def test_provisioner_mount(client, core_api, storage_class, pvc, pod): # NOQA
"""
Test that a StorageClass provisioned volume can be created, mounted,
unmounted, and deleted properly on the Kubernetes cluster.
Fixtures are torn down here in reverse order that they are specified as a
parameter. Take caution when reordering test fixtures.
"""
# Prepare pod and volume specs.
pod_name = 'provisioner-mount-test'
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [
create_pvc_spec(pvc['metadata']['name'])
]
pvc['spec']['storageClassName'] = DEFAULT_STORAGECLASS_NAME
storage_class['metadata']['name'] = DEFAULT_STORAGECLASS_NAME
volume_size = DEFAULT_VOLUME_SIZE * Gi
create_storage(core_api, storage_class, pvc)
create_and_wait_pod(core_api, pod)
pvc_volume_name = get_volume_name(core_api, pvc['metadata']['name'])
# Confirm that the volume has all the correct parameters we gave it.
volumes = client.list_volume()
assert len(volumes) == 1
assert volumes[0]["name"] == pvc_volume_name
assert volumes[0]["size"] == str(volume_size)
assert volumes[0]["numberOfReplicas"] == \
int(storage_class['parameters']['numberOfReplicas'])
assert volumes[0]["state"] == "attached"
def test_provisioner_io(client, core_api, storage_class, pvc, pod): # NOQA
"""
Test that input and output on a StorageClass provisioned
PersistentVolumeClaim works as expected.
Fixtures are torn down here in reverse order that they are specified as a
parameter. Take caution when reordering test fixtures.
"""
# Prepare pod and volume specs.
pod_name = 'provisioner-io-test'
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [
create_pvc_spec(pvc['metadata']['name'])
]
pvc['spec']['storageClassName'] = DEFAULT_STORAGECLASS_NAME
storage_class['metadata']['name'] = DEFAULT_STORAGECLASS_NAME
test_data = generate_random_data(VOLUME_RWTEST_SIZE)
create_storage(core_api, storage_class, pvc)
create_and_wait_pod(core_api, pod)
pvc_volume_name = get_volume_name(core_api, pvc['metadata']['name'])
write_volume_data(core_api, pod_name, test_data)
delete_and_wait_pod(core_api, pod_name)
common.wait_for_volume_detached(client, pvc_volume_name)
pod_name = 'flexvolume-provisioner-io-test-2'
pod['metadata']['name'] = pod_name
create_and_wait_pod(core_api, pod)
resp = read_volume_data(core_api, pod_name)
assert resp == test_data
def csi_io_test(client, core_api, csi_pv, pvc, pod_make, base_image=""): # NOQA
pv_name = generate_volume_name()
pod_name = 'csi-io-test'
create_and_wait_csi_pod_named_pv(pv_name, pod_name, client, core_api,
csi_pv, pvc, pod_make, base_image, "")
test_data = generate_random_data(VOLUME_RWTEST_SIZE)
write_volume_data(core_api, pod_name, test_data)
delete_and_wait_pod(core_api, pod_name)
common.wait_for_volume_detached(client, csi_pv['metadata']['name'])
pod_name = 'csi-io-test-2'
pod = pod_make(name=pod_name)
pod['spec']['volumes'] = [
create_pvc_spec(pv_name)
]
csi_pv['metadata']['name'] = pv_name
csi_pv['spec']['csi']['volumeHandle'] = pv_name
pvc['metadata']['name'] = pv_name
pvc['spec']['volumeName'] = pv_name
update_storageclass_references(CSI_PV_TEST_STORAGE_NAME, csi_pv, pvc)
create_and_wait_pod(core_api, pod)
resp = read_volume_data(core_api, pod_name)
assert resp == test_data
def create_and_wait_csi_pod_named_pv(pv_name, pod_name, client, core_api, csi_pv, pvc, pod_make, base_image, from_backup): # NOQA
pod = pod_make(name=pod_name)
pod['spec']['volumes'] = [
create_pvc_spec(pv_name)
]
csi_pv['metadata']['name'] = pv_name
csi_pv['spec']['csi']['volumeHandle'] = pv_name
csi_pv['spec']['csi']['volumeAttributes']['fromBackup'] = from_backup
pvc['metadata']['name'] = pv_name
pvc['spec']['volumeName'] = pv_name
update_storageclass_references(CSI_PV_TEST_STORAGE_NAME, csi_pv, pvc)
create_pv_storage(core_api, client, csi_pv, pvc, base_image, from_backup)
create_and_wait_pod(core_api, pod)
def flexvolume_mount_test(client, core_api, flexvolume, pod, volume_size): # NOQA
pod_name = 'flexvolume-mount-test'
pod['metadata']['name'] = pod_name
pod['spec']['containers'][0]['volumeMounts'][0]['name'] = \
flexvolume['name']
pod['spec']['volumes'] = [
flexvolume
]
create_and_wait_pod(core_api, pod)
volumes = client.list_volume()
assert len(volumes) == 1
assert volumes[0]["name"] == flexvolume['name']
assert volumes[0]["size"] == str(volume_size)
assert volumes[0]["numberOfReplicas"] == int(
flexvolume["flexVolume"]["options"]["numberOfReplicas"])
assert volumes[0]["state"] == "attached"
def flexvolume_io_test(client, core_api, flexvolume, pod): # NOQA
pod_name = 'flexvolume-io-test'
pod['metadata']['name'] = pod_name
pod['spec']['containers'][0]['volumeMounts'][0]['name'] = \
flexvolume['name']
pod['spec']['volumes'] = [
flexvolume
]
test_data = generate_random_data(VOLUME_RWTEST_SIZE)
create_and_wait_pod(core_api, pod)
common.write_volume_data(core_api, pod_name, test_data)
delete_and_wait_pod(core_api, pod_name)
wait_for_volume_detached(client, flexvolume["name"])
pod_name = 'volume-driver-io-test-2'
pod['metadata']['name'] = pod_name
create_and_wait_pod(core_api, pod)
resp = read_volume_data(core_api, pod_name)
assert resp == test_data
def test_provisioner_params(client, core_api, storage_class, pvc, pod): # NOQA
"""
Test that substituting different StorageClass parameters is reflected in
the resulting PersistentVolumeClaim.
Fixtures are torn down here in reverse order that they are specified as a
parameter. Take caution when reordering test fixtures.
"""
# Prepare pod and volume specs.
pod_name = 'provisioner-params-test'
volume_size = 2 * Gi
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [
create_pvc_spec(pvc['metadata']['name'])
]
pvc['spec']['resources']['requests']['storage'] = \
size_to_string(volume_size)
pvc['spec']['storageClassName'] = DEFAULT_STORAGECLASS_NAME
storage_class['metadata']['name'] = DEFAULT_STORAGECLASS_NAME
storage_class['parameters'] = {
'numberOfReplicas': '2',
'staleReplicaTimeout': '20'
}
create_storage(core_api, storage_class, pvc)
create_and_wait_pod(core_api, pod)
pvc_volume_name = get_volume_name(core_api, pvc['metadata']['name'])
# Confirm that the volume has all the correct parameters we gave it.
volumes = client.list_volume()
assert len(volumes) == 1
assert volumes[0]["name"] == pvc_volume_name
assert volumes[0]["size"] == str(volume_size)
assert volumes[0]["numberOfReplicas"] == \
int(storage_class['parameters']['numberOfReplicas'])
assert volumes[0]["state"] == "attached"
def test_provisioner_tags(client, core_api, node_default_tags, storage_class,
pvc, pod): # NOQA
"""
Test that a StorageClass can properly provision a volume with requested
Tags.
"""
# Prepare pod and volume specs.
pod_name = 'provisioner-tags-test'
tag_spec = {
"disk": ["ssd", "nvme"],
"expected": 1,
"node": ["storage", "main"]
}
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [create_pvc_spec(pvc['metadata']['name'])]
pvc['spec']['storageClassName'] = DEFAULT_STORAGECLASS_NAME
storage_class['metadata']['name'] = DEFAULT_STORAGECLASS_NAME
storage_class['parameters']['diskSelector'] = 'ssd,nvme'
storage_class['parameters']['nodeSelector'] = 'storage,main'
volume_size = DEFAULT_VOLUME_SIZE * Gi
create_storage(core_api, storage_class, pvc)
create_and_wait_pod(core_api, pod)
pvc_volume_name = get_volume_name(core_api, pvc['metadata']['name'])
# Confirm that the volume has all the correct parameters we gave it.
volumes = client.list_volume()
assert len(volumes) == 1
assert volumes[0]["name"] == pvc_volume_name
assert volumes[0]["size"] == str(volume_size)
assert volumes[0]["numberOfReplicas"] == \
int(storage_class['parameters']['numberOfReplicas'])
assert volumes[0]["state"] == "attached"
check_volume_replicas(volumes[0], tag_spec, node_default_tags)
def test_rwx_parallel_writing(core_api, statefulset, pod): # NOQA
"""
Test parallel writing of data
1. Create a StatefulSet of 1 pod with VolumeClaimTemplate where accessMode
is 'RWX'.
2. Wait for StatefulSet to come up healthy.
3. Create another statefulSet with same pvc which got created with first
statefulSet.
4. Wait for statefulSet to come up healthy.
5. Start writing 800 MB data in first statefulSet `file 1` and start
writing 500 MB data in second statefulSet `file 2`.
6. Compute md5sum.
7. Check the data md5sum in share manager pod volume
"""
statefulset_name = 'statefulset-rwx-parallel-writing-test'
statefulset['metadata']['name'] = \
statefulset['spec']['selector']['matchLabels']['app'] = \
statefulset['spec']['serviceName'] = \
statefulset['spec']['template']['metadata']['labels']['app'] = \
statefulset_name
statefulset['spec']['replicas'] = 1
statefulset['spec']['volumeClaimTemplates'][0]['spec']['storageClassName']\
= 'longhorn'
statefulset['spec']['volumeClaimTemplates'][0]['spec']['accessModes'] \
= ['ReadWriteMany']
create_and_wait_statefulset(statefulset)
statefulset_pod_name = statefulset_name + '-0'
pvc_name = \
statefulset['spec']['volumeClaimTemplates'][0]['metadata']['name'] \
+ '-' + statefulset_name + '-0'
pv_name = get_volume_name(core_api, pvc_name)
share_manager_name = 'share-manager-' + pv_name
pod_name = 'pod-parallel-write-test'
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [create_pvc_spec(pvc_name)]
create_and_wait_pod(core_api, pod)
with Pool(2) as p:
p.map(
write_data_into_pod,
[statefulset_pod_name + ':/data/test1', pod_name + ':/data/test2'])
md5sum1 = get_pod_data_md5sum(core_api, statefulset_pod_name, 'data/test1')
md5sum2 = get_pod_data_md5sum(core_api, pod_name, 'data/test2')
command1 = 'md5sum /export' + '/' + pv_name + '/' + 'test1' + \
" | awk '{print $1}'"
share_manager_data1 = exec_command_in_pod(core_api, command1,
share_manager_name,
LONGHORN_NAMESPACE)
assert md5sum1 == share_manager_data1
command2 = 'md5sum /export' + '/' + pv_name + '/' + 'test2' + \
" | awk '{print $1}'"
share_manager_data2 = exec_command_in_pod(core_api, command2,
share_manager_name,
LONGHORN_NAMESPACE)
assert md5sum2 == share_manager_data2
def generate_load(request):
index = get_random_suffix()
longhorn_api_client = get_longhorn_api_client()
k8s_api_client = get_core_api_client()
check_and_set_backupstore(longhorn_api_client)
volume_name = STRESS_VOLUME_NAME_PREFIX + index
pv_name = STRESS_PV_NAME_PREFIX + index
pvc_name = STRESS_PVC_NAME_PREFIX + index
pod_name = STRESS_POD_NAME_PREFIX + index
atexit.register(remove_datafile, pod_name)
atexit.register(delete_and_wait_longhorn, longhorn_api_client, volume_name)
atexit.register(delete_and_wait_pv, k8s_api_client, pv_name)
atexit.register(delete_and_wait_pvc, k8s_api_client, pvc_name)
atexit.register(delete_and_wait_pod, k8s_api_client, pod_name)
longhorn_volume = create_and_check_volume(longhorn_api_client,
volume_name,
size=VOLUME_SIZE)
wait_for_volume_detached(longhorn_api_client, volume_name)
pod_manifest = generate_pod_with_pvc_manifest(pod_name, pvc_name)
create_pv_for_volume(longhorn_api_client, k8s_api_client, longhorn_volume,
pv_name)
create_pvc_for_volume(longhorn_api_client, k8s_api_client, longhorn_volume,
pvc_name)
create_and_wait_pod(k8s_api_client, pod_manifest)
snapshots_md5sum = dict()
write_data(k8s_api_client, pod_name)
create_recurring_jobs(longhorn_api_client, volume_name)
global N_RANDOM_ACTIONS
for round in range(N_RANDOM_ACTIONS):
action = randrange(0, 8)
if action == 0:
print("write data started: " + time_now(), end=', ')
write_data(k8s_api_client, pod_name)
print("ended: " + time_now())
elif action == 1:
print("delete data started: " + time_now(), end=', ')
delete_data(k8s_api_client, pod_name)
print("ended: " + time_now())
elif action == 2:
print("create snapshot started: " + time_now(), end=', ')
snapshot_create_and_record_md5sum(longhorn_api_client,
k8s_api_client, volume_name,
pod_name, snapshots_md5sum)
print("ended: " + time_now())
elif action == 3:
print("delete random snapshot started: " + time_now(), end=', ')
delete_random_snapshot(longhorn_api_client, volume_name,
snapshots_md5sum)
print("ended: " + time_now())
elif action == 4:
print("revert random snapshot started: " + time_now(), end=', ')
revert_random_snapshot(longhorn_api_client, k8s_api_client,
volume_name, pod_manifest, snapshots_md5sum)
print("ended: " + time_now())
elif action == 5:
print("create backup started: " + time_now(), end=', ')
backup_create_and_record_md5sum(longhorn_api_client,
k8s_api_client, volume_name,
pod_name, snapshots_md5sum)
print("ended: " + time_now())
elif action == 6:
print("delete replica started: " + time_now(), end=', ')
delete_replica(longhorn_api_client, volume_name)
print("ended: " + time_now())
elif action == 7:
print("restore random backup started: " + time_now(), end=', ')
restore_and_check_random_backup(longhorn_api_client,
k8s_api_client, volume_name,
pod_name, snapshots_md5sum)
print("ended: " + time_now())
clean_volume_backups(longhorn_api_client, volume_name)
def test_csi_volumesnapshot_restore_existing_backup(
set_random_backupstore, # NOQA
client, # NOQA
core_api, # NOQA
volume_name, # NOQA
csi_pv, # NOQA
pvc, # NOQA
pod_make, # NOQA
volumesnapshotclass, # NOQA
volumesnapshotcontent,
volumesnapshot, # NOQA
volsnapshotclass_delete_policy, # NOQA
backup_is_deleted): # NOQA
"""
Test retention of a backup while deleting the associated `VolumeSnapshot`
via the csi snapshotter
Context:
We want to allow the user to programmatically create/restore/delete
longhorn backups via the csi snapshot mechanism
ref: https://kubernetes.io/docs/concepts/storage/volume-snapshots/
Setup:
1. Make sure your cluster contains the below crds
https://github.com/kubernetes-csi/external-snapshotter
/tree/master/client/config/crd
2. Make sure your cluster contains the snapshot controller
https://github.com/kubernetes-csi/external-snapshotter
/tree/master/deploy/kubernetes/snapshot-controller
Steps:
1. create new snapshotClass with deletionPolicy set to Retain
2. call csi_volumesnapshot_creation_test(snapshotClass=custom)
3. call csi_volumesnapshot_restore_test()
4. call csi_volumesnapshot_deletion_test(deletionPolicy='Retain'):
5. cleanup
"""
csisnapclass = \
volumesnapshotclass(name="snapshotclass",
deletepolicy=volsnapshotclass_delete_policy)
pod_name, pv_name, pvc_name, md5sum = \
prepare_pod_with_data_in_mb(client, core_api,
csi_pv, pvc, pod_make,
volume_name,
data_path="/data/test")
volume = client.by_id_volume(volume_name)
snap = create_snapshot(client, volume_name)
volume.snapshotBackup(name=snap.name)
wait_for_backup_completion(client, volume_name, snap.name)
bv, b = find_backup(client, volume_name, snap.name)
csivolsnap_name = volume_name + "-volumesnapshot"
csivolsnap_namespace = "default"
volsnapcontent = \
volumesnapshotcontent("volsnapcontent",
csisnapclass["metadata"]["name"],
"Delete",
"bs://" + volume_name + "/" + b.name,
csivolsnap_name,
csivolsnap_namespace)
csivolsnap = volumesnapshot(csivolsnap_name, csivolsnap_namespace,
csisnapclass["metadata"]["name"],
"volumeSnapshotContentName",
volsnapcontent["metadata"]["name"])
restore_pvc_name = pvc["metadata"]["name"] + "-restore"
restore_pvc_size = pvc["spec"]["resources"]["requests"]["storage"]
restore_csi_volume_snapshot(core_api, client, csivolsnap, restore_pvc_name,
restore_pvc_size)
restore_pod = pod_make()
restore_pod_name = restore_pod["metadata"]["name"]
restore_pod['spec']['volumes'] = [create_pvc_spec(restore_pvc_name)]
create_and_wait_pod(core_api, restore_pod)
restore_md5sum = \
get_pod_data_md5sum(core_api, restore_pod_name, path="/data/test")
assert restore_md5sum == md5sum
# Delete volumeSnapshot test
delete_volumesnapshot(csivolsnap["metadata"]["name"], "default")
if backup_is_deleted is False:
find_backup(client, volume_name, b["snapshotName"])
else:
wait_for_backup_delete(client, volume_name, b["name"])
def test_cloning_interrupted(client, core_api, pvc, pod, clone_pvc, clone_pod): # NOQA
"""
1. Create a PVC:
```yaml
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: source-pvc
spec:
storageClassName: longhorn
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 3Gi
```
2. Specify the `source-pvc` in a pod yaml and start the pod
3. Wait for the pod to be running, write 500MB of data to the mount
path of the volume
4. Clone a volume by creating the PVC:
```yaml
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: cloned-pvc
spec:
storageClassName: longhorn
dataSource:
name: source-pvc
kind: PersistentVolumeClaim
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 3Gi
```
5. Wait for the `CloneStatus.State` in `cloned-pvc` to be `initiated`
6. Kill all replicas process of the `source-pvc`
7. Wait for the `CloneStatus.State` in `cloned-pvc` to be `failed`
8. Clean up `clone-pvc`
9. Redeploy `cloned-pvc` and clone pod
10. In 3-min retry loop, verify cloned pod become running
11. `cloned-pvc` has the same data as `source-pvc`
12. In 2-min retry loop, verify the volume of the `clone-pvc`
eventually becomes healthy.
