PARTITION BY d ORDER BY d;"""
with When("Create test table"):
clickhouse.query(chi, create_table)
with And("Insert some data"):
clickhouse.query(
chi,
"INSERT INTO test(team, user) values('team1', 'user1'),('team2', 'user2'),('team3', 'user3'),('team4', 'user4')"
)
with Then(
"Make another query for different users. It should be restricted to corresponding team by row-level security"
):
for user in ['user1', 'user2', 'user3', 'user4']:
out = clickhouse.query(chi,
"select user from test",
user=user,
pwd=user)
assert out == user, error()
with Then(
"Make a count() query for different users. It should be restricted to corresponding team by row-level security"
):
for user in ['user1', 'user2', 'user3', 'user4']:
out = clickhouse.query(chi,
"select count() from test",
user=user,
pwd=user)
assert out == "1", error()
kubectl.delete_chi(chi)
def test_zookeeper_rescale(self):
"""
test scenario for ZK
CH 1 -> 2 wait complete + ZK 1 -> 3 nowait
CH 2 -> 1 wait complete + ZK 3 -> 1 nowait
CH 1 -> 2 wait complete + ZK 1 -> 3 nowait
"""
def insert_replicated_data(chi, create_tables, insert_tables):
with When(f'create if not exists replicated tables {create_tables}'):
for table in create_tables:
clickhouse.create_table_on_cluster(
chi,
'all-sharded',
f'default.{table}',
f'(id UInt64) ENGINE=ReplicatedMergeTree(\'/clickhouse/tables/default.{table}/{{shard}}\',\'{{replica}}\') ORDER BY (id)',
if_not_exists=True,
)
with When(f'insert tables data {insert_tables}'):
for table in insert_tables:
clickhouse.query(
chi['metadata']['name'],
f'INSERT INTO default.{table} SELECT rand()+number FROM numbers(1000)',
pod="chi-test-cluster-for-zk-default-0-1-0")
def check_zk_root_znode(chi, pod_count, zk_retry=5):
for pod_num in range(pod_count):
out = ""
for i in range(zk_retry):
out = kubectl.launch(
f"exec zookeeper-{pod_num} -- bash -ce './bin/zkCli.sh ls /'",
ns=settings.test_namespace,
ok_to_fail=True)
if "[clickhouse, zookeeper]" in out:
break
else:
with Then(
f"Zookeeper ROOT NODE not ready, wait { (i+1)*3} sec"
):
time.sleep((i + 1) * 3)
assert "[clickhouse, zookeeper]" in out, "Unexpected `zkCli.sh ls /` output"
out = clickhouse.query(
chi["metadata"]["name"],
"SELECT count() FROM system.zookeeper WHERE path='/'")
assert "2" == out.strip(
" \t\r\n"
), f"Unexpected `SELECT count() FROM system.zookeeper WHERE path='/'` output {out}"
def rescale_zk_and_clickhouse(ch_node_count,
zk_node_count,
first_install=False):
zk_manifest = 'zookeeper-1-node-1GB-for-tests-only.yaml' if zk_node_count == 1 else 'zookeeper-3-nodes-1GB-for-tests-only.yaml'
_, chi = util.install_clickhouse_and_zookeeper(
chi_file=
f'manifests/chi/test-cluster-for-zookeeper-{ch_node_count}.yaml',
chi_template_file='manifests/chit/tpl-clickhouse-latest.yaml',
chi_name='test-cluster-for-zk',
zk_manifest=zk_manifest,
clean_ns=first_install,
force_zk_install=True,
zk_install_first=first_install,
make_object_count=False,
)
return chi
with When("Clean exists ClickHouse and Zookeeper"):
kubectl.delete_all_zookeeper(settings.test_namespace)
kubectl.delete_all_chi(settings.test_namespace)
with When("Install CH 1 node ZK 1 node"):
chi = rescale_zk_and_clickhouse(ch_node_count=1,
zk_node_count=1,
first_install=True)
util.wait_clickhouse_cluster_ready(chi)
