def test_restart_gluster_block_provisioner_pod(self):
"""Restart gluster-block provisioner pod
"""
# create heketi block volume
vol_info = heketi_blockvolume_create(self.heketi_client_node,
self.heketi_server_url,
size=5,
json=True)
self.assertTrue(vol_info, "Failed to create heketi block"
"volume of size 5")
self.addCleanup(heketi_blockvolume_delete, self.heketi_client_node,
self.heketi_server_url, vol_info['id'])
# restart gluster-block-provisioner-pod
dc_name = "glusterblock-%s-provisioner-dc" % self.storage_project_name
pod_name = get_pod_name_from_dc(self.ocp_master_node, dc_name)
oc_delete(self.ocp_master_node, 'pod', pod_name)
wait_for_resource_absence(self.ocp_master_node, 'pod', pod_name)
# new gluster-pod name
pod_name = get_pod_name_from_dc(self.ocp_master_node, dc_name)
wait_for_pod_be_ready(self.ocp_master_node, pod_name)
# create new heketi block volume
vol_info = heketi_blockvolume_create(self.heketi_client_node,
self.heketi_server_url,
size=2,
json=True)
self.assertTrue(vol_info, "Failed to create heketi block"
"volume of size 2")
heketi_blockvolume_delete(self.heketi_client_node,
self.heketi_server_url, vol_info['id'])
def test_pvc_deletion_while_pod_is_running(self):
"""Validate PVC deletion while pod is running"""
# Create DC with POD and attached PVC to it
sc_name = self.create_storage_class()
pvc_name = self.create_and_wait_for_pvc(sc_name=sc_name)
dc_name, pod_name = self.create_dc_with_pvc(pvc_name)
# Delete PVC
oc_delete(self.node, 'pvc', self.pvc_name)
with self.assertRaises(ExecutionError):
wait_for_resource_absence(self.node,
'pvc',
self.pvc_name,
interval=3,
timeout=30)
# Make sure we are able to work with files on the mounted volume
# after deleting pvc.
filepath = "/mnt/file_for_testing_volume.log"
cmd = "dd if=/dev/urandom of=%s bs=1K count=100" % filepath
ret, out, err = oc_rsh(self.node, pod_name, cmd)
self.assertEqual(
ret, 0, "Failed to execute command %s on %s" % (cmd, self.node))
def test_dynamic_provisioning_glusterfile_glusterpod_failure(self):
"""Validate dynamic provisioning for gluster file when gluster pod down
"""
mount_path = "/mnt"
datafile_path = '%s/fake_file_for_%s' % (mount_path, self.id())
# Create secret and storage class
self.create_storage_class()
# Create PVC
pvc_name = self.create_and_wait_for_pvc()
# Create app POD with attached volume
pod_name = oc_create_tiny_pod_with_volume(
self.node, pvc_name, "test-pvc-mount-on-app-pod",
mount_path=mount_path)
self.addCleanup(
wait_for_resource_absence, self.node, 'pod', pod_name)
self.addCleanup(oc_delete, self.node, 'pod', pod_name)
# Wait for app POD be up and running
wait_for_pod_be_ready(
self.node, pod_name, timeout=60, wait_step=2)
# Run IO in background
io_cmd = "oc rsh %s dd if=/dev/urandom of=%s bs=1000K count=900" % (
pod_name, datafile_path)
async_io = g.run_async(self.node, io_cmd, "root")
# Pick up one of the hosts which stores PV brick (4+ nodes case)
gluster_pod_data = get_gluster_pod_names_by_pvc_name(
