frostfs-testcases/pytest_tests/testsuites/metrics/test_shard_metrics.py

import random
import re

import allure
import pytest
from frostfs_testlib import reporter
from frostfs_testlib.resources.error_patterns import OBJECT_NOT_FOUND
from frostfs_testlib.steps.cli.object import get_object, put_object
from frostfs_testlib.steps.metrics import check_metrics_counter
from frostfs_testlib.steps.node_management import node_shard_list, node_shard_set_mode
from frostfs_testlib.steps.storage_policy import get_nodes_with_object
from frostfs_testlib.storage.cluster import Cluster, ClusterNode
from frostfs_testlib.storage.controllers import ShardsWatcher
from frostfs_testlib.storage.dataclasses.wallet import WalletInfo
from frostfs_testlib.testing import parallel, wait_for_success
from frostfs_testlib.testing.cluster_test_base import ClusterTestBase
from frostfs_testlib.utils.file_utils import generate_file

from ...helpers.container_request import PUBLIC_WITH_POLICY, requires_container


@pytest.mark.nightly
@pytest.mark.metrics
class TestShardMetrics(ClusterTestBase):
    @pytest.fixture()
    @allure.title("Get two shards for set mode")
    def two_shards_and_node(self, cluster: Cluster) -> tuple[str, str, ClusterNode]:
        node = random.choice(cluster.cluster_nodes)
        shards = node_shard_list(node.storage_node)
        two_shards = random.sample(shards, k=2)

        yield two_shards[0], two_shards[1], node

        for shard in two_shards:
            node_shard_set_mode(node.storage_node, shard, "read-write")

        node_shard_list(node.storage_node)

    @pytest.fixture()
    @allure.title("Revert all shards mode")
    def revert_all_shards_mode(self):
        yield
        parallel(self.set_shard_rw_mode, self.cluster.cluster_nodes)

    def set_shard_rw_mode(self, node: ClusterNode):
        watcher = ShardsWatcher(node)
        shards = watcher.get_shards()
        for shard in shards:
            watcher.set_shard_mode(shard["shard_id"], mode="read-write")
        watcher.await_for_all_shards_status(status="read-write")

    @staticmethod
    def get_error_count_from_logs(cluster_node: ClusterNode, object_path: str, object_name: str):
        error_count = 0
        try:
            logs = cluster_node.host.get_filtered_logs("error count", unit="frostfs-storage")
            # search error logs for current object
            for error_line in logs.split("\n"):
                if object_path in error_line and object_name in error_line:
                    result = re.findall(r'"error\scount":\s(\d+)', error_line)
                    error_count += sum(map(int, result))
        except RuntimeError as e:
            ...

        return error_count

    @staticmethod
    @wait_for_success(180, 30)
    def get_object_path_and_name_file(oid: str, cid: str, node: ClusterNode) -> tuple[str, str]:
        oid_path = f"{oid[0]}/{oid[1]}/{oid[2]}/{oid[3]}"
        object_path = None

        with reporter.step("Search object file"):
            node_shell = node.storage_node.host.get_shell()
            data_path = node.storage_node.get_data_directory()
            all_datas = node_shell.exec(f"ls -la {data_path}/data | awk '{{ print $9 }}'").stdout.strip()
            for data_dir in all_datas.replace(".", "").strip().split("\n"):
                check_dir = node_shell.exec(f" [ -d {data_path}/data/{data_dir}/data/{oid_path} ] && echo 1 || echo 0").stdout
                if "1" in check_dir:
                    object_path = f"{data_path}/data/{data_dir}/data/{oid_path}"
                    object_name = f"{oid[4:]}.{cid}"
                    break

        assert object_path is not None, f"{oid} object not found in directory - {data_path}/data"
        return object_path, object_name

    @allure.title("Metric for shard mode")
    def test_shard_metrics_set_mode(self, two_shards_and_node: tuple[str, str, ClusterNode]):
        metrics_counter = 1
        shard1, shard2, node = two_shards_and_node

        with reporter.step("Shard1 set to mode 'read-only'"):
            node_shard_set_mode(node.storage_node, shard1, "read-only")

        with reporter.step(f"Check shard metrics, 'the mode will change to 'READ_ONLY'"):
            check_metrics_counter(
                [node],
                counter_exp=metrics_counter,
                command="frostfs_node_engine_mode_info",
                mode="READ_ONLY",
                shard_id=shard1,
            )

        with reporter.step("Shard2 set to mode 'degraded-read-only'"):
            node_shard_set_mode(node.storage_node, shard2, "degraded-read-only")

        with reporter.step(f"Check shard metrics, 'the mode will change to 'DEGRADED_READ_ONLY'"):
            check_metrics_counter(
                [node],
                counter_exp=metrics_counter,
                command="frostfs_node_engine_mode_info",
                mode="DEGRADED_READ_ONLY",
                shard_id=shard2,
            )

        with reporter.step("Both shards set to mode 'read-write'"):
            for shard in [shard1, shard2]:
                node_shard_set_mode(node.storage_node, shard, "read-write")

        with reporter.step(f"Check shard metrics, 'the mode will change to 'READ_WRITE'"):
            for shard in [shard1, shard2]:
                check_metrics_counter(
                    [node],
                    counter_exp=metrics_counter,
                    command="frostfs_node_engine_mode_info",
                    mode="READ_WRITE",
                    shard_id=shard,
                )

    @allure.title("Metric for error count on shard")
    @requires_container(PUBLIC_WITH_POLICY("REP 1 CBF 1", short_name="REP 1"))
    def test_shard_metrics_error_count(
        self, max_object_size: int, default_wallet: WalletInfo, cluster: Cluster, container: str, revert_all_shards_mode
    ):
        file_path = generate_file(round(max_object_size * 0.8))

        with reporter.step("Put object"):
            oid = put_object(default_wallet, file_path, container, self.shell, cluster.default_rpc_endpoint)

        with reporter.step("Get object nodes"):
            object_storage_nodes = get_nodes_with_object(container, oid, self.shell, cluster.storage_nodes)
            object_nodes = [cluster_node for cluster_node in cluster.cluster_nodes if cluster_node.storage_node in object_storage_nodes]
            node = random.choice(object_nodes)

        with reporter.step("Search object in system."):
            object_path, object_name = self.get_object_path_and_name_file(oid, container, node)

        with reporter.step("Block read file"):
            node.host.get_shell().exec(f"chmod a-r {object_path}/{object_name}")

        with reporter.step("Get object, expect error"):
            with pytest.raises(RuntimeError, match=OBJECT_NOT_FOUND):
                get_object(
                    wallet=default_wallet,
                    cid=container,
                    oid=oid,
                    shell=self.shell,
                    endpoint=node.storage_node.get_rpc_endpoint(),
                )

        with reporter.step(f"Get shard error count from logs"):
            counter = self.get_error_count_from_logs(node, object_path, object_name)

        with reporter.step(f"Check shard error metrics"):
            check_metrics_counter([node], counter_exp=counter, command="frostfs_node_engine_errors_total")