def backup_test(
dev,
address, # NOQA
volume_name,
engine_name,
backup_target):
offset = 0
length = 128
snap1_data = random_string(length)
verify_data(dev, offset, snap1_data)
snap1_checksum = checksum_dev(dev)
snap1 = cmd.snapshot_create(address)
backup1_info = create_backup(address, snap1, backup_target)
assert backup1_info["VolumeName"] == volume_name
assert backup1_info["Size"] == BLOCK_SIZE_STR
snap2_data = random_string(length)
verify_data(dev, offset, snap2_data)
snap2_checksum = checksum_dev(dev)
snap2 = cmd.snapshot_create(address)
backup2_info = create_backup(address, snap2, backup_target)
assert backup2_info["VolumeName"] == volume_name
assert backup2_info["Size"] == BLOCK_SIZE_STR
snap3_data = random_string(length)
verify_data(dev, offset, snap3_data)
snap3_checksum = checksum_dev(dev)
snap3 = cmd.snapshot_create(address)
backup3_info = create_backup(address, snap3, backup_target)
assert backup3_info["VolumeName"] == volume_name
assert backup3_info["Size"] == BLOCK_SIZE_STR
restore_with_frontend(address, engine_name, backup3_info["URL"])
readed = read_dev(dev, offset, length)
assert readed == snap3_data
c = checksum_dev(dev)
assert c == snap3_checksum
rm_backups(address, engine_name, [backup3_info["URL"]])
restore_with_frontend(address, engine_name, backup1_info["URL"])
readed = read_dev(dev, offset, length)
assert readed == snap1_data
c = checksum_dev(dev)
assert c == snap1_checksum
rm_backups(address, engine_name, [backup1_info["URL"]])
restore_with_frontend(address, engine_name, backup2_info["URL"])
readed = read_dev(dev, offset, length)
assert readed == snap2_data
c = checksum_dev(dev)
assert c == snap2_checksum
rm_backups(address, engine_name, [backup2_info["URL"]])
def test_ha_remove_extra_disks(
grpc_controller, # NOQA
grpc_replica1,
grpc_replica2): # NOQA
address = grpc_controller.address
prepare_backup_dir(BACKUP_DIR)
open_replica(grpc_replica1)
replicas = grpc_controller.replica_list()
assert len(replicas) == 0
r1_url = grpc_replica1.url
v = grpc_controller.volume_start(replicas=[r1_url])
assert v.name == VOLUME_NAME
assert v.replicaCount == 1
replicas = grpc_controller.replica_list()
assert len(replicas) == 1
assert replicas[0].mode == "RW"
dev = get_blockdev(VOLUME_NAME)
wasted_data = random_string(128)
data_offset = 1024
verify_data(dev, data_offset, wasted_data)
# now replica1 contains extra data in a snapshot
cmd.snapshot_create(address)
cleanup_controller(grpc_controller)
open_replica(grpc_replica2)
replicas = grpc_controller.replica_list()
assert len(replicas) == 0
r2_url = grpc_replica2.url
v = grpc_controller.volume_start(replicas=[r2_url])
assert v.name == VOLUME_NAME
assert v.replicaCount == 1
replicas = grpc_controller.replica_list()
assert len(replicas) == 1
assert replicas[0].mode == "RW"
dev = get_blockdev(VOLUME_NAME)
data = random_string(128)
data_offset = 1024
verify_data(dev, data_offset, data)
r1 = grpc_replica1.replica_reload()
print(r1)
cmd.add_replica(address, r1_url)
wait_for_rebuild_complete(address)
verify_data(dev, data_offset, data)
def backup_hole_with_backing_file_test(
backup_target, # NOQA
grpc_backing_controller, # NOQA
grpc_backing_replica1, # NOQA
grpc_backing_replica2): # NOQA
address = grpc_backing_controller.address
dev = get_backing_dev(grpc_backing_replica1, grpc_backing_replica2,
grpc_backing_controller)
offset1 = 512
length1 = 256
offset2 = 640
length2 = 256
boundary_offset = 0
boundary_length = 4100 # just pass 4096 into next 4k
hole_offset = 2 * 1024 * 1024
hole_length = 1024
snap1_data = random_string(length1)
verify_data(dev, offset1, snap1_data)
snap1_checksum = checksum_dev(dev)
snap1 = cmd.snapshot_create(address)
boundary_data_backup1 = read_dev(dev, boundary_offset, boundary_length)
hole_data_backup1 = read_dev(dev, hole_offset, hole_length)
backup1_info = create_backup(address, snap1, backup_target)
snap2_data = random_string(length2)
verify_data(dev, offset2, snap2_data)
snap2_checksum = checksum_dev(dev)
snap2 = cmd.snapshot_create(address)
boundary_data_backup2 = read_dev(dev, boundary_offset, boundary_length)
hole_data_backup2 = read_dev(dev, hole_offset, hole_length)
backup2_info = create_backup(address, snap2, backup_target)
restore_with_frontend(address, ENGINE_BACKING_NAME, backup1_info["URL"])
readed = read_dev(dev, boundary_offset, boundary_length)
assert readed == boundary_data_backup1
readed = read_dev(dev, hole_offset, hole_length)
assert readed == hole_data_backup1
c = checksum_dev(dev)
assert c == snap1_checksum
restore_with_frontend(address, ENGINE_BACKING_NAME, backup2_info["URL"])
readed = read_dev(dev, boundary_offset, boundary_length)
assert readed == boundary_data_backup2
readed = read_dev(dev, hole_offset, hole_length)
assert readed == hole_data_backup2
c = checksum_dev(dev)
assert c == snap2_checksum
def test_basic_rw(dev): # NOQA
for i in range(0, 10):
base = random.randint(1, SIZE - PAGE_SIZE)
offset = (base // PAGE_SIZE) * PAGE_SIZE
length = base - offset
data = random_string(length)
verify_data(dev, offset, data)
def test_backup_volume_deletion(
grpc_replica1,
grpc_replica2, # NOQA
grpc_controller,
backup_targets): # NOQA
offset = 0
length = 128
address = grpc_controller.address