wait_zookeeper_ready(pod_count=1)
check_zk_root_znode(chi, pod_count=1)
util.wait_clickhouse_cluster_ready(chi)
wait_clickhouse_no_readonly_replicas(chi)
insert_replicated_data(chi,
create_tables=['test_repl1'],
insert_tables=['test_repl1'])
total_iterations = 5
for iteration in range(total_iterations):
with When(f"ITERATION {iteration}"):
with Then("CH 1 -> 2 wait complete + ZK 1 -> 3 nowait"):
chi = rescale_zk_and_clickhouse(ch_node_count=2,
zk_node_count=3)
wait_zookeeper_ready(pod_count=3)
check_zk_root_znode(chi, pod_count=3)
util.wait_clickhouse_cluster_ready(chi)
insert_replicated_data(
chi,
create_tables=['test_repl2'],
insert_tables=['test_repl1', 'test_repl2'])
with Then("CH 2 -> 1 wait complete + ZK 3 -> 1 nowait"):
chi = rescale_zk_and_clickhouse(ch_node_count=1,
zk_node_count=1)
wait_zookeeper_ready(pod_count=1)
check_zk_root_znode(chi, pod_count=1)
util.wait_clickhouse_cluster_ready(chi)
insert_replicated_data(
chi,
create_tables=['test_repl3'],
insert_tables=['test_repl1', 'test_repl2', 'test_repl3'])
with When("CH 1 -> 2 wait complete + ZK 1 -> 3 nowait"):
chi = rescale_zk_and_clickhouse(ch_node_count=2, zk_node_count=3)
check_zk_root_znode(chi, pod_count=3)
with Then('check data in tables'):
for table, exptected_rows in {
"test_repl1": str(1000 + 2000 * total_iterations),
"test_repl2": str(2000 * total_iterations),
"test_repl3": str(1000 * total_iterations)
}.items():
actual_rows = clickhouse.query(
chi['metadata']['name'],
f'SELECT count() FROM default.{table}',
pod="chi-test-cluster-for-zk-default-0-1-0")
assert actual_rows == exptected_rows, f"Invalid rows counter after inserts {table} expected={exptected_rows} actual={actual_rows}"
with Then('drop all created tables'):
for i in range(3):
clickhouse.drop_table_on_cluster(chi, 'all-sharded',
f'default.test_repl{i+1}')
def test_ch_001(self):
util.require_zookeeper()
quorum_template = "manifests/chit/tpl-clickhouse-21.8.yaml"
chit_data = yaml_manifest.get_manifest_data(
util.get_full_path(quorum_template))
kubectl.launch(f"delete chit {chit_data['metadata']['name']}",
ns=settings.test_namespace,
ok_to_fail=True)
kubectl.create_and_check("manifests/chi/test-ch-001-insert-quorum.yaml", {
"apply_templates": {quorum_template},
"pod_count": 2,
"do_not_delete": 1,
})
chi = yaml_manifest.get_chi_name(
util.get_full_path("manifests/chi/test-ch-001-insert-quorum.yaml"))
chi_data = kubectl.get("chi", ns=settings.test_namespace, name=chi)
util.wait_clickhouse_cluster_ready(chi_data)
host0 = "chi-test-ch-001-insert-quorum-default-0-0"
host1 = "chi-test-ch-001-insert-quorum-default-0-1"
create_table = """
create table t1 on cluster default (a Int8, d Date default today())
Engine = ReplicatedMergeTree('/clickhouse/tables/{table}', '{replica}')
partition by d order by a
TTL d + interval 5 second
SETTINGS merge_with_ttl_timeout=5""".replace('\r', '').replace('\n', '')
create_mv_table2 = """
create table t2 on cluster default (a Int8)
Engine = ReplicatedMergeTree('/clickhouse/tables/{table}', '{replica}')
partition by tuple() order by a""".replace('\r', '').replace('\n', '')
create_mv_table3 = """
create table t3 on cluster default (a Int8)
Engine = ReplicatedMergeTree('/clickhouse/tables/{table}', '{replica}')
