Bring your towel and join one of the fastest growing data communities. Welcome to our second-generation open source orchestration platform, a completely rethought approach to dataflow automation.

Prefect Community

I'm getting familiar with prefect 2.0b3 and I've run into a problem with using classes that own a python logger.  I get the following error:

`ValueError: [TypeError("'_thread.RLock' object is not iterable"), TypeError('vars() argument must have __dict__ attribute')]`

Code to reproduce in thread

Prefect 1.3 didn't seem to care about loggers being a part of a class used in a flow - any  patterns worth adopting to replace this or does this merit a fix?

```import logging
from prefect import flow, task


class MyModel:
    def __init__(self, name: str):
        self.logger = logging.getLogger(f"model.{name}")
        self.name = name

    def model_things(self, a: int):
        <http://self.logger.info|self.logger.info>(f"Running on value {a}")
        return a &gt; 1


@task
def model_task(model: MyModel):
    outputs = []
    for i in range(10):
        outputs.append(model.model_things(i))
    return outputs


@flow
def model_flow(model):
    model_task(model)


model_flow(MyModel("testing"))```


you could set extra loggers via environment variable `PREFECT_LOGGING_EXTRA_LOGGERS` - <https://www.prefect.io/blog/logs-the-prefect-way/|this blog post> explains it in detail

This looks to be because you are passing the model to the flow as an argument. Not sure what’s going on behind the scenes (something to do with flow parameters maybe?); however, if you re-structure it a bit to create the model in the flow function instead it seems to work:

```@flow
def model_flow():
    model = MyModel("testing")
    model_task(model)

model_flow()```

Yeah, there seems to be a problem when serializing the model when passed as a parameter.  <@U02H1A95XDW> I suppose I could pass a `prefect` logger to my objects, but this seems like an extra step that requires classes to accept loggers instead of controlling them independently.  I still want  that class's logger around for non-prefect testing and experimentation.

Just to clarify, you can pass your model with the logger between tasks fine. It’s just passing it to a flow that doesn’t work.

Not sure about your use cases; however, I believe flows typically have more typical parameter data types (strings, ints, dicts, etc). So maybe you could create the “MyModel” class in the flow from the simple parameter types and then use that between tasks?

Fair point, I'm sure I can get some sort of factory `task` or function implemented.  Thanks, Kevin