[#232] add test object metrics

This commit is contained in:
Ilyas Niyazov 2024-05-23 10:33:20 +03:00
parent 3941619431
commit a95aab62dc
2 changed files with 324 additions and 2 deletions

View file

@ -23,7 +23,7 @@ from frostfs_testlib.resources.common import (
)
from frostfs_testlib.s3 import AwsCliClient, Boto3ClientWrapper, S3ClientWrapper, VersioningStatus
from frostfs_testlib.shell import LocalShell, Shell
from frostfs_testlib.steps.cli.container import DEFAULT_PLACEMENT_RULE, DEFAULT_EC_PLACEMENT_RULE
from frostfs_testlib.steps.cli.container import DEFAULT_EC_PLACEMENT_RULE, DEFAULT_PLACEMENT_RULE
from frostfs_testlib.steps.cli.object import get_netmap_netinfo
from frostfs_testlib.steps.s3 import s3_helper
from frostfs_testlib.storage import get_service_registry
@ -210,6 +210,7 @@ def object_size(
return complex_object_size
@pytest.fixture(scope="session")
def rep_placement_policy() -> PlacementPolicy:
return PlacementPolicy("rep", DEFAULT_PLACEMENT_RULE)
@ -234,6 +235,7 @@ def placement_policy(
return ec_placement_policy
@pytest.fixture(scope="session")
def cluster(temp_directory: str, hosting: Hosting, client_shell: Shell) -> Cluster:
cluster = Cluster(hosting)
@ -266,10 +268,12 @@ def healthcheck(cluster: Cluster) -> Healthcheck:
return healthcheck_cls()
@pytest.fixture(scope="session")
@pytest.fixture
def cluster_state_controller(client_shell: Shell, cluster: Cluster, healthcheck: Healthcheck) -> ClusterStateController:
controller = ClusterStateController(client_shell, cluster, healthcheck)
yield controller
controller.start_stopped_hosts()
controller.start_all_stopped_services()
@pytest.fixture(scope="session")