"""
# Step-1
source_pvc_name = 'source-pvc' + generate_random_suffix()
pvc['metadata']['name'] = source_pvc_name
pvc['spec']['storageClassName'] = 'longhorn'
core_api.create_namespaced_persistent_volume_claim(
body=pvc, namespace='default')
wait_for_pvc_phase(core_api, source_pvc_name, "Bound")
# Step-2
pod_name = 'source-pod' + generate_random_suffix()
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [create_pvc_spec(source_pvc_name)]
create_and_wait_pod(core_api, pod)
# Step-3
write_pod_volume_random_data(core_api, pod_name,
'/data/test', DATA_SIZE_IN_MB_3)
source_data = get_pod_data_md5sum(core_api, pod_name, '/data/test')
source_volume_name = get_volume_name(core_api, source_pvc_name)
# Step-4
clone_pvc_name = 'clone-pvc' + generate_random_suffix()
clone_pvc['metadata']['name'] = clone_pvc_name
clone_pvc['spec']['storageClassName'] = 'longhorn'
clone_pvc['spec']['dataSource'] = {
'name': source_pvc_name,
'kind': 'PersistentVolumeClaim'
}
core_api.create_namespaced_persistent_volume_claim(
body=clone_pvc, namespace='default')
# Step-5
clone_volume_name = get_clone_volume_name(client, source_volume_name)
wait_for_volume_clone_status(client, clone_volume_name, VOLUME_FIELD_STATE,
'initiated')
# Step-6
crash_replica_processes(client, core_api, source_volume_name)
# Step-7
wait_for_volume_faulted(client, source_volume_name)
wait_for_volume_clone_status(client, clone_volume_name, VOLUME_FIELD_STATE,
'failed')
# Step-8
delete_and_wait_pvc(core_api, clone_pvc_name)
# Step-9
clone_pvc_name = 'clone-pvc-2' + generate_random_suffix()
clone_pvc['metadata']['name'] = clone_pvc_name
clone_pvc['spec']['storageClassName'] = 'longhorn'
clone_pvc['spec']['dataSource'] = {
'name': source_pvc_name,
'kind': 'PersistentVolumeClaim'
}
core_api.create_namespaced_persistent_volume_claim(
body=clone_pvc, namespace='default')
wait_for_pvc_phase(core_api, clone_pvc_name, "Bound")
# Step-9
clone_pod_name = 'clone-pod' + generate_random_suffix()
clone_pod['metadata']['name'] = clone_pod_name
clone_pod['spec']['volumes'] = [create_pvc_spec(clone_pvc_name)]
create_and_wait_pod(core_api, clone_pod)
# Step-10
clone_volume_name = get_volume_name(core_api, clone_pvc_name)
wait_for_volume_clone_status(client, clone_volume_name, VOLUME_FIELD_STATE,
VOLUME_FIELD_CLONE_COMPLETED)
# Step-11
clone_data = get_pod_data_md5sum(core_api, clone_pod_name, '/data/test')
assert source_data == clone_data
# Step-12
wait_for_volume_healthy(client, clone_volume_name)
def test_kubernetes_status(
client,
core_api,
storage_class, # NOQA
statefulset,
csi_pv,
pvc,
pod): # NOQA
statefulset_name = 'kubernetes-status-test'
update_statefulset_manifests(statefulset, storage_class, statefulset_name)
storage_class['reclaimPolicy'] = 'Retain'
create_storage_class(storage_class)
create_and_wait_statefulset(statefulset)
pod_info = get_statefulset_pod_info(core_api, statefulset)
volume_info = [p['pv_name'] for p in pod_info]
extra_pod_name = 'extra-pod-using-' + volume_info[1]
pod['metadata']['name'] = extra_pod_name
p2 = core_api.read_namespaced_pod(name=pod_info[1]['pod_name'],
namespace='default')
pod['spec']['nodeName'] = p2.spec.node_name
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': pod_info[1]['pvc_name'],
},
}]
create_and_wait_pod(core_api, pod)
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert k_status['pvName'] == p['pv_name']
assert k_status['pvStatus'] == 'Bound'
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == p['pvc_name']
assert not k_status['lastPVCRefAt']
assert not k_status['lastPodRefAt']
if i == 0:
assert len(workloads) == 1
assert workloads[0]['podName'] == p['pod_name']
assert workloads[0]['workloadName'] == statefulset_name
assert workloads[0]['workloadType'] == 'StatefulSet'
for _ in range(RETRY_COUNTS):
if workloads[0]['podStatus'] == 'Running':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert workloads[0]['podStatus'] == 'Running'
if i == 1:
assert len(k_status['workloadsStatus']) == 2
if workloads[0]['podName'] == pod_info[i]['pod_name']:
assert workloads[1]['podName'] == extra_pod_name
assert workloads[0]['workloadName'] == statefulset_name
assert workloads[0]['workloadType'] == 'StatefulSet'
assert not workloads[1]['workloadName']
assert not workloads[1]['workloadType']
else:
assert workloads[1]['podName'] == pod_info[i]['pod_name']
assert workloads[0]['podName'] == extra_pod_name
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
assert workloads[1]['workloadName'] == statefulset_name
assert workloads[1]['workloadType'] == 'StatefulSet'
for _ in range(RETRY_COUNTS):
if workloads[0]['podStatus'] == 'Running' and \
workloads[1]['podStatus'] == 'Running':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert len(workloads) == 2
assert workloads[0]['podStatus'] == 'Running'
assert workloads[1]['podStatus'] == 'Running'
ks_list = [{}, {}]
delete_and_wait_statefulset_only(core_api, statefulset)
# the extra pod is still using the 2nd volume
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
ks_list[i]['pvName'] = p['pv_name']
ks_list[i]['pvStatus'] = 'Bound'
ks_list[i]['namespace'] = 'default'
ks_list[i]['pvcName'] = p['pvc_name']
ks_list[i]['lastPVCRefAt'] = ''
if i == 0:
ks_list[i]['lastPodRefAt'] = 'not empty'
ks_list[i]['workloadsStatus'] = [
{
'podName': p['pod_name'],
'podStatus': 'Running',
'workloadName': statefulset_name,
'workloadType': 'StatefulSet',
},
]
if i == 1:
ks_list[i]['lastPodRefAt'] = ''
ks_list[i]['workloadsStatus'] = [{
'podName': extra_pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': '',
}]
wait_volume_kubernetes_status(client, volume_name, ks_list[i])
# deleted extra_pod, all volumes have no workload
delete_and_wait_pod(core_api, pod['metadata']['name'])
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
ks_list[i]['lastPodRefAt'] = 'not empty'
wait_volume_kubernetes_status(client, volume_name, ks_list[i])
# deleted pvc only.
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
delete_and_wait_pvc(core_api, p['pvc_name'])
ks_list[i]['pvStatus'] = 'Released'
ks_list[i]['lastPVCRefAt'] = 'not empty'
wait_volume_kubernetes_status(client, volume_name, ks_list[i])
# deleted pv only.
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
delete_and_wait_pv(core_api, p['pv_name'])
ks_list[i]['pvName'] = ''
ks_list[i]['pvStatus'] = ''
wait_volume_kubernetes_status(client, volume_name, ks_list[i])
# reuse that volume
for p, volume_name in zip(pod_info, volume_info):
p['pod_name'] = p['pod_name'].replace('kubernetes-status-test',
'kubernetes-status-test-reuse')
p['pvc_name'] = p['pvc_name'].replace('kubernetes-status-test',
'kubernetes-status-test-reuse')
p['pv_name'] = p['pvc_name']
csi_pv['metadata']['name'] = p['pv_name']
csi_pv['spec']['csi']['volumeHandle'] = volume_name
core_api.create_persistent_volume(csi_pv)
pvc['metadata']['name'] = p['pvc_name']
pvc['spec']['volumeName'] = p['pv_name']
core_api.create_namespaced_persistent_volume_claim(body=pvc,
namespace='default')
pod['metadata']['name'] = p['pod_name']
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': p['pvc_name'],
},
}]
create_and_wait_pod(core_api, pod)
ks = {
'pvName':
p['pv_name'],
'pvStatus':
'Bound',
'namespace':
'default',
'pvcName':
p['pvc_name'],
'lastPVCRefAt':
'',
'lastPodRefAt':
'',
'workloadsStatus': [
{
'podName': p['pod_name'],
'podStatus': 'Running',
'workloadName': '',
'workloadType': '',
},
],
}
wait_volume_kubernetes_status(client, volume_name, ks)
delete_and_wait_pod(core_api, p['pod_name'])
# Since persistentVolumeReclaimPolicy of csi_pv is `Delete`,
# we don't need to delete bounded pv manually
delete_and_wait_pvc(core_api, p['pvc_name'])
wait_delete_pv(core_api, p['pv_name'])
def test_replica_auto_balance_zone_best_effort_with_data_locality(
client, core_api, volume_name, pod): # NOQA
"""
Background:
Given set `replica-soft-anti-affinity` to `true`.
And set `replica-zone-soft-anti-affinity` to `true`.
And set `default-data-locality` to `best-effort`.
And set `replicaAutoBalance` to `best-effort`.
And set node-1 to zone-1.
set node-2 to zone-1.
set node-3 to zone-2.
And create volume with 2 replicas.
And create pv for volume.
And create pvc for volume.
Scenario Outline: replica auto-balance zones with best-effort should
not remove pod local replicas when data locality is
enabled (best-effort).
Given create and wait pod on <pod-node>.
And disable scheduling and evict node-3.
And count replicas on each nodes.
And 1 replica running on <pod-node>.
1 replica running on <duplicate-node>.
0 replica running on node-3.
When enable scheduling for node-3.
Then count replicas on each nodes.
And 1 replica running on <pod-node>.
0 replica running on <duplicate-node>.
1 replica running on node-3.
And count replicas in each zones.
And 1 replica running in zone-1.
1 replica running in zone-2.
And loop 3 times with each wait 5 seconds and count replicas on each nodes.
To ensure no addition scheduling is happening.
1 replica running on <pod-node>.
0 replica running on <duplicate-node>.
1 replica running on node-3.
And delete pod.
Examples:
| pod-node | duplicate-node |
| node-1 | node-2 |
| node-2 | node-1 |
| node-1 | node-2 |
"""
common.update_setting(client, SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY,
"true")
common.update_setting(client, SETTING_REPLICA_ZONE_SOFT_ANTI_AFFINITY,
"true")
common.update_setting(client, SETTING_DEFAULT_DATA_LOCALITY, "best-effort")
common.update_setting(client, SETTING_REPLICA_AUTO_BALANCE, "best-effort")
n1, n2, n3 = client.list_node()
set_k8s_node_zone_label(core_api, n1.name, ZONE1)
set_k8s_node_zone_label(core_api, n2.name, ZONE1)
set_k8s_node_zone_label(core_api, n3.name, ZONE2)
wait_longhorn_node_zone_updated(client)
n_replicas = 2
volume = create_and_check_volume(client,
volume_name,
num_of_replicas=n_replicas)
common.create_pv_for_volume(client, core_api, volume, volume_name)
common.create_pvc_for_volume(client, core_api, volume, volume_name)
pod['spec']['volumes'] = [{
"name": "pod-data",
"persistentVolumeClaim": {
"claimName": volume_name
}
}]
for i in range(1, 4):
pod_node_name = n2.name if i % 2 == 0 else n1.name
pod['spec']['nodeSelector'] = {"kubernetes.io/hostname": pod_node_name}
common.create_and_wait_pod(core_api, pod)
client.update(n3, allowScheduling=False, evictionRequested=True)
duplicate_node = [n1.name, n2.name]
duplicate_node.remove(pod_node_name)
for _ in range(RETRY_COUNTS):
pod_node_r_count = common.get_host_replica_count(client,
volume_name,
pod_node_name,
chk_running=True)
duplicate_node_r_count = common.get_host_replica_count(
client, volume_name, duplicate_node[0], chk_running=True)
balance_node_r_count = common.get_host_replica_count(
client, volume_name, n3.name, chk_running=False)
if pod_node_r_count == duplicate_node_r_count == 1 and \
balance_node_r_count == 0:
break
time.sleep(RETRY_INTERVAL)
assert pod_node_r_count == 1
assert duplicate_node_r_count == 1
assert balance_node_r_count == 0
client.update(n3, allowScheduling=True)
for _ in range(RETRY_COUNTS):
pod_node_r_count = common.get_host_replica_count(client,
volume_name,
pod_node_name,
chk_running=True)
duplicate_node_r_count = common.get_host_replica_count(
client, volume_name, duplicate_node[0], chk_running=False)
balance_node_r_count = common.get_host_replica_count(
client, volume_name, n3.name, chk_running=True)
if pod_node_r_count == balance_node_r_count == 1 and \
duplicate_node_r_count == 0:
break
time.sleep(RETRY_INTERVAL)
assert pod_node_r_count == 1
assert duplicate_node_r_count == 0
assert balance_node_r_count == 1
z1_r_count = get_zone_replica_count(client,
volume_name,
ZONE1,
chk_running=True)
z2_r_count = get_zone_replica_count(client,
volume_name,
ZONE2,
chk_running=True)
