Azure Search分区机制及无分区键场景下的读写疑问

Great questions about how Azure Cognitive Search handles partitioning when you're pushing data directly (no indexers involved)! Let me break this down clearly for you:

How Documents Are Routed to Partitions When Pushing #

When you push documents to your 3-partition index, Azure Search uses consistent hashing based on the document's unique key field to determine which partition stores the document. Here's the breakdown:

• Every document you push must include a unique key field (this is a mandatory requirement for all Azure Search indexes).
• The service computes a hash value from this key, then maps that hash to one of your 3 partitions.
• This mapping is consistent: the same document key will always land in the same partition, which ensures updates or deletions target the correct data without duplication.
• You don't need to specify any partition key manually—this entire process runs behind the scenes, completely transparent to your application.

How Queries Are Handled Across Partitions #

The query behavior varies based on the type of query you're running:

• Point lookups (by document key): If you use the lookup API to fetch a single document by its key, the service will compute the hash of the key to locate the exact partition where the document lives, then only query that partition. This is highly efficient and avoids unnecessary load on other partitions.
• Range/Full-text/Filtered queries: For broader queries (like full-text search, filtered results, or sorting across the entire index), the service will broadcast the query to all 3 partitions. Each partition processes the query locally, returns its matching results, and a coordinating node merges, sorts, and paginates these results into a single, unified response sent back to you. This coordination is handled automatically—you'll receive a seamless result set as if it's coming from a single data store.

内容的提问来源于stack exchange，提问作者vrcks