View file

@ -0,0 +1,318 @@
import random
import re
import allure
import pytest
from frostfs_testlib import reporter
from frostfs_testlib.steps.cli.container import create_container, delete_container, search_nodes_with_container
from frostfs_testlib.steps.cli.object import (
delete_object,
get_object_nodes,
lock_object,
put_object,
put_object_to_random_node,
)
from frostfs_testlib.storage.cluster import Cluster, ClusterNode
from frostfs_testlib.storage.controllers.cluster_state_controller import ClusterStateController
from frostfs_testlib.storage.dataclasses.object_size import ObjectSize
from frostfs_testlib.storage.dataclasses.wallet import WalletInfo
from frostfs_testlib.testing.cluster_test_base import ClusterTestBase
from frostfs_testlib.testing.test_control import wait_for_success
from frostfs_testlib.utils.file_utils import generate_file
class TestObjectMetrics(ClusterTestBase):
@wait_for_success(interval=10)
def check_metrics_by_type(
self, cluster_nodes: list[ClusterNode], metric_command: str, grep_by: str, metric_type: str, counter_exp: int
):
counter_act = 0
for cluster_node in cluster_nodes:
try:
metric_result = cluster_node.metrics.storage.get_metrics_search_by_greps(
command=metric_command, grep_by=grep_by
)
counter_act += self.calc_metrics_count_from_stdout(metric_result.stdout, metric_type)
except RuntimeError as e:
...
assert (
counter_act == counter_exp
), f"Expected metric {metric_type}={counter_exp}, Actual: {counter_act} in nodes: {cluster_nodes}"
@staticmethod
def calc_metrics_count_from_stdout(metric_result_stdout: str, metric_type: str):
result = re.findall(rf'type="{metric_type}"}}\s(\d+)', metric_result_stdout)
return sum(map(int, result))
@wait_for_success(interval=10)
def check_object_metrics_total_and_container(
self, cluster_nodes: list[ClusterNode], cid: str, objects_metric_total: int, objects_metric_container: int
):
self.check_metrics_by_type(
cluster_nodes,
"frostfs_node_engine_objects_total",
grep_by="user",
metric_type="user",
counter_exp=objects_metric_total,
)
objects_metric_container_act = 0
for node in cluster_nodes:
try:
metrics_container = node.metrics.storage.get_metrics_search_by_greps(
command="frostfs_node_engine_container_objects_total", cid=cid, type="user"
)
objects_metric_container_act += self.calc_metrics_count_from_stdout(
metrics_container.stdout, metric_type="user"
)
except RuntimeError as e:
...
assert (
objects_metric_container_act == objects_metric_container
), f"Expected {objects_metric_container} objects in container"
@wait_for_success(max_wait_time=120, interval=10)
def check_object_metrics_container(
self, cluster_nodes: list[ClusterNode], cid: str, objects_metric_container_exp: int
):
objects_metric_container_act = 0
for node in cluster_nodes:
try:
metrics_container = node.metrics.storage.get_metrics_search_by_greps(
command="frostfs_node_engine_container_objects_total", cid=cid, type="user"
)
objects_metric_container_act += self.calc_metrics_count_from_stdout(
metrics_container.stdout, metric_type="user"
)
except RuntimeError as e:
...
assert (
objects_metric_container_act == objects_metric_container_exp
), f"Expected {objects_metric_container_exp} objects in container"
@allure.title("Object metrics of removed container")
def test_object_metrics_removed_container(
self, object_size: ObjectSize, default_wallet: WalletInfo, cluster: Cluster
):
file_path = generate_file(object_size.value)
placement_policy = "REP 2 IN X CBF 2 SELECT 2 FROM * AS X"
copies = 2
with reporter.step(f"Create container with policy {placement_policy}"):
cid = create_container(
default_wallet,
rule=placement_policy,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
)
with reporter.step("Put object to random node"):
storage_object_id = put_object_to_random_node(
wallet=default_wallet,
path=file_path,
cid=cid,
shell=self.shell,
cluster=cluster,
)
with reporter.step("Check metric appears in node where the object is located"):
object_nodes = get_object_nodes(
cluster=cluster, cid=cid, oid=storage_object_id, alive_node=cluster.cluster_nodes[0]
)
self.check_metrics_by_type(object_nodes, "frostfs_node_engine_container_objects_total", cid, "user", copies)
with reporter.step("Delete container"):
delete_container(default_wallet, cid, shell=self.shell, endpoint=self.cluster.default_rpc_endpoint)
with reporter.step("Tick Epoch"):
self.tick_epochs(epochs_to_tick=2, wait_block=2)
with reporter.step("Check metrics of removed containers doesn't appear in the storage node"):
self.check_metrics_by_type(object_nodes, "frostfs_node_engine_container_objects_total", cid, "user", 0)
for node in object_nodes:
with pytest.raises(RuntimeError):
node.metrics.storage.get_metric_container(f"frostfs_node_engine_container_size_byte", cid)
@allure.title("Object metrics, locked object, (policy={placement_policy})")
@pytest.mark.parametrize(
"placement_policy", ["REP 1 IN X CBF 1 SELECT 1 FROM * AS X", "REP 2 IN X CBF 2 SELECT 2 FROM * AS X"]
)
def test_object_metrics_blocked_object(
self, object_size: ObjectSize, default_wallet: WalletInfo, cluster: Cluster, placement_policy: str
):
file_path = generate_file(object_size.value)
metric_step = int(re.search(r"REP\s(\d+)", placement_policy).group(1))
with reporter.step(f"Create container with policy {placement_policy}"):
cid = create_container(
wallet=default_wallet,
rule=placement_policy,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
)
with reporter.step("Search container nodes"):