assert z1_r_count == z2_r_count == 1
# loop 3 times and each to wait 5 seconds to ensure there is no
# re-scheduling happening.
for _ in range(3):
time.sleep(5)
assert pod_node_r_count == common.get_host_replica_count(
client, volume_name, pod_node_name, chk_running=True)
assert duplicate_node_r_count == common.get_host_replica_count(
client, volume_name, duplicate_node[0], chk_running=False)
assert balance_node_r_count == common.get_host_replica_count(
client, volume_name, n3.name, chk_running=True)
common.delete_and_wait_pod(core_api, pod['metadata']['name'])
def restore_inc_test(client, core_api, volume_name, pod): # NOQA
std_volume = create_and_check_volume(client, volume_name, 2, SIZE)
lht_host_id = get_self_host_id()
std_volume.attach(hostId=lht_host_id)
std_volume = common.wait_for_volume_healthy(client, volume_name)
with pytest.raises(Exception) as e:
std_volume.activate(frontend="blockdev")
assert "already in active mode" in str(e.value)
data0 = {'len': 4 * 1024, 'pos': 0}
data0['content'] = common.generate_random_data(data0['len'])
bv, backup0, _, data0 = create_backup(client, volume_name, data0)
sb_volume0_name = "sb-0-" + volume_name
sb_volume1_name = "sb-1-" + volume_name
sb_volume2_name = "sb-2-" + volume_name
client.create_volume(name=sb_volume0_name,
size=SIZE,
numberOfReplicas=2,
fromBackup=backup0['url'],
frontend="",
standby=True)
client.create_volume(name=sb_volume1_name,
size=SIZE,
numberOfReplicas=2,
fromBackup=backup0['url'],
frontend="",
standby=True)
client.create_volume(name=sb_volume2_name,
size=SIZE,
numberOfReplicas=2,
fromBackup=backup0['url'],
frontend="",
standby=True)
common.wait_for_volume_restoration_completed(client, sb_volume0_name)
common.wait_for_volume_restoration_completed(client, sb_volume1_name)
common.wait_for_volume_restoration_completed(client, sb_volume2_name)
sb_volume0 = common.wait_for_volume_healthy(client, sb_volume0_name)
sb_volume1 = common.wait_for_volume_healthy(client, sb_volume1_name)
sb_volume2 = common.wait_for_volume_healthy(client, sb_volume2_name)
for i in range(RETRY_COUNTS):
sb_volume0 = client.by_id_volume(sb_volume0_name)
sb_volume1 = client.by_id_volume(sb_volume1_name)
sb_volume2 = client.by_id_volume(sb_volume2_name)
sb_engine0 = get_volume_engine(sb_volume0)
sb_engine1 = get_volume_engine(sb_volume1)
sb_engine2 = get_volume_engine(sb_volume2)
if sb_volume0["lastBackup"] != backup0["name"] or \
sb_volume1["lastBackup"] != backup0["name"] or \
sb_volume2["lastBackup"] != backup0["name"] or \
sb_engine0["lastRestoredBackup"] != backup0["name"] or \
sb_engine1["lastRestoredBackup"] != backup0["name"] or \
sb_engine2["lastRestoredBackup"] != backup0["name"]:
time.sleep(RETRY_INTERVAL)
else:
break
assert sb_volume0["standby"] is True
assert sb_volume0["lastBackup"] == backup0["name"]
assert sb_volume0["frontend"] == ""
assert sb_volume0["disableFrontend"] is True
assert sb_volume0["initialRestorationRequired"] is False
sb_engine0 = get_volume_engine(sb_volume0)
assert sb_engine0["lastRestoredBackup"] == backup0["name"]
assert sb_engine0["requestedBackupRestore"] == backup0["name"]
assert sb_volume1["standby"] is True
assert sb_volume1["lastBackup"] == backup0["name"]
assert sb_volume1["frontend"] == ""
assert sb_volume1["disableFrontend"] is True
assert sb_volume1["initialRestorationRequired"] is False
sb_engine1 = get_volume_engine(sb_volume1)
assert sb_engine1["lastRestoredBackup"] == backup0["name"]
assert sb_engine1["requestedBackupRestore"] == backup0["name"]
assert sb_volume2["standby"] is True
assert sb_volume2["lastBackup"] == backup0["name"]
assert sb_volume2["frontend"] == ""
assert sb_volume2["disableFrontend"] is True
assert sb_volume2["initialRestorationRequired"] is False
sb_engine2 = get_volume_engine(sb_volume2)
assert sb_engine2["lastRestoredBackup"] == backup0["name"]
assert sb_engine2["requestedBackupRestore"] == backup0["name"]
sb0_snaps = sb_volume0.snapshotList()
assert len(sb0_snaps) == 2
for s in sb0_snaps:
if s['name'] != "volume-head":
sb0_snap = s
assert sb0_snaps
with pytest.raises(Exception) as e:
sb_volume0.snapshotCreate()
assert "cannot create snapshot for standby volume" in str(e.value)
with pytest.raises(Exception) as e:
sb_volume0.snapshotRevert(name=sb0_snap["name"])
assert "cannot revert snapshot for standby volume" in str(e.value)
with pytest.raises(Exception) as e:
sb_volume0.snapshotDelete(name=sb0_snap["name"])
assert "cannot delete snapshot for standby volume" in str(e.value)
with pytest.raises(Exception) as e:
sb_volume0.snapshotBackup(name=sb0_snap["name"])
assert "cannot create backup for standby volume" in str(e.value)
with pytest.raises(Exception) as e:
sb_volume0.pvCreate(pvName=sb_volume0_name)
assert "cannot create PV for standby volume" in str(e.value)
with pytest.raises(Exception) as e:
sb_volume0.pvcCreate(pvcName=sb_volume0_name)
assert "cannot create PVC for standby volume" in str(e.value)
setting = client.by_id_setting(common.SETTING_BACKUP_TARGET)
with pytest.raises(Exception) as e:
client.update(setting, value="random.backup.target")
assert "cannot modify BackupTarget " \
"since there are existing standby volumes" in str(e.value)
with pytest.raises(Exception) as e:
sb_volume0.activate(frontend="wrong_frontend")
assert "invalid frontend" in str(e.value)
activate_standby_volume(client, sb_volume0_name)
sb_volume0 = client.by_id_volume(sb_volume0_name)
sb_volume0.attach(hostId=lht_host_id)
sb_volume0 = common.wait_for_volume_healthy(client, sb_volume0_name)
check_volume_data(sb_volume0, data0, False)
zero_string = b'\x00'.decode('utf-8')
_, backup1, _, data1 = create_backup(client, volume_name, {
'len': 2 * 1024,
'pos': 0,
'content': zero_string * 2 * 1024
})
# use this api to update field `last backup`
client.list_backupVolume()
check_volume_last_backup(client, sb_volume1_name, backup1['name'])
activate_standby_volume(client, sb_volume1_name)
sb_volume1 = client.by_id_volume(sb_volume1_name)
sb_volume1.attach(hostId=lht_host_id)
sb_volume1 = common.wait_for_volume_healthy(client, sb_volume1_name)
data0_modified = {
'len': data0['len'] - data1['len'],
'pos': data1['len'],
'content': data0['content'][data1['len']:],
}
check_volume_data(sb_volume1, data0_modified, False)
check_volume_data(sb_volume1, data1)
data2 = {'len': 1 * 1024 * 1024, 'pos': 0}
data2['content'] = common.generate_random_data(data2['len'])
_, backup2, _, data2 = create_backup(client, volume_name, data2)
client.list_backupVolume()
check_volume_last_backup(client, sb_volume2_name, backup2['name'])
activate_standby_volume(client, sb_volume2_name)
sb_volume2 = client.by_id_volume(sb_volume2_name)
sb_volume2.attach(hostId=lht_host_id)
sb_volume2 = common.wait_for_volume_healthy(client, sb_volume2_name)
check_volume_data(sb_volume2, data2)
# allocated this active volume to a pod
sb_volume2.detach()
sb_volume2 = common.wait_for_volume_detached(client, sb_volume2_name)
create_pv_for_volume(client, core_api, sb_volume2, sb_volume2_name)
create_pvc_for_volume(client, core_api, sb_volume2, sb_volume2_name)
sb_volume2_pod_name = "pod-" + sb_volume2_name
pod['metadata']['name'] = sb_volume2_pod_name
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': sb_volume2_name,
},
}]
create_and_wait_pod(core_api, pod)
sb_volume2 = client.by_id_volume(sb_volume2_name)
k_status = sb_volume2["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert k_status['pvName'] == sb_volume2_name
assert k_status['pvStatus'] == 'Bound'
assert len(workloads) == 1
for i in range(RETRY_COUNTS):
if workloads[0]['podStatus'] == 'Running':
break
time.sleep(RETRY_INTERVAL)
sb_volume2 = client.by_id_volume(sb_volume2_name)
k_status = sb_volume2["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert len(workloads) == 1
assert workloads[0]['podName'] == sb_volume2_pod_name
assert workloads[0]['podStatus'] == 'Running'
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == sb_volume2_name
assert not k_status['lastPVCRefAt']
assert not k_status['lastPodRefAt']
delete_and_wait_pod(core_api, sb_volume2_pod_name)
delete_and_wait_pvc(core_api, sb_volume2_name)
delete_and_wait_pv(core_api, sb_volume2_name)
# cleanup
std_volume.detach()
sb_volume0.detach()
sb_volume1.detach()
std_volume = common.wait_for_volume_detached(client, volume_name)
sb_volume0 = common.wait_for_volume_detached(client, sb_volume0_name)
sb_volume1 = common.wait_for_volume_detached(client, sb_volume1_name)
sb_volume2 = common.wait_for_volume_detached(client, sb_volume2_name)
bv.backupDelete(name=backup2["name"])
bv.backupDelete(name=backup1["name"])
bv.backupDelete(name=backup0["name"])
client.delete(std_volume)
client.delete(sb_volume0)
client.delete(sb_volume1)
client.delete(sb_volume2)
wait_for_volume_delete(client, volume_name)
wait_for_volume_delete(client, sb_volume0_name)
wait_for_volume_delete(client, sb_volume1_name)
wait_for_volume_delete(client, sb_volume2_name)
volumes = client.list_volume()
assert len(volumes) == 0
def test_kubernetes_status(client, core_api, storage_class, # NOQA
statefulset, csi_pv, pvc, pod): # NOQA
statefulset_name = 'kubernetes-status-test'
update_statefulset_manifests(statefulset, storage_class, statefulset_name)
storage_class['reclaimPolicy'] = 'Retain'
create_storage_class(storage_class)
create_and_wait_statefulset(statefulset)
pod_info = get_statefulset_pod_info(core_api, statefulset)
volume_info = [p['pv_name'] for p in pod_info]
extra_pod_name = 'extra-pod-using-' + volume_info[1]
pod['metadata']['name'] = extra_pod_name
p2 = core_api.read_namespaced_pod(name=pod_info[1]['pod_name'],
namespace='default')
pod['spec']['nodeName'] = p2.spec.node_name
pod['spec']['volumes'] = [{
'name': pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': pod_info[1]['pvc_name'],
},
}]
create_and_wait_pod(core_api, pod)
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert k_status['pvName'] == p['pv_name']
assert k_status['pvStatus'] == 'Bound'
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == p['pvc_name']
assert not k_status['lastPVCRefAt']
assert not k_status['lastPodRefAt']
if i == 0:
assert len(workloads) == 1
assert workloads[0]['podName'] == p['pod_name']
assert workloads[0]['workloadName'] == statefulset_name
assert workloads[0]['workloadType'] == 'StatefulSet'
for _ in range(RETRY_COUNTS):
if workloads[0]['podStatus'] == 'Running':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert workloads[0]['podStatus'] == 'Running'
if i == 1:
assert len(k_status['workloadsStatus']) == 2
if workloads[0]['podName'] == pod_info[i]['pod_name']:
assert workloads[1]['podName'] == extra_pod_name
assert workloads[0]['workloadName'] == statefulset_name
assert workloads[0]['workloadType'] == 'StatefulSet'
assert not workloads[1]['workloadName']
assert not workloads[1]['workloadType']
else:
assert workloads[1]['podName'] == pod_info[i]['pod_name']
assert workloads[0]['podName'] == extra_pod_name
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
assert workloads[1]['workloadName'] == statefulset_name
assert workloads[1]['workloadType'] == 'StatefulSet'
for _ in range(RETRY_COUNTS):
if workloads[0]['podStatus'] == 'Running' and \
workloads[1]['podStatus'] == 'Running':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert len(workloads) == 2
assert workloads[0]['podStatus'] == 'Running'
assert workloads[1]['podStatus'] == 'Running'
# the extra pod is still using the 2nd volume
delete_and_wait_statefulset_only(core_api, statefulset)
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert k_status['pvName'] == p['pv_name']
assert k_status['pvStatus'] == 'Bound'
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == p['pvc_name']
assert not k_status['lastPVCRefAt']
assert len(workloads) == 1
if i == 0:
assert workloads[0]['podName'] == p['pod_name']
assert workloads[0]['workloadName'] == statefulset_name
assert workloads[0]['workloadType'] == 'StatefulSet'
assert k_status['lastPodRefAt']
if i == 1:
assert workloads[0]['podName'] == extra_pod_name
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
assert not k_status['lastPodRefAt']
# deleted extra_pod, all volumes have no workload
delete_and_wait_pod(core_api, pod['metadata']['name'])
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert k_status['pvName'] == p['pv_name']
assert k_status['pvStatus'] == 'Bound'
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == p['pvc_name']
assert not k_status['lastPVCRefAt']
assert k_status['lastPodRefAt']
assert len(workloads) == 1
if i == 0:
assert workloads[0]['podName'] == p['pod_name']
assert workloads[0]['workloadName'] == statefulset_name
assert workloads[0]['workloadType'] == 'StatefulSet'
if i == 1:
assert workloads[0]['podName'] == extra_pod_name
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
# deleted pvc only.
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
delete_and_wait_pvc(core_api, p['pvc_name'])
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
for _ in range(RETRY_COUNTS):
if k_status['pvStatus'] == 'Released':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert k_status['pvName'] == p['pv_name']
assert k_status['pvStatus'] == 'Released'
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == p['pvc_name']
assert k_status['lastPVCRefAt']
assert k_status['lastPodRefAt']
assert len(workloads) == 1
if i == 0:
assert workloads[0]['podName'] == p['pod_name']
assert workloads[0]['workloadName'] == statefulset_name
assert workloads[0]['workloadType'] == 'StatefulSet'
if i == 1:
