def reboot_metrics_exporter():
clickhouse_operator_pod = clickhouse_operator_spec["items"][0][
"metadata"]["name"]
kubectl.launch(
f"exec -n {settings.operator_namespace} {clickhouse_operator_pod} -c metrics-exporter -- reboot",
ok_to_fail=True,
)
def restart_zookeeper():
kubectl.launch(
f"exec -n {kubectl.namespace} zookeeper-0 -- sh -c \"kill 1\"",
ok_to_fail=True,
)
clickhouse.query_with_error(chi_name, "SELECT name, path FROM system.zookeeper WHERE path='/'", host=svc1)
clickhouse.query_with_error(chi_name, "SELECT name, path FROM system.zookeeper WHERE path='/'", host=svc2)
def make_too_many_connection():
long_cmd = ""
for _ in range(120):
port = random.choice(["8123", "3306", "9000"])
if port == "8123":
# HTTPConnection metric increase after full parsing of HTTP Request, we can't provide pause between CONNECT and QUERY running
# long_cmd += f"nc -vv 127.0.0.1 {port} <( printf \"POST / HTTP/1.1\\r\\nHost: 127.0.0.1:8123\\r\\nContent-Length: 34\\r\\n\\r\\nTEST\\r\\nTEST\\r\\nTEST\\r\\nTEST\\r\\nTEST\\r\\nTEST\");"
long_cmd += 'wget -qO- "http://127.0.0.1:8123?query=SELECT sleepEachRow(1),number,now() FROM numbers(30)";'
elif port == "9000":
long_cmd += 'clickhouse-client --idle_connection_timeout 70 --receive_timeout 70 -q "SELECT sleepEachRow(1),number,now() FROM numbers(30)";'
# elif port == "3306":
# long_cmd += 'mysql -u default -h 127.0.0.1 -e "SELECT sleepEachRow(1),number, now() FROM numbers(30)";'
else:
long_cmd += f"printf \"1\\n1\" | nc -q 5 -i 30 -vv 127.0.0.1 {port};"
nc_cmd = f"echo '{long_cmd} exit 0' | xargs --verbose -i'{{}}' --no-run-if-empty -d ';' -P 120 bash -c '{{}}' 1>/dev/null"
with open("/tmp/nc_cmd.sh", "w") as f:
f.write(nc_cmd)
kubectl.launch(
f"cp /tmp/nc_cmd.sh {too_many_connection_pod}:/tmp/nc_cmd.sh -c clickhouse"
)
kubectl.launch(
f"exec -n {kubectl.namespace} {too_many_connection_pod} -c clickhouse -- bash /tmp/nc_cmd.sh",
timeout=600,
)
def test_metrics_exporter_reboot():
def check_monitoring_chi(operator_namespace, operator_pod, expect_result, max_retries=10):
with And(f"metrics-exporter /chi enpoint result should return {expect_result}"):
for i in range(1, max_retries):
# check /metrics for try to refresh monitored instances
kubectl.launch(
f"exec {operator_pod} -c metrics-exporter -- wget -O- -q http://127.0.0.1:8888/metrics",
ns=operator_namespace
)
# check /chi after refresh monitored instances
out = kubectl.launch(
f"exec {operator_pod} -c metrics-exporter -- wget -O- -q http://127.0.0.1:8888/chi",
ns=operator_namespace
)
out = json.loads(out)
if out == expect_result:
break
with Then("Not ready. Wait for " + str(i * 5) + " seconds"):
time.sleep(i * 5)
assert out == expect_result, error()
with Given("clickhouse-operator is installed"):
kubectl.wait_field("pods", "-l app=clickhouse-operator", ".status.containerStatuses[*].ready", "true,true",
ns=settings.operator_namespace)
assert kubectl.get_count("pod", ns='--all-namespaces', label="-l app=clickhouse-operator") > 0, error()
out = kubectl.launch("get pods -l app=clickhouse-operator", ns=settings.operator_namespace).splitlines()[1]
operator_pod = re.split(r'[\t\r\n\s]+', out)[0]
operator_namespace = settings.operator_namespace
kubectl.delete_ns(kubectl.namespace)
kubectl.create_ns(kubectl.namespace)
check_monitoring_chi(operator_namespace, operator_pod, [])
with And("created simple clickhouse installation"):
config = util.get_full_path("../docs/chi-examples/01-simple-layout-01-1shard-1repl.yaml")
kubectl.create_and_check(
config=config,
check={
"object_counts": {
"statefulset": 1,
"pod": 1,
"service": 2,
},
