AWS Step Functions Error Handling

For AWS Step Functions

• Error can happen in variety of ways
  • State machine definition error (example: no matching name for a state)
  • Task failures (example: exception in lambda)
  • Transient issues (example: network partition event)
• Use Retry to retry failed state and Catch to transition the state machine to failure path
  • Note: try not to handle the error in the Application layer because it increases the complexity of our application
• Some of the predefined error codes
  • States.ALL: matches any error name
  • States.Timeout: task ran longer than TimeoutSeconds eor no heartbeat recevied
  • States.TaskFailed: execution failure
  • States.Permissions: insufficient privileges to execute code
• The state may report is own errors and you can catch them in step functions

Retry (for Task State or Parallel State)

• ErrorEquals: specify the error type
• IntervalSeconds: how long should we delay after each retry
• BackoffRate: multiple with the delay after each retry for Exponential Backoff (any AWS service)
• MaxAttempts: default to 3. Set to 0 to never retry
  • When max attempts are reached. The Catch block kicks in

Catch (for Task State or Parallel State)

• ErrorEquals: match a specific kind of error
• Next: state to send to
• ResultPath: A path that determines what input is sent to the state specified in Next field
  • the $.error puts the error inside the output. For example