def e2e_milvus(host, c_name):
# 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())
# collection_w.init_collection(name=c_name)
# 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, 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
# 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_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 connection(self, host, port):
connections.add_connection(default={"host": host, "port": port})
conn = connections.connect(alias='default')
if conn is None:
raise Exception("no connections")
self.host = host
self.port = port
return conn
def connection(self):
connections.add_connection(default={
"host": "192.168.1.239",
"port": 19530
})
conn = connections.connect(alias='default')
if conn is None:
raise Exception("no connections")
return conn
def test_expand_query_node(self):
release_name = "scale-query"
env = HelmEnv(release_name=release_name)
env.helm_install_cluster_milvus()
# connect
connections.add_connection(default={
"host": '10.98.0.8',
"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 __init__(self, name):
connections.add_connection(default={
'host': '127.0.0.1',
'port': '19530',
})
connections.connect(alias=name)
id_field = FieldSchema(name='id',
is_primary=True,
dtype=DataType.INT64,
description='id')
image_field = FieldSchema(name='image',
dtype=DataType.FLOAT_VECTOR,
dim=512,
description='image')
schema = CollectionSchema(fields=[id_field, image_field],
primary_field='id')
collection = Collection(name=name, schema=schema)
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-test"
env = HelmEnv(release_name=release_name, dataNode=2)
env.helm_install_cluster_milvus(image_pull_policy=constants.IF_NOT_PRESENT)
# connect
connections.add_connection(default={"host": '10.98.0.8', "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_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)
env.helm_install_cluster_milvus(
image_pull_policy=constants.IF_NOT_PRESENT)
# connect
connections.add_connection(default={
"host": '10.98.0.8',
"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
pdb.set_trace()
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]