assert workloads[0]['podName'] == extra_pod_name
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
# deleted pv only.
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
delete_and_wait_pv(core_api, p['pv_name'])
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert k_status['pvName'] == ''
assert k_status['pvStatus'] == ''
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == p['pvc_name']
assert k_status['lastPVCRefAt']
assert k_status['lastPodRefAt']
assert len(workloads) == 1
if i == 0:
assert workloads[0]['podName'] == p['pod_name']
assert workloads[0]['workloadName'] == statefulset_name
assert workloads[0]['workloadType'] == 'StatefulSet'
if i == 1:
assert workloads[0]['podName'] == extra_pod_name
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
# reuse that volume
for p, volume_name in zip(pod_info, volume_info):
p['pod_name'] = p['pod_name'].replace('kubernetes-status-test',
'kubernetes-status-test-reuse')
p['pvc_name'] = p['pvc_name'].replace('kubernetes-status-test',
'kubernetes-status-test-reuse')
p['pv_name'] = p['pvc_name']
csi_pv['metadata']['name'] = p['pv_name']
csi_pv['spec']['csi']['volumeHandle'] = volume_name
core_api.create_persistent_volume(csi_pv)
pvc['metadata']['name'] = p['pvc_name']
pvc['spec']['volumeName'] = p['pv_name']
core_api.create_namespaced_persistent_volume_claim(
body=pvc, namespace='default')
pod['metadata']['name'] = p['pod_name']
pod['spec']['volumes'] = [{
'name': pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': p['pvc_name'],
},
}]
create_and_wait_pod(core_api, pod)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert len(workloads) == 1
assert k_status['pvName'] == p['pv_name']
for _ in range(RETRY_COUNTS):
if k_status['pvStatus'] == 'Bound':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert len(workloads) == 1
assert k_status['pvStatus'] == 'Bound'
for _ in range(RETRY_COUNTS):
if workloads[0]['podStatus'] == 'Running':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert len(workloads) == 1
assert workloads[0]['podStatus'] == 'Running'
assert workloads[0]['podName'] == p['pod_name']
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == p['pvc_name']
assert not k_status['lastPVCRefAt']
assert not k_status['lastPodRefAt']
delete_and_wait_pod(core_api, p['pod_name'])
# Since persistentVolumeReclaimPolicy of csi_pv is `Delete`,
# we don't need to delete bounded pv manually
delete_and_wait_pvc(core_api, p['pvc_name'])
wait_delete_pv(core_api, p['pv_name'])
def test_pvc_creation(client, core_api, pod): # NOQA
volume_name = "test-pvc-creation"
client.create_volume(name=volume_name, size=SIZE,
numberOfReplicas=2)
volume = wait_for_volume_detached(client, volume_name)
pv_name = "pv-" + volume_name
pvc_name = "pvc-" + volume_name
pod_name = "pod-" + volume_name
# try to create pvc without pv for the volume
with pytest.raises(Exception) as e:
volume.pvcCreate(namespace="default", pvcName=pvc_name)
assert "connot find existing PV for volume" in str(e.value)
volume.pvCreate(pvName=pv_name)
for i in range(RETRY_COUNTS):
if check_pv_existence(core_api, pv_name):
break
time.sleep(RETRY_INTERVAL)
assert check_pv_existence(core_api, pv_name)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
for i in range(RETRY_COUNTS):
if k_status['pvName'] and k_status['pvStatus'] == 'Available':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
assert k_status['pvName'] == pv_name
assert k_status['pvStatus'] == 'Available'
assert not k_status['namespace']
assert not k_status['pvcName']
assert not k_status['workloadsStatus']
assert not k_status['lastPVCRefAt']
assert not k_status['lastPodRefAt']
volume.pvcCreate(namespace="default", pvcName=pvc_name)
for i in range(RETRY_COUNTS):
if check_pvc_existence(core_api, pvc_name):
break
time.sleep(RETRY_INTERVAL)
assert check_pvc_existence(core_api, pvc_name)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
for i in range(RETRY_COUNTS):
if k_status['pvcName'] and k_status['namespace']:
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
assert k_status['pvName'] == pv_name
assert k_status['pvStatus'] == 'Bound'
assert k_status['namespace'] == "default"
assert k_status['pvcName'] == pvc_name
assert not k_status['workloadsStatus']
assert not k_status['lastPVCRefAt']
assert not k_status['lastPodRefAt']
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [{
'name': pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': pvc_name,
},
}]
create_and_wait_pod(core_api, pod)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert k_status['pvName'] == pv_name
assert k_status['pvStatus'] == 'Bound'
assert len(workloads) == 1
for i in range(RETRY_COUNTS):
if workloads[0]['podStatus'] == 'Running':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume["kubernetesStatus"]
workloads = k_status['workloadsStatus']
assert len(workloads) == 1
assert workloads[0]['podName'] == pod_name
assert workloads[0]['podStatus'] == 'Running'
assert not workloads[0]['workloadName']
assert not workloads[0]['workloadType']
assert k_status['namespace'] == 'default'
assert k_status['pvcName'] == pvc_name
assert not k_status['lastPVCRefAt']
assert not k_status['lastPodRefAt']
delete_and_wait_pod(core_api, pod_name)
delete_and_wait_pvc(core_api, pvc_name)
wait_delete_pv(core_api, pv_name)
def test_engine_live_upgrade_with_intensive_data_writing(
client, core_api, volume_name, pod_make): # NOQA
"""
Test engine live upgrade with intensive data writing
1. Deploy a compatible new engine image
2. Create a volume(with the old default engine image) with /PV/PVC/Pod
and wait for pod to be deployed.
3. Write data to a tmp file in the pod and get the md5sum
4. Upgrade the volume to the new engine image without waiting.
5. Keep copying data from the tmp file to the volume
during the live upgrade.
6. Wait until the upgrade completed, verify the volume engine image changed
7. Wait for new replica mode update then check the engine status.
8. Verify all engine and replicas' engine image changed
9. Verify the reference count of the new engine image changed
10. Check the existing data.
Then write new data to the upgraded volume and get the md5sum.
11. Delete the pod and wait for the volume detached.
Then check engine and replicas's engine image again.
12. Recreate the pod.
13. Check if the attached volume is state `healthy`
rather than `degraded`.
14. Check the data.
"""
default_img = common.get_default_engine_image(client)
default_img_name = default_img.name
default_img = wait_for_engine_image_ref_count(client, default_img_name, 0)
cli_v = default_img.cliAPIVersion
cli_minv = default_img.cliAPIMinVersion
ctl_v = default_img.controllerAPIVersion
ctl_minv = default_img.controllerAPIMinVersion
data_v = default_img.dataFormatVersion
data_minv = default_img.dataFormatMinVersion
engine_upgrade_image = common.get_upgrade_test_image(
cli_v, cli_minv, ctl_v, ctl_minv, data_v, data_minv)
new_img = client.create_engine_image(image=engine_upgrade_image)
new_img_name = new_img.name
ei_status_value = get_engine_image_status_value(client, new_img_name)
new_img = wait_for_engine_image_state(client, new_img_name,
ei_status_value)
assert new_img.refCount == 0
assert new_img.noRefSince != ""
default_img = common.get_default_engine_image(client)
default_img_name = default_img.name
pod_name = volume_name + "-pod"
pv_name = volume_name + "-pv"
pvc_name = volume_name + "-pvc"
pod = pod_make(name=pod_name)
volume = create_and_check_volume(client,
volume_name,
num_of_replicas=3,
size=str(1 * Gi))
original_engine_image = volume.engineImage
assert original_engine_image != engine_upgrade_image
create_pv_for_volume(client, core_api, volume, pv_name)
create_pvc_for_volume(client, core_api, volume, pvc_name)
pod['spec']['volumes'] = [create_pvc_spec(pvc_name)]
create_and_wait_pod(core_api, pod)
volume = client.by_id_volume(volume_name)
assert volume.engineImage == original_engine_image
assert volume.currentImage == original_engine_image
engine = get_volume_engine(volume)
assert engine.engineImage == original_engine_image
assert engine.currentImage == original_engine_image
for replica in volume.replicas:
assert replica.engineImage == original_engine_image
assert replica.currentImage == original_engine_image
data_path0 = "/tmp/test"
data_path1 = "/data/test1"
write_pod_volume_random_data(core_api, pod_name, data_path0,
RANDOM_DATA_SIZE_LARGE)
original_md5sum1 = get_pod_data_md5sum(core_api, pod_name, data_path0)
volume.engineUpgrade(image=engine_upgrade_image)
# Keep writing data to the volume during the live upgrade
copy_pod_volume_data(core_api, pod_name, data_path0, data_path1)
# Wait for live upgrade complete
wait_for_volume_current_image(client, volume_name, engine_upgrade_image)
volume = wait_for_volume_replicas_mode(client, volume_name, "RW")
engine = get_volume_engine(volume)
assert engine.engineImage == engine_upgrade_image
check_volume_endpoint(volume)
wait_for_engine_image_ref_count(client, default_img_name, 0)
wait_for_engine_image_ref_count(client, new_img_name, 1)
volume_file_md5sum1 = get_pod_data_md5sum(core_api, pod_name, data_path1)
assert volume_file_md5sum1 == original_md5sum1
data_path2 = "/data/test2"
write_pod_volume_random_data(core_api, pod_name, data_path2,
RANDOM_DATA_SIZE_SMALL)
original_md5sum2 = get_pod_data_md5sum(core_api, pod_name, data_path2)
delete_and_wait_pod(core_api, pod_name)
volume = wait_for_volume_detached(client, volume_name)
assert len(volume.replicas) == 3
assert volume.engineImage == engine_upgrade_image
engine = get_volume_engine(volume)
assert engine.engineImage == engine_upgrade_image
for replica in volume.replicas:
assert replica.engineImage == engine_upgrade_image
create_and_wait_pod(core_api, pod)
common.wait_for_volume_healthy(client, volume_name)
volume_file_md5sum1 = get_pod_data_md5sum(core_api, pod_name, data_path1)
assert volume_file_md5sum1 == original_md5sum1
volume_file_md5sum2 = get_pod_data_md5sum(core_api, pod_name, data_path2)
assert volume_file_md5sum2 == original_md5sum2
def test_rwx_multi_statefulset_with_same_pvc(core_api, pvc, statefulset,
pod): # NOQA
"""
Test writing of data into a volume from multiple pods using same PVC
1. Create a volume with 'accessMode' rwx.
2. Create a PV and a PVC with access mode 'readwritemany' and attach to the
volume.
3. Deploy a StatefulSet of 2 pods with the existing PVC above created.
4. Wait for both pods to come up.
5. Create a pod with the existing PVC above created.
6. Wait for StatefulSet to come up healthy.
7. Write data all three pods and compute md5sum.
8. Check the data md5sum in the share manager pod.
"""
pvc_name = 'pvc-multi-pods-test'
statefulset_name = 'statefulset-rwx-same-pvc-test'
pod_name = 'pod-rwx-same-pvc-test'
pvc['metadata']['name'] = pvc_name
pvc['spec']['storageClassName'] = 'longhorn'
pvc['spec']['accessModes'] = ['ReadWriteMany']
core_api.create_namespaced_persistent_volume_claim(body=pvc,
namespace='default')
statefulset['metadata']['name'] = \
statefulset['spec']['selector']['matchLabels']['app'] = \
statefulset['spec']['serviceName'] = \
statefulset['spec']['template']['metadata']['labels']['app'] = \
statefulset_name
statefulset['spec']['template']['spec']['volumes'] = \
[create_pvc_spec(pvc_name)]
del statefulset['spec']['volumeClaimTemplates']
create_and_wait_statefulset(statefulset)
pv_name = get_volume_name(core_api, pvc_name)
share_manager_name = 'share-manager-' + pv_name
test_data = generate_random_data(VOLUME_RWTEST_SIZE)
write_pod_volume_data(core_api,
statefulset_name + '-0',
test_data,
filename='test1')
assert test_data == read_volume_data(core_api,
statefulset_name + '-1',
filename='test1')
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [create_pvc_spec(pvc_name)]
create_and_wait_pod(core_api, pod)
assert test_data == read_volume_data(core_api, pod_name, filename='test1')
test_data_2 = generate_random_data(VOLUME_RWTEST_SIZE)
write_pod_volume_data(core_api, pod_name, test_data_2, filename='test2')
command1 = 'cat /export' + '/' + pv_name + '/' + 'test1'
command2 = 'cat /export' + '/' + pv_name + '/' + 'test2'
assert test_data == exec_command_in_pod(core_api, command1,
share_manager_name,
LONGHORN_NAMESPACE)
assert test_data_2 == exec_command_in_pod(core_api, command2,
share_manager_name,
LONGHORN_NAMESPACE)
def test_csi_block_volume(client, core_api, storage_class, pvc, pod_manifest): # NOQA
"""
Test CSI feature: raw block volume
1. Create a PVC with `volumeMode = Block`
2. Create a pod using the PVC to dynamic provision a volume
3. Verify the pod creation
4. Generate `test_data` and write to the block volume directly in the pod
5. Read the data back for validation
6. Delete the pod and create `pod2` to use the same volume
7. Validate the data in `pod2` is consistent with `test_data`
"""
pod_name = 'csi-block-volume-test'
pvc_name = pod_name + "-pvc"
device_path = "/dev/longhorn/longhorn-test-blk"
storage_class['reclaimPolicy'] = 'Retain'
pvc['metadata']['name'] = pvc_name
pvc['spec']['volumeMode'] = 'Block'
pvc['spec']['storageClassName'] = storage_class['metadata']['name']
pvc['spec']['resources'] = {
'requests': {
'storage': size_to_string(1 * Gi)
}
}
pod_manifest['metadata']['name'] = pod_name
pod_manifest['spec']['volumes'] = [{
'name': 'longhorn-blk',
'persistentVolumeClaim': {
'claimName': pvc_name,
},
}]
pod_manifest['spec']['containers'][0]['volumeMounts'] = []
pod_manifest['spec']['containers'][0]['volumeDevices'] = [
{'name': 'longhorn-blk', 'devicePath': device_path}
]
create_storage_class(storage_class)
create_pvc(pvc)
pv_name = wait_and_get_pv_for_pvc(core_api, pvc_name).metadata.name
create_and_wait_pod(core_api, pod_manifest)
test_data = generate_random_data(VOLUME_RWTEST_SIZE)
test_offset = random.randint(0, VOLUME_RWTEST_SIZE)
write_pod_block_volume_data(
core_api, pod_name, test_data, test_offset, device_path)
returned_data = read_pod_block_volume_data(
core_api, pod_name, len(test_data), test_offset, device_path
)
assert test_data == returned_data
md5_sum = get_pod_data_md5sum(
core_api, pod_name, device_path)
delete_and_wait_pod(core_api, pod_name)
common.wait_for_volume_detached(client, pv_name)
pod_name_2 = 'csi-block-volume-test-reuse'
pod_manifest['metadata']['name'] = pod_name_2
create_and_wait_pod(core_api, pod_manifest)
returned_data = read_pod_block_volume_data(
core_api, pod_name_2, len(test_data), test_offset, device_path
)
assert test_data == returned_data
md5_sum_2 = get_pod_data_md5sum(
core_api, pod_name_2, device_path)
assert md5_sum == md5_sum_2
delete_and_wait_pod(core_api, pod_name_2)
delete_and_wait_pvc(core_api, pvc_name)
delete_and_wait_pv(core_api, pv_name)
def test_backup_kubernetes_status(client, core_api, pod): # NOQA
"""
Test that Backups have KubernetesStatus stored properly when there is an
associated PersistentVolumeClaim and Pod.
