def test_expand_data_node(self):
"""
target: test create and insert api after expand dataNode pod
method: 1.create collection a and insert df
2.expand dataNode pod from 1 to 2
3.verify collection a property and verify create and insert of new collection
expected: two collection create and insert op are both correctly
"""
# deploy all nodes one pod cluster milvus with helm
release_name = "scale-test"
env = HelmEnv(release_name=release_name)
env.helm_install_cluster_milvus()
host = env.get_svc_external_ip()
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create
c_name = cf.gen_unique_str(prefix)
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name, schema=cf.gen_default_collection_schema())
# # insert
data = cf.gen_default_list_data(ct.default_nb)
mutation_res, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
# scale dataNode to 2 pods
env.helm_upgrade_cluster_milvus(dataNode=2)
# after scale, assert data consistent
assert utility.has_collection(c_name)
assert collection_w.num_entities == ct.default_nb
# assert new operations
new_cname = cf.gen_unique_str(prefix)
new_collection_w = ApiCollectionWrapper()
new_collection_w.init_collection(name=new_cname, schema=cf.gen_default_collection_schema())
new_mutation_res, _ = new_collection_w.insert(data)
assert new_mutation_res.insert_count == ct.default_nb
assert new_collection_w.num_entities == ct.default_nb
# assert old collection ddl
mutation_res_2, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
assert collection_w.num_entities == ct.default_nb*2
collection_w.drop()
new_collection_w.drop()
def test_shrink_data_node(self):
"""
target: test shrink dataNode from 2 to 1
method: 1.create collection and insert df 2. shrink dataNode 3.insert df
expected: verify the property of collection which channel on shrink pod
"""
release_name = "scale-data"
env = HelmEnv(release_name=release_name, dataNode=2)
host = env.helm_install_cluster_milvus(image_pull_policy=constants.IF_NOT_PRESENT)
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
c_name = "data_scale_one"
data = cf.gen_default_list_data(ct.default_nb)
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name, schema=cf.gen_default_collection_schema())
mutation_res, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
assert collection_w.num_entities == ct.default_nb
c_name_2 = "data_scale_two"
collection_w2 = ApiCollectionWrapper()
collection_w2.init_collection(name=c_name_2, schema=cf.gen_default_collection_schema())
mutation_res2, _ = collection_w2.insert(data)
assert mutation_res2.insert_count == ct.default_nb
assert collection_w2.num_entities == ct.default_nb
env.helm_upgrade_cluster_milvus(dataNode=1)
assert collection_w.num_entities == ct.default_nb
mutation_res2, _ = collection_w2.insert(data)
assert collection_w2.num_entities == ct.default_nb*2
collection_w.drop()
collection_w2.drop()
def test_simd_compat_e2e(self, request, simd):
log.info(f"start to e2e verification: {simd}")
# parse results from previous results
results = request.config.cache.get(simd, None)
alias = results.get('alias', simd)
conn = connections.connect(alias=alias)
assert conn is not None
simd_cache = request.config.cache.get(simd, None)
log.info(f"simd_cache: {simd_cache}")
# create
name = cf.gen_unique_str("compat")
t0 = time.time()
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=name,
schema=cf.gen_default_collection_schema(),
using=alias,
timeout=40)
tt = time.time() - t0
assert collection_w.name == name
entities = collection_w.num_entities
log.info(f"assert create collection: {tt}, init_entities: {entities}")
# insert
data = cf.gen_default_list_data()
t0 = time.time()
_, res = collection_w.insert(data)
tt = time.time() - t0
log.info(f"assert insert: {tt}")
assert res
# flush
t0 = time.time()
assert collection_w.num_entities == len(data[0]) + entities
tt = time.time() - t0
entities = collection_w.num_entities
log.info(f"assert flush: {tt}, entities: {entities}")
# search
collection_w.load()
search_vectors = cf.gen_vectors(1, ct.default_dim)
search_params = {"metric_type": "L2", "params": {"nprobe": 16}}
t0 = time.time()
res_1, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param=search_params,
limit=1)
tt = time.time() - t0
log.info(f"assert search: {tt}")
assert len(res_1) == 1
collection_w.release()
# index
d = cf.gen_default_list_data()
collection_w.insert(d)
log.info(f"assert index entities: {collection_w.num_entities}")
_index_params = {
"index_type": "IVF_SQ8",
"params": {
"nlist": 64
},
"metric_type": "L2"
}
t0 = time.time()
index, _ = collection_w.create_index(
field_name=ct.default_float_vec_field_name,
index_params=_index_params,
name=cf.gen_unique_str())
tt = time.time() - t0
log.info(f"assert index: {tt}")
assert len(collection_w.indexes) == 1
# search
t0 = time.time()
collection_w.load()
tt = time.time() - t0
log.info(f"assert load: {tt}")
search_vectors = cf.gen_vectors(1, ct.default_dim)
t0 = time.time()
res_1, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param=search_params,
limit=1)
tt = time.time() - t0
log.info(f"assert search: {tt}")
# query
term_expr = f'{ct.default_int64_field_name} in [1001,1201,4999,2999]'
t0 = time.time()
res, _ = collection_w.query(term_expr)
tt = time.time() - t0
log.info(f"assert query result {len(res)}: {tt}")
class Base:
""" Initialize class object """
connection_wrap = None
collection_wrap = None
partition_wrap = None
index_wrap = None
utility_wrap = None
collection_schema_wrap = None
field_schema_wrap = None
collection_object_list = []
def setup_class(self):
log.info("[setup_class] Start setup class...")
def teardown_class(self):
log.info("[teardown_class] Start teardown class...")
pass
def setup_method(self, method):
log.info(("*" * 35) + " setup " + ("*" * 35))
log.info("[setup_method] Start setup test case %s..." % method.__name__)
self.connection_wrap = ApiConnectionsWrapper()
self.utility_wrap = ApiUtilityWrapper()
self.collection_wrap = ApiCollectionWrapper()
self.partition_wrap = ApiPartitionWrapper()
self.index_wrap = ApiIndexWrapper()
self.collection_schema_wrap = ApiCollectionSchemaWrapper()
self.field_schema_wrap = ApiFieldSchemaWrapper()
def teardown_method(self, method):
log.info(("*" * 35) + " teardown " + ("*" * 35))
log.info("[teardown_method] Start teardown test case %s..." % method.__name__)
try:
""" Drop collection before disconnect """
if self.connection_wrap.get_connection(alias=DefaultConfig.DEFAULT_USING)[0] is None:
self.connection_wrap.connect(alias=DefaultConfig.DEFAULT_USING, host=param_info.param_host,
port=param_info.param_port)
if self.collection_wrap.collection is not None:
self.collection_wrap.drop(check_task=ct.CheckTasks.check_nothing)
for collection_object in self.collection_object_list:
if collection_object.collection is not None \
and collection_object.name in self.utility_wrap.list_collections()[0]:
collection_object.drop(check_task=ct.CheckTasks.check_nothing)
except Exception as e:
log.debug(str(e))
try:
""" Delete connection and reset configuration"""
res = self.connection_wrap.list_connections()
for i in res[0]:
self.connection_wrap.remove_connection(i[0])
# because the connection is in singleton mode, it needs to be restored to the original state after teardown
self.connection_wrap.add_connection(default={"host": DefaultConfig.DEFAULT_HOST,
"port": DefaultConfig.DEFAULT_PORT})
except Exception as e:
log.debug(str(e))
@pytest.fixture(scope="module", autouse=True)
def initialize_env(self, request):
""" clean log before testing """
host = request.config.getoption("--host")
port = request.config.getoption("--port")
handler = request.config.getoption("--handler")
clean_log = request.config.getoption("--clean_log")
""" params check """
assert ip_check(host) and number_check(port)
""" modify log files """
cf.modify_file(file_path_list=[log_config.log_debug, log_config.log_info, log_config.log_err], is_modify=clean_log)
log.info("#" * 80)
log.info("[initialize_milvus] Log cleaned up, start testing...")