self.node, pvc_name)[0]
# Delete glusterfs POD from chosen host and wait for spawn of new one
oc_delete(self.node, 'pod', gluster_pod_data["pod_name"])
cmd = ("oc get pods -o wide | grep glusterfs | grep %s | "
"grep -v Terminating | awk '{print $1}'") % (
gluster_pod_data["host_name"])
for w in Waiter(600, 30):
out = self.cmd_run(cmd)
new_gluster_pod_name = out.strip().split("\n")[0].strip()
if not new_gluster_pod_name:
continue
else:
break
if w.expired:
error_msg = "exceeded timeout, new gluster pod not created"
g.log.error(error_msg)
raise ExecutionError(error_msg)
new_gluster_pod_name = out.strip().split("\n")[0].strip()
g.log.info("new gluster pod name is %s" % new_gluster_pod_name)
wait_for_pod_be_ready(self.node, new_gluster_pod_name)
# Check that async IO was not interrupted
ret, out, err = async_io.async_communicate()
self.assertEqual(ret, 0, "IO %s failed on %s" % (io_cmd, self.node))
def test_restart_heketi_pod(self):
"""Validate restarting heketi pod"""
# create heketi volume
vol_info = heketi_volume_create(self.heketi_client_node,
self.heketi_server_url,
size=1, json=True)
self.assertTrue(vol_info, "Failed to create heketi volume of size 1")
self.addCleanup(
heketi_volume_delete, self.heketi_client_node,
self.heketi_server_url, vol_info['id'], raise_on_error=False)
topo_info = heketi_topology_info(self.heketi_client_node,
self.heketi_server_url,
json=True)
# get heketi-pod name
heketi_pod_name = get_pod_name_from_dc(self.ocp_master_node[0],
self.heketi_dc_name)
# delete heketi-pod (it restarts the pod)
oc_delete(self.ocp_master_node[0], 'pod', heketi_pod_name)
wait_for_resource_absence(self.ocp_master_node[0],
'pod', heketi_pod_name)
# get new heketi-pod name
heketi_pod_name = get_pod_name_from_dc(self.ocp_master_node[0],
self.heketi_dc_name)
wait_for_pod_be_ready(self.ocp_master_node[0],
heketi_pod_name)
# check heketi server is running
self.assertTrue(
hello_heketi(self.heketi_client_node, self.heketi_server_url),
"Heketi server %s is not alive" % self.heketi_server_url
)
# compare the topology
new_topo_info = heketi_topology_info(self.heketi_client_node,
self.heketi_server_url,
json=True)
self.assertEqual(new_topo_info, topo_info, "topology info is not same,"
" difference - %s" % diff(topo_info, new_topo_info))
# create new volume
vol_info = heketi_volume_create(self.heketi_client_node,
self.heketi_server_url,
size=2, json=True)
self.assertTrue(vol_info, "Failed to create heketi volume of size 20")
heketi_volume_delete(
self.heketi_client_node, self.heketi_server_url, vol_info['id'])
def test_pv_resize_with_prefix_for_name(self,
create_vol_name_prefix=False):
"""Validate PV resize with and without name prefix"""
dir_path = "/mnt/"
node = self.ocp_client[0]
# Create PVC
self.create_storage_class(
allow_volume_expansion=True,
create_vol_name_prefix=create_vol_name_prefix)
pvc_name = self.create_and_wait_for_pvc()
