def test_daemon_kill_during_pvc_pod_creation_and_io(
self, interface, resource_name, setup, multi_pvc_factory,
pod_factory
):
"""
Kill 'resource_name' daemon while PVCs creation, pods
creation and IO operation are progressing.
"""
num_of_new_pvcs = 5
pvc_objs, io_pods, pvc_objs_new_pods, access_modes = setup
proj_obj = pvc_objs[0].project
storageclass = pvc_objs[0].storageclass
pod_functions = {
'mds': partial(get_mds_pods), 'mon': partial(get_mon_pods),
'mgr': partial(get_mgr_pods), 'osd': partial(get_osd_pods),
'rbdplugin': partial(get_plugin_pods, interface=interface),
'cephfsplugin': partial(get_plugin_pods, interface=interface),
'cephfsplugin_provisioner': partial(get_cephfsplugin_provisioner_pods),
'rbdplugin_provisioner': partial(get_rbdfsplugin_provisioner_pods),
'operator': partial(get_operator_pods)
}
executor = ThreadPoolExecutor(max_workers=len(io_pods))
disruption = disruption_helpers.Disruptions()
disruption.set_resource(resource=resource_name)
# Get number of pods of type 'resource_name'
resource_pods_num = len(pod_functions[resource_name]())
# Do setup for running IO on pods
log.info("Setting up pods for running IO")
for pod_obj in io_pods:
if pod_obj.pvc.volume_mode == 'Block':
storage_type = 'block'
else:
storage_type = 'fs'
executor.submit(pod_obj.workload_setup, storage_type=storage_type)
# Wait for setup on pods to complete
for pod_obj in io_pods:
log.info(f"Waiting for IO setup to complete on pod {pod_obj.name}")
for sample in TimeoutSampler(
180, 2, getattr, pod_obj, 'wl_setup_done'
):
if sample:
log.info(
f"Setup for running IO is completed on pod "
f"{pod_obj.name}."
)
break
log.info("Setup for running IO is completed on pods")
# Set daemon to be killed
disruption.select_daemon()
# Start creating new pods
log.info("Start creating new pods.")
bulk_pod_create = executor.submit(
helpers.create_pods, pvc_objs_new_pods, pod_factory, interface, 2
)
# Start creation of new PVCs
log.info("Start creating new PVCs.")
bulk_pvc_create = executor.submit(
multi_pvc_factory, interface=interface,
project=proj_obj, storageclass=storageclass, size=self.pvc_size,
access_modes=access_modes,
access_modes_selection='distribute_random',
status="", num_of_pvc=num_of_new_pvcs, wait_each=False
)
# Start IO on each pod
log.info("Start IO on pods")
for pod_obj in io_pods:
if pod_obj.pvc.volume_mode == 'Block':
storage_type = 'block'
else:
storage_type = 'fs'
pod_obj.run_io(
storage_type=storage_type, size='1G', runtime=10,
fio_filename=f'{pod_obj.name}_io_file1'
)
log.info("IO started on all pods.")
# Kill daemon
disruption.kill_daemon()
# Getting result of PVC creation as list of PVC objects
pvc_objs_new = bulk_pvc_create.result()
# Confirm PVCs are Bound
for pvc_obj in pvc_objs_new:
helpers.wait_for_resource_state(
resource=pvc_obj, state=constants.STATUS_BOUND, timeout=180
)
pvc_obj.reload()
log.info("Verified: New PVCs are Bound.")
# Getting result of pods creation as list of Pod objects
pod_objs_new = bulk_pod_create.result()
# Verify new pods are Running
for pod_obj in pod_objs_new:
helpers.wait_for_resource_state(
resource=pod_obj, state=constants.STATUS_RUNNING
)
pod_obj.reload()
log.info("Verified: All new pods are Running.")