param_info.prepare_param_info(host, port, handler)
def test_expand_data_node(self):
"""
target: test create and insert api after expand dataNode pod
method: 1.create collection a and insert df
2.expand dataNode pod from 1 to 2
3.verify collection a property and verify create and insert of new collection
expected: two collection create and insert op are both correctly
"""
# deploy all nodes one pod cluster milvus with helm
release_name = "scale-data"
# env = HelmEnv(release_name=release_name)
# host = env.helm_install_cluster_milvus()
# deploy cluster milvus with dataNode 1 replicas
default_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.components.image':
'milvusdb/milvus-dev:master-20211020-b40513b',
'spec.components.proxy.serviceType': 'LoadBalancer',
'dependencies.etcd.inCluster.deletionPolicy': 'Delete',
'dependencies.etcd.inCluster.pvcDeletion': 'true',
'dependencies.pulsar.inCluster.deletionPolicy': 'Delete',
'dependencies.pulsar.inCluster.pvcDeletion': 'true',
'dependencies.storage.inCluster.deletionPolicy': 'Delete',
'dependencies.storage.inCluster.pvcDeletion': 'true',
}
milvusOp = MilvusOperator()
milvusOp.install(default_config)
if milvusOp.wait_for_healthy(release_name,
namespace=constants.NAMESPACE):
endpoint = milvusOp.endpoint(release_name, constants.NAMESPACE)
endpoint = endpoint.split(':')
host = endpoint[0]
port = int(endpoint[-1])
else:
raise Exception(f"Failed to install {release_name}")
# connect
connections.add_connection(default={"host": host, "port": port})
connections.connect(alias='default')
# create
c_name = cf.gen_unique_str(prefix)
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
# # insert
data = cf.gen_default_list_data(ct.default_nb)
mutation_res, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
# scale dataNode to 2 pods
milvusOp.upgrade(release_name,
{'spec.components.dataNode.replicas': 2},
constants.NAMESPACE)
# env.helm_upgrade_cluster_milvus(dataNode=2)
# after scale, assert data consistent
assert utility.has_collection(c_name)
assert collection_w.num_entities == ct.default_nb
# assert new operations
new_cname = cf.gen_unique_str(prefix)
new_collection_w = ApiCollectionWrapper()
new_collection_w.init_collection(
name=new_cname, schema=cf.gen_default_collection_schema())
new_mutation_res, _ = new_collection_w.insert(data)
assert new_mutation_res.insert_count == ct.default_nb
assert new_collection_w.num_entities == ct.default_nb
# assert old collection ddl
mutation_res_2, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
assert collection_w.num_entities == ct.default_nb * 2
collection_w.drop()
new_collection_w.drop()
def test_shrink_index_node(self):
"""
target: test shrink indexNode from 2 to 1
method: 1.deploy two indexNode
2.create index with two indexNode
3.shrink indexNode from 2 to 1
4.create index with 1 indexNode
expected: The cost of one indexNode is about twice that of two indexNodes
"""
release_name = "shrink-index"
image_tag = get_latest_tag()
image = f'{constants.IMAGE_REPOSITORY}:{image_tag}'
data_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.indexNode.replicas': 2,
'spec.components.dataNode.replicas': 2,
'spec.config.dataCoord.enableCompaction': True,
'spec.config.dataCoord.enableGarbageCollection': True
}
mic = MilvusOperator()
mic.install(data_config)
if mic.wait_for_healthy(release_name,
constants.NAMESPACE,
timeout=1800):
host = mic.endpoint(release_name,
constants.NAMESPACE).split(':')[0]
else:
raise MilvusException(message=f'Milvus healthy timeout 1800s')
try:
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
data = cf.gen_default_dataframe_data(nb)
# create
c_name = "index_scale_one"
collection_w = ApiCollectionWrapper()
# collection_w.init_collection(name=c_name)
collection_w.init_collection(
name=c_name, schema=cf.gen_default_collection_schema())
# insert
loop = 10
for i in range(loop):
collection_w.insert(data)
assert collection_w.num_entities == nb * loop
# create index on collection one and two
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t0 = datetime.datetime.now() - start
log.info(f'Create index on 2 indexNode cost t0: {t0}')
collection_w.drop_index()
assert not collection_w.has_index()[0]
# shrink indexNode from 2 to 1
mic.upgrade(release_name,
{'spec.components.indexNode.replicas': 1},
constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t1 = datetime.datetime.now() - start
log.info(f'Create index on 1 indexNode cost t1: {t1}')
collection_w.drop_index()
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t2 = datetime.datetime.now() - start
log.info(f'Create index on 1 indexNode cost t2: {t2}')
log.debug(f'one indexNode: {t2}')
log.debug(f't2 is {t2}, t0 is {t0}, t2/t0 is {t2 / t0}')
# assert round(t2 / t0) == 2
except Exception as e:
raise Exception(str(e))
finally:
label = f"app.kubernetes.io/instance={release_name}"
log.info('Start to export milvus pod logs')
read_pod_log(namespace=constants.NAMESPACE,
label_selector=label,
release_name=release_name)
mic.uninstall(release_name, namespace=constants.NAMESPACE)
def test_shrink_index_node(self):
"""
target: test shrink indexNode from 2 to 1
method: 1.deploy two indexNode
2.create index with two indexNode
3.shrink indexNode from 2 to 1
4.create index with 1 indexNode
expected: The cost of one indexNode is about twice that of two indexNodes
"""
release_name = "scale-index"
env = HelmEnv(release_name=release_name, indexNode=2)
env.helm_install_cluster_milvus()
# connect
connections.add_connection(default={
"host": '10.98.0.8',
"port": 19530
})
connections.connect(alias='default')
data = cf.gen_default_dataframe_data(nb)
# create
c_name = "index_scale_one"
collection_w = ApiCollectionWrapper()
# collection_w.init_collection(name=c_name)
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
# insert
loop = 10
for i in range(loop):
collection_w.insert(data)
assert collection_w.num_entities == nb * loop
# create index on collection one and two
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t0 = datetime.datetime.now() - start
log.debug(f'two indexNodes: {t0}')
collection_w.drop_index()
assert not collection_w.has_index()[0]
# expand indexNode from 1 to 2
# pdb.set_trace()
env.helm_upgrade_cluster_milvus(indexNode=1)
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t1 = datetime.datetime.now() - start
log.debug(f'one indexNode: {t1}')
log.debug(t1 / t0)
assert round(t1 / t0) == 2
def test_scale_proxy(self):
"""
target: test milvus operation after proxy expand
method: 1.deploy 1 proxy replicas
2.milvus e2e test in parallel
3.expand proxy pod from 1 to 5
4.milvus e2e test
5.shrink proxy from 5 to 2
expected: 1.verify data consistent and func work
"""
# deploy milvus cluster with one proxy
fail_count = 0
release_name = "scale-proxy"
image_tag = get_latest_tag()
image = f'{constants.IMAGE_REPOSITORY}:{image_tag}'
data_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.mode': 'cluster',
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.proxy.replicas': 1,
'spec.components.dataNode.replicas': 2,
'spec.config.common.retentionDuration': 60
}
mic = MilvusOperator()
mic.install(data_config)
if mic.wait_for_healthy(release_name, constants.NAMESPACE, timeout=1800):
host = mic.endpoint(release_name, constants.NAMESPACE).split(':')[0]
else:
raise MilvusException(message=f'Milvus healthy timeout 1800s')
try:
c_name = cf.gen_unique_str("proxy_scale")
e2e_milvus_parallel(2, host, c_name)
log.info('Milvus test before expand')
# expand proxy replicas from 1 to 5
mic.upgrade(release_name, {'spec.components.proxy.replicas': 5}, constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE, f"app.kubernetes.io/instance={release_name}")
