Milvus Lite on Unix, lock file and refuse to connect

[christy]

Trying to use Milvus Lite on Unix, a user repeatedly gets a lock file and error message they cannot connect to Milvus Lite.

Code used:
https://milvus.io/docs/quickstart.md

Error message:
E20240722 17:51:48.802269 71268 collection_meta.cpp:94] [SERVER][CreateTable][milvus] Create table failed, errs: database is locked
E20240722 17:51:48.802600 71268 server.cpp:73] [SERVER][main][milvus] Init milvus failed
Failed to create new connection using: 6aa54c5d05b945488b9f5fc0bc5500c1

File /anaconda/envs/test/lib/python3.11/site-packages/pymilvus/client/grpc_handler.py:150, in GrpcHandler._wait_for_channel_ready(self, timeout)
148 self._setup_identifier_interceptor(self._user, timeout=timeout)
149 except grpc.FutureTimeoutError as e:
--> 150 raise MilvusException(
151 code=Status.CONNECT_FAILED,
152 message=f"Fail connecting to server on {self._address}, illegal connection params or server unavailable",
153 ) from e
154 except Exception as e:
155 raise e from e

MilvusException: <MilvusException: (code=2, message=Fail connecting to server on unix:/tmp/tmp4da_2ec6_test.db.sock, illegal connection params or server unavailable)>"
}:cry:

[christy]

[junjiejiangjjj]

This is an error message from sqlite3. There may be a limitation or problem with the underlying file system that is causing the database to be locked. Try running it in a different directory.

[arielfaur]

I'm having the same issue on mac os, the lock is not removed.

[alibabadoufu]

I am having the same issue!

[leonardozilli]

Same problem here on Fedora 40.

[xiaofan-luan]

@junjiejiangjjj
is there anything we can do here?

[junjiejiangjjj]

@junjiejiangjjj is there anything we can do here?

I have run the script thousands of times on both Mac and Ubuntu, and I have not been able to reproduce this issue.

[leonardozilli]

More context, i'm working with jupyter notebooks in vscode.
platform: fedora 40
python version: 3.12.6
pymilvus version: 2.4.9

Populate the database

from pymilvus import MilvusClient, connections, utility, FieldSchema, CollectionSchema, DataType, Collection
from pymilvus.model.hybrid import BGEM3EmbeddingFunction


MILVUS_URL = "../vec_db/definitions_vectors.db"

client = MilvusClient(
    uri=MILVUS_URL
)

connections.connect(uri=MILVUS_URL)

fields = [
    FieldSchema(name="id", dtype=DataType.INT64,
                is_primary=True, auto_id=True, max_length=100),
    FieldSchema(name="text", dtype=DataType.VARCHAR, max_length=5000),
    FieldSchema(name="dense_vector", dtype=DataType.FLOAT_VECTOR,
                dim=1024),
]

schema = CollectionSchema(fields, "Definitions embeddings")

COLLECTION_NAME = "Definitions"
if utility.has_collection(COLLECTION_NAME):
    Collection(COLLECTION_NAME).drop()
collection = Collection(COLLECTION_NAME, schema, consistency_level="Strong")

dense_index = {"index_type": "AUTOINDEX", "metric_type": "IP"}
collection.create_index("dense_vector", dense_index)
collection.load()

for i in range(0, len(doc_list), 50):
    batched_entities = [
        doc_list[i : i + 50],
        defs_embeddings["dense"][i : i + 50],
    ]
    collection.insert(batched_entities)
print("Number of entities inserted:", collection.num_entities)

The code above creates the definitions_vectors.db and definitions_vectors.db.lock files in the /vec_db folder. I am able to connect to the db with no problems from the same jupyter notebook. However, if I open a new jupyter notebook and try to connect to the database like so:

from langchain_milvus import Milvus
from pymilvus.model.hybrid import BGEM3EmbeddingFunction
from langchain.embeddings.base import Embeddings

class BGEMilvusEmbeddings(Embeddings):
    def __init__(self):
        self.model = BGEM3EmbeddingFunction(
            model_name='BAAI/bge-m3',
            device='cpu',
            use_fp16=False #set to false if device='cpu'
        )

    def embed_documents(self, texts):
        embeddings = self.model.encode_documents(texts)
        return [i.tolist() for i in embeddings["dense"]]

    def embed_query(self, text):
        embedding = self.model.encode_queries([text])
        return embedding["dense"][0].tolist()