container_nodes = search_nodes_with_container(
wallet=default_wallet,
cid=cid,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
cluster=cluster,
)
with reporter.step("Get current metrics for metric_type=user"):
objects_metric_counter = 0
for node in container_nodes:
metric_objects_total = node.metrics.storage.get_metrics_search_by_greps(
command="frostfs_node_engine_objects_total", type="user"
)
objects_metric_counter += self.calc_metrics_count_from_stdout(
metric_objects_total.stdout, metric_type="user"
)
with reporter.step("Put object to container node"):
oid = put_object(
wallet=default_wallet,
path=file_path,
cid=cid,
shell=self.shell,
endpoint=container_nodes[0].storage_node.get_rpc_endpoint(),
)
with reporter.step(f"Check metric user 'the counter should increase by {metric_step}'"):
objects_metric_counter += metric_step
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, metric_step)
with reporter.step("Delete object"):
delete_object(default_wallet, cid, oid, self.shell, self.cluster.default_rpc_endpoint)
with reporter.step(f"Check metric user 'the counter should decrease by {metric_step}'"):
objects_metric_counter -= metric_step
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, 0)
with reporter.step("Put object and lock it"):
oid = put_object(
default_wallet, file_path, cid, self.shell, container_nodes[0].storage_node.get_rpc_endpoint()
)
current_epoch = self.get_epoch()
lock_object(
default_wallet,
cid,
oid,
self.shell,
container_nodes[0].storage_node.get_rpc_endpoint(),
expire_at=current_epoch + 1,
)
with reporter.step(f"Check metric user 'the counter should increase by {metric_step}'"):
objects_metric_counter += metric_step
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, metric_step)
with reporter.step(f"Wait until remove locking 'the counter doesn't change'"):
self.tick_epochs(epochs_to_tick=2)
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, metric_step)
with reporter.step("Delete object"):
delete_object(default_wallet, cid, oid, self.shell, self.cluster.default_rpc_endpoint)
with reporter.step(f"Check metric user 'the counter should decrease by {metric_step}'"):
objects_metric_counter -= metric_step
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, 0)
with reporter.step("Put object with expire_at"):
current_epoch = self.get_epoch()
oid = put_object(
default_wallet,
file_path,
cid,
self.shell,
container_nodes[0].storage_node.get_rpc_endpoint(),
expire_at=current_epoch + 1,
)
with reporter.step(f"Check metric user 'the counter should increase by {metric_step}'"):
objects_metric_counter += metric_step
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, metric_step)
with reporter.step("Tick Epoch"):
self.tick_epochs(epochs_to_tick=2)
with reporter.step(f"Check metric user 'the counter should decrease by {metric_step}'"):
objects_metric_counter -= metric_step
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, 0)
with reporter.step("Delete container"):
delete_container(default_wallet, cid, shell=self.shell, endpoint=self.cluster.default_rpc_endpoint)
@allure.title("Object metrics, stop the node")
def test_object_metrics_stop_node(
self,
object_size: ObjectSize,
max_object_size: int,
default_wallet: WalletInfo,
cluster_state_controller: ClusterStateController,
):
placement_policy = "REP 2 IN X CBF 2 SELECT 2 FROM * AS X"
file_path = generate_file(object_size.value)
copies = 2
with reporter.step(f"Create container with policy {placement_policy}"):
cid = create_container(
wallet=default_wallet,
rule=placement_policy,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
)
with reporter.step("Search container nodes"):
container_nodes = search_nodes_with_container(
wallet=default_wallet,
cid=cid,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
cluster=self.cluster,
)
with reporter.step("Get current metrics for container nodes"):
objects_metric_counter = 0
for node in container_nodes:
metric_objects_total = node.metrics.storage.get_metrics_search_by_greps(
command="frostfs_node_engine_objects_total", type="user"
)
objects_metric_counter += self.calc_metrics_count_from_stdout(
metric_objects_total.stdout, metric_type="user"
)
with reporter.step("Put object to container node"):
oid = put_object(
wallet=default_wallet,
path=file_path,
cid=cid,
shell=self.shell,
endpoint=container_nodes[0].storage_node.get_rpc_endpoint(),
)
with reporter.step(f"Check metric in container nodes 'the counter should increase by {copies}'"):
objects_metric_counter += copies
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, copies)
with reporter.step(f"Select node to stop"):
node_to_stop = container_nodes[0]
alive_nodes = [node for node in container_nodes if node != node_to_stop]
with reporter.step(f"Stop the node, wait until the object is replicated to another node"):
cluster_state_controller.stop_node_host(node_to_stop, "hard")
with reporter.step(f"Check metric in alive nodes 'the counter should increase by 1'"):
self.check_object_metrics_container(alive_nodes, cid, copies)
with reporter.step("Start node"):
cluster_state_controller.start_node_host(node_to_stop)
with reporter.step(f"Check metric in container nodes 'the counter doesn't change'"):
self.check_object_metrics_total_and_container(container_nodes, cid, objects_metric_counter, copies)
with reporter.step("Delete container"):
delete_container(default_wallet, cid, shell=self.shell, endpoint=self.cluster.default_rpc_endpoint)