# Verify IO
log.info("Fetching IO results from IO pods.")
for pod_obj in io_pods:
fio_result = pod_obj.get_fio_results()
err_count = fio_result.get('jobs')[0].get('error')
assert err_count == 0, (
f"FIO error on pod {pod_obj.name}. FIO result: {fio_result}"
)
log.info(f"IOPs after FIO on pod {pod_obj.name}:")
log.info(
f"Read: {fio_result.get('jobs')[0].get('read').get('iops')}"
)
log.info(
f"Write: {fio_result.get('jobs')[0].get('write').get('iops')}"
)
log.info("Verified IO result on IO pods.")
all_pod_objs = io_pods + pod_objs_new
# Fetch volume details from pods for the purpose of verification
node_pv_dict = {}
for pod in all_pod_objs:
pod_info = pod.get()
node = pod_info['spec']['nodeName']
pvc = pod_info['spec']['volumes'][0]['persistentVolumeClaim']['claimName']
for pvc_obj in pvc_objs:
if pvc_obj.name == pvc:
pvc_obj.reload()
pv = pvc_obj.backed_pv
break
if node in node_pv_dict:
node_pv_dict[node].append(pv)
else:
node_pv_dict[node] = [pv]
# Delete pods
for pod_obj in all_pod_objs:
pod_obj.delete(wait=False)
# Verify pods are deleted
for pod_obj in all_pod_objs:
pod_obj.ocp.wait_for_delete(resource_name=pod_obj.name)
# Verify number of 'resource_name' type pods
final_resource_pods_num = len(pod_functions[resource_name]())
assert final_resource_pods_num == resource_pods_num, (
f"Total number of {resource_name} pods is not matching with "
f"initial value. Total number of pods before daemon kill: "
f"{resource_pods_num}. Total number of pods present now: "
f"{final_resource_pods_num}"
)
# Verify volumes are unmapped from nodes after deleting the pods
node_pv_mounted = helpers.verify_pv_mounted_on_node(node_pv_dict)
for node, pvs in node_pv_mounted.items():
assert not pvs, (
f"PVs {pvs} is still present on node {node} after "
f"deleting the pods."
)
log.info(
"Verified: mount points are removed from nodes after deleting "
"the pods"
)
# Set volume mode on PVC objects
for pvc_obj in pvc_objs_new:
pvc_info = pvc_obj.get()
setattr(pvc_obj, 'volume_mode', pvc_info['spec']['volumeMode'])
# Verify that PVCs are reusable by creating new pods
all_pvc_objs = pvc_objs + pvc_objs_new
pod_objs_re = helpers.create_pods(
all_pvc_objs, pod_factory, interface, 2
)
# Verify pods are Running
for pod_obj in pod_objs_re:
helpers.wait_for_resource_state(
resource=pod_obj, state=constants.STATUS_RUNNING
)
pod_obj.reload()
log.info("Successfully created new pods using all PVCs.")
# Run IO on each of the newly created pods
for pod_obj in pod_objs_re:
if pod_obj.pvc.volume_mode == 'Block':
storage_type = 'block'
else:
storage_type = 'fs'
pod_obj.run_io(
storage_type=storage_type, size='1G', runtime=10,
fio_filename=f'{pod_obj.name}_io_file2'
)
log.info("Fetching IO results from newly created pods")
for pod_obj in pod_objs_re:
fio_result = pod_obj.get_fio_results()
err_count = fio_result.get('jobs')[0].get('error')
assert err_count == 0, (
f"FIO error on pod {pod_obj.name}. FIO result: {fio_result}"
)
log.info(f"IOPs after FIO on pod {pod_obj.name}:")
log.info(
f"Read: {fio_result.get('jobs')[0].get('read').get('iops')}"
)
log.info(
f"Write: {fio_result.get('jobs')[0].get('write').get('iops')}"
)
log.info("Verified IO result on newly created pods.")