for backup_target in backup_targets:
dev = get_dev(grpc_replica1, grpc_replica2, grpc_controller)
snap_data = random_string(length)
verify_data(dev, offset, snap_data)
snap = cmd.snapshot_create(address)
backup_info = create_backup(address, snap, backup_target)
assert backup_info["VolumeName"] == VOLUME_NAME
assert backup_info["Size"] == BLOCK_SIZE_STR
assert snap in backup_info["SnapshotName"]
cmd.backup_volume_rm(address, VOLUME_NAME, backup_target)
info = cmd.backup_volume_list(address, VOLUME_NAME, backup_target)
assert "cannot find" in info[VOLUME_NAME]["Messages"]["error"]
cmd.sync_agent_server_reset(address)
cleanup_controller(grpc_controller)
cleanup_replica(grpc_replica1)
cleanup_replica(grpc_replica2)
def test_rw_with_metric(
grpc_controller, # NOQA
grpc_replica1,
grpc_replica2): # NOQA
rw_dev = get_dev(grpc_replica1, grpc_replica2, grpc_controller)
replies = grpc_controller.metric_get()
# skip the first metric since its fields are 0
next(replies).metric
for i in range(0, 5):
base = random.randint(1, SIZE - PAGE_SIZE)
offset = (base // PAGE_SIZE) * PAGE_SIZE
length = base - offset
data = random_string(length)
verify_data(rw_dev, offset, data)
while 1:
try:
metric = next(replies).metric
# it's hard to confirm the accurate value of metric
assert metric.readBandwidth != 0
assert metric.writeBandwidth != 0
assert metric.iOPS != 0
break
except StopIteration:
time.sleep(1)
def test_frontend_switch(
grpc_controller_no_frontend, # NOQA
grpc_replica1,
grpc_replica2): # NOQA
open_replica(grpc_replica1)
open_replica(grpc_replica2)
replicas = grpc_controller_no_frontend.replica_list()
assert len(replicas) == 0
r1_url = grpc_replica1.url
r2_url = grpc_replica2.url
v = grpc_controller_no_frontend.volume_start(replicas=[r1_url, r2_url])
assert v.name == VOLUME_NO_FRONTEND_NAME
assert v.replicaCount == 2
assert v.frontend == ""
start_engine_frontend(ENGINE_NO_FRONTEND_NAME)
dev = get_blockdev(volume=VOLUME_NO_FRONTEND_NAME)
data = random_string(128)
data_offset = 1024
verify_data(dev, data_offset, data)
shutdown_engine_frontend(ENGINE_NO_FRONTEND_NAME)
start_engine_frontend(ENGINE_NO_FRONTEND_NAME)
dev = get_blockdev(volume=VOLUME_NO_FRONTEND_NAME)
verify_read(dev, data_offset, data)
shutdown_engine_frontend(ENGINE_NO_FRONTEND_NAME)
def test_ha_single_replica_failure(
grpc_controller, # NOQA
grpc_replica1,
grpc_replica2): # NOQA
open_replica(grpc_replica1)
open_replica(grpc_replica2)
replicas = grpc_controller.replica_list()
assert len(replicas) == 0
r1_url = grpc_replica1.url
r2_url = grpc_replica2.url
v = grpc_controller.volume_start(replicas=[r1_url, r2_url])
assert v.replicaCount == 2
replicas = grpc_controller.replica_list()
assert len(replicas) == 2
assert replicas[0].mode == "RW"
assert replicas[1].mode == "RW"
dev = get_blockdev(VOLUME_NAME)
data = random_string(128)
data_offset = 1024
verify_data(dev, data_offset, data)
cleanup_replica(grpc_replica2)
verify_async(dev, 10, 128, 1)
verify_replica_state(grpc_controller, 1, "ERR")
verify_read(dev, data_offset, data)
def test_snapshot_tree_rebuild(
grpc_controller, # NOQA
grpc_replica1,
grpc_replica2): # NOQA
address = grpc_controller.address
offset = 0
length = 128
open_replica(grpc_replica1)
open_replica(grpc_replica2)
replicas = grpc_controller.replica_list()
assert len(replicas) == 0
r1_url = grpc_replica1.url
r2_url = grpc_replica2.url
v = grpc_controller.volume_start(replicas=[r1_url, r2_url])
assert v.name == VOLUME_NAME
assert v.replicaCount == 2
replicas = grpc_controller.replica_list()
assert len(replicas) == 2
assert replicas[0].mode == "RW"
assert replicas[1].mode == "RW"
dev = get_blockdev(VOLUME_NAME)
snap, snap_data = snapshot_tree_build(dev, address, ENGINE_NAME, offset,
length)
data = random_string(128)
data_offset = 1024
verify_data(dev, data_offset, data)
# Cleanup replica2
cleanup_replica(grpc_replica2)
verify_async(dev, 10, 128, 1)
verify_replica_state(grpc_controller, 1, "ERR")
verify_read(dev, data_offset, data)
grpc_controller.replica_delete(replicas[1].address)
# Rebuild replica2
open_replica(grpc_replica2)
cmd.add_replica(address, r2_url)
wait_for_rebuild_complete(address)
verify_async(dev, 10, 128, 1)
verify_replica_state(grpc_controller, 1, "RW")
snapshot_tree_verify(dev, address, ENGINE_NAME, offset, length, snap,
snap_data)
def test_beyond_boundary(dev): # NOQA
# check write at the boundary
data = random_string(128)
verify_data(dev, SIZE - 1 - 128, data)
# out of bounds
with pytest.raises(EnvironmentError) as err:
write_dev(dev, SIZE, "1")
assert 'No space left' in str(err.value)
assert len(read_dev(dev, SIZE, 1)) == 0
# normal writes to verify controller/replica survival
test_basic_rw(dev)
def restore_to_file_with_backing_file_test(
backup_target, # NOQA
grpc_backing_controller, # NOQA
grpc_backing_replica1, # NOQA
grpc_backing_replica2): # NOQA
address = grpc_backing_controller.address
backing_dev = get_backing_dev(grpc_backing_replica1, grpc_backing_replica2,
grpc_backing_controller)
length0 = 4 * 1024
length1 = 256
length2 = 128
offset0 = 0
offset1 = length1 + offset0
offset2 = length2 + offset0
output_raw_path = file(OUTPUT_FILE_RAW)
output_qcow2_path = file(OUTPUT_FILE_QCOW2)