e2e_milvus_parallel(5, host, c_name)
log.info('Milvus test after expand')
# expand proxy replicas from 5 to 2
mic.upgrade(release_name, {'spec.components.proxy.replicas': 2}, constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE, f"app.kubernetes.io/instance={release_name}")
e2e_milvus_parallel(2, host, c_name)
log.info('Milvus test after shrink')
connections.connect('default', host=host, port=19530)
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name)
"""
total start 2+5+2 process to run e2e, each time insert default_nb data, But one of the 2 processes started
for the first time did not insert due to collection creation exception. So actually insert eight times
"""
assert collection_w.num_entities == 8 * default_nb
except Exception as e:
log.error(str(e))
fail_count += 1
# raise Exception(str(e))
finally:
log.info(f'Test finished with {fail_count} fail request')
assert fail_count <= 1
label = f"app.kubernetes.io/instance={release_name}"
log.info('Start to export milvus pod logs')
read_pod_log(namespace=constants.NAMESPACE, label_selector=label, release_name=release_name)
mic.uninstall(release_name, namespace=constants.NAMESPACE)
def test_customize_segment_size(self, seg_size, seg_count):
"""
steps
"""
log.info(f"start to install milvus with segment size {seg_size}")
release_name, host, port = _install_milvus(seg_size)
self.release_name = release_name
assert host is not None
conn = connections.connect("default", host=host, port=port)
assert conn is not None
mil = MilvusSys(alias="default")
log.info(f"milvus build version: {mil.build_version}")
log.info(f"start to e2e verification: {seg_size}")
# create
name = cf.gen_unique_str("segsiz")
t0 = time.time()
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=name,
schema=cf.gen_default_collection_schema(),
timeout=40)
tt = time.time() - t0
assert collection_w.name == name
entities = collection_w.num_entities
log.info(f"assert create collection: {tt}, init_entities: {entities}")
# insert
nb = 50000
data = cf.gen_default_list_data(nb=nb)
t0 = time.time()
_, res = collection_w.insert(data)
tt = time.time() - t0
log.info(f"assert insert: {tt}")
assert res
# insert 2 million entities
rounds = 40
for _ in range(rounds - 1):
_, res = collection_w.insert(data)
entities = collection_w.num_entities
assert entities == nb * rounds
# load
collection_w.load()
utility_wrap = ApiUtilityWrapper()
segs, _ = utility_wrap.get_query_segment_info(collection_w.name)
log.info(f"assert segments: {len(segs)}")
assert len(segs) == seg_count
# search
search_vectors = cf.gen_vectors(1, ct.default_dim)
search_params = {"metric_type": "L2", "params": {"nprobe": 16}}
t0 = time.time()
res_1, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param=search_params,
limit=1,
timeout=30)
tt = time.time() - t0
log.info(f"assert search: {tt}")
assert len(res_1) == 1
collection_w.release()
# index
d = cf.gen_default_list_data()
collection_w.insert(d)
log.info(f"assert index entities: {collection_w.num_entities}")
_index_params = {
"index_type": "IVF_SQ8",
"params": {
"nlist": 64
},
"metric_type": "L2"
}
t0 = time.time()
index, _ = collection_w.create_index(
field_name=ct.default_float_vec_field_name,
index_params=_index_params,
name=cf.gen_unique_str(),
timeout=120)
tt = time.time() - t0
log.info(f"assert index: {tt}")
assert len(collection_w.indexes) == 1
# search
t0 = time.time()
collection_w.load()
tt = time.time() - t0
log.info(f"assert load: {tt}")
search_vectors = cf.gen_vectors(1, ct.default_dim)
t0 = time.time()
res_1, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param=search_params,
limit=1,
timeout=30)
tt = time.time() - t0
log.info(f"assert search: {tt}")
# query
term_expr = f'{ct.default_int64_field_name} in [1001,1201,4999,2999]'
t0 = time.time()
res, _ = collection_w.query(term_expr, timeout=30)
tt = time.time() - t0
log.info(f"assert query result {len(res)}: {tt}")
def test_scale_query_node(self):
"""
target: test scale queryNode
method: 1.deploy milvus cluster with 1 queryNode
2.prepare work (connect, create, insert, index and load)
3.continuously search (daemon thread)
4.expand queryNode from 2 to 5
5.continuously insert new data (daemon thread)
6.shrink queryNode from 5 to 3
expected: Verify milvus remains healthy and search successfully during scale
"""
release_name = "scale-query"
image_tag = get_latest_tag()
image = f'{constants.IMAGE_REPOSITORY}:{image_tag}'
query_config = {
'metadata.namespace': constants.NAMESPACE,
'spec.mode': 'cluster',
'metadata.name': release_name,
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.queryNode.replicas': 1,
'spec.config.common.retentionDuration': 60
}
mic = MilvusOperator()
mic.install(query_config)
if mic.wait_for_healthy(release_name,
constants.NAMESPACE,
timeout=1800):
host = mic.endpoint(release_name,
constants.NAMESPACE).split(':')[0]
else:
raise MilvusException(message=f'Milvus healthy timeout 1800s')
try:
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create
c_name = cf.gen_unique_str("scale_query")
# c_name = 'scale_query_DymS7kI4'
collection_w = ApiCollectionWrapper()
collection_w.init_collection(
name=c_name,
schema=cf.gen_default_collection_schema(),
shards_num=2)
# insert two segments
for i in range(3):
df = cf.gen_default_dataframe_data(nb)
collection_w.insert(df)
log.debug(collection_w.num_entities)
# create index
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
assert collection_w.index()[0] == Index(
collection_w.collection, ct.default_float_vec_field_name,
default_index_params)
# load
collection_w.load()
# scale queryNode to 5
mic.upgrade(release_name,
{'spec.components.queryNode.replicas': 5},
constants.NAMESPACE)
@counter
def do_search():
""" do search """
search_res, is_succ = collection_w.search(
cf.gen_vectors(1, ct.default_dim),
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit,
check_task=CheckTasks.check_nothing)
assert len(search_res) == 1
return search_res, is_succ
def loop_search():
""" continuously search """
while True:
do_search()
threading.Thread(target=loop_search, args=(), daemon=True).start()
# wait new QN running, continuously insert
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
@counter
def do_insert():
""" do insert """
return collection_w.insert(cf.gen_default_dataframe_data(1000),
check_task=CheckTasks.check_nothing)
def loop_insert():
""" loop insert """
while True:
do_insert()
threading.Thread(target=loop_insert, args=(), daemon=True).start()
log.debug(collection_w.num_entities)
time.sleep(20)
log.debug("Expand querynode test finished")
mic.upgrade(release_name,
{'spec.components.queryNode.replicas': 3},
constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
log.debug(collection_w.num_entities)
time.sleep(60)
scale_common.check_succ_rate(do_search)
scale_common.check_succ_rate(do_insert)
log.debug("Shrink querynode test finished")
except Exception as e:
raise Exception(str(e))
finally:
label = f"app.kubernetes.io/instance={release_name}"
log.info('Start to export milvus pod logs')
read_pod_log(namespace=constants.NAMESPACE,
label_selector=label,
release_name=release_name)
mic.uninstall(release_name, namespace=constants.NAMESPACE)
def __init__(self):
super().__init__()
self.c_wrapper = ApiCollectionWrapper()
def test_scale_in_query_node_less_than_replicas(self):
"""
target: test scale in cluster and querynode < replica
method: 1.Deploy cluster with 3 querynodes
2.Create and insert data, flush
3.Load collection with 2 replica number
4.Scale in querynode from 3 to 1 and query
5.Scale out querynode from 1 back to 3
expected: Verify search successfully after scale out
"""
release_name = "scale-in-query"
image_tag = get_latest_tag()
