Azure Search索引Azure Blob Storage：最佳唯一键选型及实现问询

Great question! Let's walk through the best options and real-world practices for setting a unique key when indexing Azure Blob Storage with Azure Search.

First off, you don't need to add custom properties just to get a valid unique key—Azure Search's blob indexer generates a perfect candidate automatically: metadata_storage_path.

Here's why this is the best choice:

• It's naturally unique: Every blob has a distinct storage path, and this field encodes that path in Base64, which strips out any problematic characters like / or . that you were worried about.
• No extra work required: The indexer creates this field by default, so you don't have to modify your blob upload process or add custom metadata.
• Officially recommended: This built-in field is designed explicitly to uniquely identify blobs in the index, making it the lowest-friction, most reliable option.

I've used this approach in multiple production projects, and it's never failed me—no duplicate keys, no special character issues, just seamless indexing. To set it up, simply configure your Azure Search index to use metadata_storage_path as the key field (it defaults to a string type, which meets the requirement).

If you need a human-readable key (like an order number or document ID) instead of a Base64 string, adding a custom metadata property to your blobs is a solid alternative. Here's how to implement it properly:

• Add metadata during upload: When uploading a blob, include a unique string property (like a GUID or business-specific ID) in the blob's metadata. For example, using the .NET SDK:

  var blobClient = containerClient.GetBlobClient("my-document.pdf");
  blobClient.Upload("path/to/file.pdf", new BlobUploadOptions
  {
      Metadata = { { "documentId", Guid.NewGuid().ToString() } }
  });
  

• Map the metadata to your index key: Create a string-type key field in your Azure Search index (e.g., id), then set up a field mapping in your indexer to pull the documentId metadata into this key field.

A critical note here: You must ensure every blob has this metadata property before indexing—Azure Search will throw an error if the key field is missing. Using a GUID is a safe way to guarantee uniqueness without extra validation.

• Prioritize metadata_storage_path unless you have a specific need: It's the lowest-effort option, and the lack of readability rarely matters for search indexing workflows.
• Keys are immutable: Once a document is indexed with a key, you can't change that key later. If you use custom metadata, pick a value that won't need updating. If you use metadata_storage_path, moving a blob will create a new key (and thus a new indexed document), so you'll need to clean up old entries if you reorganize blobs.
• Validate uniqueness upfront: For custom metadata, consider adding checks during upload (like using Azure Functions) to avoid duplicate keys that would break indexing.

内容的提问来源于stack exchange，提问作者Ekaterina Ermilova