Getting a surprising-to-me error running a super-s...
# prefect-community
Getting a surprising-to-me error running a super-simple flow using the "Non-Docker Storage for Containerized Environments" recipe - something that was working in 12.x. If I start my agent with
and try to run the registered flow via the UI. Agent picks up the job fine but reports:
Copy code
[2020-08-06 23:43:16,058] ERROR - agent | Error while deploying flow: AttributeError("Can't pickle local object 'NpipeHTTPAdapter.__init__.<locals>.<lambda>'")
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "C:\Python38\lib\multiprocessing\", line 116, in spawn_main
    exitcode = _main(fd, parent_sentinel)
  File "C:\Python38\lib\multiprocessing\", line 126, in _main
    self = reduction.pickle.load(from_parent)
EOFError: Ran out of input
I stripped anything out of the flow that was unique to the container just in case that was the issue. It runs fine under the standard agent, but fails with the docker agent.
Hi @Skip Breidbach - not sure what might be going on here, can you provide a min reproducible example for us to take a look at?
@nicholas sure thing! Sorry for the delay. Here's a bare-bones example.
Copy code
from prefect import Flow, task
from prefect.environments import LocalEnvironment
from import S3, Docker

def do_it():
    return 1

storage = Docker(

with Flow("simple",
          storage=S3(bucket="skip-sandbox", key="flows/simple"),
          environment=LocalEnvironment(metadata={"image": "simple-container"})) as flow:
    x = do_it()

flow.register(project_name="Skip's Sandbox")
And I start the agent with
start prefect agent start docker -l s3-flow-storage -l <machine name tag> --show-flow-logs
Edit* in my example above I was initially pulling
- I ran the same sample against 13.1 just in case there was a change but got the same results.
Thanks @Skip Breidbach, I'm going to open an issue for the Core team to take a look at this behavior.
@Marvin open "Can't pickle local object when using Non-Docker Storage for Containerized Environments"
Hey @Skip Breidbach I posted a comment on that issue 🙂
💥 1