def test_disruptive_during_pod_pvc_deletion_and_io(
self, interface, resource_to_delete,
setup_base
):
"""
Delete ceph/rook pod while PVCs deletion, pods deletion and IO are
progressing
"""
pvc_objs, pod_objs, rwx_pod_objs = setup_base
sc_obj = pvc_objs[0].storageclass
namespace = pvc_objs[0].project.namespace
num_of_pods_to_delete = 10
num_of_io_pods = 5
# Select pods to be deleted
pods_to_delete = pod_objs[:num_of_pods_to_delete]
pods_to_delete.extend(
[pod for pod in rwx_pod_objs for pod_obj in pods_to_delete if (
pod_obj.pvc == pod.pvc
)]
)
# Select pods to run IO
io_pods = pod_objs[num_of_pods_to_delete:num_of_pods_to_delete + num_of_io_pods]
io_pods.extend(
[pod for pod in rwx_pod_objs for pod_obj in io_pods if (
pod_obj.pvc == pod.pvc
)]
)
# Select pods which are having PVCs to delete
pods_for_pvc = pod_objs[num_of_pods_to_delete + num_of_io_pods:]
pvcs_to_delete = [pod_obj.pvc for pod_obj in pods_for_pvc]
pods_for_pvc.extend(
[pod for pod in rwx_pod_objs for pod_obj in pods_for_pvc if (
pod_obj.pvc == pod.pvc
)]
)
log.info(
f"{len(pods_to_delete)} pods selected for deletion in which "
f"{len(pods_to_delete) - num_of_pods_to_delete} pairs of pod "
f"share same RWX PVC"
)
log.info(
f"{len(io_pods)} pods selected for running IO in which "
f"{len(io_pods) - num_of_io_pods} pairs of pod share same "
f"RWX PVC"
)
no_of_rwx_pvcs_delete = len(pods_for_pvc) - len(pvcs_to_delete)
log.info(
f"{len(pvcs_to_delete)} PVCs selected for deletion. "
f"RWO PVCs: {len(pvcs_to_delete) - no_of_rwx_pvcs_delete}, "
f"RWX PVCs: {no_of_rwx_pvcs_delete}"
)
pod_functions = {
'mds': get_mds_pods, 'mon': get_mon_pods, 'mgr': get_mgr_pods,
'osd': get_osd_pods
}
disruption = disruption_helpers.Disruptions()
disruption.set_resource(resource=resource_to_delete)
executor = ThreadPoolExecutor(max_workers=len(pod_objs))
# Get number of pods of type 'resource_to_delete'
num_of_resource_to_delete = len(pod_functions[resource_to_delete]())
# Fetch the number of Pods and PVCs
initial_num_of_pods = len(get_all_pods(namespace=namespace))
initial_num_of_pvc = len(
get_all_pvcs(namespace=namespace)['items']
)
# Fetch PV names to verify after deletion
pv_objs = []
for pvc_obj in pvcs_to_delete:
pvc_obj.reload()
pv_objs.append(pvc_obj.backed_pv_obj)
# Fetch volume details from pods for the purpose of verification
node_pv_dict = {}
for pod_obj in pods_to_delete:
pod_info = pod_obj.get()
node = pod_info['spec']['nodeName']
pvc = pod_info['spec']['volumes'][0]['persistentVolumeClaim']['claimName']
for pvc_obj in pvc_objs:
if pvc_obj.name == pvc:
pvc_obj.reload()
pv = pvc_obj.backed_pv
break
if node in node_pv_dict:
node_pv_dict[node].append(pv)
else:
node_pv_dict[node] = [pv]
# Fetch image uuid associated with PVCs to be deleted
pvc_uuid_map = {}
for pvc_obj in pvcs_to_delete:
pvc_uuid_map[pvc_obj.name] = pvc_obj.image_uuid
log.info("Fetched image uuid associated with each PVC")
# Do setup on pods for running IO
log.info("Setting up pods for running IO.")
for pod_obj in pod_objs + rwx_pod_objs:
executor.submit(pod_obj.workload_setup, storage_type='fs')
# Wait for setup on pods to complete
for pod_obj in pod_objs + rwx_pod_objs:
for sample in TimeoutSampler(
180, 2, getattr, pod_obj, 'wl_setup_done'
):
if sample:
log.info(
f"Setup for running IO is completed on pod "
f"{pod_obj.name}."
)
break
log.info("Setup for running IO is completed on all pods.")
# Start IO on pods having PVCs to delete to load data
log.info("Starting IO on pods having PVCs to delete.")
self.run_io_on_pods(pods_for_pvc)
log.info("IO started on pods having PVCs to delete.")
log.info("Fetching IO results from the pods having PVCs to delete.")
for pod_obj in pods_for_pvc:
get_fio_rw_iops(pod_obj)
log.info("Verified IO result on pods having PVCs to delete.")
# Delete pods having PVCs to delete.
assert self.delete_pods(pods_for_pvc), (
"Couldn't delete pods which are having PVCs to delete."
