Azure storage model
01Namespace choice changes the operation itself.
An Azure Blob route spans account, container, virtual directory, blob type, and metadata decisions. Hierarchical namespace, large-object behavior, and destination policy influence how work is listed, resumed, and reconciled across Azure and non-Azure endpoints.
Account, container, path
Define the account, container, prefix or path, listing depth, inclusion rules, and expected properties. If a hierarchical namespace is involved, make directory assumptions explicit instead of treating every key as flat.
Copy state is not completion
Choose a recoverable unit for large blobs and define how interrupted reads, writes, and service-side operations are observed. A retry must identify a complete destination separately from a partial or pending result.
Properties travel with purpose
Specify destination names, overwrite policy, metadata and headers, access tier expectations, retention, archive, quarantine, and cleanup. Preserve the properties downstream systems actually use, not only the bytes.
Mode-aware Flower flow
02Preserve Azure semantics while the destination changes.
Flower can inspect properties and metadata, stream or range-read blobs, execute local or service-side copies, manage access tiers, and use native filesystem operations where the Azure namespace supports them.
Discover in the right namespace
A flat prefix walk and an ADLS directory traversal express different intent. Flower scopes listing depth, versions, snapshots, deleted entries, and metadata according to the route instead of assuming one universal directory model.
Choose copy or byte pipeline
A same-service copy can avoid moving bytes through compute; a cross-platform or transforming route needs a controlled stream. Flower keeps pending copy state, retries, validation, and final evidence in the plan.
Rename only where it is real
With hierarchical namespace, native rename can publish a file or directory atomically. In flat Blob storage, movement may require copy and delete, so the flow waits for destination proof before treating the source as disposable.
Azure production decisions
03Control identity, cold data, and publish semantics.
Azure features interact: hierarchical namespace affects operations, versions affect listing, archive tiers affect availability, and authorization may differ between blob and dfs endpoints. The route must state those dependencies.
Identity and endpoint
Choose managed identity, service principal, or another approved mechanism with least privilege. Test the exact blob or dfs endpoint, private networking, container scope, and ACL or RBAC boundary used in production.
Tier and rehydration
An archived blob is not an ordinary retry: it may need rehydration and a different service objective. Define whether the flow waits, alerts, skips, or launches a controlled retrieval path, including time and cost limits.
Headers, metadata, and versions
List the properties downstream systems depend on: content type, encoding, custom metadata, tier, version, timestamps, and naming. Validate them after delivery rather than proving only that the byte count matches.