def test_soft_anti_affinity_scheduling(client, volume_name): # NOQA
"""
Test that volumes with Soft Anti-Affinity work as expected.
With Soft Anti-Affinity, a new replica should still be scheduled on a node
with an existing replica, which will result in "Healthy" state but limited
redundancy.
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_SOFT_ANTI_AFFINITY)
client.update(setting, value="true")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
replica_names = map(lambda replica: replica.name, volume["replicas"])
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica["name"])
wait_new_replica_ready(client, volume_name, replica_names)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def test_soft_anti_affinity_detach(client, volume_name): # NOQA
"""
Test that volumes with Soft Anti-Affinity can detach and reattach to a
node properly.
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_SOFT_ANTI_AFFINITY)
client.update(setting, value="true")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
replica_names = map(lambda replica: replica.name, volume["replicas"])
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica["name"])
wait_new_replica_ready(client, volume_name, replica_names)
volume = wait_for_volume_healthy(client, volume_name)
volume.detach()
volume = wait_for_volume_detached(client, volume_name)
assert len(volume["replicas"]) == 3
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def test_hard_anti_affinity_offline_rebuild(client, volume_name): # NOQA
"""
Test that volumes with Hard Anti-Affinity can build new replicas during
the attaching process once a valid node is available.
Once a new replica has been built as part of the attaching process, the
volume should be Healthy again.
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_SOFT_ANTI_AFFINITY)
client.update(setting, value="false")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
replica_names = map(lambda replica: replica.name, volume["replicas"])
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica["name"])
volume = wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
volume.detach()
volume = wait_for_volume_detached(client, volume_name)
client.update(node, allowScheduling=True)
volume.attach(hostId=host_id)
wait_new_replica_ready(client, volume_name, replica_names)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def test_hard_anti_affinity_live_rebuild(client, volume_name): # NOQA
"""
Test that volumes with Hard Anti-Affinity can build new replicas live once
a valid node is available.
If no nodes without existing replicas are available, the volume should
remain in "Degraded" state. However, once one is available, the replica
should now be scheduled successfully, with the volume returning to
"Healthy" state.
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_SOFT_ANTI_AFFINITY)
client.update(setting, value="false")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
replica_names = map(lambda replica: replica.name, volume["replicas"])
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica["name"])
wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
# Allow scheduling on host node again
client.update(node, allowScheduling=True)
wait_new_replica_ready(client, volume_name, replica_names)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def test_hard_anti_affinity_detach(client, volume_name): # NOQA
"""
Test that volumes with Hard Anti-Affinity are still able to detach and
reattach to a node properly, even in degraded state.
1. Create a volume and attach to the current node
2. Generate and write `data` to the volume.
3. Set `soft anti-affinity` to false
4. Disable current node's scheduling.
5. Remove the replica on the current node
1. Verify volume will be in degraded state.
2. Verify volume reports condition `scheduled == false`
6. Detach the volume.
7. Verify that volume only have 2 replicas
1. Unhealthy replica will be removed upon detach.
8. Attach the volume again.
1. Verify volume will be in degraded state.
2. Verify volume reports condition `scheduled == false`
3. Verify only two of three replicas of volume are healthy.
4. Verify the remaining replica doesn't have `replica.HostID`, meaning
it's unscheduled
9. Check volume `data`
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
client.update(setting, value="false")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica.name)
volume = wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
volume.detach()
volume = wait_for_volume_detached(client, volume_name)
assert len(volume.replicas) == 2
volume.attach(hostId=host_id)
# Make sure we're still not getting another successful replica.
volume = wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
assert sum([
1 for replica in volume.replicas
if replica.running and replica.mode == "RW"
]) == 2
assert sum([1 for replica in volume.replicas if not replica.hostId]) == 1
assert len(volume.replicas) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def test_hard_anti_affinity_scheduling(client, volume_name): # NOQA
"""
Test that volumes with Hard Anti-Affinity work as expected.
With Hard Anti-Affinity, scheduling on nodes with existing replicas should
be forbidden, resulting in "Degraded" state.
1. Create a volume and attach to the current node
2. Generate and write `data` to the volume.
3. Set `soft anti-affinity` to false
4. Disable current node's scheduling.
5. Remove the replica on the current node
1. Verify volume will be in degraded state.
2. Verify volume reports condition `scheduled == false`
3. Verify only two of three replicas of volume are healthy.
4. Verify the remaining replica doesn't have `replica.HostID`, meaning
it's unscheduled
6. Check volume `data`
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
client.update(setting, value="false")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica.name)
# Instead of waiting for timeout and lengthening the tests a significant
# amount we can make sure the scheduling isn't working by making sure the
# volume becomes Degraded and reports a scheduling error.
wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
# While there are three replicas that should exist to meet the Volume's
# request, only two of those volumes should actually be Healthy.
volume = client.by_id_volume(volume_name)
assert sum([
1 for replica in volume.replicas
if replica.running and replica.mode == "RW"
]) == 2
# Confirm that the final volume is an unscheduled volume.
assert sum([1 for replica in volume.replicas if not replica.hostId]) == 1
# Three replicas in total should still exist.
assert len(volume.replicas) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def check_volume_data(client, volume_name, data): # NOQA
"""
Attaches the volume to the current node
then compares the volumes data
against the passed data.