image = f'{constants.IMAGE_REPOSITORY}:{image_tag}'
query_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.mode': 'cluster',
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.queryNode.replicas': 2,
'spec.config.common.retentionDuration': 60
}
mic = MilvusOperator()
mic.install(query_config)
if mic.wait_for_healthy(release_name,
constants.NAMESPACE,
timeout=1800):
host = mic.endpoint(release_name,
constants.NAMESPACE).split(':')[0]
else:
raise MilvusException(message=f'Milvus healthy timeout 1800s')
try:
# prepare collection
connections.connect("scale-in", host=host, port=19530)
utility_w = ApiUtilityWrapper()
collection_w = ApiCollectionWrapper()
collection_w.init_collection(
name=cf.gen_unique_str("scale_in"),
schema=cf.gen_default_collection_schema(),
using="scale-in")
collection_w.insert(cf.gen_default_dataframe_data())
assert collection_w.num_entities == ct.default_nb
# load multi replicas and search success
collection_w.load(replica_number=2)
search_res, is_succ = collection_w.search(
cf.gen_vectors(1, ct.default_dim),
ct.default_float_vec_field_name, ct.default_search_params,
ct.default_limit)
assert len(search_res[0].ids) == ct.default_limit
log.info("Search successfully after load with 2 replicas")
log.debug(collection_w.get_replicas()[0])
log.debug(
utility_w.get_query_segment_info(collection_w.name,
using="scale-in"))
# scale in querynode from 2 to 1, less than replica number
log.debug("Scale in querynode from 2 to 1")
mic.upgrade(release_name,
{'spec.components.queryNode.replicas': 1},
constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
# search and not assure success
collection_w.search(cf.gen_vectors(1, ct.default_dim),
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit,
check_task=CheckTasks.check_nothing)
log.debug(
collection_w.get_replicas(
check_task=CheckTasks.check_nothing)[0])
# scale querynode from 1 back to 2
mic.upgrade(release_name,
{'spec.components.queryNode.replicas': 2},
constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
# verify search success
collection_w.search(cf.gen_vectors(1, ct.default_dim),
ct.default_float_vec_field_name,
ct.default_search_params, ct.default_limit)
# Verify replica info is correct
replicas = collection_w.get_replicas()[0]
assert len(replicas.groups) == 2
for group in replicas.groups:
assert len(group.group_nodes) == 1
# Verify loaded segment info is correct
seg_info = utility_w.get_query_segment_info(collection_w.name,
using="scale-in")[0]
num_entities = 0
for seg in seg_info:
assert len(seg.nodeIds) == 2
num_entities += seg.num_rows
assert num_entities == ct.default_nb
except Exception as e:
raise Exception(str(e))
finally:
label = f"app.kubernetes.io/instance={release_name}"
log.info('Start to export milvus pod logs')
read_pod_log(namespace=constants.NAMESPACE,
label_selector=label,
release_name=release_name)
mic.uninstall(release_name, namespace=constants.NAMESPACE)
def test_scale_query_node_replicas(self):
"""
target: test scale out querynode when load multi replicas
method: 1.Deploy cluster with 5 querynodes
2.Create collection with 2 shards
3.Insert 10 segments and flushed
4.Load collection with 2 replicas
5.Scale out querynode from 5 to 6 while search and insert growing data
expected: Verify search succ rate is 100%
"""
release_name = "scale-replica"
image_tag = get_latest_tag()
image = f'{constants.IMAGE_REPOSITORY}:{image_tag}'
query_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.mode': 'cluster',
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.queryNode.replicas': 5,
'spec.config.common.retentionDuration': 60
}
mic = MilvusOperator()
mic.install(query_config)
if mic.wait_for_healthy(release_name,
constants.NAMESPACE,
timeout=1800):
host = mic.endpoint(release_name,
constants.NAMESPACE).split(':')[0]
else:
raise MilvusException(message=f'Milvus healthy timeout 1800s')
try:
scale_querynode = random.choice([6, 7, 4, 3])
connections.connect("scale-replica", host=host, port=19530)
collection_w = ApiCollectionWrapper()
collection_w.init_collection(
name=cf.gen_unique_str("scale_out"),
schema=cf.gen_default_collection_schema(),
using='scale-replica',
shards_num=3)
# insert 10 sealed segments
for i in range(5):
df = cf.gen_default_dataframe_data(nb=nb, start=i * nb)
collection_w.insert(df)
assert collection_w.num_entities == (i + 1) * nb
collection_w.load(replica_number=2)
@counter
def do_search():
""" do search """
search_res, is_succ = collection_w.search(
cf.gen_vectors(1, ct.default_dim),
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit,
check_task=CheckTasks.check_nothing)
assert len(search_res) == 1
return search_res, is_succ
def loop_search():
""" continuously search """
while True:
do_search()
threading.Thread(target=loop_search, args=(), daemon=True).start()
# scale out
mic.upgrade(
release_name,
{'spec.components.queryNode.replicas': scale_querynode},
constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
log.debug("Scale out querynode success")
time.sleep(100)
scale_common.check_succ_rate(do_search)
log.debug("Scale out test finished")
except Exception as e:
raise Exception(str(e))
finally:
label = f"app.kubernetes.io/instance={release_name}"
log.info('Start to export milvus pod logs')
read_pod_log(namespace=constants.NAMESPACE,
label_selector=label,
release_name=release_name)
mic.uninstall(release_name, namespace=constants.NAMESPACE)
def test_chaos_data_consist(self, connection, chaos_yaml):
c_name = cf.gen_unique_str('chaos_collection_')
nb = 5000
i_name = cf.gen_unique_str('chaos_index_')
index_params = {
"index_type": "IVF_SQ8",
"metric_type": "L2",
"params": {
"nlist": 64
}
}
# create
t0 = datetime.datetime.now()
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
tt = datetime.datetime.now() - t0
log.debug(f"assert create: {tt}")
assert collection_w.name == c_name
# insert
data = cf.gen_default_list_data(nb=nb)
t0 = datetime.datetime.now()
_, res = collection_w.insert(data)
tt = datetime.datetime.now() - t0
log.debug(f"assert insert: {tt}")
assert res
# flush
t0 = datetime.datetime.now()
assert collection_w.num_entities == nb
tt = datetime.datetime.now() - t0
log.debug(f"assert flush: {tt}")
# search
collection_w.load()
search_vectors = cf.gen_vectors(1, ct.default_dim)
t0 = datetime.datetime.now()
search_res, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param={"nprobe": 16},
limit=1)
tt = datetime.datetime.now() - t0
log.debug(f"assert search: {tt}")
assert len(search_res) == 1
# index
t0 = datetime.datetime.now()
index, _ = collection_w.create_index(
field_name=ct.default_float_vec_field_name,
index_params=index_params,
name=i_name)
tt = datetime.datetime.now() - t0
log.debug(f"assert index: {tt}")
assert len(collection_w.indexes) == 1
# query
term_expr = f'{ct.default_int64_field_name} in [3001,4001,4999,2999]'
t0 = datetime.datetime.now()
query_res, _ = collection_w.query(term_expr)
tt = datetime.datetime.now() - t0
log.debug(f"assert query: {tt}")
assert len(query_res) == 4
# reboot a pod
reboot_pod(chaos_yaml)
# reconnect if needed
sleep(constants.WAIT_PER_OP * 4)
reconnect(connections, self.host, self.port)
# verify collection persists
assert utility.has_collection(c_name)
log.debug("assert collection persists")
collection_w2 = ApiCollectionWrapper()
collection_w2.init_collection(c_name)
# verify data persist
assert collection_w2.num_entities == nb
log.debug("assert data persists")
# verify index persists
assert collection_w2.has_index(i_name)
log.debug("assert index persists")
# verify search results persist
# verify query results persist
query_res2, _ = collection_w2.query(term_expr)
assert query_res2 == query_res
log.debug("assert query result persists")
