frostfs-testcases/pytest_tests/testsuites/failovers/test_frostfs_failover_network.py
Dmitriy Zayakin c8b95d98f4
Some checks reported warnings
DCO check / Commits Check (pull_request) Has been cancelled
[#291] Change error message for network test
Signed-off-by: Dmitriy Zayakin <d.zayakin@yadro.com>
2024-08-16 12:07:50 +03:00

438 lines
19 KiB
Python

import logging
import random
from time import sleep
import allure
import pytest
from frostfs_testlib import reporter
from frostfs_testlib.healthcheck.interfaces import Healthcheck
from frostfs_testlib.resources.wellknown_acl import EACL_PUBLIC_READ_WRITE, PUBLIC_ACL
from frostfs_testlib.steps.cli.container import create_container
from frostfs_testlib.steps.cli.object import get_object, get_object_nodes, neo_go_query_height, put_object, put_object_to_random_node
from frostfs_testlib.steps.storage_object import delete_objects
from frostfs_testlib.storage.cluster import ClusterNode
from frostfs_testlib.storage.controllers import ClusterStateController
from frostfs_testlib.storage.dataclasses.object_size import ObjectSize
from frostfs_testlib.storage.dataclasses.storage_object_info import Interfaces, StorageObjectInfo
from frostfs_testlib.storage.dataclasses.wallet import WalletInfo
from frostfs_testlib.testing.cluster_test_base import ClusterTestBase
from frostfs_testlib.testing.parallel import parallel
from frostfs_testlib.utils.failover_utils import wait_object_replication
from frostfs_testlib.utils.file_utils import generate_file, get_file_hash
logger = logging.getLogger("NeoLogger")
STORAGE_NODE_COMMUNICATION_PORT = "8080"
STORAGE_NODE_COMMUNICATION_PORT_TLS = "8082"
PORTS_TO_BLOCK = [STORAGE_NODE_COMMUNICATION_PORT, STORAGE_NODE_COMMUNICATION_PORT_TLS]
blocked_nodes: list[ClusterNode] = []
OBJECT_ATTRIBUTES = [
None,
{"key1": 1, "key2": "abc", "common_key": "common_value"},
{"key1": 2, "common_key": "common_value"},
]
@pytest.mark.failover
@pytest.mark.failover_network
class TestFailoverNetwork(ClusterTestBase):
@pytest.fixture(autouse=True)
@allure.title("Restore network")
def restore_network(self, healthcheck: Healthcheck, cluster_state_controller: ClusterStateController):
yield
with reporter.step(f"Count blocked nodes {len(blocked_nodes)}"):
not_empty = len(blocked_nodes) != 0
for node in list(blocked_nodes):
with reporter.step(f"Restore network for {node}"):
cluster_state_controller.restore_traffic(node=node)
blocked_nodes.remove(node)
if not_empty:
parallel(healthcheck.storage_healthcheck, self.cluster.cluster_nodes)
@pytest.fixture()
@allure.title("Restore drop traffic to system")
def restore_down_interfaces(self, cluster_state_controller: ClusterStateController):
yield
cluster_state_controller.restore_interfaces()
@pytest.fixture()
def storage_objects(
self,
simple_object_size: ObjectSize,
default_wallet: WalletInfo,
) -> list[StorageObjectInfo]:
file_path = generate_file(simple_object_size.value)
file_hash = get_file_hash(file_path)
with reporter.step("Create container"):
placement_rule = "REP 1 CBF 1"
cid = create_container(
wallet=default_wallet,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
rule=placement_rule,
await_mode=True,
basic_acl=EACL_PUBLIC_READ_WRITE,
)
storage_objects = []
with reporter.step("Put object"):
for attribute in OBJECT_ATTRIBUTES:
oid = put_object_to_random_node(
wallet=default_wallet,
path=file_path,
cid=cid,
shell=self.shell,
cluster=self.cluster,
)
storage_object = StorageObjectInfo(cid=cid, oid=oid)
storage_object.size = simple_object_size.value
storage_object.wallet = default_wallet
storage_object.file_path = file_path
storage_object.file_hash = file_hash
storage_object.attributes = attribute
storage_objects.append(storage_object)
return storage_objects
@allure.title("Block Storage node traffic")
def test_block_storage_node_traffic(
self,
default_wallet: WalletInfo,
require_multiple_hosts,
simple_object_size: ObjectSize,
cluster_state_controller: ClusterStateController,
):
"""
Block storage nodes traffic using iptables and wait for replication for objects.