"""
volume = client.by_id_volume(volume_name)
volume.attach(hostId=common.get_self_host_id())
volume = common.wait_for_volume_healthy(client, volume_name)
common.check_volume_data(volume, data)
volume.detach(hostId="")
volume = common.wait_for_volume_detached(client, volume_name)
def test_csi_expansion_with_size_round_up(client, core_api): # NOQA
"""
test expand longhorn volume
1. Create longhorn volume with size '1Gi'
2. Attach, write data, and detach
3. Expand volume size to '2000000000/2G' and
check if size round up '2000683008'
4. Attach, write data, and detach
5. Expand volume size to '2Gi' and check if size is '2147483648'
6. Attach, write data, and detach
"""
volume_name = generate_volume_name()
volume = create_and_check_volume(client, volume_name, 2, str(1 * Gi))
self_hostId = get_self_host_id()
volume.attach(hostId=self_hostId, disableFrontend=False)
volume = wait_for_volume_healthy(client, volume_name)
test_data = write_volume_random_data(volume)
volume.detach(hostId="")
volume = wait_for_volume_detached(client, volume_name)
volume.expand(size="2000000000")
wait_for_volume_expansion(client, volume_name)
volume = client.by_id_volume(volume_name)
assert volume.size == "2000683008"
self_hostId = get_self_host_id()
volume.attach(hostId=self_hostId, disableFrontend=False)
volume = wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, test_data, False)
test_data = write_volume_random_data(volume)
volume.detach(hostId="")
volume = wait_for_volume_detached(client, volume_name)
volume.expand(size=str(2 * Gi))
wait_for_volume_expansion(client, volume_name)
volume = client.by_id_volume(volume_name)
assert volume.size == "2147483648"
self_hostId = get_self_host_id()
volume.attach(hostId=self_hostId, disableFrontend=False)
volume = wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, test_data, False)
volume.detach(hostId="")
volume = wait_for_volume_detached(client, volume_name)
client.delete(volume)
wait_for_volume_delete(client, volume_name)
def ha_salvage_test(client, volume_name, base_image=""): # NOQA
volume = client.create_volume(name=volume_name,
size=SIZE,
numberOfReplicas=2,
baseImage=base_image)
volume = common.wait_for_volume_detached(client, volume_name)
assert volume["name"] == volume_name
assert volume["size"] == SIZE
assert volume["numberOfReplicas"] == 2
assert volume["state"] == "detached"
assert volume["created"] != ""
assert volume["baseImage"] == base_image
host_id = get_self_host_id()
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 2
replica0_name = volume["replicas"][0]["name"]
replica1_name = volume["replicas"][1]["name"]
data = write_volume_random_data(volume)
common.k8s_delete_replica_pods_for_volume(volume_name)
volume = common.wait_for_volume_faulted(client, volume_name)
assert len(volume["replicas"]) == 2
assert volume["replicas"][0]["failedAt"] != ""
assert volume["replicas"][1]["failedAt"] != ""
volume.salvage(names=[replica0_name, replica1_name])
volume = common.wait_for_volume_detached(client, volume_name)
assert len(volume["replicas"]) == 2
assert volume["replicas"][0]["failedAt"] == ""
assert volume["replicas"][1]["failedAt"] == ""
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, data)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
client.delete(volume)
common.wait_for_volume_delete(client, volume_name)
volumes = client.list_volume()
assert len(volumes) == 0
def backupstore_test(client, host_id, volname, size):
volume = client.by_id_volume(volname)
volume.snapshotCreate()
data = write_volume_random_data(volume)
snap2 = volume.snapshotCreate()
volume.snapshotCreate()
volume.snapshotBackup(name=snap2["name"])
bv, b = common.find_backup(client, volname, snap2["name"])
new_b = bv.backupGet(name=b["name"])
assert new_b["name"] == b["name"]
assert new_b["url"] == b["url"]
assert new_b["snapshotName"] == b["snapshotName"]
assert new_b["snapshotCreated"] == b["snapshotCreated"]
assert new_b["created"] == b["created"]
assert new_b["volumeName"] == b["volumeName"]
assert new_b["volumeSize"] == b["volumeSize"]
assert new_b["volumeCreated"] == b["volumeCreated"]
# test restore
restoreName = generate_volume_name()
volume = client.create_volume(name=restoreName,
size=size,
numberOfReplicas=2,
fromBackup=b["url"])
volume = common.wait_for_volume_detached(client, restoreName)
assert volume["name"] == restoreName
assert volume["size"] == size
assert volume["numberOfReplicas"] == 2
assert volume["state"] == "detached"
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, restoreName)
check_volume_data(volume, data)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, restoreName)
client.delete(volume)
volume = wait_for_volume_delete(client, restoreName)
bv.backupDelete(name=b["name"])
backups = bv.backupList()
found = False
for b in backups:
if b["snapshotName"] == snap2["name"]:
found = True
break
assert not found
def test_hard_anti_affinity_offline_rebuild(client, volume_name): # NOQA
"""
Test that volumes with Hard Anti-Affinity can build new replicas during
the attaching process once a valid node is available.
Once a new replica has been built as part of the attaching process, the
volume should be Healthy again.
1. Create a volume and attach to the current node
2. Generate and write `data` to the volume.
3. Set `soft anti-affinity` to false
4. Disable current node's scheduling.
5. Remove the replica on the current node
1. Verify volume will be in degraded state.
2. Verify volume reports condition `scheduled == false`
6. Detach the volume.
7. Enable current node's scheduling.
8. Attach the volume again.
9. Wait for volume to become healthy with 3 replicas
10. Check volume `data`
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
client.update(setting, value="false")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
replica_names = map(lambda replica: replica.name, volume.replicas)
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica.name)
volume = wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
volume.detach()
volume = wait_for_volume_detached(client, volume_name)
client.update(node, allowScheduling=True)
volume.attach(hostId=host_id)
wait_new_replica_ready(client, volume_name, replica_names)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def ha_salvage_test(client, volume_name, base_image=""): # NOQA
volume = client.create_volume(name=volume_name, size=SIZE,
numberOfReplicas=2, baseImage=base_image)
volume = common.wait_for_volume_detached(client, volume_name)
assert volume["name"] == volume_name
assert volume["size"] == SIZE
assert volume["numberOfReplicas"] == 2
assert volume["state"] == "detached"
assert volume["created"] != ""
assert volume["baseImage"] == base_image
host_id = get_self_host_id()
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 2
replica0_name = volume["replicas"][0]["name"]
replica1_name = volume["replicas"][1]["name"]
data = write_volume_random_data(volume)
common.k8s_delete_replica_pods_for_volume(volume_name)
volume = common.wait_for_volume_faulted(client, volume_name)
assert len(volume["replicas"]) == 2
assert volume["replicas"][0]["failedAt"] != ""
assert volume["replicas"][1]["failedAt"] != ""
volume.salvage(names=[replica0_name, replica1_name])
volume = common.wait_for_volume_detached(client, volume_name)
assert len(volume["replicas"]) == 2
assert volume["replicas"][0]["failedAt"] == ""
assert volume["replicas"][1]["failedAt"] == ""
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, data)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
client.delete(volume)
common.wait_for_volume_delete(client, volume_name)
volumes = client.list_volume()
assert len(volumes) == 0
def test_attach_without_frontend(clients, volume_name): # NOQA
for host_id, client in clients.iteritems():
break
volume = create_and_check_volume(client, volume_name)
lht_hostId = get_self_host_id()
volume.attach(hostId=lht_hostId, disableFrontend=False)
common.wait_for_volume_healthy(client, volume_name)
volume = client.by_id_volume(volume_name)
assert volume["disableFrontend"] is False
assert volume["frontend"] == "blockdev"
snap1_data = write_volume_random_data(volume)
snap1 = volume.snapshotCreate()
write_volume_random_data(volume)
volume.snapshotCreate()
volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
volume.attach(hostId=lht_hostId, disableFrontend=True)
common.wait_for_volume_healthy(client, volume_name)
volume = client.by_id_volume(volume_name)
engine = get_volume_engine(volume)
assert volume["disableFrontend"] is True
assert volume["frontend"] == "blockdev"
assert engine["endpoint"] == ""
volume.snapshotRevert(name=snap1["name"])
volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
volume.attach(hostId=lht_hostId, disableFrontend=False)
common.wait_for_volume_healthy(client, volume_name)
volume = client.by_id_volume(volume_name)
assert volume["disableFrontend"] is False
assert volume["frontend"] == "blockdev"
check_volume_data(volume, snap1_data)
client.delete(volume)
wait_for_volume_delete(client, volume_name)
def test_hard_anti_affinity_live_rebuild(client, volume_name): # NOQA
"""
Test that volumes with Hard Anti-Affinity can build new replicas live once
a valid node is available.