# create 1 empty snapshot.
# data in output image == data in backing
check_backing()
check_empty_volume(backing_dev)
snap0 = cmd.snapshot_create(address)
backup = create_backup(address, snap0, backup_target)["URL"]
volume_data = read_dev(backing_dev, offset0, length0)
backing_data = read_from_backing_file(offset0, length0)
dev_checksum = checksum_dev(backing_dev)
assert volume_data != ""
assert volume_data == backing_data
cmd.restore_to_file(address, backup, file(BACKING_FILE_QCOW),
output_raw_path, IMAGE_FORMAT_RAW)
output0_raw = read_file(output_raw_path, offset0, length0)
output0_checksum = checksum_data(
read_file(output_raw_path, 0, SIZE).encode('utf-8'))
assert output0_raw == backing_data
assert output0_checksum == dev_checksum
os.remove(output_raw_path)
assert not os.path.exists(output_raw_path)
cmd.restore_to_file(address, backup, file(BACKING_FILE_QCOW),
output_qcow2_path, IMAGE_FORMAT_QCOW2)
output0_qcow2 = read_qcow2_file_without_backing_file(
output_qcow2_path, offset0, length0)
output0_checksum = checksum_data(
read_qcow2_file_without_backing_file(output_qcow2_path, 0, SIZE))
assert output0_qcow2.decode('utf-8') == backing_data
assert output0_qcow2.decode('utf-8') == volume_data
assert output0_checksum == dev_checksum
os.remove(output_qcow2_path)
assert not os.path.exists(output_qcow2_path)
rm_backups(address, ENGINE_BACKING_NAME, [backup])
# create 1 snapshot with 256B data.
# output = snap1(offset0, length1) + backing(offset1, ...)
snap1_data = random_string(length1)
verify_data(backing_dev, offset0, snap1_data)
snap1 = cmd.snapshot_create(address)
backup = create_backup(address, snap1, backup_target)["URL"]
volume_data = read_dev(backing_dev, offset0, length0)
backing_data = read_from_backing_file(offset1, length0 - offset1)
dev_checksum = checksum_dev(backing_dev)
cmd.restore_to_file(address, backup, file(BACKING_FILE_QCOW),
output_raw_path, IMAGE_FORMAT_RAW)
output1_raw_snap1 = read_file(output_raw_path, offset0, length1)
output1_raw_backing = read_file(output_raw_path, offset1,
length0 - offset1)
output1_checksum = checksum_data(
read_file(output_raw_path, 0, SIZE).encode('utf-8'))
assert output1_raw_snap1 == snap1_data
assert output1_raw_backing == backing_data
assert output1_raw_snap1 + output1_raw_backing == volume_data
assert output1_checksum == dev_checksum
os.remove(output_raw_path)
assert not os.path.exists(output_raw_path)
cmd.restore_to_file(address, backup, file(BACKING_FILE_QCOW),
output_qcow2_path, IMAGE_FORMAT_QCOW2)
output1_qcow2_snap1 = read_qcow2_file_without_backing_file(
output_qcow2_path, offset0, length1)
output1_qcow2_backing = read_qcow2_file_without_backing_file(
output_qcow2_path, offset1, length0 - offset1)
output1_checksum = checksum_data(
read_qcow2_file_without_backing_file(output_qcow2_path, 0, SIZE))
assert output1_qcow2_snap1.decode('utf-8') == snap1_data
assert output1_qcow2_backing.decode('utf-8') == backing_data
assert output1_qcow2_snap1.decode('utf-8') + \
output1_qcow2_backing.decode('utf-8') == volume_data
assert output1_checksum == dev_checksum
os.remove(output_qcow2_path)
assert not os.path.exists(output_qcow2_path)
snapshot_revert_with_frontend(address, ENGINE_BACKING_NAME, snap0)
rm_snaps(address, [snap1])
rm_backups(address, ENGINE_BACKING_NAME, [backup])
check_backing()
check_empty_volume(backing_dev)
# create 2 snapshots with 256B data and 128B data
# output = snap2(offset0, length1 - length2) +
# snap1(offset2, length2) + backing(offset2, ...)
snap1_data = random_string(length1)
verify_data(backing_dev, offset0, snap1_data)
snap1 = cmd.snapshot_create(address)
snap2_data = random_string(length2)
verify_data(backing_dev, offset0, snap2_data)
snap2 = cmd.snapshot_create(address)
backup = create_backup(address, snap2, backup_target)["URL"]
volume_data = read_dev(backing_dev, offset0, length0)
backing_data = read_from_backing_file(offset1, length0 - offset1)
dev_checksum = checksum_dev(backing_dev)
cmd.restore_to_file(address, backup, file(BACKING_FILE_QCOW),
output_raw_path, IMAGE_FORMAT_RAW)
output2_raw_snap2 = read_file(output_raw_path, offset0, length2)
output2_raw_snap1 = read_file(output_raw_path, offset2, length1 - length2)
output2_raw_backing = read_file(output_raw_path, offset1,
length0 - offset1)
output2_checksum = checksum_data(
read_file(output_raw_path, 0, SIZE).encode('utf-8'))
assert output2_raw_snap2 == snap2_data
assert output2_raw_snap1 == snap1_data[offset2:length1]
assert output2_raw_backing == backing_data
assert \
volume_data == \
output2_raw_snap2 + output2_raw_snap1 + output2_raw_backing
assert output2_checksum == dev_checksum
os.remove(output_raw_path)
assert not os.path.exists(output_raw_path)
cmd.restore_to_file(address, backup, file(BACKING_FILE_QCOW),
output_qcow2_path, IMAGE_FORMAT_QCOW2)
output2_qcow2_snap2 = read_qcow2_file_without_backing_file(
output_qcow2_path, offset0, length2)
output2_qcow2_snap1 = read_qcow2_file_without_backing_file(
output_qcow2_path, offset2, length1 - length2)
output2_qcow2_backing = read_qcow2_file_without_backing_file(