def test_chaos_memory_stress_indexnode(self, connection, chaos_yaml):
"""
target: test inject memory stress into indexnode
method: 1.Deploy milvus and limit indexnode memory resource 1Gi
2.Create collection and insert some data
3.Create index
4.Inject memory stress chaos 512Mi
expected:
"""
# init collection and insert 250 nb
# nb = 50000 # vector size: 512*4*nb about 100Mi and create index need 600Mi memory
nb = 256000
dim = 512
# c_name = cf.gen_unique_str('chaos_memory')
c_name = 'chaos_memory_gKs8aSUu'
index_params = {
"index_type": "IVF_SQ8",
"metric_type": "L2",
"params": {
"nlist": 128
}
}
collection_w = ApiCollectionWrapper()
collection_w.init_collection(
name=c_name,
schema=cf.gen_default_collection_schema(dim=dim),
shards_num=1)
# insert 256000 512 dim entities, size 512Mi
for i in range(2):
t0_insert = datetime.datetime.now()
df = cf.gen_default_dataframe_data(nb=nb // 2, dim=dim)
res = collection_w.insert(df)[0]
assert res.insert_count == nb // 2
# log.info(f'After {i + 1} insert, num_entities: {collection_w.num_entities}')
tt_insert = datetime.datetime.now() - t0_insert
log.info(f"{i} insert data cost: {tt_insert}")
# flush
t0_flush = datetime.datetime.now()
assert collection_w.num_entities == nb
tt_flush = datetime.datetime.now() - t0_flush
log.info(f'flush {nb * 10} entities cost: {tt_flush}')
# create index
t0_index = datetime.datetime.now()
index, _ = collection_w.create_index(
field_name=ct.default_float_vec_field_name,
index_params=index_params)
tt_index = datetime.datetime.now() - t0_index
log.info(f"create index cost: {tt_index}")
log.info(collection_w.indexes)
# indexNode start build index, inject chaos memory stress
chaos_config = gen_experiment_config(chaos_yaml)
log.debug(chaos_config)
chaos_res = CusResource(kind=chaos_config['kind'],
group=constants.CHAOS_GROUP,
version=constants.CHAOS_VERSION,
namespace=constants.CHAOS_NAMESPACE)
chaos_res.create(chaos_config)
log.debug("inject chaos")
def collection_wrap_4_flush(self, connection):
c_wrap = ApiCollectionWrapper()
c_wrap.init_collection(name=cf.gen_unique_str("collection_4_insert"),
schema=cf.gen_default_collection_schema(),
check_task="check_nothing")
return c_wrap
def test_scale_query_node(self):
"""
target: test scale queryNode
method: 1.deploy milvus cluster with 1 queryNode
2.prepare work (connect, create, insert, index and load)
3.continuously search (daemon thread)
4.expand queryNode from 2 to 5
5.continuously insert new data (daemon thread)
6.shrink queryNode from 5 to 3
expected: Verify milvus remains healthy and search successfully during scale
"""
release_name = "scale-query"
query_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.components.image': 'harbor.zilliz.cc/milvus/milvus:master-20211202-ed546d0',
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.queryNode.replicas': 1,
'spec.config.dataCoord.enableCompaction': True,
'spec.config.dataCoord.enableGarbageCollection': True
}
mic = MilvusOperator()
mic.install(query_config)
healthy = mic.wait_for_healthy(release_name, constants.NAMESPACE, timeout=1200)
log.info(f"milvus healthy: {healthy}")
host = mic.endpoint(release_name, constants.NAMESPACE).split(':')[0]
# host = "10.98.0.8"
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create
c_name = cf.gen_unique_str("scale_query")
# c_name = 'scale_query_DymS7kI4'
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name, schema=cf.gen_default_collection_schema(), shards_num=2)
# insert two segments
for i in range(3):
df = cf.gen_default_dataframe_data(nb)
collection_w.insert(df)
log.debug(collection_w.num_entities)
# create index
collection_w.create_index(ct.default_float_vec_field_name, default_index_params)
assert collection_w.has_index()[0]
assert collection_w.index()[0] == Index(collection_w.collection, ct.default_float_vec_field_name,
default_index_params)
# load
collection_w.load()
# scale queryNode to 5
mic.upgrade(release_name, {'spec.components.queryNode.replicas': 5}, constants.NAMESPACE)
# continuously search
def do_search():
while True:
search_res, _ = collection_w.search(cf.gen_vectors(1, ct.default_dim),
ct.default_float_vec_field_name,
ct.default_search_params, ct.default_limit)
log.debug(search_res[0].ids)
assert len(search_res[0].ids) == ct.default_limit
t_search = threading.Thread(target=do_search, args=(), daemon=True)
t_search.start()
# wait new QN running, continuously insert
# time.sleep(10)
healthy = mic.wait_for_healthy(release_name, constants.NAMESPACE, timeout=1200)
log.info(f"milvus healthy after scale up: {healthy}")
# wait_pods_ready(constants.NAMESPACE, f"app.kubernetes.io/instance={release_name}")
def do_insert():
while True:
tmp_df = cf.gen_default_dataframe_data(1000)
collection_w.insert(tmp_df)
t_insert = threading.Thread(target=do_insert, args=(), daemon=True)
t_insert.start()
log.debug(collection_w.num_entities)
time.sleep(20)
log.debug("Expand querynode test finished")
mic.upgrade(release_name, {'spec.components.queryNode.replicas': 3}, constants.NAMESPACE)
time.sleep(60)
wait_pods_ready(constants.NAMESPACE, f"app.kubernetes.io/instance={release_name}")
log.debug(collection_w.num_entities)
time.sleep(60)
log.debug("Shrink querynode test finished")
def test_scale_data_node(self):
"""
target: test scale dataNode
method: 1.deploy milvus cluster with 2 dataNode
2.create collection with shards_num=5
3.continuously insert new data (daemon thread)
4.expand dataNode from 2 to 5
5.create new collection with shards_num=2
6.continuously insert new collection new data (daemon thread)
7.shrink dataNode from 5 to 3
expected: Verify milvus remains healthy, Insert and flush successfully during scale
Average dataNode memory usage
"""
release_name = "scale-data"
image = f'{constants.IMAGE_REPOSITORY}:{constants.IMAGE_TAG}'
data_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.dataNode.replicas': 2,
'spec.config.dataCoord.enableCompaction': True,
'spec.config.dataCoord.enableGarbageCollection': True
}
mic = MilvusOperator()
mic.install(data_config)
healthy = mic.wait_for_healthy(release_name,
constants.NAMESPACE,
timeout=1200)
log.info(f"milvus healthy: {healthy}")
host = mic.endpoint(release_name, constants.NAMESPACE).split(':')[0]
# host = '10.98.0.4'
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create
c_name = cf.gen_unique_str("scale_query")
# c_name = 'scale_query_DymS7kI4'
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema(),
shards_num=5)
tmp_nb = 10000
def do_insert():
while True:
tmp_df = cf.gen_default_dataframe_data(tmp_nb)
collection_w.insert(tmp_df)
log.debug(collection_w.num_entities)
t_insert = threading.Thread(target=do_insert, args=(), daemon=True)
t_insert.start()
# scale dataNode to 5
mic.upgrade(release_name, {'spec.components.dataNode.replicas': 5},
constants.NAMESPACE)
time.sleep(300)
log.debug("Expand dataNode test finished")
# create new collection and insert
new_c_name = cf.gen_unique_str("scale_query")
collection_w_new = ApiCollectionWrapper()
collection_w_new.init_collection(
name=new_c_name,
schema=cf.gen_default_collection_schema(),
shards_num=2)
def do_new_insert():
while True:
tmp_df = cf.gen_default_dataframe_data(tmp_nb)
collection_w_new.insert(tmp_df)
log.debug(collection_w_new.num_entities)
t_insert_new = threading.Thread(target=do_new_insert,
args=(),
daemon=True)
t_insert_new.start()
# scale dataNode to 3
mic.upgrade(release_name, {'spec.components.dataNode.replicas': 3},
constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
log.debug(collection_w.num_entities)
time.sleep(300)
log.debug("Shrink dataNode test finished")
def test_simd_compat_e2e(self, simd_id):
"""
steps
1. [test_milvus_install]: set up milvus with customized simd configured
