def test_query_preempted(self): priority_pod, priority_svc, _, _ = alerts.random_pod_choice_for_callbacks( chi) 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, ) with Then("check ClickHouseQueryPreempted firing"): fired = alerts.wait_alert_state( "ClickHouseQueryPreempted", "firing", True, labels={"hostname": priority_svc}, time_range='30s', sleep_time=settings.prometheus_scrape_interval, callback=run_queries_with_priority) assert fired, error( "can't get ClickHouseQueryPreempted alert in firing state") with Then("check ClickHouseQueryPreempted gone away"): resolved = alerts.wait_alert_state("ClickHouseQueryPreempted", "firing", False, labels={"hostname": priority_svc}) assert resolved, error( "can't check ClickHouseQueryPreempted alert is gone away")
def test_longest_running_query(self): 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_distributed_connection_exceptions(self): delayed_pod, delayed_svc, restarted_pod, restarted_svc = alerts.random_pod_choice_for_callbacks( chi) clickhouse.create_distributed_table_on_cluster(chi) 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) with When("check ClickHouseDistributedConnectionExceptions firing"): fired = alerts.wait_alert_state( "ClickHouseDistributedConnectionExceptions", "firing", True, labels={ "hostname": delayed_svc, "chi": chi["metadata"]["name"] }, time_range='30s', callback=reboot_clickhouse_and_distributed_exection, ) assert fired, error( "can't get ClickHouseDistributedConnectionExceptions alert in firing state" ) with Then("check DistributedConnectionExpections gone away"): resolved = alerts.wait_alert_state( "ClickHouseDistributedConnectionExceptions", "firing", False, labels={"hostname": delayed_svc}) assert resolved, error( "can't check ClickHouseDistributedConnectionExceptions alert is gone away" ) kubectl.wait_pod_status(restarted_pod, "Running", ns=kubectl.namespace) kubectl.wait_jsonpath("pod", restarted_pod, "{.status.containerStatuses[0].ready}", "true", ns=kubectl.namespace) clickhouse.drop_distributed_table_on_cluster(chi)
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_system_settings_changed(self): changed_pod, changed_svc, _, _ = alerts.random_pod_choice_for_callbacks( chi) with When("apply changed settings"): kubectl.create_and_check( config="configs/test-cluster-for-alerts-changed-settings.yaml", check={ "apply_templates": [ "templates/tpl-clickhouse-stable.yaml", "templates/tpl-persistent-volume-100Mi.yaml" ], "object_counts": { "statefulset": 2, "pod": 2, "service": 3, }, "do_not_delete": 1 }) with Then("check ClickHouseSystemSettingsChanged firing"): fired = alerts.wait_alert_state("ClickHouseSystemSettingsChanged", "firing", True, labels={"hostname": changed_svc}, time_range="30s") assert fired, error( "can't get ClickHouseTooManyConnections alert in firing state") with When("rollback changed settings"): kubectl.create_and_check( config="configs/test-cluster-for-alerts.yaml", check={ "apply_templates": [ "templates/tpl-clickhouse-latest.yaml", "templates/tpl-clickhouse-alerts.yaml", "templates/tpl-persistent-volume-100Mi.yaml" ], "object_counts": { "statefulset": 2, "pod": 2, "service": 3, }, "do_not_delete": 1 }) with Then("check ClickHouseSystemSettingsChanged gone away"): resolved = alerts.wait_alert_state("ClickHouseSystemSettingsChanged", "firing", False, labels={"hostname": changed_svc}, sleep_time=30) assert resolved, error( "can't check ClickHouseTooManyConnections 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 test_detached_parts(self): clickhouse.create_table_on_cluster(chi) detached_pod, detached_svc, _, _ = alerts.random_pod_choice_for_callbacks( chi) 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 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) with When("check ClickHouseDetachedParts firing"): fired = alerts.wait_alert_state("ClickHouseDetachedParts", "firing", True, labels={ "hostname": detached_svc, "chi": chi["metadata"]["name"] }, time_range='30s', callback=create_part_and_detach) assert fired, error( "can't get ClickHouseDetachedParts alert in firing state") with Then("check ClickHouseDetachedParts gone away"): resolved = alerts.wait_alert_state("ClickHouseDetachedParts", "firing", False, labels={"hostname": detached_svc}, callback=attach_all_parts) assert resolved, error( "can't check ClickHouseDetachedParts alert is gone away") clickhouse.drop_table_on_cluster(chi)
