54 lines
1.8 KiB
Python
54 lines
1.8 KiB
Python
import logging
|
|
from time import sleep
|
|
|
|
import allure
|
|
from cluster import Cluster, StorageNode
|
|
from neofs_testlib.shell import Shell
|
|
from python_keywords.node_management import storage_node_healthcheck
|
|
from storage_policy import get_nodes_with_object
|
|
|
|
logger = logging.getLogger("NeoLogger")
|
|
|
|
|
|
@allure.step("Wait for object replication")
|
|
def wait_object_replication(
|
|
cid: str,
|
|
oid: str,
|
|
expected_copies: int,
|
|
shell: Shell,
|
|
nodes: list[StorageNode],
|
|
) -> list[StorageNode]:
|
|
sleep_interval, attempts = 15, 20
|
|
nodes_with_object = []
|
|
for _ in range(attempts):
|
|
nodes_with_object = get_nodes_with_object(cid, oid, shell=shell, nodes=nodes)
|
|
if len(nodes_with_object) >= expected_copies:
|
|
return nodes_with_object
|
|
sleep(sleep_interval)
|
|
raise AssertionError(
|
|
f"Expected {expected_copies} copies of object, but found {len(nodes_with_object)}. "
|
|
f"Waiting time {sleep_interval * attempts}"
|
|
)
|
|
|
|
|
|
@allure.step("Wait for storage nodes returned to cluster")
|
|
def wait_all_storage_nodes_returned(cluster: Cluster) -> None:
|
|
sleep_interval, attempts = 15, 20
|
|
for __attempt in range(attempts):
|
|
if is_all_storage_nodes_returned(cluster):
|
|
return
|
|
sleep(sleep_interval)
|
|
raise AssertionError("Storage node(s) is broken")
|
|
|
|
|
|
def is_all_storage_nodes_returned(cluster: Cluster) -> bool:
|
|
with allure.step("Run health check for all storage nodes"):
|
|
for node in cluster.storage_nodes:
|
|
try:
|
|
health_check = storage_node_healthcheck(node)
|
|
except Exception as err:
|
|
logger.warning(f"Node healthcheck fails with error {err}")
|
|
return False
|
|
if health_check.health_status != "READY" or health_check.network_status != "ONLINE":
|
|
return False
|
|
return True
|