If no nodes without existing replicas are available, the volume should
remain in "Degraded" state. However, once one is available, the replica
should now be scheduled successfully, with the volume returning to
"Healthy" state.
1. Create a volume and attach to the current node
2. Generate and write `data` to the volume.
3. Set `soft anti-affinity` to false
4. Disable current node's scheduling.
5. Remove the replica on the current node
1. Verify volume will be in degraded state.
2. Verify volume reports condition `scheduled == false`
6. Enable the current node's scheduling
7. Wait for volume to start rebuilding and become healthy again
8. Check volume `data`
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
client.update(setting, value="false")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
replica_names = map(lambda replica: replica.name, volume.replicas)
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica.name)
wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
# Allow scheduling on host node again
client.update(node, allowScheduling=True)
wait_new_replica_ready(client, volume_name, replica_names)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def ha_rebuild_replica_test(client, volname): # NOQA
volume = client.by_id_volume(volname)
assert get_volume_endpoint(volume) == DEV_PATH + volname
assert len(volume["replicas"]) == 2
replica0 = volume["replicas"][0]
assert replica0["name"] != ""
replica1 = volume["replicas"][1]
assert replica1["name"] != ""
data = write_volume_random_data(volume)
volume = volume.replicaRemove(name=replica0["name"])
# wait until we saw a replica starts rebuilding
new_replica_found = False
for i in range(RETRY_COUNTS):
v = client.by_id_volume(volname)
for r in v["replicas"]:
if r["name"] != replica0["name"] and \
r["name"] != replica1["name"]:
new_replica_found = True
break
if new_replica_found:
break
time.sleep(RETRY_INTERVAL)
assert new_replica_found
volume = common.wait_for_volume_healthy(client, volname)
volume = client.by_id_volume(volname)
assert volume["state"] == common.VOLUME_STATE_ATTACHED
assert volume["robustness"] == common.VOLUME_ROBUSTNESS_HEALTHY
assert len(volume["replicas"]) >= 2
found = False
for replica in volume["replicas"]:
if replica["name"] == replica1["name"]:
found = True
break
assert found
check_volume_data(volume, data)
def ha_rebuild_replica_test(client, volname): # NOQA
volume = client.by_id_volume(volname)
assert get_volume_endpoint(volume) == DEV_PATH + volname
assert len(volume["replicas"]) == 2
replica0 = volume["replicas"][0]
assert replica0["name"] != ""
replica1 = volume["replicas"][1]
assert replica1["name"] != ""
data = write_volume_random_data(volume)
volume = volume.replicaRemove(name=replica0["name"])
# wait until we saw a replica starts rebuilding
new_replica_found = False
for i in range(RETRY_COUNTS):
v = client.by_id_volume(volname)
for r in v["replicas"]:
if r["name"] != replica0["name"] and \
r["name"] != replica1["name"]:
new_replica_found = True
break
if new_replica_found:
break
time.sleep(RETRY_INTERVAL)
assert new_replica_found
volume = common.wait_for_volume_healthy(client, volname)
volume = client.by_id_volume(volname)
assert volume["state"] == common.VOLUME_STATE_ATTACHED
assert volume["robustness"] == common.VOLUME_ROBUSTNESS_HEALTHY
assert len(volume["replicas"]) >= 2
found = False
for replica in volume["replicas"]:
if replica["name"] == replica1["name"]:
found = True
break
assert found
check_volume_data(volume, data)
def test_soft_anti_affinity_detach(client, volume_name): # NOQA
"""
Test that volumes with Soft Anti-Affinity can detach and reattach to a
node properly.
1. Create a volume and attach to the current node.
2. Generate and write `data` to the volume
3. Set `soft anti-affinity` to true
4. Disable current node's scheduling.
5. Remove the replica on the current node
6. Wait for the new replica to be rebuilt
7. Detach the volume.
8. Verify there are 3 replicas
9. Attach the volume again. Verify there are still 3 replicas
10. Verify the `data`.
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
client.update(setting, value="true")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
replica_names = list(map(lambda replica: replica.name, volume.replicas))
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica.name)
wait_new_replica_ready(client, volume_name, replica_names)
volume = wait_for_volume_healthy(client, volume_name)
volume.detach()
volume = wait_for_volume_detached(client, volume_name)
assert len(volume.replicas) == 3
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def setup_migration_test(clients, volume_name, backing_image=""): # NOQA
"""
Creates a new migratable volume then attaches it to the
current node to write some test data on it.
"""
client = get_random_client(clients)
volume = client.create_volume(name=volume_name, size=SIZE,
numberOfReplicas=REPLICA_COUNT,
backingImage=backing_image,
accessMode="rwx", migratable=True)
volume = common.wait_for_volume_detached(client, volume_name)
volume.attach(hostId=common.get_self_host_id())
volume = common.wait_for_volume_healthy(client, volume_name)
# write test data
data = common.write_volume_random_data(volume)
common.check_volume_data(volume, data)
volume.detach(hostId="")
volume = common.wait_for_volume_detached(client, volume_name)
return client, volume, data
def test_hard_anti_affinity_scheduling(client, volume_name): # NOQA
"""
Test that volumes with Hard Anti-Affinity work as expected.
With Hard Anti-Affinity, scheduling on nodes with existing replicas should
be forbidden, resulting in "Degraded" state.
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_SOFT_ANTI_AFFINITY)
client.update(setting, value="false")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica["name"])
# Instead of waiting for timeout and lengthening the tests a significant
# amount we can make sure the scheduling isn't working by making sure the
# volume becomes Degraded and reports a scheduling error.
volume = wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
# While there are three replicas that should exist to meet the Volume's
# request, only two of those volumes should actually be Healthy.
assert sum([
1 for replica in volume["replicas"]
if replica["running"] and replica["mode"] == "RW"
]) == 2
# Confirm that the final volume is an unscheduled volume.
assert sum([1 for replica in volume["replicas"]
if not replica["hostId"]]) == 1
# Three replicas in total should still exist.
assert len(volume["replicas"]) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def backupstore_test(client, host_id, volname, size):
bv, b, snap2, data = create_backup(client, volname)
# test restore
restoreName = generate_volume_name()
volume = client.create_volume(name=restoreName,
size=size,
numberOfReplicas=2,
fromBackup=b["url"])
volume = common.wait_for_volume_restoration_completed(client, restoreName)
volume = common.wait_for_volume_detached(client, restoreName)
assert volume["name"] == restoreName
assert volume["size"] == size
assert volume["numberOfReplicas"] == 2
assert volume["state"] == "detached"
assert volume["initialRestorationRequired"] is False
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, restoreName)
check_volume_data(volume, data)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, restoreName)
bv.backupDelete(name=b["name"])
backups = bv.backupList()
found = False
for b in backups:
if b["snapshotName"] == snap2["name"]:
found = True
break
assert not found
volume = wait_for_volume_status(client, volume["name"], "lastBackup", "")
assert volume["lastBackupAt"] == ""
client.delete(volume)
volume = wait_for_volume_delete(client, restoreName)
def test_hard_anti_affinity_detach(client, volume_name): # NOQA
"""
Test that volumes with Hard Anti-Affinity are still able to detach and
reattach to a node properly, even in degraded state.
