def test_operator_restart(config, version=settings.operator_version):
with Given(f"clickhouse-operator {version}"):
set_operator_version(version)
config = util.get_full_path(config)
chi = manifest.get_chi_name(config)
cluster = chi
kubectl.create_and_check(config=config,
check={
"object_counts": {
"statefulset": 1,
"pod": 1,
"service": 2,
},
"do_not_delete": 1,
})
start_time = kubectl.get_field("pod", f"chi-{chi}-{cluster}-0-0-0",
".status.startTime")
with When("Restart operator"):
restart_operator()
time.sleep(5)
kubectl.wait_chi_status(chi, "Completed")
kubectl.wait_objects(chi, {
"statefulset": 1,
"pod": 1,
"service": 2,
})
new_start_time = kubectl.get_field("pod",
f"chi-{chi}-{cluster}-0-0-0",
".status.startTime")
assert start_time == new_start_time
kubectl.delete_chi(chi)
def test_operator_upgrade(config,
version_from,
version_to=settings.operator_version):
version_to = settings.operator_version
with Given(f"clickhouse-operator {version_from}"):
set_operator_version(version_from)
config = util.get_full_path(config)
chi = manifest.get_chi_name(config)
kubectl.create_and_check(config=config,
check={
"object_counts": {
"statefulset": 1,
"pod": 1,
"service": 2,
},
"do_not_delete": 1,
})
start_time = kubectl.get_field("pod", f"chi-{chi}-{chi}-0-0-0",
".status.startTime")
with When(f"upgrade operator to {version_to}"):
set_operator_version(version_to, timeout=120)
time.sleep(5)
kubectl.wait_chi_status(chi, "Completed", retries=6)
kubectl.wait_objects(chi, {
"statefulset": 1,
"pod": 1,
"service": 2
})
new_start_time = kubectl.get_field("pod", f"chi-{chi}-{chi}-0-0-0",
".status.startTime")
assert start_time == new_start_time
kubectl.delete_chi(chi)
def test_020(config="configs/test-020-multi-volume.yaml"):
chi = manifest.get_chi_name(util.get_full_path(config))
kubectl.create_and_check(
config=config,
check={
"pod_count": 1,
"pod_volumes": {
"/var/lib/clickhouse",
"/var/lib/clickhouse2",
},
"do_not_delete": 1,
})
with When("Create a table and insert 1 row"):
clickhouse.query(chi, "create table test_disks(a Int8) Engine = MergeTree() order by a")
clickhouse.query(chi, "insert into test_disks values (1)")
with Then("Data should be placed on default disk"):
out = clickhouse.query(chi, "select disk_name from system.parts where table='test_disks'")
assert out == 'default'
with When("alter table test_disks move partition tuple() to disk 'disk2'"):
clickhouse.query(chi, "alter table test_disks move partition tuple() to disk 'disk2'")
with Then("Data should be placed on disk2"):
out = clickhouse.query(chi, "select disk_name from system.parts where table='test_disks'")
assert out == 'disk2'
kubectl.delete_chi(chi)
def test_022(config="configs/test-022-broken-image.yaml"):
chi = manifest.get_chi_name(util.get_full_path(config))
kubectl.create_and_check(config=config,
check={
"pod_count": 1,
"do_not_delete": 1,
"chi_status": "InProgress",
})
with When("ClickHouse image can not be retrieved"):
kubectl.wait_field(
"pod", "chi-test-022-broken-image-default-0-0-0",
".status.containerStatuses[0].state.waiting.reason",
"ErrImagePull")
kubectl.delete_chi(chi)
def test_019(config="configs/test-019-retain-volume.yaml"):
require_zookeeper()
chi = manifest.get_chi_name(util.get_full_path(config))
kubectl.create_and_check(
config=config,
check={
"pod_count": 1,
"do_not_delete": 1,
})
create_non_replicated_table = "create table t1 Engine = Log as select 1 as a"
create_replicated_table = """
create table t2
Engine = ReplicatedMergeTree('/clickhouse/{installation}/{cluster}/tables/{shard}/{database}/{table}', '{replica}')