"""
host_id = get_self_host_id()
static_sc_name = "longhorn-static-test"
setting = client.by_id_setting(SETTING_DEFAULT_LONGHORN_STATIC_SC)
setting = client.update(setting, value=static_sc_name)
assert setting["value"] == static_sc_name
volume_name = "test-backup-kubernetes-status-pod"
client.create_volume(name=volume_name, size=SIZE, numberOfReplicas=2)
volume = wait_for_volume_detached(client, volume_name)
pod_name = "pod-" + volume_name
pv_name = "pv-" + volume_name
pvc_name = "pvc-" + volume_name
create_pv_for_volume(client, core_api, volume, pv_name)
create_pvc_for_volume(client, core_api, volume, pvc_name)
ret = core_api.list_namespaced_persistent_volume_claim(namespace='default')
pvc_found = False
for item in ret.items:
if item.metadata.name == pvc_name:
pvc_found = item
break
assert pvc_found
assert pvc_found.spec.storage_class_name == static_sc_name
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': pvc_name,
},
}]
create_and_wait_pod(core_api, pod)
ks = {
'lastPodRefAt':
'',
'lastPVCRefAt':
'',
'namespace':
'default',
'pvcName':
pvc_name,
'pvName':
pv_name,
'pvStatus':
'Bound',
'workloadsStatus': [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
}
wait_volume_kubernetes_status(client, volume_name, ks)
volume = wait_for_volume_healthy(client, volume_name)
# Create Backup manually instead of calling create_backup since Kubernetes
# is not guaranteed to mount our Volume to the test host.
snap = volume.snapshotCreate()
volume.snapshotBackup(name=snap["name"])
bv, b = find_backup(client, volume_name, snap["name"])
new_b = bv.backupGet(name=b["name"])
status = loads(new_b["labels"].get(KUBERNETES_STATUS_LABEL))
assert status == ks
restore_name = generate_volume_name()
client.create_volume(name=restore_name,
size=SIZE,
numberOfReplicas=2,
fromBackup=b["url"])
wait_for_volume_restoration_completed(client, restore_name)
wait_for_volume_detached(client, restore_name)
snapshot_created = b["snapshotCreated"]
ks = {
'lastPodRefAt':
b["snapshotCreated"],
'lastPVCRefAt':
b["snapshotCreated"],
'namespace':
'default',
'pvcName':
pvc_name,
# Restoration should not apply PersistentVolume data.
'pvName':
'',
'pvStatus':
'',
'workloadsStatus': [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
}
wait_volume_kubernetes_status(client, restore_name, ks)
restore = client.by_id_volume(restore_name)
# We need to compare LastPodRefAt and LastPVCRefAt manually since
# wait_volume_kubernetes_status only checks for empty or non-empty state.
assert restore["kubernetesStatus"]["lastPodRefAt"] == ks["lastPodRefAt"]
assert restore["kubernetesStatus"]["lastPVCRefAt"] == ks["lastPVCRefAt"]
bv.backupDelete(name=b["name"])
client.delete(restore)
wait_for_volume_delete(client, restore_name)
delete_and_wait_pod(core_api, pod_name)
delete_and_wait_pvc(core_api, pvc_name)
delete_and_wait_pv(core_api, pv_name)
# With the Pod, PVC, and PV deleted, the Volume should have both Ref
# fields set. Check that a new Backup and Restore will use this instead of
# manually populating the Ref fields.
ks = {
'lastPodRefAt':
'NOT NULL',
'lastPVCRefAt':
'NOT NULL',
'namespace':
'default',
'pvcName':
pvc_name,
'pvName':
'',
'pvStatus':
'',
'workloadsStatus': [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
}
wait_volume_kubernetes_status(client, volume_name, ks)
volume = wait_for_volume_detached(client, volume_name)
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
snap = volume.snapshotCreate()
volume.snapshotBackup(name=snap["name"])
bv, b = find_backup(client, volume_name, snap["name"])
new_b = bv.backupGet(name=b["name"])
status = loads(new_b["labels"].get(KUBERNETES_STATUS_LABEL))
# Check each field manually, we have no idea what the LastPodRefAt or the
# LastPVCRefAt will be. We just know it shouldn't be SnapshotCreated.
assert status["lastPodRefAt"] != snapshot_created
assert status["lastPVCRefAt"] != snapshot_created
assert status["namespace"] == "default"
assert status["pvcName"] == pvc_name
assert status["pvName"] == ""
assert status["pvStatus"] == ""
assert status["workloadsStatus"] == [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
restore_name = generate_volume_name()
client.create_volume(name=restore_name,
size=SIZE,
numberOfReplicas=2,
fromBackup=b["url"])
wait_for_volume_restoration_completed(client, restore_name)
wait_for_volume_detached(client, restore_name)
ks = {
'lastPodRefAt':
status["lastPodRefAt"],
'lastPVCRefAt':
status["lastPVCRefAt"],
'namespace':
'default',
'pvcName':
pvc_name,
'pvName':
'',
'pvStatus':
'',
'workloadsStatus': [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
}
wait_volume_kubernetes_status(client, restore_name, ks)
restore = client.by_id_volume(restore_name)
assert restore["kubernetesStatus"]["lastPodRefAt"] == ks["lastPodRefAt"]
assert restore["kubernetesStatus"]["lastPVCRefAt"] == ks["lastPVCRefAt"]
bv.backupDelete(name=b["name"])
client.delete(restore)
cleanup_volume(client, volume)
def test_csi_expansion_with_replica_failure(client, core_api, storage_class, pvc, pod_manifest): # NOQA
"""
Test expansion success but with one replica expansion failure
1. Create a new `storage_class` with `allowVolumeExpansion` set
2. Create PVC and Pod with dynamic provisioned volume from the StorageClass
3. Create an empty directory with expansion snapshot tmp meta file path
for one replica so that the replica expansion will fail
4. Generate `test_data` and write to the pod
5. Delete the pod and wait for volume detachment
6. Update pvc.spec.resources to expand the volume
7. Check expansion result using Longhorn API. There will be expansion error
caused by the failed replica but overall the expansion should succeed.
8. Create a new pod and
check if the volume will rebuild the failed replica
9. Validate the volume content, then check if data writing looks fine
"""
create_storage_class(storage_class)
pod_name = 'csi-expansion-with-replica-failure-test'
pvc_name = pod_name + "-pvc"
pvc['metadata']['name'] = pvc_name
pvc['spec']['storageClassName'] = storage_class['metadata']['name']
create_pvc(pvc)
pod_manifest['metadata']['name'] = pod_name
pod_manifest['spec']['volumes'] = [{
'name':
pod_manifest['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {'claimName': pvc_name},
}]
create_and_wait_pod(core_api, pod_manifest)
expand_size = str(EXPANDED_VOLUME_SIZE*Gi)
pv = wait_and_get_pv_for_pvc(core_api, pvc_name)
assert pv.status.phase == "Bound"
volume_name = pv.spec.csi.volume_handle
volume = client.by_id_volume(volume_name)
failed_replica = volume.replicas[0]
fail_replica_expansion(client, core_api,
volume_name, expand_size, [failed_replica])
test_data = generate_random_data(VOLUME_RWTEST_SIZE)
write_pod_volume_data(core_api, pod_name, test_data)
delete_and_wait_pod(core_api, pod_name)
wait_for_volume_detached(client, volume_name)
# There will be replica expansion error info
# but the expansion should succeed.
pvc['spec']['resources'] = {
'requests': {
'storage': size_to_string(EXPANDED_VOLUME_SIZE*Gi)
}
}
expand_and_wait_for_pvc(core_api, pvc)
wait_for_expansion_failure(client, volume_name)
wait_for_volume_expansion(client, volume_name)
volume = client.by_id_volume(volume_name)
assert volume.state == "detached"
assert volume.size == expand_size
for r in volume.replicas:
if r.name == failed_replica.name:
assert r.failedAt != ""
else:
assert r.failedAt == ""
# Check if the replica will be rebuilded
# and if the volume still works fine.
create_and_wait_pod(core_api, pod_manifest)
volume = wait_for_volume_healthy(client, volume_name)
for r in volume.replicas:
if r.name == failed_replica.name:
assert r.mode == ""
else:
assert r.mode == "RW"
resp = read_volume_data(core_api, pod_name)
assert resp == test_data
test_data = generate_random_data(VOLUME_RWTEST_SIZE)
write_pod_volume_data(core_api, pod_name, test_data)
resp = read_volume_data(core_api, pod_name)
assert resp == test_data
def test_cloning_basic(client, core_api, pvc, pod, clone_pvc, clone_pod, storage_class_name='longhorn'): # NOQA
"""
1. Create a PVC:
```yaml
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: source-pvc
spec:
storageClassName: longhorn
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 3Gi
```
2. Specify the `source-pvc` in a pod yaml and start the pod
3. Wait for the pod to be running, write some data to the mount
path of the volume
4. Clone a volume by creating the PVC:
```yaml
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: cloned-pvc
spec:
storageClassName: longhorn
dataSource:
name: source-pvc
kind: PersistentVolumeClaim
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 3Gi
```
5. Wait for the `CloneStatus.State` in `cloned-pvc` to be `completed`
6. Clone volume should get detached after cloning completion, wait for it.
7. Specify the `cloned-pvc` in a cloned pod yaml and deploy the cloned pod
8. In 3-min retry loop, wait for the cloned pod to be running
9. Verify the data in `cloned-pvc` is the same as in `source-pvc`
10. In 2-min retry loop, verify the volume of the `clone-pvc` eventually
becomes healthy
"""
# Step-1
source_pvc_name = 'source-pvc' + generate_random_suffix()
pvc['metadata']['name'] = source_pvc_name
pvc['spec']['storageClassName'] = storage_class_name
core_api.create_namespaced_persistent_volume_claim(
body=pvc, namespace='default')
wait_for_pvc_phase(core_api, source_pvc_name, "Bound")
# Step-2
pod_name = 'source-pod' + generate_random_suffix()
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [create_pvc_spec(source_pvc_name)]
create_and_wait_pod(core_api, pod)
# Step-3
write_pod_volume_random_data(core_api, pod_name,
'/data/test', DATA_SIZE_IN_MB_2)
source_data = get_pod_data_md5sum(core_api, pod_name, '/data/test')
# Step-4
clone_pvc_name = 'clone-pvc' + generate_random_suffix()
clone_pvc['metadata']['name'] = clone_pvc_name
clone_pvc['spec']['storageClassName'] = storage_class_name
clone_pvc['spec']['dataSource'] = {
'name': source_pvc_name,
'kind': 'PersistentVolumeClaim'
}
core_api.create_namespaced_persistent_volume_claim(
body=clone_pvc, namespace='default')
wait_for_pvc_phase(core_api, clone_pvc_name, "Bound")
# Step-5
clone_volume_name = get_volume_name(core_api, clone_pvc_name)
wait_for_volume_clone_status(client, clone_volume_name, VOLUME_FIELD_STATE,
VOLUME_FIELD_CLONE_COMPLETED)
# Step-6
wait_for_volume_detached(client, clone_volume_name)
# Step-7,8
clone_pod_name = 'clone-pod' + generate_random_suffix()
clone_pod['metadata']['name'] = clone_pod_name
clone_pod['spec']['volumes'] = [create_pvc_spec(clone_pvc_name)]
create_and_wait_pod(core_api, clone_pod)
clone_data = get_pod_data_md5sum(core_api, clone_pod_name, '/data/test')
# Step-9
assert source_data == clone_data
# Step-10
wait_for_volume_healthy(client, clone_volume_name)
def test_pvc_creation_with_default_sc_set(client, core_api, storage_class,
pod): # NOQA
"""
Test creating PVC with default StorageClass set
The target is to make sure the newly create PV/PVC won't use default
StorageClass, and if there is no default StorageClass, PV/PVC can still be
created.
1. Create a StorageClass and set it to be the default StorageClass
2. Update static StorageClass to `longhorn-static-test`
3. Create volume then PV/PVC.
4. Make sure the newly created PV/PVC using StorageClass
`longhorn-static-test`
5. Create pod with PVC.
6. Verify volume's Kubernetes Status
7. Remove PVC and Pod.
8. Verify volume's Kubernetes Status only contains current PV and history
9. Wait for volume to detach (since pod is deleted)
10. Reuse the volume on a new pod. Wait for the pod to start
11. Verify volume's Kubernetes Status reflect the new pod.
12. Delete PV/PVC/Pod.
13. Verify volume's Kubernetes Status only contains history
14. Delete the default StorageClass.
15. Create PV/PVC for the volume.
16. Make sure the PV's StorageClass is static StorageClass
"""
# set default storage class
storage_class['metadata']['annotations'] = \
{"storageclass.kubernetes.io/is-default-class": "true"}
create_storage_class(storage_class)
static_sc_name = "longhorn-static-test"
setting = client.by_id_setting(SETTING_DEFAULT_LONGHORN_STATIC_SC)
setting = client.update(setting, value=static_sc_name)
assert setting.value == static_sc_name
volume_name = "test-pvc-creation-with-sc" # NOQA
pod_name = "pod-" + volume_name
client.create_volume(name=volume_name, size=SIZE, numberOfReplicas=2)
volume = wait_for_volume_detached(client, volume_name)
pv_name = "pv-" + volume_name
pvc_name = "pvc-" + volume_name
pvc_name_extra = "pvc-" + volume_name + "-extra"
create_pv_for_volume(client, core_api, volume, pv_name)
create_pvc_for_volume(client, core_api, volume, pvc_name)
ret = core_api.list_namespaced_persistent_volume_claim(namespace='default')
for item in ret.items:
if item.metadata.name == pvc_name:
pvc_found = item
break
assert pvc_found
assert pvc_found.spec.storage_class_name == static_sc_name
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': pvc_name,
},
}]
create_and_wait_pod(core_api, pod)
ks = {
'pvName':
pv_name,
'pvStatus':
'Bound',
'namespace':
'default',
'pvcName':
pvc_name,
'lastPVCRefAt':
'',
'lastPodRefAt':
'',
'workloadsStatus': [
{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': '',
},
],
}
wait_volume_kubernetes_status(client, volume_name, ks)
delete_and_wait_pod(core_api, pod_name)
delete_and_wait_pvc(core_api, pvc_name)
ks = {
'pvName': pv_name,
'pvStatus': 'Released',
'namespace': 'default',
'pvcName': pvc_name,
'lastPVCRefAt': 'not empty',
'lastPodRefAt': 'not empty',
}
wait_volume_kubernetes_status(client, volume_name, ks)
# try to reuse the pv
volume = wait_for_volume_detached(client, volume_name)
create_pvc_for_volume(client, core_api, volume, pvc_name_extra)
pod['spec']['volumes'][0]['persistentVolumeClaim']['claimName'] = \
pvc_name_extra
create_and_wait_pod(core_api, pod)
ks = {
'pvName':
pv_name,
'pvStatus':
'Bound',
'namespace':
'default',
'pvcName':
pvc_name_extra,
'lastPVCRefAt':
'',
'lastPodRefAt':
'',
'workloadsStatus': [
{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': '',
},
],
}
wait_volume_kubernetes_status(client, volume_name, ks)
delete_and_wait_pod(core_api, pod_name)
delete_and_wait_pvc(core_api, pvc_name_extra)
delete_and_wait_pv(core_api, pv_name)
ks = {
'pvName': '',
'pvStatus': '',
'namespace': 'default',
'pvcName': pvc_name_extra,
'lastPVCRefAt': 'not empty',
'lastPodRefAt': 'not empty',
}
wait_volume_kubernetes_status(client, volume_name, ks)
# without default storage class
delete_storage_class(storage_class['metadata']['name'])
create_pv_for_volume(client, core_api, volume, pv_name)
create_pvc_for_volume(client, core_api, volume, pvc_name)
ret = core_api.list_namespaced_persistent_volume_claim(namespace='default')
for item in ret.items:
if item.metadata.name == pvc_name:
pvc2 = item
break
assert pvc2
assert pvc2.spec.storage_class_name == static_sc_name
delete_and_wait_pvc(core_api, pvc_name)
delete_and_wait_pv(core_api, pv_name)
def test_csi_minimal_volume_size(
client, core_api, csi_pv, pvc, pod_make): # NOQA
"""
Test CSI Minimal Volume Size
1. Create a PVC requesting size 5MiB. Check the PVC requested size is
5MiB and capacity size get is 10MiB.
2. Remove the PVC.
3. Create a PVC requesting size 10MiB. Check the PVC requested size and
capacity size get are both 10MiB.