output_qcow2_path, offset1, length0 - offset1)
output2_checksum = checksum_data(
read_qcow2_file_without_backing_file(output_qcow2_path, 0, SIZE))
assert output2_qcow2_snap2.decode('utf-8') == snap2_data
assert output2_qcow2_snap1.decode('utf-8') == snap1_data[offset2:length1]
assert output2_qcow2_backing.decode('utf-8') == backing_data
assert \
volume_data == \
output2_qcow2_snap2.decode('utf-8') + \
output2_qcow2_snap1.decode('utf-8') + \
output1_qcow2_backing.decode('utf-8')
assert output2_checksum == dev_checksum
os.remove(output_qcow2_path)
assert not os.path.exists(output_qcow2_path)
snapshot_revert_with_frontend(address, ENGINE_BACKING_NAME, snap0)
rm_snaps(address, [snap1, snap2])
rm_backups(address, ENGINE_BACKING_NAME, [backup])
check_backing()
check_empty_volume(backing_dev)
def test_ha_double_replica_rebuild(
grpc_controller, # NOQA
grpc_replica1,
grpc_replica2): # NOQA
open_replica(grpc_replica1)
open_replica(grpc_replica2)
replicas = grpc_controller.replica_list()
assert len(replicas) == 0
r1_url = grpc_replica1.url
r2_url = grpc_replica2.url
v = grpc_controller.volume_start(replicas=[r1_url, r2_url])
assert v.name == VOLUME_NAME
assert v.replicaCount == 2
replicas = grpc_controller.replica_list()
assert len(replicas) == 2
assert replicas[0].mode == "RW"
assert replicas[1].mode == "RW"
dev = get_blockdev(VOLUME_NAME)
data1 = random_string(128)
data1_offset = 1024
verify_data(dev, data1_offset, data1)
# Close replica2
r2 = grpc_replica2.replica_get()
assert r2.revisionCounter == 1
grpc_replica2.replica_close()
verify_async(dev, 10, 128, 1)
verify_replica_state(grpc_controller, 1, "ERR")
verify_read(dev, data1_offset, data1)
data2 = random_string(128)
data2_offset = 512
verify_data(dev, data2_offset, data2)
# Close replica1
r1 = grpc_replica1.replica_get()
assert r1.revisionCounter == 12 # 1 + 10 + 1
grpc_replica1.replica_close()
# Restart volume
cleanup_controller(grpc_controller)
replicas = grpc_controller.replica_list()
assert len(replicas) == 0
# NOTE the order is reversed here
r1_url = grpc_replica1.url
r2_url = grpc_replica2.url
v = grpc_controller.volume_start(replicas=[r2_url, r1_url])
assert v.replicaCount == 2
# replica2 is out because of lower revision counter
replicas = grpc_controller.replica_list()
assert len(replicas) == 2
assert replicas[0].mode == "ERR"
assert replicas[1].mode == "RW"
verify_read(dev, data1_offset, data1)
verify_read(dev, data2_offset, data2)
# Rebuild replica2
r2 = grpc_replica2.replica_get()
assert r2.revisionCounter == 1
grpc_replica2.replica_close()
grpc_controller.replica_delete(replicas[0].address)
cmd.add_replica(grpc_controller.address, r2_url)
wait_for_rebuild_complete(grpc_controller.address)
verify_async(dev, 10, 128, 1)
verify_replica_state(grpc_controller, 1, "RW")
verify_read(dev, data1_offset, data1)
verify_read(dev, data2_offset, data2)
r1 = grpc_replica1.replica_get()
r2 = grpc_replica2.replica_get()
assert r1.revisionCounter == 22 # 1 + 10 + 1 + 10
assert r2.revisionCounter == 22 # must be in sync with r1
def test_upgrade(
grpc_engine_manager, # NOQA
grpc_controller, # NOQA
grpc_fixed_dir_replica1,
grpc_fixed_dir_replica2, # NOQA
grpc_extra_replica1,
grpc_extra_replica2): # NOQA
dev = get_dev(grpc_fixed_dir_replica1, grpc_fixed_dir_replica2,
grpc_controller)
offset = 0
length = 128
data = random_string(length)
verify_data(dev, offset, data)
# both set pointed to the same volume underlying
r1_url = grpc_fixed_dir_replica1.url
r2_url = grpc_fixed_dir_replica2.url
upgrade_r1_url = grpc_extra_replica1.url
upgrade_r2_url = grpc_extra_replica2.url
v = grpc_controller.volume_start(replicas=[r1_url, r2_url])
assert v.replicaCount == 2
upgrade_e = grpc_engine_manager.engine_upgrade(
ENGINE_NAME, LONGHORN_UPGRADE_BINARY, SIZE,
[upgrade_r1_url, upgrade_r2_url])
assert upgrade_e.spec.binary == LONGHORN_UPGRADE_BINARY
verify_data(dev, offset, data)
grpc_controller.client_upgrade(upgrade_e.spec.listen)
wait_for_process_running(grpc_engine_manager, ENGINE_NAME,
INSTANCE_MANAGER_TYPE_ENGINE)
# cannot start with same binary
with pytest.raises(grpc.RpcError):
grpc_engine_manager.engine_upgrade(ENGINE_NAME,
LONGHORN_UPGRADE_BINARY, SIZE,
[r1_url, r2_url])
verify_data(dev, offset, data)
# cannot start with wrong replica, would trigger rollback
with pytest.raises(grpc.RpcError):
grpc_engine_manager.engine_upgrade(ENGINE_NAME,
LONGHORN_UPGRADE_BINARY, SIZE,
["random"])
verify_data(dev, offset, data)
grpc_fixed_dir_replica1 = cleanup_replica(grpc_fixed_dir_replica1)
grpc_fixed_dir_replica2 = cleanup_replica(grpc_fixed_dir_replica2)
open_replica(grpc_fixed_dir_replica1)
open_replica(grpc_fixed_dir_replica2)
e = grpc_engine_manager.engine_upgrade(ENGINE_NAME, LONGHORN_BINARY, SIZE,
[r1_url, r2_url])
assert e.spec.binary == LONGHORN_BINARY
verify_data(dev, offset, data)
grpc_controller.client_upgrade(e.spec.listen)
wait_for_process_running(grpc_engine_manager, ENGINE_NAME,
INSTANCE_MANAGER_TYPE_ENGINE)
def restore_inc_test(
grpc_engine_manager, # NOQA
grpc_controller, # NOQA
grpc_replica1,
grpc_replica2, # NOQA
grpc_dr_controller, # NOQA