partition by tuple() order by a""".replace('\r', '').replace('\n', '')
create_mv2 = "create materialized view t_mv2 on cluster default to t2 as select a from t1"
create_mv3 = "create materialized view t_mv3 on cluster default to t3 as select a from t1"
with Given("Tables t1, t2, t3 and MVs t1->t2, t1-t3 are created"):
clickhouse.query(chi, create_table)
clickhouse.query(chi, create_mv_table2)
clickhouse.query(chi, create_mv_table3)
clickhouse.query(chi, create_mv2)
clickhouse.query(chi, create_mv3)
with When("Add a row to an old partition"):
clickhouse.query(chi,
"insert into t1(a,d) values(6, today()-1)",
host=host0)
with When("Stop fetches for t1 at replica1"):
clickhouse.query(chi, "system stop fetches default.t1", host=host1)
with Then("Wait 10 seconds and the data should be dropped by TTL"):
time.sleep(10)
out = clickhouse.query(chi,
"select count() from t1 where a=6",
host=host0)
assert out == "0", error()
with When("Resume fetches for t1 at replica1"):
clickhouse.query(chi,
"system start fetches default.t1",
host=host1)
time.sleep(5)
with Then("Inserts should resume"):
clickhouse.query(chi,
"insert into t1(a) values(7)",
host=host0)
clickhouse.query(chi, "insert into t1(a) values(1)")
with When("Stop fetches for t2 at replica1"):
clickhouse.query(chi, "system stop fetches default.t2", host=host1)
with Then("Insert should fail since it can not reach the quorum"):
out = clickhouse.query_with_error(
chi, "insert into t1(a) values(2)", host=host0)
assert "Timeout while waiting for quorum" in out, error()
# kubectl(f"exec {host0}-0 -n test -- cp /var/lib//clickhouse/data/default/t2/all_1_1_0/a.mrk2 /var/lib//clickhouse/data/default/t2/all_1_1_0/a.bin")
# with Then("Corrupt data part in t2"):
# kubectl(f"exec {host0}-0 -n test -- sed -i \"s/b/c/\" /var/lib/clickhouse/data/default/t2/all_1_1_0/a.bin")
with When("Resume fetches for t2 at replica1"):
clickhouse.query(chi,
"system start fetches default.t2",
host=host1)
i = 0
while "2" != clickhouse.query(
chi,
"select active_replicas from system.replicas where database='default' and table='t1'",
pod=host0) and i < 10:
with Then("Not ready, wait 5 seconds"):
time.sleep(5)
i += 1
with Then(
"Inserts should fail with an error regarding not satisfied quorum"
):
out = clickhouse.query_with_error(
chi, "insert into t1(a) values(3)", host=host0)
assert "Quorum for previous write has not been satisfied yet" in out, error(
)
with And("Second insert of the same block should pass"):
clickhouse.query(chi,
"insert into t1(a) values(3)",
host=host0)
with And("Insert of the new block should fail"):
out = clickhouse.query_with_error(
chi, "insert into t1(a) values(4)", host=host0)
assert "Quorum for previous write has not been satisfied yet" in out, error(
)
with And(
"Second insert of the same block with 'deduplicate_blocks_in_dependent_materialized_views' setting should fail"
):
out = clickhouse.query_with_error(
chi,
"set deduplicate_blocks_in_dependent_materialized_views=1; insert into t1(a) values(5)",
host=host0)
assert "Quorum for previous write has not been satisfied yet" in out, error(
)
out = clickhouse.query_with_error(
chi,
"select t1.a t1_a, t2.a t2_a from t1 left outer join t2 using (a) order by t1_a settings join_use_nulls=1"