4. Create a pod to use this PVC.
5. Write some data to the volume and read it back to compare.
"""
vol_name = generate_volume_name()
create_and_check_volume(client, vol_name, size=str(100*Mi))
low_storage = str(5*Mi)
min_storage = str(10*Mi)
pv_name = vol_name + "-pv"
csi_pv['metadata']['name'] = pv_name
csi_pv['spec']['csi']['volumeHandle'] = vol_name
csi_pv['spec']['capacity']['storage'] = min_storage
core_api.create_persistent_volume(csi_pv)
pvc_name = vol_name + "-pvc"
pvc['metadata']['name'] = pvc_name
pvc['spec']['volumeName'] = pv_name
pvc['spec']['resources']['requests']['storage'] = low_storage
pvc['spec']['storageClassName'] = ''
core_api.create_namespaced_persistent_volume_claim(body=pvc,
namespace='default')
claim = common.wait_for_pvc_phase(core_api, pvc_name, "Bound")
assert claim.spec.resources.requests['storage'] == low_storage
assert claim.status.capacity['storage'] == min_storage
common.delete_and_wait_pvc(core_api, pvc_name)
common.delete_and_wait_pv(core_api, pv_name)
wait_for_volume_detached(client, vol_name)
core_api.create_persistent_volume(csi_pv)
pvc['spec']['resources']['requests']['storage'] = min_storage
core_api.create_namespaced_persistent_volume_claim(body=pvc,
namespace='default')
claim = common.wait_for_pvc_phase(core_api, pvc_name, "Bound")
assert claim.spec.resources.requests['storage'] == min_storage
assert claim.status.capacity['storage'] == min_storage
pod_name = vol_name + '-pod'
pod = pod_make(name=pod_name)
pod['spec']['volumes'] = [create_pvc_spec(pvc_name)]
create_and_wait_pod(core_api, pod)
test_data = "longhorn-integration-test"
test_file = "test"
write_pod_volume_data(core_api, pod_name, test_data, test_file)
read_data = read_volume_data(core_api, pod_name, test_file)
assert read_data == test_data
def test_pvc_creation_with_default_sc_set(client, core_api, storage_class,
pod): # NOQA
# set default storage class
storage_class['metadata']['annotations'] = \
{"storageclass.kubernetes.io/is-default-class": "true"}
create_storage_class(storage_class)
static_sc_name = "longhorn-static-test"
setting = client.by_id_setting(SETTING_DEFAULT_LONGHORN_STATIC_SC)
setting = client.update(setting, value=static_sc_name)
assert setting["value"] == static_sc_name
volume_name = "test-pvc-creation-with-sc"
pod_name = "pod-" + volume_name
client.create_volume(name=volume_name, size=SIZE, numberOfReplicas=2)
volume = wait_for_volume_detached(client, volume_name)
pv_name = "pv-" + volume_name
pvc_name = "pvc-" + volume_name
pvc_name_extra = "pvc-" + volume_name + "-extra"
create_pv_for_volume(client, core_api, volume, pv_name)
create_pvc_for_volume(client, core_api, volume, pvc_name)
ret = core_api.list_namespaced_persistent_volume_claim(namespace='default')
for item in ret.items:
if item.metadata.name == pvc_name:
pvc_found = item
break
assert pvc_found
assert pvc_found.spec.storage_class_name == static_sc_name
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': pvc_name,
},
}]
create_and_wait_pod(core_api, pod)
ks = {
'pvName':
pv_name,
'pvStatus':
'Bound',
'namespace':
'default',
'pvcName':
pvc_name,
'lastPVCRefAt':
'',
'lastPodRefAt':
'',
'workloadsStatus': [
{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': '',
},
],
}
wait_volume_kubernetes_status(client, volume_name, ks)
delete_and_wait_pod(core_api, pod_name)
delete_and_wait_pvc(core_api, pvc_name)
# try to reuse the pv
volume = wait_for_volume_detached(client, volume_name)
create_pvc_for_volume(client, core_api, volume, pvc_name_extra)
pod['spec']['volumes'][0]['persistentVolumeClaim']['claimName'] = \
pvc_name_extra
create_and_wait_pod(core_api, pod)
ks['pvcName'] = pvc_name_extra
wait_volume_kubernetes_status(client, volume_name, ks)
delete_and_wait_pod(core_api, pod_name)
delete_and_wait_pvc(core_api, pvc_name_extra)
delete_and_wait_pv(core_api, pv_name)
# without default storage class
delete_storage_class(storage_class['metadata']['name'])
create_pv_for_volume(client, core_api, volume, pv_name)
create_pvc_for_volume(client, core_api, volume, pvc_name)
ret = core_api.list_namespaced_persistent_volume_claim(namespace='default')
for item in ret.items:
if item.metadata.name == pvc_name:
pvc2 = item
break
assert pvc2
assert pvc2.spec.storage_class_name == static_sc_name
delete_and_wait_pvc(core_api, pvc_name)
delete_and_wait_pv(core_api, pv_name)
def test_csi_offline_expansion(client, core_api, storage_class, pvc, pod_manifest): # NOQA
"""
Test CSI feature: offline expansion
1. Create a new `storage_class` with `allowVolumeExpansion` set
2. Create PVC and Pod with dynamic provisioned volume from the StorageClass
3. Generate `test_data` and write to the pod
4. Delete the pod
5. Update pvc.spec.resources to expand the volume
6. Verify the volume expansion done using Longhorn API
7. Create a new pod and validate the volume content
"""
create_storage_class(storage_class)
pod_name = 'csi-offline-expand-volume-test'
pvc_name = pod_name + "-pvc"
pvc['metadata']['name'] = pvc_name
pvc['spec']['storageClassName'] = storage_class['metadata']['name']
create_pvc(pvc)
pod_manifest['metadata']['name'] = pod_name
pod_manifest['spec']['volumes'] = [{
'name':
pod_manifest['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {'claimName': pvc_name},
}]
create_and_wait_pod(core_api, pod_manifest)
test_data = generate_random_data(VOLUME_RWTEST_SIZE)
write_pod_volume_data(core_api, pod_name, test_data)
delete_and_wait_pod(core_api, pod_name)
pv = wait_and_get_pv_for_pvc(core_api, pvc_name)
assert pv.status.phase == "Bound"
volume_name = pv.spec.csi.volume_handle
wait_for_volume_detached(client, volume_name)
pvc['spec']['resources'] = {
'requests': {
'storage': size_to_string(EXPANDED_VOLUME_SIZE*Gi)
}
}
expand_and_wait_for_pvc(core_api, pvc)
wait_for_volume_expansion(client, volume_name)
volume = client.by_id_volume(volume_name)
assert volume.state == "detached"
assert volume.size == str(EXPANDED_VOLUME_SIZE*Gi)
pod_manifest['metadata']['name'] = pod_name
pod_manifest['spec']['volumes'] = [{
'name':
pod_manifest['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {'claimName': pvc_name},
}]
create_and_wait_pod(core_api, pod_manifest)
resp = read_volume_data(core_api, pod_name)
assert resp == test_data
volume = client.by_id_volume(volume_name)
engine = get_volume_engine(volume)
assert volume.size == str(EXPANDED_VOLUME_SIZE*Gi)
assert volume.size == engine.size
def test_upgrade(upgrade_image_tag, settings_reset, volume_name, pod_make, statefulset, storage_class): # NOQA
"""
Test Longhorn upgrade
Prerequisite:
- Disable Auto Salvage Setting
1. Find the upgrade image tag
2. Create a volume, generate and write data into the volume.
3. Create a Pod using a volume, generate and write data
4. Create a StatefulSet with 2 replicas,
generate and write data to their volumes
5. Keep all volumes attached
6. Upgrade Longhorn system.
7. Check Pod and StatefulSet didn't restart after upgrade
8. Check All volumes data
9. Write data to StatefulSet pods, and Attached volume
10. Check data written to StatefulSet pods, and attached volume.
11. Detach the volume, and Delete Pod, and
StatefulSet to detach theirvolumes
12. Upgrade all volumes engine images.
13. Attach the volume, and recreate Pod, and StatefulSet
14. Check All volumes data
"""
new_ei_name = "longhornio/longhorn-engine:" + upgrade_image_tag
client = get_longhorn_api_client()
core_api = get_core_api_client()
host_id = get_self_host_id()
pod_data_path = "/data/test"
pod_volume_name = generate_volume_name()
auto_salvage_setting = client.by_id_setting(SETTING_AUTO_SALVAGE)
setting = client.update(auto_salvage_setting, value="false")
assert setting.name == SETTING_AUTO_SALVAGE
assert setting.value == "false"
# Create Volume attached to a node.
volume1 = create_and_check_volume(client,
volume_name,
size=SIZE)
volume1.attach(hostId=host_id)
volume1 = wait_for_volume_healthy(client, volume_name)
volume1_data = write_volume_random_data(volume1)
# Create Volume used by Pod
pod_name, pv_name, pvc_name, pod_md5sum = \
prepare_pod_with_data_in_mb(client, core_api,
pod_make, pod_volume_name,
data_path=pod_data_path,
add_liveness_prope=False)
# Create multiple volumes used by StatefulSet
statefulset_name = 'statefulset-upgrade-test'
update_statefulset_manifests(statefulset,
storage_class,
statefulset_name)
create_storage_class(storage_class)
create_and_wait_statefulset(statefulset)
statefulset_pod_info = get_statefulset_pod_info(core_api, statefulset)
for sspod_info in statefulset_pod_info:
sspod_info['data'] = generate_random_data(VOLUME_RWTEST_SIZE)
write_pod_volume_data(core_api,
sspod_info['pod_name'],
sspod_info['data'])
# upgrade Longhorn
assert longhorn_upgrade(upgrade_image_tag)
client = get_longhorn_api_client()
# wait for 1 minute before checking pod restarts
time.sleep(60)
pod = core_api.read_namespaced_pod(name=pod_name,
namespace='default')
assert pod.status.container_statuses[0].restart_count == 0
for sspod_info in statefulset_pod_info:
sspod = core_api.read_namespaced_pod(name=sspod_info['pod_name'],
namespace='default')
assert \
sspod.status.container_statuses[0].restart_count == 0
for sspod_info in statefulset_pod_info:
resp = read_volume_data(core_api, sspod_info['pod_name'])
assert resp == sspod_info['data']
res_pod_md5sum = get_pod_data_md5sum(core_api, pod_name, pod_data_path)
assert res_pod_md5sum == pod_md5sum
check_volume_data(volume1, volume1_data)
for sspod_info in statefulset_pod_info:
sspod_info['data'] = generate_random_data(VOLUME_RWTEST_SIZE)
write_pod_volume_data(core_api,
sspod_info['pod_name'],
sspod_info['data'])
for sspod_info in statefulset_pod_info:
resp = read_volume_data(core_api, sspod_info['pod_name'])
assert resp == sspod_info['data']
volume1 = client.by_id_volume(volume_name)
volume1_data = write_volume_random_data(volume1)
check_volume_data(volume1, volume1_data)
statefulset['spec']['replicas'] = replicas = 0
apps_api = get_apps_api_client()
apps_api.patch_namespaced_stateful_set(
name=statefulset_name,
namespace='default',
body={
'spec': {
'replicas': replicas
}
})
delete_and_wait_pod(core_api, pod_name)
volume = client.by_id_volume(volume_name)
volume.detach()
volumes = client.list_volume()
for v in volumes:
wait_for_volume_detached(client, v.name)
engineimages = client.list_engine_image()
for ei in engineimages:
if ei.image == new_ei_name:
new_ei = ei
volumes = client.list_volume()
for v in volumes:
volume = client.by_id_volume(v.name)
volume.engineUpgrade(image=new_ei.image)
statefulset['spec']['replicas'] = replicas = 2
apps_api = get_apps_api_client()
apps_api.patch_namespaced_stateful_set(
name=statefulset_name,
namespace='default',
body={
'spec': {
'replicas': replicas
}
})
wait_statefulset(statefulset)
pod = pod_make(name=pod_name)
pod['spec']['volumes'] = [create_pvc_spec(pvc_name)]
create_and_wait_pod(core_api, pod)
volume1 = client.by_id_volume(volume_name)
volume1.attach(hostId=host_id)
volume1 = wait_for_volume_healthy(client, volume_name)
for sspod_info in statefulset_pod_info:
resp = read_volume_data(core_api, sspod_info['pod_name'])
assert resp == sspod_info['data']
res_pod_md5sum = get_pod_data_md5sum(core_api, pod_name, pod_data_path)
assert res_pod_md5sum == pod_md5sum
check_volume_data(volume1, volume1_data)
def test_kubernetes_status(
client,
core_api,
storage_class, # NOQA
statefulset,
csi_pv,
pvc,
pod): # NOQA
"""
Test Volume feature: Kubernetes Status
1. Create StorageClass with `reclaimPolicy = Retain`
2. Create a statefulset `kubernetes-status-test` with the StorageClass
1. The statefulset has scale of 2.
3. Get the volume name from the SECOND pod of the StateufulSet pod and
create an `extra_pod` with the same volume on the same node
4. Check the volumes that used by the StatefulSet
1. The volume used by the FIRST StatefulSet pod will have one workload
2. The volume used by the SECOND StatefulSet pod will have two
workloads
3. Validate related status, e.g. pv/pod name/state, workload
name/type
5. Check the volumes again
1. PV/PVC should still be bound
2. The volume used by the FIRST pod should have history data
3. The volume used by the SECOND and extra pod should have current data
point to the extra pod
6. Delete the extra pod
1. Now all the volume's should only have history data(`lastPodRefAt`
set)
7. Delete the PVC
1. PVC should be updated with status `Released` and become history data
8. Delete PV
1. All the Kubernetes status information should be cleaned up.
9. Reuse the two Longhorn volumes to create new pods
1. Since the `reclaimPolicy == Retain`, volume won't be deleted by
Longhorn
2. Check the Kubernetes status now updated, with pod info but empty
workload
3. Default Longhorn Static StorageClass will remove the PV with PVC,
but leave Longhorn volume
"""
statefulset_name = 'kubernetes-status-test'
update_statefulset_manifests(statefulset, storage_class, statefulset_name)
storage_class['reclaimPolicy'] = 'Retain'
create_storage_class(storage_class)
create_and_wait_statefulset(statefulset)
pod_info = get_statefulset_pod_info(core_api, statefulset)
volume_info = [p['pv_name'] for p in pod_info]
extra_pod_name = 'extra-pod-using-' + volume_info[1]
pod['metadata']['name'] = extra_pod_name
p2 = core_api.read_namespaced_pod(name=pod_info[1]['pod_name'],
namespace='default')
pod['spec']['nodeName'] = p2.spec.node_name
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': pod_info[1]['pvc_name'],
},
}]
create_and_wait_pod(core_api, pod)
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i] # NOQA
volume = client.by_id_volume(volume_name)
k_status = volume.kubernetesStatus
workloads = k_status.workloadsStatus
assert k_status.pvName == p['pv_name']
assert k_status.pvStatus == 'Bound'
assert k_status.namespace == 'default'
assert k_status.pvcName == p['pvc_name']
assert not k_status.lastPVCRefAt
assert not k_status.lastPodRefAt
if i == 0:
assert len(workloads) == 1
assert workloads[0].podName == p['pod_name']
assert workloads[0].workloadName == statefulset_name
assert workloads[0].workloadType == 'StatefulSet'
for _ in range(RETRY_COUNTS):
if workloads[0].podStatus == 'Running':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume.kubernetesStatus
workloads = k_status.workloadsStatus
assert workloads[0].podStatus == 'Running'
if i == 1:
assert len(k_status.workloadsStatus) == 2
if workloads[0].podName == pod_info[i]['pod_name']:
assert workloads[1].podName == extra_pod_name
assert workloads[0].workloadName == statefulset_name
assert workloads[0].workloadType == 'StatefulSet'
assert not workloads[1].workloadName
assert not workloads[1].workloadType
else:
assert workloads[1].podName == pod_info[i]['pod_name']
assert workloads[0].podName == extra_pod_name
assert not workloads[0].workloadName
assert not workloads[0].workloadType
assert workloads[1].workloadName == statefulset_name
assert workloads[1].workloadType == 'StatefulSet'
for _ in range(RETRY_COUNTS):
if workloads[0].podStatus == 'Running' and \
workloads[1].podStatus == 'Running':
break
time.sleep(RETRY_INTERVAL)
volume = client.by_id_volume(volume_name)
k_status = volume.kubernetesStatus
workloads = k_status.workloadsStatus
assert len(workloads) == 2
assert workloads[0].podStatus == 'Running'
assert workloads[1].podStatus == 'Running'
ks_list = [{}, {}]
delete_and_wait_statefulset_only(core_api, statefulset)
# the extra pod is still using the 2nd volume
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
ks_list[i]['pvName'] = p['pv_name']
ks_list[i]['pvStatus'] = 'Bound'
ks_list[i]['namespace'] = 'default'
ks_list[i]['pvcName'] = p['pvc_name']
ks_list[i]['lastPVCRefAt'] = ''
if i == 0:
ks_list[i]['lastPodRefAt'] = 'not empty'
ks_list[i]['workloadsStatus'] = [
{
'podName': p['pod_name'],
'podStatus': 'Running',
'workloadName': statefulset_name,
'workloadType': 'StatefulSet',
},
]
if i == 1:
ks_list[i]['lastPodRefAt'] = ''
ks_list[i]['workloadsStatus'] = [{
'podName': extra_pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': '',
}]
wait_volume_kubernetes_status(client, volume_name, ks_list[i])
# deleted extra_pod, all volumes have no workload
delete_and_wait_pod(core_api, pod['metadata']['name'])
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
ks_list[i]['lastPodRefAt'] = 'not empty'
wait_volume_kubernetes_status(client, volume_name, ks_list[i])
