Edit

Storage containers and storage assets in Microsoft Discovery

Microsoft Discovery uses storage containers and storage assets to organize data used by workspaces and projects. A storage container references an Azure Blob Storage account or Azure NetApp Files volume. Storage assets reference paths within that storage backend. Together, they provide the data layer that agents, tools, and shared sessions use to read input and write output.

Key concepts

Concept Description Azure resource type
Storage container A resource-group-level resource that references an Azure Blob Storage account or Azure NetApp Files volume. Microsoft.Discovery/storageContainers
Storage asset A child resource of a storage container that references a path in the backing storage. Microsoft.Discovery/storageContainers/storageAssets
Storage account The underlying Azure Blob Storage account that holds the actual data. Microsoft.Storage/storageAccounts

Resource hierarchy

Storage containers are Azure Resource Manager (ARM) resources at resource group scope. Storage assets are their children. Projects reference these shared resources; neither a workspace nor a project is their ARM parent. Project access doesn't grant access to the referenced storage automatically. See Projects and shared sessions.

Diagram showing storage container and storage asset hierarchy within a resource group, with workspace and project references.

Projects in a workspace can reference multiple storage containers, each pointing to a storage backend. A storage container can have multiple storage assets, each pointing to a different path within that backend.

How storage containers work

Registration, not data movement

Creating a storage container registers an existing storage backend with Microsoft Discovery. It doesn't move or copy data. A Discovery storage container isn't an Azure Blob Storage container. The actual data remains in your storage backend. Projects, agents, and tools use the Discovery reference with the required identity permissions and network access.

Supported storage kinds

Discovery supports the following storage kinds:

Kind Value Description
Azure Blob Storage AzureStorageBlob Standard Azure Blob Storage account. Hierarchical namespace (Data Lake Storage Gen2) isn't required.
Azure NetApp Files AzureNetAppFiles Azure NetApp Files volume for high-performance file storage. Specify the netAppVolumeId instead of storageAccountId.

Azure Blob Storage is the most common choice for general-purpose data ingestion and output. Azure NetApp Files is available for workloads that require high-throughput NFS-based file access.

Authentication

A storage container is a reference, not an identity. Grant storage access to the identities used by its consumers: for example, the workspace's user-assigned managed identity (UAMI) for workspace file operations or the Bookshelf workload identity for indexing. For Azure Blob Storage, Storage Blob Data Contributor provides read/write data access on the backing storage account. For identity configuration and required roles, see Managed identities in Microsoft Discovery.

How storage assets work

A storage asset represents a path within the backend referenced by its parent storage container. For Azure Blob Storage, the path includes the blob container name and an optional file or prefix, such as datasets/experiment-001/. Storage assets are the primary way data enters and exits Discovery shared sessions.

Input data

When you create a storage asset with a path pointing to an existing blob or blob prefix, tools and agents running on the supercomputer can read that data. For example, you might create a storage asset pointing to a blob prefix containing CSV files for an experiment.

Output data

When agents and tools produce output, the platform tracks the files in the conversation's resource context. Agent-generated files and shared tool outputs are represented by storage assets. Tool outputs can remain intermediate resources until shared. See Files and storage assets and Share storage assets with users.

Addressing

Agents use conversation resource URIs to address files within the platform:

discovery://resources/{storageContainerName}/paths/{blobPath}

Agents use built-in tools like GetResourceContext and PreviewResource to discover and read files through these URIs. The platform translates discovery:// URIs to the physical blob storage location at runtime.

These conversation URIs aren't the same as the storageUri values used for REST job mounts. For that API's storage asset URI format, see DataMount object. Don't substitute one URI format for the other.

How Discovery resources use storage

Resource How it uses storage
Project References one or more storage containers as its data sources. Tools and investigations within the project read from and write to these containers.
Tool runs Mount storage containers as input and output data volumes. Input mounts provide read access; output mounts collect results.
Agents Use built-in file tools (WriteResource, PreviewResource) to create and read storage assets during investigations.
Investigations Accumulate storage assets as tasks complete. The root task collects all output file references through upward propagation.
Bookshelf Uses storage to ingest documents for knowledge-base indexing. Source data for indexing comes from storage assets.

Storage container lifecycle

Phase What happens
Create You register a storage backend by creating a storage container (PUT).
Use Projects, tools, and agents read from and write to the storage account through the registered container.
Update In API version 2026-06-01, the backing storage reference and kind are immutable after creation. To use a different backend, create a new storage container and reference it in new project configurations.
Delete Deleting a storage container removes the Discovery reference. The underlying Azure Blob Storage account and its data are not deleted. Delete child storage assets before deleting the parent container.

Important

Deleting a storage container doesn't delete your data. The actual blobs remain in your Azure Storage account. Only the Discovery resource reference is removed.

Networking considerations

The storage account backing a storage container must be network-accessible to the Discovery platform:

  • The storage account must allow access from the virtual network subnets used by the supercomputer, workspace, and agents.
  • If you restrict public access on the storage account, configure virtual network rules or private endpoints so the platform can reach it.
  • For CORS requirements (needed for Discovery Studio file browsing), see Azure Blob Storage in Microsoft Discovery.

Limitations

  • In API version 2026-06-01, the storage backend reference and kind are immutable after creation. Check the versioned storage container reference when using other API versions.
  • A storage container can reference only one storage backend (storage account or NetApp volume). To access multiple backends, create multiple storage containers.
  • A storage asset path can reference a file or a prefix containing multiple blobs.
  • The consuming workload identities need the appropriate storage permissions before accessing data.