Handling launch errors

Prior to launching a Step Job the engine should be in possession of the RunningWorkflow record and will have created a RunningWorkflowStep record.

When the engine launches the step corresponding to the RunningWorkflowStep record (using InstanceLauncher.launch()) it receives LaunchResponse object. If the object's error_num is anything other than 0 (zero) it can assume that the Step's Job Instance did not start. No PodMessage will be received.

Required Actions

The engine MUST call the API adapter's set_running_workflow_step_done() method to update the state of the step. It must set the method's success value to False, set error_num to the error_num in the launch response, and provide a suitable error_msg.

The engine MUST also call the API adapter's set_running_workflow_done() to update the state of the workflow run. It must set the method's success value to False, set error_num to the error_num in the launch response, and provide a suitable error_msg.

- NOTE -

The engine MUST handle any PodMessages for the same running workflow but must not continue with any further step executions, and must not alter the success state of the RunningWorkflow. More than one step instance may be running concurrently so the engine must handle their responses as normal but must not proceed with any further launches.