partition by tuple() order by a
as select 1 as a""".replace('\r', '').replace('\n', '')
with Given("ClickHouse has some data in place"):
clickhouse.query(chi, sql=create_non_replicated_table)
clickhouse.query(chi, sql=create_replicated_table)
with When("CHI with retained volume is deleted"):
pvc_count = kubectl.get_count("pvc")
pv_count = kubectl.get_count("pv")
kubectl.delete_chi(chi)
with Then("PVC should be retained"):
assert kubectl.get_count("pvc") == pvc_count
assert kubectl.get_count("pv") == pv_count
with When("Re-create CHI"):
kubectl.create_and_check(
config=config,
check={
"pod_count": 1,
"do_not_delete": 1,
})
with Then("PVC should be re-mounted"):
with And("Non-replicated table should have data"):
out = clickhouse.query(chi, sql="select a from t1")
assert out == "1"
with And("Replicated table should have data"):
out = clickhouse.query(chi, sql="select a from t2")
assert out == "1"
kubectl.delete_chi(chi)
def create_and_check(config, check, ns=namespace, timeout=600):
config = util.get_full_path(config)
chi_name = manifest.get_chi_name(config)
if "apply_templates" in check:
print("Need to apply additional templates")
for t in check["apply_templates"]:
print("Applying template:" + t)
apply(util.get_full_path(t), ns=ns)
time.sleep(5)
apply(config, ns=ns, timeout=timeout)
if "object_counts" in check:
wait_objects(chi_name, check["object_counts"], ns=ns)
if "pod_count" in check:
wait_object("pod",
"",
label=f"-l clickhouse.altinity.com/chi={chi_name}",
count=check["pod_count"],
ns=ns)
if "chi_status" in check:
wait_chi_status(chi_name, check["chi_status"], ns=ns)
else:
wait_chi_status(chi_name, "Completed", ns=ns)
if "pod_image" in check:
check_pod_image(chi_name, check["pod_image"], ns=ns)
if "pod_volumes" in check:
check_pod_volumes(chi_name, check["pod_volumes"], ns=ns)
if "pod_podAntiAffinity" in check:
check_pod_antiaffinity(chi_name, ns=ns)
if "pod_ports" in check:
check_pod_ports(chi_name, check["pod_ports"], ns=ns)
if "service" in check:
check_service(check["service"][0], check["service"][1], ns=ns)
if "configmaps" in check:
check_configmaps(chi_name, ns=ns)
if "do_not_delete" not in check:
delete_chi(chi_name, ns=ns)
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 test_014():
require_zookeeper()
create_table = """
CREATE TABLE test_local(a Int8)
Engine = ReplicatedMergeTree('/clickhouse/{installation}/{cluster}/tables/{shard}/{database}/{table}', '{replica}')
PARTITION BY tuple()
ORDER BY a
""".replace('\r', '').replace('\n', '')
config = "configs/test-014-replication-1.yaml"
chi = manifest.get_chi_name(util.get_full_path(config))
cluster = "default"
kubectl.create_and_check(
config=config,
check={
"apply_templates": {
settings.clickhouse_template,
"templates/tpl-persistent-volume-100Mi.yaml",
},
"object_counts": {
"statefulset": 2,
"pod": 2,
"service": 3,
},
"do_not_delete": 1,
})
start_time = kubectl.get_field("pod", f"chi-{chi}-{cluster}-0-0-0",
".status.startTime")
schema_objects = ['test_local', 'test_view', 'test_mv', 'a_view']
with Given("Create schema objects"):
clickhouse.query(chi, create_table, host=f"chi-{chi}-{cluster}-0-0")
clickhouse.query(chi,
"CREATE VIEW test_view as SELECT * from test_local",
host=f"chi-{chi}-{cluster}-0-0")
clickhouse.query(chi,
"CREATE VIEW a_view as SELECT * from test_view",
host=f"chi-{chi}-{cluster}-0-0")
clickhouse.query(
chi,
"CREATE MATERIALIZED VIEW test_mv Engine = Log as SELECT * from test_local",
host=f"chi-{chi}-{cluster}-0-0")