"do_not_delete": True,
})
expected_chi = [{
"namespace": "test", "name": "simple-01",
"hostnames": ["chi-simple-01-cluster-0-0.test.svc.cluster.local"]
}]
check_monitoring_chi(operator_namespace, operator_pod, expected_chi)
with When("reboot metrics exporter"):
kubectl.launch(f"exec -n {operator_namespace} {operator_pod} -c metrics-exporter -- reboot")
time.sleep(15)
kubectl.wait_field("pods", "-l app=clickhouse-operator",
".status.containerStatuses[*].ready", "true,true",
ns=settings.operator_namespace)
with Then("check metrics exporter still contains chi objects"):
check_monitoring_chi(operator_namespace, operator_pod, expected_chi)
kubectl.delete(config)
check_monitoring_chi(operator_namespace, operator_pod, [])
def create_success_backup():
backup_name = backup_prefix + "-" + str(random.randint(1, 4096))
kubectl.launch(
f"exec -n {settings.test_namespace} {backup_pod} -c clickhouse-backup -- curl -X POST -sL http://127.0.0.1:7171/backup/create?name={backup_name}",
)
wait_backup_command_status(backup_pod,
command_name=f'create {backup_name}',
expected_status='success')
def restart_zookeeper():
kubectl.launch(
f"exec -n {kubectl.namespace} zookeeper-0 -- sh -c \"kill 1\"",
ok_to_fail=True,
)
clickhouse.query_with_error(
chi_name,
"INSERT INTO default.test_repl VALUES(now(),rand())",
host=read_only_svc)
def test_backup_not_run(self):
not_run_pod, _, _, _ = alerts.random_pod_choice_for_callbacks(chi)
apply_fake_backup("prepare fake backup for time metric")
with Then(f"wait {not_run_pod} ready"):
kubectl.wait_field("pod", not_run_pod, ".spec.containers[1].image",
"nginx:latest")
kubectl.wait_field("pod", not_run_pod,
".status.containerStatuses[1].ready", "true")
with Then(f"setup {not_run_pod} backup create end time"):
kubectl.launch(
f'exec {not_run_pod} -c clickhouse-backup -- bash -xc \''
'echo "# HELP clickhouse_backup_last_create_finish Last backup create finish timestamp" > /usr/share/nginx/html/metrics && '
'echo "# TYPE clickhouse_backup_last_create_finish gauge" >> /usr/share/nginx/html/metrics && '
f'echo "clickhouse_backup_last_create_finish {int((datetime.datetime.now() - datetime.timedelta(days=2)).timestamp())}" >> /usr/share/nginx/html/metrics '
'\'')
fired = alerts.wait_alert_state(
"ClickhouseBackupDoesntRunTooLong",
"firing",
expected_state=True,
sleep_time=settings.prometheus_scrape_interval,
labels={"pod_name": not_run_pod},
time_range='60s')
assert fired, error(
"can't get ClickhouseBackupDoesntRunTooLong alert in firing state")
apply_normal_backup()
backup_name = prepare_table_for_backup(not_run_pod)
wait_backup_pod_ready_and_curl_installed(not_run_pod)
with When('Backup is success'):
exec_on_backup_container(
not_run_pod,
f'curl -X POST -sL "http://127.0.0.1:7171/backup/create?name={backup_name}"'
)
wait_backup_command_status(not_run_pod,
f'create {backup_name}',
expected_status='success')
exec_on_backup_container(
not_run_pod,
f'curl -X POST -sL "http://127.0.0.1:7171/backup/upload/{backup_name}"'
)
wait_backup_command_status(not_run_pod,
f'upload {backup_name}',
expected_status='success')
with Then("check ClickhouseBackupDoesntRunTooLong gone away"):
resolved = alerts.wait_alert_state("ClickhouseBackupDoesntRunTooLong",
"firing",
expected_state=False,
labels={"pod_name": not_run_pod})
assert resolved, error(
"can't get ClickhouseBackupDoesntRunTooLong alert is gone away")
def test_too_many_connections(self):
too_many_connection_pod, too_many_connection_svc, _, _ = alerts.random_pod_choice_for_callbacks(
chi)
cmd = "export DEBIAN_FRONTEND=noninteractive; apt-get update; apt-get install -y netcat mysql-client"