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_SOFT_ANTI_AFFINITY)
client.update(setting, value="false")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica["name"])
volume = wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
volume.detach()
volume = wait_for_volume_detached(client, volume_name)
assert len(volume["replicas"]) == 2
volume.attach(hostId=host_id)
# Make sure we're still not getting another successful replica.
volume = wait_for_volume_degraded(client, volume_name)
wait_scheduling_failure(client, volume_name)
assert sum([
1 for replica in volume["replicas"]
if replica["running"] and replica["mode"] == "RW"
]) == 2
assert sum([1 for replica in volume["replicas"]
if not replica["hostId"]]) == 1
assert len(volume["replicas"]) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def test_soft_anti_affinity_scheduling(client, volume_name): # NOQA
"""
Test that volumes with Soft Anti-Affinity work as expected.
With Soft Anti-Affinity, a new replica should still be scheduled on a node
with an existing replica, which will result in "Healthy" state but limited
redundancy.
1. Create a volume and attach to the current node
2. Generate and write `data` to the volume.
3. Set `soft anti-affinity` to true
4. Disable current node's scheduling.
5. Remove the replica on the current node
6. Wait for the volume to complete rebuild. Volume should have 3 replicas.
7. Verify `data`
"""
volume = create_and_check_volume(client, volume_name)
host_id = get_self_host_id()
volume.attach(hostId=host_id)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
data = write_volume_random_data(volume)
setting = client.by_id_setting(SETTING_REPLICA_NODE_SOFT_ANTI_AFFINITY)
client.update(setting, value="true")
node = client.by_id_node(host_id)
client.update(node, allowScheduling=False)
replica_names = list(map(lambda replica: replica.name, volume.replicas))
host_replica = get_host_replica(volume, host_id)
volume.replicaRemove(name=host_replica.name)
wait_new_replica_ready(client, volume_name, replica_names)
volume = wait_for_volume_healthy(client, volume_name)
assert len(volume.replicas) == 3
check_volume_data(volume, data)
cleanup_volume(client, volume)
def ha_salvage_test(client, volume_name, base_image=""): # NOQA
volume = create_and_check_volume(client,
volume_name,
2,
base_image=base_image)
host_id = get_self_host_id()
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
assert len(volume["replicas"]) == 2
replica0_name = volume["replicas"][0]["name"]
replica1_name = volume["replicas"][1]["name"]
data = write_volume_random_data(volume)
common.k8s_delete_replica_pods_for_volume(volume_name)
volume = common.wait_for_volume_faulted(client, volume_name)
assert len(volume["replicas"]) == 2
assert volume["replicas"][0]["failedAt"] != ""
assert volume["replicas"][1]["failedAt"] != ""
volume.salvage(names=[replica0_name, replica1_name])
volume = common.wait_for_volume_detached(client, volume_name)
assert len(volume["replicas"]) == 2
assert volume["replicas"][0]["failedAt"] == ""
assert volume["replicas"][1]["failedAt"] == ""
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, data)
cleanup_volume(client, volume)
def engine_live_upgrade_test(client, volume_name, base_image=""): # NOQA
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"]
new_img = wait_for_engine_image_state(client, new_img_name, "ready")
assert new_img["refCount"] == 0
assert new_img["noRefSince"] != ""
default_img = common.get_default_engine_image(client)
default_img_name = default_img["name"]
volume = client.create_volume(name=volume_name,
size=SIZE,
numberOfReplicas=2,
baseImage=base_image)
volume = common.wait_for_volume_detached(client, volume_name)
default_img = wait_for_engine_image_ref_count(client, default_img_name, 1)
assert volume["name"] == volume_name
assert volume["baseImage"] == base_image
original_engine_image = volume["engineImage"]
assert original_engine_image != engine_upgrade_image
host_id = get_self_host_id()
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, 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 = write_volume_random_data(volume)
volume.engineUpgrade(image=engine_upgrade_image)
volume = wait_for_volume_current_image(client, volume_name,
engine_upgrade_image)
engine = get_volume_engine(volume)
assert engine["engineImage"] == engine_upgrade_image
default_img = wait_for_engine_image_ref_count(client, default_img_name, 0)
new_img = wait_for_engine_image_ref_count(client, new_img_name, 1)
count = 0
# old replica may be in deletion process
for replica in volume["replicas"]:
if replica["currentImage"] == engine_upgrade_image:
count += 1
assert count == REPLICA_COUNT
check_volume_data(volume, data)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
assert len(volume["replicas"]) == REPLICA_COUNT
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
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
assert volume["engineImage"] == engine_upgrade_image
assert volume["currentImage"] == engine_upgrade_image
engine = get_volume_engine(volume)
assert engine["engineImage"] == engine_upgrade_image
assert engine["currentImage"] == engine_upgrade_image
for replica in volume["replicas"]:
