Zanie Blue

I'm going to mark this as deferred for now, as I do not think we can undertake this in the near future. Loosely, I think that creating a `prefect-openapi-client` is...

We don't have internal consensus on alternative package manager to use an `pip` / `setuptools` has all the features we need. What are you using to detect the OpenAPI errors?

> about errors - I tried generator - https://github.com/openapi-generators/openapi-python-client Thanks! @abrookins should we open an issue to ensure our OpenAPI specification conforms to the latest standards and include some sort...

Interesting. For reference, this is failing here https://github.com/PrefectHQ/prefect/blob/38758c6527314abd97da3e791f227ca5afc947e7/src/prefect/engine.py#L955-L960 I'm a bit confused as from your code this task should never be completed already. Are your flows running on the same...

@caleb-recursion Can you share the `TaskRun` and `State` entries for that task run id?

Yeah the task results aren't available on another machine, but why is this task being loaded from another machine? Are you using the `cache_key_fn`?

Can you share details per https://github.com/PrefectHQ/prefect/issues/6086#issuecomment-1208336327 for the task run that fails

This exception is occurring outside your code so retries aren't handled. I believe there's an open issue with retrieving logs from Dask. This occurs without the Dask task runner, correct?...

Thanks for continuing to share updates. That's a _very_ weird error since we explicitly pass the flow run context to the logger retrieval call. I'll look into that one separately.

> Is there any chance the dask workers are being killed before a file is being flushed to disk? Is there any logic to confirm the state is written before...