MILVUS_URI = "../vec_db/definitions_vectors.db"
MILVUS_COLLECTION_NAME = 'Definitions'

vector_store = Milvus(
    embedding_function=BGEMilvusEmbeddings(),
    connection_args={"uri": MILVUS_URI},
    collection_name=MILVUS_COLLECTION_NAME,
    vector_field="dense_vector",
    text_field="text",
)

I get the following error:

Failed to create new connection using: acd5d70fea4c46beb2a1a5bab35edf75

---------------------------------------------------------------------------
MilvusException: <MilvusException: (code=2, message=Fail connecting to server on unix:/tmp/tmp5nlcf3pw_definitions_vectors.db.sock, illegal connection params or server unavailable)>

[junjiejiangjjj]

More context, i'm working with jupyter notebooks in vscode. platform: fedora 40 python version: 3.12.6 pymilvus version: 2.4.9

Populate the database

from pymilvus import MilvusClient, connections, utility, FieldSchema, CollectionSchema, DataType, Collection
from pymilvus.model.hybrid import BGEM3EmbeddingFunction


MILVUS_URL = "../vec_db/definitions_vectors.db"

client = MilvusClient(
    uri=MILVUS_URL
)

connections.connect(uri=MILVUS_URL)

fields = [
    FieldSchema(name="id", dtype=DataType.INT64,
                is_primary=True, auto_id=True, max_length=100),
    FieldSchema(name="text", dtype=DataType.VARCHAR, max_length=5000),
    FieldSchema(name="dense_vector", dtype=DataType.FLOAT_VECTOR,
                dim=1024),
]

schema = CollectionSchema(fields, "Definitions embeddings")

COLLECTION_NAME = "Definitions"
if utility.has_collection(COLLECTION_NAME):
    Collection(COLLECTION_NAME).drop()
collection = Collection(COLLECTION_NAME, schema, consistency_level="Strong")

dense_index = {"index_type": "AUTOINDEX", "metric_type": "IP"}
collection.create_index("dense_vector", dense_index)
collection.load()

for i in range(0, len(doc_list), 50):
    batched_entities = [
        doc_list[i : i + 50],
        defs_embeddings["dense"][i : i + 50],
    ]
    collection.insert(batched_entities)
print("Number of entities inserted:", collection.num_entities)

The code above creates the definitions_vectors.db and definitions_vectors.db.lock files in the /vec_db folder. I am able to connect to the db with no problems from the same jupyter notebook. However, if I open a new jupyter notebook and try to connect to the database like so:

from langchain_milvus import Milvus
from pymilvus.model.hybrid import BGEM3EmbeddingFunction
from langchain.embeddings.base import Embeddings

class BGEMilvusEmbeddings(Embeddings):
    def __init__(self):
        self.model = BGEM3EmbeddingFunction(
            model_name='BAAI/bge-m3',
            device='cpu',
            use_fp16=False #set to false if device='cpu'
        )

    def embed_documents(self, texts):
        embeddings = self.model.encode_documents(texts)
        return [i.tolist() for i in embeddings["dense"]]

    def embed_query(self, text):
        embedding = self.model.encode_queries([text])
        return embedding["dense"][0].tolist()

MILVUS_URI = "../vec_db/definitions_vectors.db"
MILVUS_COLLECTION_NAME = 'Definitions'

vector_store = Milvus(
    embedding_function=BGEMilvusEmbeddings(),
    connection_args={"uri": MILVUS_URI},
    collection_name=MILVUS_COLLECTION_NAME,
    vector_field="dense_vector",
    text_field="text",
)

I get the following error:

Failed to create new connection using: acd5d70fea4c46beb2a1a5bab35edf75

---------------------------------------------------------------------------
MilvusException: <MilvusException: (code=2, message=Fail connecting to server on unix:/tmp/tmp5nlcf3pw_definitions_vectors.db.sock, illegal connection params or server unavailable)>

hi @leonardozilli the db file of milvus-lite does't support multiple processes to open, otherwise it will cause data confusion

[leonardozilli]

Okay that makes sense, I'll make sure to have only one process working on it. Thanks