# Create DC with POD and attached PVC to it.
dc_name = oc_create_app_dc_with_io(node, pvc_name)
self.addCleanup(oc_delete, node, 'dc', dc_name)
self.addCleanup(scale_dc_pod_amount_and_wait, node, dc_name, 0)
pod_name = get_pod_name_from_dc(node, dc_name)
wait_for_pod_be_ready(node, pod_name)
if create_vol_name_prefix:
ret = heketi_ops.verify_volume_name_prefix(
node, self.sc['volumenameprefix'], self.sc['secretnamespace'],
pvc_name, self.heketi_server_url)
self.assertTrue(ret, "verify volnameprefix failed")
cmd = ("dd if=/dev/urandom of=%sfile " "bs=100K count=1000") % dir_path
ret, out, err = oc_rsh(node, pod_name, cmd)
self.assertEqual(ret, 0,
"failed to execute command %s on %s" % (cmd, node))
cmd = ("dd if=/dev/urandom of=%sfile2 "
"bs=100K count=10000") % dir_path
ret, out, err = oc_rsh(node, pod_name, cmd)
self.assertNotEqual(
ret, 0, " This IO did not fail as expected "
"command %s on %s" % (cmd, node))
pvc_size = 2
resize_pvc(node, pvc_name, pvc_size)
verify_pvc_size(node, pvc_name, pvc_size)
pv_name = get_pv_name_from_pvc(node, pvc_name)
verify_pv_size(node, pv_name, pvc_size)
oc_delete(node, 'pod', pod_name)
wait_for_resource_absence(node, 'pod', pod_name)
pod_name = get_pod_name_from_dc(node, dc_name)
wait_for_pod_be_ready(node, pod_name)
cmd = ("dd if=/dev/urandom of=%sfile_new "
"bs=50K count=10000") % dir_path
ret, out, err = oc_rsh(node, pod_name, cmd)
self.assertEqual(ret, 0,
"failed to execute command %s on %s" % (cmd, node))
def test_dynamic_provisioning_glusterfile_heketidown_pvc_delete(self):
"""Validate deletion of PVC's when heketi is down"""
# Create storage class, secret and PVCs
self.create_storage_class()
self.pvc_name_list = self.create_and_wait_for_pvcs(
1, 'pvc-heketi-down', 3)
# remove heketi-pod
scale_dc_pod_amount_and_wait(self.ocp_client[0], self.heketi_dc_name,
0, self.storage_project_name)
try:
# delete pvc
for pvc in self.pvc_name_list:
oc_delete(self.ocp_client[0], 'pvc', pvc)
for pvc in self.pvc_name_list:
with self.assertRaises(ExecutionError):
wait_for_resource_absence(self.ocp_client[0],
'pvc',
pvc,
interval=3,
timeout=30)
finally:
# bring back heketi-pod
scale_dc_pod_amount_and_wait(self.ocp_client[0],
self.heketi_dc_name, 1,
self.storage_project_name)
# verify PVC's are deleted
for pvc in self.pvc_name_list:
wait_for_resource_absence(self.ocp_client[0],
'pvc',
pvc,
interval=1,
timeout=120)
# create a new PVC
self.create_and_wait_for_pvc()
def test_dynamic_provisioning_glusterfile_reclaim_policy_retain(self):
"""Validate retain policy for glusterfs after deletion of pvc"""
self.create_storage_class(reclaim_policy='Retain')
self.create_and_wait_for_pvc()
# get the name of the volume
pv_name = get_pv_name_from_pvc(self.node, self.pvc_name)
custom = [
r':.metadata.annotations.'
r'"gluster\.kubernetes\.io\/heketi\-volume\-id"',
r':.spec.persistentVolumeReclaimPolicy'
]
vol_id, reclaim_policy = oc_get_custom_resource(
self.node, 'pv', custom, pv_name)
self.assertEqual(reclaim_policy, 'Retain')