2. [test_simd_compat_e2e]: verify milvus is working well
4. [test_milvus_cleanup]: delete milvus instances in teardown
"""
simd = supported_simd_types[simd_id]
log.info(f"start to install milvus with simd {simd}")
release_name, host, port = _install_milvus(simd)
self.release_name = release_name
assert host is not None
conn = connections.connect("default", host=host, port=port)
assert conn is not None
mil = MilvusSys(alias="default")
log.info(f"milvus build version: {mil.build_version}")
log.info(f"milvus simdType: {mil.simd_type}")
assert str(mil.simd_type).lower() in [
simd_type.lower() for simd_type in supported_simd_types[simd_id:]
]
log.info(f"start to e2e verification: {simd}")
# create
name = cf.gen_unique_str("compat")
t0 = time.time()
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=name,
schema=cf.gen_default_collection_schema(),
timeout=40)
tt = time.time() - t0
assert collection_w.name == name
entities = collection_w.num_entities
log.info(f"assert create collection: {tt}, init_entities: {entities}")
# insert
data = cf.gen_default_list_data()
t0 = time.time()
_, res = collection_w.insert(data)
tt = time.time() - t0
log.info(f"assert insert: {tt}")
assert res
# flush
t0 = time.time()
assert collection_w.num_entities == len(data[0]) + entities
tt = time.time() - t0
entities = collection_w.num_entities
log.info(f"assert flush: {tt}, entities: {entities}")
# search
collection_w.load()
search_vectors = cf.gen_vectors(1, ct.default_dim)
search_params = {"metric_type": "L2", "params": {"nprobe": 16}}
t0 = time.time()
res_1, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param=search_params,
limit=1)
tt = time.time() - t0
log.info(f"assert search: {tt}")
assert len(res_1) == 1
collection_w.release()
# index
d = cf.gen_default_list_data()
collection_w.insert(d)
log.info(f"assert index entities: {collection_w.num_entities}")
_index_params = {
"index_type": "IVF_SQ8",
"params": {
"nlist": 64
},
"metric_type": "L2"
}
t0 = time.time()
index, _ = collection_w.create_index(
field_name=ct.default_float_vec_field_name,
index_params=_index_params,
name=cf.gen_unique_str())
tt = time.time() - t0
log.info(f"assert index: {tt}")
assert len(collection_w.indexes) == 1
# search
t0 = time.time()
collection_w.load()
tt = time.time() - t0
log.info(f"assert load: {tt}")
search_vectors = cf.gen_vectors(1, ct.default_dim)
t0 = time.time()
res_1, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param=search_params,
limit=1)
tt = time.time() - t0
log.info(f"assert search: {tt}")
# query
term_expr = f'{ct.default_int64_field_name} in [1001,1201,4999,2999]'
t0 = time.time()
res, _ = collection_w.query(term_expr)
tt = time.time() - t0
log.info(f"assert query result {len(res)}: {tt}")
def test_expand_index_node(self):
"""
target: test expand indexNode from 1 to 2
method: 1.deploy two indexNode
2.create index with two indexNode
3.expand indexNode from 1 to 2
4.create index with one indexNode
expected: The cost of one indexNode is about twice that of two indexNodes
"""
release_name = "scale-index"
image = f'{constants.IMAGE_REPOSITORY}:{constants.IMAGE_TAG}'
data_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.indexNode.replicas': 1,
'spec.components.dataNode.replicas': 2,
'spec.config.dataCoord.enableCompaction': True,
'spec.config.dataCoord.enableGarbageCollection': True
}
mic = MilvusOperator()
mic.install(data_config)
healthy = mic.wait_for_healthy(release_name,
constants.NAMESPACE,
timeout=1200)
log.info(f"milvus healthy: {healthy}")
host = mic.endpoint(release_name, constants.NAMESPACE).split(':')[0]
# host = '10.98.0.8'
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
data = cf.gen_default_dataframe_data(nb)
# create
c_name = "index_scale_one"
collection_w = ApiCollectionWrapper()
# collection_w.init_collection(name=c_name)
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
# insert
loop = 100
for i in range(loop):
collection_w.insert(data, timeout=60)
assert collection_w.num_entities == nb * loop
# create index on collection
# note that the num of segments and the num of indexNode are related to indexing time
collection_w.drop_index()
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t0 = datetime.datetime.now() - start
log.debug(f't0: {t0}')
collection_w.drop_index()
assert not collection_w.has_index()[0]
# expand indexNode from 1 to 2
mic.upgrade(release_name, {'spec.components.indexNode.replicas': 2},
constants.NAMESPACE)
time.sleep(60)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t1 = datetime.datetime.now() - start
log.debug(f't1: {t1}')
assert round(t0 / t1) == 2
def e2e_milvus(host, c_name, collection_exist=False):
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create
collection_w = ApiCollectionWrapper()
if collection_exist:
collection_w.init_collection(name=c_name)
else:
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
# insert
data = cf.gen_default_list_data(ct.default_nb)
mutation_res, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
log.debug(collection_w.num_entities)
# create index
collection_w.create_index(ct.default_float_vec_field_name,
ct.default_index)
assert collection_w.has_index()[0]
assert collection_w.index()[0] == Index(collection_w.collection,
ct.default_float_vec_field_name,
ct.default_index)
# search
collection_w.load()
search_res, _ = collection_w.search(data[-1][:ct.default_nq],
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit)
assert len(search_res[0]) == ct.default_limit
log.debug(search_res[0][0].id)
# query
ids = search_res[0].ids[0]
term_expr = f'{ct.default_int64_field_name} in [{ids}]'
query_res, _ = collection_w.query(term_expr, output_fields=["*", "%"])
assert query_res[0][ct.default_int64_field_name] == ids
def test_expand_index_node(self):
"""
target: test expand indexNode from 1 to 2
method: 1.deploy two indexNode
2.create index with two indexNode
3.expand indexNode from 1 to 2
4.create index with one indexNode
expected: The cost of one indexNode is about twice that of two indexNodes
"""
release_name = "scale-index"
milvusOp, host, port = scale_common.deploy_default_milvus(release_name)
# connect
connections.add_connection(default={"host": host, "port": port})
connections.connect(alias='default')
data = cf.gen_default_dataframe_data(nb)
# create
c_name = "index_scale_one"
collection_w = ApiCollectionWrapper()
# collection_w.init_collection(name=c_name)
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
# insert
loop = 10
for i in range(loop):
collection_w.insert(data)
assert collection_w.num_entities == nb * loop
# create index on collection one and two
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t0 = datetime.datetime.now() - start
log.debug(f't0: {t0}')
collection_w.drop_index()
assert not collection_w.has_index()[0]
# expand indexNode from 1 to 2
milvusOp.upgrade(release_name,
{'spec.components.indexNode.replicas': 2},
constants.NAMESPACE)
milvusOp.wait_for_healthy(release_name, constants.NAMESPACE)
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name,
default_index_params)
assert collection_w.has_index()[0]
t1 = datetime.datetime.now() - start
log.debug(f't1: {t1}')
assert round(t0 / t1) == 2
def test_scale_data_node(self):
"""
target: test scale dataNode
method: 1.deploy milvus cluster with 2 dataNode
2.create collection with shards_num=5
3.continuously insert new data (daemon thread)
4.expand dataNode from 2 to 5
5.create new collection with shards_num=2
6.continuously insert new collection new data (daemon thread)
7.shrink dataNode from 5 to 3
expected: Verify milvus remains healthy, Insert and flush successfully during scale
Average dataNode memory usage
"""
release_name = "scale-data"