grpc_dr_replica1,
grpc_dr_replica2, # NOQA
backup_target): # NOQA
address = grpc_controller.address
dev = get_dev(grpc_replica1, grpc_replica2, grpc_controller)
zero_string = b'\x00'.decode('utf-8')
# backup0: 256 random data in 1st block
length0 = 256
snap0_data = random_string(length0)
verify_data(dev, 0, snap0_data)
verify_data(dev, BLOCK_SIZE, snap0_data)
snap0 = cmd.snapshot_create(address)
backup0 = create_backup(address, snap0, backup_target)["URL"]
backup0_name = cmd.backup_inspect(address, backup0)['Name']
# backup1: 32 random data + 32 zero data + 192 random data in 1st block
length1 = 32
offset1 = 32
snap1_data = zero_string * length1
verify_data(dev, offset1, snap1_data)
snap1 = cmd.snapshot_create(address)
backup1 = create_backup(address, snap1, backup_target)["URL"]
backup1_name = cmd.backup_inspect(address, backup1)['Name']
# backup2: 32 random data + 256 random data in 1st block,
# 256 random data in 2nd block
length2 = 256
offset2 = 32
snap2_data = random_string(length2)
verify_data(dev, offset2, snap2_data)
verify_data(dev, BLOCK_SIZE, snap2_data)
snap2 = cmd.snapshot_create(address)
backup2 = create_backup(address, snap2, backup_target)["URL"]
backup2_name = cmd.backup_inspect(address, backup2)['Name']
# backup3: 64 zero data + 192 random data in 1st block
length3 = 64
offset3 = 0
verify_data(dev, offset3, zero_string * length3)
verify_data(dev, length2, zero_string * offset2)
verify_data(dev, BLOCK_SIZE, zero_string * length2)
snap3 = cmd.snapshot_create(address)
backup3 = create_backup(address, snap3, backup_target)["URL"]
backup3_name = cmd.backup_inspect(address, backup3)['Name']
# backup4: 256 random data in 1st block
length4 = 256
offset4 = 0
snap4_data = random_string(length4)
verify_data(dev, offset4, snap4_data)
snap4 = cmd.snapshot_create(address)
backup4 = create_backup(address, snap4, backup_target)["URL"]
backup4_name = cmd.backup_inspect(address, backup4)['Name']
# start no-frontend volume
# start dr volume (no frontend)
dr_address = grpc_dr_controller.address
start_no_frontend_volume(grpc_engine_manager, grpc_dr_controller,
grpc_dr_replica1, grpc_dr_replica2)
cmd.backup_restore(dr_address, backup0)
wait_for_restore_completion(dr_address, backup0)
verify_no_frontend_data(grpc_engine_manager, 0, snap0_data,
grpc_dr_controller)
# mock restore crash/error
delta_file1 = "volume-delta-" + backup0_name + ".img"
if "vfs" in backup_target:
command = ["find", VFS_DIR, "-type", "d", "-name", VOLUME_NAME]
backup_volume_path = subprocess.check_output(command).strip()
command = ["find", backup_volume_path, "-name", "*blk"]
blocks = subprocess.check_output(command).split()
assert len(blocks) != 0
for blk in blocks:
command = ["mv", blk, blk + ".tmp".encode('utf-8')]
subprocess.check_output(command).strip()
# should fail
is_failed = False
cmd.restore_inc(dr_address, backup1, backup0_name)
for i in range(RETRY_COUNTS):
rs = cmd.restore_status(dr_address)
for status in rs.values():
if status['backupURL'] != backup1:
break
if 'error' in status.keys():
if status['error'] != "":
assert 'no such file or directory' in \
status['error']
is_failed = True
if is_failed:
break
time.sleep(RETRY_INTERVAL)
assert is_failed
assert path.exists(FIXED_REPLICA_PATH1 + delta_file1)
assert path.exists(FIXED_REPLICA_PATH2 + delta_file1)
for blk in blocks:
command = ["mv", blk + ".tmp".encode('utf-8'), blk]
subprocess.check_output(command)
data1 = \
snap0_data[0:offset1] + snap1_data + \
snap0_data[offset1+length1:]
# race condition: last restoration has failed
# but `isRestoring` hasn't been cleanup
for i in range(RETRY_COUNTS):
try:
restore_incrementally(dr_address, backup1, backup0_name)
break
except subprocess.CalledProcessError as e:
if "already in progress" not in e.output:
time.sleep(RETRY_INTERVAL)
else:
raise e
verify_no_frontend_data(grpc_engine_manager, 0, data1, grpc_dr_controller)
assert not path.exists(FIXED_REPLICA_PATH1 + delta_file1)
assert not path.exists(FIXED_REPLICA_PATH2 + delta_file1)
status = cmd.restore_status(dr_address)
compare_last_restored_with_backup(status, backup1_name)
data2 = \
data1[0:offset2] + snap2_data + \
zero_string * (BLOCK_SIZE - length2 - offset2) + snap2_data
restore_incrementally(dr_address, backup2, backup1_name)
verify_no_frontend_data(grpc_engine_manager, 0, data2, grpc_dr_controller)
delta_file2 = "volume-delta-" + backup1_name + ".img"
assert not path.exists(FIXED_REPLICA_PATH1 + delta_file2)
assert not path.exists(FIXED_REPLICA_PATH2 + delta_file2)
status = cmd.restore_status(dr_address)
compare_last_restored_with_backup(status, backup2_name)
# mock race condition
with pytest.raises(subprocess.CalledProcessError) as e:
restore_incrementally(dr_address, backup1, backup0_name)
assert "doesn't match lastRestored" in e
data3 = zero_string * length3 + data2[length3:length2]
restore_incrementally(dr_address, backup3, backup2_name)
verify_no_frontend_data(grpc_engine_manager, 0, data3, grpc_dr_controller)
delta_file3 = "volume-delta-" + backup3_name + ".img"
assert not path.exists(FIXED_REPLICA_PATH1 + delta_file3)
