Step Functions

coordination, orchestration for distributed applications

Document | Workshop |

Overview

• Similar to Microsoft Power Automate. A GUI workflow.

• Help build visual workflow to translate business requirements to technical requirements.

• Build apps fast, manage logic with low-code.

• Built-in try/catch, retry, rollback.

• Keywords: orchestration

Features

Workflow configuration

Built-in service primitives

AWS service integration

• Lambda, ECS, Fargate

• DynamoDB

• SNS, SQS

• AWS Batch, AWS Glue

• SageMaker

Coordination of distributed apps

Can make http connection between app hosted on EC2, on-premises, mobile devices.

Components reuse

Built-in error handling

• Retry failed or time-out tasks.

• Response differently depend on type of error.

History execution

Cost

Pay for each state transition.

Concepts

• Amazon State Language: is a JSON-based, structure language use to define state machine.

• "Type": "Parallel" use to create parallel branches state machine.

• Task: is a single unit of work and choice.

• State: a step in the workflow

  • Pass state: debugging (just pass the input to its output, or inject some fixed data)

  • Task state: do some works in your state machine. eg.Lambda

  • Choice state: if-else statement

  • Wait state: delay for a certain amount of time

  • Success state: stop & mark as success

  • Fail state: stop & mark as failure.

  • Parallel state: create parallel branches of the same input

  • Map state: as LOOP statement (execute the same step for multiple entries)

• HeartbeatSeconds: maximum interval that the task will wait for a hearbeat signal. If an activity worker fails to send heartbeats within this interval, the state is failed. A retry policy on the state allows another activity worker to attempt to complete the state.

HeartbeatSeconds vs Wait State:

• HeartbeatSeconds: This is a parameter used within a Task state to specify the maximum interval the task will wait for a heartbeat signal. If no heartbeat is received within this interval, the task is marked as failed and can retry based on a defined policy.

• Wait State: This is a state type used to introduce a deliberate delay in the execution of a state machine for a specified amount of time before transitioning to the next state. It is useful for adding pauses or delays in the workflow.

In essence, HeartbeatSeconds is about keeping a Task alive, while Wait state is about intentionally pausing the workflow.