kubectl.launch(
f"exec -n {kubectl.namespace} {too_many_connection_pod} -c clickhouse-pod -- bash -c \"{cmd}\"",
timeout=120,
)
def make_too_many_connection():
long_cmd = ""
for _ in range(120):
port = random.choice(["8123", "3306", "3306", "3306", "9000"])
if port == "8123":
# HTTPConnection metric increase after full parsing of HTTP Request, we can't provide pause between CONNECT and QUERY running
# long_cmd += f"nc -vv 127.0.0.1 {port} <( printf \"POST / HTTP/1.1\\r\\nHost: 127.0.0.1:8123\\r\\nContent-Length: 34\\r\\n\\r\\nTEST\\r\\nTEST\\r\\nTEST\\r\\nTEST\\r\\nTEST\\r\\nTEST\");"
long_cmd += 'wget -qO- "http://127.0.0.1:8123?query=SELECT sleepEachRow(1),number,now() FROM numbers(30)";'
elif port == "9000":
long_cmd += 'clickhouse-client --send_logs_level information --idle_connection_timeout 70 --receive_timeout 70 -q "SELECT sleepEachRow(1),number,now() FROM numbers(30)";'
# elif port == "3306":
# long_cmd += 'mysql -u default -h 127.0.0.1 -e "SELECT sleepEachRow(1),number, now() FROM numbers(30)";'
else:
long_cmd += f"printf \"1\\n1\" | nc -q 5 -i 30 -vv 127.0.0.1 {port};"
nc_cmd = f"echo '{long_cmd} whereis nc; exit 0' | xargs --verbose -i'{{}}' --no-run-if-empty -d ';' -P 120 bash -c '{{}}' 1>/dev/null"
with open("/tmp/nc_cmd.sh", "w") as f:
f.write(nc_cmd)
kubectl.launch(
f"cp /tmp/nc_cmd.sh {too_many_connection_pod}:/tmp/nc_cmd.sh -c clickhouse-pod"
)
kubectl.launch(
f"exec -n {kubectl.namespace} {too_many_connection_pod} -c clickhouse-pod -- bash /tmp/nc_cmd.sh",
timeout=600,
)
with Then("check ClickHouseTooManyConnections firing"):
fired = alerts.wait_alert_state(
"ClickHouseTooManyConnections",
"firing",
True,
labels={"hostname": too_many_connection_svc},
time_range='90s',
callback=make_too_many_connection)
assert fired, error(
"can't get ClickHouseTooManyConnections alert in firing state")
with Then("check ClickHouseTooManyConnections gone away"):
resolved = alerts.wait_alert_state(
"ClickHouseTooManyConnections",
"firing",
False,
labels={"hostname": too_many_connection_svc})
assert resolved, error(
"can't check ClickHouseTooManyConnections alert is gone away")
def rewrite_dns_on_clickhouse_server(write_new=True):
dns = new_dns if write_new else old_dns
kubectl.launch(
f"exec -n {kubectl.namespace} {clickhouse_pod} -c clickhouse -- bash -c \"printf \\\"{dns}\\\" > /etc/resolv.conf\"",
ok_to_fail=False,
)
kubectl.launch(
f"exec -n {kubectl.namespace} {clickhouse_pod} -c clickhouse -- clickhouse-client --echo -mn -q \"SYSTEM DROP DNS CACHE; SELECT count() FROM cluster('all-sharded',system,metrics)\"",
ok_to_fail=True,
)
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 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 -- clickhouse-client -q \"{sql}\"",
ok_to_fail=True,
)
sql = "INSERT INTO default.test_repl SELECT now(), number FROM numbers(100000)"
kubectl.launch(
f"exec -n {kubectl.namespace} {insert_pod} -c clickhouse -- clickhouse-client -q \"{sql}\"",
)
def test_too_much_running_queries(self):
_, _, too_many_queries_pod, too_many_queries_svc = alerts.random_pod_choice_for_callbacks(
chi)
cmd = "export DEBIAN_FRONTEND=noninteractive; apt-get update; apt-get install -y mysql-client"
kubectl.launch(
f"exec -n {kubectl.namespace} {too_many_queries_pod} -c clickhouse-pod -- bash -c \"{cmd}\"",
ok_to_fail=True,
)
def make_too_many_queries():
long_cmd = ""
for _ in range(90):
port = random.choice(["8123", "3306", "9000"])
if port == "9000":
long_cmd += 'clickhouse-client -q "SELECT sleepEachRow(1),now() FROM numbers(60)";'
if port == "3306":
