Explain an Azure background-job architecture with Mermaid

Draw an illustrative Azure job-processing architecture with API, queue, worker, and storage boundaries. Explain acknowledgment, completion, and failure handling.

Updated November 19, 2026 · 2 min read

An Azure architecture diagram is easier to understand when it tells a specific story. For a background-job system, that story is the difference between accepting a request and completing the work.

The example below uses familiar Azure service roles to explain that distinction. It is a logical design sketch, not an Azure deployment template or a production-ready reference architecture.

Draw acceptance and processing separately

flowchart LR
  Client[Client]
  API[API hosted on App Service]
  Queue[Service Bus queue]
  Worker[Background worker]
  Blob[Blob Storage]
  Status[(Job status store)]
  Client -->|submit job| API
  API -->|enqueue job| Queue
  API -->|create status record| Status
  Queue -->|receive message| Worker
  Worker -->|write output| Blob
  Worker -->|update completion| Status
  Client -->|request status| API
  API -->|read status| Status

“Job status store” deliberately avoids selecting a database. The right storage choice depends on access patterns and other requirements. Similarly, the worker box describes a role; it does not assert which Azure compute service hosts it.

State the contract beside the arrows

When the API responds that a job was accepted, the output may not exist yet. Document how the caller obtains a job ID and learns about completion. The diagram should not imply that enqueueing and completing are the same event.

There is also a consistency question between creating a status record and publishing the message. What happens if one succeeds and the other fails? A diagram cannot solve that design problem, but making the two writes visible helps the team discuss its chosen approach.

Add the failure view

Ask what happens when processing fails repeatedly. A dead-letter path may belong in an operational version of the diagram, along with who inspects and retries those messages. Azure Service Bus provides dead-letter queue behavior, but the application still needs handling and ownership. Consult the Service Bus dead-letter documentation.

Also decide how to handle duplicate work. The worker should have a defined policy for an already-completed job. Do not describe an arrow as “exactly once” unless the full application behavior supports that promise.

Use provider icons without losing meaning

Flowgraph Studio includes the official Azure icons. Find them with the Rich node builder in the editor, which writes the @icon: line for you, and keep descriptive labels such as “job queue” or “rendered reports.” Service names explain the technology; role names explain why it is present. For other Azure shapes, the Templates page has serverless order processing, a static web app, streaming analytics, and a hub-and-spoke network to adapt.

Use grouping for boundaries only when you can name the boundary accurately. A resource group, network, subscription, and application tier are different things. A generic blue rectangle should not stand in for all four.

Create the example in Flowgraph Studio, then add only the deployment details your audience needs. For Mermaid grouping and labels, see the flowchart guide. The result should make the request contract clearer before it makes the slide prettier.

More guides