workflow

Durable workflows are long-running processes that endure restarts, pause for external events, and coordinate multi-step execution across distributed services. You need durable workflows when orchestrating processes that require reliable retries and external event-driven pauses.

You start a workflow using start(myWorkflow) and resume it by using hooks designed for external event-driven pauses. The Workflow SDK maintains serialization and sandboxing rules automatically to ensure resumable execution across services.

Hooks enable external event-driven pauses and resumes within a durable workflow. By configuring hooks, the workflow suspends execution until an external event triggers, allowing multi-step coordination across distributed services without losing state.

Yes, the Workflow SDK supports built-in observability for multi-step orchestration. You can start, resume, and observe workflows directly, ensuring end-to-end orchestration across services maintains reliable fault tolerance and state visibility.

The best way to handle restarts and retries in distributed orchestration is using a durable workflow SDK. It maintains serialization and sandboxing rules, ensuring long-running processes endure restarts and execute reliable retries across distributed services.

You should avoid durable workflows for short-lived, stateless computations that do not require external event coordination or restart tolerance. Overhead from serialization, sandboxing rules, and observability makes them unsuitable for simple, synchronous tasks.