Tune consistency

Milvus supports setting consistency level while creating a collection, conducting a vector query, and conducting a vector search (only on PyMilvus currently). Milvus supports four levels of consistency: Strong, Eventual, Bounded, and Session. By default, a collection created without specifying the consistency level is set with Bounded consistency level. This topic describes how to tune consistency.

Configure parameter

By default, the consistency level is set as Bounded, under which Milvus reads a less updated data view (usually several seconds earlier) when a search or query request comes. You can set the consistency level by configuring the parameter consistency_level while creating a collection and conducting a search or query. See Guarantee Timestamp in Search Requests for the mechanism behind.

Parameter	Description	Option
consistency_level	Consistency level of the collection to create.	Strong Bounded Session Eventually

Example

The following examples set the consistency level as Strong, meaning Milvus will read the most updated data view at the exact time point when a search or query request comes. The consistency level set in the search or query requests overwrites the one set while creating the collection. Without specifying the consistency level during a search or query, Milvus adopts the original consistency level of the collection.

• Create a collection

from pymilvus import Collection
collection = Collection(
    name=collection_name, 
    schema=schema, 
    using='default', 
    shards_num=2,
    consistency_level="Strong"
    )

• Conduct a vector search

result = hello_milvus.search(
        vectors_to_search,
        "embeddings",
        search_params,
        limit=3,
        output_fields=["random"],
        # search will scan all entities inserted into Milvus.
        consistency_level="Strong",
        )

• Conduct a vector query

res = collection.query(
  expr = "book_id in [2,4,6,8]", 
  output_fields = ["book_id", "book_intro"],
  consistency_level="Strong"
)