assert not path.exists(FIXED_REPLICA_PATH2 + delta_file3)
status = cmd.restore_status(dr_address)
compare_last_restored_with_backup(status, backup3_name)
# mock corner case: invalid last-restored backup
rm_backups(address, ENGINE_NAME, [backup3])
# actually it is full restoration
restore_incrementally(dr_address, backup4, backup3_name)
verify_no_frontend_data(grpc_engine_manager, 0, snap4_data,
grpc_dr_controller)
status = cmd.restore_status(dr_address)
compare_last_restored_with_backup(status, backup4_name)
if "vfs" in backup_target:
command = ["find", VFS_DIR, "-type", "d", "-name", VOLUME_NAME]
backup_volume_path = subprocess.check_output(command).strip()
command = ["find", backup_volume_path, "-name", "*tempoary"]
tmp_files = subprocess.check_output(command).split()
assert len(tmp_files) == 0
cleanup_no_frontend_volume(grpc_engine_manager, grpc_dr_controller,
grpc_dr_replica1, grpc_dr_replica2)
rm_backups(address, ENGINE_NAME, [backup0, backup1, backup2, backup4])
cmd.sync_agent_server_reset(address)
cleanup_controller(grpc_controller)
cleanup_replica(grpc_replica1)
cleanup_replica(grpc_replica2)
def volume_expansion_with_backup_test(
grpc_engine_manager, # NOQA
grpc_controller, # NOQA
grpc_dr_controller, # NOQA
grpc_replica1,
grpc_replica2, # NOQA
grpc_dr_replica1, # NOQA
grpc_dr_replica2, # NOQA
volume_name,
engine_name,
backup_target): # NOQA
address = grpc_controller.address
dr_address = grpc_dr_controller.address
dev = get_dev(grpc_replica1, grpc_replica2, grpc_controller)
start_no_frontend_volume(grpc_engine_manager, grpc_dr_controller,
grpc_dr_replica1, grpc_dr_replica2)
try:
cmd.backup_volume_rm(address, volume_name, backup_target)
except Exception:
pass
data0_len = random_length(PAGE_SIZE)
data0 = Data(random.randrange(0, SIZE - 2 * PAGE_SIZE, PAGE_SIZE),
data0_len, random_string(data0_len))
snap0 = Snapshot(dev, data0, address)
backup0_info = create_backup(address, snap0.name, backup_target)
assert backup0_info["VolumeName"] == volume_name
assert backup0_info["Size"] == str(BLOCK_SIZE)
cmd.backup_restore(dr_address, backup0_info["URL"])
wait_for_restore_completion(dr_address, backup0_info["URL"])
verify_no_frontend_data(grpc_engine_manager, data0.offset, data0.content,
grpc_dr_controller)
grpc_controller.volume_expand(EXPAND_SIZE)
wait_for_volume_expansion(grpc_controller, EXPAND_SIZE)
check_block_device_size(volume_name, EXPAND_SIZE)
data1_len = random_length(PAGE_SIZE)
data1 = Data(random.randrange(SIZE, EXPAND_SIZE - PAGE_SIZE, PAGE_SIZE),
data1_len, random_string(data1_len))
snap1 = Snapshot(dev, data1, address)
backup1_info = create_backup(address, snap1.name, backup_target,
EXPAND_SIZE_STR)
assert backup1_info["VolumeName"] == volume_name
assert backup1_info["Size"] == str(2 * BLOCK_SIZE)
backup_volumes = cmd.backup_volume_list(address, volume_name,
backup_target)
assert volume_name in backup_volumes
assert backup_volumes[volume_name]["Size"] == EXPAND_SIZE_STR
# incremental restoration will implicitly expand the volume first
restore_incrementally(dr_address, backup1_info["URL"],
backup0_info["Name"])
check_dr_volume_block_device_size(grpc_engine_manager, grpc_dr_controller,
EXPAND_SIZE)
verify_no_frontend_data(grpc_engine_manager, data1.offset, data1.content,
grpc_dr_controller)
cmd.backup_volume_rm(grpc_controller.address, volume_name, backup_target)
def volume_expansion_with_snapshots_test(
dev,
grpc_controller, # NOQA
volume_name,
engine_name,
original_data):
# the default size is 4MB, will expand it to 8MB
address = grpc_controller.address
zero_char = b'\x00'.decode('utf-8')
# write the data to the original part then do expansion
data1_len = random_length(PAGE_SIZE)
data1 = Data(random.randrange(0, SIZE - 2 * PAGE_SIZE, PAGE_SIZE),
data1_len, random_string(data1_len))
snap1 = Snapshot(dev, data1, address)
grpc_controller.volume_expand(EXPAND_SIZE)
wait_for_volume_expansion(grpc_controller, EXPAND_SIZE)
check_block_device_size(volume_name, EXPAND_SIZE)
snap1.verify_data()
assert \
dev.readat(0, SIZE) == \
original_data[0:data1.offset] + data1.content + \
original_data[data1.offset+data1.length:]
assert dev.readat(SIZE, SIZE) == zero_char * SIZE
# write the data to both the original part and the expanded part
data2_len = random_length(PAGE_SIZE)
data2 = Data(SIZE - PAGE_SIZE, data2_len, random_string(data2_len))
snap2 = Snapshot(dev, data2, address)
data3_len = random_length(PAGE_SIZE)
data3 = Data(random.randrange(SIZE, EXPAND_SIZE - PAGE_SIZE, PAGE_SIZE),
data3_len, random_string(data3_len))
snap3 = Snapshot(dev, data3, address)
snap1.verify_data()
snap2.verify_data()
snap3.verify_data()
assert \
dev.readat(SIZE, SIZE) == zero_char*(data3.offset-SIZE) + \
data3.content + zero_char*(EXPAND_SIZE-data3.offset-data3.length)
data4_len = random_length(PAGE_SIZE)
data4 = Data(data1.offset, data4_len, random_string(data4_len))
snap4 = Snapshot(dev, data4, address)
snap4.verify_data()
# revert to snap1 then see if we can still r/w the existing data
# and expanded part
snapshot_revert_with_frontend(address, engine_name, snap1.name)