long_cmd += 'mysql -h 127.0.0.1 -P 3306 -u default -e "SELECT sleepEachRow(1),now() FROM numbers(60)";'
if port == "8123":
long_cmd += 'wget -qO- "http://127.0.0.1:8123?query=SELECT sleepEachRow(1),now() FROM numbers(60)";'
long_cmd = f"echo '{long_cmd}' | xargs --verbose -i'{{}}' --no-run-if-empty -d ';' -P 100 bash -c '{{}}' 1>/dev/null"
with open("/tmp/long_cmd.sh", "w") as f:
f.write(long_cmd)
kubectl.launch(
f"cp /tmp/long_cmd.sh {too_many_queries_pod}:/tmp/long_cmd.sh -c clickhouse-pod"
)
kubectl.launch(
f"exec -n {kubectl.namespace} {too_many_queries_pod} -c clickhouse-pod -- bash /tmp/long_cmd.sh",
timeout=90,
)
with Then("check ClickHouseTooManyRunningQueries firing"):
fired = alerts.wait_alert_state(
"ClickHouseTooManyRunningQueries",
"firing",
True,
labels={"hostname": too_many_queries_svc},
callback=make_too_many_queries,
time_range="30s")
assert fired, error(
"can't get ClickHouseTooManyConnections alert in firing state")
with Then("check ClickHouseTooManyConnections gone away"):
resolved = alerts.wait_alert_state(
"ClickHouseTooManyRunningQueries",
"firing",
False,
labels={"hostname": too_many_queries_svc},
sleep_time=settings.prometheus_scrape_interval)
assert resolved, error(
"can't check ClickHouseTooManyConnections alert is gone away")
def create_fail_backup():
backup_name = backup_prefix + "-" + str(random.randint(1, 4096))
backup_dir = f"/var/lib/clickhouse/backup/{backup_name}/shadow/default/test_backup"
kubectl.launch(
f"exec -n {settings.test_namespace} {backup_pod} -c clickhouse-backup -- bash -c 'mkdir -v -m 0400 -p {backup_dir}'",
)
kubectl.launch(
f"exec -n {settings.test_namespace} {backup_pod} -c clickhouse-backup -- curl -X POST -sL http://127.0.0.1:7171/backup/create?name={backup_name}",
)
wait_backup_command_status(backup_pod,
command_name=f'create {backup_name}',
expected_status='error')
def restart_operator(ns=settings.operator_namespace, timeout=60):
pod_name = kubectl.get(
"pod", name="", ns=ns,
label="-l app=clickhouse-operator")["items"][0]["metadata"]["name"]
kubectl.launch(f"delete pod {pod_name}", ns=ns, timeout=timeout)
kubectl.wait_object("pod",
name="",
ns=ns,
label="-l app=clickhouse-operator")
pod_name = kubectl.get(
"pod", name="", ns=ns,
label="-l app=clickhouse-operator")["items"][0]["metadata"]["name"]
kubectl.wait_pod_status(pod_name, "Running", ns=ns)
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 -- kill 1",
ok_to_fail=True,
)
with And("Insert to distributed table"):
clickhouse.query(chi["metadata"]["name"], insert_sql, host=delayed_pod, ns=kubectl.namespace)
with And("Select from distributed table"):
clickhouse.query_with_error(chi["metadata"]["name"], select_sql, host=delayed_pod,
ns=kubectl.namespace)
def test_distributed_files_to_insert():
delayed_pod, delayed_svc, restarted_pod, restarted_svc = random_pod_choice_for_callbacks()
create_distributed_table_on_cluster()
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 -- 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 -- 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 = 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 = wait_alert_state("ClickHouseDistributedFilesToInsertHigh", "firing", False, labels={"hostname": delayed_svc})
assert resolved, error("can't check ClickHouseDistributedFilesToInsertHigh alert is gone away")
drop_distributed_table_on_cluster()
def test_clickhouse_dns_errors():
random_idx = random.randint(0, 1)
clickhouse_pod = chi["status"]["pods"][random_idx]
clickhouse_svc = chi["status"]["fqdns"][random_idx]
old_dns = kubectl.launch(
f"exec -n {kubectl.namespace} {clickhouse_pod} -c clickhouse -- cat /etc/resolv.conf",