assert replica["engineImage"] == engine_upgrade_image
assert replica["currentImage"] == engine_upgrade_image
# Make sure detaching didn't somehow interfere with the data.
check_volume_data(volume, data)
volume.engineUpgrade(image=original_engine_image)
volume = wait_for_volume_current_image(client, volume_name,
original_engine_image)
engine = get_volume_engine(volume)
assert engine["engineImage"] == original_engine_image
default_img = wait_for_engine_image_ref_count(client, default_img_name, 1)
new_img = wait_for_engine_image_ref_count(client, new_img_name, 0)
assert volume["engineImage"] == original_engine_image
engine = get_volume_engine(volume)
assert engine["engineImage"] == original_engine_image
count = 0
# old replica may be in deletion process
for replica in volume["replicas"]:
if replica["engineImage"] == original_engine_image:
count += 1
assert count == REPLICA_COUNT
check_volume_data(volume, data)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
assert len(volume["replicas"]) == REPLICA_COUNT
assert volume["engineImage"] == original_engine_image
engine = get_volume_engine(volume)
assert engine["engineImage"] == original_engine_image
for replica in volume["replicas"]:
assert replica["engineImage"] == original_engine_image
client.delete(volume)
wait_for_volume_delete(client, volume_name)
client.delete(new_img)
def snapshot_test(clients, volume_name, base_image): # NOQA
for host_id, client in clients.iteritems():
break
volume = client.create_volume(name=volume_name,
size=SIZE,
numberOfReplicas=2,
baseImage=base_image)
volume = common.wait_for_volume_detached(client, volume_name)
assert volume["name"] == volume_name
assert volume["size"] == SIZE
assert volume["numberOfReplicas"] == 2
assert volume["state"] == "detached"
assert volume["baseImage"] == base_image
lht_hostId = get_self_host_id()
volume = volume.attach(hostId=lht_hostId)
volume = common.wait_for_volume_healthy(client, volume_name)
volume = client.by_id_volume(volume_name)
positions = {}
snap1 = volume.snapshotCreate()
snap2_data = write_volume_random_data(volume, positions)
snap2 = volume.snapshotCreate()
snap3_data = write_volume_random_data(volume, positions)
snap3 = volume.snapshotCreate()
snapshots = volume.snapshotList()
snapMap = {}
for snap in snapshots:
snapMap[snap["name"]] = snap
assert snapMap[snap1["name"]]["name"] == snap1["name"]
assert snapMap[snap1["name"]]["removed"] is False
assert snapMap[snap2["name"]]["name"] == snap2["name"]
assert snapMap[snap2["name"]]["parent"] == snap1["name"]
assert snapMap[snap2["name"]]["removed"] is False
assert snapMap[snap3["name"]]["name"] == snap3["name"]
assert snapMap[snap3["name"]]["parent"] == snap2["name"]
assert snapMap[snap3["name"]]["removed"] is False
volume.snapshotDelete(name=snap3["name"])
check_volume_data(volume, snap3_data)
snapshots = volume.snapshotList(volume=volume_name)
snapMap = {}
for snap in snapshots:
snapMap[snap["name"]] = snap
assert snapMap[snap1["name"]]["name"] == snap1["name"]
assert snapMap[snap1["name"]]["removed"] is False
assert snapMap[snap2["name"]]["name"] == snap2["name"]
assert snapMap[snap2["name"]]["parent"] == snap1["name"]
assert snapMap[snap2["name"]]["removed"] is False
assert snapMap[snap3["name"]]["name"] == snap3["name"]
assert snapMap[snap3["name"]]["parent"] == snap2["name"]
assert len(snapMap[snap3["name"]]["children"]) == 1
assert "volume-head" in snapMap[snap3["name"]]["children"]
assert snapMap[snap3["name"]]["removed"] is True
snap = volume.snapshotGet(name=snap3["name"])
assert snap["name"] == snap3["name"]
assert snap["parent"] == snap3["parent"]
assert len(snap3["children"]) == 1
assert len(snap["children"]) == 1
assert "volume-head" in snap3["children"]
assert "volume-head" in snap["children"]
assert snap["removed"] is True
volume.snapshotRevert(name=snap2["name"])
check_volume_data(volume, snap2_data)
snapshots = volume.snapshotList(volume=volume_name)
snapMap = {}
for snap in snapshots:
snapMap[snap["name"]] = snap
assert snapMap[snap1["name"]]["name"] == snap1["name"]
assert snapMap[snap1["name"]]["removed"] is False
assert snapMap[snap2["name"]]["name"] == snap2["name"]
assert snapMap[snap2["name"]]["parent"] == snap1["name"]
assert "volume-head" in snapMap[snap2["name"]]["children"]
assert snap3["name"] in snapMap[snap2["name"]]["children"]
assert snapMap[snap2["name"]]["removed"] is False
assert snapMap[snap3["name"]]["name"] == snap3["name"]
assert snapMap[snap3["name"]]["parent"] == snap2["name"]
assert len(snapMap[snap3["name"]]["children"]) == 0
assert snapMap[snap3["name"]]["removed"] is True
volume.snapshotDelete(name=snap1["name"])
volume.snapshotDelete(name=snap2["name"])
volume.snapshotPurge()
wait_for_snapshot_purge(volume, snap1["name"], snap3["name"])
snapshots = volume.snapshotList(volume=volume_name)
snapMap = {}
for snap in snapshots:
snapMap[snap["name"]] = snap
assert snap1["name"] not in snapMap
assert snap3["name"] not in snapMap
# it's the parent of volume-head, so it cannot be purged at this time
assert snapMap[snap2["name"]]["name"] == snap2["name"]
assert snapMap[snap2["name"]]["parent"] == ""
assert "volume-head" in snapMap[snap2["name"]]["children"]
assert snapMap[snap2["name"]]["removed"] is True
check_volume_data(volume, snap2_data)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
client.delete(volume)
volume = wait_for_volume_delete(client, volume_name)
volumes = client.list_volume()
assert len(volumes) == 0
def test_setting_toleration():
"""
Test toleration setting
1. Set `taint-toleration` to "key1=value1:NoSchedule; key2:InvalidEffect".
2. Verify the request fails.
3. Create a volume and attach it.
4. Set `taint-toleration` to "key1=value1:NoSchedule; key2:NoExecute".
5. Verify that cannot update toleration setting when any volume is
attached.
6. Generate and write `data1` into the volume.
7. Detach the volume.
8. Set `taint-toleration` to "key1=value1:NoSchedule; key2:NoExecute".
9. Wait for all the Longhorn system components to restart with new
toleration.
10. Verify that UI, manager, and drive deployer don't restart and
don't have new toleration.
11. Attach the volume again and verify the volume `data1`.
12. Generate and write `data2` to the volume.
13. Detach the volume.
14. Clean the `toleration` setting.
15. Wait for all the Longhorn system components to restart with no
toleration.