assert \
dev.readat(0, SIZE) == \
original_data[0:data1.offset] + data1.content + \
original_data[data1.offset+data1.length:]
assert dev.readat(SIZE, SIZE) == zero_char * SIZE
data5_len = random_length(PAGE_SIZE)
data5 = Data(random.randrange(SIZE, EXPAND_SIZE - PAGE_SIZE, PAGE_SIZE),
data5_len, random_string(data5_len))
snap5 = Snapshot(dev, data5, address)
snap5.verify_data()
assert \
dev.readat(SIZE, SIZE) == zero_char*(data5.offset-SIZE) + \
data5.content + zero_char*(EXPAND_SIZE-data5.offset-data5.length)
# delete and purge the snap1. it will coalesce with the larger snap2
cmd.snapshot_rm(address, snap1.name)
cmd.snapshot_purge(address)
wait_for_purge_completion(address)
assert \
dev.readat(0, SIZE) == \
original_data[0:data1.offset] + data1.content + \
original_data[data1.offset+data1.length:]
assert \
dev.readat(SIZE, SIZE) == zero_char*(data5.offset-SIZE) + \
data5.content + zero_char*(EXPAND_SIZE-data5.offset-data5.length)
def test_backup_type(
grpc_replica1,
grpc_replica2, # NOQA
grpc_controller,
backup_targets): # NOQA
for backup_target in backup_targets:
address = grpc_controller.address
block_size = 2 * 1024 * 1024
dev = get_dev(grpc_replica1, grpc_replica2, grpc_controller)
zero_string = b'\x00'.decode('utf-8')
# backup0: 256 random data in 1st block
length0 = 256
snap0_data = random_string(length0)
verify_data(dev, 0, snap0_data)
verify_data(dev, block_size, snap0_data)
snap0 = cmd.snapshot_create(address)
backup0 = create_backup(address, snap0, backup_target)
backup0_url = backup0["URL"]
assert backup0['IsIncremental'] is False
# backup1: 32 random data + 32 zero data + 192 random data in 1st block
length1 = 32
offset1 = 32
snap1_data = zero_string * length1
verify_data(dev, offset1, snap1_data)
snap1 = cmd.snapshot_create(address)
backup1 = create_backup(address, snap1, backup_target)
backup1_url = backup1["URL"]
assert backup1['IsIncremental'] is True
# backup2: 32 random data + 256 random data in 1st block,
# 256 random data in 2nd block
length2 = 256
offset2 = 32
snap2_data = random_string(length2)
verify_data(dev, offset2, snap2_data)
verify_data(dev, block_size, snap2_data)
snap2 = cmd.snapshot_create(address)
backup2 = create_backup(address, snap2, backup_target)
backup2_url = backup2["URL"]
assert backup2['IsIncremental'] is True
rm_backups(address, ENGINE_NAME, [backup2_url])
# backup3: 64 zero data + 192 random data in 1st block
length3 = 64
offset3 = 0
verify_data(dev, offset3, zero_string * length3)
verify_data(dev, length2, zero_string * offset2)
verify_data(dev, block_size, zero_string * length2)
snap3 = cmd.snapshot_create(address)
backup3 = create_backup(address, snap3, backup_target)
backup3_url = backup3["URL"]
assert backup3['IsIncremental'] is False
# backup4: 256 random data in 1st block
length4 = 256
offset4 = 0
snap4_data = random_string(length4)
verify_data(dev, offset4, snap4_data)
snap4 = cmd.snapshot_create(address)
backup4 = create_backup(address, snap4, backup_target)
backup4_url = backup4["URL"]
assert backup4['IsIncremental'] is True
rm_backups(address, ENGINE_NAME,
[backup0_url, backup1_url, backup3_url, backup4_url])
cmd.sync_agent_server_reset(address)
cleanup_replica(grpc_replica1)
cleanup_replica(grpc_replica2)
cleanup_controller(grpc_controller)
def test_expansion_with_rebuild(
grpc_controller, # NOQA
grpc_replica1,
grpc_replica2): # NOQA
address = grpc_controller.address
dev = get_dev(grpc_replica1, grpc_replica2, grpc_controller)
replicas = grpc_controller.replica_list()
assert len(replicas) == 2
assert replicas[0].mode == "RW"
assert replicas[1].mode == "RW"
# the default size is 4MB, will expand it to 8MB
address = grpc_controller.address
zero_char = b'\x00'.decode('utf-8')
original_data = zero_char * SIZE
# write the data to the original part then do expansion
data1_len = random_length(PAGE_SIZE)
data1 = Data(random.randrange(0, SIZE - 2 * PAGE_SIZE, PAGE_SIZE),
data1_len, random_string(data1_len))
snap1 = Snapshot(dev, data1, address)
grpc_controller.volume_expand(EXPAND_SIZE)
wait_for_volume_expansion(grpc_controller, EXPAND_SIZE)
check_block_device_size(VOLUME_NAME, EXPAND_SIZE)
snap1.verify_data()
assert \
dev.readat(0, SIZE) == \
original_data[0:data1.offset] + data1.content + \
original_data[data1.offset+data1.length:]
assert dev.readat(SIZE, SIZE) == zero_char * SIZE
# write the data to both the original part and the expanded part
data2_len = random_length(PAGE_SIZE)
data2 = Data(SIZE - PAGE_SIZE, data2_len, random_string(data2_len))
snap2 = Snapshot(dev, data2, address)
data3_len = random_length(PAGE_SIZE)
data3 = Data(random.randrange(SIZE, EXPAND_SIZE - PAGE_SIZE, PAGE_SIZE),
data3_len, random_string(data3_len))
snap3 = Snapshot(dev, data3, address)
snap1.verify_data()
snap2.verify_data()
snap3.verify_data()
assert \
dev.readat(SIZE, SIZE) == zero_char*(data3.offset-SIZE) + \
data3.content + zero_char*(EXPAND_SIZE-data3.offset-data3.length)
# Cleanup replica2
cleanup_replica(grpc_replica2)
verify_replica_state(grpc_controller, 1, "ERR")
grpc_controller.replica_delete(replicas[1].address)