"""
wallet = default_wallet
placement_rule = "REP 2 IN X CBF 2 SELECT 2 FROM * AS X"
wakeup_node_timeout = 10 # timeout to let nodes detect that traffic has blocked
nodes_to_block_count = 2
source_file_path = generate_file(simple_object_size.value)
cid = create_container(
wallet,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
rule=placement_rule,
basic_acl=PUBLIC_ACL,
)
oid = put_object_to_random_node(wallet, source_file_path, cid, shell=self.shell, cluster=self.cluster)
nodes = wait_object_replication(cid, oid, 2, shell=self.shell, nodes=self.cluster.storage_nodes)
logger.info(f"Nodes are {nodes}")
nodes_to_block = nodes
if nodes_to_block_count > len(nodes):
# TODO: the intent of this logic is not clear, need to revisit
nodes_to_block = random.choices(nodes, k=2)
nodes_non_block = list(set(self.cluster.storage_nodes) - set(nodes_to_block))
nodes_non_block_cluster = [
cluster_node for cluster_node in self.cluster.cluster_nodes if cluster_node.storage_node in nodes_non_block
]
with reporter.step("Block traffic and check corrupted object"):
for node in nodes_non_block_cluster:
with reporter.step(f"Block incoming traffic at node {node}"):
blocked_nodes.append(node)
cluster_state_controller.drop_traffic(
node=node, wakeup_timeout=wakeup_node_timeout, name_interface="data", block_nodes=nodes_to_block
)
with reporter.step(f"Check object is not stored on node {node}"):
new_nodes = wait_object_replication(
cid,
oid,
2,
shell=self.shell,
nodes=list(set(self.cluster.storage_nodes) - set(nodes_non_block)),
)
assert node.storage_node not in new_nodes
with reporter.step("Check object data is not corrupted"):
got_file_path = get_object(wallet, cid, oid, endpoint=new_nodes[0].get_rpc_endpoint(), shell=self.shell)
assert get_file_hash(source_file_path) == get_file_hash(got_file_path)
with reporter.step(f"Unblock incoming traffic"):
for node in nodes_non_block_cluster:
with reporter.step(f"Unblock at host {node}"):
cluster_state_controller.restore_traffic(node=node)
block_node = [
cluster_node for cluster_node in self.cluster.cluster_nodes if cluster_node.storage_node == node.storage_node
]
blocked_nodes.remove(*block_node)
sleep(wakeup_node_timeout)
with reporter.step("Check object data is not corrupted"):
new_nodes = wait_object_replication(cid, oid, 2, shell=self.shell, nodes=self.cluster.storage_nodes)
got_file_path = get_object(wallet, cid, oid, shell=self.shell, endpoint=new_nodes[0].get_rpc_endpoint())
assert get_file_hash(source_file_path) == get_file_hash(got_file_path)
@pytest.mark.interfaces
@allure.title("Block DATA interface node")
def test_block_data_interface(
self,
cluster_state_controller: ClusterStateController,
default_wallet: WalletInfo,
restore_down_interfaces: None,
storage_objects: list[StorageObjectInfo],
):
storage_object = storage_objects[0]
with reporter.step("Search nodes with object"):
nodes_with_object = get_object_nodes(
cluster=self.cluster,
cid=storage_object.cid,
oid=storage_object.oid,
alive_node=self.cluster.cluster_nodes[0],
)
with reporter.step("Get data interface to node"):
config_interfaces = list(nodes_with_object[0].host.config.interfaces.keys())
with reporter.step(f"Get data in {config_interfaces}"):
data_interfaces = [interface for interface in config_interfaces if "data" in interface]
with reporter.step("Block data interfaces for node"):
for interface in data_interfaces:
cluster_state_controller.down_interface(nodes=nodes_with_object, interface=interface)
with reporter.step("Tick epoch and wait 2 block"):
nodes_without_an_object = list(set(self.cluster.cluster_nodes) - set(nodes_with_object))
self.tick_epochs(1, alive_node=nodes_without_an_object[0].storage_node, wait_block=2)