ok_to_fail=False,
)
new_dns = re.sub(r'^nameserver (.+)', 'nameserver 1.1.1.1', old_dns)
def rewrite_dns_on_clickhouse_server(write_new=True):
dns = new_dns if write_new else old_dns
kubectl.launch(
f"exec -n {kubectl.namespace} {clickhouse_pod} -c clickhouse -- bash -c \"printf \\\"{dns}\\\" > /etc/resolv.conf\"",
ok_to_fail=False,
)
kubectl.launch(
f"exec -n {kubectl.namespace} {clickhouse_pod} -c clickhouse -- clickhouse-client --echo -mn -q \"SYSTEM DROP DNS CACHE; SELECT count() FROM cluster('all-sharded',system,metrics)\"",
ok_to_fail=True,
)
with When("rewrite /etc/resolv.conf in clickhouse-server pod"):
fired = wait_alert_state("ClickHouseDNSErrors", "firing", True, labels={"hostname": clickhouse_svc},
time_range='20s', callback=rewrite_dns_on_clickhouse_server, sleep_time=5)
assert fired, error("can't get ClickHouseDNSErrors alert in firing state")
with Then("check ClickHouseDNSErrors gone away"):
rewrite_dns_on_clickhouse_server(write_new=False)
resolved = wait_alert_state("ClickHouseDNSErrors", "firing", False, labels={"hostname": clickhouse_svc})
assert resolved, error("can't check ClickHouseDNSErrors alert is gone away")
def check_alert_state(alert_name,
prometheus_pod,
alert_state="firing",
labels=None,
time_range="10s"):
with Then(
f"check {alert_name} for state {alert_state} and {labels} labels in {time_range}"
):
cmd = f"exec -n {settings.prometheus_namespace} {prometheus_pod} -c prometheus -- "
cmd += "wget -qO- 'http://127.0.0.1:9090/api/v1/query?query=ALERTS{"
if labels is None:
labels = {}
if not isinstance(labels, dict):
fail(f"Invalid labels={labels}")
labels.update({"alertname": alert_name, "alertstate": alert_state})
cmd += ",".join(
[f"{name}=\"{value}\"" for name, value in labels.items()])
cmd += f"}}[{time_range}]' 2>/dev/null"
out = kubectl.launch(cmd)
out = json.loads(out)
if not ("status" in out and out["status"] == "success"):
fail("wrong response from prometheus query API")
if len(out["data"]["result"]) == 0:
with Then("not present, empty result"):
return False
result_labels = out["data"]["result"][0]["metric"].items()
exists = all(item in result_labels for item in labels.items())
with Then("got result and contains labels"
if exists else "got result, but doesn't contain labels"):
return exists
def test_metrics_exporter_with_multiple_clickhouse_version():
def check_monitoring_metrics(operator_namespace, operator_pod, expect_result, max_retries=10):
with And(f"metrics-exporter /metrics enpoint result should match with {expect_result}"):
for i in range(1, max_retries):
out = kubectl.launch(
f"exec {operator_pod} -c metrics-exporter -- wget -O- -q http://127.0.0.1:8888/metrics",
ns=operator_namespace
)
all_strings_expected_done = True
for string, exists in expect_result.items():
all_strings_expected_done = (exists == (string in out))
if not all_strings_expected_done:
break
if all_strings_expected_done:
break
with Then("Not ready. Wait for " + str(i * 5) + " seconds"):
time.sleep(i * 5)
assert all_strings_expected_done, error()
with Given("clickhouse-operator pod exists"):
out = kubectl.launch("get pods -l app=clickhouse-operator", ns='kube-system').splitlines()[1]
operator_pod = re.split(r'[\t\r\n\s]+', out)[0]
operator_namespace = "kube-system"
with Then("check empty /metrics"):
kubectl.delete_ns(kubectl.namespace, ok_to_fail=True)
kubectl.create_ns(kubectl.namespace)
check_monitoring_metrics(operator_namespace, operator_pod, expect_result={
'chi_clickhouse_metric_VersionInteger': False,
})
with Then("Install multiple clickhouse version"):