# Create DC with POD and attached PVC to it.
try:
dc_name = oc_create_app_dc_with_io(self.node, self.pvc_name)
pod_name = get_pod_name_from_dc(self.node, dc_name)
wait_for_pod_be_ready(self.node, pod_name)
finally:
scale_dc_pod_amount_and_wait(self.node, dc_name, 0)
oc_delete(self.node, 'dc', dc_name)
wait_for_resource_absence(self.node, 'pod', pod_name)
oc_delete(self.node, 'pvc', self.pvc_name)
with self.assertRaises(ExecutionError):
wait_for_resource_absence(self.node,
'pvc',
self.pvc_name,
interval=3,
timeout=30)
heketi_volume_delete(self.heketi_client_node, self.heketi_server_url,
vol_id)
vol_list = heketi_volume_list(self.heketi_client_node,
self.heketi_server_url)
self.assertNotIn(vol_id, vol_list)
oc_delete(self.node, 'pv', pv_name)
wait_for_resource_absence(self.node, 'pv', pv_name)
def _pv_resize(self, exceed_free_space):
dir_path = "/mnt"
pvc_size_gb, min_free_space_gb = 1, 3
# Get available free space disabling redundant devices and nodes
heketi_url = self.heketi_server_url
node_id_list = heketi_ops.heketi_node_list(self.heketi_client_node,
heketi_url)
self.assertTrue(node_id_list)
nodes = {}
min_free_space = min_free_space_gb * 1024**2
for node_id in node_id_list:
node_info = heketi_ops.heketi_node_info(self.heketi_client_node,
heketi_url,
node_id,
json=True)
if (node_info['state'].lower() != 'online'
or not node_info['devices']):
continue
if len(nodes) > 2:
out = heketi_ops.heketi_node_disable(self.heketi_client_node,
heketi_url, node_id)
self.assertTrue(out)
self.addCleanup(heketi_ops.heketi_node_enable,
self.heketi_client_node, heketi_url, node_id)
for device in node_info['devices']:
if device['state'].lower() != 'online':
continue
free_space = device['storage']['free']
if (node_id in nodes.keys() or free_space < min_free_space):
out = heketi_ops.heketi_device_disable(
self.heketi_client_node, heketi_url, device['id'])
self.assertTrue(out)
self.addCleanup(heketi_ops.heketi_device_enable,
self.heketi_client_node, heketi_url,
device['id'])
continue
nodes[node_id] = free_space
if len(nodes) < 3:
raise self.skipTest("Could not find 3 online nodes with, "
"at least, 1 online device having free space "
"bigger than %dGb." % min_free_space_gb)
# Calculate maximum available size for PVC
available_size_gb = int(min(nodes.values()) / (1024**2))
# Create PVC
self.create_storage_class(allow_volume_expansion=True)
pvc_name = self.create_and_wait_for_pvc(pvc_size=pvc_size_gb)
# Create DC with POD and attached PVC to it
dc_name = oc_create_app_dc_with_io(self.node, pvc_name)
self.addCleanup(oc_delete, self.node, 'dc', dc_name)
self.addCleanup(scale_dc_pod_amount_and_wait, self.node, dc_name, 0)
pod_name = get_pod_name_from_dc(self.node, dc_name)
wait_for_pod_be_ready(self.node, pod_name)
if exceed_free_space:
# Try to expand existing PVC exceeding free space
resize_pvc(self.node, pvc_name, available_size_gb)
wait_for_events(self.node,
obj_name=pvc_name,
event_reason='VolumeResizeFailed')
# Check that app POD is up and runnig then try to write data
wait_for_pod_be_ready(self.node, pod_name)
cmd = ("dd if=/dev/urandom of=%s/autotest bs=100K count=1" %
dir_path)
ret, out, err = oc_rsh(self.node, pod_name, cmd)
self.assertEqual(
ret, 0,
"Failed to write data after failed attempt to expand PVC.")
else:
# Expand existing PVC using all the available free space
expand_size_gb = available_size_gb - pvc_size_gb
resize_pvc(self.node, pvc_name, expand_size_gb)
verify_pvc_size(self.node, pvc_name, expand_size_gb)
pv_name = get_pv_name_from_pvc(self.node, pvc_name)
verify_pv_size(self.node, pv_name, expand_size_gb)
wait_for_events(self.node,
obj_name=pvc_name,
event_reason='VolumeResizeSuccessful')
# Recreate app POD
oc_delete(self.node, 'pod', pod_name)
wait_for_resource_absence(self.node, 'pod', pod_name)
pod_name = get_pod_name_from_dc(self.node, dc_name)