# deleted pvc only.
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
delete_and_wait_pvc(core_api, p['pvc_name'])
ks_list[i]['pvStatus'] = 'Released'
ks_list[i]['lastPVCRefAt'] = 'not empty'
wait_volume_kubernetes_status(client, volume_name, ks_list[i])
# deleted pv only.
for i in range(len(volume_info)):
p, volume_name = pod_info[i], volume_info[i]
delete_and_wait_pv(core_api, p['pv_name'])
ks_list[i]['pvName'] = ''
ks_list[i]['pvStatus'] = ''
wait_volume_kubernetes_status(client, volume_name, ks_list[i])
# reuse that volume
for p, volume_name in zip(pod_info, volume_info):
p['pod_name'] = p['pod_name'].replace('kubernetes-status-test',
'kubernetes-status-test-reuse')
p['pvc_name'] = p['pvc_name'].replace('kubernetes-status-test',
'kubernetes-status-test-reuse')
p['pv_name'] = p['pvc_name']
csi_pv['metadata']['name'] = p['pv_name']
csi_pv['spec']['csi']['volumeHandle'] = volume_name
csi_pv['spec']['storageClassName'] = \
DEFAULT_LONGHORN_STATIC_STORAGECLASS_NAME
core_api.create_persistent_volume(csi_pv)
pvc['metadata']['name'] = p['pvc_name']
pvc['spec']['volumeName'] = p['pv_name']
pvc['spec']['storageClassName'] = \
DEFAULT_LONGHORN_STATIC_STORAGECLASS_NAME
core_api.create_namespaced_persistent_volume_claim(body=pvc,
namespace='default')
pod['metadata']['name'] = p['pod_name']
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': p['pvc_name'],
},
}]
create_and_wait_pod(core_api, pod)
ks = {
'pvName':
p['pv_name'],
'pvStatus':
'Bound',
'namespace':
'default',
'pvcName':
p['pvc_name'],
'lastPVCRefAt':
'',
'lastPodRefAt':
'',
'workloadsStatus': [
{
'podName': p['pod_name'],
'podStatus': 'Running',
'workloadName': '',
'workloadType': '',
},
],
}
wait_volume_kubernetes_status(client, volume_name, ks)
delete_and_wait_pod(core_api, p['pod_name'])
# Since persistentVolumeReclaimPolicy of csi_pv is `Delete`,
# we don't need to delete bounded pv manually
delete_and_wait_pvc(core_api, p['pvc_name'])
wait_delete_pv(core_api, p['pv_name'])
def test_csi_volumesnapshot_basic(
set_random_backupstore, # NOQA
volumesnapshotclass, # NOQA
volumesnapshot, # NOQA
client, # NOQA
core_api, # NOQA
volume_name, # NOQA
csi_pv, # NOQA
pvc, # NOQA
pod_make, # NOQA
volsnapshotclass_delete_policy, # NOQA
backup_is_deleted): # NOQA
"""
Test creation / restoration / deletion of a backup via the csi snapshotter
Context:
We want to allow the user to programmatically create/restore/delete
longhorn backups via the csi snapshot mechanism
ref: https://kubernetes.io/docs/concepts/storage/volume-snapshots/
Setup:
1. Make sure your cluster contains the below crds
https://github.com/kubernetes-csi/external-snapshotter
/tree/master/client/config/crd
2. Make sure your cluster contains the snapshot controller
https://github.com/kubernetes-csi/external-snapshotter
/tree/master/deploy/kubernetes/snapshot-controller
Steps:
def csi_volumesnapshot_creation_test(snapshotClass=longhorn|custom):
1. create volume(1)
2. write data to volume(1)
3. create a kubernetes `VolumeSnapshot` object
the `VolumeSnapshot.uuid` will be used to identify a
**longhorn snapshot** and the associated `VolumeSnapshotContent` object
4. check creation of a new longhorn snapshot named `snapshot-uuid`
5. check for `VolumeSnapshotContent` named `snapcontent-uuid`
6. wait for `VolumeSnapshotContent.readyToUse` flag to be set to **true**
7. check for backup existance on the backupstore
# the csi snapshot restore sets the fromBackup field same as
# the StorageClass based restore approach.
def csi_volumesnapshot_restore_test():
8. create a `PersistentVolumeClaim` object where the `dataSource` field
references the `VolumeSnapshot` object by name
9. verify creation of a new volume(2) bound to the pvc created in step(8)
10. verify data of new volume(2) equals data
from backup (ie old data above)
# default longhorn snapshot class is set to Delete
# add a second test with a custom snapshot class with deletionPolicy
# set to Retain you can reuse these methods for that and other tests
def csi_volumesnapshot_deletion_test(deletionPolicy='Delete|Retain'):
11. delete `VolumeSnapshot` object
12. if deletionPolicy == Delete:
13. verify deletion of `VolumeSnapshot` and
`VolumeSnapshotContent` objects
14. verify deletion of backup from backupstore
12. if deletionPolicy == Retain:
13. verify deletion of `VolumeSnapshot`
14. verify retention of `VolumeSnapshotContent`
and backup on backupstore
15. cleanup
"""
csisnapclass = \
volumesnapshotclass(name="snapshotclass",
deletepolicy=volsnapshotclass_delete_policy)
pod_name, pv_name, pvc_name, md5sum = \
prepare_pod_with_data_in_mb(client, core_api,
csi_pv, pvc, pod_make,
volume_name,
data_path="/data/test")
# Create volumeSnapshot test
csivolsnap = volumesnapshot(volume_name + "-volumesnapshot", "default",
csisnapclass["metadata"]["name"],
"persistentVolumeClaimName", pvc_name)
volume = client.by_id_volume(volume_name)
for i in range(RETRY_COUNTS):
snapshots = volume.snapshotList()
if len(snapshots) == 2:
break
time.sleep(RETRY_INTERVAL)
lh_snapshot = None
snapshots = volume.snapshotList()
for snapshot in snapshots:
if snapshot["name"] == "snapshot-" + csivolsnap["metadata"]["uid"]:
lh_snapshot = snapshot
assert lh_snapshot is not None
wait_for_volumesnapshot_ready(csivolsnap["metadata"]["name"],
csivolsnap["metadata"]["namespace"])
bv1, b = find_backup(client, volume_name, lh_snapshot["name"])
assert b["snapshotName"] == lh_snapshot["name"]
restore_pvc_name = pvc["metadata"]["name"] + "-restore"
restore_pvc_size = pvc["spec"]["resources"]["requests"]["storage"]
restore_csi_volume_snapshot(core_api, client, csivolsnap, restore_pvc_name,
restore_pvc_size)
restore_pod = pod_make()
restore_pod_name = restore_pod["metadata"]["name"]
restore_pod['spec']['volumes'] = [create_pvc_spec(restore_pvc_name)]
create_and_wait_pod(core_api, restore_pod)
restore_md5sum = \
get_pod_data_md5sum(core_api, restore_pod_name, path="/data/test")
assert restore_md5sum == md5sum
# Delete volumeSnapshot test
delete_volumesnapshot(csivolsnap["metadata"]["name"], "default")
if backup_is_deleted is False:
find_backup(client, volume_name, b["snapshotName"])
else:
wait_for_backup_delete(client, volume_name, b["name"])
def test_backup_kubernetes_status(set_random_backupstore, client, core_api,
pod): # NOQA
"""
Test that Backups have KubernetesStatus stored properly when there is an
associated PersistentVolumeClaim and Pod.
1. Setup a random backupstore
2. Set settings Longhorn Static StorageClass to `longhorn-static-test`
3. Create a volume and PV/PVC. Verify the StorageClass of PVC
4. Create a Pod using the PVC.
5. Check volume's Kubernetes status to reflect PV/PVC/Pod correctly.
6. Create a backup for the volume.
7. Verify the labels of created backup reflect PV/PVC/Pod status.
8. Restore the backup to a volume. Wait for restoration to complete.
9. Check the volume's Kubernetes Status
1. Make sure the `lastPodRefAt` and `lastPVCRefAt` is snapshot created
time
10. Delete the backup and restored volume.
11. Delete PV/PVC/Pod.
12. Verify volume's Kubernetes Status updated to reflect history data.
13. Attach the volume and create another backup. Verify the labels
14. Verify the volume's Kubernetes status.
15. Restore the previous backup to a new volume. Wait for restoration.
16. Verify the restored volume's Kubernetes status.
1. Make sure `lastPodRefAt` and `lastPVCRefAt` matched volume on step
12
"""
host_id = get_self_host_id()
static_sc_name = "longhorn-static-test"
setting = client.by_id_setting(SETTING_DEFAULT_LONGHORN_STATIC_SC)
setting = client.update(setting, value=static_sc_name)
assert setting.value == static_sc_name
volume_name = "test-backup-kubernetes-status-pod" # NOQA
client.create_volume(name=volume_name, size=SIZE, numberOfReplicas=2)
volume = wait_for_volume_detached(client, volume_name)
pod_name = "pod-" + volume_name
pv_name = "pv-" + volume_name
pvc_name = "pvc-" + volume_name
create_pv_for_volume(client, core_api, volume, pv_name)
create_pvc_for_volume(client, core_api, volume, pvc_name)
ret = core_api.list_namespaced_persistent_volume_claim(namespace='default')
pvc_found = False
for item in ret.items:
if item.metadata.name == pvc_name:
pvc_found = item
break
assert pvc_found
assert pvc_found.spec.storage_class_name == static_sc_name
pod['metadata']['name'] = pod_name
pod['spec']['volumes'] = [{
'name':
pod['spec']['containers'][0]['volumeMounts'][0]['name'],
'persistentVolumeClaim': {
'claimName': pvc_name,
},
}]
create_and_wait_pod(core_api, pod)
ks = {
'lastPodRefAt':
'',
'lastPVCRefAt':
'',
'namespace':
'default',
'pvcName':
pvc_name,
'pvName':
pv_name,
'pvStatus':
'Bound',
'workloadsStatus': [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
}
wait_volume_kubernetes_status(client, volume_name, ks)
volume = wait_for_volume_healthy(client, volume_name)
# Create Backup manually instead of calling create_backup since Kubernetes
# is not guaranteed to mount our Volume to the test host.
snap = create_snapshot(client, volume_name)
volume.snapshotBackup(name=snap.name)
wait_for_backup_completion(client, volume_name, snap.name)
_, b = find_backup(client, volume_name, snap.name)
# Check backup label
status = loads(b.labels.get(KUBERNETES_STATUS_LABEL))
assert status == ks
# Check backup volume label
for _ in range(RETRY_COUNTS):
bv = client.by_id_backupVolume(volume_name)
if bv is not None and bv.labels is not None:
break
time.sleep(RETRY_INTERVAL)
assert bv is not None and bv.labels is not None
status = loads(bv.labels.get(KUBERNETES_STATUS_LABEL))
assert status == ks
restore_name = generate_volume_name()
client.create_volume(name=restore_name,
size=SIZE,
numberOfReplicas=2,
fromBackup=b.url)
wait_for_volume_restoration_completed(client, restore_name)
wait_for_volume_detached(client, restore_name)
snapshot_created = b.snapshotCreated
ks = {
'lastPodRefAt':
b.snapshotCreated,
'lastPVCRefAt':
b.snapshotCreated,
'namespace':
'default',
'pvcName':
pvc_name,
# Restoration should not apply PersistentVolume data.
'pvName':
'',
'pvStatus':
'',
'workloadsStatus': [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
}
wait_volume_kubernetes_status(client, restore_name, ks)
restore = client.by_id_volume(restore_name)
# We need to compare LastPodRefAt and LastPVCRefAt manually since
# wait_volume_kubernetes_status only checks for empty or non-empty state.
assert restore.kubernetesStatus.lastPodRefAt == ks["lastPodRefAt"]
assert restore.kubernetesStatus.lastPVCRefAt == ks["lastPVCRefAt"]
delete_backup(client, bv.name, b.name)
client.delete(restore)
wait_for_volume_delete(client, restore_name)
delete_and_wait_pod(core_api, pod_name)
delete_and_wait_pvc(core_api, pvc_name)
delete_and_wait_pv(core_api, pv_name)
# With the Pod, PVC, and PV deleted, the Volume should have both Ref
# fields set. Check that a new Backup and Restore will use this instead of
# manually populating the Ref fields.
ks = {
'lastPodRefAt':
'NOT NULL',
'lastPVCRefAt':
'NOT NULL',
'namespace':
'default',
'pvcName':
pvc_name,
'pvName':
'',
'pvStatus':
'',
'workloadsStatus': [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
}
wait_volume_kubernetes_status(client, volume_name, ks)
volume = wait_for_volume_detached(client, volume_name)
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
snap = create_snapshot(client, volume_name)
volume.snapshotBackup(name=snap.name)
volume = wait_for_backup_completion(client, volume_name, snap.name)
bv, b = find_backup(client, volume_name, snap.name)
new_b = bv.backupGet(name=b.name)
status = loads(new_b.labels.get(KUBERNETES_STATUS_LABEL))
# Check each field manually, we have no idea what the LastPodRefAt or the
# LastPVCRefAt will be. We just know it shouldn't be SnapshotCreated.
assert status['lastPodRefAt'] != snapshot_created
assert status['lastPVCRefAt'] != snapshot_created
assert status['namespace'] == "default"
assert status['pvcName'] == pvc_name
assert status['pvName'] == ""
assert status['pvStatus'] == ""
assert status['workloadsStatus'] == [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
restore_name = generate_volume_name()
client.create_volume(name=restore_name,
size=SIZE,
numberOfReplicas=2,
fromBackup=b.url)
wait_for_volume_restoration_completed(client, restore_name)
wait_for_volume_detached(client, restore_name)
ks = {
'lastPodRefAt':
status['lastPodRefAt'],
'lastPVCRefAt':
status['lastPVCRefAt'],
'namespace':
'default',
'pvcName':
pvc_name,
'pvName':
'',
'pvStatus':
'',
'workloadsStatus': [{
'podName': pod_name,
'podStatus': 'Running',
'workloadName': '',
'workloadType': ''
}]
}
wait_volume_kubernetes_status(client, restore_name, ks)
restore = client.by_id_volume(restore_name)
assert restore.kubernetesStatus.lastPodRefAt == ks["lastPodRefAt"]
assert restore.kubernetesStatus.lastPVCRefAt == ks["lastPVCRefAt"]
# cleanup
backupstore_cleanup(client)
client.delete(restore)
cleanup_volume(client, volume)
def test_data_locality_basic(client, core_api, volume_name, pod,
settings_reset): # NOQA
"""
Test data locality basic feature
Context:
Data Locality feature allows users to have an option to keep a local
replica on the same node as the consuming pod.
Longhorn is currently supporting 2 modes:
- disabled: Longhorn does not try to keep a local replica
- best-effort: Longhorn try to keep a local replica
See manual tests at:
https://github.com/longhorn/longhorn/issues/1045#issuecomment-680706283
Steps:
Case 1: Test that Longhorn builds a local replica on the engine node
1. Create a volume(1) with 1 replica and dataLocality set to disabled
2. Find node where the replica is located on.
Let's call the node is replica-node
3. Attach the volume to a node different than replica-node.
Let call the node is engine-node
4. Write 200MB data to volume(1)
5. Use a retry loop to verify that Longhorn does not create
a replica on the engine-node
6. Update dataLocality to best-effort for volume(1)
7. Use a retry loop to verify that Longhorn creates and rebuilds
a replica on the engine-node and remove the other replica
8. detach the volume(1) and attach it to a different node.