config = util.get_full_path("configs/test-017-multi-version.yaml")
kubectl.create_and_check(
config=config,
check={
"object_counts": {
"statefulset": 4,
"pod": 4,
"service": 5,
},
"do_not_delete": True,
})
with And("Check not empty /metrics"):
check_monitoring_metrics(operator_namespace, operator_pod, expect_result={
'# HELP chi_clickhouse_metric_VersionInteger': True,
'# TYPE chi_clickhouse_metric_VersionInteger gauge': True,
'chi_clickhouse_metric_VersionInteger{chi="test-017-multi-version",hostname="chi-test-017-multi-version-default-0-0': True,
'chi_clickhouse_metric_VersionInteger{chi="test-017-multi-version",hostname="chi-test-017-multi-version-default-1-0': True,
'chi_clickhouse_metric_VersionInteger{chi="test-017-multi-version",hostname="chi-test-017-multi-version-default-2-0': True,
'chi_clickhouse_metric_VersionInteger{chi="test-017-multi-version",hostname="chi-test-017-multi-version-default-3-0': True,
})
with Then("check empty /metrics after delete namespace"):
kubectl.delete_ns(kubectl.namespace)
check_monitoring_metrics(operator_namespace, operator_pod, expect_result={
'chi_clickhouse_metric_VersionInteger': False,
})
def exec_on_backup_container(backup_pod,
cmd,
ns=settings.test_namespace,
ok_to_fail=False,
timeout=60,
container='clickhouse-backup'):
return kubectl.launch(f'exec -n {ns} {backup_pod} -c {container} -- {cmd}',
ok_to_fail=ok_to_fail,
timeout=timeout)
def query(
chi_name,
sql,
with_error=False,
host="127.0.0.1",
port="9000",
user="",
pwd="",
ns=settings.test_namespace,
timeout=60,
advanced_params="",
pod="",
):
pod_names = kubectl.get_pod_names(chi_name, ns)
pod_name = pod_names[0]
for p in pod_names:
if host in p or p == pod:
pod_name = p
break
pwd_str = "" if pwd == "" else f"--password={pwd}"
user_str = "" if user == "" else f"--user={user}"
if with_error:
return kubectl.launch(
f"exec {pod_name}"
f" --"
f" clickhouse-client -mn -h {host} --port={port} {user_str} {pwd_str} {advanced_params}"
f" --query=\"{sql}\""
f" 2>&1",
timeout=timeout,
ns=ns,
ok_to_fail=True,
)
else:
return kubectl.launch(
f"exec {pod_name} -n {ns}"
f" -- "
f"clickhouse-client -mn -h {host} --port={port} {user_str} {pwd_str} {advanced_params}"
f"--query=\"{sql}\"",
timeout=timeout,
ns=ns,
)
def check_monitoring_chi(operator_namespace, operator_pod, expect_result, max_retries=10):
with Then(f"metrics-exporter /chi enpoint result should return {expect_result}"):
for i in range(1, max_retries):
# check /metrics for try to refresh monitored instances
kubectl.launch(
f"exec {operator_pod} -c metrics-exporter -- wget -O- -q http://127.0.0.1:8888/metrics",
ns=operator_namespace
)
# check /chi after refresh monitored instances
out = kubectl.launch(
f"exec {operator_pod} -c metrics-exporter -- wget -O- -q http://127.0.0.1:8888/chi",
ns=operator_namespace
)
out = json.loads(out)
if out == expect_result:
break
with Then("Not ready. Wait for " + str(i * 5) + " seconds"):
time.sleep(i * 5)
assert out == expect_result, error()
def set_operator_version(version, ns=settings.operator_namespace, timeout=60):
operator_image = f"{settings.operator_docker_repo}:{version}"
metrics_exporter_image = f"{settings.metrics_exporter_docker_repo}:{version}"
kubectl.launch(f"set image deployment.v1.apps/clickhouse-operator clickhouse-operator={operator_image}", ns=ns)
kubectl.launch(f"set image deployment.v1.apps/clickhouse-operator metrics-exporter={metrics_exporter_image}", ns=ns)
kubectl.launch("rollout status deployment.v1.apps/clickhouse-operator", ns=ns, timeout=timeout)
assert kubectl.get_count("pod", ns=ns, label="-l app=clickhouse-operator") > 0, error()