image_tag = get_latest_tag()
image = f'{constants.IMAGE_REPOSITORY}:{image_tag}'
fail_count = 0
data_config = {
'metadata.namespace': constants.NAMESPACE,
'metadata.name': release_name,
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.dataNode.replicas': 2,
'spec.config.dataCoord.enableCompaction': True,
'spec.config.dataCoord.enableGarbageCollection': True
}
mic = MilvusOperator()
mic.install(data_config)
if mic.wait_for_healthy(release_name,
constants.NAMESPACE,
timeout=1200):
host = mic.endpoint(release_name,
constants.NAMESPACE).split(':')[0]
else:
# log.warning(f'Deploy {release_name} timeout and ready to uninstall')
# mic.uninstall(release_name, namespace=constants.NAMESPACE)
raise BaseException(f'Milvus healthy timeout 1200s')
try:
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create
c_name = cf.gen_unique_str("scale_query")
# c_name = 'scale_query_DymS7kI4'
collection_w = ApiCollectionWrapper()
collection_w.init_collection(
name=c_name,
schema=cf.gen_default_collection_schema(),
shards_num=5)
tmp_nb = 10000
def do_insert():
while True:
tmp_df = cf.gen_default_dataframe_data(tmp_nb)
collection_w.insert(tmp_df)
log.debug(collection_w.num_entities)
t_insert = threading.Thread(target=do_insert, args=(), daemon=True)
t_insert.start()
# scale dataNode to 5
mic.upgrade(release_name, {'spec.components.dataNode.replicas': 5},
constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
log.debug("Expand dataNode test finished")
# create new collection and insert
new_c_name = cf.gen_unique_str("scale_query")
collection_w_new = ApiCollectionWrapper()
collection_w_new.init_collection(
name=new_c_name,
schema=cf.gen_default_collection_schema(),
shards_num=2)
def do_new_insert():
while True:
tmp_df = cf.gen_default_dataframe_data(tmp_nb)
collection_w_new.insert(tmp_df)
log.debug(collection_w_new.num_entities)
t_insert_new = threading.Thread(target=do_new_insert,
args=(),
daemon=True)
t_insert_new.start()
# scale dataNode to 3
mic.upgrade(release_name, {'spec.components.dataNode.replicas': 3},
constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE,
f"app.kubernetes.io/instance={release_name}")
log.debug(collection_w.num_entities)
time.sleep(300)
log.debug("Shrink dataNode test finished")
except Exception as e:
log.error(str(e))
fail_count += 1
# raise Exception(str(e))
finally:
log.info(f'Test finished with {fail_count} fail request')
assert fail_count <= 1
label = f"app.kubernetes.io/instance={release_name}"
log.info('Start to export milvus pod logs')
read_pod_log(namespace=constants.NAMESPACE,
label_selector=label,
release_name=release_name)
mic.uninstall(release_name, namespace=constants.NAMESPACE)
def test_expand_query_node(self):
release_name = "scale-query"
env = HelmEnv(release_name=release_name)
host = env.helm_install_cluster_milvus()
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create
c_name = "query_scale_one"
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
# insert
data = cf.gen_default_list_data(ct.default_nb)
mutation_res, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
# # create index
# collection_w.create_index(ct.default_float_vec_field_name, default_index_params)
# assert collection_w.has_index()[0]
# assert collection_w.index()[0] == Index(collection_w.collection, ct.default_float_vec_field_name,
# default_index_params)
collection_w.load()
# vectors = [[random.random() for _ in range(ct.default_dim)] for _ in range(5)]
res1, _ = collection_w.search(data[-1][:5],
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit)
# scale queryNode pod
env.helm_upgrade_cluster_milvus(queryNode=2)
c_name_2 = "query_scale_two"
collection_w2 = ApiCollectionWrapper()
collection_w2.init_collection(
name=c_name_2, schema=cf.gen_default_collection_schema())
collection_w2.insert(data)
assert collection_w2.num_entities == ct.default_nb
collection_w2.load()
res2, _ = collection_w2.search(data[-1][:5],
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit)
assert res1[0].ids == res2[0].ids
def test_expand_data_node(self):
"""
target: test create and insert api after expand dataNode pod
method: 1.create collection a and insert df
2.expand dataNode pod from 1 to 2
3.verify collection a property and verify create and insert of new collection
expected: two collection create and insert op are both correctly
"""
release_name = "scale-data"
milvusOp, host, port = scale_common.deploy_default_milvus(release_name)
# connect
connections.add_connection(default={"host": host, "port": port})
connections.connect(alias='default')
# create
c_name = cf.gen_unique_str(prefix)
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name, schema=cf.gen_default_collection_schema())
# # insert
data = cf.gen_default_list_data()
mutation_res, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
# scale dataNode to 2 pods
milvusOp.upgrade(release_name, {'spec.components.dataNode.replicas': 2}, constants.NAMESPACE)
milvusOp.wait_for_healthy(release_name, constants.NAMESPACE)
# after scale, assert data consistent
assert utility.has_collection(c_name)
assert collection_w.num_entities == ct.default_nb
# assert new operations
new_cname = cf.gen_unique_str(prefix)
new_collection_w = ApiCollectionWrapper()
new_collection_w.init_collection(name=new_cname, schema=cf.gen_default_collection_schema())
new_mutation_res, _ = new_collection_w.insert(data)
assert new_mutation_res.insert_count == ct.default_nb
assert new_collection_w.num_entities == ct.default_nb
# assert old collection ddl
mutation_res_2, _ = collection_w.insert(data)
assert mutation_res.insert_count == ct.default_nb
assert collection_w.num_entities == ct.default_nb*2
collection_w.drop()
new_collection_w.drop()
def test_shrink_query_node(self):
"""
target: test shrink queryNode from 2 to 1
method: 1.deploy two queryNode
2.search two collections in two queryNode
3.upgrade queryNode from 2 to 1
4.search second collection
expected: search result is correct
"""
# deploy
release_name = "scale-query"
env = HelmEnv(release_name=release_name, queryNode=2)
host = env.helm_install_cluster_milvus(
image_pull_policy=constants.IF_NOT_PRESENT)
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# collection one
data = cf.gen_default_list_data(nb)
c_name = "query_scale_one"
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
collection_w.insert(data)
assert collection_w.num_entities == nb
collection_w.load()
res1, _ = collection_w.search(data[-1][:nq],
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit)
assert res1[0].ids[0] == data[0][0]
# collection two
c_name_2 = "query_scale_two"
collection_w2 = ApiCollectionWrapper()
collection_w2.init_collection(
name=c_name_2, schema=cf.gen_default_collection_schema())
collection_w2.insert(data)
assert collection_w2.num_entities == nb
collection_w2.load()
res2, _ = collection_w2.search(data[-1][:nq],
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit)
assert res2[0].ids[0] == data[0][0]
# scale queryNode pod
env.helm_upgrade_cluster_milvus(queryNode=1)
# search
res1, _ = collection_w.search(data[-1][:nq],
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit)
assert res1[0].ids[0] == data[0][0]
res2, _ = collection_w2.search(data[-1][:nq],
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit)
assert res2[0].ids[0] == data[0][0]
class Base:
""" Initialize class object """
connection_wrap = None
collection_wrap = None
partition_wrap = None
index_wrap = None
utility_wrap = None
collection_schema_wrap = None
field_schema_wrap = None
collection_object_list = []
def setup_class(self):
log.info("[setup_class] Start setup class...")
def teardown_class(self):
log.info("[teardown_class] Start teardown class...")