Let's call the new node is new-engine-node and the old
node is old-engine-node
9. Wait for volume(1) to finish attaching
10. Use a retry loop to verify that Longhorn creates and rebuilds
a replica on the new-engine-node and remove the replica on
old-engine-node
Case 2: Test that Longhorn prioritizes deleting replicas on the same node
1. Add the tag AVAIL to node-1 and node-2
2. Set node soft anti-affinity to `true`.
3. Create a volume(2) with 3 replicas and dataLocality set to best-effort
4. Use a retry loop to verify that all 3 replicas are on node-1 and
node-2, no replica is on node-3
5. Attach volume(2) to node-3
6. User a retry loop to verify that there is no replica on node-3 and
we can still read/write to volume(2)
7. Find the node which contains 2 replicas.
Let call the node is most-replica-node
8. Set the replica count to 2 for volume(2)
9. Verify that Longhorn remove one replica from most-replica-node
Case 3: Test that the volume is not corrupted if there is an unexpected
detachment during building local replica
1. Remove the tag AVAIL from node-1 and node-2
Set node soft anti-affinity to `false`.
2. Create a volume(3) with 1 replicas and dataLocality set to best-effort
3. Attach volume(3) to node-3.
4. Use a retry loop to verify that volume(3) has only 1 replica on node-3
5. Write 800MB data to volume(3)
6. Detach volume(3)
7. Attach volume(3) to node-1
8. Use a retry loop to:
Wait until volume(3) finishes attaching.
Wait until Longhorn start rebuilding a replica on node-1
Immediately detach volume(3)
9. Verify that the replica on node-1 is in ERR state.
10. Attach volume(3) to node-1
11. Wait until volume(3) finishes attaching.
12. Use a retry loop to verify the Longhorn cleanup the ERR replica,
rebuild a new replica on node-1, and remove the replica on node-3
Case 4: Make sure failed to schedule local replica doesn't block the
the creation of other replicas.
1. Disable scheduling for node-3
2. Create a vol with 1 replica, `dataLocality = best-effort`.
The replica is scheduled on a node (say node-1)
3. Attach vol to node-3. There is a fail-to-schedule
replica with Spec.HardNodeAffinity=node-3
4. Increase numberOfReplica to 3. Verify that the replica set contains:
one on node-1, one on node-2, one failed replica
with Spec.HardNodeAffinity=node-3.
5. Decrease numberOfReplica to 2. Verify that the replica set contains:
one on node-1, one on node-2, one failed replica
with Spec.HardNodeAffinity=node-3.
6. Decrease numberOfReplica to 1. Verify that the replica set contains:
one on node-1 or node-2, one failed replica
with Spec.HardNodeAffinity=node-3.
7. Decrease numberOfReplica to 2. Verify that the replica set contains:
one on node-1, one on node-2, one failed replica
with Spec.HardNodeAffinity=node-3.
8. Turn off data locality by set `dataLocality=disabled` for the vol.
Verify that the replica set contains: one on node-1, one on node-2
9. clean up
"""
# Case 1: Test that Longhorn builds a local replica on the engine node
nodes = client.list_node()
default_data_locality_setting = \
client.by_id_setting(SETTING_DEFAULT_DATA_LOCALITY)
try:
client.update(default_data_locality_setting, value="disabled")
except Exception as e:
print("Exception when update Default Data Locality setting",
default_data_locality_setting, e)
volume1_name = volume_name + "-1"
volume1_size = str(500 * Mi)
volume1_data_path = "/data/test"
pv1_name = volume1_name + "-pv"
pvc1_name = volume1_name + "-pvc"
pod1_name = volume1_name + "-pod"
pod1 = pod
pod1['metadata']['name'] = pod1_name
volume1 = create_and_check_volume(client,
volume1_name,
num_of_replicas=1,
size=volume1_size)
volume1 = client.by_id_volume(volume1_name)
create_pv_for_volume(client, core_api, volume1, pv1_name)
create_pvc_for_volume(client, core_api, volume1, pvc1_name)
volume1 = client.by_id_volume(volume1_name)
volume1_replica_node = volume1.replicas[0]['hostId']
volume1_attached_node = None
for node in nodes:
if node.name != volume1_replica_node:
volume1_attached_node = node.name
break
assert volume1_attached_node is not None
pod1['spec']['volumes'] = [{
"name": "pod-data",
"persistentVolumeClaim": {
"claimName": pvc1_name
}
}]
pod1['spec']['nodeSelector'] = \
{"kubernetes.io/hostname": volume1_attached_node}
create_and_wait_pod(core_api, pod1)
write_pod_volume_random_data(core_api, pod1_name, volume1_data_path,
DATA_SIZE_IN_MB_2)
for i in range(10):
volume1 = client.by_id_volume(volume1_name)
assert len(volume1.replicas) == 1
assert volume1.replicas[0]['hostId'] != volume1_attached_node
time.sleep(1)
volume1 = client.by_id_volume(volume1_name)
volume1.updateDataLocality(dataLocality="best-effort")
for _ in range(RETRY_COUNTS):
volume1 = client.by_id_volume(volume1_name)
assert volume1[VOLUME_FIELD_ROBUSTNESS] == VOLUME_ROBUSTNESS_HEALTHY
if len(volume1.replicas) == 1 and \
volume1.replicas[0]['hostId'] == volume1_attached_node:
break
time.sleep(RETRY_INTERVAL)
assert len(volume1.replicas) == 1
assert volume1.replicas[0]['hostId'] == volume1_attached_node
delete_and_wait_pod(core_api, pod1_name)
volume1 = wait_for_volume_detached(client, volume1_name)
volume1_replica_node = volume1.replicas[0]['hostId']
volume1_attached_node = None
for node in nodes:
if node.name != volume1_replica_node:
volume1_attached_node = node.name
break
assert volume1_attached_node is not None
pod1['spec']['nodeSelector'] = \
{"kubernetes.io/hostname": volume1_attached_node}
create_and_wait_pod(core_api, pod1)
for _ in range(RETRY_COUNTS):
volume1 = client.by_id_volume(volume1_name)
assert volume1[VOLUME_FIELD_ROBUSTNESS] == VOLUME_ROBUSTNESS_HEALTHY
if len(volume1.replicas) == 1 and \
volume1.replicas[0]['hostId'] == volume1_attached_node:
break
time.sleep(RETRY_INTERVAL)
assert len(volume1.replicas) == 1
assert volume1.replicas[0]['hostId'] == volume1_attached_node
delete_and_wait_pod(core_api, pod1_name)
wait_for_volume_detached(client, volume1_name)
# Case 2: Test that Longhorn prioritizes deleting replicas on the same node
node1 = nodes[0]
node2 = nodes[1]
node3 = nodes[2]
client.update(node1, allowScheduling=True, tags=["AVAIL"])
client.update(node2, allowScheduling=True, tags=["AVAIL"])
replica_node_soft_anti_affinity_setting = \
client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
try:
client.update(replica_node_soft_anti_affinity_setting, value="true")
except Exception as e:
print(
"Exception when update "
"Replica Node Level Soft Anti-Affinity setting",
replica_node_soft_anti_affinity_setting, e)
volume2_name = volume_name + "-2"
volume2_size = str(500 * Mi)
pv2_name = volume2_name + "-pv"
pvc2_name = volume2_name + "-pvc"
pod2_name = volume2_name + "-pod"
pod2 = pod
pod2['metadata']['name'] = pod2_name
volume2 = client.create_volume(name=volume2_name,
size=volume2_size,
numberOfReplicas=3,
nodeSelector=["AVAIL"],
dataLocality="best-effort")
volume2 = wait_for_volume_detached(client, volume2_name)
volume2 = client.by_id_volume(volume2_name)
create_pv_for_volume(client, core_api, volume2, pv2_name)
create_pvc_for_volume(client, core_api, volume2, pvc2_name)
volume2 = client.by_id_volume(volume2_name)
pod2['spec']['volumes'] = [{
"name": "pod-data",
"persistentVolumeClaim": {
"claimName": pvc2_name
}
}]
pod2['spec']['nodeSelector'] = {"kubernetes.io/hostname": node3.name}
create_and_wait_pod(core_api, pod2)
volume2 = wait_for_volume_healthy(client, volume2_name)
for replica in volume2.replicas:
assert replica["hostId"] != node3.name
volume2.updateReplicaCount(replicaCount=2)
# 2 Healthy replicas and 1 replica failed to schedule
# The failed to schedule replica is the local replica on node3
volume2 = wait_for_volume_replica_count(client, volume2_name, 3)
volume2 = client.by_id_volume(volume2_name)
volume2_healthy_replicas = []
for replica in volume2.replicas:
if replica.running is True:
volume2_healthy_replicas.append(replica)
assert len(volume2_healthy_replicas) == 2
volume2_rep1 = volume2_healthy_replicas[0]
volume2_rep2 = volume2_healthy_replicas[1]
assert volume2_rep1["hostId"] != volume2_rep2["hostId"]
delete_and_wait_pod(core_api, pod2_name)
wait_for_volume_detached(client, volume2_name)
# Case 3: Test that the volume is not corrupted if there is an unexpected
# detachment during building local replica
client.update(node1, allowScheduling=True, tags=[])
client.update(node2, allowScheduling=True, tags=[])
replica_node_soft_anti_affinity_setting = \
client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
try:
client.update(replica_node_soft_anti_affinity_setting, value="false")
except Exception as e:
print(
"Exception when update "
"Replica Node Level Soft Anti-Affinity setting",
replica_node_soft_anti_affinity_setting, e)
volume3_name = volume_name + "-3"
volume3_size = str(1 * Gi)
volume3_data_path = "/data/test"
pv3_name = volume3_name + "-pv"
pvc3_name = volume3_name + "-pvc"
pod3_name = volume3_name + "-pod"
pod3 = pod
pod3['metadata']['name'] = pod3_name
volume3 = client.create_volume(name=volume3_name,
size=volume3_size,
numberOfReplicas=1)
volume3 = wait_for_volume_detached(client, volume3_name)
volume3 = client.by_id_volume(volume3_name)
create_pv_for_volume(client, core_api, volume3, pv3_name)
create_pvc_for_volume(client, core_api, volume3, pvc3_name)
volume3 = client.by_id_volume(volume3_name)
pod3['spec']['volumes'] = [{
"name": "pod-data",
"persistentVolumeClaim": {
"claimName": pvc3_name
}
}]
pod3['spec']['nodeSelector'] = {"kubernetes.io/hostname": node3.name}
create_and_wait_pod(core_api, pod3)
volume3 = wait_for_volume_healthy(client, volume3_name)
write_pod_volume_random_data(core_api, pod3_name, volume3_data_path,
DATA_SIZE_IN_MB_4)
volume3.updateDataLocality(dataLocality="best-effort")
volume3 = client.by_id_volume(volume3_name)
if volume3.replicas[0]['hostId'] != node3.name:
wait_for_rebuild_start(client, volume3_name)
volume3 = client.by_id_volume(volume3_name)
assert len(volume3.replicas) == 2
wait_for_rebuild_complete(client, volume3_name)
volume3 = wait_for_volume_replica_count(client, volume3_name, 1)
assert volume3.replicas[0]["hostId"] == node3.name
delete_and_wait_pod(core_api, pod3_name)
pod3['spec']['nodeSelector'] = {"kubernetes.io/hostname": node1.name}
create_and_wait_pod(core_api, pod3)
wait_for_rebuild_start(client, volume3_name)
crash_engine_process_with_sigkill(client, core_api, volume3_name)
delete_and_wait_pod(core_api, pod3_name)
wait_for_volume_detached(client, volume3_name)
volume3 = client.by_id_volume(volume3_name)
assert len(volume3.replicas) == 1
assert volume3.replicas[0]["hostId"] == node3.name
create_and_wait_pod(core_api, pod3)
wait_for_rebuild_start(client, volume3_name)
volume3 = client.by_id_volume(volume3_name)
assert len(volume3.replicas) == 2
wait_for_rebuild_complete(client, volume3_name)
# Wait for deletion of extra replica
volume3 = wait_for_volume_replica_count(client, volume3_name, 1)
assert volume3.replicas[0]["hostId"] == node1.name
assert volume3.replicas[0]["mode"] == "RW"
assert volume3.replicas[0]["running"] is True
delete_and_wait_pod(core_api, pod3_name)
wait_for_volume_detached(client, volume3_name)
# Case 4: Make sure failed to schedule local replica doesn't block the
# the creation of other replicas.
replica_node_soft_anti_affinity_setting = \
client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
try:
client.update(replica_node_soft_anti_affinity_setting, value="false")
except Exception as e:
print(
"Exception when update "
"Replica Node Level Soft Anti-Affinity setting",
replica_node_soft_anti_affinity_setting, e)
client.update(node3, allowScheduling=False)
volume4_name = volume_name + "-4"
volume4_size = str(1 * Gi)
volume4 = client.create_volume(name=volume4_name,
size=volume4_size,
numberOfReplicas=1,
dataLocality="best-effort")
volume4 = wait_for_volume_detached(client, volume4_name)
volume4 = client.by_id_volume(volume4_name)
volume4_replica_name = volume4.replicas[0]["name"]
volume4.attach(hostId=node3.name)
wait_for_volume_healthy(client, volume4_name)
volume4 = client.by_id_volume(volume4_name)
assert len(volume4.replicas) == 2
for replica in volume4.replicas:
if replica["name"] == volume4_replica_name:
assert replica["running"] is True
assert replica["mode"] == "RW"
else:
assert replica["running"] is False
assert replica["mode"] == ""
assert volume4.conditions.scheduled.reason == \
"LocalReplicaSchedulingFailure"
volume4 = volume4.updateReplicaCount(replicaCount=3)
volume4 = wait_for_volume_degraded(client, volume4_name)
v4_node1_replica_count = 0
v4_node2_replica_count = 0
v4_failed_replica_count = 0
for replica in volume4.replicas:
if replica["hostId"] == node1.name:
v4_node1_replica_count += 1
elif replica["hostId"] == node2.name:
v4_node2_replica_count += 1
elif replica["hostId"] == "":
v4_failed_replica_count += 1
assert v4_node1_replica_count == 1
assert v4_node2_replica_count == 1
assert v4_failed_replica_count > 0
volume4 = volume4.updateReplicaCount(replicaCount=2)
volume4 = wait_for_volume_replica_count(client, volume4_name, 3)
v4_node1_replica_count = 0
v4_node2_replica_count = 0
v4_failed_replica_count = 0
for replica in volume4.replicas:
if replica["hostId"] == node1.name:
v4_node1_replica_count += 1
elif replica["hostId"] == node2.name:
v4_node2_replica_count += 1
elif replica["hostId"] == "":
v4_failed_replica_count += 1
assert v4_node1_replica_count == 1
assert v4_node2_replica_count == 1
assert v4_failed_replica_count > 0
volume4 = volume4.updateReplicaCount(replicaCount=1)
volume4 = wait_for_volume_replica_count(client, volume4_name, 2)
v4_node1_replica_count = 0
v4_node2_replica_count = 0
v4_failed_replica_count = 0
for replica in volume4.replicas:
if replica["hostId"] == node1.name:
v4_node1_replica_count += 1
elif replica["hostId"] == node2.name:
v4_node2_replica_count += 1
elif replica["hostId"] == "":
v4_failed_replica_count += 1
assert v4_node1_replica_count + v4_node2_replica_count == 1
assert v4_failed_replica_count == 1
volume4 = volume4.updateDataLocality(dataLocality="disabled")
volume4 = volume4.updateReplicaCount(replicaCount=2)
running_replica_count = 0
for _ in range(RETRY_COUNTS):
volume4 = client.by_id_volume(volume4_name)
running_replica_count = 0
for r in volume4.replicas:
if r.failedAt == "" and r.running is True:
running_replica_count += 1
if running_replica_count == 2:
break
time.sleep(RETRY_INTERVAL)
assert running_replica_count == 2
v4_node1_replica_count = 0
v4_node2_replica_count = 0
v4_node3_replica_count = 0
for replica in volume4.replicas:
wait_for_replica_running(client, volume4_name, replica["name"])
if replica["hostId"] == node1.name:
v4_node1_replica_count += 1
elif replica["hostId"] == node2.name:
v4_node2_replica_count += 1
elif replica["hostId"] == node3.name:
v4_node3_replica_count += 1
assert v4_node1_replica_count == 1
assert v4_node2_replica_count == 1
assert v4_node3_replica_count == 0