)
for pod_obj in pods_for_pvc:
pod_obj.ocp.wait_for_delete(pod_obj.name)
log.info("Verified: Deleted pods which are having PVCs to delete.")
# Start IO on pods to be deleted
log.info("Starting IO on pods to be deleted.")
self.run_io_on_pods(pods_to_delete)
log.info("IO started on pods to be deleted.")
# Start deleting PVCs
pvc_bulk_delete = executor.submit(delete_pvcs, pvcs_to_delete)
log.info("Started deleting PVCs")
# Start deleting pods
pod_bulk_delete = executor.submit(self.delete_pods, pods_to_delete)
log.info("Started deleting pods")
# Start IO on IO pods
self.run_io_on_pods(io_pods)
log.info("Started IO on IO pods")
# Verify pvc deletion has started
pvc_deleting = executor.submit(
wait_for_resource_count_change, func_to_use=get_all_pvcs,
previous_num=initial_num_of_pvc, namespace=namespace,
change_type='decrease', min_difference=1, timeout=30, interval=0.01
)
# Verify pod deletion has started
pod_deleting = executor.submit(
wait_for_resource_count_change, func_to_use=get_all_pods,
previous_num=initial_num_of_pods, namespace=namespace,
change_type='decrease', min_difference=1, timeout=30, interval=0.01
)
assert pvc_deleting.result(), (
"Wait timeout: PVCs are not being deleted."
)
log.info("PVCs deletion has started.")
assert pod_deleting.result(), (
"Wait timeout: Pods are not being deleted."
)
log.info("Pods deletion has started.")
# Delete pod of type 'resource_to_delete'
disruption.delete_resource()
pods_deleted = pod_bulk_delete.result()
assert pods_deleted, "Deletion of pods failed."
# Verify pods are deleted
for pod_obj in pods_to_delete:
pod_obj.ocp.wait_for_delete(pod_obj.name)
log.info("Verified: Pods are deleted.")
# Verify that the mount point is removed from nodes after deleting pod
node_pv_mounted = verify_pv_mounted_on_node(node_pv_dict)
for node, pvs in node_pv_mounted.items():
assert not pvs, (
f"PVs {pvs} is still present on node {node} after "
f"deleting the pods."
)
log.info(
"Verified: mount points are removed from nodes after deleting "
"the pods"
)
pvcs_deleted = pvc_bulk_delete.result()
assert pvcs_deleted, "Deletion of PVCs failed."
# Verify PVCs are deleted
for pvc_obj in pvcs_to_delete:
pvc_obj.ocp.wait_for_delete(pvc_obj.name)
logging.info("Verified: PVCs are deleted.")
# Verify PVs are deleted
for pv_obj in pv_objs:
pv_obj.ocp.wait_for_delete(resource_name=pv_obj.name, timeout=300)
logging.info("Verified: PVs are deleted.")
# Verify PV using ceph toolbox. Image/Subvolume should be deleted.
for pvc_name, uuid in pvc_uuid_map.items():
if interface == constants.CEPHBLOCKPOOL:
ret = verify_volume_deleted_in_backend(
interface=interface, image_uuid=uuid,
pool_name=sc_obj.ceph_pool.name
)
if interface == constants.CEPHFILESYSTEM:
ret = verify_volume_deleted_in_backend(
interface=interface, image_uuid=uuid
)
assert ret, (
f"Volume associated with PVC {pvc_name} still exists "
f"in backend"
)
log.info("Fetching IO results from the pods.")
for pod_obj in io_pods:
fio_result = pod_obj.get_fio_results()
err_count = fio_result.get('jobs')[0].get('error')
assert err_count == 0, (
f"FIO error on pod {pod_obj.name}. FIO result: {fio_result}"
)
log.info("Verified IO result on pods.")
# Verify number of pods of type 'resource_to_delete'
final_num_resource_to_delete = len(pod_functions[resource_to_delete]())
assert final_num_resource_to_delete == num_of_resource_to_delete, (
f"Total number of {resource_to_delete} pods is not matching with "
f"initial value. Total number of pods before deleting a pod: "
f"{num_of_resource_to_delete}. Total number of pods present now: "
f"{final_num_resource_to_delete}"
)
# Check ceph status
ceph_health_check(namespace=config.ENV_DATA['cluster_namespace'])
log.info("Ceph cluster health is OK")