16. Attach the volume and validate `data2`.
17. Generate and write `data3` to the volume.
"""
client = get_longhorn_api_client() # NOQA
apps_api = get_apps_api_client() # NOQA
core_api = get_core_api_client() # NOQA
count = len(client.list_node())
setting = client.by_id_setting(SETTING_TAINT_TOLERATION)
with pytest.raises(Exception) as e:
client.update(setting,
value="key1=value1:NoSchedule; key2:InvalidEffect")
assert 'invalid effect' in str(e.value)
volume_name = "test-toleration-vol" # NOQA
volume = create_and_check_volume(client, volume_name)
volume.attach(hostId=get_self_host_id())
volume = wait_for_volume_healthy(client, volume_name)
setting_value_str = "key1=value1:NoSchedule; key2:NoExecute"
setting_value_dicts = [
{
"key": "key1",
"value": "value1",
"operator": "Equal",
"effect": "NoSchedule"
},
{
"key": "key2",
"value": None,
"operator": "Exists",
"effect": "NoExecute"
},
]
with pytest.raises(Exception) as e:
client.update(setting, value=setting_value_str)
assert 'cannot modify toleration setting before all volumes are detached' \
in str(e.value)
data1 = write_volume_random_data(volume)
check_volume_data(volume, data1)
volume.detach(hostId="")
wait_for_volume_detached(client, volume_name)
setting = client.update(setting, value=setting_value_str)
assert setting.value == setting_value_str
wait_for_toleration_update(core_api, apps_api, count, setting_value_dicts)
client, node = wait_for_longhorn_node_ready()
volume = client.by_id_volume(volume_name)
volume.attach(hostId=node)
volume = wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, data1)
data2 = write_volume_random_data(volume)
check_volume_data(volume, data2)
volume.detach(hostId="")
wait_for_volume_detached(client, volume_name)
# cleanup
setting_value_str = ""
setting_value_dicts = []
setting = client.by_id_setting(SETTING_TAINT_TOLERATION)
setting = client.update(setting, value=setting_value_str)
assert setting.value == setting_value_str
wait_for_toleration_update(core_api, apps_api, count, setting_value_dicts)
client, node = wait_for_longhorn_node_ready()
volume = client.by_id_volume(volume_name)
volume.attach(hostId=node)
volume = wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, data2)
data3 = write_volume_random_data(volume)
check_volume_data(volume, data3)
cleanup_volume(client, volume)
def test_setting_priority_class(core_api, apps_api, scheduling_api,
priority_class, volume_name): # NOQA
"""
Test that the Priority Class setting is validated and utilized correctly.
1. Verify that the name of a non-existent Priority Class cannot be used
for the Setting.
2. Create a new Priority Class in Kubernetes.
3. Create and attach a Volume.
4. Verify that the Priority Class Setting cannot be updated with an
attached Volume.
5. Generate and write `data1`.
6. Detach the Volume.
7. Update the Priority Class Setting to the new Priority Class.
8. Wait for all the Longhorn system components to restart with the new
Priority Class.
9. Verify that UI, manager, and drive deployer don't have Priority Class
10. Attach the Volume and verify `data1`.
11. Generate and write `data2`.
12. Unset the Priority Class Setting.
13. Wait for all the Longhorn system components to restart with the new
Priority Class.
14. Verify that UI, manager, and drive deployer don't have Priority Class
15. Attach the Volume and verify `data2`.
16. Generate and write `data3`.
Note: system components are workloads other than UI, manager, driver
deployer
"""
client = get_longhorn_api_client() # NOQA
count = len(client.list_node())
name = priority_class['metadata']['name']
setting = client.by_id_setting(SETTING_PRIORITY_CLASS)
with pytest.raises(Exception) as e:
client.update(setting, value=name)
assert 'failed to get priority class ' in str(e.value)
scheduling_api.create_priority_class(priority_class)
volume = create_and_check_volume(client, volume_name)
volume.attach(hostId=get_self_host_id())
volume = wait_for_volume_healthy(client, volume_name)
with pytest.raises(Exception) as e:
client.update(setting, value=name)
assert 'cannot modify priority class setting before all volumes are ' \
'detached' in str(e.value)
data1 = write_volume_random_data(volume)
check_volume_data(volume, data1)
volume.detach(hostId="")
wait_for_volume_detached(client, volume_name)
setting = client.update(setting, value=name)
assert setting.value == name
wait_for_priority_class_update(core_api, apps_api, count, priority_class)
client, node = wait_for_longhorn_node_ready()
volume = client.by_id_volume(volume_name)
volume.attach(hostId=node)
volume = wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, data1)
data2 = write_volume_random_data(volume)
check_volume_data(volume, data2)
volume.detach(hostId="")
wait_for_volume_detached(client, volume_name)
setting = client.by_id_setting(SETTING_PRIORITY_CLASS)
setting = client.update(setting, value='')
assert setting.value == ''
wait_for_priority_class_update(core_api, apps_api, count)
client, node = wait_for_longhorn_node_ready()
volume = client.by_id_volume(volume_name)
volume.attach(hostId=node)
volume = wait_for_volume_healthy(client, volume_name)
check_volume_data(volume, data2)
data3 = write_volume_random_data(volume)
check_volume_data(volume, data3)
cleanup_volume(client, volume)
def test_instance_manager_cpu_reservation(client, core_api): # NOQA
"""
Test if the CPU requests of instance manager pods are controlled by
the settings and the node specs correctly.
1. Try to change the deprecated setting `Guaranteed Engine CPU`.
--> The setting update should fail.
2. Pick up node 1, set `node.engineManagerCPURequest` and
`node.replicaManagerCPURequest` to 150 and 250, respectively.
--> The IM pods on this node will be restarted. And the CPU requests
of these IM pods matches the above milli value.
3. Change the new settings `Guaranteed Engine Manager CPU` and
`Guaranteed Replica Manager CPU` to 10 and 20, respectively.
Then wait for all IM pods except for the pods on node 1 restarting.
--> The CPU requests of the restarted IM pods equals to
the new setting value multiply the kube node allocatable CPU.
4. Set the both new settings to 0.
--> All IM pods except for the pod on node 1 will be restarted without
CPU requests.
5. Set the fields on node 1 to 0.
--> The IM pods on node 1 will be restarted without CPU requests.
6. Set the both new settings to 2 random values,
and the sum of the 2 values is small than 40.
Then wait for all IM pods restarting.
--> The CPU requests of all IM pods equals to
the new setting value multiply the kube node allocatable CPU.
7. Set the both new settings to 2 random values,
and the single value or the sum of the 2 values is greater than 40.
--> The setting update should fail.