)
note(out)
def test_insert_related_alerts(self, prometheus_operator_spec, clickhouse_operator_spec, chi):
clickhouse.create_table_on_cluster(chi)
delayed_pod, delayed_svc, rejected_pod, rejected_svc = alerts.random_pod_choice_for_callbacks(chi)
prometheus_scrape_interval = settings.prometheus_scrape_interval
# default values in system.merge_tree_settings
parts_to_throw_insert = 300
parts_to_delay_insert = 150
chi_name = chi["metadata"]["name"]
parts_limits = parts_to_delay_insert
selected_svc = delayed_svc
def insert_many_parts_to_clickhouse():
stop_merges = "SYSTEM STOP MERGES default.test;"
min_block = "SET max_block_size=1; SET max_insert_block_size=1; SET min_insert_block_size_rows=1;"
with When(f"Insert to MergeTree table {parts_limits} parts"):
r = parts_limits
sql = stop_merges + min_block + f"INSERT INTO default.test(event_time, test) SELECT now(),number FROM system.numbers LIMIT {r};"
clickhouse.query(chi_name, sql, host=selected_svc, ns=kubectl.namespace)
sql = min_block + "INSERT INTO default.test(event_time, test) SELECT now(), number FROM system.numbers LIMIT 1;"
clickhouse.query_with_error(chi_name, sql, host=selected_svc, ns=kubectl.namespace)
with Then(f"wait prometheus_scrape_interval={prometheus_scrape_interval}*2 sec"):
time.sleep(prometheus_scrape_interval * 2)
with Then("after 21.8 InsertedRows include system.* rows"):
for i in range(35):
sql = min_block + "INSERT INTO default.test(event_time, test) SELECT now(), number FROM system.numbers LIMIT 1;"
clickhouse.query_with_error(chi_name, sql, host=selected_svc, ns=kubectl.namespace)
insert_many_parts_to_clickhouse()
with Then("check ClickHouseDelayedInsertThrottling firing"):
fired = alerts.wait_alert_state(
"ClickHouseDelayedInsertThrottling", "firing", True, labels={"hostname": delayed_svc}, time_range="60s"
)
assert fired, error("can't get ClickHouseDelayedInsertThrottling alert in firing state")
with Then("check ClickHouseMaxPartCountForPartition firing"):
fired = alerts.wait_alert_state(
"ClickHouseMaxPartCountForPartition", "firing", True, labels={"hostname": delayed_svc}, time_range="90s"
)
assert fired, error("can't get ClickHouseMaxPartCountForPartition alert in firing state")
with Then("check ClickHouseLowInsertedRowsPerQuery firing"):
fired = alerts.wait_alert_state(
"ClickHouseLowInsertedRowsPerQuery", "firing", True, labels={"hostname": delayed_svc}, time_range="120s",
)
assert fired, error("can't get ClickHouseLowInsertedRowsPerQuery alert in firing state")
clickhouse.query(chi_name, "SYSTEM START MERGES default.test", host=selected_svc, ns=kubectl.namespace)
with Then("check ClickHouseDelayedInsertThrottling gone away"):
resolved = alerts.wait_alert_state("ClickHouseDelayedInsertThrottling", "firing", False, labels={"hostname": delayed_svc})
assert resolved, error("can't check ClickHouseDelayedInsertThrottling alert is gone away")
with Then("check ClickHouseMaxPartCountForPartition gone away"):
resolved = alerts.wait_alert_state("ClickHouseMaxPartCountForPartition", "firing", False, labels={"hostname": delayed_svc})
assert resolved, error("can't check ClickHouseMaxPartCountForPartition alert is gone away")
with Then("check ClickHouseLowInsertedRowsPerQuery gone away"):
