How Milvus Deletes Streaming Data in a Distributed Cluster

What this post added

This post details the implementation of the DELETE function in Milvus 2.0, enabling logical deletion of entities. It explains how data nodes handle INSERT and DELETE messages, including the use of bloom filters and Delta binlogs. It also describes how query nodes process these messages, distinguishing between growing and sealed segments, and how deleted entities are omitted during search operations using bitsets. The post also touches upon the impact of consistency levels on the visibility of deleted data.