8. Create a volume, verify everything works as normal
Note: use fixture to restore the setting into the original state
"""
instance_managers = client.list_instance_manager()
deprecated_setting = client.by_id_setting(SETTING_GUARANTEED_ENGINE_CPU)
with pytest.raises(Exception) as e:
client.update(deprecated_setting, value="0.1")
host_node_name = get_self_host_id()
host_node = client.by_id_node(host_node_name)
other_ems, other_rms = [], []
for im in instance_managers:
if im.managerType == "engine":
if im.nodeID == host_node_name:
em_on_host = im
else:
other_ems.append(im)
else:
if im.nodeID == host_node_name:
rm_on_host = im
else:
other_rms.append(im)
assert em_on_host and rm_on_host
host_kb_node = core_api.read_node(host_node_name)
if host_kb_node.status.allocatable["cpu"].endswith('m'):
allocatable_millicpu = int(host_kb_node.status.allocatable["cpu"][:-1])
else:
allocatable_millicpu = int(
host_kb_node.status.allocatable["cpu"]) * 1000
client.update(host_node,
allowScheduling=True,
engineManagerCPURequest=150,
replicaManagerCPURequest=250)
time.sleep(5)
guaranteed_engine_cpu_setting_check(client, core_api, [em_on_host],
"Running", True, "150m")
guaranteed_engine_cpu_setting_check(client, core_api, [rm_on_host],
"Running", True, "250m")
em_setting = client.by_id_setting(SETTING_GUARANTEED_ENGINE_MANAGER_CPU)
client.update(em_setting, value="10")
rm_setting = client.by_id_setting(SETTING_GUARANTEED_REPLICA_MANAGER_CPU)
client.update(rm_setting, value="20")
time.sleep(5)
guaranteed_engine_cpu_setting_check(
client, core_api, other_ems, "Running", True,
str(int(allocatable_millicpu * 10 / 100)) + "m")
guaranteed_engine_cpu_setting_check(
client, core_api, other_rms, "Running", True,
str(int(allocatable_millicpu * 20 / 100)) + "m")
em_setting = client.by_id_setting(SETTING_GUARANTEED_ENGINE_MANAGER_CPU)
client.update(em_setting, value="0")
rm_setting = client.by_id_setting(SETTING_GUARANTEED_REPLICA_MANAGER_CPU)
client.update(rm_setting, value="0")
time.sleep(5)
guaranteed_engine_cpu_setting_check(client, core_api, other_ems, "Running",
True, "")
guaranteed_engine_cpu_setting_check(client, core_api, other_rms, "Running",
True, "")
ems, rms = other_ems, other_rms
ems.append(em_on_host)
rms.append(rm_on_host)
host_node = client.by_id_node(host_node_name)
client.update(host_node,
allowScheduling=True,
engineManagerCPURequest=0,
replicaManagerCPURequest=0)
time.sleep(5)
guaranteed_engine_cpu_setting_check(client, core_api, ems, "Running", True,
"")
guaranteed_engine_cpu_setting_check(client, core_api, rms, "Running", True,
"")
client.update(em_setting, value="20")
rm_setting = client.by_id_setting(SETTING_GUARANTEED_REPLICA_MANAGER_CPU)
client.update(rm_setting, value="15")
time.sleep(5)
guaranteed_engine_cpu_setting_check(
client, core_api, ems, "Running", True,
str(int(allocatable_millicpu * 20 / 100)) + "m")
guaranteed_engine_cpu_setting_check(
client, core_api, rms, "Running", True,
str(int(allocatable_millicpu * 15 / 100)) + "m")
with pytest.raises(Exception) as e:
client.update(em_setting, value="41")
assert "should be between 0 to 40" in \
str(e.value)
em_setting = client.by_id_setting(SETTING_GUARANTEED_ENGINE_MANAGER_CPU)
with pytest.raises(Exception) as e:
client.update(em_setting, value="35")
assert "The sum should not be smaller than 0% or greater than 40%" in \
str(e.value)
# Create a volume to test
vol_name = generate_volume_name()
volume = create_and_check_volume(client, vol_name)
volume.attach(hostId=get_self_host_id())
volume = wait_for_volume_healthy(client, vol_name)
assert len(volume.replicas) == 3
data = write_volume_random_data(volume)
check_volume_data(volume, data)
cleanup_volume(client, volume)
def engine_live_upgrade_rollback_test(client,
volume_name,
base_image=""): # NOQA
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"]
wrong_engine_upgrade_image = common.get_compatibility_test_image(
cli_v, cli_minv, ctl_v, ctl_minv, data_v, data_minv)
new_img = client.create_engine_image(image=wrong_engine_upgrade_image)
new_img_name = new_img["name"]
new_img = wait_for_engine_image_state(client, new_img_name, "ready")
assert new_img["refCount"] == 0
assert new_img["noRefSince"] != ""
default_img = common.get_default_engine_image(client)
default_img_name = default_img["name"]
volume = client.create_volume(name=volume_name,
size=SIZE,
numberOfReplicas=2,
baseImage=base_image)
volume = common.wait_for_volume_detached(client, volume_name)
default_img = wait_for_engine_image_ref_count(client, default_img_name, 1)
assert volume["baseImage"] == base_image
original_engine_image = volume["engineImage"]
assert original_engine_image != wrong_engine_upgrade_image
host_id = get_self_host_id()
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
data = write_volume_random_data(volume)
volume.engineUpgrade(image=wrong_engine_upgrade_image)
volume = client.by_id_volume(volume["name"])
assert volume["engineImage"] == wrong_engine_upgrade_image
assert volume["currentImage"] == original_engine_image
with pytest.raises(Exception):
# this will timeout
wait_for_volume_current_image(client, volume_name,
wrong_engine_upgrade_image)
# rollback
volume.engineUpgrade(image=original_engine_image)
volume = wait_for_volume_current_image(client, volume_name,
original_engine_image)
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
volume = common.wait_for_volume_replica_count(client, volume_name,
REPLICA_COUNT)
check_volume_data(volume, data)
assert volume["state"] == common.VOLUME_STATE_ATTACHED
assert volume["robustness"] == common.VOLUME_ROBUSTNESS_HEALTHY
# try again, this time let's try detach
volume.engineUpgrade(image=wrong_engine_upgrade_image)
volume = client.by_id_volume(volume["name"])
assert volume["engineImage"] == wrong_engine_upgrade_image
assert volume["currentImage"] == original_engine_image
with pytest.raises(Exception):
# this will timeout
wait_for_volume_current_image(client, volume_name,
wrong_engine_upgrade_image)
volume = volume.detach()
volume = wait_for_volume_current_image(client, volume_name,
wrong_engine_upgrade_image)
# all the images would be updated
assert volume["engineImage"] == wrong_engine_upgrade_image
engine = get_volume_engine(volume)
assert engine["engineImage"] == wrong_engine_upgrade_image
volume = common.wait_for_volume_replica_count(client, volume_name,
REPLICA_COUNT)
for replica in volume["replicas"]:
assert replica["engineImage"] == wrong_engine_upgrade_image
# upgrade to the correct image when offline
volume.engineUpgrade(image=original_engine_image)
volume = client.by_id_volume(volume["name"])
assert volume["engineImage"] == original_engine_image
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, 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
check_volume_data(volume, data)
client.delete(volume)
wait_for_volume_delete(client, volume_name)
client.delete(new_img)
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 backing_image_content_test(client, volume_name_prefix, bi_name, bi_url): # NOQA
"""
Verify the content of the Backing Image is accessible and read-only for
all volumes.
1. Create a backing image. (Done by the caller)
2. Create a Volume with the backing image set then attach it to host node.
3. Verify that the all disk states in the backing image are "downloaded".
4. Verify volume can be directly mounted and there is already data in the
filesystem due to the backing image.
5. Verify the volume r/w.
6. Launch one more volume with the same backing image.
7. Verify the data content of the new volume is the same as the data in
step 4.