resolved = alerts.wait_alert_state("ClickHouseLowInsertedRowsPerQuery", "firing", False, labels={"hostname": delayed_svc})
assert resolved, error("can't check ClickHouseLowInsertedRowsPerQuery alert is gone away")
parts_limits = parts_to_throw_insert
selected_svc = rejected_svc
insert_many_parts_to_clickhouse()
with Then("check ClickHouseRejectedInsert firing"):
fired = alerts.wait_alert_state("ClickHouseRejectedInsert", "firing", True, labels={"hostname": rejected_svc}, time_range="30s",
sleep_time=settings.prometheus_scrape_interval)
assert fired, error("can't get ClickHouseRejectedInsert alert in firing state")
with Then("check ClickHouseRejectedInsert gone away"):
resolved = alerts.wait_alert_state("ClickHouseRejectedInsert", "firing", False, labels={"hostname": rejected_svc})
assert resolved, error("can't check ClickHouseRejectedInsert alert is gone away")
clickhouse.query(chi_name, "SYSTEM START MERGES default.test", host=selected_svc, ns=kubectl.namespace)
clickhouse.drop_table_on_cluster(chi)
def test_keeper_outline(
self,
keeper_type="zookeeper",
pod_for_insert_data="chi-test-cluster-for-zk-default-0-1-0",
keeper_manifest_1_node='zookeeper-1-node-1GB-for-tests-only.yaml',
keeper_manifest_3_node='zookeeper-3-nodes-1GB-for-tests-only.yaml',
):
"""
test scenario for Zoo/Clickhouse Keeper
CH 1 -> 2 wait complete + Keeper 1 -> 3 nowait
CH 2 -> 1 wait complete + Keeper 3 -> 1 nowait
CH 1 -> 2 wait complete + Keeper 1 -> 3 nowait
"""
def insert_replicated_data(chi, create_tables, insert_tables):
with When(f'create if not exists replicated tables {create_tables}'):
for table in create_tables:
clickhouse.create_table_on_cluster(
chi,
'all-sharded',
f'default.{table}',
f'(id UInt64) ENGINE=ReplicatedMergeTree(\'/clickhouse/tables/default.{table}/{{shard}}\',\'{{replica}}\') ORDER BY (id)',
if_not_exists=True,
)
with When(f'insert tables data {insert_tables}'):
for table in insert_tables:
clickhouse.query(
chi['metadata']['name'],
f'INSERT INTO default.{table} SELECT rand()+number FROM numbers(1000)',
pod=pod_for_insert_data)
def check_zk_root_znode(chi, pod_count, retry_count=5):
for pod_num in range(pod_count):
out = ""
expected_out = ""
for i in range(retry_count):
if keeper_type == "zookeeper-operator":
expected_out = "[clickhouse, zookeeper, zookeeper-operator]"
keeper_cmd = './bin/zkCli.sh ls /'
pod_prefix = "zookeeper"
elif keeper_type == "zookeeper":
expected_out = "[clickhouse, zookeeper]"
keeper_cmd = './bin/zkCli.sh ls /'
pod_prefix = "zookeeper"
else:
expected_out = "clickhouse"
keeper_cmd = "if [[ ! $(command -v zookeepercli) ]]; then "
keeper_cmd += "wget -q -O /tmp/zookeepercli.deb https://github.com/outbrain/zookeepercli/releases/download/v1.0.12/zookeepercli_1.0.12_amd64.deb; "
keeper_cmd += "dpkg -i /tmp/zookeepercli.deb; "
keeper_cmd += "fi; "
keeper_cmd += "zookeepercli -servers 127.0.0.1:2181 -c ls /"
pod_prefix = "clickhouse-keeper"
out = kubectl.launch(
f"exec {pod_prefix}-{pod_num} -- bash -ce '{keeper_cmd}'",
ns=settings.test_namespace,
ok_to_fail=True)
if expected_out in out:
break
else:
with Then(
f"{keeper_type} ROOT NODE not ready, wait {(i + 1) * 3} sec"
):
time.sleep((i + 1) * 3)
assert expected_out in out, f"Unexpected {keeper_type} `ls /` output"
out = clickhouse.query(
