I successfully ran a flow locally. however, when I deploy the flow to prefect-cloud, I got this warning message at the

healthcheck

step.

/opt/prefect/healthcheck.py:149: UserWarning: Task <Task: fetch_runnable_models> has retry settings but some upstream dependencies do not have result types. See <https://docs.prefect.io/core/concepts/results.html> for more details.
  result_check(flows)

please see the thread for more details

Hi @Hui Zheng - this isn’t necessarily a problem, depending on what scenarios you hope to cover with this Flow. Assuming you are using

Docker

storage, this Flow is not configured to store your task outputs anywhere. This means that, if

fetch_runnable_models

were to fail, and the process / job were to unexpectedly die before the retry could proceed, you would not be able to successfully retry the task because it would not be able to recreate its upstream data dependency. In normal operating scenarios though, the retry will occur within the same process and you should be fine - it’s the edge case when the process / job unexpectedly dies that this warning is highlighting

Thank you, Chris. yes, we are using

Docker

storage.. Because this flow is high critical, I do hope to recover from the worst edge cases, such as job were to unexpectedly die. How do I store your task outputs outside the docker container? I already set the flow result with the following settings to build out the container.

flow.result = GCSResult(bucket=...)

it seems not cover everything given to the warning message?

Hmmm, are you setting that result prior to registering the flow / building its storage?

yes, I set the result prior to registering the flow / building its storage step. It has been done like that, and I never see those warning message before. There is one change recently though around the time these warning message appears. I declared some variables in my

flow.py

file at the global level outside the scope of tasks and flow. and then use those variables in some tasks. Are those variables the cause of the warning?

schedule_type = getenv('SCHEDULE_TYPE')
target_release_u = getenv('TARGET_RELEASE_VERSION').replace('.', '_')
environment = getenv("ENVIRONMENT", default="stage")
fs_collection_name = '_'.join([environment, target_release_u, schedule_type])

No that wouldn’t do it, the check is incredibly simple: https://github.com/PrefectHQ/prefect/blob/master/src/prefect/environments/storage/_healthcheck.py#L67 You can see that if the flow has a result, the check is skipped; so for some reason, your Flow object within your Docker image does not have a result attribute

Could that related to our another conversation, https://prefect-community.slack.com/archives/CL09KU1K7/p1603838678211000 my t`ask_B` in this case is the

SkipNoneDbtShellTask

task below. You suggested that the 

task_B

 could check if the provided value is 

None

 and 

SKIP

 if so. However, actually, when I implement it, I did it a bit differently. I say ``task_B` that is

SkipNoneDbtShellTask

, return None if the provided value is

None

.

class SkipNoneDbtShellTask(DbtShellTask):
    @defaults_from_attrs("command", "env", "dbt_kwargs")
    def run(
        self,
        command: str = None,
        # env: dict = None,
        # dbt_kwargs: dict = None
        **kwargs: Any
    ) -> str:
        if command is None:
            return None
        else:
            return super(SkipNoneDbtShellTask, self).run(
                command=command, **kwargs)