def make_too_many_queries():
long_cmd = ""
for _ in range(90):
port = random.choice(["8123", "3306", "9000"])
if port == "9000":
long_cmd += 'clickhouse-client -q "SELECT sleepEachRow(1),now() FROM numbers(60)";'
if port == "3306":
long_cmd += 'mysql -h 127.0.0.1 -P 3306 -u default -e "SELECT sleepEachRow(1),now() FROM numbers(60)";'
if port == "8123":
long_cmd += 'wget -qO- "http://127.0.0.1:8123?query=SELECT sleepEachRow(1),now() FROM numbers(60)";'
long_cmd = f"echo '{long_cmd}' | xargs --verbose -i'{{}}' --no-run-if-empty -d ';' -P 100 bash -c '{{}}' 1>/dev/null"
with open("/tmp/long_cmd.sh", "w") as f:
f.write(long_cmd)
kubectl.launch(
f"cp /tmp/long_cmd.sh {too_many_queries_pod}:/tmp/long_cmd.sh -c clickhouse"
)
kubectl.launch(
f"exec -n {kubectl.namespace} {too_many_queries_pod} -c clickhouse -- bash /tmp/long_cmd.sh",
timeout=70,
)
def set_operator_version(version, ns=settings.operator_namespace, timeout=60):
operator_image = f"{settings.operator_docker_repo}:{version}"
metrics_exporter_image = f"{settings.metrics_exporter_docker_repo}:{version}"
kubectl.launch(f"set image deployment.v1.apps/clickhouse-operator clickhouse-operator={operator_image}", ns=ns)
kubectl.launch(f"set image deployment.v1.apps/clickhouse-operator metrics-exporter={metrics_exporter_image}", ns=ns)
kubectl.launch("rollout status deployment.v1.apps/clickhouse-operator", ns=ns, timeout=timeout)
if kubectl.get_count("pod", ns=ns, label=operator_label) == 0:
fail("invalid clickhouse-operator pod count")
def check_monitoring_metrics(operator_namespace, operator_pod, expect_result, max_retries=10):
with Then(f"metrics-exporter /metrics enpoint result should match with {expect_result}"):
for i in range(1, max_retries):
out = kubectl.launch(
f"exec {operator_pod} -c metrics-exporter -- wget -O- -q http://127.0.0.1:8888/metrics",
ns=operator_namespace
)
all_strings_expected_done = True
for string, exists in expect_result.items():
all_strings_expected_done = (exists == (string in out))
if not all_strings_expected_done:
break
if all_strings_expected_done:
break
with Then("Not ready. Wait for " + str(i * 5) + " seconds"):
time.sleep(i * 5)
assert all_strings_expected_done, error()
def test_ch_001(self):
util.require_zookeeper()
chit_data = manifest.get_chit_data(
util.get_full_path("templates/tpl-clickhouse-19.11.yaml"))
kubectl.launch(f"delete chit {chit_data['metadata']['name']}",
ns=settings.test_namespace)
kubectl.create_and_check(
"configs/test-ch-001-insert-quorum.yaml", {
"apply_templates": {"templates/tpl-clickhouse-20.8.yaml"},
"pod_count": 2,
"do_not_delete": 1,
})
chi = manifest.get_chi_name(
util.get_full_path("configs/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"
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
# 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
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
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
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"
)
print(out)
def restart_zookeeper():
kubectl.launch(
f"exec -n {kubectl.namespace} {zookeeper_pod} -- sh -c \"kill 1\"",
ok_to_fail=True,
)
def set_metrics_exporter_version(version, ns=settings.operator_namespace):
kubectl.launch(f"set image deployment.v1.apps/clickhouse-operator metrics-exporter=altinity/metrics-exporter:{version}", ns=ns)
kubectl.launch("rollout status deployment.v1.apps/clickhouse-operator", ns=ns)
def reboot_clickhouse_server():
kubectl.launch(
f"exec -n {kubectl.namespace} {clickhouse_pod} -c clickhouse -- kill 1",
ok_to_fail=True,
)