clickhouse.query(
chi,
"CREATE DICTIONARY test_dict (a Int8, b Int8) PRIMARY KEY a SOURCE(CLICKHOUSE(host 'localhost' port 9000 table 'test_local' user 'default')) LAYOUT(FLAT()) LIFETIME(0)",
host=f"chi-{chi}-{cluster}-0-0")
with Given(
"Replicated table is created on a first replica and data is inserted"
):
clickhouse.query(chi,
"INSERT INTO test_local values(1)",
host=f"chi-{chi}-{cluster}-0-0")
with When("Table is created on the second replica"):
clickhouse.query(chi,
create_table,
host=f"chi-{chi}-{cluster}-0-1")
# Give some time for replication to catch up
time.sleep(10)
with Then("Data should be replicated"):
out = clickhouse.query(chi,
"SELECT a FROM test_local",
host=f"chi-{chi}-{cluster}-0-1")
assert out == "1"
with When("Add one more replica"):
kubectl.create_and_check(config="configs/test-014-replication-2.yaml",
check={
"pod_count": 3,
"do_not_delete": 1,
})
# Give some time for replication to catch up
time.sleep(10)
new_start_time = kubectl.get_field("pod", f"chi-{chi}-{cluster}-0-0-0",
".status.startTime")
assert start_time == new_start_time
with Then("Schema objects should be migrated to the new replica"):
for obj in schema_objects:
out = clickhouse.query(
chi,
f"SELECT count() FROM system.tables WHERE name = '{obj}'",
host=f"chi-{chi}-{cluster}-0-2")
assert out == "1"
# Check dictionary
out = clickhouse.query(
chi,
f"SELECT count() FROM system.dictionaries WHERE name = 'test_dict'",
host=f"chi-{chi}-{cluster}-0-2")
assert out == "1"
with And("Replicated table should have the data"):
out = clickhouse.query(chi,
"SELECT a FROM test_local",
host=f"chi-{chi}-{cluster}-0-2")
assert out == "1"
with When("Remove replica"):
kubectl.create_and_check(config=config,
check={
"pod_count": 1,
"do_not_delete": 1,
})
new_start_time = kubectl.get_field("pod", f"chi-{chi}-{cluster}-0-0-0",
".status.startTime")
assert start_time == new_start_time
with Then("Replica needs to be removed from the Zookeeper as well"):
out = clickhouse.query(
chi,
"SELECT count() FROM system.replicas WHERE table='test_local'")
assert out == "1"
with When("Restart Zookeeper pod"):
with Then("Delete Zookeeper pod"):
kubectl.launch("delete pod zookeeper-0")
time.sleep(1)
with Then(
"Insert into the table while there is no Zookeeper -- table should be in readonly mode"
):
out = clickhouse.query_with_error(
chi, "INSERT INTO test_local values(2)")
assert "Table is in readonly mode" in out
with Then("Wait for Zookeeper pod to come back"):
kubectl.wait_object("pod", "zookeeper-0")
kubectl.wait_pod_status("zookeeper-0", "Running")
with Then(
"Wait for ClickHouse to reconnect to Zookeeper and switch to read-write mode"
):
time.sleep(30)
# with Then("Restart clickhouse pods"):
# kubectl("delete pod chi-test-014-replication-default-0-0-0")
# kubectl("delete pod chi-test-014-replication-default-0-1-0")
with Then("Table should be back to normal"):
clickhouse.query(chi, "INSERT INTO test_local values(3)")
kubectl.delete_chi("test-014-replication")
def test_013():
config = "configs/test-013-add-shards-1.yaml"
chi = manifest.get_chi_name(util.get_full_path(config))
cluster = "default"
kubectl.create_and_check(config=config,
check={
"apply_templates": {
settings.clickhouse_template,
},
"object_counts": {
"statefulset": 1,
"pod": 1,
"service": 2,
},
"do_not_delete": 1,
})
start_time = kubectl.get_field("pod", f"chi-{chi}-{cluster}-0-0-0",
".status.startTime")
schema_objects = [
'test_local',
'test_distr',
'events-distr',
]
with Then("Create local and distributed tables"):