def test_replicas_max_abosulute_delay(self): 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_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 test_backup_size(self): 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, 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 test_backup_is_down(self): reboot_pod, _, _, _ = alerts.random_pod_choice_for_callbacks(chi) def reboot_backup_container(): kubectl.launch( f"exec -n {settings.test_namespace} {reboot_pod} -c clickhouse-backup -- kill 1", ok_to_fail=True, ) with When("reboot clickhouse-backup"): fired = alerts.wait_alert_state("ClickHouseBackupDown", "firing", expected_state=True, callback=reboot_backup_container, labels={"pod_name": reboot_pod}, time_range='60s') assert fired, error( "can't get ClickHouseBackupDown alert in firing state") with Then("check ClickHouseBackupDown gone away"): resolved = alerts.wait_alert_state( "ClickHouseBackupDown", "firing", expected_state=False, sleep_time=settings.prometheus_scrape_interval, labels={"pod_name": reboot_pod}, ) assert resolved, error( "can't get ClickHouseBackupDown alert is gone away") with When("reboot clickhouse-backup again"): fired = alerts.wait_alert_state("ClickHouseBackupRecentlyRestart", "firing", expected_state=True, callback=reboot_backup_container, labels={"pod_name": reboot_pod}, time_range='60s') assert fired, error( "can't get ClickHouseBackupRecentlyRestart alert in firing state") with Then("check ClickHouseBackupRecentlyRestart gone away"): resolved = alerts.wait_alert_state("ClickHouseBackupRecentlyRestart", "firing", expected_state=False, time_range='30s', labels={"pod_name": reboot_pod}, sleep_time=30) assert resolved, error( "can't get ClickHouseBackupRecentlyRestart alert is gone away")
def test_zookeeper_hardware_exceptions(self): pod1, svc1, pod2, svc2 = alerts.random_pod_choice_for_callbacks(chi) chi_name = chi["metadata"]["name"] 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) with Then("check ClickHouseZooKeeperHardwareExceptions firing"): for svc in (svc1, svc2): fired = alerts.wait_alert_state( "ClickHouseZooKeeperHardwareExceptions", "firing", True, labels={"hostname": svc}, time_range='40s', sleep_time=settings.prometheus_scrape_interval, callback=restart_zookeeper) assert fired, error( "can't get ClickHouseZooKeeperHardwareExceptions alert in firing state" ) kubectl.wait_pod_status("zookeeper-0", "Running", ns=kubectl.namespace) kubectl.wait_jsonpath("pod", "zookeeper-0", "{.status.containerStatuses[0].ready}", "true", ns=kubectl.namespace) with Then("check ClickHouseZooKeeperHardwareExceptions gone away"): for svc in (svc1, svc2): resolved = alerts.wait_alert_state( "ClickHouseZooKeeperHardwareExceptions", "firing", False, labels={"hostname": svc}) assert resolved, error( "can't check ClickHouseZooKeeperHardwareExceptions alert is gone away" )