with reporter.step("Get object for target nodes to data interfaces, expect false"):
with pytest.raises(RuntimeError, match="can't create API client: can't init SDK client: gRPC dial: context deadline exceeded"):
get_object(
wallet=default_wallet,
cid=storage_object.cid,
oid=storage_object.oid,
shell=self.shell,
endpoint=nodes_with_object[0].storage_node.get_rpc_endpoint(),
)
with reporter.step(f"Get object others nodes, expect true"):
input_file = get_object(
wallet=default_wallet,
cid=storage_object.cid,
oid=storage_object.oid,
shell=self.shell,
endpoint=nodes_without_an_object[0].storage_node.get_rpc_endpoint(),
)
with reporter.step("Restore interface and tick 1 epoch, wait 2 block"):
cluster_state_controller.restore_interfaces()
self.tick_epochs(1, alive_node=nodes_without_an_object[0].storage_node, wait_block=2)
@pytest.mark.interfaces
@allure.title("Block INTERNAL interface node")
def test_block_internal_interface(
self,
cluster_state_controller: ClusterStateController,
default_wallet: WalletInfo,
restore_down_interfaces: None,
storage_objects: list[StorageObjectInfo],
simple_object_size: ObjectSize,
):
storage_object = storage_objects[0]
with reporter.step("Search nodes with object"):
nodes_with_object = get_object_nodes(
cluster=self.cluster,
cid=storage_object.cid,
oid=storage_object.oid,
alive_node=self.cluster.cluster_nodes[0],
)
with reporter.step("Get internal interface to node"):
config_interfaces = list(nodes_with_object[0].host.config.interfaces.keys())
with reporter.step(f"Get internal in {config_interfaces}"):
internal_interfaces = [interface for interface in config_interfaces if "internal" in interface]
with reporter.step("Block internal interfaces for node"):
for interface in internal_interfaces:
cluster_state_controller.down_interface(nodes=nodes_with_object, interface=interface)
with reporter.step("Tick epoch and wait 2 block"):
nodes_without_an_object = list(set(self.cluster.cluster_nodes) - set(nodes_with_object))
self.tick_epochs(1, alive_node=nodes_without_an_object[0].storage_node, wait_block=2)
with reporter.step("Get object others node, expect false"):
with pytest.raises(RuntimeError, match="rpc error"):
get_object(
wallet=default_wallet,
cid=storage_object.cid,
oid=storage_object.oid,
shell=self.shell,
endpoint=nodes_without_an_object[0].storage_node.get_rpc_endpoint(),
)
with reporter.step("Put object, others node, expect false"):
with pytest.raises(RuntimeError, match="rpc error"):
put_object(
wallet=default_wallet,
path=storage_object.file_path,
cid=storage_object.cid,
shell=self.shell,
endpoint=nodes_without_an_object[0].storage_node.get_rpc_endpoint(),
)
with reporter.step(f"Get object nodes with object, expect true"):
input_file = get_object(
wallet=default_wallet,
cid=storage_object.cid,
oid=storage_object.oid,
shell=self.shell,
endpoint=nodes_with_object[0].storage_node.get_rpc_endpoint(),
)
with reporter.step(f"Put object nodes with object, expect true"):
temp_file_path = generate_file(simple_object_size.value)
_ = put_object(
wallet=default_wallet,
path=temp_file_path,
cid=storage_object.cid,
shell=self.shell,
endpoint=nodes_with_object[0].storage_node.get_rpc_endpoint(),
)
with reporter.step("Restore interface and tick 1 epoch, wait 2 block"):
cluster_state_controller.restore_interfaces()
self.tick_epochs(1, alive_node=nodes_without_an_object[0].storage_node, wait_block=2)
@pytest.mark.interfaces
@pytest.mark.failover_baremetal
@pytest.mark.parametrize(
"block_interface, other_interface",
[(Interfaces.DATA_O, Interfaces.DATA_1), (Interfaces.DATA_1, Interfaces.DATA_O)],
)
@allure.title("Down data interfaces to all nodes(interface={block_interface})")
def test_down_data_interface(
self,
require_multiple_interfaces,
