[DISCUSS] change the warning to debug in _sanity_check providers_manager.py

[Ping Zhang <pi...@umich.edu>]

Hi all,

I am thinking if we can make the warning level to debug level or make it
configurable to ignore those warnings from:
```
    except ImportError as e:
        # When there is an ImportError we turn it into debug warnings as
this is
        # an expected case when only some providers are installed
        log.warning(
            "Exception when importing '%s' from '%s' package",
            class_name,
            provider_package,
            exc_info=e,
        )
        return None
```
in `def _sanity_check`

https://github.com/apache/airflow/blob/main/airflow/providers_manager.py#L171-L176


It is distracting when launching the webserver.

What's your thoughts?

Thanks,

Ping

Re: [DISCUSS] change the warning to debug in _sanity_check providers_manager.py

[Ping Zhang <pi...@umich.edu>]

Hi Jarek,

Appreciate the background of this move. That makes sense to keep it as
warning.

Thanks,

Ping


On Thu, Feb 10, 2022 at 1:40 PM Jarek Potiuk <ja...@potiuk.com> wrote:

> This has been changed BACK to warning in this change
> https://github.com/apache/airflow/pull/21074 after this issue reported:
> https://github.com/apache/airflow/issues/20709
>
> The change introduces a new type of error that a provider can raise if
> there is an "optional feature" that depends on not installed dependency.
>
> Previously we silenced it and it caused the #20709  kind of problems where
> imports for providers failed "silently"
>
> The #21704 strikes the right balance (IMHO) between "warnings on
> accidental import errors" and "silencing expected import errors when
> provider decides so".
>
> J.
>
>
>
>
>

Re: [DISCUSS] change the warning to debug in _sanity_check providers_manager.py

[Jarek Potiuk <ja...@potiuk.com>]

This has been changed BACK to warning in this change
https://github.com/apache/airflow/pull/21074 after this issue reported:
https://github.com/apache/airflow/issues/20709

The change introduces a new type of error that a provider can raise if
there is an "optional feature" that depends on not installed dependency.

Previously we silenced it and it caused the #20709  kind of problems where
imports for providers failed "silently"

The #21704 strikes the right balance (IMHO) between "warnings on
accidental import errors" and "silencing expected import errors when
provider decides so".

J.