# Rebuild replica2.
open_replica(grpc_replica2)
# The newly opened replica2 will be expanded automatically
cmd.add_replica(address, grpc_replica2.url)
wait_for_rebuild_complete(address)
verify_replica_state(grpc_controller, 1, "RW")
# Cleanup replica1 then check if the rebuilt replica2 works fine
cleanup_replica(grpc_replica1)
verify_replica_state(grpc_controller, 0, "ERR")
grpc_controller.replica_delete(replicas[0].address)
assert \
dev.readat(0, SIZE) == \
original_data[0:data1.offset] + data1.content + \
original_data[data1.offset+data1.length:data2.offset] + \
data2.content + \
original_data[data2.offset+data2.length:]
assert \
dev.readat(SIZE, SIZE) == zero_char*(data3.offset-SIZE) + \
data3.content + zero_char*(EXPAND_SIZE-data3.offset-data3.length)
data4_len = random_length(PAGE_SIZE)
data4 = Data(data1.offset, data4_len, random_string(data4_len))
snap4 = Snapshot(dev, data4, address)
snap4.verify_data()
def restore_to_file_without_backing_file_test(
backup_target, # NOQA
grpc_controller, # NOQA
grpc_replica1, # NOQA
grpc_replica2): # NOQA
address = grpc_controller.address
dev = get_dev(grpc_replica1, grpc_replica2, grpc_controller)
length0 = 256
length1 = 128
offset0 = 0
offset1 = length1 + offset0
output_raw_path = file(OUTPUT_FILE_RAW)
output_qcow2_path = file(OUTPUT_FILE_QCOW2)
# create 1 empty snapshot for converting to init state.
snap0 = cmd.snapshot_create(address)
# create 1 snapshot with 256B data.
# output = snap2(offset0, length1)
snap1_data = random_string(length0)
verify_data(dev, offset0, snap1_data)
snap1 = cmd.snapshot_create(address)
backup = create_backup(address, snap1, backup_target)["URL"]
cmd.restore_to_file(address, backup, "", output_raw_path, IMAGE_FORMAT_RAW)
output1_raw = read_file(output_raw_path, offset0, length0)
assert output1_raw == snap1_data
os.remove(output_raw_path)
assert not os.path.exists(output_raw_path)
cmd.restore_to_file(address, backup, "", output_qcow2_path,
IMAGE_FORMAT_QCOW2)
output1_qcow2 = read_qcow2_file_without_backing_file(
output_qcow2_path, offset0, length0)
assert output1_qcow2.decode('utf-8') == snap1_data
os.remove(output_qcow2_path)
assert not os.path.exists(output_qcow2_path)
snapshot_revert_with_frontend(address, ENGINE_NAME, snap0)
rm_snaps(address, [snap1])
rm_backups(address, ENGINE_NAME, [backup])
# create 2 snapshots with 256B data and 128B data
# output = snap2(offset0, length1 - length2) +
# snap1(offset2, length2)
snap1_data = random_string(length0)
verify_data(dev, offset0, snap1_data)
snap1 = cmd.snapshot_create(address)
snap2_data = random_string(length1)
verify_data(dev, offset0, snap2_data)
snap2 = cmd.snapshot_create(address)
backup = create_backup(address, snap2, backup_target)["URL"]
cmd.restore_to_file(address, backup, "", output_raw_path, IMAGE_FORMAT_RAW)
output2_raw_snap2 = read_file(output_raw_path, offset0, length1)
output2_raw_snap1 = read_file(output_raw_path, offset1, length0 - length1)
assert output2_raw_snap2 == snap2_data
assert output2_raw_snap1 == snap1_data[offset1:length0]
cmd.restore_to_file(address, backup, "", output_qcow2_path,
IMAGE_FORMAT_QCOW2)
output2_qcow2_snap2 = read_qcow2_file_without_backing_file(
output_qcow2_path, offset0, length1)
output2_qcow2_snap1 = read_qcow2_file_without_backing_file(
output_qcow2_path, offset1, length0 - length1)
assert output2_qcow2_snap2.decode('utf-8') == snap2_data
assert output2_qcow2_snap1.decode('utf-8') == snap1_data[offset1:length0]
os.remove(output_qcow2_path)
assert not os.path.exists(output_qcow2_path)
snapshot_revert_with_frontend(address, ENGINE_NAME, snap0)
rm_snaps(address, [snap1, snap2])
rm_backups(address, ENGINE_NAME, [backup])
def test_ha_single_replica_rebuild(
grpc_controller, # NOQA
grpc_replica1,
grpc_replica2): # NOQA
address = grpc_controller.address
open_replica(grpc_replica1)
open_replica(grpc_replica2)
replicas = grpc_controller.replica_list()
assert len(replicas) == 0
r1_url = grpc_replica1.url
r2_url = grpc_replica2.url
v = grpc_controller.volume_start(replicas=[r1_url, r2_url])
assert v.replicaCount == 2
replicas = grpc_controller.replica_list()
assert len(replicas) == 2
assert replicas[0].mode == "RW"
assert replicas[1].mode == "RW"
dev = get_blockdev(VOLUME_NAME)
data = random_string(128)
data_offset = 1024
verify_data(dev, data_offset, data)
# Cleanup replica2
cleanup_replica(grpc_replica2)
verify_async(dev, 10, 128, 1)
verify_replica_state(grpc_controller, 1, "ERR")
verify_read(dev, data_offset, data)
grpc_controller.replica_delete(replicas[1].address)
# Rebuild replica2
open_replica(grpc_replica2)
cmd.add_replica(address, r2_url)
wait_for_rebuild_complete(address)
verify_async(dev, 10, 128, 1)
verify_replica_state(grpc_controller, 1, "RW")
verify_read(dev, data_offset, data)
# WORKAROUND for unable to remove the parent of volume head
newsnap = cmd.snapshot_create(address)
info = cmd.snapshot_info(address)
assert len(info) == 3
sysnap = info[newsnap]["parent"]
assert info[sysnap]["parent"] == ""
assert newsnap in info[sysnap]["children"]
assert info[sysnap]["usercreated"] is False
assert info[sysnap]["removed"] is False
cmd.snapshot_purge(address)
wait_for_purge_completion(address)
info = cmd.snapshot_info(address)
assert len(info) == 2
assert info[newsnap] is not None
assert info[VOLUME_HEAD] is not None
def snapshot_tree_create_node(dev, address, offset, length, snap, data, name):
data[name] = random_string(length)
verify_data(dev, offset, data[name])
snap[name] = cmd.snapshot_create(address)