cluster_state_controller: ClusterStateController,
default_wallet: WalletInfo,
simple_object_size: ObjectSize,
restore_down_interfaces: None,
block_interface: Interfaces,
other_interface: Interfaces,
):
cluster_nodes = self.cluster.cluster_nodes
with reporter.step(f"Block {block_interface.value} interfaces"):
cluster_state_controller.down_interface(cluster_nodes, block_interface.value)
with reporter.step("Tick 1 epoch and wait 2 block for sync all nodes"):
self.tick_epochs(1, alive_node=cluster_nodes[0].storage_node, wait_block=2)
with reporter.step("Create container"):
cid = create_container(
wallet=default_wallet,
shell=self.shell,
endpoint=f"{cluster_nodes[0].get_data_interface(other_interface.value)[0]}:8080",
rule="REP 4 CBF 1",
)
with reporter.step("Put object"):
file_path = generate_file(simple_object_size.value)
oid = put_object(
wallet=default_wallet,
path=file_path,
cid=cid,
shell=self.shell,
endpoint=f"{cluster_nodes[0].get_data_interface(other_interface.value)[0]}:8080",
)
with reporter.step("Get object"):
file_get_path = get_object(
wallet=default_wallet,
cid=cid,
oid=oid,
shell=self.shell,
endpoint=f"{cluster_nodes[0].get_data_interface(other_interface.value)[0]}:8080",
)
with reporter.step("Restore interfaces all nodes"):
cluster_state_controller.restore_interfaces()
self.tick_epochs(1, alive_node=cluster_nodes[0].storage_node, wait_block=2)
@pytest.mark.interfaces
@pytest.mark.failover_baremetal
@pytest.mark.parametrize("interface", [Interfaces.INTERNAL_0, Interfaces.INTERNAL_1])
@allure.title("Down internal interfaces to all nodes(interface={interface})")
def test_down_internal_interface(
self,
require_multiple_interfaces,
cluster_state_controller: ClusterStateController,
default_wallet: WalletInfo,
simple_object_size: ObjectSize,
restore_down_interfaces: None,
interface: Interfaces,
):
cluster_nodes = self.cluster.cluster_nodes
latest_block = {}
with reporter.step("Get block all nodes"):
for cluster_node in cluster_nodes:
latest_block[cluster_node] = neo_go_query_height(
shell=cluster_node.host.get_shell(), endpoint=cluster_node.morph_chain.get_http_endpoint()
)
with reporter.step(f"Block {interface} interfaces"):
cluster_state_controller.down_interface(cluster_nodes, interface.value)
with reporter.step("Tick 1 epoch and wait 2 block for sync all nodes"):
self.tick_epochs(1, alive_node=cluster_nodes[0].storage_node, wait_block=2)
with reporter.step("Create container"):
cid = create_container(
wallet=default_wallet,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
rule="REP 4 CBF 1",
)
with reporter.step(f"Put object, after down {interface}"):
file_path = generate_file(simple_object_size.value)
oid = put_object(
wallet=default_wallet,
path=file_path,
cid=cid,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
)
with reporter.step("Get object"):
file_get_path = get_object(
wallet=default_wallet,
cid=cid,
oid=oid,
shell=self.shell,
endpoint=self.cluster.default_rpc_endpoint,
)
now_block = {}
with reporter.step("Get actual block"):
for cluster_node in cluster_nodes:
now_block[cluster_node] = neo_go_query_height(
shell=cluster_node.host.get_shell(), endpoint=cluster_node.morph_chain.get_http_endpoint()
)
with reporter.step(f"Compare block"):
for cluster_node, items in now_block.items():
with reporter.step(
f"Node - {cluster_node.host_ip}, old block - {latest_block[cluster_node]['Latest block']}, "
f"now block - {now_block[cluster_node]['Latest block']}"
):
assert latest_block[cluster_node]["Latest block"] < now_block[cluster_node]["Latest block"]
with reporter.step("Restore interfaces all nodes"):
cluster_state_controller.restore_interfaces()
self.tick_epochs(1, alive_node=self.cluster.cluster_nodes[0].storage_node, wait_block=2)