def setup_method(self, method):
log.info(("*" * 35) + " setup " + ("*" * 35))
log.info("[setup_method] Start setup test case %s." % method.__name__)
self.connection_wrap = ApiConnectionsWrapper()
self.utility_wrap = ApiUtilityWrapper()
self.collection_wrap = ApiCollectionWrapper()
self.partition_wrap = ApiPartitionWrapper()
self.index_wrap = ApiIndexWrapper()
self.collection_schema_wrap = ApiCollectionSchemaWrapper()
self.field_schema_wrap = ApiFieldSchemaWrapper()
def teardown_method(self, method):
log.info(("*" * 35) + " teardown " + ("*" * 35))
log.info("[teardown_method] Start teardown test case %s..." %
method.__name__)
try:
""" Drop collection before disconnect """
if self.connection_wrap.get_connection(
alias=DefaultConfig.DEFAULT_USING)[0] is None:
self.connection_wrap.connect(alias=DefaultConfig.DEFAULT_USING,
host=param_info.param_host,
port=param_info.param_port)
if self.collection_wrap.collection is not None:
self.collection_wrap.drop(
check_task=ct.CheckTasks.check_nothing)
collection_list = self.utility_wrap.list_collections()[0]
for collection_object in self.collection_object_list:
if collection_object.collection is not None and collection_object.name in collection_list:
collection_object.drop(
check_task=ct.CheckTasks.check_nothing)
except Exception as e:
log.debug(str(e))
try:
""" Delete connection and reset configuration"""
res = self.connection_wrap.list_connections()
for i in res[0]:
self.connection_wrap.remove_connection(i[0])
# because the connection is in singleton mode, it needs to be restored to the original state after teardown
self.connection_wrap.add_connection(
default={
"host": DefaultConfig.DEFAULT_HOST,
"port": DefaultConfig.DEFAULT_PORT
})
except Exception as e:
log.debug(str(e))
def test_chaos_data_consist(self, connection, chaos_yaml):
"""
target: verify data consistence after chaos injected and recovered
method: 1. create a collection, insert some data, search and query
2. inject a chaos object
3. reconnect to service
4. verify a) data entities persists, index persists,
b) search and query results persist
expected: collection data and results persist
"""
c_name = cf.gen_unique_str('chaos_collection_')
nb = 5000
i_name = cf.gen_unique_str('chaos_index_')
index_params = {
"index_type": "IVF_SQ8",
"metric_type": "L2",
"params": {
"nlist": 64
}
}
# create
t0 = datetime.datetime.now()
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
tt = datetime.datetime.now() - t0
log.info(f"assert create: {tt}")
assert collection_w.name == c_name
# insert
data = cf.gen_default_list_data(nb=nb)
t0 = datetime.datetime.now()
_, res = collection_w.insert(data)
tt = datetime.datetime.now() - t0
log.info(f"assert insert: {tt}")
assert res
# flush
t0 = datetime.datetime.now()
assert collection_w.num_entities == nb
tt = datetime.datetime.now() - t0
log.info(f"assert flush: {tt}")
# search
collection_w.load()
search_vectors = cf.gen_vectors(1, ct.default_dim)
t0 = datetime.datetime.now()
search_params = {"metric_type": "L2", "params": {"nprobe": 16}}
search_res, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param=search_params,
limit=1)
tt = datetime.datetime.now() - t0
log.info(f"assert search: {tt}")
assert len(search_res) == 1
# index
t0 = datetime.datetime.now()
index, _ = collection_w.create_index(
field_name=ct.default_float_vec_field_name,
index_params=index_params,
name=i_name)
tt = datetime.datetime.now() - t0
log.info(f"assert index: {tt}")
assert len(collection_w.indexes) == 1
# query
term_expr = f'{ct.default_int64_field_name} in [1001,1201,999,99]'
t0 = datetime.datetime.now()
query_res, _ = collection_w.query(term_expr)
tt = datetime.datetime.now() - t0
log.info(f"assert query: {tt}")
assert len(query_res) == 4
# reboot a pod
reboot_pod(chaos_yaml)
# parse chaos object
chaos_config = cc.gen_experiment_config(chaos_yaml)
meta_name = chaos_config.get('metadata', None).get('name', None)
# wait all pods ready
log.info(
f"wait for pods in namespace {constants.CHAOS_NAMESPACE} with label app.kubernetes.io/instance={meta_name}"
)
wait_pods_ready(constants.CHAOS_NAMESPACE,
f"app.kubernetes.io/instance={meta_name}")
log.info(
f"wait for pods in namespace {constants.CHAOS_NAMESPACE} with label release={meta_name}"
)
wait_pods_ready(constants.CHAOS_NAMESPACE, f"release={meta_name}")
log.info("all pods are ready")
# reconnect if needed
sleep(constants.WAIT_PER_OP * 3)
reconnect(connections, alias='default')
# verify collection persists
assert utility.has_collection(c_name)
log.info("assert collection persists")
collection_w2 = ApiCollectionWrapper()
collection_w2.init_collection(c_name)
# verify data persist
assert collection_w2.num_entities == nb
log.info("assert data persists")
# verify index persists
assert collection_w2.has_index(i_name)
log.info("assert index persists")
# verify search results persist
collection_w2.load()
search_res, _ = collection_w.search(
data=search_vectors,
anns_field=ct.default_float_vec_field_name,
param=search_params,
limit=1)
tt = datetime.datetime.now() - t0
log.info(f"assert search: {tt}")
assert len(search_res) == 1
# verify query results persist
query_res2, _ = collection_w2.query(term_expr)
assert len(query_res2) == len(query_res)
log.info("assert query result persists")
def test_expand_index_node(self):
"""
target: test expand indexNode from 1 to 2
method: 1.deploy two indexNode
2.create index with two indexNode
3.expand indexNode from 1 to 2
4.create index with one indexNode
expected: The cost of one indexNode is about twice that of two indexNodes
"""
release_name = "expand-index"
image_tag = get_latest_tag()
image = f'{constants.IMAGE_REPOSITORY}:{image_tag}'
init_replicas = 1
expand_replicas = 2
data_config = {
'metadata.namespace': constants.NAMESPACE,
'spec.mode': 'cluster',
'metadata.name': release_name,
'spec.components.image': image,
'spec.components.proxy.serviceType': 'LoadBalancer',
'spec.components.indexNode.replicas': init_replicas,
'spec.components.dataNode.replicas': 2,
'spec.config.common.retentionDuration': 60
}
mic = MilvusOperator()
mic.install(data_config)
if mic.wait_for_healthy(release_name, constants.NAMESPACE, timeout=1800):
host = mic.endpoint(release_name, constants.NAMESPACE).split(':')[0]
else:
# If deploy failed and want to uninsatll mic
# log.warning(f'Deploy {release_name} timeout and ready to uninstall')
# mic.uninstall(release_name, namespace=constants.NAMESPACE)
raise MilvusException(message=f'Milvus healthy timeout 1800s')
try:
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create collection
c_name = "index_scale_one"
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name, schema=cf.gen_default_collection_schema())
# insert data
data = cf.gen_default_dataframe_data(nb)
loop = 100
for i in range(loop):
collection_w.insert(data, timeout=60)
assert collection_w.num_entities == nb * loop
# create index
# Note that the num of segments and the num of indexNode are related to indexing time
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name, default_index_params)
assert collection_w.has_index()[0]
t0 = datetime.datetime.now() - start
log.info(f'Create index on {init_replicas} indexNode cost t0: {t0}')
# drop index
collection_w.drop_index()
assert not collection_w.has_index()[0]
# expand indexNode
mic.upgrade(release_name, {'spec.components.indexNode.replicas': expand_replicas}, constants.NAMESPACE)
mic.wait_for_healthy(release_name, constants.NAMESPACE)
wait_pods_ready(constants.NAMESPACE, f"app.kubernetes.io/instance={release_name}")
# create index again
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name, default_index_params)
assert collection_w.has_index()[0]
t1 = datetime.datetime.now() - start
log.info(f'Create index on {expand_replicas} indexNode cost t1: {t1}')
collection_w.drop_index()
start = datetime.datetime.now()
collection_w.create_index(ct.default_float_vec_field_name, default_index_params)
assert collection_w.has_index()[0]
t2 = datetime.datetime.now() - start
log.info(f'Create index on {expand_replicas} indexNode cost t2: {t2}')
log.debug(f't2 is {t2}, t0 is {t0}, t0/t2 is {t0 / t2}')
# assert round(t0 / t2) == 2
except Exception as e:
raise Exception(str(e))
finally:
label = f"app.kubernetes.io/instance={release_name}"
log.info('Start to export milvus pod logs')
read_pod_log(namespace=constants.NAMESPACE, label_selector=label, release_name=release_name)
mic.uninstall(release_name, namespace=constants.NAMESPACE)
def e2e_milvus(host, c_name):
""" e2e milvus """
log.debug(f'pid: {os.getpid()}')
# connect
connections.add_connection(default={"host": host, "port": 19530})
connections.connect(alias='default')
# create
collection_w = ApiCollectionWrapper()
collection_w.init_collection(name=c_name,
schema=cf.gen_default_collection_schema())
# insert
df = cf.gen_default_dataframe_data()
mutation_res, _ = collection_w.insert(df)
assert mutation_res.insert_count == ct.default_nb
log.debug(collection_w.num_entities)
# create index
collection_w.create_index(ct.default_float_vec_field_name,
ct.default_index)
assert collection_w.has_index()[0]
assert collection_w.index()[0] == Index(collection_w.collection,
ct.default_float_vec_field_name,
ct.default_index)
# search
collection_w.load()
search_res, _ = collection_w.search(cf.gen_vectors(1, dim=ct.default_dim),
ct.default_float_vec_field_name,
ct.default_search_params,
ct.default_limit)
assert len(search_res[0]) == ct.default_limit
log.debug(search_res[0].ids)
# query
ids = search_res[0].ids[0]
term_expr = f'{ct.default_int64_field_name} in [{ids}]'
query_res, _ = collection_w.query(term_expr, output_fields=["*", "%"])
assert query_res[0][ct.default_int64_field_name] == ids