wait_for_pod_be_ready(self.node, pod_name)
# Write data on the expanded PVC
cmd = ("dd if=/dev/urandom of=%s/autotest "
"bs=1M count=1025" % dir_path)
ret, out, err = oc_rsh(self.node, pod_name, cmd)
self.assertEqual(ret, 0,
"Failed to write data on the expanded PVC")
def test_create_delete_pvcs_to_make_gluster_reuse_released_space(self):
"""Validate reuse of volume space after deletion of PVCs"""
min_storage_gb = 10
# Set arbiter:disabled tags to the first 2 nodes
data_nodes = []
biggest_disks = []
self.assertGreater(len(self.node_id_list), 2)
for node_id in self.node_id_list[0:2]:
node_info = heketi_ops.heketi_node_info(self.heketi_client_node,
self.heketi_server_url,
node_id,
json=True)
biggest_disk_free_space = 0
for device in node_info['devices']:
disk_free_space = int(device['storage']['free'])
if disk_free_space < (min_storage_gb * 1024**2):
self.skipTest("Devices are expected to have more than "
"%sGb of free space" % min_storage_gb)
if disk_free_space > biggest_disk_free_space:
biggest_disk_free_space = disk_free_space
self._set_arbiter_tag_with_further_revert(
self.heketi_client_node,
self.heketi_server_url,
'device',
device['id'],
'disabled',
revert_to=device.get('tags', {}).get('arbiter'))
biggest_disks.append(biggest_disk_free_space)
self._set_arbiter_tag_with_further_revert(self.heketi_client_node,
self.heketi_server_url,
'node',
node_id,
'disabled',
revert_to=node_info.get(
'tags',
{}).get('arbiter'))
data_nodes.append(node_info)
# Set arbiter:required tag to all other nodes and their devices
arbiter_nodes = []
for node_id in self.node_id_list[2:]:
node_info = heketi_ops.heketi_node_info(self.heketi_client_node,
self.heketi_server_url,
node_id,
json=True)
for device in node_info['devices']:
self._set_arbiter_tag_with_further_revert(
self.heketi_client_node,
self.heketi_server_url,
'device',
device['id'],
'required',
revert_to=device.get('tags', {}).get('arbiter'))
self._set_arbiter_tag_with_further_revert(self.heketi_client_node,
self.heketi_server_url,
'node',
node_id,
'required',
revert_to=node_info.get(
'tags',
{}).get('arbiter'))
arbiter_nodes.append(node_info)
# Calculate size and amount of volumes to be created
pvc_size = int(min(biggest_disks) / 1024**2)
pvc_amount = max([len(n['devices']) for n in data_nodes]) + 1
# Create sc with gluster arbiter info
self.create_storage_class(is_arbiter_vol=True)
# Create and delete 3 small volumes concurrently
pvc_names = []
for i in range(3):
pvc_name = oc_create_pvc(self.node,
self.sc_name,
pvc_name_prefix='arbiter-pvc',
pvc_size=int(pvc_size / 3))
pvc_names.append(pvc_name)
exception_exists = False
for pvc_name in pvc_names:
try:
verify_pvc_status_is_bound(self.node, pvc_name)
except Exception:
for pvc_name in pvc_names:
self.addCleanup(wait_for_resource_absence, self.node,
'pvc', pvc_name)
for pvc_name in pvc_names:
self.addCleanup(oc_delete, self.node, 'pvc', pvc_name)
exception_exists = True
if exception_exists:
raise
for pvc_name in pvc_names:
oc_delete(self.node, 'pvc', pvc_name)
for pvc_name in pvc_names:
wait_for_resource_absence(self.node, 'pvc', pvc_name)
# Create and delete big volumes in a loop
for i in range(pvc_amount):
pvc_name = oc_create_pvc(self.node,
self.sc_name,
pvc_name_prefix='arbiter-pvc',
pvc_size=pvc_size)
try:
verify_pvc_status_is_bound(self.node, pvc_name)
except Exception:
self.addCleanup(wait_for_resource_absence, self.node, 'pvc',
pvc_name)
self.addCleanup(oc_delete, self.node, 'pvc', pvc_name)
raise
oc_delete(self.node, 'pvc', pvc_name)
wait_for_resource_absence(self.node, 'pvc', pvc_name)
def delete_pvc(self, ocp_node):
oc_delete(ocp_node, 'pvc', self.pvc_name)
def delete_storageclass(ocp_node, sc_name, timeout=120):
wait_for_sc_unused(ocp_node, sc_name, timeout)
oc_delete(ocp_node, 'storageclass', sc_name)