def test_distributed_sync_insertion_timeout(self): sync_pod, sync_svc, restarted_pod, restarted_svc = alerts.random_pod_choice_for_callbacks( chi) clickhouse.create_distributed_table_on_cluster( chi, local_engine='ENGINE Null()') def insert_distributed_sync(): with When("Insert to distributed table SYNC"): # look to https://github.com/ClickHouse/ClickHouse/pull/14260#issuecomment-683616862 # insert_sql = 'INSERT INTO default.test_distr SELECT now(), number FROM numbers(toUInt64(5e9))' insert_sql = "INSERT INTO FUNCTION remote('127.1', currentDatabase(), test_distr) SELECT now(), number FROM numbers(toUInt64(5e9))" insert_params = '--insert_distributed_timeout=1 --insert_distributed_sync=1' error = clickhouse.query_with_error(chi["metadata"]["name"], insert_sql, pod=sync_pod, host=sync_pod, ns=kubectl.namespace, advanced_params=insert_params) assert 'Code: 159' in error with When( "check ClickHouseDistributedSyncInsertionTimeoutExceeded firing"): fired = alerts.wait_alert_state( "ClickHouseDistributedSyncInsertionTimeoutExceeded", "firing", True, labels={ "hostname": sync_svc, "chi": chi["metadata"]["name"] }, time_range='30s', callback=insert_distributed_sync) assert fired, error( "can't get ClickHouseDistributedSyncInsertionTimeoutExceeded alert in firing state" ) with Then( "check ClickHouseDistributedSyncInsertionTimeoutExceeded gone away" ): resolved = alerts.wait_alert_state( "ClickHouseDistributedSyncInsertionTimeoutExceeded", "firing", False, labels={"hostname": sync_svc}) assert resolved, error( "can't check ClickHouseDistributedSyncInsertionTimeoutExceeded alert is gone away" ) clickhouse.drop_distributed_table_on_cluster(chi)
def test_backup_failed(self): backup_pod, _, _, _ = alerts.random_pod_choice_for_callbacks(chi) backup_prefix = prepare_table_for_backup(backup_pod) wait_backup_pod_ready_and_curl_installed(backup_pod) 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 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') with When("clickhouse-backup create failed"): fired = alerts.wait_alert_state("ClickHouseBackupFailed", "firing", expected_state=True, callback=create_fail_backup, labels={"pod_name": backup_pod}, time_range='60s') assert fired, error( "can't get ClickHouseBackupFailed alert in firing state") with Then("check ClickHouseBackupFailed gone away"): resolved = alerts.wait_alert_state("ClickHouseBackupFailed", "firing", expected_state=False, callback=create_success_backup, labels={"pod_name": backup_pod}, sleep_time=30) assert resolved, error( "can't get ClickHouseBackupFailed alert is gone away")
def test_backup_is_success(self): _, _, backup_pod, _ = alerts.random_pod_choice_for_callbacks(chi) backup_name = prepare_table_for_backup(backup_pod) wait_backup_pod_ready_and_curl_installed(backup_pod) with When('Backup is success'): backup_successful_before = get_backup_metric_value( backup_pod, 'clickhouse_backup_successful_backups|clickhouse_backup_successful_creates' ) list_before = exec_on_backup_container( backup_pod, 'curl -sL http://127.0.0.1:7171/backup/list') 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') exec_on_backup_container( backup_pod, f'curl -X POST -sL "http://127.0.0.1:7171/backup/upload/{backup_name}"' ) wait_backup_command_status(backup_pod, f'upload {backup_name}', expected_status='success') with Then('list of backups shall changed'): list_after = exec_on_backup_container( backup_pod, 'curl -sL http://127.0.0.1:7171/backup/list') assert list_before != list_after, error("backup is not created") with Then('successful backup count shall increased'): backup_successful_after = get_backup_metric_value( backup_pod, 'clickhouse_backup_successful_backups|clickhouse_backup_successful_creates' ) assert backup_successful_before != backup_successful_after, error( "clickhouse_backup_successful_backups shall increased")
def test_insert_related_alerts(self): 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) # @TODO we need only one query after resolve https://github.com/ClickHouse/ClickHouse/issues/11384 and switch to 21.3+ 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) 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_distributed_files_to_insert(self): 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 test_backup_duration(self): short_pod, _, long_pod, _ = alerts.random_pod_choice_for_callbacks(chi) apply_fake_backup("prepare fake backup duration metric") for pod in [short_pod, long_pod]: with Then(f"wait {pod} ready"): kubectl.wait_field("pod", pod, ".spec.containers[1].image", "nginx:latest") kubectl.wait_field("pod", pod, ".status.containerStatuses[1].ready", "true") fired = alerts.wait_alert_state( "ClickHouseBackupTooLong", "firing", expected_state=True, sleep_time=settings.prometheus_scrape_interval, labels={"pod_name": pod}, time_range='60s') assert fired, error( f"can't get ClickHouseBackupTooLong alert in firing state for {pod}" ) with Then(f"wait when prometheus will scrape fake data"): time.sleep(70) with Then(f"decrease {short_pod} backup duration"): kubectl.launch( f'exec {short_pod} -c clickhouse-backup -- bash -xc \'' 'echo "# HELP clickhouse_backup_last_create_duration Backup create duration in nanoseconds" > /usr/share/nginx/html/metrics && ' 'echo "# TYPE clickhouse_backup_last_create_duration gauge" >> /usr/share/nginx/html/metrics && ' 'echo "clickhouse_backup_last_create_duration 7000000000000" >> /usr/share/nginx/html/metrics && ' 'echo "# HELP clickhouse_backup_last_create_status Last backup create status: 0=failed, 1=success, 2=unknown" >> /usr/share/nginx/html/metrics && ' 'echo "# TYPE clickhouse_backup_last_create_status gauge" >> /usr/share/nginx/html/metrics && ' 'echo "clickhouse_backup_last_create_status 1" >> /usr/share/nginx/html/metrics' '\'') fired = alerts.wait_alert_state( "ClickHouseBackupTooShort", "firing", expected_state=True, sleep_time=settings.prometheus_scrape_interval, labels={"pod_name": short_pod}, time_range='60s') assert fired, error( "can't get ClickHouseBackupTooShort alert in firing state") apply_normal_backup() with Then("check ClickHouseBackupTooShort gone away"): resolved = alerts.wait_alert_state("ClickHouseBackupTooShort", "firing", expected_state=False, labels={"pod_name": short_pod}) assert resolved, error( "can't get ClickHouseBackupTooShort alert is gone away") with Then("check ClickHouseBackupTooLong gone away"): resolved = alerts.wait_alert_state("ClickHouseBackupTooLong", "firing", expected_state=False, labels={"pod_name": long_pod}) assert resolved, error( "can't get ClickHouseBackupTooLong alert is gone away")
def test_read_only_replica(self): read_only_pod, read_only_svc, other_pod, other_svc = alerts.random_pod_choice_for_callbacks( chi) chi_name = chi["metadata"]["name"] 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()' ) 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) with Then("check ClickHouseReadonlyReplica firing"): fired = alerts.wait_alert_state( "ClickHouseReadonlyReplica", "firing", True, labels={"hostname": read_only_svc}, time_range='30s', sleep_time=settings.prometheus_scrape_interval, callback=restart_zookeeper) assert fired, error( "can't get ClickHouseReadonlyReplica alert in firing state") with Then("check ClickHouseReadonlyReplica gone away"): resolved = alerts.wait_alert_state("ClickHouseReadonlyReplica", "firing", False, labels={"hostname": read_only_svc}) assert resolved, error( "can't check ClickHouseReadonlyReplica alert is gone away") kubectl.wait_pod_status("zookeeper-0", "Running", ns=kubectl.namespace) kubectl.wait_jsonpath("pod", "zookeeper-0", "{.status.containerStatuses[0].ready}", "true", ns=kubectl.namespace) for i in range(11): zookeeper_status = kubectl.launch( f"exec -n {kubectl.namespace} zookeeper-0 -- sh -c \"echo ruok | nc 127.0.0.1 2181\"", ok_to_fail=True) if "imok" in zookeeper_status: break elif i == 10: fail(f"invalid zookeeper status after {i} retries") with Then("zookeper is not ready, wait 2 seconds"): time.sleep(2) clickhouse.query_with_error( chi_name, "SYSTEM RESTART REPLICAS; SYSTEM SYNC REPLICA default.test_repl", host=read_only_svc, timeout=240) clickhouse.query_with_error( chi_name, "SYSTEM RESTART REPLICAS; SYSTEM SYNC REPLICA default.test_repl", host=other_svc, timeout=240) clickhouse.drop_table_on_cluster(chi, 'all-replicated', 'default.test_repl')