chi["metadata"]["name"],
"SELECT count() FROM system.zookeeper WHERE path='/'")
expected_out = {
"zookeeper": "2",
"zookeeper-operator": "3",
"clickhouse-keeper": "1",
}
assert expected_out[keeper_type] == out.strip(
" \t\r\n"
), f"Unexpected `SELECT count() FROM system.zookeeper WHERE path='/'` output {out}"
def rescale_zk_and_clickhouse(ch_node_count,
keeper_node_count,
first_install=False):
keeper_manifest = keeper_manifest_1_node if keeper_node_count == 1 else keeper_manifest_3_node
_, chi = util.install_clickhouse_and_keeper(
chi_file=
f'manifests/chi/test-cluster-for-{keeper_type}-{ch_node_count}.yaml',
chi_template_file='manifests/chit/tpl-clickhouse-latest.yaml',
chi_name='test-cluster-for-zk',
keeper_manifest=keeper_manifest,
keeper_type=keeper_type,
clean_ns=first_install,
force_keeper_install=True,
keeper_install_first=first_install,
make_object_count=False,
)
return chi
with When("Clean exists ClickHouse Keeper and ZooKeeper"):
kubectl.delete_all_chi(settings.test_namespace)
kubectl.delete_all_keeper(settings.test_namespace)
with When("Install CH 1 node ZK 1 node"):
chi = rescale_zk_and_clickhouse(ch_node_count=1,
keeper_node_count=1,
first_install=True)
util.wait_clickhouse_cluster_ready(chi)
wait_keeper_ready(keeper_type=keeper_type, pod_count=1)
check_zk_root_znode(chi, pod_count=1)
util.wait_clickhouse_cluster_ready(chi)
wait_clickhouse_no_readonly_replicas(chi)
insert_replicated_data(chi,
create_tables=['test_repl1'],
insert_tables=['test_repl1'])
total_iterations = 3
for iteration in range(total_iterations):
with When(f"ITERATION {iteration}"):
with Then("CH 1 -> 2 wait complete + ZK 1 -> 3 nowait"):
chi = rescale_zk_and_clickhouse(ch_node_count=2,
keeper_node_count=3)
wait_keeper_ready(keeper_type=keeper_type, pod_count=3)
check_zk_root_znode(chi, pod_count=3)
util.wait_clickhouse_cluster_ready(chi)
wait_clickhouse_no_readonly_replicas(chi)
insert_replicated_data(
chi,
create_tables=['test_repl2'],
insert_tables=['test_repl1', 'test_repl2'])
with Then("CH 2 -> 1 wait complete + ZK 3 -> 1 nowait"):
chi = rescale_zk_and_clickhouse(ch_node_count=1,
keeper_node_count=1)
wait_keeper_ready(keeper_type=keeper_type, pod_count=1)
check_zk_root_znode(chi, pod_count=1)
util.wait_clickhouse_cluster_ready(chi)
wait_clickhouse_no_readonly_replicas(chi)
insert_replicated_data(
chi,
create_tables=['test_repl3'],
insert_tables=['test_repl1', 'test_repl2', 'test_repl3'])
with When("CH 1 -> 2 wait complete + ZK 1 -> 3 nowait"):
chi = rescale_zk_and_clickhouse(ch_node_count=2, keeper_node_count=3)
check_zk_root_znode(chi, pod_count=3)
with Then('check data in tables'):
for table_name, exptected_rows in {
"test_repl1": str(1000 + 2000 * total_iterations),
"test_repl2": str(2000 * total_iterations),
"test_repl3": str(1000 * total_iterations)
}.items():
actual_rows = clickhouse.query(
chi['metadata']['name'],
f'SELECT count() FROM default.{table_name}',
pod="chi-test-cluster-for-zk-default-0-1-0")
assert actual_rows == exptected_rows, f"Invalid rows counter after inserts {table_name} expected={exptected_rows} actual={actual_rows}"
with Then('drop all created tables'):
for i in range(3):
clickhouse.drop_table_on_cluster(chi, 'all-sharded',
f'default.test_repl{i + 1}')