5. Do cleanup. (Done by the caller)
"""
lht_host_id = get_self_host_id()
volume_name1 = volume_name_prefix + "-1"
volume1 = create_and_check_volume(
client, volume_name1, 3,
str(BACKING_IMAGE_EXT4_SIZE), bi_name)
volume1.attach(hostId=lht_host_id)
volume1 = wait_for_volume_healthy(client, volume_name1)
assert volume1.backingImage == bi_name
assert volume1.size == str(BACKING_IMAGE_EXT4_SIZE)
backing_image = client.by_id_backing_image(bi_name)
assert backing_image.imageURL == bi_url
assert not backing_image.deletionTimestamp
assert len(backing_image.diskStateMap) == 3
for disk_id, state in iter(backing_image.diskStateMap.items()):
assert state == "downloaded"
# Since there is already a filesystem with data in the backing image,
# we can directly mount and access the volume without `mkfs`.
dev1 = get_volume_endpoint(volume1)
mount_path1 = os.path.join(DIRECTORY_PATH, volume_name1)
mount_disk(dev1, mount_path1)
output1 = subprocess.check_output(["ls", mount_path1])
# The following random write may crash the filesystem of volume1,
# need to umount it here
cleanup_host_disk(volume_name1)
# Verify r/w for the volume with a backing image
data = write_volume_random_data(volume1)
check_volume_data(volume1, data)
volume_name2 = volume_name_prefix + "-2"
volume2 = create_and_check_volume(
client, volume_name2, 3,
str(BACKING_IMAGE_EXT4_SIZE), bi_name)
volume2.attach(hostId=lht_host_id)
volume2 = wait_for_volume_healthy(client, volume_name2)
assert volume1.backingImage == bi_name
assert volume1.size == str(BACKING_IMAGE_EXT4_SIZE)
dev2 = get_volume_endpoint(volume2)
mount_path2 = os.path.join(DIRECTORY_PATH, volume_name2)
mount_disk(dev2, mount_path2)
output2 = subprocess.check_output(["ls", mount_path2])
# The output is the content of the backing image, which should keep
# unchanged
assert output2 == output1
cleanup_host_disk(volume_name2)
def engine_offline_upgrade_test(client, volume_name, base_image=""): # NOQA
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"]
new_img = wait_for_engine_image_state(client, new_img_name, "ready")
assert new_img["refCount"] == 0
assert new_img["noRefSince"] != ""
default_img = common.get_default_engine_image(client)
default_img_name = default_img["name"]
volume = client.create_volume(name=volume_name,
size=SIZE,
numberOfReplicas=REPLICA_COUNT,
baseImage=base_image)
volume = common.wait_for_volume_detached(client, volume_name)
default_img = wait_for_engine_image_ref_count(client, default_img_name, 1)
original_engine_image = default_img["image"]
assert volume["name"] == volume_name
assert volume["engineImage"] == original_engine_image
assert volume["currentImage"] == original_engine_image
assert volume["baseImage"] == base_image
# Before our upgrade, write data to the volume first.
host_id = get_self_host_id()
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
data = write_volume_random_data(volume)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
volume.engineUpgrade(image=engine_upgrade_image)
volume = wait_for_volume_current_image(client, volume_name,
engine_upgrade_image)
default_img = wait_for_engine_image_ref_count(client, default_img_name, 0)
new_img = wait_for_engine_image_ref_count(client, new_img_name, 1)
# cannot delete a image in use
with pytest.raises(Exception) as e:
client.delete(new_img)
assert "while being used" in str(e.value)
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
engine = get_volume_engine(volume)
assert engine["engineImage"] == engine_upgrade_image
assert engine["currentImage"] == engine_upgrade_image
for replica in volume["replicas"]:
assert replica["engineImage"] == engine_upgrade_image
assert replica["currentImage"] == engine_upgrade_image
check_volume_data(volume, data)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
volume.engineUpgrade(image=original_engine_image)
volume = wait_for_volume_current_image(client, volume_name,
original_engine_image)
engine = get_volume_engine(volume)
assert volume["engineImage"] == original_engine_image
assert engine["engineImage"] == original_engine_image
for replica in volume["replicas"]:
assert replica["engineImage"] == original_engine_image
default_img = wait_for_engine_image_ref_count(client, default_img_name, 1)
new_img = wait_for_engine_image_ref_count(client, new_img_name, 0)
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
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
check_volume_data(volume, data)
client.delete(volume)
wait_for_volume_delete(client, volume_name)
client.delete(new_img)
def ha_backup_deletion_recovery_test(client, volume_name, size, base_image=""): # NOQA
volume = client.create_volume(name=volume_name, size=size,
numberOfReplicas=2, baseImage=base_image)
volume = common.wait_for_volume_detached(client, volume_name)
host_id = get_self_host_id()
volume = volume.attach(hostId=host_id)
volume = common.wait_for_volume_healthy(client, volume_name)
setting = client.by_id_setting(common.SETTING_BACKUP_TARGET)
# test backupTarget for multiple settings
backupstores = common.get_backupstore_url()
for backupstore in backupstores:
if common.is_backupTarget_s3(backupstore):
backupsettings = backupstore.split("$")
setting = client.update(setting, value=backupsettings[0])
assert setting["value"] == backupsettings[0]
credential = client.by_id_setting(
common.SETTING_BACKUP_TARGET_CREDENTIAL_SECRET)
credential = client.update(credential, value=backupsettings[1])
assert credential["value"] == backupsettings[1]
else:
setting = client.update(setting, value=backupstore)
assert setting["value"] == backupstore
credential = client.by_id_setting(
common.SETTING_BACKUP_TARGET_CREDENTIAL_SECRET)
credential = client.update(credential, value="")
assert credential["value"] == ""
data = write_volume_random_data(volume)
snap2 = volume.snapshotCreate()
volume.snapshotCreate()
volume.snapshotBackup(name=snap2["name"])
_, b = common.find_backup(client, volume_name, snap2["name"])
res_name = common.generate_volume_name()
res_volume = client.create_volume(name=res_name, size=size,
numberOfReplicas=2,
fromBackup=b["url"])
res_volume = common.wait_for_volume_detached(client, res_name)
res_volume = res_volume.attach(hostId=host_id)
res_volume = common.wait_for_volume_healthy(client, res_name)
check_volume_data(res_volume, data)
snapshots = res_volume.snapshotList()
# only the backup snapshot + volume-head
assert len(snapshots) == 2
backup_snapshot = ""
for snap in snapshots:
if snap["name"] != "volume-head":
backup_snapshot = snap["name"]
assert backup_snapshot != ""
res_volume.snapshotCreate()
snapshots = res_volume.snapshotList()
assert len(snapshots) == 3
res_volume.snapshotDelete(name=backup_snapshot)
res_volume.snapshotPurge()
snapshots = res_volume.snapshotList()
assert len(snapshots) == 2
ha_rebuild_replica_test(client, res_name)
res_volume = res_volume.detach()
res_volume = common.wait_for_volume_detached(client, res_name)
client.delete(res_volume)
common.wait_for_volume_delete(client, res_name)
volume = volume.detach()
volume = common.wait_for_volume_detached(client, volume_name)
client.delete(volume)
common.wait_for_volume_delete(client, volume_name)
volumes = client.list_volume()
assert len(volumes) == 0