clickhouse.query(
chi,
"CREATE TABLE test_local Engine = Log as SELECT * FROM system.one")
clickhouse.query(
chi,
"CREATE TABLE test_distr as test_local Engine = Distributed('default', default, test_local)"
)
clickhouse.query(chi, "CREATE DATABASE \\\"test-db\\\"")
clickhouse.query(
chi,
"CREATE TABLE \\\"test-db\\\".\\\"events-distr\\\" as system.events "
"ENGINE = Distributed('all-sharded', system, events)")
with Then("Add shards"):
kubectl.create_and_check(config="configs/test-013-add-shards-2.yaml",
check={
"object_counts": {
"statefulset": 3,
"pod": 3,
"service": 4,
},
"do_not_delete": 1,
})
# Give some time for replication to catch up
time.sleep(10)
with Then("Unaffected pod should not be restarted"):
new_start_time = kubectl.get_field("pod", f"chi-{chi}-{cluster}-0-0-0",
".status.startTime")
assert start_time == new_start_time
with And("Schema objects should be migrated to new shards"):
for obj in schema_objects:
out = clickhouse.query(
chi,
f"SELECT count() FROM system.tables WHERE name = '{obj}'",
host=f"chi-{chi}-{cluster}-1-0")
assert out == "1"
out = clickhouse.query(
chi,
f"SELECT count() FROM system.tables WHERE name = '{obj}'",
host=f"chi-{chi}-{cluster}-2-0")
assert out == "1"
with When("Remove shards"):
kubectl.create_and_check(config=config,
check={
"object_counts": {
"statefulset": 1,
"pod": 1,
"service": 2,
},
"do_not_delete": 1,
})
time.sleep(10)
with Then("Unaffected pod should not be restarted"):
new_start_time = kubectl.get_field("pod",
f"chi-{chi}-{cluster}-0-0-0",
".status.startTime")
assert start_time == new_start_time
kubectl.delete_chi(chi)
def test_021(config="configs/test-021-rescale-volume-01.yaml"):
with Given("Default storage class is expandable"):
default_storage_class = kubectl.get_default_storage_class()
assert default_storage_class is not None
assert len(default_storage_class) > 0
allow_volume_expansion = kubectl.get_field("storageclass",
default_storage_class,
".allowVolumeExpansion")
if allow_volume_expansion != "true":
kubectl.launch(
f"patch storageclass {default_storage_class} -p '{{\"allowVolumeExpansion\":true}}'"
)
chi = manifest.get_chi_name(util.get_full_path(config))
kubectl.create_and_check(config=config,
check={
"pod_count": 1,
"do_not_delete": 1,
})
with Then("Storage size should be 100Mi"):
size = kubectl.get_pvc_size(
"disk1-chi-test-021-rescale-volume-simple-0-0-0")
assert size == "100Mi"
with When("Re-scale volume configuration to 200Mb"):
kubectl.create_and_check(
config="configs/test-021-rescale-volume-02-enlarge-disk.yaml",
check={
"pod_count": 1,
"do_not_delete": 1,
})
with Then("Storage size should be 200Mi"):
size = kubectl.get_pvc_size(
"disk1-chi-test-021-rescale-volume-simple-0-0-0")
assert size == "200Mi"
with When("Add second disk 50Mi"):
kubectl.create_and_check(
config="configs/test-021-rescale-volume-03-add-disk.yaml",
check={
"pod_count": 1,
"pod_volumes": {
"/var/lib/clickhouse",
"/var/lib/clickhouse2",
},
"do_not_delete": 1,
})
with Then("There should be two PVC"):
size = kubectl.get_pvc_size(
"disk1-chi-test-021-rescale-volume-simple-0-0-0")
assert size == "200Mi"
size = kubectl.get_pvc_size(
"disk2-chi-test-021-rescale-volume-simple-0-0-0")
assert size == "50Mi"
with And("There should be two disks recognized by ClickHouse"):
# ClickHouse requires some time to mount volume. Race conditions.
time.sleep(120)
out = clickhouse.query(chi, "SELECT count() FROM system.disks")
print("SELECT count() FROM system.disks RETURNED:")
print(out)
assert out == "2"
kubectl.delete_chi(chi)